Version bump (Changes update)

Version bump
Commentary: Update ChangeLog
2026-10-03 08:33:21 +02:00 · 2022-06-19 15:10:23 -04:00 · 2022-06-19 15:01:17 -04:00 · 2022-06-19 15:00:10 -04:00 · 2022-06-15 18:49:32 -04:00 · 2022-06-15 18:11:03 -04:00
1183 changed files with 63727 additions and 33008 deletions
@@ -1,6 +1,6 @@
 ---
 name: Bug report
-about: Something isn't working as expected, and it isn't "Unsupported." (Note our contributor agreement at https://github.com/verilator/verilator/blob/master/docs/CONTRIBUTING.rst)
+about: Use this to report that something isn't working as expected, and it isn't "Unsupported." (Note our contributor agreement at https://github.com/verilator/verilator/blob/master/docs/CONTRIBUTING.rst)
 title: ''
 labels: new
 assignees: ''
@@ -15,4 +15,4 @@ What 'verilator --version' are you using?  Did you try it with the git master ve

 What OS and distribution are you using?

-Would you be willing to try to fix Verilator yourself with assistance?
+May we assist you in trying to fix this in Verilator yourself?
@@ -1,6 +1,6 @@
 ---
 name: Feature Request
-about: Request something should be supported, or a new feature added. (Note our contributor agreement at https://github.com/verilator/verilator/blob/master/docs/CONTRIBUTING.rst)
+about: Use this to request something should be supported, or a new feature added. (Note our contributor agreement at https://github.com/verilator/verilator/blob/master/docs/CONTRIBUTING.rst)
 title: ''
 labels: new
 assignees: ''
@@ -13,4 +13,4 @@ What would you like added/supported?

 Can you attach an example that runs on other simulators?  (Must be openly licensed, ideally in test_regress format.)

-May we assist you in trying to fix this yourself?
+May we assist you in trying to fix this in Verilator yourself?
@@ -1,6 +1,6 @@
 ---
 name: Q and A, or Other
-about: Ask a question, not related to a specific bug or feature request. (Note our contributor agreement at https://github.com/verilator/verilator/blob/master/docs/CONTRIBUTING.rst)
+about: Use this to ask a question, not related to a specific bug nor feature request. (Note our contributor agreement at https://github.com/verilator/verilator/blob/master/docs/CONTRIBUTING.rst)
 title: ''
 labels: new
 assignees: ''
@@ -29,7 +29,7 @@ jobs:
    strategy:
      fail-fast: false
      matrix:
-        os: [ubuntu-20.04, ubuntu-18.04]
+        os: [ubuntu-22.04, ubuntu-20.04, ubuntu-18.04]
        compiler:
          - { cc: clang, cxx: clang++ }
          - { cc: gcc,   cxx: g++     }
@@ -37,9 +37,16 @@ jobs:
        exclude:
          # Build pull requests only with ubuntu-20.04 and without m32
          - os:  ${{ github.event_name == 'pull_request' && 'ubuntu-18.04' || 'do-not-exclude' }}
+          - os:  ${{ github.event_name == 'pull_request' && 'ubuntu-22.04' || 'do-not-exclude' }}
          - m32: ${{ github.event_name == 'pull_request' && 1              || 'do-not-exclude' }}
          # Build -m32 only on ubuntu-20.04
          - {os: ubuntu-18.04, m32: 1}
+          - {os: ubuntu-22.04, m32: 1}
+        include:
+          # Build GCC 10 on ubuntu-20.04
+          - os: ubuntu-20.04
+            compiler: { cc: gcc-10, cxx: g++-10 }
+            m32: 0
    runs-on: ${{ matrix.os }}
    name: Build | ${{ matrix.os }} | ${{ matrix.compiler.cc }} ${{ matrix.m32 && '| -m32' || '' }}
    env:
@@ -90,7 +97,7 @@ jobs:
    strategy:
      fail-fast: false
      matrix:
-        os: [ubuntu-20.04, ubuntu-18.04]
+        os: [ubuntu-22.04, ubuntu-20.04, ubuntu-18.04]
        compiler:
          - { cc: clang, cxx: clang++ }
          - { cc: gcc,   cxx: g++     }
@@ -99,9 +106,18 @@ jobs:
        exclude:
          # Build pull requests only with ubuntu-20.04 and without m32
          - os:  ${{ github.event_name == 'pull_request' && 'ubuntu-18.04' || 'do-not-exclude' }}
+          - os:  ${{ github.event_name == 'pull_request' && 'ubuntu-22.04' || 'do-not-exclude' }}
          - m32: ${{ github.event_name == 'pull_request' && 1              || 'do-not-exclude' }}
          # Build -m32 only on ubuntu-20.04
          - {os: ubuntu-18.04, m32: 1}
+          - {os: ubuntu-22.04, m32: 1}
+        include:
+          # Test with GCC 10 on ubuntu-20.04 without m32
+          - {os: ubuntu-20.04, compiler: { cc: gcc-10, cxx: g++-10 }, m32: 0, suite: dist-vlt-0}
+          - {os: ubuntu-20.04, compiler: { cc: gcc-10, cxx: g++-10 }, m32: 0, suite: dist-vlt-1}
+          - {os: ubuntu-20.04, compiler: { cc: gcc-10, cxx: g++-10 }, m32: 0, suite: dist-vlt-2}
+          - {os: ubuntu-20.04, compiler: { cc: gcc-10, cxx: g++-10 }, m32: 0, suite: vltmt-0}
+          - {os: ubuntu-20.04, compiler: { cc: gcc-10, cxx: g++-10 }, m32: 0, suite: vltmt-1}
    runs-on: ${{ matrix.os }}
    name: Test | ${{ matrix.os }} | ${{ matrix.compiler.cc }} | ${{ matrix.suite }} ${{ matrix.m32 && '| -m32' || '' }}
    env:
@@ -110,7 +126,7 @@ jobs:
      CI_M32: ${{ matrix.m32 }}
      CC: ${{ matrix.compiler.cc }}
      CXX: ${{ matrix.compiler.cxx }}
-      CACHE_BASE_KEY: test-${{ matrix.os }}-${{ matrix.compiler.cc }}-m32=${{ matrix.m32 }}-${ matrix.suite }}
+      CACHE_BASE_KEY: test-${{ matrix.os }}-${{ matrix.compiler.cc }}-m32=${{ matrix.m32 }}-${{ matrix.suite }}
      CCACHE_MAXSIZE: 64M # Per build matrix entry (2160M in total)
      VERILATOR_ARCHIVE: verilator-${{ github.sha }}-${{ matrix.os }}-${{ matrix.compiler.cc }}${{ matrix.m32 && '-m32' || '' }}.tar.gz
    steps:
@@ -0,0 +1,17 @@
+set noparent
+linelength=100
+filter=-build/c++11
+filter=-build/header_guard
+filter=-build/include_order
+filter=-build/include_subdir
+filter=-readability/inheritance
+filter=-readability/nolint
+filter=-readability/todo
+filter=-runtime/references
+filter=-whitespace/blank_line
+filter=-whitespace/empty_if_body
+filter=-whitespace/indent
+filter=-whitespace/newline
+filter=-whitespace/braces
+filter=-whitespace/parens
+filter=-whitespace/todo
@@ -8,6 +8,126 @@ The changes in each Verilator version are described below.  The
 contributors that suggested a given feature are shown in []. Thanks!


+Verilator 4.224 2022-06-19
+==========================
+
+**Major:**
+
+* VCD tracing is now parallelized with --threads (#3449). [Geza Lore, Shunyao CAD]
+
+**Minor:**
+
+* Add -f<optimization> options to replace -O<letter> options (#3436).
+* Changed --no-merge-const-pool to -fno-merge-const-pool (#3436).
+* Changed --no-decoration to remove output whitespace (#3460). [Kamil Rakoczy]
+* Support compile time trace signal selection with tracing_on/off (#3323). [Shunyao CAD]
+* Support non-ANSI interface port declarations (#3439). [Geza Lore, Shunyao CAD]
+* Support concat assignment to packed array (#3446).
+* Improve conditional merging optimization (#3125). [Geza Lore, Shunyao CAD]
+* Define VM_TRACE_VCD when tracing in VCD format. [Geza Lore, Shunyao CAD]
+* Add assert when VerilatedContext is mis-deleted (#3121). [Rupert Swarbrick]
+* Internal prep work towards timing control. [Krzysztof Bieganski]
+* Fix hang with large case statement optimization (#3405). [Mike Urbach]
+* Fix UNOPTFLAT warning from initial static var (#3406). [Kamil Rakoczy]
+* Fix compile error when enable VL_LEAK_CHECKS (#3411). [HungMingWu]
+* Fix cmake rules to support higher-level targets (#3377) (#3386). [Martin Stadler]
+* Fix BLKANDNBLK on $readmem/$writemem (#3379). [Alex Solomatnikov]
+* Fix 'with' operator with type casting (#3387). [xiak95]
+* Fix incorrect conditional merging (#3409). [Raynard Qiao]
+* Fix passing VL_TRACE_FST_WRITER_THREAD in CMake build. [Geza Lore, Shunyao CAD]
+* Fix compile error under strict C++11 mode (#3463). [Kevin Kiningham]
+* Fix public unpacked input ports (#3465). [Todd Strader]
+
+
+Verilator 4.222 2022-05-02
+==========================
+
+**Minor:**
+
+* Split --prof-threads into --prof-exec and --prof-pgo (#3365). [Geza Lore, Shunyao CAD]
+* Deprecate 'vluint64_t' and similar types (#3255).
+* Raise error on assignment to const in initial blocks. [Geza Lore, Shunyao CAD]
+* Issue INITIALDLY/COMBDLY/BLKSEQ warnings consistent with Verilator execution. [Geza Lore, Shunyao CAD]
+* Support LoongArch ISA multithreading (#3353) (#3354). [Xi Zhang]
+* Fix MSVC localtime_s (#3124).
+* Fix Bison 3.8.2 error (#3366). [elike-ypq]
+* Fix rare bug in -Oz (V3Localize) (#3286). [Geza Lore, Shunyao CAD]
+* Fix tracing interfaces inside interfaces (#3309). [Kevin Millis]
+* Fix filenames with dots overwriting debug .vpp files (#3373).
+* Fix including VK_USER_OBJS in make library (#3370) (#3382). [Julien Margetts]
+* Fix hang in generate symbol references (#3391) (#3398). [Yoda Lee]
+* Fix missing #include <memory> (#3392). [Aliaksei Chapyzhenka]
+* Fix crash in recursive module inlining (#3393). [david-sawatzke]
+* Fix --protect-ids mangling names of library methods. [Geza Lore, Shunyao CAD]
+* Fix foreach segmentation fault (#3400). [Kamil Rakoczy]
+
+
+Verilator 4.220 2022-03-12
+==========================
+
+**Minor:**
+
+* Removed the deprecated lint_off flag -msg; use -rule instead.
+* Removed the deprecated "fl" attribute in XML output; use "loc" attribute instead.
+* Suppress WIDTH warning on negate using carry bit (#3295). [Peter Monsson]
+* Add trace dumpvars() call for selective runtime tracing (#3322). [Shunyao CAD]
+* Add VERILATOR_VERSION_INTEGER for determining API (#3343). [Larry Doolittle]
+* Improve various V3Combine algorithm details (#3328). [Yutetsu TAKATSUKASA]
+* Improve various V3Order algorithm details. [Geza Lore]
+* Fix MacOS arm64 build (#3285) (#3291). [Guokai Chen]
+* Fix signed number operation (#3294) (#3308). [Raynard Qiao]
+* Fix FST traces to include vector range (#3296) (#3297). [Jamie Iles]
+* Fix skipping public enum values with four-state values (#3303).
+* Fix $readmem file not found to be warning not error (#3310). [Alexander Grobman]
+* Fix class stringification on wide arrays (#3312). [Iru Cai]
+* Fix $fscanf etc to return -1 on EOF (#3313). [Jose Tejada]
+* Fix public function arguments that are arrayed (#3316). [pawel256]
+* Fix unnamedblk error on foreach (#3321). [Aliaksei Chapyzhenka]
+* Fix crash in recursive module inlining (#3324). [Larry Doolittle]
+* Fix VL_RESTORER behavior on passing a lvalue reference (#3326). [HungMingWu]
+* Fix compile error with --trace-fst --sc (#3332). [leavinel]
+* Fix cast to array types (#3333). [Todd Strader]
+* Fix Vdeeptemp error with --threads and --compiler clang (#3338). [Per Karlsson]
+
+
+Verilator 4.218 2022-01-17
+==========================
+
+**Major:**
+
+* Primary inputs and outputs (VL_INW/VL_OUTW) now use VlWide type.
+  In general this should be backward compatible, but may lead to some
+  wrapper code needing changes.
+* Option --cdc is deprecated and is planned for removal,
+  file a bug if this is still being used.
+
+**Minor:**
+
+* Support class static members (#2233).
+* Support force/release (#2431) (#2593). [Shunyao CAD]
+* Add 'forceable' attribute to allow forcing from C++. (#3272). [Geza Lore, Shunyao CAD]
+* Support lower dimension looping in foreach loops (#3172). [Ehab Ibrahim]
+* Support up to 64 bit enums for .next/.prev/.name (#3244). [Alexander Grobman]
+* Reduce .rodata footprint of trace initialization (#3250). [Geza Lore, Shunyao CAD]
+* Support FST tracing in hierarchical Verilation (#3251). [Yutetsu TAKATSUKASA]
+* Use C++11 standard types for MacOS portability (#3254) (#3257). [Adrien Le Masle]
+* Fix make support for BSD ar (#2999) (#3256). [Julie Schwartz]
+* Fix bad ending address on $readmem (#3205). [Julie Schwartz]
+* Fix MSWIN compile error (#2681). [Unai Martinez-Corral]
+* Fix break under foreach loop (#3230).
+* Fix VL_STREAML_FAST_QQI with 64 bit left-hand-side (#3232) (#3235). [Adrien Le Masle]
+* Fix $sformat of inputs/outputs (#3236). [Adrien Le Masle]
+* Fix associative array first method as statement (#3228). [Adrien Le Masle]
+* Fix associative array foreach loop (#3229).
+* Fix $fclose not accepting expressions (#3237). [Julie Schwartz]
+* Fix $random not updating seed (#3238). [Julie Schwartz]
+* Fix top level param overwrite when package has same param (#3241) (#3247). [Adrien Le Masle]
+* Fix spurious UNUSED by ignoring inout pin connections (#3242). [Julie Schwartz]
+* Fix splitting of _eval and other top level functions. [Geza Lore, Shunyao CAD]
+* Fix internal error by inout port (#3258). [Yutetsu TAKATSUKASA]
+* Fix GCC 11 compile error (#3273). [HungMingWu]
+
+
 Verilator 4.216 2021-12-05
 ==========================

@@ -19,7 +139,7 @@ Verilator 4.216 2021-12-05
 **Minor:**

 * Internal code cleanups and improvements. [Geza Lore]
-* Improve --thread verilation-time performance.
+* Improve --thread Verilation-time performance.
 * Support task name in $display %m (#3211). [Julie Schwartz]
 * Make 'bit', 'logic' and 'time' types unsigned by default. [Geza Lore]
 * Optimize $random concatenates/selects (#3114).
@@ -43,7 +163,7 @@ Verilator 4.214 2021-10-17

 **Major:**

-* Add profile-guided optmization of mtasks (#3150).
+* Add profile-guided optimization of mtasks (#3150).

 **Minor:**

@@ -109,7 +229,7 @@ Verilator 4.210 2021-07-07

 * Add --prof-c to pass profiling to compiler (#3059). [Alexander Grobman]
 * Optimize a lot more model variables into function locals (#3027). [Geza Lore]
-* Support middle-of-design nested topmodules (#3026). [Dan Petrisko]
+* Support middle-of-design nested top modules (#3026). [Dan Petrisko]
 * Remove deprecated --no-relative-cfuncs option (#3024). [Geza Lore]
 * Remove deprecated --inhibit-sim option (#3035). [Geza Lore]
 * Merge const static data globally into a new constant pool (#3013). [Geza Lore]
@@ -148,10 +268,10 @@ Verilator 4.204 2021-06-12
 * Fix to emit 'else if' without nesting (#2944). [Geza Lore]
 * Fix part select issues in LATCH warning (#2948) (#2938). [Julien Margetts]
 * Fix to not emit empty files with low split limits (#2961). [Geza Lore]
-* Fix merging of assignments in C++ code (#2970). [Ruper Swarbrick]
+* Fix merging of assignments in C++ code (#2970). [Rupert Swarbrick]
 * Fix unused variable warnings (#2991). [Pieter Kapsenberg]
 * Fix --protect-ids when using SV classes (#2994). [Geza Lore]
-* Fix constant function calls with uninit value (#2995). [yanx21]
+* Fix constant function calls with uninitialized value (#2995). [yanx21]
 * Fix Makefiles to support Windows EXEEXT usage (#3008). [Miodrag Milanovic]


@@ -189,7 +309,7 @@ Verilator 4.202 2021-04-24
 * Fix Cygwin example compile issues (#2856). [Mark Shaw]
 * Fix select of with index variable (#2880). [Alexander Grobman]
 * Fix cmake version number to be numeric (#2881). [Yuri Victorovich]
-* Fix MinGW not supportting 'localtime_r' (#2882). [HyungKi Jeong]
+* Fix MinGW not supporting 'localtime_r' (#2882). [HyungKi Jeong]
 * Fix cast from packed, typedef'ed interface signal (#2884). [Todd Strader]
 * Fix VPI package reported as vpiModule (#2885). [Todd Strader]
 * Fix dumping waveforms to multiple FST files (#2889). [David Metz]
@@ -336,7 +456,7 @@ Verilator 4.102 2020-10-15
 * Support # as a comment in -f files (#2497). [phantom-killua]
 * Support 'this' (#2585). [Rafal Kapuscik]
 * Support defines for FST tracing (#2592). [Markus Krause]
-* Support non-overapping implication inside properties (#1292). [Peter Monsson]
+* Support non-overlapping implication inside properties (#1292). [Peter Monsson]
 * Fix timescale with --hierarchical (#2554). [Yutetsu TAKATSUKASA]
 * Fix cmake build with --hierarchical (#2560). [Yutetsu TAKATSUKASA]
 * Fix -G dropping public indication (#2561). [Andrew Goessling]
@@ -362,7 +482,7 @@ Verilator 4.100 2020-09-07
 * Support (with limitations) class extern, class extends, virtual class.
 * Support $urandom, $urandom_range without stability.
 * Support assume property. [Peter Monsson]
-* Support non-overapping implication inside properties (#1292). [Peter Monsson]
+* Support non-overlapping implication inside properties (#1292). [Peter Monsson]
 * Fix false DECLFILENAME on black-boxed modules (#2430). [Philipp Wagner]
 * Fix naming of "id : begin" blocks.
 * Fix class constructor error on assignments to const.
@@ -3762,7 +3882,7 @@ Verilator 0.0 1994-07-08
 Copyright
 =========

-Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -7,7 +7,7 @@
 #
 #*****************************************************************************
 #
-# Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -21,7 +21,7 @@ Welcome to Verilator
 .. list-table::

   * - **Welcome to Verilator, the fastest Verilog/SystemVerilog simulator.**
-        * Accepts synthesizable Verilog or SystemVerilog
+        * Accepts Verilog or SystemVerilog
        * Performs lint code-quality checks
        * Compiles into multithreaded C++, or SystemC
        * Creates XML to front-end your own tools
@@ -57,23 +57,24 @@ files, the "Verilated" code.

 The user writes a little C++/SystemC wrapper file, which instantiates the
 "Verilated" model of the user's top level module. These C++/SystemC files
-are then compiled by a C++ compiler (gcc/clang/MSVC++). The resulting
-executable performs the design simulation. Verilator also supports linking
-its generated libraries, optionally encrypted, into other simulators.
+are then compiled by a C++ compiler (gcc/clang/MSVC++). Executing the
+resulting executable performs the design simulation. Verilator also
+supports linking Verilated generated libraries, optionally encrypted, into
+other simulators.

 Verilator may not be the best choice if you are expecting a full featured
-replacement for NC-Verilog, VCS or another commercial Verilog simulator, or
-if you are looking for a behavioral Verilog simulator e.g. for a quick
-class project (we recommend `Icarus Verilog`_ for this.) However, if you
-are looking for a path to migrate SystemVerilog to C++ or SystemC, or your
-team is comfortable writing just a touch of C++ code, Verilator is the tool
-for you.
+replacement for Incisive, ModelSim/Questa, VCS or another commercial
+Verilog simulator, or if you are looking for a behavioral Verilog simulator
+e.g. for a quick class project (we recommend `Icarus Verilog`_ for this.)
+However, if you are looking for a path to migrate SystemVerilog to C++ or
+SystemC, or your team is comfortable writing just a touch of C++ code,
+Verilator is the tool for you.


 Performance
 ===========

-Verilator does not simply convert Verilog HDL to C++ or SystemC. Rather,
+Verilator does not directly translate Verilog HDL to C++ or SystemC. Rather,
 Verilator compiles your code into a much faster optimized and optionally
 thread-partitioned model, which is in turn wrapped inside a C++/SystemC
 module. The results are a compiled Verilog model that executes even on a
@@ -137,7 +138,7 @@ Related Projects
 Open License
 ============

-Verilator is Copyright 2003-2021 by Wilson Snyder. (Report bugs to
+Verilator is Copyright 2003-2022 by Wilson Snyder. (Report bugs to
 `Verilator Issues <https://verilator.org/issues>`_.)

 Verilator is free software; you can redistribute it and/or modify it under
@@ -1,7 +1,7 @@
 #!/usr/bin/env perl
 ######################################################################
 #
-# Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -319,6 +319,7 @@ detailed descriptions of these arguments.
     -f <file>                  Parse arguments from a file
     -FI <file>                 Force include of a file
    --flatten                   Force inlining of all modules, tasks and functions
+     -fno-<optimization>        Disable internal optimization stage
     -G<name>=<value>           Overwrite top-level parameter
    --gdb                       Run Verilator under GDB interactively
    --gdbbt                     Run Verilator under GDB for backtrace
@@ -344,11 +345,10 @@ detailed descriptions of these arguments.
    --MMD                       Create .d dependency files
    --MP                        Create phony dependency targets
    --Mdir <directory>          Name of output object directory
-    --no-merge-const-pool       Disable merging of different types in const pool
    --mod-prefix <topname>      Name to prepend to lower classes
    --no-clk <signal-name>      Prevent marking specified signal as clock
    --no-decoration             Disable comments and symbol decorations
-    --no-pins64                 Don't use vluint64_t's for 33-64 bit sigs
+    --no-pins64                 Don't use uint64_t's for 33-64 bit sigs
    --no-skip-identical         Disable skipping identical output
     +notimingchecks            Ignored
     -O0                        Disable optimizations
@@ -370,7 +370,8 @@ detailed descriptions of these arguments.
    --prefix <topname>          Name of top level class
    --prof-c                    Compile C++ code with profiling
    --prof-cfuncs               Name functions for profiling
-    --prof-threads              Enable generating gantt chart data for threads
+    --prof-exec                 Enable generating execution profile for gantt chart
+    --prof-pgo                  Enable generating profiling data for PGO
    --protect-key <key>         Key for symbol protection
    --protect-ids               Hash identifier names for obscurity
    --protect-lib <name>        Create a DPI protected library
@@ -404,7 +405,7 @@ detailed descriptions of these arguments.
    --trace-max-width <width>   Maximum array depth for tracing
    --trace-params              Enable tracing of parameters
    --trace-structs             Enable tracing structure names
-    --trace-threads <threads>   Enable waveform creation on separate threads
+    --trace-threads <threads>   Enable FST waveform creation on separate threads
    --trace-underscore          Enable tracing of _signals
     -U<var>                    Undefine preprocessor define
    --unroll-count <loops>      Tune maximum loop iterations
@@ -445,10 +446,10 @@ description of these arguments.
     +verilator+error+limit+<value>    Set error limit
     +verilator+help                   Display help
     +verilator+noassert               Disable assert checking
-     +verilator+prof+threads+file+<filename>  Set profile filename
-     +verilator+prof+threads+start+<value>    Set profile starting point
-     +verilator+prof+threads+window+<value>   Set profile duration
-     +verilator+prof+vlt+file+<filename>      Set profile guided filename
+     +verilator+prof+exec+file+<filename>  Set execution profile filename
+     +verilator+prof+exec+start+<value>    Set execution profile starting point
+     +verilator+prof+exec+window+<value>   Set execution profile duration
+     +verilator+prof+vlt+file+<filename>   Set PGO profile filename
     +verilator+rand+reset+<value>     Set random reset technique
     +verilator+seed+<value>           Set random seed
     +verilator+V                      Verbose version and config
@@ -459,7 +460,7 @@ description of these arguments.

 The latest version is available from L<https://verilator.org>.

-Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 redistribute it and/or modify the Verilator internals under the terms of
 either the GNU Lesser General Public License Version 3 or the Perl Artistic
 License Version 2.0.
@@ -14,7 +14,7 @@ parser = argparse.ArgumentParser(
    formatter_class=argparse.RawDescriptionHelpFormatter,
    description="""Report ccache behavior of a Verilated model build.""",
    epilog=
-    """Copyright 2002-2021 by Wilson Snyder. This program is free software; you
+    """Copyright 2002-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -1,7 +1,7 @@
 #!/usr/bin/env perl
 ######################################################################
 #
-# Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -187,7 +187,7 @@ L<https://verilator.org/guide/latest/exe_verilator_coverage.html>.

 The latest version is available from L<https://verilator.org>.

-Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify the Verilator internals under the terms
 of either the GNU Lesser General Public License Version 3 or the Perl
 Artistic License Version 2.0.
@@ -38,10 +38,10 @@ def diff_dir(a, b):

    anyfile = False
    for base in sorted(files.keys()):
+        if (not 'a' in files[base]) or (not 'b' in files[base]):
+            continue
        a = files[base]['a']
        b = files[base]['b']
-        if not a or not b:
-            continue
        print("=" * 70)
        print("= %s <-> %s" % (a, b))
        diff_file(a, b)
@@ -113,7 +113,7 @@ directories, ignoring irrelevant pointer differences.
 For documentation see
 https://verilator.org/guide/latest/exe_verilator_difftree.html

-Copyright 2005-2021 by Wilson Snyder. This program is free software; you
+Copyright 2005-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -9,7 +9,7 @@ import re
 import statistics
 # from pprint import pprint

-Threads = collections.defaultdict(lambda: {})
+Threads = collections.defaultdict(lambda: collections.defaultdict(lambda: {}))
 Mtasks = collections.defaultdict(lambda: {})
 Evals = collections.defaultdict(lambda: {})
 EvalLoops = collections.defaultdict(lambda: {})
@@ -30,12 +30,12 @@ def process(filename):

 def read_data(filename):
    with open(filename) as fh:
-        re_prof = re.compile(
-            r'^VLPROF mtask\s(\d+)\sstart\s(\d+)\selapsed\s(\d+)\spredict_start\s(\d+)\spredict_cost\s(\d+)\scpu\s(\d+)\son thread (\d+)'
-        )
-        re_eval = re.compile(r'^VLPROF eval\sstart\s(\d+)\selapsed\s(\d+)')
-        re_loop = re.compile(
-            r'^VLPROF eval_loop\sstart\s(\d+)\selapsed\s(\d+)')
+        re_thread = re.compile(r'^VLPROFTHREAD (\d+)$')
+        re_record = re.compile(r'^VLPROFEXEC (\S+) (\d+)(.*)$')
+        re_payload_mtaskBegin = re.compile(
+            r'id (\d+) predictStart (\d+) cpu (\d+)')
+        re_payload_mtaskEnd = re.compile(r'id (\d+) predictCost (\d+)')
+
        re_arg1 = re.compile(r'VLPROF arg\s+(\S+)\+([0-9.]*)\s*')
        re_arg2 = re.compile(r'VLPROF arg\s+(\S+)\s+([0-9.]*)\s*$')
        re_stat = re.compile(r'VLPROF stat\s+(\S+)\s+([0-9.]+)')
@@ -43,47 +43,60 @@ def read_data(filename):
        re_proc_cpu = re.compile(r'VLPROFPROC processor\s*:\s*(\d+)\s*$')
        re_proc_dat = re.compile(r'VLPROFPROC ([a-z_ ]+)\s*:\s*(.*)$')
        cpu = None
+        thread = None
+
+        lastEvalBeginTick = None
+        lastEvalLoopBeginTick = None

        for line in fh:
-            if re_prof.match(line):
-                match = re_prof.match(line)
-                mtask = int(match.group(1))
-                start = int(match.group(2))
-                elapsed_time = int(match.group(3))
-                end = start + elapsed_time
-                predict_start = int(match.group(4))
-                predict_cost = int(match.group(5))
-                cpu = int(match.group(6))
-                thread = int(match.group(7))
-                if start not in Threads[thread]:
-                    Threads[thread][start] = {}
-                Threads[thread][start]['mtask'] = mtask
-                Threads[thread][start]['end'] = end
-                Threads[thread][start]['cpu'] = cpu
-                Threads[thread][start]['predict_start'] = predict_start
-                Threads[thread][start]['predict_cost'] = predict_cost
-
-                if 'elapsed' not in Mtasks[mtask]:
-                    Mtasks[mtask] = {'end': 0, 'elapsed': 0}
-                Mtasks[mtask]['thread'] = thread
-                Mtasks[mtask]['elapsed'] += elapsed_time
-                Mtasks[mtask]['predict_start'] = predict_start
-                Mtasks[mtask]['predict_cost'] = predict_cost
-                Mtasks[mtask]['end'] = max(Mtasks[mtask]['end'], end)
-            elif re_eval.match(line):
-                match = re_eval.match(line)
-                start = int(match.group(1))
-                elapsed_time = int(match.group(2))
-                Evals[start]['start'] = start
-                Evals[start]['end'] = start + elapsed_time
-            elif re_loop.match(line):
-                match = re_loop.match(line)
-                start = int(match.group(1))
-                elapsed_time = int(match.group(2))
-                EvalLoops[start]['start'] = start
-                EvalLoops[start]['end'] = start + elapsed_time
-            elif re.match(r'^VLPROFTHREAD', line):
-                None  # pylint: disable=pointless-statement
+            recordMatch = re_record.match(line)
+            if recordMatch:
+                kind, tick, payload = recordMatch.groups()
+                tick = int(tick)
+                payload = payload.strip()
+                if kind == "EVAL_BEGIN":
+                    Evals[tick]['start'] = tick
+                    lastEvalBeginTick = tick
+                elif kind == "EVAL_END":
+                    Evals[lastEvalBeginTick]['end'] = tick
+                    lastEvalBeginTick = None
+                elif kind == "EVAL_LOOP_BEGIN":
+                    EvalLoops[tick]['start'] = tick
+                    lastEvalLoopBeginTick = tick
+                elif kind == "EVAL_LOOP_END":
+                    EvalLoops[lastEvalLoopBeginTick]['end'] = tick
+                    lastEvalLoopBeginTick = None
+                elif kind == "MTASK_BEGIN":
+                    mtask, predict_start, ecpu = re_payload_mtaskBegin.match(
+                        payload).groups()
+                    mtask = int(mtask)
+                    predict_start = int(predict_start)
+                    ecpu = int(ecpu)
+                    Threads[thread][tick]['mtask'] = mtask
+                    Threads[thread][tick]['predict_start'] = predict_start
+                    Threads[thread][tick]['cpu'] = ecpu
+                    if 'elapsed' not in Mtasks[mtask]:
+                        Mtasks[mtask] = {'end': 0, 'elapsed': 0}
+                    Mtasks[mtask]['begin'] = tick
+                    Mtasks[mtask]['thread'] = thread
+                    Mtasks[mtask]['predict_start'] = predict_start
+                elif kind == "MTASK_END":
+                    mtask, predict_cost = re_payload_mtaskEnd.match(
+                        payload).groups()
+                    mtask = int(mtask)
+                    predict_cost = int(predict_cost)
+                    begin = Mtasks[mtask]['begin']
+                    Threads[thread][begin]['end'] = tick
+                    Threads[thread][begin]['predict_cost'] = predict_cost
+                    Mtasks[mtask]['elapsed'] += tick - begin
+                    Mtasks[mtask]['predict_cost'] = predict_cost
+                    Mtasks[mtask]['end'] = max(Mtasks[mtask]['end'], tick)
+                elif Args.debug:
+                    print("-Unknown execution trace record: %s" % line)
+            elif re_thread.match(line):
+                thread = int(re_thread.match(line).group(1))
+            elif re.match(r'^VLPROF(THREAD|VERSION)', line):
+                pass
            elif re_arg1.match(line):
                match = re_arg1.match(line)
                Global['args'][match.group(1)] = match.group(2)
@@ -105,7 +118,7 @@ def read_data(filename):
                value = re.sub(r'\s+$', '', value)
                Global['cpuinfo'][cpu][term] = value
            elif re.match(r'^#', line):
-                None  # pylint: disable=pointless-statement
+                pass
            elif Args.debug:
                print("-Unk: %s" % line)
            # TODO -- this is parsing text printed by a client.
@@ -131,11 +144,12 @@ def report():
        plus = "+" if re.match(r'^\+', arg) else " "
        print("  %s%s%s" % (arg, plus, Global['args'][arg]))

-    nthreads = len(Threads)
+    nthreads = int(Global['stats']['threads'])
    Global['cpus'] = {}
    for thread in Threads:
        # Make potentially multiple characters per column
        for start in Threads[thread]:
+            if not Threads[thread][start]: continue
            cpu = Threads[thread][start]['cpu']
            elapsed = Threads[thread][start]['end'] - start
            if cpu not in Global['cpus']:
@@ -169,74 +183,79 @@ def report():
    print("\nAnalysis:")
    print("  Total threads             = %d" % nthreads)
    print("  Total mtasks              = %d" % len(Mtasks))
-    ncpus = len(Global['cpus'])
+    ncpus = max(len(Global['cpus']), 1)
    print("  Total cpus used           = %d" % ncpus)
-    print("  Total yields              = %d" % int(Global['stats']['yields']))
+    print("  Total yields              = %d" %
+          int(Global['stats'].get('yields', 0)))
    print("  Total evals               = %d" % len(Evals))
    print("  Total eval loops          = %d" % len(EvalLoops))
-    print("  Total eval time           = %d rdtsc ticks" %
-          Global['measured_last_end'])
-    print("  Longest mtask time        = %d rdtsc ticks" % long_mtask_time)
-    print("  All-thread mtask time     = %d rdtsc ticks" %
-          measured_mt_mtask_time)
-    long_efficiency = long_mtask_time / (Global.get('measured_last_end', 1)
-                                         or 1)
-    print("  Longest-thread efficiency = %0.1f%%" % (long_efficiency * 100.0))
-    mt_efficiency = measured_mt_mtask_time / (
-        Global.get('measured_last_end', 1) * nthreads or 1)
-    print("  All-thread efficiency     = %0.1f%%" % (mt_efficiency * 100.0))
-    print("  All-thread speedup        = %0.1f" % (mt_efficiency * nthreads))
-    if Global['rdtsc_cycle_time'] > 0:
-        ut = measured_mt_mtask_time / Global['rdtsc_cycle_time']
-        print("tot_mtask_cpu=" + measured_mt_mtask_time + " cyc=" +
-              Global['rdtsc_cycle_time'] + " ut=" + ut)
+    if Mtasks:
+        print("  Total eval time           = %d rdtsc ticks" %
+              Global['measured_last_end'])
+        print("  Longest mtask time        = %d rdtsc ticks" % long_mtask_time)
+        print("  All-thread mtask time     = %d rdtsc ticks" %
+              measured_mt_mtask_time)
+        long_efficiency = long_mtask_time / (Global.get(
+            'measured_last_end', 1) or 1)
+        print("  Longest-thread efficiency = %0.1f%%" %
+              (long_efficiency * 100.0))
+        mt_efficiency = measured_mt_mtask_time / (
+            Global.get('measured_last_end', 1) * nthreads or 1)
+        print("  All-thread efficiency     = %0.1f%%" %
+              (mt_efficiency * 100.0))
+        print("  All-thread speedup        = %0.1f" %
+              (mt_efficiency * nthreads))
+        if Global['rdtsc_cycle_time'] > 0:
+            ut = measured_mt_mtask_time / Global['rdtsc_cycle_time']
+            print("tot_mtask_cpu=" + measured_mt_mtask_time + " cyc=" +
+                  Global['rdtsc_cycle_time'] + " ut=" + ut)

-    predict_mt_efficiency = predict_mt_mtask_time / (
-        Global.get('predict_last_end', 1) * nthreads or 1)
-    print("\nPrediction (what Verilator used for scheduling):")
-    print("  All-thread efficiency     = %0.1f%%" %
-          (predict_mt_efficiency * 100.0))
-    print("  All-thread speedup        = %0.1f" %
-          (predict_mt_efficiency * nthreads))
+        predict_mt_efficiency = predict_mt_mtask_time / (
+            Global.get('predict_last_end', 1) * nthreads or 1)
+        print("\nPrediction (what Verilator used for scheduling):")
+        print("  All-thread efficiency     = %0.1f%%" %
+              (predict_mt_efficiency * 100.0))
+        print("  All-thread speedup        = %0.1f" %
+              (predict_mt_efficiency * nthreads))

-    p2e_ratios = []
-    min_p2e = 1000000
-    min_mtask = None
-    max_p2e = -1000000
-    max_mtask = None
+        p2e_ratios = []
+        min_p2e = 1000000
+        min_mtask = None
+        max_p2e = -1000000
+        max_mtask = None

-    for mtask in sorted(Mtasks.keys()):
-        if Mtasks[mtask]['elapsed'] > 0:
-            if Mtasks[mtask]['predict_cost'] == 0:
-                Mtasks[mtask]['predict_cost'] = 1  # don't log(0) below
-            p2e_ratio = math.log(Mtasks[mtask]['predict_cost'] /
-                                 Mtasks[mtask]['elapsed'])
-            p2e_ratios.append(p2e_ratio)
+        for mtask in sorted(Mtasks.keys()):
+            if Mtasks[mtask]['elapsed'] > 0:
+                if Mtasks[mtask]['predict_cost'] == 0:
+                    Mtasks[mtask]['predict_cost'] = 1  # don't log(0) below
+                p2e_ratio = math.log(Mtasks[mtask]['predict_cost'] /
+                                     Mtasks[mtask]['elapsed'])
+                p2e_ratios.append(p2e_ratio)

-            if p2e_ratio > max_p2e:
-                max_p2e = p2e_ratio
-                max_mtask = mtask
-            if p2e_ratio < min_p2e:
-                min_p2e = p2e_ratio
-                min_mtask = mtask
+                if p2e_ratio > max_p2e:
+                    max_p2e = p2e_ratio
+                    max_mtask = mtask
+                if p2e_ratio < min_p2e:
+                    min_p2e = p2e_ratio
+                    min_mtask = mtask

-    print("\nStatistics:")
-    print("  min log(p2e) = %0.3f" % min_p2e, end="")
-    print("  from mtask %d (predict %d," %
-          (min_mtask, Mtasks[min_mtask]['predict_cost']),
-          end="")
-    print(" elapsed %d)" % Mtasks[min_mtask]['elapsed'])
-    print("  max log(p2e) = %0.3f" % max_p2e, end="")
-    print("  from mtask %d (predict %d," %
-          (max_mtask, Mtasks[max_mtask]['predict_cost']),
-          end="")
-    print(" elapsed %d)" % Mtasks[max_mtask]['elapsed'])
+        print("\nMTask statistics:")
+        print("  min log(p2e) = %0.3f" % min_p2e, end="")
+        print("  from mtask %d (predict %d," %
+              (min_mtask, Mtasks[min_mtask]['predict_cost']),
+              end="")
+        print(" elapsed %d)" % Mtasks[min_mtask]['elapsed'])
+        print("  max log(p2e) = %0.3f" % max_p2e, end="")
+        print("  from mtask %d (predict %d," %
+              (max_mtask, Mtasks[max_mtask]['predict_cost']),
+              end="")
+        print(" elapsed %d)" % Mtasks[max_mtask]['elapsed'])

-    stddev = statistics.pstdev(p2e_ratios)
-    mean = statistics.mean(p2e_ratios)
-    print("  mean = %0.3f" % mean)
-    print("  stddev = %0.3f" % stddev)
-    print("  e ^ stddev = %0.3f" % math.exp(stddev))
+        stddev = statistics.pstdev(p2e_ratios)
+        mean = statistics.mean(p2e_ratios)
+        print("  mean = %0.3f" % mean)
+        print("  stddev = %0.3f" % stddev)
+        print("  e ^ stddev = %0.3f" % math.exp(stddev))

    report_cpus()

@@ -375,44 +394,45 @@ def write_vcd(filename):
            vcd['values'][eval_start][elcode] = n
            vcd['values'][eval_end][elcode] = None

-        # Predicted graph
-        for eval_start in EvalLoops:
-            eval_end = EvalLoops[eval_start]['end']
-            # Compute scale so predicted graph is of same width as eval
-            measured_scaling = (eval_end -
-                                eval_start) / Global['predict_last_end']
-            # Predict mtasks that fill the time the eval occupied
-            for mtask in Mtasks:
-                thread = Mtasks[mtask]['thread']
-                pred_scaled_start = eval_start + int(
-                    Mtasks[mtask]['predict_start'] * measured_scaling)
-                pred_scaled_end = eval_start + int(
-                    (Mtasks[mtask]['predict_start'] +
-                     Mtasks[mtask]['predict_cost']) * measured_scaling)
-                if pred_scaled_start == pred_scaled_end:
-                    continue
+        if Mtasks:
+            # Predicted graph
+            for eval_start in EvalLoops:
+                eval_end = EvalLoops[eval_start]['end']
+                # Compute scale so predicted graph is of same width as eval
+                measured_scaling = (eval_end -
+                                    eval_start) / Global['predict_last_end']
+                # Predict mtasks that fill the time the eval occupied
+                for mtask in Mtasks:
+                    thread = Mtasks[mtask]['thread']
+                    pred_scaled_start = eval_start + int(
+                        Mtasks[mtask]['predict_start'] * measured_scaling)
+                    pred_scaled_end = eval_start + int(
+                        (Mtasks[mtask]['predict_start'] +
+                         Mtasks[mtask]['predict_cost']) * measured_scaling)
+                    if pred_scaled_start == pred_scaled_end:
+                        continue

-                sig = "predicted_thread%d_mtask" % thread
-                if sig not in vcd['sigs']['predicted_threads']:
-                    vcd['sigs']['predicted_threads'][sig] = code
-                    code += 1
-                mcode = vcd['sigs']['predicted_threads'][sig]
+                    sig = "predicted_thread%d_mtask" % thread
+                    if sig not in vcd['sigs']['predicted_threads']:
+                        vcd['sigs']['predicted_threads'][sig] = code
+                        code += 1
+                    mcode = vcd['sigs']['predicted_threads'][sig]

-                vcd['values'][pred_scaled_start][mcode] = mtask
-                vcd['values'][pred_scaled_end][mcode] = None
+                    vcd['values'][pred_scaled_start][mcode] = mtask
+                    vcd['values'][pred_scaled_end][mcode] = None

-                parallelism['predicted'][pred_scaled_start] += 1
-                parallelism['predicted'][pred_scaled_end] -= 1
+                    parallelism['predicted'][pred_scaled_start] += 1
+                    parallelism['predicted'][pred_scaled_end] -= 1

-        # Parallelism graph
-        for measpred in ('measured', 'predicted'):
-            vcd['sigs']['Stats']["%s_parallelism" % measpred] = code
-            pcode = code
-            code += 1
-            value = 0
-            for time in sorted(parallelism[measpred].keys()):
-                value += parallelism[measpred][time]
-                vcd['values'][time][pcode] = value
+            # Parallelism graph
+            for measpred in ('measured', 'predicted'):
+                vcd['sigs']['Stats']["%s_parallelism" % measpred] = code
+                pcode = code
+                code += 1
+                value = 0
+                for time in sorted(parallelism[measpred].keys()):
+                    value += parallelism[measpred][time]
+                    vcd['values'][time][pcode] = value

        # Create output file
        fh.write("$version Generated by verilator_gantt $end\n")
@@ -463,7 +483,7 @@ parser = argparse.ArgumentParser(
 For documentation see
 https://verilator.org/guide/latest/exe_verilator_gantt.html

-Copyright 2018-2021 by Wilson Snyder. This program is free software; you
+Copyright 2018-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -476,10 +496,10 @@ parser.add_argument('--no-vcd',
                    action='store_true')
 parser.add_argument('--vcd',
                    help='filename for vcd outpue',
-                    default='profile_threads.vcd')
+                    default='profile_exec.vcd')
 parser.add_argument('filename',
-                    help='input profile_threads.dat filename to process',
-                    default='profile_threads.dat')
+                    help='input profile_exec.dat filename to process',
+                    default='profile_exec.dat')

 Args = parser.parse_args()

@@ -1,7 +1,7 @@
 #!/usr/bin/env perl
 # DESCRIPTION: Print include statements for each ARGV
 #
-# Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -182,7 +182,7 @@ in each Verilog block.
 For documentation see
 https://verilator.org/guide/latest/exe_verilator_profcfunc.html

-Copyright 2002-2021 by Wilson Snyder. This program is free software; you
+Copyright 2002-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -54,8 +54,12 @@ if [ "$CI_BUILD_STAGE_NAME" = "build" ]; then

  if [ "$CI_OS_NAME" = "linux" ]; then
    sudo apt-get update
-    sudo apt-get install libfl-dev libgoogle-perftools-dev ccache
-    if [ "$CI_RUNS_ON" = "ubuntu-20.04" ]; then
+    sudo apt-get install libfl-dev ccache
+    if [ "$CI_RUNS_ON" != "ubuntu-22.04" ]; then
+      # Some conflict of libunwind verison on 22.04, can live without it for now
+      sudo apt-get install libgoogle-perftools-dev
+    fi
+    if [ "$CI_RUNS_ON" = "ubuntu-20.04" ] || [ "$CI_RUNS_ON" = "ubuntu-22.04" ]; then
      sudo apt-get install libsystemc libsystemc-dev
    fi
    if [ "$COVERAGE" = 1 ]; then
@@ -85,7 +89,7 @@ elif [ "$CI_BUILD_STAGE_NAME" = "test" ]; then
    sudo apt-get update
    # libfl-dev needed for internal coverage's test runs
    sudo apt-get install gdb gtkwave lcov libfl-dev ccache
-    if [ "$CI_RUNS_ON" = "ubuntu-20.04" ]; then
+    if [ "$CI_RUNS_ON" = "ubuntu-20.04" ] || [ "$CI_RUNS_ON" = "ubuntu-22.04" ]; then
      sudo apt-get install libsystemc-dev
    fi
    if [ "$CI_M32" = 1 ]; then
@@ -1,6 +1,6 @@
 # DESCRIPTION: codecov.io config
 #
-# Copyright 2020-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2020-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -1,20 +1,20 @@
 # DESCRIPTION: Process this file with autoconf to produce a configure script.
 #
-# Copyright 2003-2021 by Wilson Snyder. Verilator is free software; you
+# Copyright 2003-2022 by Wilson Snyder. Verilator is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU Lesser
 # General Public License Version 3 or the Perl Artistic License Version 2.0.
 # SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 #AC_INIT([Verilator],[#.### YYYY-MM-DD])
 #AC_INIT([Verilator],[#.### devel])
-AC_INIT([Verilator],[4.216 2021-12-05],
+AC_INIT([Verilator],[4.224 2022-06-19],
        [https://verilator.org],
        [verilator],[https://verilator.org])
 # When releasing, also update header of Changes file
 # and commit using "devel release" or "Version bump" message
 # Then 'make maintainer-dist'

-AC_CONFIG_HEADER(src/config_build.h)
+AC_CONFIG_HEADERS(src/config_build.h)
 AC_CONFIG_FILES(Makefile src/Makefile src/Makefile_obj include/verilated.mk include/verilated_config.h verilator.pc verilator-config.cmake verilator-config-version.cmake)

 # Version
@@ -22,6 +22,8 @@ AC_MSG_RESULT([configuring for $PACKAGE_STRING])
 PACKAGE_VERSION_NUMBER=`AS_ECHO("$PACKAGE_VERSION") | sed 's/ .*//g'`
 AC_SUBST(PACKAGE_VERSION_NUMBER)
 AC_DEFINE_UNQUOTED([PACKAGE_VERSION_NUMBER_STRING],["$PACKAGE_VERSION_NUMBER"],[Package version as a number])
+VERILATOR_VERSION_INTEGER=`AS_ECHO("$PACKAGE_VERSION") | [sed 's/\([0-9]\)\.\([0-9][0-9][0-9]\) .*/\1\2000/g']`
+AC_SUBST(VERILATOR_VERSION_INTEGER)

 # Ignore automake flags passed by Ubuntu builds
 AC_ARG_ENABLE([dependency-tracking],
@@ -143,10 +145,10 @@ AC_SUBST(CFG_WITH_LONGTESTS)
 AC_MSG_RESULT($CFG_WITH_LONGTESTS)

 # Compiler flags (ensure they are not empty to avoid configure defaults)
-CFLAGS+=" "
-CPPFLAGS+=" "
-CXXFLAGS+=" "
-LDFLAGS+=" "
+CFLAGS="$CFLAGS "
+CPPFLAGS="$CPPFLAGS "
+CXXFLAGS="$CXXFLAGS "
+LDFLAGS="$LDFLAGS "

 # Checks for programs.
 AC_PROG_CC
@@ -346,14 +348,18 @@ AC_SUBST(CFG_CXXFLAGS_PROFILE)

 # Flag to select newest language standard supported
 # Macros work such that first option that passes is the one we take
-# Currently enabled c++14 due to packaged SystemC dependency
-# c++14 is the newest that Verilator is regressed to support
+# Currently enable c++17/c++14 due to packaged SystemC dependency
+# c++17 is the newest that Verilator is regularly tested to support
 # c++11 is the oldest that Verilator supports
 # gnu is requried for Cygwin to compile verilated.h successfully
 #_MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=gnu++20)
 #_MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=c++20)
-#_MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=gnu++17)
-#_MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=c++17)
+case "$(which lsb_release 2>&1 > /dev/null && lsb_release -d)" in
+*Ubuntu*22.04*)
+_MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=gnu++17)
+_MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=c++17)
+;;
+esac
 _MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=gnu++14)
 _MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=c++14)
 _MY_CXX_CHECK_SET(CFG_CXXFLAGS_STD_NEWEST,-std=gnu++11)
@@ -2,3 +2,4 @@
 *.html
 *.pdf
 _build
+guide/spelling.txt
@@ -3,12 +3,15 @@ under the Developer Certificate of Origin (https://developercertificate.org/).

 Please see the Verilator manual for 200+ additional contributors. Thanks to all.

+Adrien Le Masle
 Ahmed El-Mahmoudy
 Alex Chadwick
 Àlex Torregrosa
+Aliaksei Chapyzhenka
 Ameya Vikram Singh
 Andreas Kuster
 Chris Randall
+Chuxuan Wang
 Conor McCullough
 Dan Petrisko
 Daniel Bates
@@ -16,6 +19,7 @@ David Horton
 David Metz
 David Stanford
 David Turner
+Drew Ranck
 Drew Taussig
 Driss Hafdi
 Edgar E. Iglesias
@@ -27,9 +31,12 @@ Geza Lore
 Gianfranco Costamagna
 Glen Gibb
 Graham Rushton
+Guokai Chen
 Harald Heckmann
 Howard Su
 Huang Rui
+Huanghuang Zhou
+HungMingWu
 HyungKi Jeong
 Iru Cai
 Ivan Vnučec
@@ -38,6 +45,7 @@ James Hanlon
 James Hutchinson
 James Pallister
 Jamey Hicks
+Jamie Iles
 Jan Van Winkel
 Jean Berniolles
 Jeremy Bennett
@@ -48,11 +56,13 @@ Josh Redford
 Julie Schwartz
 Julien Margetts
 Kaleb Barrett
+Kamil Rakoczy
 Kanad Kanhere
 Keith Colbert
 Kevin Kiningham
 Krzysztof Bieganski
 Kuba Ober
+Larry Doolittle
 Ludwig Rogiers
 Lukasz Dalek
 Maarten De Braekeleer
@@ -62,6 +72,7 @@ Markus Krause
 Marlon James
 Marshal Qiao
 Martin Schmidt
+Martin Stadler
 Matthew Ballance
 Michael Killough
 Michaël Lefebvre
@@ -81,6 +92,7 @@ Pieter Kapsenberg
 Piotr Binkowski
 Qingyao Sun
 Rafal Kapuscik
+Raynard Qiao
 Richard Myers
 Rupert Swarbrick
 Samuel Riedel
@@ -103,6 +115,8 @@ Vassilis Papaefstathiou
 Veripool API Bot
 Victor Besyakov
 Wilson Snyder
+Xi Zhang
+Yoda Lee
 Yossi Nivin
 Yuri Victorovich
 Yutetsu TAKATSUKASA
@@ -5,7 +5,7 @@
 #
 # Code available from: https://verilator.org
 #
-# Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -35,6 +35,7 @@ default:
 # Intermediate rules

 vl-extract: ../bin/verilator ../Changes
+	ln -sf ../spelling.txt guide/spelling.txt
 	mkdir -p _build/gen
 	$(PYTHON3) bin/vl_sphinx_extract ../bin/verilator
 	sed 's/`/\&96;/g' < ../Changes > _build/gen/Changes
@@ -77,6 +78,7 @@ clean mostlyclean distclean maintainer-clean::
 	rm -f *.pg *.pgs *.toc *.tp *.tps *.vr *.vrs *.idx
 	rm -f *.ev *.evs *.ov *.ovs *.cv *.cvs *.ma *.mas
 	rm -f *.tex
+	rm -f guide/spelling.txt

 distclean maintainer-clean::
 	rm -f *.info* *.1 *.html *.pdf $(INFOS)
@@ -40,7 +40,7 @@ parser = argparse.ArgumentParser(
    formatter_class=argparse.RawDescriptionHelpFormatter,
    description="""Read a file and extract documentation data.""",
    epilog=
-    """ Copyright 2021-2021 by Wilson Snyder.  This package is free software;
+    """ Copyright 2021-2022 by Wilson Snyder.  This package is free software;
 you can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -55,7 +55,7 @@ parser = argparse.ArgumentParser(
    formatter_class=argparse.RawDescriptionHelpFormatter,
    description="""Post-process Sphinx HTML.""",
    epilog=
-    """ Copyright 2021-2021 by Wilson Snyder.  This package is free software;
+    """ Copyright 2021-2022 by Wilson Snyder.  This package is free software;
 you can redistribute it and/or modify it under the terms of either the GNU
 Lesser General Public License Version 3 or the Perl Artistic License
 Version 2.0.
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 ****************
@@ -36,7 +36,7 @@ def setup(app):
 # -- Project information

 project = 'Verilator'
-copyright = '2021 by Wilson Snyder, under LGPL-3.0 or Artistic-2.0'
+copyright = '2022 by Wilson Snyder, under LGPL-3.0 or Artistic-2.0'
 author = 'Wilson Snyder'

 # The master toctree document.
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Connecting:
@@ -96,7 +96,7 @@ model.  Here is a simple example:

         Vtop *top;                      // Instantiation of model

-         vluint64_t main_time = 0;       // Current simulation time
+         uint64_t main_time = 0;       // Current simulation time
         // This is a 64-bit integer to reduce wrap over issues and
         // allow modulus.  This is in units of the timeprecision
         // used in Verilog (or from --timescale-override)
@@ -110,6 +110,9 @@ model.  Here is a simple example:
             Verilated::commandArgs(argc, argv);   // Remember args

             top = new Vtop;             // Create model
+             // Do not instead make Vtop as a file-scope static
+             // variable, as the "C++ static initialization order fiasco"
+             // may cause a crash

             top->reset_l = 0;           // Set some inputs

@@ -150,7 +153,7 @@ netlist as an instantiation.
 The SC_MODULE gets the same pinout as the Verilog module, with the
 following type conversions: Pins of a single bit become bool.  Pins 2-32
 bits wide become uint32_t's.  Pins 33-64 bits wide become sc_bv's or
-vluint64_t's depending on the :vlopt:`--no-pins64` option.  Wider pins
+uint64_t's depending on the :vlopt:`--no-pins64` option.  Wider pins
 become sc_bv's.  (Uints simulate the fastest so are used where possible.)

 Model internals, including lower level sub-modules are not pure SystemC
@@ -200,7 +203,7 @@ DPI System Task/Functions
 -------------------------

 Verilator extends the DPI format to allow using the same scheme to
-efficiently add system functions.  Simply use a dollar-sign prefixed system
+efficiently add system functions.  Use a dollar-sign prefixed system
 function name for the import, but note it must be escaped.

 .. code-block:: sv
@@ -257,7 +260,9 @@ with respect to that top level module, then the scope could be set with

     #include "svdpi.h"
     ...
-     svSetScope(svGetScopeFromName("TOP.dut"));
+     const svScope scope = svGetScopeFromName("TOP.dut");
+     assert(scope);  // Check for nullptr if scope not found
+     svSetScope(scope);

 (Remember that Verilator adds a "TOP" to the top of the module hierarchy.)

@@ -422,12 +427,12 @@ accesses the above signal "readme" would be:

 .. code-block:: bash

-     cat >sim_main.cpp <<'<<EOF'
+     cat >sim_main.cpp <<'EOF'
       #include "Vour.h"
       #include "verilated.h"
       #include "verilated_vpi.h"  // Required to get definitions

-       vluint64_t main_time = 0;   // See comments in first example
+       uint64_t main_time = 0;   // See comments in first example
       double sc_time_stamp() { return main_time; }

       void read_and_check() {
@@ -506,7 +511,7 @@ structure.  If a ``VerilatedContext`` is not created prior to creating a
 model, a default global one is created automatically.

 The ``Verilated::`` methods, including the ``Verilated::commandArgs`` call
-shown above, simply call VerilatedContext methods using the default global
+shown above, call VerilatedContext methods using the default global
 VerilatedContext.  (Technically they operate on the last one used by a
 given thread.)  If you are using multiple simulation contexts you should
 not use the Verilated:: methods, and instead always use VerilatedContext
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 *******************************
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 ************************
@@ -23,14 +23,14 @@ Contributors
 Many people have provided ideas and other assistance with Verilator.

 Verilator is receiving major development support from the `CHIPS Alliance
-<https://chipsalliance.org>`_.
+<https://chipsalliance.org>`_ and `Shunyao CAD <https://shunyaocad.com>`_.

 Previous major corporate sponsors of Verilator, by providing significant
 contributions of time or funds included include: Atmel Corporation, Cavium
 Inc., Compaq Corporation, Digital Equipment Corporation, Embecosm Ltd.,
 Hicamp Systems, Intel Corporation, Mindspeed Technologies Inc., MicroTune
 Inc., picoChip Designs Ltd., Sun Microsystems Inc., Nauticus Networks Inc.,
-and SiCortex Inc.
+SiCortex Inc, and Shunyao CAD.

 The people who have contributed major functionality are: Byron Bradley,
 Jeremy Bennett, Lane Brooks, John Coiner, Duane Galbi, Geza Lore, Todd
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 *********
@@ -8,7 +8,7 @@ Copyright
 The latest version of Verilator is available from `https://verilator.org
 <https://verilator.org>`_.

-Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 can redistribute it and/or modify the Verilator internals under the terms
 of either the GNU Lesser General Public License Version 3 or the Perl
 Artistic License Version 2.0.
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 Deprecations
@@ -16,10 +16,22 @@ Verilated_heavy.h
  "verilated.h". Verilated_heavy.h is planned for removal no sooner than
  July 2022.

-Configuration File -msg
-  The :vlopt:`lint_off` "-msg" option has been replaced with the "-rule"
-  option.  "-msg" is planned for removal no sooner than January 2021.
+Option `--cdc`
+  The experimental `--cdc` option is believed to be generally unused and is
+  planned for removal no sooner than January 2023.

-XML locations
-  The XML "fl" attribute has been replaced with the "loc" attribute.  "fl"
-  is planned for removal no sooner than January 2021.
+Option `-O<letter>`
+  The debug `-O<letter>` options have been replaced with
+  `-fno-<optimization>` debug options to match GCC. The old options are
+  planned for removal no sooner than June 2023.
+
+Option `--prof-threads`
+  The `--prof-threads` option has been superseded by the `--prof-exec` and
+  `--prof-pgo` options and is planned for removal no sooner than April 2023.
+
+Verilated model options `+verilator+prof+threads+*`
+  The `+verilator+prof+threads+start`, `+verilator+prof+threads+window` and
+  `+verilator+prof+threads+file` options have been superseded by the
+  `+verilator+prof+exec+start`, `+verilator+prof+exec+window` and
+  `+verilator+prof+exec+file` options respectively and are planned for removal
+  no sooner than April 2023.
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 Environment
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Example C++ Execution:
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 First you need Verilator installed, see :ref:`Installation`. In brief, if
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Examples in the Distribution:
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Example SystemC Execution:
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Examples:
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Simulation Runtime Arguments:
@@ -38,33 +38,45 @@ Summary:

   Display help and exit.

-.. option:: +verilator+prof+threads+file+<filename>
+.. option:: +verilator+prof+exec+file+<filename>

-   When a model was Verilated using :vlopt:`--prof-threads`, sets the
+   When a model was Verilated using :vlopt:`--prof-exec`, sets the
   simulation runtime filename to dump to.  Defaults to
-   :file:`profile_threads.dat`.
+   :file:`profile_exec.dat`.

-.. option:: +verilator+prof+threads+start+<value>
+.. option:: +verilator+prof+exec+start+<value>

-   When a model was Verilated using :vlopt:`--prof-threads`, the simulation
+   When a model was Verilated using :vlopt:`--prof-exec`, the simulation
   runtime will wait until $time is at this value (expressed in units of
   the time precision), then start the profiling warmup, then
   capturing. Generally this should be set to some time that is well within
   the normal operation of the simulation, i.e. outside of reset. If 0, the
   dump is disabled. Defaults to 1.

-.. option:: +verilator+prof+threads+window+<value>
+.. option:: +verilator+prof+exec+window+<value>

-   When a model was Verilated using :vlopt:`--prof-threads`, after $time
-   reaches :vlopt:`+verilator+prof+threads+start+\<value\>`, Verilator will
+   When a model was Verilated using :vlopt:`--prof-exec`, after $time
+   reaches :vlopt:`+verilator+prof+exec+start+\<value\>`, Verilator will
   warm up the profiling for this number of eval() calls, then will capture
   the profiling of this number of eval() calls.  Defaults to 2, which
   makes sense for a single-clock-domain module where it's typical to want
   to capture one posedge eval() and one negedge eval().

+.. option:: +verilator+prof+threads+file+<filename>
+
+   Deprecated. Alias for :vlopt:`+verilator+prof+exec+file+\<filename\>`
+
+.. option:: +verilator+prof+threads+start+<value>
+
+   Deprecated. Alias for :vlopt:`+verilator+prof+exec+start+\<value\>`
+
+.. option:: +verilator+prof+threads+window+<value>
+
+   Deprecated. Alias for :vlopt:`+verilator+prof+exec+window+\<filename\>`
+
 .. option:: +verilator+prof+vlt+file+<filename>

-   When a model was Verilated using :vlopt:`--prof-threads`, sets the
+   When a model was Verilated using :vlopt:`--prof-pgo`, sets the
   profile-guided optimization data runtime filename to dump to.  Defaults
   to :file:`profile.vlt`.

@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 verilator Arguments
@@ -99,10 +99,10 @@ Summary:
 .. option:: --bbox-sys

   Black box any unknown $system task or function calls.  System tasks will
-   simply become no-operations, and system functions will be replaced with
-   unsized zero.  Arguments to such functions will be parsed, but not
-   otherwise checked.  This prevents errors when linting in the presence of
-   company specific PLI calls.
+   become no-operations, and system functions will be replaced with unsized
+   zero.  Arguments to such functions will be parsed, but not otherwise
+   checked.  This prevents errors when linting in the presence of company
+   specific PLI calls.

   Using this argument will likely cause incorrect simulation.

@@ -431,6 +431,54 @@ Summary:
   flattening large designs may require significant CPU time, memory and
   storage.

+.. option:: -fno-acyc-simp
+
+.. option:: -fno-assemble
+
+.. option:: -fno-case
+
+.. option:: -fno-combine
+
+.. option:: -fno-const
+
+.. option:: -fno-const-bit-op-tree
+
+.. option:: -fno-dedup
+
+.. option:: -fno-expand
+
+.. option:: -fno-gate
+
+.. option:: -fno-inline
+
+.. option:: -fno-life
+
+.. option:: -fno-life-post
+
+.. option:: -fno-localize
+
+.. option:: -fno-merge-cond
+
+.. option:: -fno-merge-cond-motion
+
+.. option:: -fno-merge-const-pool
+
+.. option:: -fno-reloop
+
+.. option:: -fno-reorder
+
+.. option:: -fno-split
+
+.. option:: -fno-subst
+
+.. option:: -fno-subst-const
+
+.. option:: -fno-table
+
+   Rarely needed. Disables one of the internal optimization steps. These
+   are typically used only when recommended by a maintainer to help debug
+   or work around an issue.
+
 .. option:: -G<name>=<value>

   Overwrites the given parameter of the toplevel module. The value is
@@ -648,13 +696,6 @@ Summary:
   The directory is created if it does not exist and the parent directories
   exist; otherwise manually create the Mdir before calling Verilator.

-.. option:: --no-merge-const-pool
-
-   Rarely needed.  In order to minimize cache footprint, values of different
-   data type, that are yet emitted identically in C++ are merged in the
-   constant pool.  This option disables this and causes every constant pool
-   entry with a distinct data type to be emitted separately.
-
 .. option:: --mod-prefix <topname>

   Specifies the name to prepend to all lower level classes.  Defaults to
@@ -703,9 +744,9 @@ Summary:

   Rarely needed.  Enables or disables a specific optimizations, with the
   optimization selected based on the letter passed.  A lowercase letter
-   disables an optimization, an upper case letter enables it.  This is
-   intended for debugging use only; see the source code for
-   version-dependent mappings of optimizations to -O letters.
+   disables an optimization, an upper case letter enables it.  This option
+   is deprecated and the various `-f<optimization>` arguments should be
+   used instead.

 .. option:: -o <executable>

@@ -768,7 +809,7 @@ Summary:
 .. option:: --pins-bv <width>

   Specifies SystemC inputs/outputs of greater than or equal to <width>
-   bits wide should use sc_bv's instead of uint32/vluint64_t's.  The
+   bits wide should use sc_bv's instead of uint32/uint64_t's.  The
   default is "--pins-bv 65", and the value must be less than or equal
   to 65.  Versions before Verilator 3.671 defaulted to "--pins-bv 33".
   The more sc_bv is used, the worse for performance.  Use the
@@ -845,10 +886,19 @@ Summary:

   Using :vlopt:`--prof-cfuncs` also enables :vlopt:`--prof-c`.

+.. option:: --prof-exec
+
+   Enable collection of execution trace, that can be converted into a gantt
+   chart with verilator_gantt See :ref:`Execution Profiling`.
+
+.. option:: --prof-pgo
+
+   Enable collection of profiling data for profile guided Verilation. Currently
+   this is only useful with :vlopt:`--threads`. See :ref:`Thread PGO`.
+
 .. option:: --prof-threads

-   Enable gantt chart data collection for threaded builds. See :ref:`Thread
-   Profiling` and :ref:`Thread PGO`.
+   Deprecated. Same as --prof-exec and --prof-pgo together.

 .. option:: --protect-key <key>

@@ -884,7 +934,7 @@ Summary:

   Using DPI imports/exports is allowed and generally relatively safe in
   terms of information disclosed, which is limited to the DPI function
-   prototyptes.  Use of the VPI is not recommended as many design details
+   prototypes.  Use of the VPI is not recommended as many design details
   may be exposed, and an INSECURE warning will be issued.

 .. option:: --protect-lib <name>
@@ -1032,7 +1082,8 @@ Summary:
   is not thread safe. With "--threads 1", the generated model is single
   threaded but may run in a multithreaded environment. With "--threads N",
   where N >= 2, the model is generated to run multithreaded on up to N
-   threads. See :ref:`Multithreading`.
+   threads. See :ref:`Multithreading`. This option also applies to
+   :vlopt:`--trace` (but not :vlopt:`--trace-fst`).

 .. option:: --threads-dpi all

@@ -1110,7 +1161,8 @@ Summary:
   Having tracing compiled in may result in some small performance losses,
   even when tracing is not turned on during model execution.

-   See also :vlopt:`--trace-threads` option.
+   When using :vlopt:`--threads`, VCD tracing is parallelized, using the
+   same number of threads as passed to :vlopt:`--threads`.

 .. option:: --trace-coverage

@@ -1164,12 +1216,12 @@ Summary:
 .. option:: --trace-threads *threads*

   Enable waveform tracing using separate threads. This is typically faster
-   in simulation runtime but uses more total compute. This option is
-   independent of, and works with, both :vlopt:`--trace` and
-   :vlopt:`--trace-fst`.  Different trace formats can take advantage of
-   more trace threads to varying degrees. Currently VCD tracing can utilize
-   at most "--trace-threads 1", and FST tracing can utilize at most
-   "--trace-threads 2". This overrides :vlopt:`--no-threads` .
+   in simulation runtime but uses more total compute. This option only
+   applies to :vlopt:`--trace-fst`. FST tracing can utilize at most
+   "--trace-threads 2". This overrides :vlopt:`--no-threads`.
+
+   This option is accepted, but has absolutely no effect with
+   :vlopt:`--trace`, which respects :vlopt:`--threads` instead.

 .. option:: --trace-underscore

@@ -1364,9 +1416,10 @@ Summary:
   .. note::

      This option applies only to values which are explicitly written as X
-      in the Verilog source code. Initial values of clocks are set to 0
-      unless `--x-initial-edge` is specified. Initial values of all other
-      state holding variables are controlled with `--x-initial`.
+      in modules (not classes) in the Verilog source code. Initial values
+      of clocks are set to 0 unless `--x-initial-edge` is
+      specified. Initial values of all other state holding variables are
+      controlled with `--x-initial`.

 .. option:: --x-initial 0

@@ -1513,7 +1566,7 @@ The grammar of configuration commands is as follows:
 .. option:: no_clocker -module "<modulename>" [-function "<funcname>"] -var "<signame>"

   Indicates that the signal is used as clock or not. This information is
-   used by Verilator to mark the signal and any derrived signals as
+   used by Verilator to mark the signal and any derived signals as
   clocker.  See :vlopt:`--clk`.

   Same as :option:`/*verilator&32;clocker*/` metacomment.
@@ -1528,6 +1581,14 @@ The grammar of configuration commands is as follows:

   Same as :option:`/*verilator&32;coverage_block_off*/` metacomment.

+.. option:: forceable -module "<modulename>" -var "<signame>"
+
+   Generate public `<signame>__VforceEn` and `<signame>__VforceVal` signals
+   that can be used to force/release a signal from C++ code. The force control
+   signals are created as :option:`public_flat` signals.
+
+   Same as :option:`/*verilator&32;forceable*/` metacomment.
+
 .. option:: full_case -file "<filename>" -lines <lineno>

 .. option:: parallel_case -file "<filename>" -lines <lineno>
@@ -1629,7 +1690,7 @@ The grammar of configuration commands is as follows:
 .. option:: sc_bv -module "<modulename>" [-function "<funcname>"] -var "<signame>"

   Sets the port to be of :code:`sc_bv<{width}>` type, instead of bool,
-   vluint32_t or vluint64_t.  Same as :option:`/*verilator&32;sc_bv*/`
+   uint32_t or uint64_t.  Same as :option:`/*verilator&32;sc_bv*/`
   metacomment.

 .. option:: sformat [-module "<modulename>"] [-task "<taskname>"] -var "<signame>"
@@ -1658,9 +1719,28 @@ The grammar of configuration commands is as follows:

 .. option:: tracing_off [-file "<filename>" [-lines <line> [ - <line> ]]]

-   Enable/disable waveform tracing for all future signals declared in the
-   specified filename (or wildcard with '\*' or '?', or all files if
-   omitted) and range of line numbers (or all lines if omitted).
+.. option:: tracing_on  [-scope "<scopename>" [-levels <levels> ]]

-   For tracing_off, instances below any module in the files/ranges
-   specified will also not be traced.
+.. option:: tracing_off [-scope "<scopename>" [-levels <levels> ]]
+
+   Enable/disable waveform tracing for all future signals declared in
+   all files.
+
+   With -file, enable/disable waveform tracing in the specified
+   filename (or wildcard with '\*' or '?'), and -line range of line
+   numbers (or all lines if omitted).
+
+   For tracing_off with -file, instances below any module in the
+   files/ranges specified will also not be traced.  To overcome this
+   feature, use tracing_on on the upper module declaration and on any
+   cells, or use the -scope flavor of the command.
+
+   With -scope enable/disable waveform tracing for the specified scope (or
+   wildcard with '\*' or '?'), and optional --levels number of levels
+   below.  These controls only take place after other file/line/module
+   based controls have indicated the signal should be traced.
+
+   With -levels (used with -scope), the number of levels below that
+   scope which the rule is to match, where 0 means all levels below, 1
+   the exact level as the provided scope, and 2 meaning an additional
+   level of children below the provided scope, etc.
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 verilator_coverage
@@ -67,7 +67,7 @@ which have low coverage are written to the output directory.

 Specifies if the coverage point does not include the count number of
 coverage hits, then the coverage point will be considered above the
-threashold, and the coverage report will put a "%" to indicate the coverage
+threshold, and the coverage report will put a "%" to indicate the coverage
 is not sufficient.  Defaults to 10.

 .. option:: --help
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 verilator_gantt
@@ -13,7 +13,7 @@ For an overview of use of verilator_gantt, see :ref:`Profiling`.
 Gantt Chart VCD
 ---------------

-Verilated_gnatt creates a value change dump (VCD) format dump file which
+Verilated_gantt creates a value change dump (VCD) format dump file which
 may be viewed in a waveform viewer (e.g. C<GTKWave>):

 .. figure:: figures/fig_gantt_min.png
@@ -72,7 +72,7 @@ verilator_gantt Arguments

 .. option:: <filename>

-The filename to read data from, defaults to "profile_threads.dat".
+The filename to read data from, defaults to "profile_exec.dat".

 .. option:: --help

@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 verilator_profcfunc
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 *********************************
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 *******************
@@ -185,7 +185,7 @@ or "`ifdef`"'s may break other tools.
 .. option:: /*verilator&32;no_clocker*/

   Specifies that the signal is used as clock or not. This information is
-   used by Verilator to mark the signal and any derrived signals as
+   used by Verilator to mark the signal and any derived signals as
   clocker.  See :vlopt:`--clk`.

   Same as :option:`clocker` and :option:`no_clocker` in configuration
@@ -209,6 +209,20 @@ or "`ifdef`"'s may break other tools.
   (if appropriate :vlopt:`--coverage` flags are passed) after being
   disabled earlier with :option:`/*verilator&32;coverage_off*/`.

+.. option:: /*verilator&32;forceable*/
+
+   Specifies that the signal (net or variable) should be made forceable from
+   C++ code by generating public `<signame>__VforceEn` and
+   `<signame>__VforceVal` signals The force control signals are created as
+   :option:`public_flat` signals.
+
+   To force a marked signal from C++, set the corresponding `__VforceVal`
+   variable to the desired value, and the `__VforceEn` signal to the bit-mask
+   indicating which bits of the signal to force. To force all bits of the
+   target signal, set `__VforceEn` to all ones. To release the signal (or part
+   thereof), set appropriate bits of the `__VforceEn` signal to zero.
+
+   Same as :option:`forceable` in configuration  files.

 .. _verilator_hier_block:

@@ -431,7 +445,7 @@ or "`ifdef`"'s may break other tools.
 .. option:: /*verilator&32;sc_bv*/

   Used after a port declaration.  It sets the port to be of
-   :code:`sc_bv<{width}>` type, instead of bool, vluint32_t or vluint64_t.
+   :code:`sc_bv<{width}>` type, instead of bool, uint32_t or uint64_t.
   This may be useful if the port width is parameterized and the
   instantiating C++ code wants to always have a sc_bv so it can accept any
   width.  In general you should avoid using this attribute when not
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 ******************************
@@ -37,15 +37,15 @@ Does Verilator run under Windows?

 Yes, ideally run Ubuntu under Windows Subsystem for Linux (WSL2).
 Alternatively use Cygwin, though this tends to be slower and is not
-regurally tested.  Verilated output also compiles under Microsoft Visual
-C++, but this is also not regurally tested.
+regularly tested.  Verilated output also compiles under Microsoft Visual
+C++, but this is also not regularly tested.


 Can you provide binaries?
 """""""""""""""""""""""""

 You can install Verilator via the system package manager (apt, yum, etc.)
-on many Linux distributions, including Debian, Ubuntu, SuSE, RedHat, and
+on many Linux distributions, including Debian, Ubuntu, SuSE, Red Hat, and
 others.  These packages are provided by the Linux distributions and
 generally will lag the version of the mainline Verilator repository.  If no
 binary package is available for your distribution, how about you set one
@@ -72,23 +72,38 @@ a good thing for getting working silicon.
 Will Verilator output remain under my own license/copyright?
 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""

-Yes, it's just like using GCC on your programs; this is why Verilator uses
-the "GNU **Lesser** Public License Version 3" instead of the more typical
-"GNU Public License".  See the licenses for details, but in brief, if you
-change Verilator itself or the header files Verilator includes, you must
-make the source code available under the GNU Lesser Public License.
-However, Verilator output (the Verilated code) only "include"s the licensed
-files, and so you are **not** required to open-source release any output
-from Verilator.
+Your SystemVerilog, VPI/DPI, or main() C++ code remains under your own license.
+
+It's just like how using GCC on your programs does not change the copyright
+of your program; this is why Verilator uses the "GNU **Lesser** Public
+License Version 3" instead of the more typical "GNU Public License".  See
+the licenses for details.
+
+Some examples:
+
+* Any SystemVerilog or other input fed into Verilator remain your own.
+
+* Any of your VPI/DPI C++ routines that Verilator calls remain your own.
+
+* Any of your main() C++ code that calls into Verilator remain your own.
+
+* If you change Verilator itself, for example changing or adding a file
+  under the src/ directory in the repository, you must make the source code
+  available under the GNU Lesser Public License.
+
+* If you change a header Verilator provides, for example under include/ in
+  the repository, you must make the source code available under the GNU
+  Lesser Public License.

 You also have the option of using the Perl Artistic License, which again
-does not require you to release your Verilog or generated code, and also
-allows you to modify Verilator for internal use without distributing the
-modified version.  But please contribute back to the community!
+does not require you to release your Verilog, C++, or generated code. This
+license also allows you to modify Verilator for internal use without
+distributing the modified version.  But please contribute back to the
+community!

-One limit is that you cannot under either license release a closed-source
-Verilog simulation product incorporating Verilator. That is you can have a
-commercial product, but must make the source code available.
+Under both license you can offer a commercial product that is based on
+Verilator either directly or embedded within.  However under both licenses,
+any changes you make to Verilator for such a product must be open sourced.

 As is standard with Open Source, contributions back to Verilator will be
 placed under the Verilator copyright and LGPL/Artistic license.  Small test
@@ -134,7 +149,8 @@ B. Or, for finer-grained control, or C++ files with multiple Verilated
          ...
          Verilated::traceEverOn(true);
          VerilatedVcdC* tfp = new VerilatedVcdC;
-          topp->trace(tfp, 99);  // Trace 99 levels of hierarchy
+          topp->trace(tfp, 99);  // Trace 99 levels of hierarchy (or see below)
+          // tfp->dumpvars(1, "t");  // trace 1 level under "t"
          tfp->open("obj_dir/t_trace_ena_cc/simx.vcd");
          ...
          while (contextp->time() < sim_time && !contextp->gotFinish()) {
@@ -353,8 +369,8 @@ also use the "import DPI" SystemVerilog feature to call C code (see the
 chapter above).  There is also limited VPI access to public signals.

 If you want something more complex, since Verilator emits standard C++
-code, you can simply write your own C++ routines that can access and modify
-signal values without needing any PLI interface code, and call it with
+code, you can write your own C++ routines that can access and modify signal
+values without needing any PLI interface code, and call it with
 $c("{any_c++_statement}").

 See the :ref:`Connecting` section.
@@ -444,7 +460,7 @@ the signal, as you would any other member variable.
 Signals are the smallest of 8-bit unsigned chars (equivalent to uint8_t),
 16-bit unsigned shorts (uint16_t), 32-bit unsigned longs (uint32_t), or
 64-bit unsigned long longs (uint64_t) that fits the width of the signal.
-Generally, you can use just uint32_t's for 1 to 32 bits, or vluint64_t for
+Generally, you can use just uint32_t's for 1 to 32 bits, or uint64_t for
 1 to 64 bits, and the compiler will properly up-convert smaller entities.
 Note even signed ports are declared as unsigned; you must sign extend
 yourself to the appropriate signal width.
@@ -481,7 +497,7 @@ by your code or you'll get strange results.
 Should a module be in Verilog or SystemC?
 """""""""""""""""""""""""""""""""""""""""

-Sometimes there is a block that just interconnects instances, and have a
+Sometimes there is a block that only interconnects instances, and have a
 choice as to if you write it in Verilog or SystemC.  Everything else being
 equal, best performance is when Verilator sees all of the design.  So, look
 at the hierarchy of your design, labeling instances as to if they are
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 *****
@@ -120,7 +120,7 @@ In certain debug and other modes, it also creates:
     - Debugging graph files (from --debug)
   * - *{prefix}{misc}*\ .tree
     - Debugging files (from --debug)
-   * - {mod_prefix}_{each_verilog_module}*{__n}*\ .vpp
+   * - {mod_prefix}_{each_verilog_base_filename}*\ .vpp
     - Pre-processed verilog (from --debug)

 After running Make, the C++ compiler may produce the following:
@@ -155,13 +155,13 @@ The Verilated executable may produce the following:
   * - gmon.out
     - GCC/clang code profiler output, often fed into :command:`verilator_profcfunc`
   * - profile.vlt
-     - -profile data file for :ref:`Thread PGO`
-   * - profile_threads.dat
-     - -profile-threads data file for :command:`verilator_gnatt`
+     - --prof-pgo data file for :ref:`Thread PGO`
+   * - profile_exec.dat
+     - --prof-exec data file for :command:`verilator_gantt`

 Verilator_gantt may produce the following:

 .. list-table::

-   * - profile_threads.vcd
+   * - profile_exec.vcd
     - Gantt report waveform output
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 ######################
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Installation:
@@ -77,7 +77,7 @@ OS Requirements
 ---------------

 Verilator is developed and has primary testing on Ubuntu, with additional
-testing on FreeBSD and Apple OS-X. Versions have also built on Redhat
+testing on FreeBSD and Apple OS-X. Versions have also built on Red Hat
 Linux, and other flavors of GNU/Linux-ish platforms. Verilator also works
 on Windows Subsystem for Linux (WSL2), Windows under Cygwin, and Windows
 under MinGW (gcc -mno-cygwin). Verilated output (not Verilator itself)
@@ -87,7 +87,7 @@ compiles under all the options above, plus using MSVC++.
 Install Prerequisites
 ---------------------

-To build or run Verilator you need these standard packages:
+To build or run Verilator, you need these standard packages:

 ::

@@ -98,13 +98,20 @@ To build or run Verilator you need these standard packages:
   sudo apt-get install libfl-dev  # Ubuntu only (ignore if gives error)
   sudo apt-get install zlibc zlib1g zlib1g-dev  # Ubuntu only (ignore if gives error)

-To build or run the following are optional but should be installed for good
-performance:
+To build or run Verilator, the following are optional but should be installed
+for good performance:

 ::

   sudo apt-get install ccache  # If present at build, needed for run
-   sudo apt-get install libgoogle-perftools-dev numactl perl-doc
+   sudo apt-get install libgoogle-perftools-dev numactl
+
+The following is optional but is recommended for nicely rendered command line
+help when running Verilator:
+
+::
+
+   sudo apt-get install perl-doc

 To build Verilator you will need to install these packages; these do not
 need to be present to run Verilator:
@@ -118,7 +125,8 @@ Those developing Verilator itself may also want these (see internals.rst):
 ::

   sudo apt-get install gdb graphviz cmake clang clang-format-11 gprof lcov
-   sudo pip3 install sphinx sphinx_rtd_theme breathe
+   sudo apt-get install yapf3
+   sudo pip3 install sphinx sphinx_rtd_theme sphinxcontrib-spelling breathe
   cpan install Pod::Perldoc
   cpan install Parallel::Forker

@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 ***************
@@ -92,7 +92,7 @@ appropriate code to detect failing cases at simulation runtime and print an

 Verilator likewise also asserts any "unique" or "priority" SystemVerilog
 keywords on case statement, as well as "unique" on if statements.  However,
-"priority if" is currently simply ignored.
+"priority if" is currently ignored.


 .. _Language Limitations:
@@ -136,7 +136,7 @@ ___05F (5F is the hex code of an underscore.)
 Bind
 ----

-sVerilator only supports bind to a target module name, not to an
+Verilator only supports bind to a target module name, not to an
 instance path.


@@ -174,9 +174,9 @@ Structures and Unions
 ---------------------

 Presently Verilator only supports packed structs and packed unions.  Rand
-and randc tags on members are simply ignored.  All structures and unions
-are represented as a single vector, which means that generating one member
-of a structure from blocking, and another from non-blocking assignments is
+and randc tags on members are ignored.  All structures and unions are
+represented as a single vector, which means that generating one member of a
+structure from blocking, and another from non-blocking assignments is
 unsupported.


@@ -336,6 +336,24 @@ Verilator does not support SEREs yet.  All assertion and coverage
 statements must be simple expressions that complete in one cycle.


+Force statement
+---------------
+
+Verilator supports the procedural `force` (and corresponding `release`)
+statement. The behavior of the `force` statement however does not entirely
+comply with the IEEE 1800 SystemVerilog standard. According to the standard,
+when a procedural statement of the form `force a = b;` is executed, the
+simulation should behave as if from that point forwards, a continuous
+assignment `assign a = b;` have been added to override the drivers of `a`.
+More specifically: the value of `a` should be updated, whenever the value of
+`b` changes, all the way until a `release a;` statement is executed.
+Verilator instead evaluates the current value of `b` at the time the `force`
+statement is executed, and forces `a` to that value, without updating it
+until a new `force` or `release` statement is encountered that applies to
+`a`. This non-standard behavior is nevertheless consistent with some other
+simulators.
+
+
 Encrypted Verilog
 -----------------

@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 ********
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 .. _Simulating:
@@ -279,26 +279,25 @@ To use profiling:
   is being spent.


-.. _Thread Profiling:
+.. _Execution Profiling:

-Thread Profiling
-================
+Execution Profiling
+===================

-When using multithreaded mode (:vlopt:`--threads`), it is useful to see
-statistics and visualize how well the multiple CPUs are being utilized.
+For performance optimization, it is useful to see statistics and visualize how
+execution time is distributed in a verilated model.

-With the :vlopt:`--prof-threads` option, Verilator will:
+With the :vlopt:`--prof-exec` option, Verilator will:

-* Add code to the Verilated model to record the start and end time of each
-  macro-task across a number of calls to eval. (What is a macro-task?  See
-  the Verilator internals document (:file:`docs/internals.rst` in the
-  distribution.)
+* Add code to the Verilated model to record execution flow.

 * Add code to save profiling data in non-human-friendly form to the file
-  specified with :vlopt:`+verilator+prof+threads+file+\<filename\>`.
+  specified with :vlopt:`+verilator+prof+exec+file+\<filename\>`.

-* Add code to save profiling data for thread profile-guided
-  optimization. See :ref:`Thread PGO`.
+* In multi-threaded models, add code to record the start and end time of each
+  macro-task across a number of calls to eval. (What is a macro-task?  See the
+  Verilator internals document (:file:`docs/internals.rst` in the
+  distribution.)

 The :command:`verilator_gantt` program may then be run to transform the
 saved profiling file into a nicer visual format and produce some related
@@ -406,8 +405,8 @@ others as they prove beneficial.
 Thread Profile-Guided Optimization
 ----------------------------------

-Verilator supports thread profile-guided optimization (Thread PGO) to
-improve multithreaded performance.
+Verilator supports profile-guided optimization (Verilation) of multi-threaded
+models (Thread PGO) to improve performance.

 When using multithreading, Verilator computes how long macro tasks take and
 tries to balance those across threads.  (What is a macro-task?  See the
@@ -417,13 +416,14 @@ balanced, leading to decreased performance.  Thread PGO allows collecting
 profiling data to replace the estimates and better optimize these
 decisions.

-To use Thread PGO, Verilate the model with the :vlopt:`--prof-threads`
-option.
+To use Thread PGO, Verilate the model with the :vlopt:`--prof-pgo` option. This
+will code to the verilated model to save profiling data for profile-guided
+optimization.

 Run the model executable. When the executable exits, it will create a
 profile.vlt file.

-Rerun Verilator, optionally omitting the :vlopt:`--prof-threads` option,
+Rerun Verilator, optionally omitting the :vlopt:`--prof-pgo` option,
 and adding the profile.vlt generated earlier to the command line.

 Note there is no Verilator equivalent to GCC's --fprofile-use. Verilator's
@@ -433,8 +433,8 @@ directly without any prefix.
 If results from multiple simulations are to be used in generating the
 optimization, multiple simulation's profile.vlt may be concatenated
 externally, or each of the files may be fed as separate command line
-options into Verilator.  Verilator will simply sum the profile results, so
-a longer running test will have proportionally more weight for optimization
+options into Verilator.  Verilator will sum the profile results, so a
+longer running test will have proportionally more weight for optimization
 than a shorter running test.

 If you provide any profile feedback data to Verilator, and it cannot use
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 **********
@@ -68,7 +68,7 @@ multi-core SoC.

 Verilator is run in hierarchical mode on the whole SoC.  Verilator will
 make two models, one for the CPU hierarchy block, and one for the SoC.  The
-Verialted code for the SoC will automatically call the CPU Verilated model.
+Verilated code for the SoC will automatically call the CPU Verilated model.

 The current hierarchical Verilation is based on :vlopt:`--lib-create`. Each
 hierarchy block is Verilated into a library. User modules of the hierarchy
@@ -221,9 +221,13 @@ model, it may be beneficial to performance to adjust the
 influences the partitioning of the model by adjusting the assumed execution
 time of DPI imports.

-The :vlopt:`--trace-threads` options can be used to produce trace dumps
-using multiple threads. If :vlopt:`--trace-threads` is set without
-:vlopt:`--threads`, then :vlopt:`--trace-threads` will imply
+When using :vlopt:`--trace` to perform VCD tracing, the VCD trace
+construction is parallelized using the same number of threads as specified
+with :vlopt:`--threads`, and is executed on the same thread pool as the model.
+
+The :vlopt:`--trace-threads` options can be used with :vlopt:`--trace-fst`
+to offload FST tracing using multiple threads. If :vlopt:`--trace-threads` is
+given without :vlopt:`--threads`, then :vlopt:`--trace-threads` will imply
 :vlopt:`--threads 1 <--threads>`, i.e.: the support libraries will be
 thread safe.

@@ -231,12 +235,12 @@ With :vlopt:`--trace-threads 0 <--trace-threads>`, trace dumps are produced
 on the main thread. This again gives the highest single thread performance.

 With :vlopt:`--trace-threads {N} <--trace-threads>`, where N is at least 1,
-N additional threads will be created and managed by the trace files (e.g.:
-VerilatedVcdC or VerilatedFstC), to generate the trace dump. The main
-thread will be released to proceed with execution as soon as possible,
-though some blocking of the main thread is still necessary while capturing
-the trace. Different trace formats can utilize a various number of
-threads. See the :vlopt:`--trace-threads` option.
+up to N additional threads will be created and managed by the trace files
+(e.g.: VerilatedFstC), to offload construction of the trace dump. The main
+thread will be released to proceed with execution as soon as possible, though
+some blocking of the main thread is still necessary while capturing the
+trace. FST tracing can utilize up to 2 offload threads, so there is no use
+of setting :vlopt:`--trace-threads` higher than 2 at the moment.

 When running a multithreaded model, the default Linux task scheduler often
 works against the model, by assuming threads are short lived, and thus
@@ -265,7 +269,7 @@ This will limit memory to socket 0, and threads to cores 0, 1, 2, 3,
 (presumably on socket 0) optimizing performance.  Of course this must be
 adjusted if you want another simulator using e.g. socket 1, or if you
 Verilated with a different number of threads.  To see what CPUs are
-actually used, use :vlopt:`--prof-threads`.
+actually used, use :vlopt:`--prof-exec`.


 Multithreaded Verilog and Library Support
@@ -441,7 +445,7 @@ SystemC include directories and link to the SystemC libraries.

 .. describe:: TRACE_THREADS

-   Optional. Generated multi-threaded trace dumping, same as
+   Optional. Generated multi-threaded FST trace dumping, same as
   "--trace-threads".

 .. describe:: TOP_MODULE
@@ -1,4 +1,4 @@
-.. Copyright 2003-2021 by Wilson Snyder.
+.. Copyright 2003-2022 by Wilson Snyder.
 .. SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0

 *******************
@@ -12,7 +12,7 @@ Disabling Warnings
 Warnings may be disabled in multiple ways:

 #. Disable the warning in the source code.  When the warning is printed it
-   will include a warning code.  Simply surround the offending line with a
+   will include a warning code.  Surround the offending line with a
   :code:`/*verilator&32;lint_off*/` and :code:`/*verilator&32;lint_on*/`
   metacomment pair:

@@ -162,7 +162,7 @@ List Of Warnings
         always @(posedge clk)  foo[0] <= ...
         always_comb foo[1] = ...

-   Simply use a different register for the flop:
+   Instead use a different register for the flop:

   .. code-block:: sv

@@ -284,7 +284,7 @@ List Of Warnings

   .. TODO better example

-   Warns that it is simply better style to use casez, and "?" in place of
+   Warns that it is better style to use casez, and "?" in place of
   "x"'s.  See
   `http://www.sunburst-design.com/papers/CummingsSNUG1999Boston_FullParallelCase_rev1_1.pdf
   <http://www.sunburst-design.com/papers/CummingsSNUG1999Boston_FullParallelCase_rev1_1.pdf>`_
@@ -447,7 +447,7 @@ List Of Warnings
              p0();
         endmodule

-   Other tools with similar warnings: Veriable's forbid_defparam_rule.
+   Other tools with similar warnings: Verible's forbid_defparam_rule.


 .. option:: DEPRECATED
@@ -518,7 +518,7 @@ List Of Warnings
   Another way DIDNOTCONVERGE may occur is if # delays are used to generate
   clocks.  Verilator ignores the delays and gives an :option:`ASSIGNDLY`
   or :option:`STMTDLY` warning.  If these were suppressed, due to the
-   absense of the delay, the code may now oscillate.
+   absence of the delay, the code may now oscillate.

   Finally, rare, more difficult cases can be debugged like a C++ program;
   either enter :command:`gdb` and use its tracing facilities, or edit the
@@ -528,7 +528,7 @@ List Of Warnings
 .. option:: ENDCAPSULATED

   Warns that a class member is declared is local or protected, but is
-   being accessed from outside that class (if local) or a derrived class
+   being accessed from outside that class (if local) or a derived class
   (if protected).

   Ignoring this warning will only suppress the lint check, it will
@@ -924,7 +924,7 @@ List Of Warnings
   simulate correctly.

   Other tools with similar warnings: Icarus Verilog's portbind, "warning:
-   Instantiating module ... with dangling impot port (...)". Slang's
+   Instantiating module ... with dangling input port (...)". Slang's
   unconnected-port, "port '...' has no connection".


@@ -1212,12 +1212,12 @@ List Of Warnings
   .. include:: ../../docs/gen/ex_STMTDLY_msg.rst

   This is a warning because Verilator does not support delayed statements.
-   It will simply ignore all such delays.  In many cases ignoring a delay
-   might be harmless, but if the delayed statement is, as in this example,
-   used to cause some important action at a later time, it might be an
-   important difference.
+   It will ignore all such delays.  In many cases ignoring a delay might be
+   harmless, but if the delayed statement is, as in this example, used to
+   cause some important action at a later time, it might be an important
+   difference.

-   Some possible work arounds:
+   Some possible workarounds:

   * Move the delayed statement into the C++ wrapper file, where the
     stimulus and clock generation can be done in C++.
@@ -1506,7 +1506,7 @@ List Of Warnings
   Error that a construct might be legal according to IEEE but is not
   currently supported by Verilator.

-   A typical workaround is to recode the construct into a simpler and more
+   A typical workaround is to rewrite the construct into a simpler and more
   common alternative language construct.

   Alternatively, check if the construct is supported by other tools, and
@@ -114,11 +114,11 @@ By convention, each function/method uses the variable ``nodep`` as a
 pointer to the ``AstNode`` currently being processed.


-``AstNVisitor``
+``VNVisitor``
 ^^^^^^^^^^^^^^^

 The passes are implemented by AST visitor classes. These are implemented by
-subclasses of the abstract class, ``AstNVisitor``. Each pass creates an
+subclasses of the abstract class, ``VNVisitor``. Each pass creates an
 instance of the visitor class, which in turn implements a method to perform
 the pass.

@@ -268,13 +268,13 @@ Estimating Logic Costs

 To compute the cost of any given path through the graph, Verilator
 estimates an execution cost for each task. Each macro-task has an execution
-cost which is simply the sum of its tasks' costs. We assume that
-communication overhead and synchronization overhead are zero, so the cost
-of any given path through the graph is simply the sum of macro-task
-execution costs. Sarkar does almost the same thing, except that he has
-nonzero estimates for synchronization costs.
+cost which is the sum of its tasks' costs. We assume that communication
+overhead and synchronization overhead are zero, so the cost of any given
+path through the graph is the sum of macro-task execution costs. Sarkar
+does almost the same thing, except that he has nonzero estimates for
+synchronization costs.

-Verilator's cost estimates are assigned by ``InstrCountCostVisitor``.  This
+Verilator's cost estimates are assigned by ``InstrCountVisitor``.  This
 class is perhaps the most fragile piece of the multithread
 implementation. It's easy to have a bug where you count something cheap
 (eg. accessing one element of a huge array) as if it were expensive (eg.
@@ -301,7 +301,7 @@ prerequisites on other threads have finished.

 The synchronization cost is cheap if the prereqs are done. If they're not,
 fragmentation (idle CPU cores waiting) is possible. This is the major
-source of overhead in this approach. The ``--prof-threads`` switch and the
+source of overhead in this approach. The ``--prof-exec`` switch and the
 ``verilator_gantt`` script can visualize the time lost to such
 fragmentation.

@@ -523,10 +523,10 @@ optimization passes. This allows separation of the pass algorithm from the
 AST on which it operates. Wikipedia provides an introduction to the concept
 at https://en.wikipedia.org/wiki/Visitor_pattern.

-As noted above, all visitors are derived classes of ``AstNVisitor``. All
+As noted above, all visitors are derived classes of ``VNVisitor``. All
 derived classes of ``AstNode`` implement the ``accept`` method, which takes
-as argument a reference to an instance or a ``AstNVisitor`` derived class
-and applies the visit method of the ``AstNVisitor`` to the invoking AstNode
+as argument a reference to an instance or a ``VNVisitor`` derived class
+and applies the visit method of the ``VNVisitor`` to the invoking AstNode
 instance (i.e. ``this``).

 One possible difficulty is that a call to ``accept`` may perform an edit
@@ -604,8 +604,8 @@ There are three ways data is passed between visitor functions.
 Iterators
 ---------

-``AstNVisitor`` provides a set of iterators to facilitate walking over
-the tree. Each operates on the current ``AstNVisitor`` class (as this)
+``VNVisitor`` provides a set of iterators to facilitate walking over
+the tree. Each operates on the current ``VNVisitor`` class (as this)
 and takes an argument type ``AstNode*``.

 ``iterate``
@@ -817,7 +817,7 @@ which you can install using cpan.

 There are some traps to avoid when running regression tests

- When checking the MANIFEST, the test will barf on unexpected code in the
+- When checking the MANIFEST, the test will fail on unexpected code in the
  Verilator tree. So make sure to keep any such code outside the tree.

 - Not all Linux systems install Perldoc by default. This is needed for the
@@ -1183,10 +1183,10 @@ Verilator ideally would support all of IEEE, and has the goal to get close
 to full support. However the following IEEE sections and features are not
 anticipated to be ever implemented for the reasons indicated.

-IEEE 1800-2017 3.3 recursive modules
+IEEE 1800-2017 3.3 modules within modules
    Little/no tool support, and arguably not a good practice.
 IEEE 1800-2017 6.12 "shortreal"
-    Little/no tool support, and easily simply promoted to real.
+    Little/no tool support, and easily promoted to real.
 IEEE 1800-2017 11.11 Min, typ, max
    No SDF support so will always use typical.
 IEEE 1800-2017 11.12 "let"
@@ -1208,7 +1208,7 @@ IEEE 1800-2017 33 Config
 Distribution
 ============

-Copyright 2008-2021 by Wilson Snyder. Verilator is free software; you can
+Copyright 2008-2022 by Wilson Snyder. Verilator is free software; you can
 redistribute it and/or modify it under the terms of either the GNU Lesser
 General Public License Version 3 or the Perl Artistic License Version 2.0.

@@ -2,9 +2,11 @@ ABCp
 Accellera
 Affe
 Aleksander
+Alexandre
 Ami
 Amir
 Anastasiadis
+Anikin
 Antonin
 Antwerpen
 Arasanipalai
@@ -70,6 +72,7 @@ Eda
 Eddleston
 Egbert
 Egil
+Ehab
 Eiler
 Eivind
 El
@@ -82,6 +85,7 @@ Fabrizio
 Fekete
 Ferrandi
 Flachs
+Flavien
 Foletto
 Forker
 Francillon
@@ -104,6 +108,7 @@ Goessling
 Gonnen
 Goorah
 Gossner
+Graybeal
 Grobman
 Gunter
 Guo
@@ -121,6 +126,7 @@ Hornung
 Hossell
 Hsu
 Hyperthreading
+Ibrahim
 Iles
 Inlines
 Inout
@@ -134,6 +140,7 @@ Jannis
 Jens
 Jeras
 Jiuyang
+Joannou
 Joly
 Jullien
 Junji
@@ -147,6 +154,7 @@ Kamendje
 Kandadi
 Kaplan
 Karge
+Karlsson
 Katz
 Katzman
 Keren
@@ -161,6 +169,7 @@ Kolecki
 Koonce
 Korteland
 Koszek
+Kouping
 Kravitz
 Krolnik
 Kruse
@@ -194,6 +203,7 @@ Mednick
 Michiels
 Microsystems
 Milanovic
+Millis
 MinW
 Mindspeed
 Miodrag
@@ -224,12 +234,14 @@ Petr
 Piechotka
 Piersall
 Plunkett
+Popolon
 Popov
 Prabhat
 Prabhu
 Prateek
 Pre
 Preprocess
+Pretet
 Priyadharshini
 Pullup
 Pulver
@@ -262,12 +274,15 @@ Shi
 Shinkarovsky
 Shirakawa
 Shuba
+Shunyao
 Slatter
 SoC
 Sobhan
 Sokorac
 Solaris
 Solomatnikov
+Solt
+Southwell
 Srini
 Stamness
 Stephane
@@ -286,6 +301,7 @@ SystemVerilog
 Takatsukasa
 Tambe
 Tariq
+Tejada
 Tengstrand
 Terpstra
 Thiede
@@ -293,8 +309,10 @@ Thierry
 Thyer
 Tichelaar
 Tomov
+Tood
 Topa
 Tota
+Trefor
 Tresidder
 Tri
 Tristate
@@ -304,11 +322,13 @@ Ubixum
 Uints
 Undefines
 Unsized
+Urbach
 Uselib
 Usha
 Usuario
 VERILATOR
 Vasu
+Vdeeptemp
 Vdly
 Vemumtab
 Vemuri
@@ -351,7 +371,9 @@ Zhang
 abirkmanis
 accessor
 accessors
+adrienlemasle
 agrobman
+ahouska
 al
 ala
 andit
@@ -372,8 +394,12 @@ backtrace
 backtraces
 basename
 bbox
+benchmarking
 biguint
 biops
+bitOpTree
+bitOpTree
+bitop
 bitstoreal
 blackbox
 bokke
@@ -389,6 +415,7 @@ casez
 casted
 cb
 ccache
+ccall
 cdc
 ceil
 celldefine
@@ -411,6 +438,8 @@ concat
 concats
 config
 const
+constexpr
+constpool
 coredump
 countbits
 countones
@@ -426,6 +455,7 @@ da
 dat
 datadir
 datafiles
+david
 ddd
 deassign
 debugi
@@ -434,6 +464,7 @@ defname
 defparam
 demangling
 der
+dereference
 desassign
 destructor
 detections
@@ -462,6 +493,7 @@ dumpvars
 dut
 dx
 elab
+elike
 elsif
 endcase
 endcelldefine
@@ -480,15 +512,18 @@ enums
 env
 envvar
 eof
+errae
 erroring
 et
 eval
+evals
 exe
 executables
 expr
 extern
 fanin
 fasttrace
+fauto
 fbranch
 fclose
 fdisplay
@@ -503,6 +538,7 @@ filt
 flto
 flushCall
 fopen
+forceable
 foreach
 fprintf
 fprofile
@@ -521,6 +557,7 @@ funcs
 fwrite
 gantt
 gcc
+gcda
 gdb
 genblk
 genvar
@@ -552,6 +589,7 @@ incdir
 includer
 inferfaces
 inhibitSim
+initarray
 initializer
 initializers
 inits
@@ -570,6 +608,7 @@ killua
 lang
 lcov
 ld
+leavinel
 len
 libext
 libgoogle
@@ -584,6 +623,7 @@ ln
 loc
 localparam
 localparams
+localtime
 logicals
 longint
 lsb
@@ -618,6 +658,7 @@ multithreaded
 multithreading
 mutexes
 mux
+myftptoyman
 mysignal
 namespace
 nand
@@ -642,15 +683,19 @@ onehot
 ooo
 oprofile
 oversubscription
+parallelized
 param
 parameterized
 params
 parens
+pawel
 pc
 pdf
 perf
 perftools
+pgo
 picoChip
+pinIndex
 pinout
 plusargs
 pmos
@@ -676,6 +721,7 @@ prev
 printf
 printtimescale
 profcfunc
+profiler
 prototyptes
 ps
 pthread
@@ -703,12 +749,14 @@ regs
 reloop
 resetall
 respecified
+rodata
 rr
 rst
 runtime
 runtimes
 rw
 sVerilator
+sawatzke
 sc
 scalared
 sccanf
@@ -724,6 +772,7 @@ specparam
 splitme
 spp
 sqrt
+src
 srcdir
 srcfile
 sscanf
@@ -743,6 +792,7 @@ subcells
 subexpressions
 submodule
 submodules
+substring
 sv
 svBitVal
 svBitVecVal
@@ -777,6 +827,7 @@ trunc
 txt
 typ
 typedef
+typedef'ed
 typedefed
 typedefs
 typename
@@ -790,6 +841,7 @@ uniquified
 unistd
 unlink
 unlinked
+unnamedblk
 unopt
 unoptflat
 unoptimizable
@@ -839,11 +891,14 @@ writeme
 writemem
 writememb
 writememh
+xiak
 xin
 xml
 xnor
 xout
 xuejiazidi
 yanx
+ypq
 yurivict
 zdave
+Øyvind
@@ -46,9 +46,6 @@ The XML document consists of 4 sections within the top level
   hierarchy. Each instance is represented with the ``<cell>`` element
   with the following attributes:

-   -  ``fl`` (deprecated): The file id and line number where the module
-      was instanced. Use ``loc`` instead.
-
   -  ``loc``: The file id, first line number, last line number, first
      column number and last column number of the identifier where the
      module was instanced, separated by commas.
@@ -73,7 +70,7 @@ The XML document consists of 4 sections within the top level
 Distribution
 ============

-Copyright 2020-2021 by Wilson Snyder. Verilator is free software; you can
+Copyright 2020-2022 by Wilson Snyder. Verilator is free software; you can
 redistribute it and/or modify it under the terms of either the GNU Lesser
 General Public License Version 3 or the Perl Artistic License Version 2.0.

@@ -33,5 +33,5 @@ add_executable(example ../make_tracing_c/sim_main.cpp)
 # Add the Verilated circuit to the target
 verilate(example COVERAGE TRACE
  INCLUDE_DIRS "../make_tracing_c"
-  VERILATOR_ARGS -f ../make_tracing_c/input.vc -Os -x-assign 0
+  VERILATOR_ARGS -f ../make_tracing_c/input.vc -x-assign fast
  SOURCES ../make_tracing_c/top.v)
@@ -45,7 +45,7 @@ set_property(
 # Add the Verilated circuit to the target
 verilate(example SYSTEMC COVERAGE TRACE
  INCLUDE_DIRS "../make_tracing_sc"
-  VERILATOR_ARGS -f ../make_tracing_sc/input.vc -Os -x-assign 0
+  VERILATOR_ARGS -f ../make_tracing_sc/input.vc -x-assign fast
  SOURCES ../make_tracing_sc/top.v)

 verilator_link_systemc(example)
@@ -33,7 +33,7 @@ VERILATOR_FLAGS =
 # Generate C++
 VERILATOR_FLAGS += -cc
 # Optimize
-VERILATOR_FLAGS += -Os -x-assign 0
+VERILATOR_FLAGS += -x-assign fast
 # Warn abount lint issues; may not want this on less solid designs
 VERILATOR_FLAGS += -Wall
 # This example does not use vl_time_stamp but rather
@@ -36,7 +36,7 @@ VERILATOR_FLAGS += -cc --exe
 # Generate makefile dependencies (not shown as complicates the Makefile)
 #VERILATOR_FLAGS += -MMD
 # Optimize
-VERILATOR_FLAGS += -Os -x-assign 0
+VERILATOR_FLAGS += -x-assign fast
 # Warn abount lint issues; may not want this on less solid designs
 VERILATOR_FLAGS += -Wall
 # Make waveforms
@@ -34,6 +34,8 @@ int main(int argc, char** argv, char** env) {
    // Using unique_ptr is similar to
    // "VerilatedContext* contextp = new VerilatedContext" then deleting at end.
    const std::unique_ptr<VerilatedContext> contextp{new VerilatedContext};
+    // Do not instead make Vtop as a file-scope static variable, as the
+    // "C++ static initialization order fiasco" may cause a crash

    // Set debug level, 0 is off, 9 is highest presently used
    // May be overridden by commandArgs argument parsing
@@ -69,7 +71,7 @@ int main(int argc, char** argv, char** env) {
        // Historical note, before Verilator 4.200 Verilated::gotFinish()
        // was used above in place of contextp->gotFinish().
        // Most of the contextp-> calls can use Verilated:: calls instead;
-        // the Verilated:: versions simply assume there's a single context
+        // the Verilated:: versions just assume there's a single context
        // being used (per thread).  It's faster and clearer to use the
        // newer contextp-> versions.

@@ -103,8 +105,8 @@ int main(int argc, char** argv, char** env) {
        top->eval();

        // Read outputs
-        VL_PRINTF("[%" VL_PRI64 "d] clk=%x rstl=%x iquad=%" VL_PRI64 "x"
-                  " -> oquad=%" VL_PRI64 "x owide=%x_%08x_%08x\n",
+        VL_PRINTF("[%" PRId64 "] clk=%x rstl=%x iquad=%" PRIx64 " -> oquad=%" PRIx64
+                  " owide=%x_%08x_%08x\n",
                  contextp->time(), top->clk, top->reset_l, top->in_quad, top->out_quad,
                  top->out_wide[2], top->out_wide[1], top->out_wide[0]);
    }
@@ -37,7 +37,7 @@ VERILATOR_FLAGS += -sc --exe
 # Generate makefile dependencies (not shown as complicates the Makefile)
 #VERILATOR_FLAGS += -MMD
 # Optimize
-VERILATOR_FLAGS += -Os -x-assign 0
+VERILATOR_FLAGS += -x-assign fast
 # Warn abount lint issues; may not want this on less solid designs
 VERILATOR_FLAGS += -Wall
 # Make waveforms
@@ -58,11 +58,11 @@ int sc_main(int argc, char* argv[]) {

    // Define interconnect
    sc_signal<bool> reset_l;
-    sc_signal<vluint32_t> in_small;
-    sc_signal<vluint64_t> in_quad;
+    sc_signal<uint32_t> in_small;
+    sc_signal<uint64_t> in_quad;
    sc_signal<sc_bv<70>> in_wide;
-    sc_signal<vluint32_t> out_small;
-    sc_signal<vluint64_t> out_quad;
+    sc_signal<uint32_t> out_small;
+    sc_signal<uint64_t> out_quad;
    sc_signal<sc_bv<70>> out_wide;

    // Construct the Verilated model, from inside Vtop.h
@@ -140,6 +140,10 @@ void **JenkinsIns(void *base_i, const unsigned char *mem, uint32_t length, uint3
 #include <sys/sysctl.h>
 #endif

+#if defined(FST_MACOSX) || defined(__MINGW32__) || defined(__OpenBSD__) || defined(__FreeBSD__)
+#define FST_UNBUFFERED_IO
+#endif
+
 #ifdef __GNUC__
 /* Boolean expression more often true than false */
 #define FST_LIKELY(x) __builtin_expect(!!(x), 1)
@@ -4577,12 +4581,11 @@ if(sectype == FST_BL_ZWRAPPER)
                if(!fcomp) { tmpfile_close(&fcomp, &xc->f_nam); return(0); }
                }

-#if defined(FST_MACOSX)
+#if defined(FST_UNBUFFERED_IO)
        setvbuf(fcomp, (char *)NULL, _IONBF, 0);   /* keeps gzip from acting weird in tandem with fopen */
 #endif

 #ifdef __MINGW32__
-        setvbuf(fcomp, (char *)NULL, _IONBF, 0);   /* keeps gzip from acting weird in tandem with fopen */
        xc->filename_unpacked = hf;
 #else
        if(hf)
@@ -4880,7 +4883,7 @@ if((!nam)||(!(xc->f=fopen(nam, "rb"))))
        char *hf = (char *)calloc(1, flen + 6);
        int rc;

-#if defined(__MINGW32__) || defined(FST_MACOSX)
+#if defined(FST_UNBUFFERED_IO)
        setvbuf(xc->f, (char *)NULL, _IONBF, 0);   /* keeps gzip from acting weird in tandem with fopen */
 #endif

@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -55,6 +55,7 @@
 #include <algorithm>
 #include <cctype>
 #include <cerrno>
+#include <cstdlib>
 #include <sstream>
 #include <sys/stat.h>  // mkdir
 #include <list>
@@ -73,10 +74,10 @@ constexpr unsigned VL_VALUE_STRING_MAX_WIDTH = 8192;
 //===========================================================================
 // Static sanity checks

-static_assert(sizeof(vluint8_t) == 1, "vluint8_t is missized");
-static_assert(sizeof(vluint16_t) == 2, "vluint8_t is missized");
-static_assert(sizeof(vluint32_t) == 4, "vluint8_t is missized");
-static_assert(sizeof(vluint64_t) == 8, "vluint8_t is missized");
+static_assert(sizeof(uint8_t) == 1, "uint8_t is missized");
+static_assert(sizeof(uint16_t) == 2, "uint8_t is missized");
+static_assert(sizeof(uint32_t) == 4, "uint8_t is missized");
+static_assert(sizeof(uint64_t) == 8, "uint8_t is missized");

 //===========================================================================
 // Global variables
@@ -168,6 +169,19 @@ void vl_stop_maybe(const char* filename, int linenum, const char* hier, bool may
 }
 #endif

+#ifndef VL_USER_WARN  ///< Define this to override the vl_warn function
+void vl_warn(const char* filename, int linenum, const char* hier, const char* msg) VL_MT_UNSAFE {
+    if (false && hier) {}
+    if (filename && filename[0]) {
+        // Not VL_PRINTF_MT, already on main thread
+        VL_PRINTF("%%Warning: %s:%d: %s\n", filename, linenum, msg);
+    } else {
+        VL_PRINTF("%%Warning: %s\n", msg);
+    }
+    Verilated::runFlushCallbacks();
+}
+#endif
+
 //===========================================================================
 // Wrapper to call certain functions via messages when multithreaded

@@ -201,6 +215,16 @@ void VL_FATAL_MT(const char* filename, int linenum, const char* hier, const char
 #endif
 }

+void VL_WARN_MT(const char* filename, int linenum, const char* hier, const char* msg) VL_MT_SAFE {
+#ifdef VL_THREADED
+    VerilatedThreadMsgQueue::post(VerilatedMsg{[=]() {  //
+        vl_warn(filename, linenum, hier, msg);
+    }});
+#else
+    vl_warn(filename, linenum, hier, msg);
+#endif
+}
+
 //===========================================================================
 // Debug prints

@@ -220,21 +244,21 @@ std::string _vl_string_vprintf(const char* formatp, va_list ap) VL_MT_SAFE {
    return out;
 }

-vluint64_t _vl_dbg_sequence_number() VL_MT_SAFE {
+uint64_t _vl_dbg_sequence_number() VL_MT_SAFE {
 #ifdef VL_THREADED
-    static std::atomic<vluint64_t> sequence;
+    static std::atomic<uint64_t> sequence;
 #else
-    static vluint64_t sequence = 0;
+    static uint64_t sequence = 0;
 #endif
    return ++sequence;
 }

-vluint32_t VL_THREAD_ID() VL_MT_SAFE {
+uint32_t VL_THREAD_ID() VL_MT_SAFE {
 #ifdef VL_THREADED
    // Alternative is to use std::this_thread::get_id, but that returns a
    // hard-to-read number and is very slow
-    static std::atomic<vluint32_t> s_nextId(0);
-    static VL_THREAD_LOCAL vluint32_t t_myId = ++s_nextId;
+    static std::atomic<uint32_t> s_nextId(0);
+    static VL_THREAD_LOCAL uint32_t t_myId = ++s_nextId;
    return t_myId;
 #else
    return 0;
@@ -248,13 +272,13 @@ void VL_DBG_MSGF(const char* formatp, ...) VL_MT_SAFE {
    va_start(ap, formatp);
    const std::string out = _vl_string_vprintf(formatp, ap);
    va_end(ap);
-    // printf("-imm-V{t%d,%" VL_PRI64 "d}%s", VL_THREAD_ID(), _vl_dbg_sequence_number(),
+    // printf("-imm-V{t%d,%" PRId64 "}%s", VL_THREAD_ID(), _vl_dbg_sequence_number(),
    // out.c_str());

    // Using VL_PRINTF not VL_PRINTF_MT so that we can call VL_DBG_MSGF
    // from within the guts of the thread execution machinery (and it goes
    // to the screen and not into the queues we're debugging)
-    VL_PRINTF("-V{t%u,%" VL_PRI64 "u}%s", VL_THREAD_ID(), _vl_dbg_sequence_number(), out.c_str());
+    VL_PRINTF("-V{t%u,%" PRIu64 "}%s", VL_THREAD_ID(), _vl_dbg_sequence_number(), out.c_str());
 }

 #ifdef VL_THREADED
@@ -272,7 +296,7 @@ void VL_PRINTF_MT(const char* formatp, ...) VL_MT_SAFE {
 //===========================================================================
 // Random -- Mostly called at init time, so not inline.

-static vluint32_t vl_sys_rand32() VL_MT_UNSAFE {
+static uint32_t vl_sys_rand32() VL_MT_UNSAFE {
    // Return random 32-bits using system library.
    // Used only to construct seed for Verilator's PNRG.
    static VerilatedMutex s_mutex;
@@ -285,9 +309,9 @@ static vluint32_t vl_sys_rand32() VL_MT_UNSAFE {
 #endif
 }

-vluint64_t vl_rand64() VL_MT_SAFE {
-    static VL_THREAD_LOCAL vluint64_t t_state[2];
-    static VL_THREAD_LOCAL vluint32_t t_seedEpoch = 0;
+uint64_t vl_rand64() VL_MT_SAFE {
+    static VL_THREAD_LOCAL uint64_t t_state[2];
+    static VL_THREAD_LOCAL uint32_t t_seedEpoch = 0;
    // For speed, we use a thread-local epoch number to know when to reseed
    // A thread always belongs to a single context, so this works out ok
    if (VL_UNLIKELY(t_seedEpoch != VerilatedContextImp::randSeedEpoch())) {
@@ -301,7 +325,7 @@ vluint64_t vl_rand64() VL_MT_SAFE {
        if (VL_COUNTONES_I(t_state[1]) < 10) t_state[1] = ~t_state[1];
    }
    // Xoroshiro128+ algorithm
-    const vluint64_t result = t_state[0] + t_state[1];
+    const uint64_t result = t_state[0] + t_state[1];
    t_state[1] ^= t_state[0];
    t_state[0] = (((t_state[0] << 55) | (t_state[0] >> 9)) ^ t_state[1] ^ (t_state[1] << 14));
    t_state[1] = (t_state[1] << 36) | (t_state[1] >> 28);
@@ -314,11 +338,17 @@ WDataOutP VL_RANDOM_W(int obits, WDataOutP outwp) VL_MT_SAFE {
    return outwp;
 }

-IData VL_RANDOM_SEEDED_II(IData seed) VL_MT_SAFE {
+IData VL_RANDOM_SEEDED_II(IData& seedr) VL_MT_SAFE {
+    // $random - seed is a new seed to apply, then we return new seed
+    Verilated::threadContextp()->randSeed(static_cast<int>(seedr));
+    seedr = VL_RANDOM_I();
+    return VL_RANDOM_I();
+}
+IData VL_URANDOM_SEEDED_II(IData seed) VL_MT_SAFE {
+    // $urandom - seed is a new seed to apply
    Verilated::threadContextp()->randSeed(static_cast<int>(seed));
    return VL_RANDOM_I();
 }
-
 IData VL_RAND_RESET_I(int obits) VL_MT_SAFE {
    if (Verilated::threadContextp()->randReset() == 0) return 0;
    IData data = ~0;
@@ -380,11 +410,11 @@ WDataOutP _vl_moddiv_w(int lbits, WDataOutP owp, const WDataInP lwp, const WData
    const int vw = VL_WORDS_I(vmsbp1);  // aka "n" in the algorithm

    if (vw == 1) {  // Single divisor word breaks rest of algorithm
-        vluint64_t k = 0;
+        uint64_t k = 0;
        for (int j = uw - 1; j >= 0; --j) {
-            const vluint64_t unw64 = ((k << 32ULL) + static_cast<vluint64_t>(lwp[j]));
-            owp[j] = unw64 / static_cast<vluint64_t>(rwp[0]);
-            k = unw64 - static_cast<vluint64_t>(owp[j]) * static_cast<vluint64_t>(rwp[0]);
+            const uint64_t unw64 = ((k << 32ULL) + static_cast<uint64_t>(lwp[j]));
+            owp[j] = unw64 / static_cast<uint64_t>(rwp[0]);
+            k = unw64 - static_cast<uint64_t>(owp[j]) * static_cast<uint64_t>(rwp[0]);
        }
        if (is_modulus) {
            owp[0] = k;
@@ -394,8 +424,8 @@ WDataOutP _vl_moddiv_w(int lbits, WDataOutP owp, const WDataInP lwp, const WData
    }

    // +1 word as we may shift during normalization
-    vluint32_t un[VL_MULS_MAX_WORDS + 1];  // Fixed size, as MSVC++ doesn't allow [words] here
-    vluint32_t vn[VL_MULS_MAX_WORDS + 1];  // v normalized
+    uint32_t un[VL_MULS_MAX_WORDS + 1];  // Fixed size, as MSVC++ doesn't allow [words] here
+    uint32_t vn[VL_MULS_MAX_WORDS + 1];  // v normalized

    // Zero for ease of debugging and to save having to zero for shifts
    // Note +1 as loop will use extra word
@@ -404,7 +434,7 @@ WDataOutP _vl_moddiv_w(int lbits, WDataOutP owp, const WDataInP lwp, const WData
    // Algorithm requires divisor MSB to be set
    // Copy and shift to normalize divisor so MSB of vn[vw-1] is set
    const int s = 31 - VL_BITBIT_I(vmsbp1 - 1);  // shift amount (0...31)
-    const vluint32_t shift_mask = s ? 0xffffffff : 0;  // otherwise >> 32 won't mask the value
+    const uint32_t shift_mask = s ? 0xffffffff : 0;  // otherwise >> 32 won't mask the value
    for (int i = vw - 1; i > 0; --i) {
        vn[i] = (rwp[i] << s) | (shift_mask & (rwp[i - 1] >> (32 - s)));
    }
@@ -424,10 +454,10 @@ WDataOutP _vl_moddiv_w(int lbits, WDataOutP owp, const WDataInP lwp, const WData
    // Main loop
    for (int j = uw - vw; j >= 0; --j) {
        // Estimate
-        const vluint64_t unw64 = (static_cast<vluint64_t>(un[j + vw]) << 32ULL
-                                  | static_cast<vluint64_t>(un[j + vw - 1]));
-        vluint64_t qhat = unw64 / static_cast<vluint64_t>(vn[vw - 1]);
-        vluint64_t rhat = unw64 - qhat * static_cast<vluint64_t>(vn[vw - 1]);
+        const uint64_t unw64
+            = (static_cast<uint64_t>(un[j + vw]) << 32ULL | static_cast<uint64_t>(un[j + vw - 1]));
+        uint64_t qhat = unw64 / static_cast<uint64_t>(vn[vw - 1]);
+        uint64_t rhat = unw64 - qhat * static_cast<uint64_t>(vn[vw - 1]);

    again:
        if (qhat >= 0x100000000ULL || ((qhat * vn[vw - 2]) > ((rhat << 32ULL) + un[j + vw - 2]))) {
@@ -436,10 +466,10 @@ WDataOutP _vl_moddiv_w(int lbits, WDataOutP owp, const WDataInP lwp, const WData
            if (rhat < 0x100000000ULL) goto again;
        }

-        vlsint64_t t = 0;  // Must be signed
-        vluint64_t k = 0;
+        int64_t t = 0;  // Must be signed
+        uint64_t k = 0;
        for (int i = 0; i < vw; ++i) {
-            const vluint64_t p = qhat * vn[i];  // Multiply by estimate
+            const uint64_t p = qhat * vn[i];  // Multiply by estimate
            t = un[i + j] - k - (p & 0xFFFFFFFFULL);  // Subtract
            un[i + j] = t;
            k = (p >> 32ULL) - (t >> 32ULL);
@@ -453,7 +483,7 @@ WDataOutP _vl_moddiv_w(int lbits, WDataOutP owp, const WDataInP lwp, const WData
            owp[j]--;
            k = 0;
            for (int i = 0; i < vw; ++i) {
-                t = static_cast<vluint64_t>(un[i + j]) + static_cast<vluint64_t>(vn[i]) + k;
+                t = static_cast<uint64_t>(un[i + j]) + static_cast<uint64_t>(vn[i]) + k;
                un[i + j] = t;
                k = t >> 32ULL;
            }
@@ -584,7 +614,7 @@ double VL_ITOR_D_W(int lbits, const WDataInP lwp) VL_PURE {
 }
 double VL_ISTOR_D_W(int lbits, const WDataInP lwp) VL_PURE {
    if (!VL_SIGN_W(lbits, lwp)) return VL_ITOR_D_W(lbits, lwp);
-    vluint32_t pos[VL_MULS_MAX_WORDS + 1];  // Fixed size, as MSVC++ doesn't allow [words] here
+    uint32_t pos[VL_MULS_MAX_WORDS + 1];  // Fixed size, as MSVC++ doesn't allow [words] here
    VL_NEGATE_W(VL_WORDS_I(lbits), pos, lwp);
    _vl_clean_inplace_w(lbits, pos);
    return -VL_ITOR_D_W(lbits, pos);
@@ -658,7 +688,7 @@ std::string _vl_vsformat_time(char* tmp, T ld, int timeunit, bool left, size_t w
        const WDataInP integer = VL_DIV_WWW(b, tmp0, shifted, fracDigitsPow10);
        const WDataInP frac = VL_MODDIV_WWW(b, tmp1, shifted, fracDigitsPow10);
        const WDataInP max64Bit
-            = VL_EXTEND_WQ(b, 0, tmp2, std::numeric_limits<vluint64_t>::max());  // breaks shifted
+            = VL_EXTEND_WQ(b, 0, tmp2, std::numeric_limits<uint64_t>::max());  // breaks shifted
        if (VL_GT_W(w, integer, max64Bit)) {
            WDataOutP v = VL_ASSIGN_W(b, tmp3, integer);  // breaks fracDigitsPow10
            VlWide<w> zero, ten;
@@ -678,18 +708,17 @@ std::string _vl_vsformat_time(char* tmp, T ld, int timeunit, bool left, size_t w
            if (!fracDigits) {
                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH, "%s%s", ptr, suffix.c_str());
            } else {
-                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH, "%s.%0*" VL_PRI64 "u%s", ptr,
+                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH, "%s.%0*" PRIu64 "%s", ptr,
                                     fracDigits, VL_SET_QW(frac), suffix.c_str());
            }
        } else {
-            const vluint64_t integer64 = VL_SET_QW(integer);
+            const uint64_t integer64 = VL_SET_QW(integer);
            if (!fracDigits) {
-                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH, "%" VL_PRI64 "u%s", integer64,
+                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH, "%" PRIu64 "%s", integer64,
                                     suffix.c_str());
            } else {
-                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH,
-                                     "%" VL_PRI64 "u.%0*" VL_PRI64 "u%s", integer64, fracDigits,
-                                     VL_SET_QW(frac), suffix.c_str());
+                digits = VL_SNPRINTF(tmp, VL_VALUE_STRING_MAX_WIDTH, "%" PRIu64 ".%0*" PRIu64 "%s",
+                                     integer64, fracDigits, VL_SET_QW(frac), suffix.c_str());
            }
        }
    } else {
@@ -850,9 +879,9 @@ void _vl_vsformat(std::string& output, const char* formatp, va_list ap) VL_MT_SA
                    int digits = 0;
                    std::string append;
                    if (lbits <= VL_QUADSIZE) {
-                        digits = VL_SNPRINTF(
-                            t_tmp, VL_VALUE_STRING_MAX_WIDTH, "%" VL_PRI64 "d",
-                            static_cast<vlsint64_t>(VL_EXTENDS_QQ(lbits, lbits, ld)));
+                        digits
+                            = VL_SNPRINTF(t_tmp, VL_VALUE_STRING_MAX_WIDTH, "%" PRId64,
+                                          static_cast<int64_t>(VL_EXTENDS_QQ(lbits, lbits, ld)));
                        append = t_tmp;
                    } else {
                        if (VL_SIGN_E(lbits, lwp[VL_WORDS_I(lbits) - 1])) {
@@ -880,8 +909,7 @@ void _vl_vsformat(std::string& output, const char* formatp, va_list ap) VL_MT_SA
                    int digits = 0;
                    std::string append;
                    if (lbits <= VL_QUADSIZE) {
-                        digits
-                            = VL_SNPRINTF(t_tmp, VL_VALUE_STRING_MAX_WIDTH, "%" VL_PRI64 "u", ld);
+                        digits = VL_SNPRINTF(t_tmp, VL_VALUE_STRING_MAX_WIDTH, "%" PRIu64, ld);
                        append = t_tmp;
                    } else {
                        append = VL_DECIMAL_NW(lbits, lwp);
@@ -1156,8 +1184,8 @@ IData _vl_vsscanf(FILE* fp,  // If a fscanf
                    _vl_vsss_skipspace(fp, floc, fromp, fstr);
                    _vl_vsss_read_str(fp, floc, fromp, fstr, t_tmp, "0123456789+-xXzZ?_");
                    if (!t_tmp[0]) goto done;
-                    vlsint64_t ld = 0;
-                    std::sscanf(t_tmp, "%30" VL_PRI64 "d", &ld);
+                    int64_t ld = 0;
+                    std::sscanf(t_tmp, "%30" PRId64, &ld);
                    VL_SET_WQ(owp, ld);
                    break;
                }
@@ -1170,7 +1198,7 @@ IData _vl_vsscanf(FILE* fp,  // If a fscanf
                    // cppcheck-suppress unusedStructMember  // It's used
                    union {
                        double r;
-                        vlsint64_t ld;
+                        int64_t ld;
                    } u;
                    u.r = std::strtod(t_tmp, nullptr);
                    VL_SET_WQ(owp, u.ld);
@@ -1182,7 +1210,7 @@ IData _vl_vsscanf(FILE* fp,  // If a fscanf
                    _vl_vsss_read_str(fp, floc, fromp, fstr, t_tmp, "0123456789+-xXzZ?_");
                    if (!t_tmp[0]) goto done;
                    QData ld = 0;
-                    std::sscanf(t_tmp, "%30" VL_PRI64 "u", &ld);
+                    std::sscanf(t_tmp, "%30" PRIu64, &ld);
                    VL_SET_WQ(owp, ld);
                    break;
                }
@@ -1470,7 +1498,7 @@ void VL_FWRITEF(IData fpi, const char* formatp, ...) VL_MT_SAFE {
 IData VL_FSCANF_IX(IData fpi, const char* formatp, ...) VL_MT_SAFE {
    // While threadsafe, each thread can only access different file handles
    FILE* const fp = VL_CVT_I_FP(fpi);
-    if (VL_UNLIKELY(!fp)) return 0;
+    if (VL_UNLIKELY(!fp)) return ~0U;  // -1

    va_list ap;
    va_start(ap, formatp);
@@ -1615,8 +1643,8 @@ IData VL_VALUEPLUSARGS_INW(int rbits, const std::string& ld, WDataOutP rwp) VL_M
    VL_ZERO_RESET_W(rbits, rwp);
    switch (std::tolower(fmt)) {
    case 'd': {
-        vlsint64_t lld = 0;
-        std::sscanf(dp, "%30" VL_PRI64 "d", &lld);
+        int64_t lld = 0;
+        std::sscanf(dp, "%30" PRId64, &lld);
        VL_SET_WQ(rwp, lld);
        break;
    }
@@ -1650,7 +1678,7 @@ IData VL_VALUEPLUSARGS_INW(int rbits, const std::string& ld, WDataOutP rwp) VL_M
        VL_SET_WQ(rwp, VL_CVT_Q_D(temp));
        break;
    }
-    default:  // Other simulators simply return 0 in these cases and don't error out
+    default:  // Other simulators return 0 in these cases and don't error out
        return 0;
    }
    _vl_clean_inplace_w(rbits, rwp);
@@ -1740,7 +1768,7 @@ std::string VL_CVT_PACK_STR_NW(int lwords, const WDataInP lwp) VL_MT_SAFE {

 std::string VL_PUTC_N(const std::string& lhs, IData rhs, CData ths) VL_PURE {
    std::string lstring = lhs;
-    const vlsint32_t rhs_s = rhs;  // To signed value
+    const int32_t rhs_s = rhs;  // To signed value
    // 6.16.2:str.putc(i, c) does not change the value when i < 0 || i >= str.len() || c == 0
    if (0 <= rhs_s && rhs < lhs.length() && ths != 0) lstring[rhs] = ths;
    return lstring;
@@ -1748,15 +1776,15 @@ std::string VL_PUTC_N(const std::string& lhs, IData rhs, CData ths) VL_PURE {

 CData VL_GETC_N(const std::string& lhs, IData rhs) VL_PURE {
    CData v = 0;
-    const vlsint32_t rhs_s = rhs;  // To signed value
+    const int32_t rhs_s = rhs;  // To signed value
    // 6.16.3:str.getc(i) returns 0 if i < 0 || i >= str.len()
    if (0 <= rhs_s && rhs < lhs.length()) v = lhs[rhs];
    return v;
 }

 std::string VL_SUBSTR_N(const std::string& lhs, IData rhs, IData ths) VL_PURE {
-    const vlsint32_t rhs_s = rhs;  // To signed value
-    const vlsint32_t ths_s = ths;  // To signed value
+    const int32_t rhs_s = rhs;  // To signed value
+    const int32_t ths_s = ths;  // To signed value
    // 6.16.8:str.substr(i, j) returns an empty string when i < 0 || j < i || j >= str.len()
    if (rhs_s < 0 || ths_s < rhs_s || ths >= lhs.length()) return "";
    // Second parameter of std::string::substr(i, n) is length, not position as in SystemVerilog
@@ -1795,7 +1823,7 @@ static const char* memhFormat(int nBits) {
    return t_buf;
 }

-static const char* formatBinary(int nBits, vluint32_t bits) {
+static const char* formatBinary(int nBits, uint32_t bits) {
    assert((nBits >= 1) && (nBits <= 32));

    static VL_THREAD_LOCAL char t_buf[64];
@@ -1817,7 +1845,7 @@ VlReadMem::VlReadMem(bool hex, int bits, const std::string& filename, QData star
    m_fp = std::fopen(filename.c_str(), "r");
    if (VL_UNLIKELY(!m_fp)) {
        // We don't report the Verilog source filename as it slow to have to pass it down
-        VL_FATAL_MT(filename.c_str(), 0, "", "$readmem file not found");
+        VL_WARN_MT(filename.c_str(), 0, "", "$readmem file not found");
        // cppcheck-suppress resourceLeak  // m_fp is nullptr - bug in cppcheck
        return;
    }
@@ -1876,6 +1904,7 @@ bool VlReadMem::get(QData& addrr, std::string& valuer) {
                ignore_to_eol = true;
            } else if (c == '@') {
                reading_addr = true;
+                m_anyAddr = true;
                m_addr = 0;
            }
            // Check for hex or binary digits as file format requests
@@ -1902,9 +1931,9 @@ bool VlReadMem::get(QData& addrr, std::string& valuer) {
        lastc = c;
    }

-    if (VL_UNLIKELY(m_end != ~0ULL && m_addr <= m_end)) {
-        VL_FATAL_MT(m_filename.c_str(), m_linenum, "",
-                    "$readmem file ended before specified final address (IEEE 2017 21.4)");
+    if (VL_UNLIKELY(m_end != ~0ULL && m_addr <= m_end && !m_anyAddr)) {
+        VL_WARN_MT(m_filename.c_str(), m_linenum, "",
+                   "$readmem file ended before specified final address (IEEE 2017 21.4)");
    }

    return false;  // EOF
@@ -1969,7 +1998,7 @@ VlWriteMem::~VlWriteMem() {
 void VlWriteMem::print(QData addr, bool addrstamp, const void* valuep) {
    if (VL_UNLIKELY(!m_fp)) return;
    if (addr != m_addr && addrstamp) {  // Only assoc has time stamps
-        fprintf(m_fp, "@%" VL_PRI64 "x\n", addr);
+        fprintf(m_fp, "@%" PRIx64 "\n", addr);
    }
    m_addr = addr + 1;
    if (m_bits <= 8) {
@@ -1998,9 +2027,9 @@ void VlWriteMem::print(QData addr, bool addrstamp, const void* valuep) {
        }
    } else if (m_bits <= 64) {
        const QData* const datap = reinterpret_cast<const QData*>(valuep);
-        const vluint64_t value = VL_MASK_Q(m_bits) & *datap;
-        const vluint32_t lo = value & 0xffffffff;
-        const vluint32_t hi = value >> 32;
+        const uint64_t value = VL_MASK_Q(m_bits) & *datap;
+        const uint32_t lo = value & 0xffffffff;
+        const uint32_t hi = value >> 32;
        if (m_hex) {
            fprintf(m_fp, memhFormat(m_bits - 32), hi);
            fprintf(m_fp, "%08x\n", lo);
@@ -2206,8 +2235,8 @@ double vl_time_multiplier(int scale) VL_PURE {
        return pow10[scale];
    }
 }
-vluint64_t vl_time_pow10(int n) {
-    static const vluint64_t pow10[20] = {
+uint64_t vl_time_pow10(int n) {
+    static const uint64_t pow10[20] = {
        1ULL,
        10ULL,
        100ULL,
@@ -2251,7 +2280,7 @@ VerilatedContext::VerilatedContext()
    : m_impdatap{new VerilatedContextImpData} {
    Verilated::lastContextp(this);
    Verilated::threadContextp(this);
-    m_ns.m_profThreadsFilename = "profile_threads.dat";
+    m_ns.m_profExecFilename = "profile_exec.dat";
    m_ns.m_profVltFilename = "profile.vlt";
    m_fdps.resize(31);
    std::fill(m_fdps.begin(), m_fdps.end(), static_cast<FILE*>(nullptr));
@@ -2260,7 +2289,17 @@ VerilatedContext::VerilatedContext()
 }

 // Must declare here not in interface, as otherwise forward declarations not known
-VerilatedContext::~VerilatedContext() {}
+VerilatedContext::~VerilatedContext() {
+    checkMagic(this);
+    m_magic = 0x1;  // Arbitrary but 0x1 is what Verilator src uses for a deleted pointer
+}
+
+void VerilatedContext::checkMagic(const VerilatedContext* contextp) {
+    if (VL_UNLIKELY(!contextp || contextp->m_magic != MAGIC)) {
+        VL_FATAL_MT("", 0, "",  // LCOV_EXCL_LINE
+                    "Attempt to create model using a bad/deleted VerilatedContext pointer");
+    }
+}

 VerilatedContext::Serialized::Serialized() {
    m_timeunit = VL_TIME_UNIT;  // Initial value until overriden by _Vconfigure
@@ -2319,21 +2358,21 @@ void VerilatedContext::gotFinish(bool flag) VL_MT_SAFE {
    const VerilatedLockGuard lock{m_mutex};
    m_s.m_gotFinish = flag;
 }
-void VerilatedContext::profThreadsStart(vluint64_t flag) VL_MT_SAFE {
+void VerilatedContext::profExecStart(uint64_t flag) VL_MT_SAFE {
    const VerilatedLockGuard lock{m_mutex};
-    m_ns.m_profThreadsStart = flag;
+    m_ns.m_profExecStart = flag;
 }
-void VerilatedContext::profThreadsWindow(vluint64_t flag) VL_MT_SAFE {
+void VerilatedContext::profExecWindow(uint64_t flag) VL_MT_SAFE {
    const VerilatedLockGuard lock{m_mutex};
-    m_ns.m_profThreadsWindow = flag;
+    m_ns.m_profExecWindow = flag;
 }
-void VerilatedContext::profThreadsFilename(const std::string& flag) VL_MT_SAFE {
+void VerilatedContext::profExecFilename(const std::string& flag) VL_MT_SAFE {
    const VerilatedLockGuard lock{m_mutex};
-    m_ns.m_profThreadsFilename = flag;
+    m_ns.m_profExecFilename = flag;
 }
-std::string VerilatedContext::profThreadsFilename() const VL_MT_SAFE {
+std::string VerilatedContext::profExecFilename() const VL_MT_SAFE {
    const VerilatedLockGuard lock{m_mutex};
-    return m_ns.m_profThreadsFilename;
+    return m_ns.m_profExecFilename;
 }
 void VerilatedContext::profVltFilename(const std::string& flag) VL_MT_SAFE {
    const VerilatedLockGuard lock{m_mutex};
@@ -2478,13 +2517,16 @@ std::pair<int, char**> VerilatedContextImp::argc_argv() VL_MT_SAFE_EXCLUDES(m_ar

 void VerilatedContextImp::commandArgVl(const std::string& arg) {
    if (0 == std::strncmp(arg.c_str(), "+verilator+", std::strlen("+verilator+"))) {
-        std::string value;
+        std::string str;
+        uint64_t u64;
        if (arg == "+verilator+debug") {
            Verilated::debug(4);
-        } else if (commandArgVlValue(arg, "+verilator+debugi+", value /*ref*/)) {
-            Verilated::debug(std::atoi(value.c_str()));
-        } else if (commandArgVlValue(arg, "+verilator+error+limit+", value /*ref*/)) {
-            errorLimit(std::atoi(value.c_str()));
+        } else if (commandArgVlUint64(arg, "+verilator+debugi+", u64, 0,
+                                      std::numeric_limits<int>::max())) {
+            Verilated::debug(static_cast<int>(u64));
+        } else if (commandArgVlUint64(arg, "+verilator+error+limit+", u64, 0,
+                                      std::numeric_limits<int>::max())) {
+            errorLimit(static_cast<int>(u64));
        } else if (arg == "+verilator+help") {
            VerilatedImp::versionDump();
            VL_PRINTF_MT("For help, please see 'verilator --help'\n");
@@ -2492,18 +2534,22 @@ void VerilatedContextImp::commandArgVl(const std::string& arg) {
                        "Exiting due to command line argument (not an error)");
        } else if (arg == "+verilator+noassert") {
            assertOn(false);
-        } else if (commandArgVlValue(arg, "+verilator+prof+threads+start+", value /*ref*/)) {
-            profThreadsStart(std::atoll(value.c_str()));
-        } else if (commandArgVlValue(arg, "+verilator+prof+threads+window+", value /*ref*/)) {
-            profThreadsWindow(std::atol(value.c_str()));
-        } else if (commandArgVlValue(arg, "+verilator+prof+threads+file+", value /*ref*/)) {
-            profThreadsFilename(value);
-        } else if (commandArgVlValue(arg, "+verilator+prof+vlt+file+", value /*ref*/)) {
-            profVltFilename(value);
-        } else if (commandArgVlValue(arg, "+verilator+rand+reset+", value /*ref*/)) {
-            randReset(std::atoi(value.c_str()));
-        } else if (commandArgVlValue(arg, "+verilator+seed+", value /*ref*/)) {
-            randSeed(std::atoi(value.c_str()));
+        } else if (commandArgVlUint64(arg, "+verilator+prof+exec+start+", u64)
+                   || commandArgVlUint64(arg, "+verilator+prof+threads+start+", u64)) {
+            profExecStart(u64);
+        } else if (commandArgVlUint64(arg, "+verilator+prof+exec+window+", u64, 1)
+                   || commandArgVlUint64(arg, "+verilator+prof+threads+window+", u64, 1)) {
+            profExecWindow(u64);
+        } else if (commandArgVlString(arg, "+verilator+prof+exec+file+", str)
+                   || commandArgVlString(arg, "+verilator+prof+threads+file+", str)) {
+            profExecFilename(str);
+        } else if (commandArgVlString(arg, "+verilator+prof+vlt+file+", str)) {
+            profVltFilename(str);
+        } else if (commandArgVlUint64(arg, "+verilator+rand+reset+", u64, 0, 2)) {
+            randReset(static_cast<int>(u64));
+        } else if (commandArgVlUint64(arg, "+verilator+seed+", u64, 1,
+                                      std::numeric_limits<int>::max())) {
+            randSeed(static_cast<int>(u64));
        } else if (arg == "+verilator+V") {
            VerilatedImp::versionDump();  // Someday more info too
            VL_FATAL_MT("COMMAND_LINE", 0, "",
@@ -2513,12 +2559,14 @@ void VerilatedContextImp::commandArgVl(const std::string& arg) {
            VL_FATAL_MT("COMMAND_LINE", 0, "",
                        "Exiting due to command line argument (not an error)");
        } else {
-            VL_PRINTF_MT("%%Warning: Unknown +verilator runtime argument: '%s'\n", arg.c_str());
+            const std::string msg = "Unknown runtime argument: " + arg;
+            VL_FATAL_MT("COMMAND_LINE", 0, "", msg.c_str());
        }
    }
 }
-bool VerilatedContextImp::commandArgVlValue(const std::string& arg, const std::string& prefix,
-                                            std::string& valuer) {
+
+bool VerilatedContextImp::commandArgVlString(const std::string& arg, const std::string& prefix,
+                                             std::string& valuer) {
    const size_t len = prefix.length();
    if (0 == std::strncmp(prefix.c_str(), arg.c_str(), len)) {
        valuer = arg.substr(len);
@@ -2528,6 +2576,30 @@ bool VerilatedContextImp::commandArgVlValue(const std::string& arg, const std::s
    }
 }

+bool VerilatedContextImp::commandArgVlUint64(const std::string& arg, const std::string& prefix,
+                                             uint64_t& valuer, uint64_t min, uint64_t max) {
+    std::string str;
+    if (commandArgVlString(arg, prefix, str)) {
+        const auto fail = [&](const std::string& extra = "") {
+            std::stringstream ss;
+            ss << "Argument '" << prefix << "' must be an unsigned integer";
+            if (min != std::numeric_limits<uint64_t>::min()) ss << ", greater than " << min - 1;
+            if (max != std::numeric_limits<uint64_t>::max()) ss << ", less than " << max + 1;
+            if (!extra.empty()) ss << ". " << extra;
+            const std::string& msg = ss.str();
+            VL_FATAL_MT("COMMAND_LINE", 0, "", msg.c_str());
+        };
+
+        if (std::any_of(str.begin(), str.end(), [](int c) { return !std::isdigit(c); })) fail();
+        char* end;
+        valuer = std::strtoull(str.c_str(), &end, 10);
+        if (errno == ERANGE) fail("Value out of range of uint64_t");
+        if (valuer < min || valuer > max) fail();
+        return true;
+    }
+    return false;
+}
+
 //======================================================================
 // VerilatedContext:: + VerilatedContextImp:: Methods - random

@@ -2536,20 +2608,19 @@ void VerilatedContext::randSeed(int val) VL_MT_SAFE {
    // and so the rand seed's mutex must also be static
    const VerilatedLockGuard lock{VerilatedContextImp::s().s_randMutex};
    m_s.m_randSeed = val;
-    const vluint64_t newEpoch = VerilatedContextImp::s().s_randSeedEpoch + 1;
+    const uint64_t newEpoch = VerilatedContextImp::s().s_randSeedEpoch + 1;
    // Obververs must see new epoch AFTER seed updated
 #ifdef VL_THREADED
    std::atomic_signal_fence(std::memory_order_release);
 #endif
    VerilatedContextImp::s().s_randSeedEpoch = newEpoch;
 }
-vluint64_t VerilatedContextImp::randSeedDefault64() const VL_MT_SAFE {
+uint64_t VerilatedContextImp::randSeedDefault64() const VL_MT_SAFE {
    if (randSeed() != 0) {
-        return ((static_cast<vluint64_t>(randSeed()) << 32)
-                ^ (static_cast<vluint64_t>(randSeed())));
+        return ((static_cast<uint64_t>(randSeed()) << 32) ^ (static_cast<uint64_t>(randSeed())));
    } else {
-        return ((static_cast<vluint64_t>(vl_sys_rand32()) << 32)
-                ^ (static_cast<vluint64_t>(vl_sys_rand32())));
+        return ((static_cast<uint64_t>(vl_sys_rand32()) << 32)
+                ^ (static_cast<uint64_t>(vl_sys_rand32())));
    }
 }

@@ -2596,6 +2667,7 @@ const VerilatedScopeNameMap* VerilatedContext::scopeNameMap() VL_MT_SAFE {

 VerilatedSyms::VerilatedSyms(VerilatedContext* contextp)
    : _vm_contextp__(contextp ? contextp : Verilated::threadContextp()) {
+    VerilatedContext::checkMagic(_vm_contextp__);
    Verilated::threadContextp(_vm_contextp__);
 #ifdef VL_THREADED
    __Vm_evalMsgQp = new VerilatedEvalMsgQueue;
@@ -2603,6 +2675,7 @@ VerilatedSyms::VerilatedSyms(VerilatedContext* contextp)
 }

 VerilatedSyms::~VerilatedSyms() {
+    VerilatedContext::checkMagic(_vm_contextp__);
 #ifdef VL_THREADED
    delete __Vm_evalMsgQp;
 #endif
@@ -2625,14 +2698,11 @@ void Verilated::debug(int level) VL_MT_SAFE {
    }
 }

-const char* Verilated::catName(const char* n1, const char* n2, int scopet,
-                               const char* delimiter) VL_MT_SAFE {
-    // Returns new'ed data
+const char* Verilated::catName(const char* n1, const char* n2, const char* delimiter) VL_MT_SAFE {
    // Used by symbol table creation to make module names
    static VL_THREAD_LOCAL char* t_strp = nullptr;
    static VL_THREAD_LOCAL size_t t_len = 0;
-    const size_t newlen
-        = std::strlen(n1) + std::strlen(n2) + std::strlen(delimiter) + (scopet > 0 ? 2 : 1);
+    const size_t newlen = std::strlen(n1) + std::strlen(n2) + std::strlen(delimiter) + 1;
    if (VL_UNLIKELY(!t_strp || newlen > t_len)) {
        if (t_strp) delete[] t_strp;
        t_strp = new char[newlen];
@@ -2640,8 +2710,6 @@ const char* Verilated::catName(const char* n1, const char* n2, int scopet,
    }
    char* dp = t_strp;
    for (const char* sp = n1; *sp;) *dp++ = *sp++;
-    // Add scope type
-    if (scopet) *dp++ = (char)(0x80 + scopet);
    for (const char* sp = delimiter; *sp;) *dp++ = *sp++;
    for (const char* sp = n2; *sp;) *dp++ = *sp++;
    *dp++ = '\0';
@@ -2798,8 +2866,8 @@ VerilatedModule::~VerilatedModule() {
 // VerilatedVar:: Methods

 // cppcheck-suppress unusedFunction  // Used by applications
-vluint32_t VerilatedVarProps::entSize() const {
-    vluint32_t size = 1;
+uint32_t VerilatedVarProps::entSize() const {
+    uint32_t size = 1;
    switch (vltype()) {
    case VLVT_PTR: size = sizeof(void*); break;
    case VLVT_UINT8: size = sizeof(CData); break;
@@ -2822,7 +2890,7 @@ void* VerilatedVarProps::datapAdjustIndex(void* datap, int dim, int indx) const
    if (VL_UNLIKELY(dim <= 0 || dim > udims())) return nullptr;
    if (VL_UNLIKELY(indx < low(dim) || indx > high(dim))) return nullptr;
    const int indxAdj = indx - low(dim);
-    vluint8_t* bytep = reinterpret_cast<vluint8_t*>(datap);
+    uint8_t* bytep = reinterpret_cast<uint8_t*>(datap);
    // If on index 1 of a 2 index array, then each index 1 is index2sz*entsz
    size_t slicesz = entSize();
    for (int d = dim + 1; d <= m_udims; ++d) slicesz *= elements(d);
@@ -2843,7 +2911,7 @@ VerilatedScope::~VerilatedScope() {
 }

 void VerilatedScope::configure(VerilatedSyms* symsp, const char* prefixp, const char* suffixp,
-                               const char* identifier, vlsint8_t timeunit,
+                               const char* identifier, int8_t timeunit,
                               const Type& type) VL_MT_UNSAFE {
    // Slowpath - called once/scope at construction
    // We don't want the space and reference-count access overhead of strings.
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -33,6 +33,7 @@

 // clang-format off
 #include "verilatedos.h"
+#include "verilated_config.h"
 #if VM_SC
 # include "verilated_sc.h"  // Get SYSTEMC_VERSION and time declarations
 #endif
@@ -86,6 +87,7 @@ class VerilatedCovContext;
 class VerilatedEvalMsgQueue;
 class VerilatedFst;
 class VerilatedFstC;
+class VerilatedFstSc;
 class VerilatedScope;
 class VerilatedScopeNameMap;
 class VerilatedVar;
@@ -99,11 +101,11 @@ class VerilatedVcdSc;

 // clang-format off
 //    P                     // Packed data of bit type (C/S/I/Q/W)
-using CData = vluint8_t;    ///< Data representing 'bit' of 1-8 packed bits
-using SData = vluint16_t;   ///< Data representing 'bit' of 9-16 packed bits
-using IData = vluint32_t;   ///< Data representing 'bit' of 17-32 packed bits
-using QData = vluint64_t;   ///< Data representing 'bit' of 33-64 packed bits
-using EData = vluint32_t;   ///< Data representing one element of WData array
+using CData = uint8_t;    ///< Data representing 'bit' of 1-8 packed bits
+using SData = uint16_t;   ///< Data representing 'bit' of 9-16 packed bits
+using IData = uint32_t;   ///< Data representing 'bit' of 17-32 packed bits
+using QData = uint64_t;   ///< Data representing 'bit' of 33-64 packed bits
+using EData = uint32_t;   ///< Data representing one element of WData array
 using WData = EData;        ///< Data representing >64 packed bits (used as pointer)
 //    F     = float;        // No typedef needed; Verilator uses float
 //    D     = double;       // No typedef needed; Verilator uses double
@@ -113,7 +115,7 @@ using WData = EData;        ///< Data representing >64 packed bits (used as poin
 using WDataInP = const WData*;  ///< 'bit' of >64 packed bits as array input to a function
 using WDataOutP = WData*;  ///< 'bit' of >64 packed bits as array output from a function

-enum VerilatedVarType : vluint8_t {
+enum VerilatedVarType : uint8_t {
    VLVT_UNKNOWN = 0,
    VLVT_PTR,  // Pointer to something
    VLVT_UINT8,  // AKA CData
@@ -141,11 +143,11 @@ enum VerilatedVarFlags {
 // Mutex and threading support

 // Return current thread ID (or 0), not super fast, cache if needed
-extern vluint32_t VL_THREAD_ID() VL_MT_SAFE;
+extern uint32_t VL_THREAD_ID() VL_MT_SAFE;

 #if VL_THREADED

-#define VL_LOCK_SPINS 50000  /// Number of times to spin for a mutex before relaxing
+#define VL_LOCK_SPINS 50000  /// Number of times to spin for a mutex before yielding

 /// Mutex, wrapped to allow -fthread_safety checks
 class VL_CAPABILITY("mutex") VerilatedMutex final {
@@ -224,7 +226,7 @@ public:
 class VerilatedAssertOneThread final {
    // MEMBERS
 #if defined(VL_THREADED) && defined(VL_DEBUG)
-    vluint32_t m_threadid;  // Thread that is legal
+    uint32_t m_threadid;  // Thread that is legal
 public:
    // CONSTRUCTORS
    // The constructor establishes the thread id for all later calls.
@@ -318,10 +320,10 @@ protected:
        bool m_fatalOnVpiError = true;  // Fatal on vpi error/unsupported
        bool m_gotError = false;  // A $finish statement executed
        bool m_gotFinish = false;  // A $finish or $stop statement executed
-        vluint64_t m_time = 0;  // Current $time (unscaled), 0=at zero, or legacy
+        uint64_t m_time = 0;  // Current $time (unscaled), 0=at zero, or legacy
        // Slow path
-        vlsint8_t m_timeunit;  // Time unit as 0..15
-        vlsint8_t m_timeprecision;  // Time precision as 0..15
+        int8_t m_timeunit;  // Time unit as 0..15
+        int8_t m_timeprecision;  // Time precision as 0..15
        int m_errorCount = 0;  // Number of errors
        int m_errorLimit = 1;  // Stop on error number
        int m_randReset = 0;  // Random reset: 0=all 0s, 1=all 1s, 2=random
@@ -342,10 +344,10 @@ protected:
    struct NonSerialized {  // Non-serialized information
        // These are reloaded from on command-line settings, so do not need to persist
        // Fast path
-        vluint64_t m_profThreadsStart = 1;  // +prof+threads starting time
-        vluint32_t m_profThreadsWindow = 2;  // +prof+threads window size
+        uint64_t m_profExecStart = 1;  // +prof+exec+start time
+        uint32_t m_profExecWindow = 2;  // +prof+exec+window size
        // Slow path
-        std::string m_profThreadsFilename;  // +prof+threads filename
+        std::string m_profExecFilename;  // +prof+exec+file filename
        std::string m_profVltFilename;  // +prof+vlt filename
    } m_ns;

@@ -372,6 +374,10 @@ protected:
    // List of free descriptors in the MCT region [4, 32)
    std::vector<IData> m_fdFreeMct VL_GUARDED_BY(m_fdMutex);

+    // Magic to check for bad construction
+    static constexpr uint64_t MAGIC = 0xC35F9A6E5298EE6EULL;  // SHA256 "VerilatedContext"
+    uint64_t m_magic = MAGIC;
+
 private:
    // CONSTRUCTORS
    VL_UNCOPYABLE(VerilatedContext);
@@ -466,16 +472,16 @@ public:
    /// timeInc, operating on the thread's context.
    ///
    /// * Else, if VL_TIME_STAMP64 is defined, time comes from the legacy
-    /// 'vluint64_t vl_time_stamp64()' which must a function be defined by
+    /// 'uint64_t vl_time_stamp64()' which must a function be defined by
    /// the user's wrapper.
    ///
    /// * Else, time comes from the legacy 'double sc_time_stamp()' which
    /// must be a function defined by the user's wrapper.
-    vluint64_t time() const VL_MT_SAFE;
+    uint64_t time() const VL_MT_SAFE;
    /// Set current simulation time. See time() for side effect details
-    void time(vluint64_t value) VL_MT_SAFE { m_s.m_time = value; }
+    void time(uint64_t value) VL_MT_SAFE { m_s.m_time = value; }
    /// Advance current simulation time. See time() for side effect details
-    void timeInc(vluint64_t add) VL_MT_UNSAFE { m_s.m_time += add; }
+    void timeInc(uint64_t add) VL_MT_UNSAFE { m_s.m_time += add; }
    /// Return time units as power-of-ten
    int timeunit() const VL_MT_SAFE { return -m_s.m_timeunit; }
    /// Set time units as power-of-ten
@@ -516,13 +522,13 @@ public:  // But for internal use only
    std::string dumpfile() const VL_MT_SAFE_EXCLUDES(m_timeDumpMutex);
    std::string dumpfileCheck() const VL_MT_SAFE_EXCLUDES(m_timeDumpMutex);

-    // Internal: --prof-threads related settings
-    void profThreadsStart(vluint64_t flag) VL_MT_SAFE;
-    vluint64_t profThreadsStart() const VL_MT_SAFE { return m_ns.m_profThreadsStart; }
-    void profThreadsWindow(vluint64_t flag) VL_MT_SAFE;
-    vluint32_t profThreadsWindow() const VL_MT_SAFE { return m_ns.m_profThreadsWindow; }
-    void profThreadsFilename(const std::string& flag) VL_MT_SAFE;
-    std::string profThreadsFilename() const VL_MT_SAFE;
+    // Internal: --prof-exec related settings
+    void profExecStart(uint64_t flag) VL_MT_SAFE;
+    uint64_t profExecStart() const VL_MT_SAFE { return m_ns.m_profExecStart; }
+    void profExecWindow(uint64_t flag) VL_MT_SAFE;
+    uint32_t profExecWindow() const VL_MT_SAFE { return m_ns.m_profExecWindow; }
+    void profExecFilename(const std::string& flag) VL_MT_SAFE;
+    std::string profExecFilename() const VL_MT_SAFE;
    void profVltFilename(const std::string& flag) VL_MT_SAFE;
    std::string profVltFilename() const VL_MT_SAFE;

@@ -533,6 +539,10 @@ public:  // But for internal use only
    // Internal: Serialization setup
    static constexpr size_t serialized1Size() VL_PURE { return sizeof(m_s); }
    void* serialized1Ptr() VL_MT_UNSAFE { return &m_s; }
+
+    // Internal: Check magic number
+    static void checkMagic(const VerilatedContext* contextp);
+    void selfTestClearMagic() { m_magic = 0x2; }
 };

 //===========================================================================
@@ -557,7 +567,7 @@ public:  // But for internal use only

 class VerilatedScope final {
 public:
-    enum Type : vluint8_t {
+    enum Type : uint8_t {
        SCOPE_MODULE,
        SCOPE_OTHER
    };  // Type of a scope, currently module is only interesting
@@ -570,21 +580,21 @@ private:
    VerilatedVarNameMap* m_varsp = nullptr;  // Variable map
    const char* m_namep = nullptr;  // Scope name (Slowpath)
    const char* m_identifierp = nullptr;  // Identifier of scope (with escapes removed)
-    vlsint8_t m_timeunit = 0;  // Timeunit in negative power-of-10
+    int8_t m_timeunit = 0;  // Timeunit in negative power-of-10
    Type m_type = SCOPE_OTHER;  // Type of the scope

 public:  // But internals only - called from VerilatedModule's
    VerilatedScope() = default;
    ~VerilatedScope();
    void configure(VerilatedSyms* symsp, const char* prefixp, const char* suffixp,
-                   const char* identifier, vlsint8_t timeunit, const Type& type) VL_MT_UNSAFE;
+                   const char* identifier, int8_t timeunit, const Type& type) VL_MT_UNSAFE;
    void exportInsert(int finalize, const char* namep, void* cb) VL_MT_UNSAFE;
    void varInsert(int finalize, const char* namep, void* datap, bool isParam,
                   VerilatedVarType vltype, int vlflags, int dims, ...) VL_MT_UNSAFE;
    // ACCESSORS
    const char* name() const { return m_namep; }
    const char* identifier() const { return m_identifierp; }
-    vlsint8_t timeunit() const { return m_timeunit; }
+    int8_t timeunit() const { return m_timeunit; }
    inline VerilatedSyms* symsp() const { return m_symsp; }
    VerilatedVar* varFind(const char* namep) const VL_MT_SAFE_POSTINIT;
    VerilatedVarNameMap* varsp() const VL_MT_SAFE_POSTINIT { return m_varsp; }
@@ -634,9 +644,9 @@ class Verilated final {
        // Fast path
        VerilatedContext* t_contextp = nullptr;  // Thread's context
 #ifdef VL_THREADED
-        vluint32_t t_mtaskId = 0;  // mtask# executing on this thread
+        uint32_t t_mtaskId = 0;  // mtask# executing on this thread
        // Messages maybe pending on thread, needs end-of-eval calls
-        vluint32_t t_endOfEvalReqd = 0;
+        uint32_t t_endOfEvalReqd = 0;
 #endif
        const VerilatedScope* t_dpiScopep = nullptr;  // DPI context scope
        const char* t_dpiFilename = nullptr;  // DPI context filename
@@ -758,11 +768,11 @@ public:
    /// Return VerilatedContext::randSeed using current thread's VerilatedContext
    static int randSeed() VL_MT_SAFE { return Verilated::threadContextp()->randSeed(); }
    /// Call VerilatedContext::time using current thread's VerilatedContext
-    static void time(vluint64_t val) VL_MT_SAFE { Verilated::threadContextp()->time(val); }
+    static void time(uint64_t val) VL_MT_SAFE { Verilated::threadContextp()->time(val); }
    /// Return VerilatedContext::time using current thread's VerilatedContext
-    static vluint64_t time() VL_MT_SAFE { return Verilated::threadContextp()->time(); }
+    static uint64_t time() VL_MT_SAFE { return Verilated::threadContextp()->time(); }
    /// Call VerilatedContext::timeInc using current thread's VerilatedContext
-    static void timeInc(vluint64_t add) VL_MT_UNSAFE { Verilated::threadContextp()->timeInc(add); }
+    static void timeInc(uint64_t add) VL_MT_UNSAFE { Verilated::threadContextp()->timeInc(add); }
    // Deprecated
    static int timeunit() VL_MT_SAFE { return Verilated::threadContextp()->timeunit(); }
    static int timeprecision() VL_MT_SAFE { return Verilated::threadContextp()->timeprecision(); }
@@ -823,8 +833,8 @@ public:
 public:
    // METHODS - INTERNAL USE ONLY (but public due to what uses it)
    // Internal: Create a new module name by concatenating two strings
-    static const char* catName(const char* n1, const char* n2, int scopet = 0,
-                               const char* delimiter = ".");  // Returns static data
+    // Returns pointer to thread-local static data (overwritten on next call)
+    static const char* catName(const char* n1, const char* n2, const char* delimiter = ".");

    // Internal: Throw signal assertion
    static void nullPointerError(const char* filename, int linenum) VL_ATTR_NORETURN VL_MT_SAFE;
@@ -848,8 +858,8 @@ public:
 #ifdef VL_THREADED
    // Internal: Set the mtaskId, called when an mtask starts
    // Per thread, so no need to be in VerilatedContext
-    static void mtaskId(vluint32_t id) VL_MT_SAFE { t_s.t_mtaskId = id; }
-    static vluint32_t mtaskId() VL_MT_SAFE { return t_s.t_mtaskId; }
+    static void mtaskId(uint32_t id) VL_MT_SAFE { t_s.t_mtaskId = id; }
+    static uint32_t mtaskId() VL_MT_SAFE { return t_s.t_mtaskId; }
    static void endOfEvalReqdInc() VL_MT_SAFE { ++t_s.t_endOfEvalReqd; }
    static void endOfEvalReqdDec() VL_MT_SAFE { --t_s.t_endOfEvalReqd; }

@@ -2,7 +2,7 @@
 ######################################################################
 # DESCRIPTION: Makefile commands for all verilated target files
 #
-# Copyright 2003-2021 by Wilson Snyder. This program is free software; you
+# Copyright 2003-2022 by Wilson Snyder. This program is free software; you
 # can redistribute it and/or modify it under the terms of either the GNU
 # Lesser General Public License Version 3 or the Perl Artistic License
 # Version 2.0.
@@ -56,6 +56,7 @@ VK_CPPFLAGS_ALWAYS += \
 		-DVM_SC=$(VM_SC) \
 		-DVM_TRACE=$(VM_TRACE) \
 		-DVM_TRACE_FST=$(VM_TRACE_FST) \
+		-DVM_TRACE_VCD=$(VM_TRACE_VCD) \
 		$(CFG_CXXFLAGS_NO_UNUSED) \

 ifeq ($(CFG_WITH_CCWARN),yes)	# Local... Else don't burden users
@@ -193,6 +194,10 @@ VK_USER_OBJS   = $(addsuffix .o, $(VM_USER_CLASSES))
 # but keeping the distinction for compatibility for now.
 VK_GLOBAL_OBJS = $(addsuffix .o, $(VM_GLOBAL_FAST) $(VM_GLOBAL_SLOW))

+# Need to re-build if the generated makefile changes, as compiler options might
+# have changed.
+$(VK_GLOBAL_OBJS): $(VM_PREFIX).mk
+
 ifneq ($(VM_PARALLEL_BUILDS),1)
  # Fast build for small designs: All .cpp files in one fell swoop. This
  # saves total compute, but can be slower if only a little changes. It is
@@ -223,7 +228,9 @@ endif
 	$(info Archive $(AR) -rcs $@ $^)
 	$(foreach L, $(filter-out %.a,$^), $(shell echo $L >>[email protected]_deplist.tmp))
 	@if test $(words $(filter %.a,$^)) -eq 0; then \
-		$(AR) -rcs $@ @[email protected]_deplist.tmp; \
+		$(RM) -f $@; \
+		cat [email protected]_deplist.tmp | xargs $(AR) -rc $@; \
+		$(AR) -s $@; \
 	else \
 		$(RM) -rf [email protected]; \
 		for archive in $(filter %.a,$^); do \
@@ -232,7 +239,9 @@ endif
 			$(AR) -x ../../$${archive}; \
 			cd ../..; \
 		done; \
-		$(AR) -rcs $@ @[email protected]_deplist.tmp [email protected]/*/*.o; \
+		$(RM) -f $@; \
+		cat [email protected]_deplist.tmp | xargs $(AR) -rc $@; \
+		$(AR) -rcs $@ [email protected]/*/*.o; \
 	fi \
 	; $(RM) -rf [email protected]_deplist.tmp [email protected]

@@ -2,7 +2,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -20,6 +20,11 @@
 // Autoconf substitutes this with the strings from AC_INIT.
 #define VERILATOR_PRODUCT "@PACKAGE_NAME@"

-/// Verilator version name, e.g. "1.000 2000-01-01"
+/// Verilator version name, e.g. "1.002 2000-01-01"
 // Autoconf substitutes this with the strings from AC_INIT.
 #define VERILATOR_VERSION "@PACKAGE_VERSION@"
+
+/// Verilator version number as integer
+/// As major * 100000 + minor * 1000, e.g. 1002000 == 1.002
+// Autoconf substitutes this with the strings from AC_INIT.
+#define VERILATOR_VERSION_INTEGER @VERILATOR_VERSION_INTEGER@
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -29,6 +29,7 @@
 #include <deque>
 #include <fstream>
 #include <map>
+#include <utility>

 //=============================================================================
 // VerilatedCovConst
@@ -58,7 +59,7 @@ public:  // But only local to this file
        }
    }
    virtual ~VerilatedCovImpItem() = default;
-    virtual vluint64_t count() const = 0;
+    virtual uint64_t count() const = 0;
    virtual void zero() const = 0;
 };

@@ -75,7 +76,7 @@ private:
 public:
    // METHODS
    // cppcheck-suppress truncLongCastReturn
-    virtual vluint64_t count() const override { return *m_countp; }
+    virtual uint64_t count() const override { return *m_countp; }
    virtual void zero() const override { *m_countp = 0; }
    // CONSTRUCTORS
    // cppcheck-suppress noExplicitConstructor
@@ -102,7 +103,7 @@ private:
    using ItemList = std::deque<VerilatedCovImpItem*>;

    // MEMBERS
-    VerilatedMutex m_mutex;  // Protects all members
+    mutable VerilatedMutex m_mutex;  // Protects all members
    ValueIndexMap m_valueIndexes VL_GUARDED_BY(m_mutex);  // Unique arbitrary value for values
    IndexValueMap m_indexValues VL_GUARDED_BY(m_mutex);  // Unique arbitrary value for keys
    ItemList m_items VL_GUARDED_BY(m_mutex);  // List of all items
@@ -122,10 +123,6 @@ public:
 protected:
    friend class VerilatedCovContext;
    virtual ~VerilatedCovImp() override { clearGuts(); }
-    static VerilatedCovImp& imp() VL_MT_SAFE {
-        static VerilatedCovImp s_singleton;
-        return s_singleton;
-    }

 private:
    // PRIVATE METHODS
@@ -231,7 +228,7 @@ private:
        // Little selftest
 #define SELF_CHECK(got, exp) \
    do { \
-        if ((got) != (exp)) VL_FATAL_MT(__FILE__, __LINE__, "", "%Error: selftest\n"); \
+        if ((got) != (exp)) VL_FATAL_MT(__FILE__, __LINE__, "", "%Error: selftest"); \
    } while (0)
        SELF_CHECK(combineHier("a.b.c", "a.b.c"), "a.b.c");
        SELF_CHECK(combineHier("a.b.c", "a.b"), "a.b*");
@@ -358,7 +355,7 @@ public:
        Verilated::quiesce();
        const VerilatedLockGuard lock{m_mutex};
 #ifndef VM_COVERAGE
-        VL_FATAL_MT("", 0, "", "%Error: Called VerilatedCov::write when VM_COVERAGE disabled\n");
+        VL_FATAL_MT("", 0, "", "%Error: Called VerilatedCov::write when VM_COVERAGE disabled");
 #endif
        selftest();

@@ -371,7 +368,7 @@ public:
        os << "# SystemC::Coverage-3\n";

        // Build list of events; totalize if collapsing hierarchy
-        std::map<const std::string, std::pair<std::string, vluint64_t>> eventCounts;
+        std::map<const std::string, std::pair<std::string, uint64_t>> eventCounts;
        for (const auto& itemp : m_items) {
            std::string name;
            std::string hier;
@@ -438,11 +435,11 @@ void VerilatedCovContext::clearNonMatch(const char* matchp) VL_MT_SAFE {
 }
 void VerilatedCovContext::zero() VL_MT_SAFE { impp()->zero(); }
 void VerilatedCovContext::write(const char* filenamep) VL_MT_SAFE { impp()->write(filenamep); }
-void VerilatedCovContext::_inserti(vluint32_t* itemp) VL_MT_SAFE {
-    impp()->inserti(new VerilatedCoverItemSpec<vluint32_t>{itemp});
+void VerilatedCovContext::_inserti(uint32_t* itemp) VL_MT_SAFE {
+    impp()->inserti(new VerilatedCoverItemSpec<uint32_t>{itemp});
 }
-void VerilatedCovContext::_inserti(vluint64_t* itemp) VL_MT_SAFE {
-    impp()->inserti(new VerilatedCoverItemSpec<vluint64_t>{itemp});
+void VerilatedCovContext::_inserti(uint64_t* itemp) VL_MT_SAFE {
+    impp()->inserti(new VerilatedCoverItemSpec<uint64_t>{itemp});
 }
 void VerilatedCovContext::_insertf(const char* filename, int lineno) VL_MT_SAFE {
    impp()->insertf(filename, lineno);
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -72,7 +72,7 @@ class VerilatedCovImp;
 ///
 /// Example:
 ///
-///     vluint32_t m_cases[10];  // Storage for coverage data
+///     uint32_t m_cases[10];  // Storage for coverage data
 ///     constructor() {
 ///         // Initialize
 ///         for (int i = 0; i < 10; ++i) m_cases[i] = 0;
@@ -126,8 +126,8 @@ public:  // But Internal use only
    // Call _insert1, followed by _insert2 and _insert3
    // Do not call directly; use VL_COVER_INSERT or higher level macros instead
    // _insert1: Remember item pointer with count.  (Not const, as may add zeroing function)
-    void _inserti(vluint32_t* itemp) VL_MT_SAFE;
-    void _inserti(vluint64_t* itemp) VL_MT_SAFE;
+    void _inserti(uint32_t* itemp) VL_MT_SAFE;
+    void _inserti(uint64_t* itemp) VL_MT_SAFE;
    // _insert2: Set default filename and line number
    void _insertf(const char* filename, int lineno) VL_MT_SAFE;
    // _insert3: Set parameters
@@ -157,8 +157,8 @@ protected:
    friend class VerilatedCovImp;
    // CONSTRUCTORS
    // Internal: Only made as part of VerilatedCovImp
-    VerilatedCovContext() {}
-    virtual ~VerilatedCovContext() {}
+    VerilatedCovContext() = default;
+    virtual ~VerilatedCovContext() = default;

    // METHODS
    // Internal: access to implementation class
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2009-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2009-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -417,7 +417,7 @@ static void _vl_svPutBitArrElem(const svOpenArrayHandle d, svBit value, int narg
 }

 //======================================================================
-// DPI accessors that simply call above functions
+// DPI accessors that call above functions

 void* svGetArrElemPtr(const svOpenArrayHandle h, int indx1, ...) {
    const VerilatedDpiOpenVar* const varp = _vl_openhandle_varp(h);
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -83,9 +83,11 @@ static_assert(static_cast<int>(FST_ST_VCD_PROGRAM) == static_cast<int>(VLT_TRACE
 //=============================================================================
 // Specialization of the generics for this trace format

-#define VL_DERIVED_T VerilatedFst
-#include "verilated_trace_imp.cpp"
-#undef VL_DERIVED_T
+#define VL_SUB_T VerilatedFst
+#define VL_BUF_T VerilatedFstBuffer
+#include "verilated_trace_imp.h"
+#undef VL_SUB_T
+#undef VL_BUF_T

 //=============================================================================
 // VerilatedFst
@@ -111,7 +113,7 @@ void VerilatedFst::open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex) {

    m_curScope.clear();

-    VerilatedTrace<VerilatedFst>::traceInit();
+    Super::traceInit();

    // Clear the scope stack
    auto it = m_curScope.begin();
@@ -122,7 +124,7 @@ void VerilatedFst::open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex) {

    // convert m_code2symbol into an array for fast lookup
    if (!m_symbolp) {
-        m_symbolp = new fstHandle[nextCode()];
+        m_symbolp = new fstHandle[nextCode()]{0};
        for (const auto& i : m_code2symbol) m_symbolp[i.first] = i.second;
    }
    m_code2symbol.clear();
@@ -133,23 +135,23 @@ void VerilatedFst::open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex) {

 void VerilatedFst::close() VL_MT_SAFE_EXCLUDES(m_mutex) {
    const VerilatedLockGuard lock{m_mutex};
-    VerilatedTrace<VerilatedFst>::closeBase();
+    Super::closeBase();
    fstWriterClose(m_fst);
    m_fst = nullptr;
 }

 void VerilatedFst::flush() VL_MT_SAFE_EXCLUDES(m_mutex) {
    const VerilatedLockGuard lock{m_mutex};
-    VerilatedTrace<VerilatedFst>::flushBase();
+    Super::flushBase();
    fstWriterFlushContext(m_fst);
 }

-void VerilatedFst::emitTimeChange(vluint64_t timeui) { fstWriterEmitTimeChange(m_fst, timeui); }
+void VerilatedFst::emitTimeChange(uint64_t timeui) { fstWriterEmitTimeChange(m_fst, timeui); }

 //=============================================================================
 // Decl

-void VerilatedFst::declDTypeEnum(int dtypenum, const char* name, vluint32_t elements,
+void VerilatedFst::declDTypeEnum(int dtypenum, const char* name, uint32_t elements,
                                 unsigned int minValbits, const char** itemNamesp,
                                 const char** itemValuesp) {
    const fstEnumHandle enumNum
@@ -157,19 +159,21 @@ void VerilatedFst::declDTypeEnum(int dtypenum, const char* name, vluint32_t elem
    m_local2fstdtype[dtypenum] = enumNum;
 }

-void VerilatedFst::declare(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
-                           fstVarType vartype, bool array, int arraynum, int msb, int lsb) {
+void VerilatedFst::declare(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+                           fstVarType vartype, bool array, int arraynum, bool bussed, int msb,
+                           int lsb) {
    const int bits = ((msb > lsb) ? (msb - lsb) : (lsb - msb)) + 1;

-    VerilatedTrace<VerilatedFst>::declCode(code, bits, false);
+    const bool enabled = Super::declCode(code, name, bits, false);
+    if (!enabled) return;

-    std::istringstream nameiss{name};
+    std::string nameasstr = namePrefix() + name;
+    std::istringstream nameiss{nameasstr};
    std::istream_iterator<std::string> beg(nameiss);
    std::istream_iterator<std::string> end;
    std::list<std::string> tokens(beg, end);  // Split name
    std::string symbol_name{tokens.back()};
    tokens.pop_back();  // Remove symbol name from hierarchy
-    tokens.insert(tokens.begin(), moduleName());  // Add current module to the hierarchy
    std::string tmpModName;

    // Find point where current and new scope diverge
@@ -195,9 +199,9 @@ void VerilatedFst::declare(vluint32_t code, const char* name, int dtypenum, fstV
            // If the scope ends with a non-ascii character, it will be 0x80 + fstScopeType
            fstWriterSetScope(m_fst, static_cast<fstScopeType>(new_it->back() & 0x7f),
                              tmpModName.c_str(), nullptr);
-        } else
+        } else {
            fstWriterSetScope(m_fst, FST_ST_VCD_SCOPE, new_it->c_str(), nullptr);
-
+        }
        m_curScope.push_back(*new_it);
        new_it = tokens.erase(new_it);
    }
@@ -205,6 +209,7 @@ void VerilatedFst::declare(vluint32_t code, const char* name, int dtypenum, fstV
    std::stringstream name_ss;
    name_ss << symbol_name;
    if (array) name_ss << "[" << arraynum << "]";
+    if (bussed) name_ss << " [" << msb << ":" << lsb << "]";
    std::string name_str = name_ss.str();

    if (dtypenum > 0) {
@@ -221,66 +226,95 @@ void VerilatedFst::declare(vluint32_t code, const char* name, int dtypenum, fstV
    }
 }

-void VerilatedFst::declBit(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+void VerilatedFst::declBit(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                           fstVarType vartype, bool array, int arraynum) {
-    declare(code, name, dtypenum, vardir, vartype, array, arraynum, 0, 0);
+    declare(code, name, dtypenum, vardir, vartype, array, arraynum, false, 0, 0);
 }
-void VerilatedFst::declBus(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+void VerilatedFst::declBus(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                           fstVarType vartype, bool array, int arraynum, int msb, int lsb) {
-    declare(code, name, dtypenum, vardir, vartype, array, arraynum, msb, lsb);
+    declare(code, name, dtypenum, vardir, vartype, array, arraynum, true, msb, lsb);
 }
-void VerilatedFst::declQuad(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+void VerilatedFst::declQuad(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                            fstVarType vartype, bool array, int arraynum, int msb, int lsb) {
-    declare(code, name, dtypenum, vardir, vartype, array, arraynum, msb, lsb);
+    declare(code, name, dtypenum, vardir, vartype, array, arraynum, true, msb, lsb);
 }
-void VerilatedFst::declArray(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+void VerilatedFst::declArray(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                             fstVarType vartype, bool array, int arraynum, int msb, int lsb) {
-    declare(code, name, dtypenum, vardir, vartype, array, arraynum, msb, lsb);
+    declare(code, name, dtypenum, vardir, vartype, array, arraynum, true, msb, lsb);
 }
-void VerilatedFst::declDouble(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+void VerilatedFst::declDouble(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                              fstVarType vartype, bool array, int arraynum) {
-    declare(code, name, dtypenum, vardir, vartype, array, arraynum, 63, 0);
+    declare(code, name, dtypenum, vardir, vartype, array, arraynum, false, 63, 0);
 }

+//=============================================================================
+// Get/commit trace buffer
+
+VerilatedFstBuffer* VerilatedFst::getTraceBuffer() { return new VerilatedFstBuffer{*this}; }
+
+void VerilatedFst::commitTraceBuffer(VerilatedFstBuffer* bufp) {
+#ifdef VL_TRACE_OFFLOAD
+    if (bufp->m_offloadBufferWritep) {
+        m_offloadBufferWritep = bufp->m_offloadBufferWritep;
+        return;  // Buffer will be deleted by the offload thread
+    }
+#endif
+    delete bufp;
+}
+
+//=============================================================================
+// VerilatedFstBuffer implementation
+
+VerilatedFstBuffer::VerilatedFstBuffer(VerilatedFst& owner)
+    : VerilatedTraceBuffer<VerilatedFst, VerilatedFstBuffer>{owner} {}
+
+//=============================================================================
+// Trace rendering primitives
+
 // Note: emit* are only ever called from one place (full* in
-// verilated_trace_imp.cpp, which is included in this file at the top),
+// verilated_trace_imp.h, which is included in this file at the top),
 // so always inline them.

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitBit(vluint32_t code, CData newval) {
+void VerilatedFstBuffer::emitBit(uint32_t code, CData newval) {
+    VL_DEBUG_IFDEF(assert(m_symbolp[code]););
    fstWriterEmitValueChange(m_fst, m_symbolp[code], newval ? "1" : "0");
 }

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitCData(vluint32_t code, CData newval, int bits) {
+void VerilatedFstBuffer::emitCData(uint32_t code, CData newval, int bits) {
    char buf[VL_BYTESIZE];
+    VL_DEBUG_IFDEF(assert(m_symbolp[code]););
    cvtCDataToStr(buf, newval << (VL_BYTESIZE - bits));
    fstWriterEmitValueChange(m_fst, m_symbolp[code], buf);
 }

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitSData(vluint32_t code, SData newval, int bits) {
+void VerilatedFstBuffer::emitSData(uint32_t code, SData newval, int bits) {
    char buf[VL_SHORTSIZE];
+    VL_DEBUG_IFDEF(assert(m_symbolp[code]););
    cvtSDataToStr(buf, newval << (VL_SHORTSIZE - bits));
    fstWriterEmitValueChange(m_fst, m_symbolp[code], buf);
 }

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitIData(vluint32_t code, IData newval, int bits) {
+void VerilatedFstBuffer::emitIData(uint32_t code, IData newval, int bits) {
    char buf[VL_IDATASIZE];
+    VL_DEBUG_IFDEF(assert(m_symbolp[code]););
    cvtIDataToStr(buf, newval << (VL_IDATASIZE - bits));
    fstWriterEmitValueChange(m_fst, m_symbolp[code], buf);
 }

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitQData(vluint32_t code, QData newval, int bits) {
+void VerilatedFstBuffer::emitQData(uint32_t code, QData newval, int bits) {
    char buf[VL_QUADSIZE];
+    VL_DEBUG_IFDEF(assert(m_symbolp[code]););
    cvtQDataToStr(buf, newval << (VL_QUADSIZE - bits));
    fstWriterEmitValueChange(m_fst, m_symbolp[code], buf);
 }

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitWData(vluint32_t code, const WData* newvalp, int bits) {
+void VerilatedFstBuffer::emitWData(uint32_t code, const WData* newvalp, int bits) {
    int words = VL_WORDS_I(bits);
    char* wp = m_strbuf;
    // Convert the most significant word
@@ -296,6 +330,6 @@ void VerilatedFst::emitWData(vluint32_t code, const WData* newvalp, int bits) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedFst::emitDouble(vluint32_t code, double newval) {
+void VerilatedFstBuffer::emitDouble(uint32_t code, double newval) {
    fstWriterEmitValueChange(m_fst, m_symbolp[code], &newval);
 }
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -31,21 +31,25 @@
 #include <string>
 #include <vector>

+class VerilatedFstBuffer;
+
 //=============================================================================
 // VerilatedFst
 // Base class to create a Verilator FST dump
 // This is an internally used class - see VerilatedFstC for what to call from applications

-class VerilatedFst final : public VerilatedTrace<VerilatedFst> {
+class VerilatedFst final : public VerilatedTrace<VerilatedFst, VerilatedFstBuffer> {
+public:
+    using Super = VerilatedTrace<VerilatedFst, VerilatedFstBuffer>;
+
 private:
-    // Give the superclass access to private bits (to avoid virtual functions)
-    friend class VerilatedTrace<VerilatedFst>;
+    friend Buffer;  // Give the buffer access to the private bits

    //=========================================================================
    // FST specific internals

    void* m_fst;
-    std::map<vluint32_t, fstHandle> m_code2symbol;
+    std::map<uint32_t, fstHandle> m_code2symbol;
    std::map<int, fstEnumHandle> m_local2fstdtype;
    std::list<std::string> m_curScope;
    fstHandle* m_symbolp = nullptr;  // same as m_code2symbol, but as an array
@@ -53,38 +57,33 @@ private:

    // CONSTRUCTORS
    VL_UNCOPYABLE(VerilatedFst);
-    void declare(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
-                 fstVarType vartype, bool array, int arraynum, int msb, int lsb);
+    void declare(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+                 fstVarType vartype, bool array, int arraynum, bool bussed, int msb, int lsb);

 protected:
    //=========================================================================
    // Implementation of VerilatedTrace interface

-    // Implementations of protected virtual methods for VerilatedTrace
-    virtual void emitTimeChange(vluint64_t timeui) override;
+    // Called when the trace moves forward to a new time point
+    virtual void emitTimeChange(uint64_t timeui) override;

    // Hooks called from VerilatedTrace
    virtual bool preFullDump() override { return isOpen(); }
    virtual bool preChangeDump() override { return isOpen(); }

-    // Implementations of duck-typed methods for VerilatedTrace. These are
-    // called from only one place (namely full*) so always inline them.
-    inline void emitBit(vluint32_t code, CData newval);
-    inline void emitCData(vluint32_t code, CData newval, int bits);
-    inline void emitSData(vluint32_t code, SData newval, int bits);
-    inline void emitIData(vluint32_t code, IData newval, int bits);
-    inline void emitQData(vluint32_t code, QData newval, int bits);
-    inline void emitWData(vluint32_t code, const WData* newvalp, int bits);
-    inline void emitDouble(vluint32_t code, double newval);
+    // Trace buffer management
+    virtual VerilatedFstBuffer* getTraceBuffer() override;
+    virtual void commitTraceBuffer(VerilatedFstBuffer*) override;

 public:
    //=========================================================================
    // External interface to client code
-    // (All must be threadsafe)

+    // CONSTRUCTOR
    explicit VerilatedFst(void* fst = nullptr);
    ~VerilatedFst();

+    // METHODS - All must be thread safe
    // Open the file; call isOpen() to see if errors
    void open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex);
    // Close the file
@@ -97,32 +96,65 @@ public:
    //=========================================================================
    // Internal interface to Verilator generated code

-    // Inside dumping routines, declare a data type
-    void declDTypeEnum(int dtypenum, const char* name, vluint32_t elements,
-                       unsigned int minValbits, const char** itemNamesp, const char** itemValuesp);
-
-    // Inside dumping routines, declare a signal
-    void declBit(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+    void declBit(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                 fstVarType vartype, bool array, int arraynum);
-    void declBus(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+    void declBus(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                 fstVarType vartype, bool array, int arraynum, int msb, int lsb);
-    void declQuad(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+    void declQuad(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                  fstVarType vartype, bool array, int arraynum, int msb, int lsb);
-    void declArray(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+    void declArray(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                   fstVarType vartype, bool array, int arraynum, int msb, int lsb);
-    void declDouble(vluint32_t code, const char* name, int dtypenum, fstVarDir vardir,
+    void declDouble(uint32_t code, const char* name, int dtypenum, fstVarDir vardir,
                    fstVarType vartype, bool array, int arraynum);
+
+    void declDTypeEnum(int dtypenum, const char* name, uint32_t elements, unsigned int minValbits,
+                       const char** itemNamesp, const char** itemValuesp);
 };

 #ifndef DOXYGEN
 // Declare specialization here as it's used in VerilatedFstC just below
-template <> void VerilatedTrace<VerilatedFst>::dump(vluint64_t timeui);
-template <> void VerilatedTrace<VerilatedFst>::set_time_unit(const char* unitp);
-template <> void VerilatedTrace<VerilatedFst>::set_time_unit(const std::string& unit);
-template <> void VerilatedTrace<VerilatedFst>::set_time_resolution(const char* unitp);
-template <> void VerilatedTrace<VerilatedFst>::set_time_resolution(const std::string& unit);
+template <> void VerilatedFst::Super::dump(uint64_t time);
+template <> void VerilatedFst::Super::set_time_unit(const char* unitp);
+template <> void VerilatedFst::Super::set_time_unit(const std::string& unit);
+template <> void VerilatedFst::Super::set_time_resolution(const char* unitp);
+template <> void VerilatedFst::Super::set_time_resolution(const std::string& unit);
+template <> void VerilatedFst::Super::dumpvars(int level, const std::string& hier);
 #endif

+//=============================================================================
+// VerilatedFstBuffer
+
+class VerilatedFstBuffer final : public VerilatedTraceBuffer<VerilatedFst, VerilatedFstBuffer> {
+    // Give the trace file access to the private bits
+    friend VerilatedFst;
+    friend VerilatedFst::Super;
+
+    // The FST file handle
+    void* const m_fst = m_owner.m_fst;
+    // code to fstHande map, as an array
+    const fstHandle* const m_symbolp = m_owner.m_symbolp;
+    // String buffer long enough to hold maxBits() chars
+    char* const m_strbuf = m_owner.m_strbuf;
+
+public:
+    // CONSTRUCTOR
+    explicit VerilatedFstBuffer(VerilatedFst& owner);
+    ~VerilatedFstBuffer() = default;
+
+    //=========================================================================
+    // Implementation of VerilatedTraceBuffer interface
+
+    // Implementations of duck-typed methods for VerilatedTraceBuffer. These are
+    // called from only one place (the full* methods), so always inline them.
+    VL_ATTR_ALWINLINE inline void emitBit(uint32_t code, CData newval);
+    VL_ATTR_ALWINLINE inline void emitCData(uint32_t code, CData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitSData(uint32_t code, SData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitIData(uint32_t code, IData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitQData(uint32_t code, QData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitWData(uint32_t code, const WData* newvalp, int bits);
+    VL_ATTR_ALWINLINE inline void emitDouble(uint32_t code, double newval);
+};
+
 //=============================================================================
 // VerilatedFstC
 /// Create a FST dump file in C standalone (no SystemC) simulations.
@@ -139,7 +171,7 @@ public:
    explicit VerilatedFstC(void* filep = nullptr)
        : m_sptrace{filep} {}
    /// Destruct, flush, and close the dump
-    ~VerilatedFstC() { close(); }
+    virtual ~VerilatedFstC() { close(); }

    // METHODS - User called

@@ -154,12 +186,12 @@ public:
    /// Write one cycle of dump data
    /// Call with the current context's time just after eval'ed,
    /// e.g. ->dump(contextp->time())
-    void dump(vluint64_t timeui) { m_sptrace.dump(timeui); }
+    void dump(uint64_t timeui) { m_sptrace.dump(timeui); }
    /// Write one cycle of dump data - backward compatible and to reduce
-    /// conversion warnings.  It's better to use a vluint64_t time instead.
-    void dump(double timestamp) { dump(static_cast<vluint64_t>(timestamp)); }
-    void dump(vluint32_t timestamp) { dump(static_cast<vluint64_t>(timestamp)); }
-    void dump(int timestamp) { dump(static_cast<vluint64_t>(timestamp)); }
+    /// conversion warnings.  It's better to use a uint64_t time instead.
+    void dump(double timestamp) { dump(static_cast<uint64_t>(timestamp)); }
+    void dump(uint32_t timestamp) { dump(static_cast<uint64_t>(timestamp)); }
+    void dump(int timestamp) { dump(static_cast<uint64_t>(timestamp)); }

    // METHODS - Internal/backward compatible
    // \protectedsection
@@ -178,9 +210,14 @@ public:
    void set_time_resolution(const std::string& unit) VL_MT_SAFE {
        m_sptrace.set_time_resolution(unit);
    }
+    // Set variables to dump, using $dumpvars format
+    // If level = 0, dump everything and hier is then ignored
+    void dumpvars(int level, const std::string& hier) VL_MT_SAFE {
+        m_sptrace.dumpvars(level, hier);
+    }

    // Internal class access
-    inline VerilatedFst* spTrace() { return &m_sptrace; };
+    inline VerilatedFst* spTrace() { return &m_sptrace; }
 };

 #endif  // guard
@@ -3,7 +3,7 @@
 //
 // THIS MODULE IS PUBLICLY LICENSED
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -27,6 +27,17 @@
 //======================================================================
 //======================================================================

+void VerilatedFstSc::open(const char* filename) {
+    if (!sc_core::sc_get_curr_simcontext()->elaboration_done()) {
+        vl_fatal(__FILE__, __LINE__, "VerilatedFstSc",
+                 ("%Error: VerilatedFstSc::open(\"" + std::string(filename)
+                  + "\") is called before sc_core::sc_start(). "
+                    "Run sc_core::sc_start(sc_core::SC_ZERO_TIME) before opening a wave file.")
+                     .c_str());
+    }
+    VerilatedFstC::open(filename);
+}
+
 //--------------------------------------------------
 // SystemC 2.1.v1
 // cppcheck-suppress unusedFunction
@@ -1,7 +1,7 @@
 // -*- mode: C++; c-file-style: "cc-mode" -*-
 //=============================================================================
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -26,6 +26,8 @@
 #include "verilated_sc.h"
 #include "verilated_fst_c.h"

+#include <string>
+
 //=============================================================================
 // VerilatedFstSc
 ///
@@ -57,6 +59,9 @@ public:
        if (!delta_cycle) { this->dump(sc_time_stamp().to_double()); }
    }

+    // Override VerilatedFstC. Must be called after starting simulation.
+    virtual void open(const char* filename) /*override*/ VL_MT_SAFE;
+
 private:
    /// Fake outs for linker

@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -28,6 +28,8 @@
 #error "verilated_funcs.h should only be included by verilated.h"
 #endif

+#include <string>
+
 //=========================================================================
 // Extern functions -- User may override -- See verilated.cpp

@@ -43,12 +45,17 @@ extern void vl_finish(const char* filename, int linenum, const char* hier);
 /// Verilator internal code must call VL_FINISH_MT instead, which eventually calls this.
 extern void vl_stop(const char* filename, int linenum, const char* hier);

-/// Routine to call for a couple of fatal messages
+/// Routine to call for fatal messages
 /// User code may wish to replace this function, to do so, define VL_USER_FATAL.
 /// This code does not have to be thread safe.
 /// Verilator internal code must call VL_FINISH_MT instead, which eventually calls this.
 extern void vl_fatal(const char* filename, int linenum, const char* hier, const char* msg);

+/// Routine to call for warning messages
+/// User code may wish to replace this function, to do so, define VL_USER_WARN.
+/// This code does not have to be thread safe.
+extern void vl_warn(const char* filename, int linenum, const char* hier, const char* msg);
+
 //=========================================================================
 // Extern functions -- Slow path

@@ -57,9 +64,12 @@ extern void VL_FINISH_MT(const char* filename, int linenum, const char* hier) VL
 /// Multithread safe wrapper for calls to $stop
 extern void VL_STOP_MT(const char* filename, int linenum, const char* hier,
                       bool maybe = true) VL_MT_SAFE;
-/// Multithread safe wrapper to call for a couple of fatal messages
+/// Multithread safe wrapper to call for fatal messages
 extern void VL_FATAL_MT(const char* filename, int linenum, const char* hier,
                        const char* msg) VL_MT_SAFE;
+/// Multithread safe wrapper to call for warning messages
+extern void VL_WARN_MT(const char* filename, int linenum, const char* hier,
+                       const char* msg) VL_MT_SAFE;

 // clang-format off
 /// Print a string, multithread safe. Eventually VL_PRINTF will get called.
@@ -77,9 +87,10 @@ extern void VL_DBG_MSGF(const char* formatp, ...) VL_ATTR_PRINTF(1) VL_MT_SAFE;
 inline IData VL_RANDOM_I() VL_MT_SAFE { return vl_rand64(); }
 inline QData VL_RANDOM_Q() VL_MT_SAFE { return vl_rand64(); }
 extern WDataOutP VL_RANDOM_W(int obits, WDataOutP outwp);
-extern IData VL_RANDOM_SEEDED_II(IData seed) VL_MT_SAFE;
+extern IData VL_RANDOM_SEEDED_II(IData& seedr) VL_MT_SAFE;
+extern IData VL_URANDOM_SEEDED_II(IData seed) VL_MT_SAFE;
 inline IData VL_URANDOM_RANGE_I(IData hi, IData lo) {
-    const vluint64_t rnd = vl_rand64();
+    const uint64_t rnd = vl_rand64();
    if (VL_LIKELY(hi > lo)) {
        // (hi - lo + 1) can be zero when hi is UINT_MAX and lo is zero
        if (VL_UNLIKELY(hi - lo + 1 == 0)) return rnd;
@@ -101,15 +112,6 @@ extern WDataOutP VL_RAND_RESET_W(int obits, WDataOutP outwp);
 /// Zero reset a signal (slow - else use VL_ZERO_W)
 extern WDataOutP VL_ZERO_RESET_W(int obits, WDataOutP outwp);

-#if VL_THREADED
-/// Return high-precision counter for profiling, or 0x0 if not available
-inline QData VL_RDTSC_Q() {
-    vluint64_t val;
-    VL_RDTSC(val);
-    return val;
-}
-#endif
-
 extern void VL_PRINTTIMESCALE(const char* namep, const char* timeunitp,
                              const VerilatedContext* contextp) VL_MT_SAFE;

@@ -151,10 +153,6 @@ extern const char* vl_mc_scan_plusargs(const char* prefixp);  // PLIish
 // Base macros

 // Return true if data[bit] set; not 0/1 return, but 0/non-zero return.
-#define VL_BITISSET_I(data, bit) ((data) & (VL_UL(1) << VL_BITBIT_I(bit)))
-#define VL_BITISSET_Q(data, bit) ((data) & (1ULL << VL_BITBIT_Q(bit)))
-#define VL_BITISSET_E(data, bit) ((data) & (VL_EUL(1) << VL_BITBIT_E(bit)))
-#define VL_BITISSET_W(data, bit) ((data)[VL_BITWORD_E(bit)] & (VL_EUL(1) << VL_BITBIT_E(bit)))
 #define VL_BITISSETLIMIT_W(data, width, bit) (((bit) < (width)) && VL_BITISSET_W(data, bit))

 // Shift appropriate word by bit. Does not account for wrapping between two words
@@ -212,29 +210,27 @@ static inline QData VL_CVT_Q_D(double lhs) VL_PURE {
 // Return double from lhs (numeric) unsigned
 double VL_ITOR_D_W(int lbits, WDataInP const lwp) VL_PURE;
 static inline double VL_ITOR_D_I(int, IData lhs) VL_PURE {
-    return static_cast<double>(static_cast<vluint32_t>(lhs));
+    return static_cast<double>(static_cast<uint32_t>(lhs));
 }
 static inline double VL_ITOR_D_Q(int, QData lhs) VL_PURE {
-    return static_cast<double>(static_cast<vluint64_t>(lhs));
+    return static_cast<double>(static_cast<uint64_t>(lhs));
 }
 // Return double from lhs (numeric) signed
 double VL_ISTOR_D_W(int lbits, WDataInP const lwp) VL_PURE;
 static inline double VL_ISTOR_D_I(int lbits, IData lhs) VL_PURE {
-    if (lbits == 32) return static_cast<double>(static_cast<vlsint32_t>(lhs));
+    if (lbits == 32) return static_cast<double>(static_cast<int32_t>(lhs));
    VlWide<VL_WQ_WORDS_E> lwp;
    VL_SET_WI(lwp, lhs);
    return VL_ISTOR_D_W(lbits, lwp);
 }
 static inline double VL_ISTOR_D_Q(int lbits, QData lhs) VL_PURE {
-    if (lbits == 64) return static_cast<double>(static_cast<vlsint64_t>(lhs));
+    if (lbits == 64) return static_cast<double>(static_cast<int64_t>(lhs));
    VlWide<VL_WQ_WORDS_E> lwp;
    VL_SET_WQ(lwp, lhs);
    return VL_ISTOR_D_W(lbits, lwp);
 }
 // Return QData from double (numeric)
-static inline IData VL_RTOI_I_D(double lhs) VL_PURE {
-    return static_cast<vlsint32_t>(VL_TRUNC(lhs));
-}
+static inline IData VL_RTOI_I_D(double lhs) VL_PURE { return static_cast<int32_t>(VL_TRUNC(lhs)); }

 // Sign extend such that if MSB set, we get ffff_ffff, else 0s
 // (Requires clean input)
@@ -283,28 +279,28 @@ extern int VL_TIME_STR_CONVERT(const char* strp) VL_PURE;
 #if defined(SYSTEMC_VERSION)
 /// Return current simulation time
 // Already defined: extern sc_time sc_time_stamp();
-inline vluint64_t vl_time_stamp64() { return sc_time_stamp().value(); }
+inline uint64_t vl_time_stamp64() { return sc_time_stamp().value(); }
 #else  // Non-SystemC
 # if !defined(VL_TIME_CONTEXT) && !defined(VL_NO_LEGACY)
 #  ifdef VL_TIME_STAMP64
 // vl_time_stamp64() may be optionally defined by the user to return time.
 // On MSVC++ weak symbols are not supported so must be declared, or define
 // VL_TIME_CONTEXT.
-extern vluint64_t vl_time_stamp64() VL_ATTR_WEAK;
+extern uint64_t vl_time_stamp64() VL_ATTR_WEAK;
 #  else
 // sc_time_stamp() may be optionally defined by the user to return time.
 // On MSVC++ weak symbols are not supported so must be declared, or define
 // VL_TIME_CONTEXT.
 extern double sc_time_stamp() VL_ATTR_WEAK;  // Verilator 4.032 and newer
-inline vluint64_t vl_time_stamp64() {
+inline uint64_t vl_time_stamp64() {
    // clang9.0.1 requires & although we really do want the weak symbol value
-    return VL_LIKELY(&sc_time_stamp) ? static_cast<vluint64_t>(sc_time_stamp()) : 0;
+    return VL_LIKELY(&sc_time_stamp) ? static_cast<uint64_t>(sc_time_stamp()) : 0;
 }
 #  endif
 # endif
 #endif

-inline vluint64_t VerilatedContext::time() const VL_MT_SAFE {
+inline uint64_t VerilatedContext::time() const VL_MT_SAFE {
    // When using non-default context, fastest path is return time
    if (VL_LIKELY(m_s.m_time)) return m_s.m_time;
 #if defined(SYSTEMC_VERSION) || (!defined(VL_TIME_CONTEXT) && !defined(VL_NO_LEGACY))
@@ -329,16 +325,22 @@ inline vluint64_t VerilatedContext::time() const VL_MT_SAFE {
 // Return time precision as multiplier of time units
 double vl_time_multiplier(int scale) VL_PURE;
 // Return power of 10. e.g. returns 100 if n==2
-vluint64_t vl_time_pow10(int n) VL_PURE;
+uint64_t vl_time_pow10(int n) VL_PURE;

 #ifdef VL_DEBUG
-/// Evaluate statement if Verilated::debug() enabled
+/// Evaluate statement if VL_DEBUG defined
+# define VL_DEBUG_IFDEF(stmt) \
+    do { \
+        stmt \
+    } while (false)
+/// Evaluate statement if VL_DEBUG defined and Verilated::debug() enabled
 # define VL_DEBUG_IF(stmt) \
    do { \
        if (VL_UNLIKELY(Verilated::debug())) {stmt} \
    } while (false)
 #else
 // We intentionally do not compile the stmt to improve compile speed
+# define VL_DEBUG_IFDEF(stmt) do {} while (false)
 # define VL_DEBUG_IF(stmt) do {} while (false)
 #endif

@@ -869,46 +871,46 @@ static inline int _vl_cmp_w(int words, WDataInP const lwp, WDataInP const rwp) V
 static inline IData VL_GTS_III(int lbits, IData lhs, IData rhs) VL_PURE {
    // For lbits==32, this becomes just a single instruction, otherwise ~5.
    // GCC 3.3.4 sign extension bugs on AMD64 architecture force us to use quad logic
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
    return lhs_signed > rhs_signed;
 }
 static inline IData VL_GTS_IQQ(int lbits, QData lhs, QData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
    return lhs_signed > rhs_signed;
 }

 static inline IData VL_GTES_III(int lbits, IData lhs, IData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
    return lhs_signed >= rhs_signed;
 }
 static inline IData VL_GTES_IQQ(int lbits, QData lhs, QData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
    return lhs_signed >= rhs_signed;
 }

 static inline IData VL_LTS_III(int lbits, IData lhs, IData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
    return lhs_signed < rhs_signed;
 }
 static inline IData VL_LTS_IQQ(int lbits, QData lhs, QData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
    return lhs_signed < rhs_signed;
 }

 static inline IData VL_LTES_III(int lbits, IData lhs, IData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);  // Q for gcc
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);  // Q for gcc
    return lhs_signed <= rhs_signed;
 }
 static inline IData VL_LTES_IQQ(int lbits, QData lhs, QData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
    return lhs_signed <= rhs_signed;
 }

@@ -1002,13 +1004,13 @@ static inline WDataOutP VL_MUL_W(int words, WDataOutP owp, WDataInP const lwp,
 }

 static inline IData VL_MULS_III(int lbits, IData lhs, IData rhs) VL_PURE {
-    const vlsint32_t lhs_signed = VL_EXTENDS_II(32, lbits, lhs);
-    const vlsint32_t rhs_signed = VL_EXTENDS_II(32, lbits, rhs);
+    const int32_t lhs_signed = VL_EXTENDS_II(32, lbits, lhs);
+    const int32_t rhs_signed = VL_EXTENDS_II(32, lbits, rhs);
    return lhs_signed * rhs_signed;
 }
 static inline QData VL_MULS_QQQ(int lbits, QData lhs, QData rhs) VL_PURE {
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(64, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(64, lbits, rhs);
    return lhs_signed * rhs_signed;
 }

@@ -1054,30 +1056,30 @@ static inline IData VL_DIVS_III(int lbits, IData lhs, IData rhs) VL_PURE {
    if (VL_UNLIKELY(rhs == 0)) return 0;
    // -MAX / -1 cannot be represented in twos complement, and will cause SIGFPE
    if (VL_UNLIKELY(lhs == 0x80000000 && rhs == 0xffffffff)) return 0;
-    const vlsint32_t lhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, lhs);
-    const vlsint32_t rhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, rhs);
+    const int32_t lhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, lhs);
+    const int32_t rhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, rhs);
    return lhs_signed / rhs_signed;
 }
 static inline QData VL_DIVS_QQQ(int lbits, QData lhs, QData rhs) VL_PURE {
    if (VL_UNLIKELY(rhs == 0)) return 0;
    // -MAX / -1 cannot be represented in twos complement, and will cause SIGFPE
    if (VL_UNLIKELY(lhs == 0x8000000000000000ULL && rhs == 0xffffffffffffffffULL)) return 0;
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, rhs);
    return lhs_signed / rhs_signed;
 }
 static inline IData VL_MODDIVS_III(int lbits, IData lhs, IData rhs) VL_PURE {
    if (VL_UNLIKELY(rhs == 0)) return 0;
    if (VL_UNLIKELY(lhs == 0x80000000 && rhs == 0xffffffff)) return 0;
-    const vlsint32_t lhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, lhs);
-    const vlsint32_t rhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, rhs);
+    const int32_t lhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, lhs);
+    const int32_t rhs_signed = VL_EXTENDS_II(VL_IDATASIZE, lbits, rhs);
    return lhs_signed % rhs_signed;
 }
 static inline QData VL_MODDIVS_QQQ(int lbits, QData lhs, QData rhs) VL_PURE {
    if (VL_UNLIKELY(rhs == 0)) return 0;
    if (VL_UNLIKELY(lhs == 0x8000000000000000ULL && rhs == 0xffffffffffffffffULL)) return 0;
-    const vlsint64_t lhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, lhs);
-    const vlsint64_t rhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, rhs);
+    const int64_t lhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, lhs);
+    const int64_t rhs_signed = VL_EXTENDS_QQ(VL_QUADSIZE, lbits, rhs);
    return lhs_signed % rhs_signed;
 }

@@ -1408,8 +1410,8 @@ static inline IData VL_STREAML_FAST_III(int lbits, IData ld, IData rd_log2) VL_P
    //   ret = 10324---
    IData ret = ld;
    if (rd_log2) {
-        const vluint32_t lbitsFloor = lbits & ~VL_MASK_I(rd_log2);  // max multiple of rd <= lbits
-        const vluint32_t lbitsRem = lbits - lbitsFloor;  // number of bits in most-sig slice (MSS)
+        const uint32_t lbitsFloor = lbits & ~VL_MASK_I(rd_log2);  // max multiple of rd <= lbits
+        const uint32_t lbitsRem = lbits - lbitsFloor;  // number of bits in most-sig slice (MSS)
        const IData msbMask = VL_MASK_I(lbitsRem) << lbitsFloor;  // mask to sel only bits in MSS
        ret = (ret & ~msbMask) | ((ret & msbMask) << ((VL_UL(1) << rd_log2) - lbitsRem));
    }
@@ -1428,9 +1430,9 @@ static inline QData VL_STREAML_FAST_QQI(int lbits, QData ld, IData rd_log2) VL_P
    // Pre-shift bits in most-significant slice (see comment in VL_STREAML_FAST_III)
    QData ret = ld;
    if (rd_log2) {
-        const vluint32_t lbitsFloor = lbits & ~VL_MASK_I(rd_log2);
-        const vluint32_t lbitsRem = lbits - lbitsFloor;
-        const QData msbMask = VL_MASK_Q(lbitsRem) << lbitsFloor;
+        const uint32_t lbitsFloor = lbits & ~VL_MASK_I(rd_log2);
+        const uint32_t lbitsRem = lbits - lbitsFloor;
+        const QData msbMask = lbitsFloor == 64 ? 0ULL : VL_MASK_Q(lbitsRem) << lbitsFloor;
        ret = (ret & ~msbMask) | ((ret & msbMask) << ((1ULL << rd_log2) - lbitsRem));
    }
    switch (rd_log2) {
@@ -1781,7 +1783,7 @@ static inline WDataOutP VL_SHIFTRS_WWW(int obits, int lbits, int rbits, WDataOut
                                       WDataInP const lwp, WDataInP const rwp) VL_MT_SAFE {
    EData overshift = 0;  // Huge shift 1>>32 or more
    for (int i = 1; i < VL_WORDS_I(rbits); ++i) overshift |= rwp[i];
-    if (VL_UNLIKELY(overshift || rwp[0] >= obits)) {
+    if (VL_UNLIKELY(overshift || rwp[0] >= static_cast<IData>(obits))) {
        const int lmsw = VL_WORDS_I(obits) - 1;
        const EData sign = VL_SIGNONES_E(lbits, lwp[lmsw]);
        for (int j = 0; j <= lmsw; ++j) owp[j] = sign;
@@ -1800,7 +1802,7 @@ static inline IData VL_SHIFTRS_IIW(int obits, int lbits, int rbits, IData lhs,
                                   WDataInP const rwp) VL_MT_SAFE {
    EData overshift = 0;  // Huge shift 1>>32 or more
    for (int i = 1; i < VL_WORDS_I(rbits); ++i) overshift |= rwp[i];
-    if (VL_UNLIKELY(overshift || rwp[0] >= obits)) {
+    if (VL_UNLIKELY(overshift || rwp[0] >= static_cast<IData>(obits))) {
        const IData sign = -(lhs >> (lbits - 1));  // ffff_ffff if negative
        return VL_CLEAN_II(obits, obits, sign);
    }
@@ -1810,7 +1812,7 @@ static inline QData VL_SHIFTRS_QQW(int obits, int lbits, int rbits, QData lhs,
                                   WDataInP const rwp) VL_MT_SAFE {
    EData overshift = 0;  // Huge shift 1>>32 or more
    for (int i = 1; i < VL_WORDS_I(rbits); ++i) overshift |= rwp[i];
-    if (VL_UNLIKELY(overshift || rwp[0] >= obits)) {
+    if (VL_UNLIKELY(overshift || rwp[0] >= static_cast<IData>(obits))) {
        const QData sign = -(lhs >> (lbits - 1));  // ffff_ffff if negative
        return VL_CLEAN_QQ(obits, obits, sign);
    }
@@ -1927,8 +1929,8 @@ static inline QData VL_RTOIROUND_Q_D(double lhs) VL_PURE {
    if (lhs == 0.0) return 0;
    const QData q = VL_CVT_Q_D(lhs);
    const int lsb = static_cast<int>((q >> 52ULL) & VL_MASK_Q(11)) - 1023 - 52;
-    const vluint64_t mantissa = (q & VL_MASK_Q(52)) | (1ULL << 52);
-    vluint64_t out = 0;
+    const uint64_t mantissa = (q & VL_MASK_Q(52)) | (1ULL << 52);
+    uint64_t out = 0;
    if (lsb < 0) {
        out = mantissa >> -lsb;
    } else if (lsb < 64) {
@@ -1948,7 +1950,7 @@ static inline WDataOutP VL_RTOIROUND_W_D(int obits, WDataOutP owp, double lhs) V
    if (lhs == 0.0) return owp;
    const QData q = VL_CVT_Q_D(lhs);
    const int lsb = static_cast<int>((q >> 52ULL) & VL_MASK_Q(11)) - 1023 - 52;
-    const vluint64_t mantissa = (q & VL_MASK_Q(52)) | (1ULL << 52);
+    const uint64_t mantissa = (q & VL_MASK_Q(52)) | (1ULL << 52);
    if (lsb < 0) {
        VL_SET_WQ(owp, mantissa >> -lsb);
    } else if (lsb < obits) {
@@ -1962,42 +1964,33 @@ static inline WDataOutP VL_RTOIROUND_W_D(int obits, WDataOutP owp, double lhs) V
 // Range assignments

 // EMIT_RULE: VL_ASSIGNRANGE:  rclean=dirty;
-static inline void VL_ASSIGNSEL_IIII(int rbits, int obits, int lsb, CData& lhsr,
-                                     IData rhs) VL_PURE {
+static inline void VL_ASSIGNSEL_II(int rbits, int obits, int lsb, CData& lhsr, IData rhs) VL_PURE {
    _vl_insert_II(lhsr, rhs, lsb + obits - 1, lsb, rbits);
 }
-static inline void VL_ASSIGNSEL_IIII(int rbits, int obits, int lsb, SData& lhsr,
-                                     IData rhs) VL_PURE {
+static inline void VL_ASSIGNSEL_II(int rbits, int obits, int lsb, SData& lhsr, IData rhs) VL_PURE {
    _vl_insert_II(lhsr, rhs, lsb + obits - 1, lsb, rbits);
 }
-static inline void VL_ASSIGNSEL_IIII(int rbits, int obits, int lsb, IData& lhsr,
-                                     IData rhs) VL_PURE {
+static inline void VL_ASSIGNSEL_II(int rbits, int obits, int lsb, IData& lhsr, IData rhs) VL_PURE {
    _vl_insert_II(lhsr, rhs, lsb + obits - 1, lsb, rbits);
 }
-static inline void VL_ASSIGNSEL_QIII(int rbits, int obits, int lsb, QData& lhsr,
-                                     IData rhs) VL_PURE {
+static inline void VL_ASSIGNSEL_QI(int rbits, int obits, int lsb, QData& lhsr, IData rhs) VL_PURE {
    _vl_insert_QQ(lhsr, rhs, lsb + obits - 1, lsb, rbits);
 }
-static inline void VL_ASSIGNSEL_QQII(int rbits, int obits, int lsb, QData& lhsr,
-                                     QData rhs) VL_PURE {
-    _vl_insert_QQ(lhsr, rhs, lsb + obits - 1, lsb, rbits);
-}
-static inline void VL_ASSIGNSEL_QIIQ(int rbits, int obits, int lsb, QData& lhsr,
-                                     QData rhs) VL_PURE {
+static inline void VL_ASSIGNSEL_QQ(int rbits, int obits, int lsb, QData& lhsr, QData rhs) VL_PURE {
    _vl_insert_QQ(lhsr, rhs, lsb + obits - 1, lsb, rbits);
 }
 // static inline void VL_ASSIGNSEL_IIIW(int obits, int lsb, IData& lhsr, WDataInP const rwp)
 // VL_MT_SAFE { Illegal, as lhs width >= rhs width
-static inline void VL_ASSIGNSEL_WIII(int rbits, int obits, int lsb, WDataOutP owp,
-                                     IData rhs) VL_MT_SAFE {
+static inline void VL_ASSIGNSEL_WI(int rbits, int obits, int lsb, WDataOutP owp,
+                                   IData rhs) VL_MT_SAFE {
    _vl_insert_WI(owp, rhs, lsb + obits - 1, lsb, rbits);
 }
-static inline void VL_ASSIGNSEL_WIIQ(int rbits, int obits, int lsb, WDataOutP owp,
-                                     QData rhs) VL_MT_SAFE {
+static inline void VL_ASSIGNSEL_WQ(int rbits, int obits, int lsb, WDataOutP owp,
+                                   QData rhs) VL_MT_SAFE {
    _vl_insert_WQ(owp, rhs, lsb + obits - 1, lsb, rbits);
 }
-static inline void VL_ASSIGNSEL_WIIW(int rbits, int obits, int lsb, WDataOutP owp,
-                                     WDataInP const rwp) VL_MT_SAFE {
+static inline void VL_ASSIGNSEL_WW(int rbits, int obits, int lsb, WDataOutP owp,
+                                   WDataInP const rwp) VL_MT_SAFE {
    _vl_insert_WW(owp, rwp, lsb + obits - 1, lsb, rbits);
 }

@@ -2036,40 +2029,40 @@ static inline WDataOutP VL_CONST_W_2X(int obits, WDataOutP o, EData d1, EData d0
 static inline WDataOutP VL_CONST_W_3X(int obits, WDataOutP o, EData d2, EData d1,
                                      EData d0) VL_MT_SAFE {
    o[0] = d0;  o[1] = d1;  o[2] = d2;
-    VL_C_END_(obits,3);
+    VL_C_END_(obits, 3);
 }
 static inline WDataOutP VL_CONST_W_4X(int obits, WDataOutP o,
                                      EData d3, EData d2, EData d1, EData d0) VL_MT_SAFE {
    o[0] = d0;  o[1] = d1;  o[2] = d2;  o[3] = d3;
-    VL_C_END_(obits,4);
+    VL_C_END_(obits, 4);
 }
 static inline WDataOutP VL_CONST_W_5X(int obits, WDataOutP o,
                                      EData d4,
                                      EData d3, EData d2, EData d1, EData d0) VL_MT_SAFE {
    o[0] = d0;  o[1] = d1;  o[2] = d2;  o[3] = d3;
    o[4] = d4;
-    VL_C_END_(obits,5);
+    VL_C_END_(obits, 5);
 }
 static inline WDataOutP VL_CONST_W_6X(int obits, WDataOutP o,
                                      EData d5, EData d4,
                                      EData d3, EData d2, EData d1, EData d0) VL_MT_SAFE {
    o[0] = d0;  o[1] = d1;  o[2] = d2;  o[3] = d3;
    o[4] = d4;  o[5] = d5;
-    VL_C_END_(obits,6);
+    VL_C_END_(obits, 6);
 }
 static inline WDataOutP VL_CONST_W_7X(int obits, WDataOutP o,
                                      EData d6, EData d5, EData d4,
                                      EData d3, EData d2, EData d1, EData d0) VL_MT_SAFE {
    o[0] = d0;  o[1] = d1;  o[2] = d2;  o[3] = d3;
    o[4] = d4;  o[5] = d5;  o[6] = d6;
-    VL_C_END_(obits,7);
+    VL_C_END_(obits, 7);
 }
 static inline WDataOutP VL_CONST_W_8X(int obits, WDataOutP o,
                                      EData d7, EData d6, EData d5, EData d4,
                                      EData d3, EData d2, EData d1, EData d0) VL_MT_SAFE {
    o[0] = d0;  o[1] = d1;  o[2] = d2;  o[3] = d3;
    o[4] = d4;  o[5] = d5;  o[6] = d6;  o[7] = d7;
-    VL_C_END_(obits,8);
+    VL_C_END_(obits, 8);
 }
 //
 static inline WDataOutP VL_CONSTHI_W_1X(int obits, int lsb, WDataOutP obase,
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2010-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2010-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2009-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2009-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -32,10 +32,15 @@
 #include "verilated.h"
 #include "verilated_syms.h"

+#include <algorithm>
 #include <deque>
-#include <set>
-#include <vector>
+#include <limits>
+#include <map>
 #include <numeric>
+#include <set>
+#include <string>
+#include <utility>
+#include <vector>
 #ifdef VL_THREADED
 # include <functional>
 # include <queue>
@@ -60,7 +65,7 @@ public:

 private:
    // MEMBERS
-    vluint32_t m_mtaskId;  // MTask that did enqueue
+    uint32_t m_mtaskId;  // MTask that did enqueue
    std::function<void()> m_cb;  // Lambda to execute when message received
 public:
    // CONSTRUCTORS
@@ -73,7 +78,7 @@ public:
    VerilatedMsg& operator=(const VerilatedMsg&) = default;
    VerilatedMsg& operator=(VerilatedMsg&&) = default;
    // METHODS
-    vluint32_t mtaskId() const { return m_mtaskId; }
+    uint32_t mtaskId() const { return m_mtaskId; }
    // Execute the lambda function
    void run() const { m_cb(); }
 };
@@ -84,9 +89,9 @@ public:
 class VerilatedEvalMsgQueue final {
    using VerilatedThreadQueue = std::multiset<VerilatedMsg, VerilatedMsg::Cmp>;

-    std::atomic<vluint64_t> m_depth;  // Current depth of queue (see comments below)
+    std::atomic<uint64_t> m_depth;  // Current depth of queue (see comments below)

-    VerilatedMutex m_mutex;  // Mutex protecting queue
+    mutable VerilatedMutex m_mutex;  // Mutex protecting queue
    VerilatedThreadQueue m_queue VL_GUARDED_BY(m_mutex);  // Message queue
 public:
    // CONSTRUCTORS
@@ -240,8 +245,8 @@ public:  // But only for verilated*.cpp
    // METHODS - extending into VerilatedContext, call via impp()->

    // Random seed handling
-    vluint64_t randSeedDefault64() const VL_MT_SAFE;
-    static vluint32_t randSeedEpoch() VL_MT_SAFE { return s().s_randSeedEpoch; }
+    uint64_t randSeedDefault64() const VL_MT_SAFE;
+    static uint32_t randSeedEpoch() VL_MT_SAFE { return s().s_randSeedEpoch; }

    // METHODS - timeformat
    int timeFormatUnits() const VL_MT_SAFE {
@@ -312,14 +317,14 @@ public:  // But only for verilated*.cpp
    IData fdSeek(IData fdi, IData offset, IData origin) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
        const VerilatedLockGuard lock{m_fdMutex};
        const VerilatedFpList fdlist = fdToFpList(fdi);
-        if (VL_UNLIKELY(fdlist.size() != 1)) return 0;
+        if (VL_UNLIKELY(fdlist.size() != 1)) return ~0U;  // -1
        return static_cast<IData>(
            std::fseek(*fdlist.begin(), static_cast<long>(offset), static_cast<int>(origin)));
    }
    IData fdTell(IData fdi) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
        const VerilatedLockGuard lock{m_fdMutex};
        const VerilatedFpList fdlist = fdToFpList(fdi);
-        if (VL_UNLIKELY(fdlist.size() != 1)) return 0;
+        if (VL_UNLIKELY(fdlist.size() != 1)) return ~0U;  // -1
        return static_cast<IData>(std::ftell(*fdlist.begin()));
    }
    void fdWrite(IData fdi, const std::string& output) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
@@ -390,7 +395,11 @@ protected:
    // METHODS - protected
    void commandArgsAddGuts(int argc, const char** argv);
    void commandArgVl(const std::string& arg);
-    bool commandArgVlValue(const std::string& arg, const std::string& prefix, std::string& valuer);
+    bool commandArgVlString(const std::string& arg, const std::string& prefix,
+                            std::string& valuer);
+    bool commandArgVlUint64(const std::string& arg, const std::string& prefix, uint64_t& valuer,
+                            uint64_t min = std::numeric_limits<uint64_t>::min(),
+                            uint64_t max = std::numeric_limits<uint64_t>::max());
    void commandArgDump() const VL_MT_SAFE_EXCLUDES(m_argMutex);
 };

@@ -480,7 +489,7 @@ public:
 public:  // But only for verilated.cpp
    // Symbol table destruction cleans up the entries for each scope.
    static void userEraseScope(const VerilatedScope* scopep) VL_MT_SAFE {
-        // Slow ok - called once/scope on destruction, so we simply iterate.
+        // Slow ok - called once/scope on destruction, so we only iterate.
        const VerilatedLockGuard lock{s().m_userMapMutex};
        for (auto it = s().m_userMap.begin(); it != s().m_userMap.end();) {
            if (it->first.first == scopep) {
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -0,0 +1,191 @@
+// -*- mode: C++; c-file-style: "cc-mode" -*-
+//=============================================================================
+//
+// Code available from: https://verilator.org
+//
+// Copyright 2012-2022 by Wilson Snyder. This program is free software; you can
+// redistribute it and/or modify it under the terms of either the GNU
+// Lesser General Public License Version 3 or the Perl Artistic License
+// Version 2.0.
+// SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0
+//
+//=============================================================================
+///
+/// \file
+/// \brief Verilated run-time profiling implementation code
+///
+//=============================================================================
+
+#include "verilatedos.h"
+#include "verilated_profiler.h"
+
+#if VL_THREADED
+#include "verilated_threads.h"
+#endif
+
+#include <fstream>
+#include <string>
+
+//=============================================================================
+// Globals
+
+// Internal note: Globals may multi-construct, see verilated.cpp top.
+
+VL_THREAD_LOCAL VlExecutionProfiler::ExecutionTrace VlExecutionProfiler::t_trace;
+
+constexpr const char* const VlExecutionRecord::s_ascii[];
+
+//=============================================================================
+// VlPgoProfiler implementation
+
+uint16_t VlExecutionRecord::getcpu() {
+#if defined(__linux)
+    return sched_getcpu();  // TODO: this is a system call. Not exactly cheap.
+#elif defined(__APPLE__) && !defined(__arm64__)
+    uint32_t info[4];
+    __cpuid_count(1, 0, info[0], info[1], info[2], info[3]);
+    // info[1] is EBX, bits 24-31 are APIC ID
+    if ((info[3] & (1 << 9)) == 0) {
+        return -1;  // no APIC on chip
+    } else {
+        return (unsigned)info[1] >> 24;
+    }
+#elif defined(_WIN32)
+    return GetCurrentProcessorNumber();
+#else
+    return 0;
+#endif
+}
+
+//=============================================================================
+// VlExecutionProfiler implementation
+
+template <size_t N> static size_t roundUptoMultipleOf(size_t value) {
+    static_assert((N & (N - 1)) == 0, "'N' must be a power of 2");
+    size_t mask = N - 1;
+    return (value + mask) & ~mask;
+}
+
+VlExecutionProfiler::VlExecutionProfiler() {
+    // Setup profiling on main thread
+    setupThread(0);
+}
+
+void VlExecutionProfiler::configure(const VerilatedContext& context) {
+    if (VL_UNLIKELY(m_enabled)) {
+        --m_windowCount;
+        if (VL_UNLIKELY(m_windowCount == context.profExecWindow())) {
+            VL_DEBUG_IF(VL_DBG_MSGF("+ profile start collection\n"););
+            clear();  // Clear the profile after the cache warm-up cycles.
+            m_tickBegin = VL_CPU_TICK();
+        } else if (VL_UNLIKELY(m_windowCount == 0)) {
+            const uint64_t tickEnd = VL_CPU_TICK();
+            VL_DEBUG_IF(VL_DBG_MSGF("+ profile end\n"););
+            const std::string& fileName = context.profExecFilename();
+            dump(fileName.c_str(), tickEnd);
+            m_enabled = false;
+        }
+        return;
+    }
+
+    const uint64_t startReq = context.profExecStart() + 1;  // + 1, so we can start at time 0
+
+    if (VL_UNLIKELY(m_lastStartReq < startReq && VL_TIME_Q() >= context.profExecStart())) {
+        VL_DEBUG_IF(VL_DBG_MSGF("+ profile start warmup\n"););
+        VL_DEBUG_IF(assert(m_windowCount == 0););
+        m_enabled = true;
+        m_windowCount = context.profExecWindow() * 2;
+        m_lastStartReq = startReq;
+    }
+}
+
+void VlExecutionProfiler::setupThread(uint32_t threadId) {
+    // Reserve some space in the thread-local profiling buffer, in order to try to avoid malloc
+    // while profiling.
+    t_trace.reserve(RESERVED_TRACE_CAPACITY);
+    // Register thread-local buffer in list of all buffers
+    {
+        const VerilatedLockGuard lock{m_mutex};
+        bool exists = !m_traceps.emplace(threadId, &t_trace).second;
+        assert(!exists);
+    }
+}
+
+void VlExecutionProfiler::clear() VL_MT_SAFE_EXCLUDES(m_mutex) {
+    const VerilatedLockGuard lock{m_mutex};
+    for (const auto& pair : m_traceps) {
+        ExecutionTrace* const tracep = pair.second;
+        const size_t reserve = roundUptoMultipleOf<RESERVED_TRACE_CAPACITY>(tracep->size());
+        tracep->clear();
+        tracep->reserve(reserve);
+    }
+}
+
+void VlExecutionProfiler::dump(const char* filenamep, uint64_t tickEnd)
+    VL_MT_SAFE_EXCLUDES(m_mutex) {
+    const VerilatedLockGuard lock{m_mutex};
+    VL_DEBUG_IF(VL_DBG_MSGF("+prof+exec writing to '%s'\n", filenamep););
+
+    FILE* const fp = std::fopen(filenamep, "w");
+    if (VL_UNLIKELY(!fp)) { VL_FATAL_MT(filenamep, 0, "", "+prof+exec+file file not writable"); }
+
+    // TODO Perhaps merge with verilated_coverage output format, so can
+    // have a common merging and reporting tool, etc.
+    fprintf(fp, "VLPROFVERSION 2.0 # Verilator execution profile version 2.0\n");
+    fprintf(fp, "VLPROF arg +verilator+prof+exec+start+%" PRIu64 "\n",
+            Verilated::threadContextp()->profExecStart());
+    fprintf(fp, "VLPROF arg +verilator+prof+exec+window+%u\n",
+            Verilated::threadContextp()->profExecWindow());
+    const unsigned threads = static_cast<unsigned>(m_traceps.size());
+    fprintf(fp, "VLPROF stat threads %u\n", threads);
+#ifdef VL_THREADED
+    fprintf(fp, "VLPROF stat yields %" PRIu64 "\n", VlMTaskVertex::yields());
+#endif
+
+    // Copy /proc/cpuinfo into this output so verilator_gantt can be run on
+    // a different machine
+    {
+        const std::unique_ptr<std::ifstream> ifp{new std::ifstream("/proc/cpuinfo")};
+        if (!ifp->fail()) {
+            std::string line;
+            while (std::getline(*ifp, line)) { fprintf(fp, "VLPROFPROC %s\n", line.c_str()); }
+        }
+    }
+
+    for (const auto& pair : m_traceps) {
+        const uint32_t threadId = pair.first;
+        ExecutionTrace* const tracep = pair.second;
+        fprintf(fp, "VLPROFTHREAD %" PRIu32 "\n", threadId);
+
+        for (const VlExecutionRecord& er : *tracep) {
+            const char* const name = VlExecutionRecord::s_ascii[static_cast<uint8_t>(er.m_type)];
+            const uint64_t time = er.m_tick - m_tickBegin;
+            fprintf(fp, "VLPROFEXEC %s %" PRIu64, name, time);
+
+            switch (er.m_type) {
+            case VlExecutionRecord::Type::EVAL_BEGIN:
+            case VlExecutionRecord::Type::EVAL_END:
+            case VlExecutionRecord::Type::EVAL_LOOP_BEGIN:
+            case VlExecutionRecord::Type::EVAL_LOOP_END:
+                // No payload
+                fprintf(fp, "\n");
+                break;
+            case VlExecutionRecord::Type::MTASK_BEGIN: {
+                const auto& payload = er.m_payload.mtaskBegin;
+                fprintf(fp, " id %u predictStart %u cpu %u\n", payload.m_id,
+                        payload.m_predictStart, payload.m_cpu);
+                break;
+            }
+            case VlExecutionRecord::Type::MTASK_END: {
+                const auto& payload = er.m_payload.mtaskEnd;
+                fprintf(fp, " id %u predictCost %u\n", payload.m_id, payload.m_predictCost);
+                break;
+            }
+            default: abort();  // LCOV_EXCL_LINE
+            }
+        }
+    }
+    fprintf(fp, "VLPROF stat ticks %" PRIu64 "\n", tickEnd - m_tickBegin);
+
+    std::fclose(fp);
+}
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2012-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2012-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -12,7 +12,7 @@
 //=============================================================================
 ///
 /// \file
-/// \brief Verilated general profiling header
+/// \brief Verilated run-time profiling header
 ///
 /// This file is not part of the Verilated public-facing API.
 /// It is only for internal use by Verilated library routines.
@@ -23,55 +23,204 @@
 #define VERILATOR_VERILATED_PROFILER_H_

 #include "verilatedos.h"
-#include "verilated.h"  // for VerilatedMutex and clang annotations

-// Profile record, private class used only by this header
-class VerilatedProfilerRec final {
-    const std::string m_name;  // Hashed name of mtask/etc
-    const size_t m_counterNumber = 0;  // Which counter has data
+#ifndef VL_PROFILER
+#error "verilated_profiler.h/cpp expects VL_PROFILER (from --prof-{exec, pgo}"
+#endif
+
+#include "verilated.h"
+
+#include <array>
+#include <atomic>
+#include <cassert>
+#include <string>
+#include <type_traits>
+#include <vector>
+
+class VlExecutionProfiler;
+
+//=============================================================================
+// Macros to simplify generated code
+
+#define VL_EXEC_TRACE_ADD_RECORD(vlSymsp) \
+    if (VL_UNLIKELY((vlSymsp)->__Vm_executionProfiler.enabled())) \
+    (vlSymsp)->__Vm_executionProfiler.addRecord()
+
+//=============================================================================
+// Return high-precision counter for profiling, or 0x0 if not available
+VL_ATTR_ALWINLINE
+inline QData VL_CPU_TICK() {
+    uint64_t val;
+    VL_GET_CPU_TICK(val);
+    return val;
+}
+
+//=============================================================================
+// Private class used by VlExecutionProfiler
+
+#define _VL_FOREACH_APPLY(macro, arg) macro(arg, #arg)
+
+// clang-format off
+#define FOREACH_VlExecutionRecord_TYPE(macro) \
+    _VL_FOREACH_APPLY(macro, EVAL_BEGIN) \
+    _VL_FOREACH_APPLY(macro, EVAL_END) \
+    _VL_FOREACH_APPLY(macro, EVAL_LOOP_BEGIN) \
+    _VL_FOREACH_APPLY(macro, EVAL_LOOP_END) \
+    _VL_FOREACH_APPLY(macro, MTASK_BEGIN) \
+    _VL_FOREACH_APPLY(macro, MTASK_END)
+// clang-format on
+
+class VlExecutionRecord final {
+    friend class VlExecutionProfiler;
+
+    // TYPES
+    enum class Type : uint8_t {
+#define VL_FOREACH_MACRO(id, name) id,
+        FOREACH_VlExecutionRecord_TYPE(VL_FOREACH_MACRO)
+#undef VL_FOREACH_MACRO
+    };
+
+    static constexpr const char* const s_ascii[] = {
+#define VL_FOREACH_MACRO(id, name) name,
+        FOREACH_VlExecutionRecord_TYPE(VL_FOREACH_MACRO)
+#undef VL_FOREACH_MACRO
+    };
+
+    union Payload {
+        struct {
+            uint32_t m_id;  // MTask id
+            uint32_t m_predictStart;  // Time scheduler predicted would start
+            uint32_t m_cpu;  // Executing CPU id
+        } mtaskBegin;
+        struct {
+            uint32_t m_id;  // MTask id
+            uint32_t m_predictCost;  // How long scheduler predicted would take
+        } mtaskEnd;
+    };
+
+    // STATE
+    // Layout below allows efficient packing.
+    const uint64_t m_tick = VL_CPU_TICK();  // Tick at construction
+    Payload m_payload;  // The record payload
+    Type m_type;  // The record type
+    static_assert(alignof(uint64_t) >= alignof(Payload), "Padding not allowed");
+    static_assert(alignof(Payload) >= alignof(Type), "Padding not allowed");
+
+    static uint16_t getcpu();  // Return currently executing CPU id
+
 public:
+    // CONSTRUCTOR
+    VlExecutionRecord() = default;
+
    // METHODS
-    VerilatedProfilerRec(size_t counterNumber, const std::string& name)
-        : m_name{name}
-        , m_counterNumber{counterNumber} {}
-    VerilatedProfilerRec() = default;
-    size_t counterNumber() const { return m_counterNumber; }
-    std::string name() const { return m_name; }
+    void evalBegin() { m_type = Type::EVAL_BEGIN; }
+    void evalEnd() { m_type = Type::EVAL_END; }
+    void evalLoopBegin() { m_type = Type::EVAL_LOOP_BEGIN; }
+    void evalLoopEnd() { m_type = Type::EVAL_LOOP_END; }
+    void mtaskBegin(uint32_t id, uint32_t predictStart) {
+        m_payload.mtaskBegin.m_id = id;
+        m_payload.mtaskBegin.m_predictStart = predictStart;
+        m_payload.mtaskBegin.m_cpu = getcpu();
+        m_type = Type::MTASK_BEGIN;
+    }
+    void mtaskEnd(uint32_t id, uint32_t predictCost) {
+        m_payload.mtaskEnd.m_id = id;
+        m_payload.mtaskEnd.m_predictCost = predictCost;
+        m_type = Type::MTASK_END;
+    }
 };

-// Create some number of bucketed profilers
-template <std::size_t T_Entries> class VerilatedProfiler final {
-    // Counters are stored packed, all together, versus in VerilatedProfilerRec to
-    // reduce cache effects
-    std::array<vluint64_t, T_Entries> m_counters{};  // Time spent on this record
-    std::deque<VerilatedProfilerRec> m_records;  // Record information
+static_assert(std::is_trivially_destructible<VlExecutionRecord>::value,
+              "VlExecutionRecord should be trivially destructible for fast buffer clearing");
+
+//=============================================================================
+// VlExecutionProfiler is for collecting profiling data about model execution
+
+class VlExecutionProfiler final {
+    // CONSTANTS
+
+    // In order to try to avoid dynamic memory allocations during the actual profiling phase,
+    // trace buffers are pre-allocated to be able to hold [a multiple] of this many records.
+    static constexpr size_t RESERVED_TRACE_CAPACITY = 4096;
+
+    // TYPES
+
+    // Execution traces are recorded into thread local vectors. We can append records of profiling
+    // events to this vector with very low overhead, and then dump them out later. This prevents
+    // the overhead of printf/malloc/IO from corrupting the profiling data. It's super cheap to
+    // append a VlProfileRec struct on the end of a pre-allocated vector; this is the only cost we
+    // pay in real-time during a profiling cycle. Internal note: Globals may multi-construct, see
+    // verilated.cpp top.
+    using ExecutionTrace = std::vector<VlExecutionRecord>;
+
+    // STATE
+    static VL_THREAD_LOCAL ExecutionTrace t_trace;  // thread-local trace buffers
+    mutable VerilatedMutex m_mutex;
+    // Map from thread id to &t_trace of given thread
+    std::map<uint32_t, ExecutionTrace*> m_traceps VL_GUARDED_BY(m_mutex);
+
+    bool m_enabled = false;  // Is profiling currently enabled
+
+    uint64_t m_tickBegin = 0;  // Sample time (rdtsc() on x86) at beginning of collection
+    uint64_t m_lastStartReq = 0;  // Last requested profiling start (in simulation time)
+    uint32_t m_windowCount = 0;  // Track our position in the cache warmup and profile window
+
+public:
+    // CONSTRUCTOR
+    VlExecutionProfiler();
+
+    // METHODS
+
+    // Is profiling enabled
+    inline bool enabled() const { return m_enabled; }
+    // Append a trace record to the trace buffer of the current thread
+    inline VlExecutionRecord& addRecord() {
+        t_trace.emplace_back();
+        return t_trace.back();
+    }
+    // Configure profiler (called in beginning of 'eval')
+    void configure(const VerilatedContext&);
+    // Setup profiling on a particular thread;
+    void setupThread(uint32_t threadId);
+    // Clear all profiling data
+    void clear() VL_MT_SAFE_EXCLUDES(m_mutex);
+    // Write profiling data into file
+    void dump(const char* filenamep, uint64_t tickEnd) VL_MT_SAFE_EXCLUDES(m_mutex);
+};
+
+//=============================================================================
+// VlPgoProfiler is for collecting profiling data for PGO
+
+template <std::size_t T_Entries> class VlPgoProfiler final {
+    // TYPES
+    struct Record final {
+        const std::string m_name;  // Hashed name of mtask/etc
+        const size_t m_counterNumber = 0;  // Which counter has data
+    };
+
+    // Counters are stored packed, all together to reduce cache effects
+    std::array<uint64_t, T_Entries> m_counters;  // Time spent on this record
+    std::vector<Record> m_records;  // Record information

 public:
    // METHODS
-    VerilatedProfiler() = default;
-    ~VerilatedProfiler() = default;
+    VlPgoProfiler() = default;
+    ~VlPgoProfiler() = default;
    void write(const char* modelp, const std::string& filename) VL_MT_SAFE;
    void addCounter(size_t counter, const std::string& name) {
        VL_DEBUG_IF(assert(counter < T_Entries););
-        m_records.emplace_back(VerilatedProfilerRec{counter, name});
+        m_records.emplace_back(Record{name, counter});
    }
    void startCounter(size_t counter) {
-        vluint64_t val;
-        VL_RDTSC(val);
-        // -= so when we add end time in stopCounter, we already subtracted
-        // out, without needing to hold another temporary
-        m_counters[counter] -= val;
-    }
-    void stopCounter(size_t counter) {
-        vluint64_t val;
-        VL_RDTSC(val);
-        m_counters[counter] += val;
+        // -= so when we add end time in stopCounter, the net effect is adding the difference,
+        // without needing to hold onto a temporary
+        m_counters[counter] -= VL_CPU_TICK();
    }
+    void stopCounter(size_t counter) { m_counters[counter] += VL_CPU_TICK(); }
 };

 template <std::size_t T_Entries>
-void VerilatedProfiler<T_Entries>::write(const char* modelp,
-                                         const std::string& filename) VL_MT_SAFE {
+void VlPgoProfiler<T_Entries>::write(const char* modelp, const std::string& filename) VL_MT_SAFE {
    static VerilatedMutex s_mutex;
    const VerilatedLockGuard lock{s_mutex};

@@ -79,20 +228,15 @@ void VerilatedProfiler<T_Entries>::write(const char* modelp,
    // So when we have multiple models in an executable, possibly even
    // running on different threads, each will have a different symtab so
    // each will collect is own data correctly.  However when each is
-    // destroid we need to get all the data, not keep overwriting and only
+    // destroyed we need to get all the data, not keep overwriting and only
    // get the last model's data.
    static bool s_firstCall = true;

    VL_DEBUG_IF(VL_DBG_MSGF("+prof+vlt+file writing to '%s'\n", filename.c_str()););

-    FILE* fp = nullptr;
-    if (!s_firstCall) fp = std::fopen(filename.c_str(), "a");
-    if (VL_UNLIKELY(!fp))
-        fp = std::fopen(filename.c_str(), "w");  // firstCall, or doesn't exist yet
+    FILE* const fp = std::fopen(filename.c_str(), s_firstCall ? "w" : "a");
    if (VL_UNLIKELY(!fp)) {
        VL_FATAL_MT(filename.c_str(), 0, "", "+prof+vlt+file file not writable");
-        // cppcheck-suppress resourceLeak   // bug, doesn't realize fp is nullptr
-        return;  // LCOV_EXCL_LINE
    }
    s_firstCall = false;

@@ -101,10 +245,9 @@ void VerilatedProfiler<T_Entries>::write(const char* modelp,
    fprintf(fp, "// Verilated model profile-guided optimization data dump file\n");
    fprintf(fp, "`verilator_config\n");

-    for (const auto& it : m_records) {
-        const std::string& name = it.name();
-        fprintf(fp, "profile_data -model \"%s\" -mtask \"%s\" -cost 64'd%" VL_PRI64 "u\n", modelp,
-                name.c_str(), m_counters[it.counterNumber()]);
+    for (const Record& rec : m_records) {
+        fprintf(fp, "profile_data -model \"%s\" -mtask \"%s\" -cost 64'd%" PRIu64 "\n", modelp,
+                rec.m_name.c_str(), m_counters[rec.m_counterNumber]);
    }

    std::fclose(fp);
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -63,8 +63,8 @@ static const char* const VLTSAVE_TRAILER_STR = "vltsaved";
 bool VerilatedDeserialize::readDiffers(const void* __restrict datap,
                                       size_t size) VL_MT_UNSAFE_ONE {
    bufferCheck();
-    const vluint8_t* __restrict dp = static_cast<const vluint8_t* __restrict>(datap);
-    vluint8_t miss = 0;
+    const uint8_t* __restrict dp = static_cast<const uint8_t* __restrict>(datap);
+    uint8_t miss = 0;
    while (size--) miss |= (*dp++ ^ *m_cp++);
    return (miss != 0);
 }
@@ -192,7 +192,7 @@ void VerilatedRestore::close() VL_MT_UNSAFE_ONE {
 void VerilatedSave::flush() VL_MT_UNSAFE_ONE {
    m_assertOne.check();
    if (VL_UNLIKELY(!isOpen())) return;
-    const vluint8_t* wp = m_bufp;
+    const uint8_t* wp = m_bufp;
    while (true) {
        const ssize_t remaining = (m_cp - wp);
        if (remaining == 0) break;
@@ -219,8 +219,8 @@ void VerilatedRestore::fill() VL_MT_UNSAFE_ONE {
    m_assertOne.check();
    if (VL_UNLIKELY(!isOpen())) return;
    // Move remaining characters down to start of buffer.  (No memcpy, overlaps allowed)
-    vluint8_t* rp = m_bufp;
-    for (vluint8_t* sp = m_cp; sp < m_endp; *rp++ = *sp++) {}  // Overlaps
+    uint8_t* rp = m_bufp;
+    for (uint8_t* sp = m_cp; sp < m_endp; *rp++ = *sp++) {}  // Overlaps
    m_endp = m_bufp + (m_endp - m_cp);
    m_cp = m_bufp;  // Reset buffer
    // Read into buffer starting at m_endp
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2000-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2000-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -40,8 +40,8 @@ class VerilatedSerialize VL_NOT_FINAL {
 protected:
    // MEMBERS
    // For speed, keep m_cp as the first member of this structure
-    vluint8_t* m_cp;  // Current pointer into m_bufp buffer
-    vluint8_t* m_bufp;  // Output buffer
+    uint8_t* m_cp;  // Current pointer into m_bufp buffer
+    uint8_t* m_bufp;  // Output buffer
    bool m_isOpen = false;  // True indicates open file/stream
    std::string m_filename;  // Filename, for error messages
    VerilatedAssertOneThread m_assertOne;  // Assert only called from single thread
@@ -58,7 +58,7 @@ protected:
 public:
    /// Construct
    VerilatedSerialize() {
-        m_bufp = new vluint8_t[bufferSize()];
+        m_bufp = new uint8_t[bufferSize()];
        m_cp = m_bufp;
    }
    /// Flish, close, and destruct
@@ -77,12 +77,12 @@ public:
    virtual void flush() VL_MT_UNSAFE_ONE {}
    /// Write data to stream
    VerilatedSerialize& write(const void* __restrict datap, size_t size) VL_MT_UNSAFE_ONE {
-        const vluint8_t* __restrict dp = (const vluint8_t* __restrict)datap;
+        const uint8_t* __restrict dp = (const uint8_t* __restrict)datap;
        while (size) {
            bufferCheck();
            size_t blk = size;
            if (blk > bufferInsertSize()) blk = bufferInsertSize();
-            const vluint8_t* __restrict maxp = dp + blk;
+            const uint8_t* __restrict maxp = dp + blk;
            for (; dp < maxp; *m_cp++ = *dp++) {}
            size -= blk;
        }
@@ -111,9 +111,9 @@ class VerilatedDeserialize VL_NOT_FINAL {
 protected:
    // MEMBERS
    // For speed, keep m_cp as the first member of this structure
-    vluint8_t* m_cp;  // Current pointer into m_bufp buffer
-    vluint8_t* m_bufp;  // Output buffer
-    vluint8_t* m_endp = nullptr;  // Last valid byte in m_bufp buffer
+    uint8_t* m_cp;  // Current pointer into m_bufp buffer
+    uint8_t* m_bufp;  // Output buffer
+    uint8_t* m_endp = nullptr;  // Last valid byte in m_bufp buffer
    bool m_isOpen = false;  // True indicates open file/stream
    std::string m_filename;  // Filename, for error messages
    VerilatedAssertOneThread m_assertOne;  // Assert only called from single thread
@@ -131,7 +131,7 @@ protected:
 public:
    /// Construct
    VerilatedDeserialize() {
-        m_bufp = new vluint8_t[bufferSize()];
+        m_bufp = new uint8_t[bufferSize()];
        m_cp = m_bufp;
    }
    /// Destruct
@@ -150,12 +150,12 @@ public:
    virtual void flush() VL_MT_UNSAFE_ONE {}
    /// Read data from stream
    VerilatedDeserialize& read(void* __restrict datap, size_t size) VL_MT_UNSAFE_ONE {
-        vluint8_t* __restrict dp = static_cast<vluint8_t* __restrict>(datap);
+        uint8_t* __restrict dp = static_cast<uint8_t* __restrict>(datap);
        while (size) {
            bufferCheck();
            size_t blk = size;
            if (blk > bufferInsertSize()) blk = bufferInsertSize();
-            const vluint8_t* __restrict maxp = dp + blk;
+            const uint8_t* __restrict maxp = dp + blk;
            for (; dp < maxp; *dp++ = *m_cp++) {}
            size -= blk;
        }
@@ -165,7 +165,7 @@ public:
    // Internal use:
    // Read a datum and compare with expected value
    VerilatedDeserialize& readAssert(const void* __restrict datap, size_t size) VL_MT_UNSAFE_ONE;
-    VerilatedDeserialize& readAssert(vluint64_t data) VL_MT_UNSAFE_ONE {
+    VerilatedDeserialize& readAssert(uint64_t data) VL_MT_UNSAFE_ONE {
        return readAssert(&data, sizeof(data));
    }

@@ -236,28 +236,28 @@ public:

 //=============================================================================

-inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const vluint64_t& rhs) {
+inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const uint64_t& rhs) {
    return os.write(&rhs, sizeof(rhs));
 }
-inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, vluint64_t& rhs) {
+inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, uint64_t& rhs) {
    return os.read(&rhs, sizeof(rhs));
 }
-inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const vluint32_t& rhs) {
+inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const uint32_t& rhs) {
    return os.write(&rhs, sizeof(rhs));
 }
-inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, vluint32_t& rhs) {
+inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, uint32_t& rhs) {
    return os.read(&rhs, sizeof(rhs));
 }
-inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const vluint16_t& rhs) {
+inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const uint16_t& rhs) {
    return os.write(&rhs, sizeof(rhs));
 }
-inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, vluint16_t& rhs) {
+inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, uint16_t& rhs) {
    return os.read(&rhs, sizeof(rhs));
 }
-inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const vluint8_t& rhs) {
+inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const uint8_t& rhs) {
    return os.write(&rhs, sizeof(rhs));
 }
-inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, vluint8_t& rhs) {
+inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, uint8_t& rhs) {
    return os.read(&rhs, sizeof(rhs));
 }
 inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const bool& rhs) {
@@ -279,12 +279,12 @@ inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, float& rhs) {
    return os.read(&rhs, sizeof(rhs));
 }
 inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const std::string& rhs) {
-    const vluint32_t len = rhs.length();
+    const uint32_t len = rhs.length();
    os << len;
    return os.write(rhs.data(), len);
 }
 inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, std::string& rhs) {
-    vluint32_t len = 0;
+    uint32_t len = 0;
    os >> len;
    rhs.resize(len);
    return os.read((void*)rhs.data(), len);
@@ -295,7 +295,7 @@ VerilatedDeserialize& operator>>(VerilatedDeserialize& os, VerilatedContext* rhs
 template <class T_Key, class T_Value>
 VerilatedSerialize& operator<<(VerilatedSerialize& os, VlAssocArray<T_Key, T_Value>& rhs) {
    os << rhs.atDefault();
-    const vluint32_t len = rhs.size();
+    const uint32_t len = rhs.size();
    os << len;
    for (const auto& i : rhs) {
        const T_Key index = i.first;  // Copy to get around const_iterator
@@ -307,10 +307,10 @@ VerilatedSerialize& operator<<(VerilatedSerialize& os, VlAssocArray<T_Key, T_Val
 template <class T_Key, class T_Value>
 VerilatedDeserialize& operator>>(VerilatedDeserialize& os, VlAssocArray<T_Key, T_Value>& rhs) {
    os >> rhs.atDefault();
-    vluint32_t len = 0;
+    uint32_t len = 0;
    os >> len;
    rhs.clear();
-    for (vluint32_t i = 0; i < len; ++i) {
+    for (uint32_t i = 0; i < len; ++i) {
        T_Key index;
        T_Value value;
        os >> index;
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2009-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2009-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -39,10 +39,10 @@
 // This class is thread safe (though most of SystemC is not).
 class VlScBvExposer final : public sc_bv_base {
 public:
-    static const vluint32_t* sp_datap(const sc_bv_base& base) VL_MT_SAFE {
+    static const uint32_t* sp_datap(const sc_bv_base& base) VL_MT_SAFE {
        return static_cast<const VlScBvExposer*>(&base)->sp_datatp();
    }
-    const vluint32_t* sp_datatp() const { return reinterpret_cast<vluint32_t*>(m_data); }
+    const uint32_t* sp_datatp() const { return reinterpret_cast<uint32_t*>(m_data); }
    // Above reads this protected element in sc_bv_base:
    //   sc_digit* m_data; // data array
 };
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -69,9 +69,9 @@ public:

 class VerilatedVarProps VL_NOT_FINAL {
    // TYPES
-    static constexpr vluint32_t MAGIC = 0xddc4f829UL;
+    static constexpr uint32_t MAGIC = 0xddc4f829UL;
    // MEMBERS
-    const vluint32_t m_magic;  // Magic number
+    const uint32_t m_magic;  // Magic number
    const VerilatedVarType m_vltype;  // Data type
    const VerilatedVarFlags m_vlflags;  // Direction
    const int m_pdims;  // Packed dimensions, 0 = none
@@ -142,7 +142,7 @@ public:
    VerilatedVarFlags vldir() const {
        return static_cast<VerilatedVarFlags>(static_cast<int>(m_vlflags) & VLVF_MASK_DIR);
    }
-    vluint32_t entSize() const;
+    uint32_t entSize() const;
    bool isPublicRW() const { return ((m_vlflags & VLVF_PUB_RW) != 0); }
    // DPI compatible C standard layout
    bool isDpiCLayout() const { return ((m_vlflags & VLVF_DPI_CLAY) != 0); }
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2012-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2012-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -24,22 +24,25 @@
 #include "verilatedos.h"
 #include "verilated_threads.h"

+#ifdef VL_PROFILER
+#include "verilated_profiler.h"
+#endif
+
 #include <cstdio>
-#include <fstream>
+#include <memory>
+#include <string>

 //=============================================================================
 // Globals

 // Internal note: Globals may multi-construct, see verilated.cpp top.

-std::atomic<vluint64_t> VlMTaskVertex::s_yields;
-
-VL_THREAD_LOCAL VlThreadPool::ProfileTrace* VlThreadPool::t_profilep = nullptr;
+std::atomic<uint64_t> VlMTaskVertex::s_yields;

 //=============================================================================
 // VlMTaskVertex

-VlMTaskVertex::VlMTaskVertex(vluint32_t upstreamDepCount)
+VlMTaskVertex::VlMTaskVertex(uint32_t upstreamDepCount)
    : m_upstreamDepsDone{0}
    , m_upstreamDepCount{upstreamDepCount} {
    assert(atomic_is_lock_free(&m_upstreamDepsDone));
@@ -48,12 +51,11 @@ VlMTaskVertex::VlMTaskVertex(vluint32_t upstreamDepCount)
 //=============================================================================
 // VlWorkerThread

-VlWorkerThread::VlWorkerThread(VlThreadPool* poolp, VerilatedContext* contextp, bool profiling)
+VlWorkerThread::VlWorkerThread(uint32_t threadId, VerilatedContext* contextp,
+                               VlExecutionProfiler* profilerp)
    : m_ready_size{0}
-    , m_poolp{poolp}
-    , m_profiling{profiling}  // Must init this last -- after setting up fields that it might read:
    , m_exiting{false}
-    , m_cthread{startWorker, this}
+    , m_cthread{startWorker, this, threadId, profilerp}
    , m_contextp{contextp} {}

 VlWorkerThread::~VlWorkerThread() {
@@ -64,8 +66,6 @@ VlWorkerThread::~VlWorkerThread() {
 }

 void VlWorkerThread::workerLoop() {
-    if (VL_UNLIKELY(m_profiling)) m_poolp->setupProfilingClientThread();
-
    ExecRec work;
    work.m_fnp = nullptr;

@@ -80,144 +80,42 @@ void VlWorkerThread::workerLoop() {
            work.m_fnp = nullptr;
        }
    }
-
-    if (VL_UNLIKELY(m_profiling)) m_poolp->tearDownProfilingClientThread();
 }

-void VlWorkerThread::startWorker(VlWorkerThread* workerp) {
+void VlWorkerThread::startWorker(VlWorkerThread* workerp, uint32_t threadId,
+                                 VlExecutionProfiler* profilerp) {
    Verilated::threadContextp(workerp->m_contextp);
+#ifdef VL_PROFILER
+    // Note: setupThread is not defined without VL_PROFILER, hence the #ifdef. Still, we might
+    // not be profiling execution (e.g.: PGO only), so profilerp might still be nullptr.
+    if (profilerp) profilerp->setupThread(threadId);
+#endif
    workerp->workerLoop();
 }

 //=============================================================================
 // VlThreadPool

-VlThreadPool::VlThreadPool(VerilatedContext* contextp, int nThreads, bool profiling)
-    : m_profiling{profiling} {
+VlThreadPool::VlThreadPool(VerilatedContext* contextp, int nThreads,
+                           VlExecutionProfiler* profiler) {
    // --threads N passes nThreads=N-1, as the "main" threads counts as 1
+    ++nThreads;
    const unsigned cpus = std::thread::hardware_concurrency();
-    if (cpus < nThreads + 1) {
+    if (cpus < nThreads) {
        static int warnedOnce = 0;
        if (!warnedOnce++) {
            VL_PRINTF_MT("%%Warning: System has %u CPUs but model Verilated with"
                         " --threads %d; may run slow.\n",
-                         cpus, nThreads + 1);
+                         cpus, nThreads);
        }
    }
-    // Create'em
-    for (int i = 0; i < nThreads; ++i) {
-        m_workers.push_back(new VlWorkerThread{this, contextp, profiling});
+    // Create worker threads
+    for (uint32_t threadId = 1; threadId < nThreads; ++threadId) {
+        m_workers.push_back(new VlWorkerThread{threadId, contextp, profiler});
    }
-    // Set up a profile buffer for the current thread too -- on the
-    // assumption that it's the same thread that calls eval and may be
-    // donated to run mtasks during the eval.
-    if (VL_UNLIKELY(m_profiling)) setupProfilingClientThread();
 }

 VlThreadPool::~VlThreadPool() {
    // Each ~WorkerThread will wait for its thread to exit.
    for (auto& i : m_workers) delete i;
-    if (VL_UNLIKELY(m_profiling)) tearDownProfilingClientThread();
-}
-
-void VlThreadPool::tearDownProfilingClientThread() {
-    assert(t_profilep);
-    delete t_profilep;
-    t_profilep = nullptr;
-}
-
-void VlThreadPool::setupProfilingClientThread() VL_MT_SAFE_EXCLUDES(m_mutex) {
-    assert(!t_profilep);
-    t_profilep = new ProfileTrace;
-    // Reserve some space in the thread-local profiling buffer;
-    // try not to malloc while collecting profiling.
-    t_profilep->reserve(4096);
-    {
-        const VerilatedLockGuard lock{m_mutex};
-        m_allProfiles.insert(t_profilep);
-    }
-}
-
-void VlThreadPool::profileAppendAll(const VlProfileRec& rec) VL_MT_SAFE_EXCLUDES(m_mutex) {
-    const VerilatedLockGuard lock{m_mutex};
-    for (const auto& profilep : m_allProfiles) {
-        // Every thread's profile trace gets a copy of rec.
-        profilep->emplace_back(rec);
-    }
-}
-
-void VlThreadPool::profileDump(const char* filenamep, vluint64_t tickStart, vluint64_t tickEnd)
-    VL_MT_SAFE_EXCLUDES(m_mutex) {
-    const VerilatedLockGuard lock{m_mutex};
-    VL_DEBUG_IF(VL_DBG_MSGF("+prof+threads writing to '%s'\n", filenamep););
-
-    FILE* const fp = std::fopen(filenamep, "w");
-    if (VL_UNLIKELY(!fp)) {
-        VL_FATAL_MT(filenamep, 0, "", "+prof+threads+file file not writable");
-        // cppcheck-suppress resourceLeak   // bug, doesn't realize fp is nullptr
-        return;  // LCOV_EXCL_LINE
-    }
-
-    // TODO Perhaps merge with verilated_coverage output format, so can
-    // have a common merging and reporting tool, etc.
-    fprintf(fp, "VLPROFTHREAD 1.1 # Verilator thread profile dump version 1.1\n");
-    fprintf(fp, "VLPROF arg --threads %" VL_PRI64 "u\n", vluint64_t(m_workers.size() + 1));
-    fprintf(fp, "VLPROF arg +verilator+prof+threads+start+%" VL_PRI64 "u\n",
-            Verilated::threadContextp()->profThreadsStart());
-    fprintf(fp, "VLPROF arg +verilator+prof+threads+window+%u\n",
-            Verilated::threadContextp()->profThreadsWindow());
-    fprintf(fp, "VLPROF stat yields %" VL_PRI64 "u\n", VlMTaskVertex::yields());
-
-    // Copy /proc/cpuinfo into this output so verilator_gantt can be run on
-    // a different machine
-    {
-        const std::unique_ptr<std::ifstream> ifp{new std::ifstream("/proc/cpuinfo")};
-        if (!ifp->fail()) {
-            std::string line;
-            while (std::getline(*ifp, line)) { fprintf(fp, "VLPROFPROC %s\n", line.c_str()); }
-        }
-    }
-
-    vluint32_t thread_id = 0;
-    for (const auto& pi : m_allProfiles) {
-        ++thread_id;
-
-        bool printing = false;  // False while in warmup phase
-        for (const auto& ei : *pi) {
-            switch (ei.m_type) {
-            case VlProfileRec::TYPE_BARRIER:  //
-                printing = true;
-                break;
-            case VlProfileRec::TYPE_EVAL:
-                if (!printing) break;
-                fprintf(fp,
-                        "VLPROF eval start %" VL_PRI64 "u elapsed %" VL_PRI64 "u"
-                        " cpu %u on thread %u\n",
-                        ei.m_startTime - tickStart, (ei.m_endTime - ei.m_startTime), ei.m_cpu,
-                        thread_id);
-                break;
-            case VlProfileRec::TYPE_EVAL_LOOP:
-                if (!printing) break;
-                fprintf(fp,
-                        "VLPROF eval_loop start %" VL_PRI64 "u elapsed %" VL_PRI64 "u"
-                        " cpu %u on thread %u\n",
-                        ei.m_startTime - tickStart, (ei.m_endTime - ei.m_startTime), ei.m_cpu,
-                        thread_id);
-                break;
-            case VlProfileRec::TYPE_MTASK_RUN:
-                if (!printing) break;
-                fprintf(fp,
-                        "VLPROF mtask %d"
-                        " start %" VL_PRI64 "u elapsed %" VL_PRI64 "u"
-                        " predict_start %u predict_cost %u cpu %u on thread %u\n",
-                        ei.m_mtaskId, ei.m_startTime - tickStart, (ei.m_endTime - ei.m_startTime),
-                        ei.m_predictStart, ei.m_predictCost, ei.m_cpu, thread_id);
-                break;
-            default: assert(false); break;  // LCOV_EXCL_LINE
-            }
-        }
-    }
-    fprintf(fp, "VLPROF stat ticks %" VL_PRI64 "u\n", tickEnd - tickStart);
-
-    std::fclose(fp);
 }
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2012-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2012-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -35,15 +35,17 @@
 #error "verilated_threads.h/cpp expected VL_THREADED (from verilator --threads)"
 #endif

+#include <atomic>
 #include <condition_variable>
 #include <set>
+#include <thread>
 #include <vector>

 // clang-format off
 #if defined(__linux)
 # include <sched.h>  // For sched_getcpu()
 #endif
-#if defined(__APPLE__)
+#if defined(__APPLE__) && !defined(__arm64__)
 # include <cpuid.h>  // For __cpuid_count()
 #endif
 // clang-format on
@@ -58,7 +60,7 @@ using VlExecFnp = void (*)(VlSelfP, bool);
 // Track dependencies for a single MTask.
 class VlMTaskVertex final {
    // MEMBERS
-    static std::atomic<vluint64_t> s_yields;  // Statistics
+    static std::atomic<uint64_t> s_yields;  // Statistics

    // On even cycles, _upstreamDepsDone increases as upstream
    // dependencies complete. When it reaches _upstreamDepCount,
@@ -76,8 +78,8 @@ class VlMTaskVertex final {
    // during done-notification. Nobody's quantified that cost though.
    // If we were really serious about shrinking this class, we could
    // use 16-bit types here...)
-    std::atomic<vluint32_t> m_upstreamDepsDone;
-    const vluint32_t m_upstreamDepCount;
+    std::atomic<uint32_t> m_upstreamDepsDone;
+    const uint32_t m_upstreamDepCount;

 public:
    // CONSTRUCTORS
@@ -85,10 +87,10 @@ public:
    // 'upstreamDepCount' is the number of upstream MTaskVertex's
    // that must notify this MTaskVertex before it will become ready
    // to run.
-    explicit VlMTaskVertex(vluint32_t upstreamDepCount);
+    explicit VlMTaskVertex(uint32_t upstreamDepCount);
    ~VlMTaskVertex() = default;

-    static vluint64_t yields() { return s_yields; }
+    static uint64_t yields() { return s_yields; }
    static void yieldThread() {
        ++s_yields;  // Statistics
        std::this_thread::yield();
@@ -99,19 +101,19 @@ public:
    // false while it's still waiting on more dependencies.
    inline bool signalUpstreamDone(bool evenCycle) {
        if (evenCycle) {
-            const vluint32_t upstreamDepsDone
+            const uint32_t upstreamDepsDone
                = 1 + m_upstreamDepsDone.fetch_add(1, std::memory_order_release);
            assert(upstreamDepsDone <= m_upstreamDepCount);
            return (upstreamDepsDone == m_upstreamDepCount);
        } else {
-            const vluint32_t upstreamDepsDone_prev
+            const uint32_t upstreamDepsDone_prev
                = m_upstreamDepsDone.fetch_sub(1, std::memory_order_release);
            assert(upstreamDepsDone_prev > 0);
            return (upstreamDepsDone_prev == 1);
        }
    }
    inline bool areUpstreamDepsDone(bool evenCycle) const {
-        const vluint32_t target = evenCycle ? m_upstreamDepCount : 0;
+        const uint32_t target = evenCycle ? m_upstreamDepCount : 0;
        return m_upstreamDepsDone.load(std::memory_order_acquire) == target;
    }
    inline void waitUntilUpstreamDone(bool evenCycle) const {
@@ -127,64 +129,7 @@ public:
    }
 };

-// Profiling support
-class VlProfileRec final {
-protected:
-    friend class VlThreadPool;
-    enum VlProfileE { TYPE_MTASK_RUN, TYPE_EVAL, TYPE_EVAL_LOOP, TYPE_BARRIER };
-    // Layout below allows efficient packing.
-    // Leave endTime first, so no math needed to calculate address in endRecord
-    vluint64_t m_endTime = 0;  // Tick at end of execution
-    vluint64_t m_startTime = 0;  // Tick at start of execution
-    vluint32_t m_mtaskId = 0;  // Mtask we're logging
-    vluint32_t m_predictStart = 0;  // Time scheduler predicted would start
-    vluint32_t m_predictCost = 0;  // How long scheduler predicted would take
-    VlProfileE m_type = TYPE_BARRIER;  // Record type
-    unsigned m_cpu;  // Execution CPU number (at start anyways)
-public:
-    class Barrier {};
-    VlProfileRec() = default;
-    explicit VlProfileRec(Barrier) { m_cpu = getcpu(); }
-    void startEval(vluint64_t time) {
-        m_type = VlProfileRec::TYPE_EVAL;
-        m_startTime = time;
-        m_cpu = getcpu();
-    }
-    void startEvalLoop(vluint64_t time) {
-        m_type = VlProfileRec::TYPE_EVAL_LOOP;
-        m_startTime = time;
-        m_cpu = getcpu();
-    }
-    void startRecord(vluint64_t time, vluint32_t mtask, vluint32_t predictStart,
-                     vluint32_t predictCost) {
-        m_type = VlProfileRec::TYPE_MTASK_RUN;
-        m_mtaskId = mtask;
-        m_predictStart = predictStart;
-        m_predictCost = predictCost;
-        m_startTime = time;
-        m_cpu = getcpu();
-    }
-    void endRecord(vluint64_t time) { m_endTime = time; }
-    static int getcpu() {  // Return current executing CPU
-#if defined(__linux)
-        return sched_getcpu();
-#elif defined(__APPLE__)
-        vluint32_t info[4];
-        __cpuid_count(1, 0, info[0], info[1], info[2], info[3]);
-        // info[1] is EBX, bits 24-31 are APIC ID
-        if ((info[3] & (1 << 9)) == 0) {
-            return -1;  // no APIC on chip
-        } else {
-            return (unsigned)info[1] >> 24;
-        }
-#elif defined(_WIN32)
-        return GetCurrentProcessorNumber();
-#else
-        return 0;
-#endif
-    }
-};
-
+class VlExecutionProfiler;
 class VlThreadPool;

 class VlWorkerThread final {
@@ -205,7 +150,7 @@ private:
    };

    // MEMBERS
-    VerilatedMutex m_mutex;
+    mutable VerilatedMutex m_mutex;
    std::condition_variable_any m_cv;
    // Only notify the condition_variable if the worker is waiting
    bool m_waiting VL_GUARDED_BY(m_mutex) = false;
@@ -217,9 +162,6 @@ private:
    // Store the size atomically, so we can spin wait
    std::atomic<size_t> m_ready_size;

-    VlThreadPool* const m_poolp;  // Our associated thread pool
-
-    const bool m_profiling;  // Is profiling enabled?
    std::atomic<bool> m_exiting;  // Worker thread should exit
    std::thread m_cthread;  // Underlying C++ thread record
    VerilatedContext* const m_contextp;  // Context for spawned thread
@@ -228,7 +170,8 @@ private:

 public:
    // CONSTRUCTORS
-    explicit VlWorkerThread(VlThreadPool* poolp, VerilatedContext* contextp, bool profiling);
+    explicit VlWorkerThread(uint32_t threadId, VerilatedContext* contextp,
+                            VlExecutionProfiler* profilerp);
    ~VlWorkerThread();

    // METHODS
@@ -265,34 +208,20 @@ public:
        if (notify) m_cv.notify_one();
    }
    void workerLoop();
-    static void startWorker(VlWorkerThread* workerp);
+    static void startWorker(VlWorkerThread* workerp, uint32_t threadId,
+                            VlExecutionProfiler* profilerp);
 };

 class VlThreadPool final {
-    // TYPES
-    using ProfileTrace = std::vector<VlProfileRec>;
-
    // MEMBERS
    std::vector<VlWorkerThread*> m_workers;  // our workers
-    const bool m_profiling;  // is profiling enabled?
-
-    // Support profiling -- we can append records of profiling events
-    // to this vector with very low overhead, and then dump them out
-    // later. This prevents the overhead of printf/malloc/IO from
-    // corrupting the profiling data. It's super cheap to append
-    // a VlProfileRec struct on the end of a pre-allocated vector;
-    // this is the only cost we pay in real-time during a profiling cycle.
-    // Internal note: Globals may multi-construct, see verilated.cpp top.
-    static VL_THREAD_LOCAL ProfileTrace* t_profilep;
-    std::set<ProfileTrace*> m_allProfiles VL_GUARDED_BY(m_mutex);
-    VerilatedMutex m_mutex;

 public:
    // CONSTRUCTORS
    // Construct a thread pool with 'nThreads' dedicated threads. The thread
    // pool will create these threads and make them available to execute tasks
    // via this->workerp(index)->addTask(...)
-    VlThreadPool(VerilatedContext* contextp, int nThreads, bool profiling);
+    VlThreadPool(VerilatedContext* contextp, int nThreads, VlExecutionProfiler* profilerp);
    ~VlThreadPool();

    // METHODS
@@ -302,17 +231,6 @@ public:
        assert(index < m_workers.size());
        return m_workers[index];
    }
-    inline VlProfileRec* profileAppend() {
-        t_profilep->emplace_back();
-        return &(t_profilep->back());
-    }
-    void profileAppendAll(const VlProfileRec& rec) VL_MT_SAFE_EXCLUDES(m_mutex);
-    void profileDump(const char* filenamep, vluint64_t tickStart, vluint64_t tickEnd)
-        VL_MT_SAFE_EXCLUDES(m_mutex);
-    // In profiling mode, each executing thread must call
-    // this once to setup profiling state:
-    void setupProfilingClientThread() VL_MT_SAFE_EXCLUDES(m_mutex);
-    void tearDownProfilingClientThread();

 private:
    VL_UNCOPYABLE(VlThreadPool);
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -24,28 +24,49 @@

 // clang-format off

+// In FST mode, VL_TRACE_THREADED enables offloading, but only if we also have
+// the FST writer thread. This means with --trace-threads 1, we get the FST
+// writer thread only, and with --trace-threads 2 we get offloading as well
+#if defined(VL_TRACE_FST_WRITER_THREAD) && defined(VL_TRACE_THREADED)
+# define VL_TRACE_OFFLOAD
+#endif
+// VCD tracing can happen fully in parallel
+#if defined(VM_TRACE_VCD) && VM_TRACE_VCD && defined(VL_TRACE_THREADED)
+# define VL_TRACE_PARALLEL
+#endif
+
+#if defined(VL_TRACE_PARALLEL) && defined(VL_TRACE_OFFLOAD)
+# error "Cannot have VL_TRACE_PARALLEL and VL_TRACE_OFFLOAD together"
+#endif
+
 #include "verilated.h"
 #include "verilated_trace_defs.h"

+#include <bitset>
+#include <condition_variable>
+#include <memory>
 #include <string>
+#include <unordered_map>
 #include <vector>

-#ifdef VL_TRACE_THREADED
-# include <condition_variable>
+#ifdef VL_TRACE_OFFLOAD
 # include <deque>
 # include <thread>
 #endif

 // clang-format on

-#ifdef VL_TRACE_THREADED
+class VlThreadPool;
+template <class T_Trace, class T_Buffer> class VerilatedTraceBuffer;
+
+#ifdef VL_TRACE_OFFLOAD
 //=============================================================================
-// Threaded tracing
+// Offloaded tracing

 // A simple synchronized first in first out queue
 template <class T> class VerilatedThreadQueue final {  // LCOV_EXCL_LINE  // lcov bug
 private:
-    VerilatedMutex m_mutex;  // Protects m_queue
+    mutable VerilatedMutex m_mutex;  // Protects m_queue
    std::condition_variable_any m_cv;
    std::deque<T> m_queue VL_GUARDED_BY(m_mutex);

@@ -86,11 +107,11 @@ public:

 // Commands used by thread tracing. Anonymous enum in class, as we want
 // it scoped, but we also want the automatic conversion to integer types.
-class VerilatedTraceCommand final {
+class VerilatedTraceOffloadCommand final {
 public:
    // These must all fit in 4 bit at the moment, as the tracing routines
    // pack parameters in the top bits.
-    enum : vluint8_t {
+    enum : uint8_t {
        CHG_BIT_0 = 0x0,
        CHG_BIT_1 = 0x1,
        CHG_CDATA = 0x2,
@@ -100,7 +121,8 @@ public:
        CHG_WDATA = 0x6,
        CHG_DOUBLE = 0x8,
        // TODO: full..
-        TIME_CHANGE = 0xd,
+        TIME_CHANGE = 0xc,
+        TRACE_BUFFER = 0xd,
        END = 0xe,  // End of buffer
        SHUTDOWN = 0xf  // Shutdown worker thread, also marks end of buffer
    };
@@ -110,16 +132,22 @@ public:
 //=============================================================================
 // VerilatedTrace

-// VerilatedTrace uses F-bounded polymorphism to access duck-typed
-// implementations in the format specific derived class, which must be passed
-// as the type parameter T_Derived
-template <class T_Derived> class VerilatedTrace VL_NOT_FINAL {
+// T_Trace is the format specific subclass of VerilatedTrace.
+// T_Buffer is the format specific subclass of VerilatedTraceBuffer.
+template <class T_Trace, class T_Buffer> class VerilatedTrace VL_NOT_FINAL {
+    // Give the buffer (both base and derived) access to the private bits
+    friend VerilatedTraceBuffer<T_Trace, T_Buffer>;
+    friend T_Buffer;
+
 public:
+    using Buffer = T_Buffer;
+
    //=========================================================================
    // Generic tracing internals

-    using initCb_t = void (*)(void*, T_Derived*, vluint32_t);  // Type of init callbacks
-    using dumpCb_t = void (*)(void*, T_Derived*);  // Type of all but init callbacks
+    using initCb_t = void (*)(void*, T_Trace*, uint32_t);  // Type of init callbacks
+    using dumpCb_t = void (*)(void*, Buffer*);  // Type of dump callbacks
+    using cleanupCb_t = void (*)(void*, T_Trace*);  // Type of cleanup callbacks

 private:
    struct CallbackRecord {
@@ -127,9 +155,10 @@ private:
        // (the one in Ubuntu 14.04 with GCC 4.8.4 in particular) use the
        // assignment operator on inserting into collections, so they don't work
        // with const fields...
-        union {
-            initCb_t m_initCb;  // The callback function
-            dumpCb_t m_dumpCb;  // The callback function
+        union {  // The callback
+            initCb_t m_initCb;
+            dumpCb_t m_dumpCb;
+            cleanupCb_t m_cleanupCb;
        };
        void* m_userp;  // The user pointer to pass to the callback (the symbol table)
        CallbackRecord(initCb_t cb, void* userp)
@@ -138,67 +167,103 @@ private:
        CallbackRecord(dumpCb_t cb, void* userp)
            : m_dumpCb{cb}
            , m_userp{userp} {}
+        CallbackRecord(cleanupCb_t cb, void* userp)
+            : m_cleanupCb{cb}
+            , m_userp{userp} {}
    };

-    vluint32_t* m_sigs_oldvalp;  // Old value store
-    vluint64_t m_timeLastDump;  // Last time we did a dump
-    std::vector<CallbackRecord> m_initCbs;  // Routines to initialize traciong
-    std::vector<CallbackRecord> m_fullCbs;  // Routines to perform full dump
-    std::vector<CallbackRecord> m_chgCbs;  // Routines to perform incremental dump
+#ifdef VL_TRACE_PARALLEL
+    struct ParallelWorkerData {
+        const dumpCb_t m_cb;  // The callback
+        void* const m_userp;  // The use pointer to pass to the callback
+        Buffer* const m_bufp;  // The buffer pointer to pass to the callback
+        std::atomic<bool> m_ready{false};  // The ready flag
+        mutable VerilatedMutex m_mutex;  // Mutex for suspension until ready
+        std::condition_variable_any m_cv;  // Condition variable for suspension
+        bool m_waiting VL_GUARDED_BY(m_mutex) = false;  // Whether a thread is suspended in wait()
+
+        void wait();
+
+        ParallelWorkerData(dumpCb_t cb, void* userp, Buffer* bufp)
+            : m_cb{cb}
+            , m_userp{userp}
+            , m_bufp{bufp} {}
+    };
+
+    // Passed a ParallelWorkerData*, second argument is ignored
+    static void parallelWorkerTask(void*, bool);
+#endif
+
+    using ParallelCallbackMap = std::unordered_map<VlThreadPool*, std::vector<CallbackRecord>>;
+
+protected:
+    uint32_t* m_sigs_oldvalp = nullptr;  // Previous value store
+    EData* m_sigs_enabledp = nullptr;  // Bit vector of enabled codes (nullptr = all on)
+private:
+    uint64_t m_timeLastDump = 0;  // Last time we did a dump
+    std::vector<bool> m_sigs_enabledVec;  // Staging for m_sigs_enabledp
+    std::vector<CallbackRecord> m_initCbs;  // Routines to initialize tracing
+    ParallelCallbackMap m_fullCbs;  // Routines to perform full dump
+    ParallelCallbackMap m_chgCbs;  // Routines to perform incremental dump
    std::vector<CallbackRecord> m_cleanupCbs;  // Routines to call at the end of dump
-    bool m_fullDump;  // Whether a full dump is required on the next call to 'dump'
-    vluint32_t m_nextCode;  // Next code number to assign
-    vluint32_t m_numSignals;  // Number of distinct signals
-    vluint32_t m_maxBits;  // Number of bits in the widest signal
-    std::string m_moduleName;  // Name of module being trace initialized now
-    char m_scopeEscape;
-    double m_timeRes;  // Time resolution (ns/ms etc)
-    double m_timeUnit;  // Time units (ns/ms etc)
+    std::vector<VlThreadPool*> m_threadPoolps;  // All thread pools, in insertion order
+    bool m_fullDump = true;  // Whether a full dump is required on the next call to 'dump'
+    uint32_t m_nextCode = 0;  // Next code number to assign
+    uint32_t m_numSignals = 0;  // Number of distinct signals
+    uint32_t m_maxBits = 0;  // Number of bits in the widest signal
+    std::vector<std::string> m_namePrefixStack{""};  // Path prefixes to add to signal names
+    std::vector<std::pair<int, std::string>> m_dumpvars;  // dumpvar() entries
+    char m_scopeEscape = '.';
+    double m_timeRes = 1e-9;  // Time resolution (ns/ms etc)
+    double m_timeUnit = 1e-0;  // Time units (ns/ms etc)
+
+    void addThreadPool(VlThreadPool* threadPoolp) VL_MT_SAFE_EXCLUDES(m_mutex);

    void addCallbackRecord(std::vector<CallbackRecord>& cbVec, CallbackRecord& cbRec)
        VL_MT_SAFE_EXCLUDES(m_mutex);

-    // Equivalent to 'this' but is of the sub-type 'T_Derived*'. Use 'self()->'
+    // Equivalent to 'this' but is of the sub-type 'T_Trace*'. Use 'self()->'
    // to access duck-typed functions to avoid a virtual function call.
-    T_Derived* self() { return static_cast<T_Derived*>(this); }
+    T_Trace* self() { return static_cast<T_Trace*>(this); }
+
+    void runParallelCallbacks(const ParallelCallbackMap& cbMap);

    // Flush any remaining data for this file
    static void onFlush(void* selfp) VL_MT_UNSAFE_ONE;
    // Close the file on termination
    static void onExit(void* selfp) VL_MT_UNSAFE_ONE;

-#ifdef VL_TRACE_THREADED
-    // Number of total trace buffers that have been allocated
-    vluint32_t m_numTraceBuffers;
-
-    // Size of trace buffers
-    size_t m_traceBufferSize;
-
+#ifdef VL_TRACE_OFFLOAD
+    // Number of total offload buffers that have been allocated
+    uint32_t m_numOffloadBuffers = 0;
+    // Size of offload buffers
+    size_t m_offloadBufferSize = 0;
    // Buffers handed to worker for processing
-    VerilatedThreadQueue<vluint32_t*> m_buffersToWorker;
+    VerilatedThreadQueue<uint32_t*> m_offloadBuffersToWorker;
    // Buffers returned from worker after processing
-    VerilatedThreadQueue<vluint32_t*> m_buffersFromWorker;
-
-    // Get a new trace buffer that can be populated. May block if none available
-    vluint32_t* getTraceBuffer();
+    VerilatedThreadQueue<uint32_t*> m_offloadBuffersFromWorker;

+protected:
    // Write pointer into current buffer
-    vluint32_t* m_traceBufferWritep;
+    uint32_t* m_offloadBufferWritep = nullptr;
+    // End of offload buffer
+    uint32_t* m_offloadBufferEndp = nullptr;

-    // End of trace buffer
-    vluint32_t* m_traceBufferEndp;
-
-    // The worker thread itself
+private:
+    // The offload worker thread itself
    std::unique_ptr<std::thread> m_workerThread;

-    // The function executed by the worker thread
-    void workerThreadMain();
+    // Get a new offload buffer that can be populated. May block if none available
+    uint32_t* getOffloadBuffer();

-    // Wait until given buffer is placed in m_buffersFromWorker
-    void waitForBuffer(const vluint32_t* bufferp);
+    // The function executed by the offload worker thread
+    void offloadWorkerThreadMain();
+
+    // Wait until given offload buffer is placed in m_offloadBuffersFromWorker
+    void waitForOffloadBuffer(const uint32_t* bufferp);

    // Shut down and join worker, if it's running, otherwise do nothing
-    void shutdownWorker();
+    void shutdownOffloadWorker();
 #endif

    // CONSTRUCTORS
@@ -208,14 +273,13 @@ protected:
    //=========================================================================
    // Internals available to format specific implementations

-    VerilatedMutex m_mutex;  // Ensure dump() etc only called from single thread
+    mutable VerilatedMutex m_mutex;  // Ensure dump() etc only called from single thread

-    vluint32_t nextCode() const { return m_nextCode; }
-    vluint32_t numSignals() const { return m_numSignals; }
-    vluint32_t maxBits() const { return m_maxBits; }
-    const std::string& moduleName() const { return m_moduleName; }
+    uint32_t nextCode() const { return m_nextCode; }
+    uint32_t numSignals() const { return m_numSignals; }
+    uint32_t maxBits() const { return m_maxBits; }
    void fullDump(bool value) { m_fullDump = value; }
-    vluint64_t timeLastDump() { return m_timeLastDump; }
+    uint64_t timeLastDump() { return m_timeLastDump; }

    double timeRes() const { return m_timeRes; }
    double timeUnit() const { return m_timeUnit; }
@@ -223,12 +287,15 @@ protected:

    void traceInit() VL_MT_UNSAFE;

-    void declCode(vluint32_t code, vluint32_t bits, bool tri);
+    // Declare new signal and return true if enabled
+    bool declCode(uint32_t code, const char* namep, uint32_t bits, bool tri);

    // Is this an escape?
    bool isScopeEscape(char c) { return std::isspace(c) || c == m_scopeEscape; }
    // Character that splits scopes.  Note whitespace are ALWAYS escapes.
    char scopeEscape() { return m_scopeEscape; }
+    // Prefix to assume in signal declarations
+    const std::string& namePrefix() const { return m_namePrefixStack.back(); }

    void closeBase();
    void flushBase();
@@ -237,13 +304,17 @@ protected:
    // Virtual functions to be provided by the format specific implementation

    // Called when the trace moves forward to a new time point
-    virtual void emitTimeChange(vluint64_t timeui) = 0;
+    virtual void emitTimeChange(uint64_t timeui) = 0;

    // These hooks are called before a full or change based dump is produced.
    // The return value indicates whether to proceed with the dump.
    virtual bool preFullDump() = 0;
    virtual bool preChangeDump() = 0;

+    // Trace buffer management
+    virtual Buffer* getTraceBuffer() = 0;
+    virtual void commitTraceBuffer(Buffer*) = 0;
+
 public:
    //=========================================================================
    // External interface to client code
@@ -257,21 +328,62 @@ public:
    // Set time resolution (s/ms, defaults to ns)
    void set_time_resolution(const char* unitp) VL_MT_SAFE;
    void set_time_resolution(const std::string& unit) VL_MT_SAFE;
+    // Set variables to dump, using $dumpvars format
+    // If level = 0, dump everything and hier is then ignored
+    void dumpvars(int level, const std::string& hier) VL_MT_SAFE;

    // Call
-    void dump(vluint64_t timeui) VL_MT_SAFE_EXCLUDES(m_mutex);
+    void dump(uint64_t timeui) VL_MT_SAFE_EXCLUDES(m_mutex);
+
+    //=========================================================================
+    // Internal interface to Verilator generated code

    //=========================================================================
    // Non-hot path internal interface to Verilator generated code

    void addInitCb(initCb_t cb, void* userp) VL_MT_SAFE;
-    void addFullCb(dumpCb_t cb, void* userp) VL_MT_SAFE;
-    void addChgCb(dumpCb_t cb, void* userp) VL_MT_SAFE;
-    void addCleanupCb(dumpCb_t cb, void* userp) VL_MT_SAFE;
+    void addFullCb(dumpCb_t cb, void* userp, VlThreadPool* = nullptr) VL_MT_SAFE;
+    void addChgCb(dumpCb_t cb, void* userp, VlThreadPool* = nullptr) VL_MT_SAFE;
+    void addCleanupCb(cleanupCb_t cb, void* userp) VL_MT_SAFE;

-    void module(const std::string& name) VL_MT_UNSAFE;
    void scopeEscape(char flag) { m_scopeEscape = flag; }

+    void pushNamePrefix(const std::string&);
+    void popNamePrefix(unsigned count = 1);
+};
+
+//=============================================================================
+// VerilatedTraceBuffer
+
+// T_Trace is the format specific subclass of VerilatedTrace.
+// T_Buffer is the format specific subclass of VerilatedTraceBuffer.
+// The format-specific hot-path methods use duck-typing via T_Buffer for performance.
+template <class T_Trace, class T_Buffer> class VerilatedTraceBuffer VL_NOT_FINAL {
+    friend T_Trace;  // Give the trace file access to the private bits
+
+protected:
+    T_Trace& m_owner;  // The VerilatedTrace subclass that owns this buffer
+
+    // Previous value store
+    uint32_t* const m_sigs_oldvalp = m_owner.m_sigs_oldvalp;
+    // Bit vector of enabled codes (nullptr = all on)
+    EData* const m_sigs_enabledp = m_owner.m_sigs_enabledp;
+
+#ifdef VL_TRACE_OFFLOAD
+    // Write pointer into current buffer
+    uint32_t* m_offloadBufferWritep = m_owner.m_offloadBufferWritep;
+    // End of offload buffer
+    uint32_t* const m_offloadBufferEndp = m_owner.m_offloadBufferEndp;
+#endif
+
+    // Equivalent to 'this' but is of the sub-type 'T_Derived*'. Use 'self()->'
+    // to access duck-typed functions to avoid a virtual function call.
+    inline T_Buffer* self() { return static_cast<T_Buffer*>(this); }
+
+    explicit VerilatedTraceBuffer(T_Trace& owner);
+    virtual ~VerilatedTraceBuffer() = default;
+
+public:
    //=========================================================================
    // Hot path internal interface to Verilator generated code

@@ -281,108 +393,112 @@ public:
    // these here, but we cannot afford dynamic dispatch for calling these as
    // this is very hot code during tracing.

-    // duck-typed void emitBit(vluint32_t code, CData newval) = 0;
-    // duck-typed void emitCData(vluint32_t code, CData newval, int bits) = 0;
-    // duck-typed void emitSData(vluint32_t code, SData newval, int bits) = 0;
-    // duck-typed void emitIData(vluint32_t code, IData newval, int bits) = 0;
-    // duck-typed void emitQData(vluint32_t code, QData newval, int bits) = 0;
-    // duck-typed void emitWData(vluint32_t code, const WData* newvalp, int bits) = 0;
-    // duck-typed void emitDouble(vluint32_t code, double newval) = 0;
+    // duck-typed void emitBit(uint32_t code, CData newval) = 0;
+    // duck-typed void emitCData(uint32_t code, CData newval, int bits) = 0;
+    // duck-typed void emitSData(uint32_t code, SData newval, int bits) = 0;
+    // duck-typed void emitIData(uint32_t code, IData newval, int bits) = 0;
+    // duck-typed void emitQData(uint32_t code, QData newval, int bits) = 0;
+    // duck-typed void emitWData(uint32_t code, const WData* newvalp, int bits) = 0;
+    // duck-typed void emitDouble(uint32_t code, double newval) = 0;

-    vluint32_t* oldp(vluint32_t code) { return m_sigs_oldvalp + code; }
+    VL_ATTR_ALWINLINE inline uint32_t* oldp(uint32_t code) { return m_sigs_oldvalp + code; }

    // Write to previous value buffer value and emit trace entry.
-    void fullBit(vluint32_t* oldp, CData newval);
-    void fullCData(vluint32_t* oldp, CData newval, int bits);
-    void fullSData(vluint32_t* oldp, SData newval, int bits);
-    void fullIData(vluint32_t* oldp, IData newval, int bits);
-    void fullQData(vluint32_t* oldp, QData newval, int bits);
-    void fullWData(vluint32_t* oldp, const WData* newvalp, int bits);
-    void fullDouble(vluint32_t* oldp, double newval);
+    void fullBit(uint32_t* oldp, CData newval);
+    void fullCData(uint32_t* oldp, CData newval, int bits);
+    void fullSData(uint32_t* oldp, SData newval, int bits);
+    void fullIData(uint32_t* oldp, IData newval, int bits);
+    void fullQData(uint32_t* oldp, QData newval, int bits);
+    void fullWData(uint32_t* oldp, const WData* newvalp, int bits);
+    void fullDouble(uint32_t* oldp, double newval);

-#ifdef VL_TRACE_THREADED
-    // Threaded tracing. Just dump everything in the trace buffer
-    inline void chgBit(vluint32_t code, CData newval) {
-        m_traceBufferWritep[0] = VerilatedTraceCommand::CHG_BIT_0 | newval;
-        m_traceBufferWritep[1] = code;
-        m_traceBufferWritep += 2;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+#ifdef VL_TRACE_OFFLOAD
+    // Offloaded tracing. Just dump everything in the offload buffer
+    inline void chgBit(uint32_t code, CData newval) {
+        m_offloadBufferWritep[0] = VerilatedTraceOffloadCommand::CHG_BIT_0 | newval;
+        m_offloadBufferWritep[1] = code;
+        m_offloadBufferWritep += 2;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }
-    inline void chgCData(vluint32_t code, CData newval, int bits) {
-        m_traceBufferWritep[0] = (bits << 4) | VerilatedTraceCommand::CHG_CDATA;
-        m_traceBufferWritep[1] = code;
-        m_traceBufferWritep[2] = newval;
-        m_traceBufferWritep += 3;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+    inline void chgCData(uint32_t code, CData newval, int bits) {
+        m_offloadBufferWritep[0] = (bits << 4) | VerilatedTraceOffloadCommand::CHG_CDATA;
+        m_offloadBufferWritep[1] = code;
+        m_offloadBufferWritep[2] = newval;
+        m_offloadBufferWritep += 3;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }
-    inline void chgSData(vluint32_t code, SData newval, int bits) {
-        m_traceBufferWritep[0] = (bits << 4) | VerilatedTraceCommand::CHG_SDATA;
-        m_traceBufferWritep[1] = code;
-        m_traceBufferWritep[2] = newval;
-        m_traceBufferWritep += 3;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+    inline void chgSData(uint32_t code, SData newval, int bits) {
+        m_offloadBufferWritep[0] = (bits << 4) | VerilatedTraceOffloadCommand::CHG_SDATA;
+        m_offloadBufferWritep[1] = code;
+        m_offloadBufferWritep[2] = newval;
+        m_offloadBufferWritep += 3;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }
-    inline void chgIData(vluint32_t code, IData newval, int bits) {
-        m_traceBufferWritep[0] = (bits << 4) | VerilatedTraceCommand::CHG_IDATA;
-        m_traceBufferWritep[1] = code;
-        m_traceBufferWritep[2] = newval;
-        m_traceBufferWritep += 3;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+    inline void chgIData(uint32_t code, IData newval, int bits) {
+        m_offloadBufferWritep[0] = (bits << 4) | VerilatedTraceOffloadCommand::CHG_IDATA;
+        m_offloadBufferWritep[1] = code;
+        m_offloadBufferWritep[2] = newval;
+        m_offloadBufferWritep += 3;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }
-    inline void chgQData(vluint32_t code, QData newval, int bits) {
-        m_traceBufferWritep[0] = (bits << 4) | VerilatedTraceCommand::CHG_QDATA;
-        m_traceBufferWritep[1] = code;
-        *reinterpret_cast<QData*>(m_traceBufferWritep + 2) = newval;
-        m_traceBufferWritep += 4;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+    inline void chgQData(uint32_t code, QData newval, int bits) {
+        m_offloadBufferWritep[0] = (bits << 4) | VerilatedTraceOffloadCommand::CHG_QDATA;
+        m_offloadBufferWritep[1] = code;
+        *reinterpret_cast<QData*>(m_offloadBufferWritep + 2) = newval;
+        m_offloadBufferWritep += 4;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }
-    inline void chgWData(vluint32_t code, const WData* newvalp, int bits) {
-        m_traceBufferWritep[0] = (bits << 4) | VerilatedTraceCommand::CHG_WDATA;
-        m_traceBufferWritep[1] = code;
-        m_traceBufferWritep += 2;
-        for (int i = 0; i < (bits + 31) / 32; ++i) { *m_traceBufferWritep++ = newvalp[i]; }
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+    inline void chgWData(uint32_t code, const WData* newvalp, int bits) {
+        m_offloadBufferWritep[0] = (bits << 4) | VerilatedTraceOffloadCommand::CHG_WDATA;
+        m_offloadBufferWritep[1] = code;
+        m_offloadBufferWritep += 2;
+        for (int i = 0; i < (bits + 31) / 32; ++i) { *m_offloadBufferWritep++ = newvalp[i]; }
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }
-    inline void chgDouble(vluint32_t code, double newval) {
-        m_traceBufferWritep[0] = VerilatedTraceCommand::CHG_DOUBLE;
-        m_traceBufferWritep[1] = code;
+    inline void chgDouble(uint32_t code, double newval) {
+        m_offloadBufferWritep[0] = VerilatedTraceOffloadCommand::CHG_DOUBLE;
+        m_offloadBufferWritep[1] = code;
        // cppcheck-suppress invalidPointerCast
-        *reinterpret_cast<double*>(m_traceBufferWritep + 2) = newval;
-        m_traceBufferWritep += 4;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
+        *reinterpret_cast<double*>(m_offloadBufferWritep + 2) = newval;
+        m_offloadBufferWritep += 4;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
    }

-#define CHG(name) chg##name##Impl
-#else
-#define CHG(name) chg##name
+#define chgBit chgBitImpl
+#define chgCData chgCDataImpl
+#define chgSData chgSDataImpl
+#define chgIData chgIDataImpl
+#define chgQData chgQDataImpl
+#define chgWData chgWDataImpl
+#define chgDouble chgDoubleImpl
 #endif

-    // In non-threaded mode, these are called directly by the trace callbacks,
-    // and are called chg*. In threaded mode, they are called by the worker
+    // In non-offload mode, these are called directly by the trace callbacks,
+    // and are called chg*. In offload mode, they are called by the worker
    // thread and are called chg*Impl

    // Check previous dumped value of signal. If changed, then emit trace entry
-    inline void CHG(Bit)(vluint32_t* oldp, CData newval) {
-        const vluint32_t diff = *oldp ^ newval;
+    VL_ATTR_ALWINLINE inline void chgBit(uint32_t* oldp, CData newval) {
+        const uint32_t diff = *oldp ^ newval;
        if (VL_UNLIKELY(diff)) fullBit(oldp, newval);
    }
-    inline void CHG(CData)(vluint32_t* oldp, CData newval, int bits) {
-        const vluint32_t diff = *oldp ^ newval;
+    VL_ATTR_ALWINLINE inline void chgCData(uint32_t* oldp, CData newval, int bits) {
+        const uint32_t diff = *oldp ^ newval;
        if (VL_UNLIKELY(diff)) fullCData(oldp, newval, bits);
    }
-    inline void CHG(SData)(vluint32_t* oldp, SData newval, int bits) {
-        const vluint32_t diff = *oldp ^ newval;
+    VL_ATTR_ALWINLINE inline void chgSData(uint32_t* oldp, SData newval, int bits) {
+        const uint32_t diff = *oldp ^ newval;
        if (VL_UNLIKELY(diff)) fullSData(oldp, newval, bits);
    }
-    inline void CHG(IData)(vluint32_t* oldp, IData newval, int bits) {
-        const vluint32_t diff = *oldp ^ newval;
+    VL_ATTR_ALWINLINE inline void chgIData(uint32_t* oldp, IData newval, int bits) {
+        const uint32_t diff = *oldp ^ newval;
        if (VL_UNLIKELY(diff)) fullIData(oldp, newval, bits);
    }
-    inline void CHG(QData)(vluint32_t* oldp, QData newval, int bits) {
-        const vluint64_t diff = *reinterpret_cast<QData*>(oldp) ^ newval;
+    VL_ATTR_ALWINLINE inline void chgQData(uint32_t* oldp, QData newval, int bits) {
+        const uint64_t diff = *reinterpret_cast<QData*>(oldp) ^ newval;
        if (VL_UNLIKELY(diff)) fullQData(oldp, newval, bits);
    }
-    inline void CHG(WData)(vluint32_t* oldp, const WData* newvalp, int bits) {
+    VL_ATTR_ALWINLINE inline void chgWData(uint32_t* oldp, const WData* newvalp, int bits) {
        for (int i = 0; i < (bits + 31) / 32; ++i) {
            if (VL_UNLIKELY(oldp[i] ^ newvalp[i])) {
                fullWData(oldp, newvalp, bits);
@@ -390,11 +506,20 @@ public:
            }
        }
    }
-    inline void CHG(Double)(vluint32_t* oldp, double newval) {
+    VL_ATTR_ALWINLINE inline void chgDouble(uint32_t* oldp, double newval) {
        // cppcheck-suppress invalidPointerCast
        if (VL_UNLIKELY(*reinterpret_cast<double*>(oldp) != newval)) fullDouble(oldp, newval);
    }

-#undef CHG
+#ifdef VL_TRACE_OFFLOAD
+#undef chgBit
+#undef chgCData
+#undef chgSData
+#undef chgIData
+#undef chgQData
+#undef chgWData
+#undef chgDouble
+#endif
 };
+
 #endif  // guard
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -1,688 +0,0 @@
-// -*- mode: C++; c-file-style: "cc-mode" -*-
-//=============================================================================
-//
-// Code available from: https://verilator.org
-//
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
-// can redistribute it and/or modify it under the terms of either the GNU
-// Lesser General Public License Version 3 or the Perl Artistic License
-// Version 2.0.
-// SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0
-//
-//=============================================================================
-///
-/// \file
-/// \brief Verilated common-format tracing implementation code
-///
-/// This file must be compiled and linked against all Verilated objects
-/// that use --trace.
-///
-/// Use "verilator --trace" to add this to the Makefile for the linker.
-///
-//=============================================================================
-
-// clang-format off
-
-#ifndef VL_CPPCHECK
-#ifndef VL_DERIVED_T
-# error "This file should be included in trace format implementations"
-#endif
-
-#include "verilated_intrinsics.h"
-#include "verilated_trace.h"
-
-#if 0
-# include <iostream>
-# define VL_TRACE_THREAD_DEBUG(msg) std::cout << "TRACE THREAD: " << msg << std::endl
-#else
-# define VL_TRACE_THREAD_DEBUG(msg)
-#endif
-
-// clang-format on
-
-//=============================================================================
-// Static utility functions
-
-static double timescaleToDouble(const char* unitp) {
-    char* endp = nullptr;
-    double value = std::strtod(unitp, &endp);
-    // On error so we allow just "ns" to return 1e-9.
-    if (value == 0.0 && endp == unitp) value = 1;
-    unitp = endp;
-    for (; *unitp && std::isspace(*unitp); unitp++) {}
-    switch (*unitp) {
-    case 's': value *= 1e0; break;
-    case 'm': value *= 1e-3; break;
-    case 'u': value *= 1e-6; break;
-    case 'n': value *= 1e-9; break;
-    case 'p': value *= 1e-12; break;
-    case 'f': value *= 1e-15; break;
-    case 'a': value *= 1e-18; break;
-    }
-    return value;
-}
-
-static std::string doubleToTimescale(double value) {
-    const char* suffixp = "s";
-    // clang-format off
-    if      (value >= 1e0)   { suffixp = "s"; value *= 1e0; }
-    else if (value >= 1e-3 ) { suffixp = "ms"; value *= 1e3; }
-    else if (value >= 1e-6 ) { suffixp = "us"; value *= 1e6; }
-    else if (value >= 1e-9 ) { suffixp = "ns"; value *= 1e9; }
-    else if (value >= 1e-12) { suffixp = "ps"; value *= 1e12; }
-    else if (value >= 1e-15) { suffixp = "fs"; value *= 1e15; }
-    else if (value >= 1e-18) { suffixp = "as"; value *= 1e18; }
-    // clang-format on
-    char valuestr[100];
-    VL_SNPRINTF(valuestr, 100, "%0.0f%s", value, suffixp);
-    return valuestr;  // Gets converted to string, so no ref to stack
-}
-
-#ifdef VL_TRACE_THREADED
-//=========================================================================
-// Buffer management
-
-template <> vluint32_t* VerilatedTrace<VL_DERIVED_T>::getTraceBuffer() {
-    vluint32_t* bufferp;
-    // Some jitter is expected, so some number of alternative trace buffers are
-    // required, but don't allocate more than 8 buffers.
-    if (m_numTraceBuffers < 8) {
-        // Allocate a new buffer if none is available
-        if (!m_buffersFromWorker.tryGet(bufferp)) {
-            ++m_numTraceBuffers;
-            // Note: over allocate a bit so pointer comparison is well defined
-            // if we overflow only by a small amount
-            bufferp = new vluint32_t[m_traceBufferSize + 16];
-        }
-    } else {
-        // Block until a buffer becomes available
-        bufferp = m_buffersFromWorker.get();
-    }
-    return bufferp;
-}
-
-template <> void VerilatedTrace<VL_DERIVED_T>::waitForBuffer(const vluint32_t* buffp) {
-    // Slow path code only called on flush/shutdown, so use a simple algorithm.
-    // Collect buffers from worker and stash them until we get the one we want.
-    std::deque<vluint32_t*> stash;
-    do { stash.push_back(m_buffersFromWorker.get()); } while (stash.back() != buffp);
-    // Now put them back in the queue, in the original order.
-    while (!stash.empty()) {
-        m_buffersFromWorker.put_front(stash.back());
-        stash.pop_back();
-    }
-}
-
-//=========================================================================
-// Worker thread
-
-template <> void VerilatedTrace<VL_DERIVED_T>::workerThreadMain() {
-    bool shutdown = false;
-
-    do {
-        vluint32_t* const bufferp = m_buffersToWorker.get();
-
-        VL_TRACE_THREAD_DEBUG("");
-        VL_TRACE_THREAD_DEBUG("Got buffer: " << bufferp);
-
-        const vluint32_t* readp = bufferp;
-
-        while (true) {
-            const vluint32_t cmd = readp[0];
-            const vluint32_t top = cmd >> 4;
-            // Always set this up, as it is almost always needed
-            vluint32_t* const oldp = m_sigs_oldvalp + readp[1];
-            // Note this increment needs to be undone on commands which do not
-            // actually contain a code, but those are the rare cases.
-            readp += 2;
-
-            switch (cmd & 0xF) {
-                //===
-                // CHG_* commands
-            case VerilatedTraceCommand::CHG_BIT_0:
-                VL_TRACE_THREAD_DEBUG("Command CHG_BIT_0 " << top);
-                chgBitImpl(oldp, 0);
-                continue;
-            case VerilatedTraceCommand::CHG_BIT_1:
-                VL_TRACE_THREAD_DEBUG("Command CHG_BIT_1 " << top);
-                chgBitImpl(oldp, 1);
-                continue;
-            case VerilatedTraceCommand::CHG_CDATA:
-                VL_TRACE_THREAD_DEBUG("Command CHG_CDATA " << top);
-                // Bits stored in bottom byte of command
-                chgCDataImpl(oldp, *readp, top);
-                readp += 1;
-                continue;
-            case VerilatedTraceCommand::CHG_SDATA:
-                VL_TRACE_THREAD_DEBUG("Command CHG_SDATA " << top);
-                // Bits stored in bottom byte of command
-                chgSDataImpl(oldp, *readp, top);
-                readp += 1;
-                continue;
-            case VerilatedTraceCommand::CHG_IDATA:
-                VL_TRACE_THREAD_DEBUG("Command CHG_IDATA " << top);
-                // Bits stored in bottom byte of command
-                chgIDataImpl(oldp, *readp, top);
-                readp += 1;
-                continue;
-            case VerilatedTraceCommand::CHG_QDATA:
-                VL_TRACE_THREAD_DEBUG("Command CHG_QDATA " << top);
-                // Bits stored in bottom byte of command
-                chgQDataImpl(oldp, *reinterpret_cast<const QData*>(readp), top);
-                readp += 2;
-                continue;
-            case VerilatedTraceCommand::CHG_WDATA:
-                VL_TRACE_THREAD_DEBUG("Command CHG_WDATA " << top);
-                chgWDataImpl(oldp, readp, top);
-                readp += VL_WORDS_I(top);
-                continue;
-            case VerilatedTraceCommand::CHG_DOUBLE:
-                VL_TRACE_THREAD_DEBUG("Command CHG_DOUBLE " << top);
-                chgDoubleImpl(oldp, *reinterpret_cast<const double*>(readp));
-                readp += 2;
-                continue;
-
-                //===
-                // Rare commands
-            case VerilatedTraceCommand::TIME_CHANGE:
-                VL_TRACE_THREAD_DEBUG("Command TIME_CHANGE " << top);
-                readp -= 1;  // No code in this command, undo increment
-                emitTimeChange(*reinterpret_cast<const vluint64_t*>(readp));
-                readp += 2;
-                continue;
-
-                //===
-                // Commands ending this buffer
-            case VerilatedTraceCommand::END: VL_TRACE_THREAD_DEBUG("Command END"); break;
-            case VerilatedTraceCommand::SHUTDOWN:
-                VL_TRACE_THREAD_DEBUG("Command SHUTDOWN");
-                shutdown = true;
-                break;
-
-            //===
-            // Unknown command
-            default: {  // LCOV_EXCL_START
-                VL_TRACE_THREAD_DEBUG("Command UNKNOWN");
-                VL_PRINTF_MT("Trace command: 0x%08x\n", cmd);
-                VL_FATAL_MT(__FILE__, __LINE__, "", "Unknown trace command");
-                break;
-            }  // LCOV_EXCL_STOP
-            }
-
-            // The above switch will execute 'continue' when necessary,
-            // so if we ever reach here, we are done with the buffer.
-            break;
-        }
-
-        VL_TRACE_THREAD_DEBUG("Returning buffer");
-
-        // Return buffer
-        m_buffersFromWorker.put(bufferp);
-    } while (VL_LIKELY(!shutdown));
-}
-
-template <> void VerilatedTrace<VL_DERIVED_T>::shutdownWorker() {
-    // If the worker thread is not running, done..
-    if (!m_workerThread) return;
-
-    // Hand an buffer with a shutdown command to the worker thread
-    vluint32_t* const bufferp = getTraceBuffer();
-    bufferp[0] = VerilatedTraceCommand::SHUTDOWN;
-    m_buffersToWorker.put(bufferp);
-    // Wait for it to return
-    waitForBuffer(bufferp);
-    // Join the thread and delete it
-    m_workerThread->join();
-    m_workerThread.reset(nullptr);
-}
-
-#endif
-
-//=============================================================================
-// Life cycle
-
-template <> void VerilatedTrace<VL_DERIVED_T>::closeBase() {
-#ifdef VL_TRACE_THREADED
-    shutdownWorker();
-    while (m_numTraceBuffers) {
-        delete[] m_buffersFromWorker.get();
-        --m_numTraceBuffers;
-    }
-#endif
-}
-
-template <> void VerilatedTrace<VL_DERIVED_T>::flushBase() {
-#ifdef VL_TRACE_THREADED
-    // Hand an empty buffer to the worker thread
-    vluint32_t* const bufferp = getTraceBuffer();
-    *bufferp = VerilatedTraceCommand::END;
-    m_buffersToWorker.put(bufferp);
-    // Wait for it to be returned. As the processing is in-order,
-    // this ensures all previous buffers have been processed.
-    waitForBuffer(bufferp);
-#endif
-}
-
-//=============================================================================
-// Callbacks to run on global events
-
-template <> void VerilatedTrace<VL_DERIVED_T>::onFlush(void* selfp) {
-    // This calls 'flush' on the derived class (which must then get any mutex)
-    reinterpret_cast<VL_DERIVED_T*>(selfp)->flush();
-}
-
-template <> void VerilatedTrace<VL_DERIVED_T>::onExit(void* selfp) {
-    // This calls 'close' on the derived class (which must then get any mutex)
-    reinterpret_cast<VL_DERIVED_T*>(selfp)->close();
-}
-
-//=============================================================================
-// VerilatedTrace
-
-template <>
-VerilatedTrace<VL_DERIVED_T>::VerilatedTrace()
-    : m_sigs_oldvalp{nullptr}
-    , m_timeLastDump{0}
-    , m_fullDump{true}
-    , m_nextCode{0}
-    , m_numSignals{0}
-    , m_maxBits{0}
-    , m_scopeEscape{'.'}
-    , m_timeRes{1e-9}
-    , m_timeUnit {
-    1e-9
-}
-#ifdef VL_TRACE_THREADED
-, m_numTraceBuffers { 0 }
-#endif
-{
-    set_time_unit(Verilated::threadContextp()->timeunitString());
-    set_time_resolution(Verilated::threadContextp()->timeprecisionString());
-}
-
-template <> VerilatedTrace<VL_DERIVED_T>::~VerilatedTrace() {
-    if (m_sigs_oldvalp) VL_DO_CLEAR(delete[] m_sigs_oldvalp, m_sigs_oldvalp = nullptr);
-    Verilated::removeFlushCb(VerilatedTrace<VL_DERIVED_T>::onFlush, this);
-    Verilated::removeExitCb(VerilatedTrace<VL_DERIVED_T>::onExit, this);
-#ifdef VL_TRACE_THREADED
-    closeBase();
-#endif
-}
-
-//=========================================================================
-// Internals available to format specific implementations
-
-template <> void VerilatedTrace<VL_DERIVED_T>::traceInit() VL_MT_UNSAFE {
-    // Note: It is possible to re-open a trace file (VCD in particular),
-    // so we must reset the next code here, but it must have the same number
-    // of codes on re-open
-    const vluint32_t expectedCodes = nextCode();
-    m_nextCode = 1;
-    m_numSignals = 0;
-    m_maxBits = 0;
-
-    // Call all initialize callbacks, which will:
-    // - Call decl* for each signal
-    // - Store the base code
-    for (vluint32_t i = 0; i < m_initCbs.size(); ++i) {
-        const CallbackRecord& cbr = m_initCbs[i];
-        cbr.m_initCb(cbr.m_userp, self(), nextCode());
-    }
-
-    if (expectedCodes && nextCode() != expectedCodes) {
-        VL_FATAL_MT(__FILE__, __LINE__, "",
-                    "Reopening trace file with different number of signals");
-    }
-
-    // Now that we know the number of codes, allocate space for the buffer
-    // holding previous signal values.
-    if (!m_sigs_oldvalp) m_sigs_oldvalp = new vluint32_t[nextCode()];
-
-    // Set callback so flush/abort will flush this file
-    Verilated::addFlushCb(VerilatedTrace<VL_DERIVED_T>::onFlush, this);
-    Verilated::addExitCb(VerilatedTrace<VL_DERIVED_T>::onExit, this);
-
-#ifdef VL_TRACE_THREADED
-    // Compute trace buffer size. we need to be able to store a new value for
-    // each signal, which is 'nextCode()' entries after the init callbacks
-    // above have been run, plus up to 2 more words of metadata per signal,
-    // plus fixed overhead of 1 for a termination flag and 3 for a time stamp
-    // update.
-    m_traceBufferSize = nextCode() + numSignals() * 2 + 4;
-
-    // Start the worker thread
-    m_workerThread.reset(new std::thread{&VerilatedTrace<VL_DERIVED_T>::workerThreadMain, this});
-#endif
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::declCode(vluint32_t code, vluint32_t bits, bool tri) {
-    if (VL_UNCOVERABLE(!code)) {
-        VL_FATAL_MT(__FILE__, __LINE__, "", "Internal: internal trace problem, code 0 is illegal");
-    }
-    // Note: The tri-state flag is not used by Verilator, but is here for
-    // compatibility with some foreign code.
-    int codesNeeded = VL_WORDS_I(bits);
-    if (tri) codesNeeded *= 2;
-    m_nextCode = std::max(m_nextCode, code + codesNeeded);
-    ++m_numSignals;
-    m_maxBits = std::max(m_maxBits, bits);
-}
-
-//=========================================================================
-// Internals available to format specific implementations
-
-template <> std::string VerilatedTrace<VL_DERIVED_T>::timeResStr() const {
-    return doubleToTimescale(m_timeRes);
-}
-
-//=========================================================================
-// External interface to client code
-
-template <> void VerilatedTrace<VL_DERIVED_T>::set_time_unit(const char* unitp) VL_MT_SAFE {
-    m_timeUnit = timescaleToDouble(unitp);
-}
-template <> void VerilatedTrace<VL_DERIVED_T>::set_time_unit(const std::string& unit) VL_MT_SAFE {
-    set_time_unit(unit.c_str());
-}
-template <> void VerilatedTrace<VL_DERIVED_T>::set_time_resolution(const char* unitp) VL_MT_SAFE {
-    m_timeRes = timescaleToDouble(unitp);
-}
-template <>
-void VerilatedTrace<VL_DERIVED_T>::set_time_resolution(const std::string& unit) VL_MT_SAFE {
-    set_time_resolution(unit.c_str());
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::dump(vluint64_t timeui) VL_MT_SAFE_EXCLUDES(m_mutex) {
-    // Not really VL_MT_SAFE but more VL_MT_UNSAFE_ONE.
-    // This does get the mutex, but if multiple threads are trying to dump
-    // chances are the data being dumped will have other problems
-    const VerilatedLockGuard lock{m_mutex};
-    if (VL_UNCOVERABLE(m_timeLastDump && timeui <= m_timeLastDump)) {  // LCOV_EXCL_START
-        VL_PRINTF_MT("%%Warning: previous dump at t=%" VL_PRI64 "u, requesting t=%" VL_PRI64
-                     "u, dump call ignored\n",
-                     m_timeLastDump, timeui);
-        return;
-    }  // LCOV_EXCL_STOP
-    m_timeLastDump = timeui;
-
-    Verilated::quiesce();
-
-    // Call hook for format specific behaviour
-    if (VL_UNLIKELY(m_fullDump)) {
-        if (!preFullDump()) return;
-    } else {
-        if (!preChangeDump()) return;
-    }
-
-#ifdef VL_TRACE_THREADED
-    // Currently only incremental dumps run on the worker thread
-    vluint32_t* bufferp = nullptr;
-    if (VL_LIKELY(!m_fullDump)) {
-        // Get the trace buffer we are about to fill
-        bufferp = getTraceBuffer();
-        m_traceBufferWritep = bufferp;
-        m_traceBufferEndp = bufferp + m_traceBufferSize;
-
-        // Tell worker to update time point
-        m_traceBufferWritep[0] = VerilatedTraceCommand::TIME_CHANGE;
-        *reinterpret_cast<vluint64_t*>(m_traceBufferWritep + 1) = timeui;
-        m_traceBufferWritep += 3;
-    } else {
-        // Update time point
-        flushBase();
-        emitTimeChange(timeui);
-    }
-#else
-    // Update time point
-    emitTimeChange(timeui);
-#endif
-
-    // Run the callbacks
-    if (VL_UNLIKELY(m_fullDump)) {
-        m_fullDump = false;  // No more need for next dump to be full
-        for (vluint32_t i = 0; i < m_fullCbs.size(); ++i) {
-            const CallbackRecord& cbr = m_fullCbs[i];
-            cbr.m_dumpCb(cbr.m_userp, self());
-        }
-    } else {
-        for (vluint32_t i = 0; i < m_chgCbs.size(); ++i) {
-            const CallbackRecord& cbr = m_chgCbs[i];
-            cbr.m_dumpCb(cbr.m_userp, self());
-        }
-    }
-
-    for (vluint32_t i = 0; i < m_cleanupCbs.size(); ++i) {
-        const CallbackRecord& cbr = m_cleanupCbs[i];
-        cbr.m_dumpCb(cbr.m_userp, self());
-    }
-
-#ifdef VL_TRACE_THREADED
-    if (VL_LIKELY(bufferp)) {
-        // Mark end of the trace buffer we just filled
-        *m_traceBufferWritep++ = VerilatedTraceCommand::END;
-
-        // Assert no buffer overflow
-        assert(m_traceBufferWritep - bufferp <= m_traceBufferSize);
-
-        // Pass it to the worker thread
-        m_buffersToWorker.put(bufferp);
-    }
-#endif
-}
-
-//=============================================================================
-// Non-hot path internal interface to Verilator generated code
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::addCallbackRecord(std::vector<CallbackRecord>& cbVec,
-                                                     CallbackRecord& cbRec)
-    VL_MT_SAFE_EXCLUDES(m_mutex) {
-    const VerilatedLockGuard lock{m_mutex};
-    if (VL_UNCOVERABLE(timeLastDump() != 0)) {  // LCOV_EXCL_START
-        const std::string msg = (std::string{"Internal: "} + __FILE__ + "::" + __FUNCTION__
-                                 + " called with already open file");
-        VL_FATAL_MT(__FILE__, __LINE__, "", msg.c_str());
-    }  // LCOV_EXCL_STOP
-    cbVec.push_back(cbRec);
-}
-
-template <> void VerilatedTrace<VL_DERIVED_T>::addInitCb(initCb_t cb, void* userp) VL_MT_SAFE {
-    CallbackRecord cbr{cb, userp};
-    addCallbackRecord(m_initCbs, cbr);
-}
-template <> void VerilatedTrace<VL_DERIVED_T>::addFullCb(dumpCb_t cb, void* userp) VL_MT_SAFE {
-    CallbackRecord cbr{cb, userp};
-    addCallbackRecord(m_fullCbs, cbr);
-}
-template <> void VerilatedTrace<VL_DERIVED_T>::addChgCb(dumpCb_t cb, void* userp) VL_MT_SAFE {
-    CallbackRecord cbr{cb, userp};
-    addCallbackRecord(m_chgCbs, cbr);
-}
-template <> void VerilatedTrace<VL_DERIVED_T>::addCleanupCb(dumpCb_t cb, void* userp) VL_MT_SAFE {
-    CallbackRecord cbr{cb, userp};
-    addCallbackRecord(m_cleanupCbs, cbr);
-}
-template <> void VerilatedTrace<VL_DERIVED_T>::module(const std::string& name) VL_MT_UNSAFE {
-    // Called via callbacks way above in call stack, which already hold m_mutex
-    m_moduleName = name;
-}
-
-//=========================================================================
-// Hot path internal interface to Verilator generated code
-
-// These functions must write the new value back into the old value store,
-// and subsequently call the format specific emit* implementations. Note
-// that this file must be included in the format specific implementation, so
-// the emit* functions can be inlined for performance.
-
-template <> void VerilatedTrace<VL_DERIVED_T>::fullBit(vluint32_t* oldp, CData newval) {
-    *oldp = newval;
-    self()->emitBit(oldp - m_sigs_oldvalp, newval);
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::fullCData(vluint32_t* oldp, CData newval, int bits) {
-    *oldp = newval;
-    self()->emitCData(oldp - m_sigs_oldvalp, newval, bits);
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::fullSData(vluint32_t* oldp, SData newval, int bits) {
-    *oldp = newval;
-    self()->emitSData(oldp - m_sigs_oldvalp, newval, bits);
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::fullIData(vluint32_t* oldp, IData newval, int bits) {
-    *oldp = newval;
-    self()->emitIData(oldp - m_sigs_oldvalp, newval, bits);
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::fullQData(vluint32_t* oldp, QData newval, int bits) {
-    *reinterpret_cast<QData*>(oldp) = newval;
-    self()->emitQData(oldp - m_sigs_oldvalp, newval, bits);
-}
-
-template <>
-void VerilatedTrace<VL_DERIVED_T>::fullWData(vluint32_t* oldp, const WData* newvalp, int bits) {
-    for (int i = 0; i < VL_WORDS_I(bits); ++i) oldp[i] = newvalp[i];
-    self()->emitWData(oldp - m_sigs_oldvalp, newvalp, bits);
-}
-
-template <> void VerilatedTrace<VL_DERIVED_T>::fullDouble(vluint32_t* oldp, double newval) {
-    // cppcheck-suppress invalidPointerCast
-    *reinterpret_cast<double*>(oldp) = newval;
-    self()->emitDouble(oldp - m_sigs_oldvalp, newval);
-}
-
-//=========================================================================
-// Primitives converting binary values to strings...
-
-// All of these take a destination pointer where the string will be emitted,
-// and a value to convert. There are a couple of variants for efficiency.
-
-static inline void cvtCDataToStr(char* dstp, CData value) {
-#ifdef VL_HAVE_SSE2
-    // Similar to cvtSDataToStr but only the bottom 8 byte lanes are used
-    const __m128i a = _mm_cvtsi32_si128(value);
-    const __m128i b = _mm_unpacklo_epi8(a, a);
-    const __m128i c = _mm_shufflelo_epi16(b, 0);
-    const __m128i m = _mm_set1_epi64x(0x0102040810204080);
-    const __m128i d = _mm_cmpeq_epi8(_mm_and_si128(c, m), m);
-    const __m128i result = _mm_sub_epi8(_mm_set1_epi8('0'), d);
-    _mm_storel_epi64(reinterpret_cast<__m128i*>(dstp), result);
-#else
-    dstp[0] = '0' | static_cast<char>((value >> 7) & 1);
-    dstp[1] = '0' | static_cast<char>((value >> 6) & 1);
-    dstp[2] = '0' | static_cast<char>((value >> 5) & 1);
-    dstp[3] = '0' | static_cast<char>((value >> 4) & 1);
-    dstp[4] = '0' | static_cast<char>((value >> 3) & 1);
-    dstp[5] = '0' | static_cast<char>((value >> 2) & 1);
-    dstp[6] = '0' | static_cast<char>((value >> 1) & 1);
-    dstp[7] = '0' | static_cast<char>(value & 1);
-#endif
-}
-
-static inline void cvtSDataToStr(char* dstp, SData value) {
-#ifdef VL_HAVE_SSE2
-    // We want each bit in the 16-bit input value to end up in a byte lane
-    // within the 128-bit XMM register. Note that x86 is little-endian and we
-    // want the MSB of the input at the low address, so we will bit-reverse
-    // at the same time.
-
-    // Put value in bottom of 128-bit register a[15:0] = value
-    const __m128i a = _mm_cvtsi32_si128(value);
-    // Interleave bytes with themselves
-    // b[15: 0] = {2{a[ 7:0]}} == {2{value[ 7:0]}}
-    // b[31:16] = {2{a[15:8]}} == {2{value[15:8]}}
-    const __m128i b = _mm_unpacklo_epi8(a, a);
-    // Shuffle bottom 64 bits, note swapping high bytes with low bytes
-    // c[31: 0] = {2{b[31:16]}} == {4{value[15:8}}
-    // c[63:32] = {2{b[15: 0]}} == {4{value[ 7:0}}
-    const __m128i c = _mm_shufflelo_epi16(b, 0x05);
-    // Shuffle whole register
-    // d[ 63: 0] = {2{c[31: 0]}} == {8{value[15:8}}
-    // d[126:54] = {2{c[63:32]}} == {8{value[ 7:0}}
-    const __m128i d = _mm_shuffle_epi32(c, 0x50);
-    // Test each bit within the bytes, this sets each byte lane to 0
-    // if the bit for that lane is 0 and to 0xff if the bit is 1.
-    const __m128i m = _mm_set1_epi64x(0x0102040810204080);
-    const __m128i e = _mm_cmpeq_epi8(_mm_and_si128(d, m), m);
-    // Convert to ASCII by subtracting the masks from ASCII '0':
-    // '0' - 0 is '0', '0' - -1 is '1'
-    const __m128i result = _mm_sub_epi8(_mm_set1_epi8('0'), e);
-    // Store the 16 characters to the un-aligned buffer
-    _mm_storeu_si128(reinterpret_cast<__m128i*>(dstp), result);
-#else
-    cvtCDataToStr(dstp, value >> 8);
-    cvtCDataToStr(dstp + 8, value);
-#endif
-}
-
-static inline void cvtIDataToStr(char* dstp, IData value) {
-#ifdef VL_HAVE_AVX2
-    // Similar to cvtSDataToStr but the bottom 16-bits are processed in the
-    // top half of the YMM registerss
-    const __m256i a = _mm256_insert_epi32(_mm256_undefined_si256(), value, 0);
-    const __m256i b = _mm256_permute4x64_epi64(a, 0);
-    const __m256i s = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2,
-                                      2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3);
-    const __m256i c = _mm256_shuffle_epi8(b, s);
-    const __m256i m = _mm256_set1_epi64x(0x0102040810204080);
-    const __m256i d = _mm256_cmpeq_epi8(_mm256_and_si256(c, m), m);
-    const __m256i result = _mm256_sub_epi8(_mm256_set1_epi8('0'), d);
-    _mm256_storeu_si256(reinterpret_cast<__m256i*>(dstp), result);
-#else
-    cvtSDataToStr(dstp, value >> 16);
-    cvtSDataToStr(dstp + 16, value);
-#endif
-}
-
-static inline void cvtQDataToStr(char* dstp, QData value) {
-    cvtIDataToStr(dstp, value >> 32);
-    cvtIDataToStr(dstp + 32, value);
-}
-
-#define cvtEDataToStr cvtIDataToStr
-
-//=============================================================================
-
-#ifdef VERILATED_VCD_TEST
-
-void verilated_trace_imp_selftest() {
-#define SELF_CHECK(got, exp) \
-    do { \
-        if ((got) != (exp)) VL_FATAL_MT(__FILE__, __LINE__, "", "%Error: selftest\n"); \
-    } while (0)
-
-#define SELF_CHECK_TS(scale) \
-    SELF_CHECK(doubleToTimescale(timescaleToDouble(scale)), std::string{scale});
-    SELF_CHECK_TS("100s");
-    SELF_CHECK_TS("10s");
-    SELF_CHECK_TS("1s");
-    SELF_CHECK_TS("100ms");
-    SELF_CHECK_TS("10ms");
-    SELF_CHECK_TS("1ms");
-    SELF_CHECK_TS("100us");
-    SELF_CHECK_TS("10us");
-    SELF_CHECK_TS("1us");
-    SELF_CHECK_TS("100ns");
-    SELF_CHECK_TS("10ns");
-    SELF_CHECK_TS("1ns");
-    SELF_CHECK_TS("100ps");
-    SELF_CHECK_TS("10ps");
-    SELF_CHECK_TS("1ps");
-    SELF_CHECK_TS("100fs");
-    SELF_CHECK_TS("10fs");
-    SELF_CHECK_TS("1fs");
-    SELF_CHECK_TS("100as");
-    SELF_CHECK_TS("10as");
-    SELF_CHECK_TS("1as");
-}
-
-#endif
-
-#endif  // VL_CPPCHECK
@@ -0,0 +1,838 @@
+// -*- mode: C++; c-file-style: "cc-mode" -*-
+//=============================================================================
+//
+// Code available from: https://verilator.org
+//
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
+// can redistribute it and/or modify it under the terms of either the GNU
+// Lesser General Public License Version 3 or the Perl Artistic License
+// Version 2.0.
+// SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0
+//
+//=============================================================================
+//
+// Verilated tracing implementation code template common to all formats.
+// This file is included by the format specific implementations and
+// should not be used otherwise.
+//
+//=============================================================================
+
+// clang-format off
+
+#ifndef VL_CPPCHECK
+#if !defined(VL_SUB_T) || !defined(VL_BUF_T)
+# error "This file should be included in trace format implementations"
+#endif
+
+#include "verilated_intrinsics.h"
+#include "verilated_trace.h"
+#ifdef VL_TRACE_PARALLEL
+# include "verilated_threads.h"
+# include <list>
+#endif
+
+#if 0
+# include <iostream>
+# define VL_TRACE_OFFLOAD_DEBUG(msg) std::cout << "TRACE OFFLOAD THREAD: " << msg << std::endl
+#else
+# define VL_TRACE_OFFLOAD_DEBUG(msg)
+#endif
+
+// clang-format on
+
+//=============================================================================
+// Static utility functions
+
+static double timescaleToDouble(const char* unitp) {
+    char* endp = nullptr;
+    double value = std::strtod(unitp, &endp);
+    // On error so we allow just "ns" to return 1e-9.
+    if (value == 0.0 && endp == unitp) value = 1;
+    unitp = endp;
+    for (; *unitp && std::isspace(*unitp); unitp++) {}
+    switch (*unitp) {
+    case 's': value *= 1e0; break;
+    case 'm': value *= 1e-3; break;
+    case 'u': value *= 1e-6; break;
+    case 'n': value *= 1e-9; break;
+    case 'p': value *= 1e-12; break;
+    case 'f': value *= 1e-15; break;
+    case 'a': value *= 1e-18; break;
+    }
+    return value;
+}
+
+static std::string doubleToTimescale(double value) {
+    const char* suffixp = "s";
+    // clang-format off
+    if      (value >= 1e0)   { suffixp = "s"; value *= 1e0; }
+    else if (value >= 1e-3 ) { suffixp = "ms"; value *= 1e3; }
+    else if (value >= 1e-6 ) { suffixp = "us"; value *= 1e6; }
+    else if (value >= 1e-9 ) { suffixp = "ns"; value *= 1e9; }
+    else if (value >= 1e-12) { suffixp = "ps"; value *= 1e12; }
+    else if (value >= 1e-15) { suffixp = "fs"; value *= 1e15; }
+    else if (value >= 1e-18) { suffixp = "as"; value *= 1e18; }
+    // clang-format on
+    char valuestr[100];
+    VL_SNPRINTF(valuestr, 100, "%0.0f%s", value, suffixp);
+    return valuestr;  // Gets converted to string, so no ref to stack
+}
+
+#ifdef VL_TRACE_OFFLOAD
+//=========================================================================
+// Buffer management
+
+template <> uint32_t* VerilatedTrace<VL_SUB_T, VL_BUF_T>::getOffloadBuffer() {
+    uint32_t* bufferp;
+    // Some jitter is expected, so some number of alternative offlaod buffers are
+    // required, but don't allocate more than 8 buffers.
+    if (m_numOffloadBuffers < 8) {
+        // Allocate a new buffer if none is available
+        if (!m_offloadBuffersFromWorker.tryGet(bufferp)) {
+            ++m_numOffloadBuffers;
+            // Note: over allocate a bit so pointer comparison is well defined
+            // if we overflow only by a small amount
+            bufferp = new uint32_t[m_offloadBufferSize + 16];
+        }
+    } else {
+        // Block until a buffer becomes available
+        bufferp = m_offloadBuffersFromWorker.get();
+    }
+    return bufferp;
+}
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::waitForOffloadBuffer(const uint32_t* buffp) {
+    // Slow path code only called on flush/shutdown, so use a simple algorithm.
+    // Collect buffers from worker and stash them until we get the one we want.
+    std::deque<uint32_t*> stash;
+    do { stash.push_back(m_offloadBuffersFromWorker.get()); } while (stash.back() != buffp);
+    // Now put them back in the queue, in the original order.
+    while (!stash.empty()) {
+        m_offloadBuffersFromWorker.put_front(stash.back());
+        stash.pop_back();
+    }
+}
+
+//=========================================================================
+// Worker thread
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::offloadWorkerThreadMain() {
+    bool shutdown = false;
+
+    do {
+        uint32_t* const bufferp = m_offloadBuffersToWorker.get();
+
+        VL_TRACE_OFFLOAD_DEBUG("");
+        VL_TRACE_OFFLOAD_DEBUG("Got buffer: " << bufferp);
+
+        const uint32_t* readp = bufferp;
+
+        std::unique_ptr<VL_BUF_T> traceBufp;  // We own the passed tracebuffer
+
+        while (true) {
+            const uint32_t cmd = readp[0];
+            const uint32_t top = cmd >> 4;
+            // Always set this up, as it is almost always needed
+            uint32_t* const oldp = m_sigs_oldvalp + readp[1];
+            // Note this increment needs to be undone on commands which do not
+            // actually contain a code, but those are the rare cases.
+            readp += 2;
+
+            switch (cmd & 0xF) {
+                //===
+                // CHG_* commands
+            case VerilatedTraceOffloadCommand::CHG_BIT_0:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_BIT_0 " << top);
+                traceBufp->chgBitImpl(oldp, 0);
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_BIT_1:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_BIT_1 " << top);
+                traceBufp->chgBitImpl(oldp, 1);
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_CDATA:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_CDATA " << top);
+                // Bits stored in bottom byte of command
+                traceBufp->chgCDataImpl(oldp, *readp, top);
+                readp += 1;
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_SDATA:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_SDATA " << top);
+                // Bits stored in bottom byte of command
+                traceBufp->chgSDataImpl(oldp, *readp, top);
+                readp += 1;
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_IDATA:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_IDATA " << top);
+                // Bits stored in bottom byte of command
+                traceBufp->chgIDataImpl(oldp, *readp, top);
+                readp += 1;
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_QDATA:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_QDATA " << top);
+                // Bits stored in bottom byte of command
+                traceBufp->chgQDataImpl(oldp, *reinterpret_cast<const QData*>(readp), top);
+                readp += 2;
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_WDATA:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_WDATA " << top);
+                traceBufp->chgWDataImpl(oldp, readp, top);
+                readp += VL_WORDS_I(top);
+                continue;
+            case VerilatedTraceOffloadCommand::CHG_DOUBLE:
+                VL_TRACE_OFFLOAD_DEBUG("Command CHG_DOUBLE " << top);
+                traceBufp->chgDoubleImpl(oldp, *reinterpret_cast<const double*>(readp));
+                readp += 2;
+                continue;
+
+                //===
+                // Rare commands
+            case VerilatedTraceOffloadCommand::TIME_CHANGE:
+                VL_TRACE_OFFLOAD_DEBUG("Command TIME_CHANGE " << top);
+                readp -= 1;  // No code in this command, undo increment
+                emitTimeChange(*reinterpret_cast<const uint64_t*>(readp));
+                readp += 2;
+                continue;
+
+            case VerilatedTraceOffloadCommand::TRACE_BUFFER:
+                VL_TRACE_OFFLOAD_DEBUG("Command TRACE_BUFFER " << top);
+                readp -= 1;  // No code in this command, undo increment
+                traceBufp.reset(*reinterpret_cast<VL_BUF_T* const*>(readp));
+                readp += 2;
+                continue;
+
+                //===
+                // Commands ending this buffer
+            case VerilatedTraceOffloadCommand::END:  //
+                VL_TRACE_OFFLOAD_DEBUG("Command END");
+                break;
+            case VerilatedTraceOffloadCommand::SHUTDOWN:
+                VL_TRACE_OFFLOAD_DEBUG("Command SHUTDOWN");
+                shutdown = true;
+                break;
+
+            //===
+            // Unknown command
+            default: {  // LCOV_EXCL_START
+                VL_TRACE_OFFLOAD_DEBUG("Command UNKNOWN " << cmd);
+                VL_FATAL_MT(__FILE__, __LINE__, "", "Unknown trace command");
+                break;
+            }  // LCOV_EXCL_STOP
+            }
+
+            // The above switch will execute 'continue' when necessary,
+            // so if we ever reach here, we are done with the buffer.
+            break;
+        }
+
+        VL_TRACE_OFFLOAD_DEBUG("Returning buffer");
+
+        // Return buffer
+        m_offloadBuffersFromWorker.put(bufferp);
+    } while (VL_LIKELY(!shutdown));
+}
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::shutdownOffloadWorker() {
+    // If the worker thread is not running, done..
+    if (!m_workerThread) return;
+
+    // Hand an buffer with a shutdown command to the worker thread
+    uint32_t* const bufferp = getOffloadBuffer();
+    bufferp[0] = VerilatedTraceOffloadCommand::SHUTDOWN;
+    m_offloadBuffersToWorker.put(bufferp);
+    // Wait for it to return
+    waitForOffloadBuffer(bufferp);
+    // Join the thread and delete it
+    m_workerThread->join();
+    m_workerThread.reset(nullptr);
+}
+
+#endif
+
+//=============================================================================
+// Life cycle
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::closeBase() {
+#ifdef VL_TRACE_OFFLOAD
+    shutdownOffloadWorker();
+    while (m_numOffloadBuffers) {
+        delete[] m_offloadBuffersFromWorker.get();
+        --m_numOffloadBuffers;
+    }
+#endif
+}
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::flushBase() {
+#ifdef VL_TRACE_OFFLOAD
+    // Hand an empty buffer to the worker thread
+    uint32_t* const bufferp = getOffloadBuffer();
+    *bufferp = VerilatedTraceOffloadCommand::END;
+    m_offloadBuffersToWorker.put(bufferp);
+    // Wait for it to be returned. As the processing is in-order,
+    // this ensures all previous buffers have been processed.
+    waitForOffloadBuffer(bufferp);
+#endif
+}
+
+//=============================================================================
+// Callbacks to run on global events
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::onFlush(void* selfp) {
+    // This calls 'flush' on the derived class (which must then get any mutex)
+    reinterpret_cast<VL_SUB_T*>(selfp)->flush();
+}
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::onExit(void* selfp) {
+    // This calls 'close' on the derived class (which must then get any mutex)
+    reinterpret_cast<VL_SUB_T*>(selfp)->close();
+}
+
+//=============================================================================
+// VerilatedTrace
+
+template <> VerilatedTrace<VL_SUB_T, VL_BUF_T>::VerilatedTrace() {
+    set_time_unit(Verilated::threadContextp()->timeunitString());
+    set_time_resolution(Verilated::threadContextp()->timeprecisionString());
+}
+
+template <> VerilatedTrace<VL_SUB_T, VL_BUF_T>::~VerilatedTrace() {
+    if (m_sigs_oldvalp) VL_DO_CLEAR(delete[] m_sigs_oldvalp, m_sigs_oldvalp = nullptr);
+    if (m_sigs_enabledp) VL_DO_CLEAR(delete[] m_sigs_enabledp, m_sigs_enabledp = nullptr);
+    Verilated::removeFlushCb(VerilatedTrace<VL_SUB_T, VL_BUF_T>::onFlush, this);
+    Verilated::removeExitCb(VerilatedTrace<VL_SUB_T, VL_BUF_T>::onExit, this);
+#ifdef VL_TRACE_OFFLOAD
+    closeBase();
+#endif
+}
+
+//=========================================================================
+// Internals available to format specific implementations
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::traceInit() VL_MT_UNSAFE {
+    // Note: It is possible to re-open a trace file (VCD in particular),
+    // so we must reset the next code here, but it must have the same number
+    // of codes on re-open
+    const uint32_t expectedCodes = nextCode();
+    m_nextCode = 1;
+    m_numSignals = 0;
+    m_maxBits = 0;
+    m_sigs_enabledVec.clear();
+
+    // Call all initialize callbacks, which will:
+    // - Call decl* for each signal (these eventually call ::declCode)
+    // - Store the base code
+    for (uint32_t i = 0; i < m_initCbs.size(); ++i) {
+        const CallbackRecord& cbr = m_initCbs[i];
+        cbr.m_initCb(cbr.m_userp, self(), nextCode());
+    }
+
+    if (expectedCodes && nextCode() != expectedCodes) {
+        VL_FATAL_MT(__FILE__, __LINE__, "",
+                    "Reopening trace file with different number of signals");
+    }
+
+    // Now that we know the number of codes, allocate space for the buffer
+    // holding previous signal values.
+    if (!m_sigs_oldvalp) m_sigs_oldvalp = new uint32_t[nextCode()];
+
+    // Apply enables
+    if (m_sigs_enabledp) VL_DO_CLEAR(delete[] m_sigs_enabledp, m_sigs_enabledp = nullptr);
+    if (!m_sigs_enabledVec.empty()) {
+        // Else if was empty, m_sigs_enabledp = nullptr to short circuit tests
+        // But it isn't, so alloc one bit for each code to indicate enablement
+        // We don't want to still use m_signs_enabledVec as std::vector<bool> is not
+        // guarenteed to be fast
+        m_sigs_enabledp = new uint32_t[1 + VL_WORDS_I(nextCode())]{0};
+        m_sigs_enabledVec.reserve(nextCode());
+        for (size_t code = 0; code < nextCode(); ++code) {
+            if (m_sigs_enabledVec[code]) {
+                m_sigs_enabledp[VL_BITWORD_I(code)] |= 1U << VL_BITBIT_I(code);
+            }
+        }
+        m_sigs_enabledVec.clear();
+    }
+
+    // Set callback so flush/abort will flush this file
+    Verilated::addFlushCb(VerilatedTrace<VL_SUB_T, VL_BUF_T>::onFlush, this);
+    Verilated::addExitCb(VerilatedTrace<VL_SUB_T, VL_BUF_T>::onExit, this);
+
+#ifdef VL_TRACE_OFFLOAD
+    // Compute offload buffer size. we need to be able to store a new value for
+    // each signal, which is 'nextCode()' entries after the init callbacks
+    // above have been run, plus up to 2 more words of metadata per signal,
+    // plus fixed overhead of 1 for a termination flag and 3 for a time stamp
+    // update.
+    m_offloadBufferSize = nextCode() + numSignals() * 2 + 4;
+
+    // Start the worker thread
+    m_workerThread.reset(
+        new std::thread{&VerilatedTrace<VL_SUB_T, VL_BUF_T>::offloadWorkerThreadMain, this});
+#endif
+}
+
+template <>
+bool VerilatedTrace<VL_SUB_T, VL_BUF_T>::declCode(uint32_t code, const char* namep, uint32_t bits,
+                                                  bool tri) {
+    if (VL_UNCOVERABLE(!code)) {
+        VL_FATAL_MT(__FILE__, __LINE__, "", "Internal: internal trace problem, code 0 is illegal");
+    }
+    // To keep it simple, this is O(enables * signals), but we expect few enables
+    std::string declName = namePrefix() + namep;
+    bool enabled = false;
+    if (m_dumpvars.empty()) enabled = true;
+    for (const auto& item : m_dumpvars) {
+        const int dumpvarsLevel = item.first;
+        const char* dvp = item.second.c_str();
+        const char* np = declName.c_str();
+        while (*dvp && *dvp == *np) {
+            ++dvp;
+            ++np;
+        }
+        if (*dvp) continue;  // Didn't match dumpvar item
+        if (*np && *np != ' ') continue;  // e.g. "t" isn't a match for "top"
+        int levels = 0;
+        while (*np) {
+            if (*np++ == ' ') ++levels;
+        }
+        if (levels > dumpvarsLevel) continue;  // Too deep
+        // We only need to set first code word if it's a multicode signal
+        // as that's all we'll check for later
+        if (m_sigs_enabledVec.size() <= code) m_sigs_enabledVec.resize((code + 1024) * 2);
+        m_sigs_enabledVec[code] = true;
+        enabled = true;
+        break;
+    }
+
+    // Note: The tri-state flag is not used by Verilator, but is here for
+    // compatibility with some foreign code.
+    int codesNeeded = VL_WORDS_I(bits);
+    if (tri) codesNeeded *= 2;
+    m_nextCode = std::max(m_nextCode, code + codesNeeded);
+    ++m_numSignals;
+    m_maxBits = std::max(m_maxBits, bits);
+    return enabled;
+}
+
+//=========================================================================
+// Internals available to format specific implementations
+
+template <> std::string VerilatedTrace<VL_SUB_T, VL_BUF_T>::timeResStr() const {
+    return doubleToTimescale(m_timeRes);
+}
+
+//=========================================================================
+// External interface to client code
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::set_time_unit(const char* unitp) VL_MT_SAFE {
+    m_timeUnit = timescaleToDouble(unitp);
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::set_time_unit(const std::string& unit) VL_MT_SAFE {
+    set_time_unit(unit.c_str());
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::set_time_resolution(const char* unitp) VL_MT_SAFE {
+    m_timeRes = timescaleToDouble(unitp);
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::set_time_resolution(const std::string& unit) VL_MT_SAFE {
+    set_time_resolution(unit.c_str());
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::dumpvars(int level, const std::string& hier) VL_MT_SAFE {
+    if (level == 0) {
+        m_dumpvars.clear();  // empty = everything on
+    } else {
+        // Convert Verilog . separators to trace space separators
+        std::string hierSpaced = hier;
+        for (auto& i : hierSpaced) {
+            if (i == '.') i = ' ';
+        }
+        m_dumpvars.push_back(std::make_pair(level, hierSpaced));
+    }
+}
+
+#ifdef VL_TRACE_PARALLEL
+template <>  //
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::parallelWorkerTask(void* datap, bool) {
+    ParallelWorkerData* const wdp = reinterpret_cast<ParallelWorkerData*>(datap);
+    // Run the task
+    wdp->m_cb(wdp->m_userp, wdp->m_bufp);
+    // Mark buffer as ready
+    const VerilatedLockGuard lock{wdp->m_mutex};
+    wdp->m_ready.store(true);
+    if (wdp->m_waiting) wdp->m_cv.notify_one();
+}
+
+template <> VL_ATTR_NOINLINE void VerilatedTrace<VL_SUB_T, VL_BUF_T>::ParallelWorkerData::wait() {
+    // Spin for a while, waiting for the buffer to become ready
+    for (int i = 0; i < VL_LOCK_SPINS; ++i) {
+        if (VL_LIKELY(m_ready.load(std::memory_order_relaxed))) return;
+        VL_CPU_RELAX();
+    }
+    // We have been spinning for a while, so yield the thread
+    VerilatedLockGuard lock{m_mutex};
+    m_waiting = true;
+    m_cv.wait(lock, [this] { return m_ready.load(std::memory_order_relaxed); });
+    m_waiting = false;
+}
+#endif
+
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::runParallelCallbacks(const ParallelCallbackMap& cbMap) {
+    for (VlThreadPool* threadPoolp : m_threadPoolps) {
+#ifdef VL_TRACE_PARALLEL
+        // If tracing in parallel, dispatch to the thread pool (if exists)
+        if (threadPoolp && threadPoolp->numThreads()) {
+            // List of work items for thread (std::list, as ParallelWorkerData is not movable)
+            std::list<ParallelWorkerData> workerData;
+            // We use the whole pool + the main thread
+            const unsigned threads = threadPoolp->numThreads() + 1;
+            // Main thread executes all jobs with index % threads == 0
+            std::vector<ParallelWorkerData*> mainThreadWorkerData;
+            // The tracing callbacks to execute on this thread-pool
+            const auto& cbVec = cbMap.at(threadPoolp);
+            // Enuque all the jobs
+            for (unsigned i = 0; i < cbVec.size(); ++i) {
+                const CallbackRecord& cbr = cbVec[i];
+                // Always get the trace buffer on the main thread
+                Buffer* const bufp = getTraceBuffer();
+                // Create new work item
+                workerData.emplace_back(cbr.m_dumpCb, cbr.m_userp, bufp);
+                // Grab the new work item
+                ParallelWorkerData* const itemp = &workerData.back();
+                // Enqueue task to thread pool, or main thread
+                if (unsigned rem = i % threads) {
+                    threadPoolp->workerp(rem - 1)->addTask(parallelWorkerTask, itemp, false);
+                } else {
+                    mainThreadWorkerData.push_back(itemp);
+                }
+            }
+            // Execute main thead jobs
+            for (ParallelWorkerData* const itemp : mainThreadWorkerData) {
+                parallelWorkerTask(itemp, false);
+            }
+            // Commit all trace buffers in order
+            for (ParallelWorkerData& item : workerData) {
+                // Wait until ready
+                item.wait();
+                // Commit the buffer
+                commitTraceBuffer(item.m_bufp);
+            }
+            continue;
+        }
+#endif
+        // Fall back on sequential execution
+        for (const CallbackRecord& cbr : cbMap.at(threadPoolp)) {
+            Buffer* const traceBufferp = getTraceBuffer();
+            cbr.m_dumpCb(cbr.m_userp, traceBufferp);
+            commitTraceBuffer(traceBufferp);
+        }
+    }
+}
+
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::dump(uint64_t timeui) VL_MT_SAFE_EXCLUDES(m_mutex) {
+    // Not really VL_MT_SAFE but more VL_MT_UNSAFE_ONE.
+    // This does get the mutex, but if multiple threads are trying to dump
+    // chances are the data being dumped will have other problems
+    const VerilatedLockGuard lock{m_mutex};
+    if (VL_UNCOVERABLE(m_timeLastDump && timeui <= m_timeLastDump)) {  // LCOV_EXCL_START
+        VL_PRINTF_MT("%%Warning: previous dump at t=%" PRIu64 ", requesting t=%" PRIu64
+                     ", dump call ignored\n",
+                     m_timeLastDump, timeui);
+        return;
+    }  // LCOV_EXCL_STOP
+    m_timeLastDump = timeui;
+
+    Verilated::quiesce();
+
+    // Call hook for format specific behaviour
+    if (VL_UNLIKELY(m_fullDump)) {
+        if (!preFullDump()) return;
+    } else {
+        if (!preChangeDump()) return;
+    }
+
+#ifdef VL_TRACE_OFFLOAD
+    // Currently only incremental dumps run on the worker thread
+    uint32_t* bufferp = nullptr;
+    if (VL_LIKELY(!m_fullDump)) {
+        // Get the offload buffer we are about to fill
+        bufferp = getOffloadBuffer();
+        m_offloadBufferWritep = bufferp;
+        m_offloadBufferEndp = bufferp + m_offloadBufferSize;
+
+        // Tell worker to update time point
+        m_offloadBufferWritep[0] = VerilatedTraceOffloadCommand::TIME_CHANGE;
+        *reinterpret_cast<uint64_t*>(m_offloadBufferWritep + 1) = timeui;
+        m_offloadBufferWritep += 3;
+    } else {
+        // Update time point
+        flushBase();
+        emitTimeChange(timeui);
+    }
+#else
+    // Update time point
+    emitTimeChange(timeui);
+#endif
+
+    // Run the callbacks
+    if (VL_UNLIKELY(m_fullDump)) {
+        m_fullDump = false;  // No more need for next dump to be full
+        runParallelCallbacks(m_fullCbs);
+    } else {
+        runParallelCallbacks(m_chgCbs);
+    }
+
+    for (uint32_t i = 0; i < m_cleanupCbs.size(); ++i) {
+        const CallbackRecord& cbr = m_cleanupCbs[i];
+        cbr.m_cleanupCb(cbr.m_userp, self());
+    }
+
+#ifdef VL_TRACE_OFFLOAD
+    if (VL_LIKELY(bufferp)) {
+        // Mark end of the offload buffer we just filled
+        *m_offloadBufferWritep++ = VerilatedTraceOffloadCommand::END;
+
+        // Assert no buffer overflow
+        assert(m_offloadBufferWritep - bufferp <= m_offloadBufferSize);
+
+        // Pass it to the worker thread
+        m_offloadBuffersToWorker.put(bufferp);
+    }
+#endif
+}
+
+//=============================================================================
+// Non-hot path internal interface to Verilator generated code
+
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::addThreadPool(VlThreadPool* threadPoolp)
+    VL_MT_SAFE_EXCLUDES(m_mutex) {
+    const VerilatedLockGuard lock{m_mutex};
+    for (VlThreadPool* const poolp : m_threadPoolps) {
+        if (poolp == threadPoolp) return;
+    }
+    m_threadPoolps.push_back(threadPoolp);
+}
+
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::addCallbackRecord(std::vector<CallbackRecord>& cbVec,
+                                                           CallbackRecord& cbRec)
+    VL_MT_SAFE_EXCLUDES(m_mutex) {
+    const VerilatedLockGuard lock{m_mutex};
+    if (VL_UNCOVERABLE(timeLastDump() != 0)) {  // LCOV_EXCL_START
+        const std::string msg = (std::string{"Internal: "} + __FILE__ + "::" + __FUNCTION__
+                                 + " called with already open file");
+        VL_FATAL_MT(__FILE__, __LINE__, "", msg.c_str());
+    }  // LCOV_EXCL_STOP
+    cbVec.push_back(cbRec);
+}
+
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::addInitCb(initCb_t cb, void* userp) VL_MT_SAFE {
+    CallbackRecord cbr{cb, userp};
+    addCallbackRecord(m_initCbs, cbr);
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::addFullCb(dumpCb_t cb, void* userp,
+                                                   VlThreadPool* threadPoolp) VL_MT_SAFE {
+    CallbackRecord cbr{cb, userp};
+    addThreadPool(threadPoolp);
+    addCallbackRecord(m_fullCbs[threadPoolp], cbr);
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::addChgCb(dumpCb_t cb, void* userp,
+                                                  VlThreadPool* threadPoolp) VL_MT_SAFE {
+    CallbackRecord cbr{cb, userp};
+    addThreadPool(threadPoolp);
+    addCallbackRecord(m_chgCbs[threadPoolp], cbr);
+}
+template <>
+void VerilatedTrace<VL_SUB_T, VL_BUF_T>::addCleanupCb(cleanupCb_t cb, void* userp) VL_MT_SAFE {
+    CallbackRecord cbr{cb, userp};
+    addCallbackRecord(m_cleanupCbs, cbr);
+}
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::pushNamePrefix(const std::string& prefix) {
+    m_namePrefixStack.push_back(m_namePrefixStack.back() + prefix);
+}
+
+template <> void VerilatedTrace<VL_SUB_T, VL_BUF_T>::popNamePrefix(unsigned count) {
+    while (count--) m_namePrefixStack.pop_back();
+    assert(!m_namePrefixStack.empty());
+}
+
+//=========================================================================
+// Primitives converting binary values to strings...
+
+// All of these take a destination pointer where the string will be emitted,
+// and a value to convert. There are a couple of variants for efficiency.
+
+static inline void cvtCDataToStr(char* dstp, CData value) {
+#ifdef VL_HAVE_SSE2
+    // Similar to cvtSDataToStr but only the bottom 8 byte lanes are used
+    const __m128i a = _mm_cvtsi32_si128(value);
+    const __m128i b = _mm_unpacklo_epi8(a, a);
+    const __m128i c = _mm_shufflelo_epi16(b, 0);
+    const __m128i m = _mm_set1_epi64x(0x0102040810204080);
+    const __m128i d = _mm_cmpeq_epi8(_mm_and_si128(c, m), m);
+    const __m128i result = _mm_sub_epi8(_mm_set1_epi8('0'), d);
+    _mm_storel_epi64(reinterpret_cast<__m128i*>(dstp), result);
+#else
+    dstp[0] = '0' | static_cast<char>((value >> 7) & 1);
+    dstp[1] = '0' | static_cast<char>((value >> 6) & 1);
+    dstp[2] = '0' | static_cast<char>((value >> 5) & 1);
+    dstp[3] = '0' | static_cast<char>((value >> 4) & 1);
+    dstp[4] = '0' | static_cast<char>((value >> 3) & 1);
+    dstp[5] = '0' | static_cast<char>((value >> 2) & 1);
+    dstp[6] = '0' | static_cast<char>((value >> 1) & 1);
+    dstp[7] = '0' | static_cast<char>(value & 1);
+#endif
+}
+
+static inline void cvtSDataToStr(char* dstp, SData value) {
+#ifdef VL_HAVE_SSE2
+    // We want each bit in the 16-bit input value to end up in a byte lane
+    // within the 128-bit XMM register. Note that x86 is little-endian and we
+    // want the MSB of the input at the low address, so we will bit-reverse
+    // at the same time.
+
+    // Put value in bottom of 128-bit register a[15:0] = value
+    const __m128i a = _mm_cvtsi32_si128(value);
+    // Interleave bytes with themselves
+    // b[15: 0] = {2{a[ 7:0]}} == {2{value[ 7:0]}}
+    // b[31:16] = {2{a[15:8]}} == {2{value[15:8]}}
+    const __m128i b = _mm_unpacklo_epi8(a, a);
+    // Shuffle bottom 64 bits, note swapping high bytes with low bytes
+    // c[31: 0] = {2{b[31:16]}} == {4{value[15:8}}
+    // c[63:32] = {2{b[15: 0]}} == {4{value[ 7:0}}
+    const __m128i c = _mm_shufflelo_epi16(b, 0x05);
+    // Shuffle whole register
+    // d[ 63: 0] = {2{c[31: 0]}} == {8{value[15:8}}
+    // d[126:54] = {2{c[63:32]}} == {8{value[ 7:0}}
+    const __m128i d = _mm_shuffle_epi32(c, 0x50);
+    // Test each bit within the bytes, this sets each byte lane to 0
+    // if the bit for that lane is 0 and to 0xff if the bit is 1.
+    const __m128i m = _mm_set1_epi64x(0x0102040810204080);
+    const __m128i e = _mm_cmpeq_epi8(_mm_and_si128(d, m), m);
+    // Convert to ASCII by subtracting the masks from ASCII '0':
+    // '0' - 0 is '0', '0' - -1 is '1'
+    const __m128i result = _mm_sub_epi8(_mm_set1_epi8('0'), e);
+    // Store the 16 characters to the un-aligned buffer
+    _mm_storeu_si128(reinterpret_cast<__m128i*>(dstp), result);
+#else
+    cvtCDataToStr(dstp, value >> 8);
+    cvtCDataToStr(dstp + 8, value);
+#endif
+}
+
+static inline void cvtIDataToStr(char* dstp, IData value) {
+#ifdef VL_HAVE_AVX2
+    // Similar to cvtSDataToStr but the bottom 16-bits are processed in the
+    // top half of the YMM registerss
+    const __m256i a = _mm256_insert_epi32(_mm256_undefined_si256(), value, 0);
+    const __m256i b = _mm256_permute4x64_epi64(a, 0);
+    const __m256i s = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2,
+                                      2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3);
+    const __m256i c = _mm256_shuffle_epi8(b, s);
+    const __m256i m = _mm256_set1_epi64x(0x0102040810204080);
+    const __m256i d = _mm256_cmpeq_epi8(_mm256_and_si256(c, m), m);
+    const __m256i result = _mm256_sub_epi8(_mm256_set1_epi8('0'), d);
+    _mm256_storeu_si256(reinterpret_cast<__m256i*>(dstp), result);
+#else
+    cvtSDataToStr(dstp, value >> 16);
+    cvtSDataToStr(dstp + 16, value);
+#endif
+}
+
+static inline void cvtQDataToStr(char* dstp, QData value) {
+    cvtIDataToStr(dstp, value >> 32);
+    cvtIDataToStr(dstp + 32, value);
+}
+
+#define cvtEDataToStr cvtIDataToStr
+
+//=========================================================================
+// VerilatedTraceBuffer
+
+template <>  //
+VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::VerilatedTraceBuffer(VL_SUB_T& owner)
+    : m_owner{owner} {
+#ifdef VL_TRACE_OFFLOAD
+    if (m_offloadBufferWritep) {
+        using This = VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>*;
+        // Tack on the buffer address
+        static_assert(2 * sizeof(uint32_t) >= sizeof(This),
+                      "This should be enough on all plafrorms");
+        *m_offloadBufferWritep++ = VerilatedTraceOffloadCommand::TRACE_BUFFER;
+        *reinterpret_cast<This*>(m_offloadBufferWritep) = this;
+        m_offloadBufferWritep += 2;
+    }
+#endif
+}
+
+// These functions must write the new value back into the old value store,
+// and subsequently call the format specific emit* implementations. Note
+// that this file must be included in the format specific implementation, so
+// the emit* functions can be inlined for performance.
+
+template <>  //
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullBit(uint32_t* oldp, CData newval) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    *oldp = newval;  // Still copy even if not tracing so chg doesn't call full
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    self()->emitBit(code, newval);
+}
+
+template <>
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullCData(uint32_t* oldp, CData newval, int bits) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    *oldp = newval;  // Still copy even if not tracing so chg doesn't call full
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    self()->emitCData(code, newval, bits);
+}
+
+template <>
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullSData(uint32_t* oldp, SData newval, int bits) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    *oldp = newval;  // Still copy even if not tracing so chg doesn't call full
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    self()->emitSData(code, newval, bits);
+}
+
+template <>
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullIData(uint32_t* oldp, IData newval, int bits) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    *oldp = newval;  // Still copy even if not tracing so chg doesn't call full
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    self()->emitIData(code, newval, bits);
+}
+
+template <>
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullQData(uint32_t* oldp, QData newval, int bits) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    *reinterpret_cast<QData*>(oldp) = newval;
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    self()->emitQData(code, newval, bits);
+}
+
+template <>
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullWData(uint32_t* oldp, const WData* newvalp,
+                                                         int bits) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    for (int i = 0; i < VL_WORDS_I(bits); ++i) oldp[i] = newvalp[i];
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    self()->emitWData(code, newvalp, bits);
+}
+
+template <>
+void VerilatedTraceBuffer<VL_SUB_T, VL_BUF_T>::fullDouble(uint32_t* oldp, double newval) {
+    const uint32_t code = oldp - m_sigs_oldvalp;
+    *reinterpret_cast<double*>(oldp) = newval;
+    if (VL_UNLIKELY(m_sigs_enabledp && !(VL_BITISSET_W(m_sigs_enabledp, code)))) return;
+    // cppcheck-suppress invalidPointerCast
+    self()->emitDouble(code, newval);
+}
+
+#endif  // VL_CPPCHECK
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2003-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2003-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -28,6 +28,14 @@
 #error "verilated_types.h should only be included by verilated.h"
 #endif

+#include <algorithm>
+#include <deque>
+#include <map>
+#include <set>
+#include <string>
+#include <unordered_set>
+#include <utility>
+
 //===================================================================
 // String formatters (required by below containers)

@@ -45,27 +53,27 @@ extern std::string VL_TO_STRING_W(int words, const WDataInP obj);
 #define VL_SIG16(name, msb, lsb) SData name  ///< Declare signal, 9-16 bits
 #define VL_SIG64(name, msb, lsb) QData name  ///< Declare signal, 33-64 bits
 #define VL_SIG(name, msb, lsb) IData name  ///< Declare signal, 17-32 bits
-#define VL_SIGW(name, msb, lsb, words) WData name[words]  ///< Declare signal, 65+ bits
+#define VL_SIGW(name, msb, lsb, words) VlWide<words> name  ///< Declare signal, 65+ bits
 #define VL_IN8(name, msb, lsb) CData name  ///< Declare input signal, 1-8 bits
 #define VL_IN16(name, msb, lsb) SData name  ///< Declare input signal, 9-16 bits
 #define VL_IN64(name, msb, lsb) QData name  ///< Declare input signal, 33-64 bits
 #define VL_IN(name, msb, lsb) IData name  ///< Declare input signal, 17-32 bits
-#define VL_INW(name, msb, lsb, words) WData name[words]  ///< Declare input signal, 65+ bits
+#define VL_INW(name, msb, lsb, words) VlWide<words> name  ///< Declare input signal, 65+ bits
 #define VL_INOUT8(name, msb, lsb) CData name  ///< Declare bidir signal, 1-8 bits
 #define VL_INOUT16(name, msb, lsb) SData name  ///< Declare bidir signal, 9-16 bits
 #define VL_INOUT64(name, msb, lsb) QData name  ///< Declare bidir signal, 33-64 bits
 #define VL_INOUT(name, msb, lsb) IData name  ///< Declare bidir signal, 17-32 bits
-#define VL_INOUTW(name, msb, lsb, words) WData name[words]  ///< Declare bidir signal, 65+ bits
+#define VL_INOUTW(name, msb, lsb, words) VlWide<words> name  ///< Declare bidir signal, 65+ bits
 #define VL_OUT8(name, msb, lsb) CData name  ///< Declare output signal, 1-8 bits
 #define VL_OUT16(name, msb, lsb) SData name  ///< Declare output signal, 9-16 bits
 #define VL_OUT64(name, msb, lsb) QData name  ///< Declare output signal, 33-64bits
 #define VL_OUT(name, msb, lsb) IData name  ///< Declare output signal, 17-32 bits
-#define VL_OUTW(name, msb, lsb, words) WData name[words]  ///< Declare output signal, 65+ bits
+#define VL_OUTW(name, msb, lsb, words) VlWide<words> name  ///< Declare output signal, 65+ bits

 //===================================================================
 // Shuffle RNG

-extern vluint64_t vl_rand64() VL_MT_SAFE;
+extern uint64_t vl_rand64() VL_MT_SAFE;

 class VlURNG final {
 public:
@@ -86,6 +94,7 @@ class VlReadMem final {
    FILE* m_fp;  // File handle for filename
    QData m_addr;  // Next address to read
    int m_linenum;  // Line number last read from file
+    bool m_anyAddr = false;  // Had address directive in the file
 public:
    VlReadMem(bool hex, int bits, const std::string& filename, QData start, QData end);
    ~VlReadMem();
@@ -232,7 +241,7 @@ public:
    int size() const { return m_deque.size(); }
    // Clear array. Verilog: function void delete([input index])
    void clear() { m_deque.clear(); }
-    void erase(vlsint32_t index) {
+    void erase(int32_t index) {
        if (VL_LIKELY(index >= 0 && index < m_deque.size()))
            m_deque.erase(m_deque.begin() + index);
    }
@@ -279,8 +288,8 @@ public:
    // Setting. Verilog: assoc[index] = v
    // Can't just overload operator[] or provide a "at" reference to set,
    // because we need to be able to insert only when the value is set
-    T_Value& at(vlsint32_t index) {
-        static T_Value s_throwAway;
+    T_Value& at(int32_t index) {
+        static VL_THREAD_LOCAL T_Value s_throwAway;
        // Needs to work for dynamic arrays, so does not use T_MaxSize
        if (VL_UNLIKELY(index < 0 || index >= m_deque.size())) {
            s_throwAway = atDefault();
@@ -290,8 +299,8 @@ public:
        }
    }
    // Accessing. Verilog: v = assoc[index]
-    const T_Value& at(vlsint32_t index) const {
-        static T_Value s_throwAway;
+    const T_Value& at(int32_t index) const {
+        static VL_THREAD_LOCAL T_Value s_throwAway;
        // Needs to work for dynamic arrays, so does not use T_MaxSize
        if (VL_UNLIKELY(index < 0 || index >= m_deque.size())) {
            return atDefault();
@@ -300,18 +309,18 @@ public:
        }
    }
    // function void q.insert(index, value);
-    void insert(vlsint32_t index, const T_Value& value) {
+    void insert(int32_t index, const T_Value& value) {
        if (VL_UNLIKELY(index < 0 || index >= m_deque.size())) return;
        m_deque.insert(m_deque.begin() + index, value);
    }

    // Return slice q[lsb:msb]
-    VlQueue slice(vlsint32_t lsb, vlsint32_t msb) const {
+    VlQueue slice(int32_t lsb, int32_t msb) const {
        VlQueue out;
        if (VL_UNLIKELY(lsb < 0)) lsb = 0;
        if (VL_UNLIKELY(lsb >= m_deque.size())) lsb = m_deque.size() - 1;
        if (VL_UNLIKELY(msb >= m_deque.size())) msb = m_deque.size() - 1;
-        for (vlsint32_t i = lsb; i <= msb; ++i) out.push_back(m_deque[i]);
+        for (int32_t i = lsb; i <= msb; ++i) out.push_back(m_deque[i]);
        return out;
    }

@@ -401,16 +410,16 @@ public:
    }
    template <typename Func> VlQueue find_last(Func with_func) const {
        IData index = m_deque.size() - 1;
-        for (auto it = m_deque.rbegin(); it != m_deque.rend(); ++it) {
-            if (with_func(index, *it)) return VlQueue::cons(*it);
+        for (auto& item : vlstd::reverse_view(m_deque)) {
+            if (with_func(index, item)) return VlQueue::cons(item);
            --index;
        }
        return VlQueue{};
    }
    template <typename Func> VlQueue<IData> find_last_index(Func with_func) const {
        IData index = m_deque.size() - 1;
-        for (auto it = m_deque.rbegin(); it != m_deque.rend(); ++it) {
-            if (with_func(index, *it)) return VlQueue<IData>::cons(index);
+        for (auto& item : vlstd::reverse_view(m_deque)) {
+            if (with_func(index, item)) return VlQueue<IData>::cons(index);
            --index;
        }
        return VlQueue<IData>{};
@@ -848,8 +857,8 @@ template <class T_Value, std::size_t T_Depth> struct VlUnpacked final {
    WData* data() { return &m_storage[0]; }
    const WData* data() const { return &m_storage[0]; }

-    T_Value& operator[](size_t index) { return m_storage[index]; };
-    const T_Value& operator[](size_t index) const { return m_storage[index]; };
+    T_Value& operator[](size_t index) { return m_storage[index]; }
+    const T_Value& operator[](size_t index) const { return m_storage[index]; }

    // Dumping. Verilog: str = $sformatf("%p", assoc)
    std::string to_string() const {
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -62,12 +62,23 @@ constexpr unsigned VL_TRACE_MAX_VCD_CODE_SIZE = 5;  // Maximum length of a VCD s
 // cache-lines.
 constexpr unsigned VL_TRACE_SUFFIX_ENTRY_SIZE = 8;  // Size of a suffix entry

+//=============================================================================
+// Utility functions: TODO: put these in a common place and share them.
+
+template <size_t N> static size_t roundUpToMultipleOf(size_t value) {
+    static_assert((N & (N - 1)) == 0, "'N' must be a power of 2");
+    size_t mask = N - 1;
+    return (value + mask) & ~mask;
+}
+
 //=============================================================================
 // Specialization of the generics for this trace format

-#define VL_DERIVED_T VerilatedVcd
-#include "verilated_trace_imp.cpp"
-#undef VL_DERIVED_T
+#define VL_SUB_T VerilatedVcd
+#define VL_BUF_T VerilatedVcdBuffer
+#include "verilated_trace_imp.h"
+#undef VL_SUB_T
+#undef VL_BUF_T

 //=============================================================================
 //=============================================================================
@@ -172,7 +183,7 @@ bool VerilatedVcd::preChangeDump() {
    return isOpen();
 }

-void VerilatedVcd::emitTimeChange(vluint64_t timeui) {
+void VerilatedVcd::emitTimeChange(uint64_t timeui) {
    printStr("#");
    printQuad(timeui);
    printStr("\n");
@@ -183,7 +194,7 @@ void VerilatedVcd::makeNameMap() {
    deleteNameMap();
    m_namemapp = new NameMap;

-    VerilatedTrace<VerilatedVcd>::traceInit();
+    Super::traceInit();

    // Though not speced, it's illegal to generate a vcd with signals
    // not under any module - it crashes at least two viewers.
@@ -218,13 +229,17 @@ VerilatedVcd::~VerilatedVcd() {
    if (m_wrBufp) VL_DO_CLEAR(delete[] m_wrBufp, m_wrBufp = nullptr);
    deleteNameMap();
    if (m_filep && m_fileNewed) VL_DO_CLEAR(delete m_filep, m_filep = nullptr);
+#ifdef VL_TRACE_PARALLEL
+    assert(m_numBuffers == m_freeBuffers.size());
+    for (auto& pair : m_freeBuffers) VL_DO_CLEAR(delete[] pair.first, pair.first = nullptr);
+#endif
 }

 void VerilatedVcd::closePrev() {
    // This function is on the flush() call path
    if (!isOpen()) return;

-    VerilatedTrace<VerilatedVcd>::flushBase();
+    Super::flushBase();
    bufferFlush();
    m_isOpen = false;
    m_filep->close();
@@ -251,14 +266,14 @@ void VerilatedVcd::close() VL_MT_SAFE_EXCLUDES(m_mutex) {
        printStr(" $end\n");
    }
    closePrev();
-    // closePrev() called VerilatedTrace<VerilatedVcd>::flush(), so we just
+    // closePrev() called Super::flush(), so we just
    // need to shut down the tracing thread here.
-    VerilatedTrace<VerilatedVcd>::closeBase();
+    Super::closeBase();
 }

 void VerilatedVcd::flush() VL_MT_SAFE_EXCLUDES(m_mutex) {
    const VerilatedLockGuard lock{m_mutex};
-    VerilatedTrace<VerilatedVcd>::flushBase();
+    Super::flushBase();
    bufferFlush();
 }

@@ -270,19 +285,19 @@ void VerilatedVcd::printStr(const char* str) {
    }
 }

-void VerilatedVcd::printQuad(vluint64_t n) {
+void VerilatedVcd::printQuad(uint64_t n) {
    constexpr size_t LEN_STR_QUAD = 40;
    char buf[LEN_STR_QUAD];
-    VL_SNPRINTF(buf, LEN_STR_QUAD, "%" VL_PRI64 "u", n);
+    VL_SNPRINTF(buf, LEN_STR_QUAD, "%" PRIu64, n);
    printStr(buf);
 }

-void VerilatedVcd::bufferResize(vluint64_t minsize) {
+void VerilatedVcd::bufferResize(size_t minsize) {
    // minsize is size of largest write.  We buffer at least 8 times as much data,
    // writing when we are 3/4 full (with thus 2*minsize remaining free)
    if (VL_UNLIKELY(minsize > m_wrChunkSize)) {
        const char* oldbufp = m_wrBufp;
-        m_wrChunkSize = minsize * 2;
+        m_wrChunkSize = roundUpToMultipleOf<1024>(minsize * 2);
        m_wrBufp = new char[m_wrChunkSize * 8];
        std::memcpy(m_wrBufp, oldbufp, m_writep - oldbufp);
        m_writep = m_wrBufp + (m_writep - oldbufp);
@@ -292,7 +307,7 @@ void VerilatedVcd::bufferResize(vluint64_t minsize) {
 }

 void VerilatedVcd::bufferFlush() VL_MT_UNSAFE_ONE {
-    // This function can be called from the trace thread
+    // This function can be called from the trace offload thread
    // This function is on the flush() call path
    // We add output data to m_writep.
    // When it gets nearly full we dump it using this routine which calls write()
@@ -328,7 +343,7 @@ void VerilatedVcd::bufferFlush() VL_MT_UNSAFE_ONE {
 //=============================================================================
 // VCD string code

-char* VerilatedVcd::writeCode(char* writep, vluint32_t code) {
+char* VerilatedVcd::writeCode(char* writep, uint32_t code) {
    *writep++ = static_cast<char>('!' + code % 94);
    code /= 94;
    while (code) {
@@ -426,8 +441,9 @@ void VerilatedVcd::dumpHeader() {
                case VLT_TRACE_SCOPE_UNION: printStr("union "); break;
                default: printStr("module ");
                }
-            } else
+            } else {
                printStr("module ");
+            }

            for (; *np && *np != ' ' && *np != '\t'; np++) {
                if (*np == '[') {
@@ -458,18 +474,22 @@ void VerilatedVcd::dumpHeader() {
    deleteNameMap();
 }

-void VerilatedVcd::declare(vluint32_t code, const char* name, const char* wirep, bool array,
+void VerilatedVcd::declare(uint32_t code, const char* name, const char* wirep, bool array,
                           int arraynum, bool tri, bool bussed, int msb, int lsb) {
    const int bits = ((msb > lsb) ? (msb - lsb) : (lsb - msb)) + 1;

-    VerilatedTrace<VerilatedVcd>::declCode(code, bits, tri);
+    const bool enabled = Super::declCode(code, name, bits, tri);

    if (m_suffixes.size() <= nextCode() * VL_TRACE_SUFFIX_ENTRY_SIZE) {
        m_suffixes.resize(nextCode() * VL_TRACE_SUFFIX_ENTRY_SIZE * 2, 0);
    }

-    // Make sure write buffer is large enough (one character per bit), plus header
-    bufferResize(bits + 1024);
+    // Keep upper bound on bytes a single signal cna emit into the buffer
+    m_maxSignalBytes = std::max<size_t>(m_maxSignalBytes, bits + 32);
+    // Make sure write buffer is large enough, plus header
+    bufferResize(m_maxSignalBytes + 1024);
+
+    if (!enabled) return;

    // Split name into basename
    // Spaces and tabs aren't legal in VCD signal names, so:
@@ -477,10 +497,7 @@ void VerilatedVcd::declare(vluint32_t code, const char* name, const char* wirep,
    // Tab separates final scope from signal name
    // Tab sorts before spaces, so signals nicely will print before scopes
    // Note the hiername may be nothing, if so we'll add "\t{name}"
-    std::string nameasstr = name;
-    if (!moduleName().empty()) {
-        nameasstr = moduleName() + scopeEscape() + nameasstr;  // Optional ->module prefix
-    }
+    std::string nameasstr = namePrefix() + name;
    std::string hiername;
    std::string basename;
    for (const char* cp = nameasstr.c_str(); *cp; cp++) {
@@ -544,44 +561,91 @@ void VerilatedVcd::declare(vluint32_t code, const char* name, const char* wirep,
    m_namemapp->emplace(hiername, decl);
 }

-void VerilatedVcd::declBit(vluint32_t code, const char* name, bool array, int arraynum) {
+void VerilatedVcd::declBit(uint32_t code, const char* name, bool array, int arraynum) {
    declare(code, name, "wire", array, arraynum, false, false, 0, 0);
 }
-void VerilatedVcd::declBus(vluint32_t code, const char* name, bool array, int arraynum, int msb,
+void VerilatedVcd::declBus(uint32_t code, const char* name, bool array, int arraynum, int msb,
                           int lsb) {
    declare(code, name, "wire", array, arraynum, false, true, msb, lsb);
 }
-void VerilatedVcd::declQuad(vluint32_t code, const char* name, bool array, int arraynum, int msb,
+void VerilatedVcd::declQuad(uint32_t code, const char* name, bool array, int arraynum, int msb,
                            int lsb) {
    declare(code, name, "wire", array, arraynum, false, true, msb, lsb);
 }
-void VerilatedVcd::declArray(vluint32_t code, const char* name, bool array, int arraynum, int msb,
+void VerilatedVcd::declArray(uint32_t code, const char* name, bool array, int arraynum, int msb,
                             int lsb) {
    declare(code, name, "wire", array, arraynum, false, true, msb, lsb);
 }
-void VerilatedVcd::declDouble(vluint32_t code, const char* name, bool array, int arraynum) {
+void VerilatedVcd::declDouble(uint32_t code, const char* name, bool array, int arraynum) {
    declare(code, name, "real", array, arraynum, false, false, 63, 0);
 }
-#ifdef VL_TRACE_VCD_OLD_API
-void VerilatedVcd::declTriBit(vluint32_t code, const char* name, bool array, int arraynum) {
-    declare(code, name, "wire", array, arraynum, true, false, 0, 0);
-}
-void VerilatedVcd::declTriBus(vluint32_t code, const char* name, bool array, int arraynum, int msb,
-                              int lsb) {
-    declare(code, name, "wire", array, arraynum, true, true, msb, lsb);
-}
-void VerilatedVcd::declTriQuad(vluint32_t code, const char* name, bool array, int arraynum,
-                               int msb, int lsb) {
-    declare(code, name, "wire", array, arraynum, true, true, msb, lsb);
-}
-void VerilatedVcd::declTriArray(vluint32_t code, const char* name, bool array, int arraynum,
-                                int msb, int lsb) {
-    declare(code, name, "wire", array, arraynum, true, true, msb, lsb);
-}
-#endif  //  VL_TRACE_VCD_OLD_API

 //=============================================================================
-// Trace rendering prinitives
+// Get/commit trace buffer
+
+VerilatedVcdBuffer* VerilatedVcd::getTraceBuffer() {
+#ifdef VL_TRACE_PARALLEL
+    // Note: This is called from VeriltedVcd::dump, which already holds the lock
+    // If no buffer available, allocate a new one
+    if (m_freeBuffers.empty()) {
+        constexpr size_t pageSize = 4096;
+        // 4 * m_maxSignalBytes, so we can reserve 2 * m_maxSignalBytes at the end for safety
+        size_t startingSize = roundUpToMultipleOf<pageSize>(4 * m_maxSignalBytes);
+        m_freeBuffers.emplace_back(new char[startingSize], startingSize);
+        ++m_numBuffers;
+    }
+    // Grab a buffer
+    const auto pair = m_freeBuffers.back();
+    m_freeBuffers.pop_back();
+    // Return the buffer
+    return new VerilatedVcdBuffer{*this, pair.first, pair.second};
+#else
+    return new VerilatedVcdBuffer{*this};
+#endif
+}
+
+void VerilatedVcd::commitTraceBuffer(VerilatedVcdBuffer* bufp) {
+#ifdef VL_TRACE_PARALLEL
+    // Note: This is called from VeriltedVcd::dump, which already holds the lock
+    // Resize output buffer. Note, we use the full size of the trace buffer, as
+    // this is a lot more stable than the actual occupancy of the trace buffer.
+    // This helps us to avoid re-allocations due to small size changes.
+    bufferResize(bufp->m_size);
+    // Compute occupancy of buffer
+    const size_t usedSize = bufp->m_writep - bufp->m_bufp;
+    // Copy to output buffer
+    std::memcpy(m_writep, bufp->m_bufp, usedSize);
+    // Adjust write pointer
+    m_writep += usedSize;
+    // Flush if necessary
+    bufferCheck();
+    // Put buffer back on free list
+    m_freeBuffers.emplace_back(bufp->m_bufp, bufp->m_size);
+#else
+    // Needs adjusting for emitTimeChange
+    m_writep = bufp->m_writep;
+#endif
+    delete bufp;
+}
+
+//=============================================================================
+// VerilatedVcdBuffer implementation
+
+#ifdef VL_TRACE_PARALLEL
+VerilatedVcdBuffer::VerilatedVcdBuffer(VerilatedVcd& owner, char* bufp, size_t size)
+    : VerilatedTraceBuffer<VerilatedVcd, VerilatedVcdBuffer>{owner}
+    , m_writep{bufp}
+    , m_bufp{bufp}
+    , m_size{size} {
+    adjustGrowp();
+}
+#else
+VerilatedVcdBuffer::VerilatedVcdBuffer(VerilatedVcd& owner)
+    : VerilatedTraceBuffer<VerilatedVcd, VerilatedVcdBuffer>{owner} {}
+#endif
+
+//=============================================================================
+// Trace rendering primitives

 static inline void
 VerilatedVcdCCopyAndAppendNewLine(char* writep, const char* suffixp) VL_ATTR_NO_SANITIZE_ALIGN;
@@ -593,7 +657,7 @@ static inline void VerilatedVcdCCopyAndAppendNewLine(char* writep, const char* s
 #ifdef VL_X86_64
    // Copy the whole 8 bytes in one go, this works on little-endian machines
    // supporting unaligned stores.
-    *reinterpret_cast<vluint64_t*>(writep) = *reinterpret_cast<const vluint64_t*>(suffixp);
+    *reinterpret_cast<uint64_t*>(writep) = *reinterpret_cast<const uint64_t*>(suffixp);
 #else
    // Portable variant
    writep[0] = suffixp[0];
@@ -606,25 +670,55 @@ static inline void VerilatedVcdCCopyAndAppendNewLine(char* writep, const char* s
 #endif
 }

-void VerilatedVcd::finishLine(vluint32_t code, char* writep) {
-    const char* const suffixp = m_suffixes.data() + code * VL_TRACE_SUFFIX_ENTRY_SIZE;
+void VerilatedVcdBuffer::finishLine(uint32_t code, char* writep) {
+    const char* const suffixp = m_suffixes + code * VL_TRACE_SUFFIX_ENTRY_SIZE;
+    VL_DEBUG_IFDEF(assert(suffixp[0]););
    VerilatedVcdCCopyAndAppendNewLine(writep, suffixp);

    // Now write back the write pointer incremented by the actual size of the
    // suffix, which was stored in the last byte of the suffix buffer entry.
    m_writep = writep + suffixp[VL_TRACE_SUFFIX_ENTRY_SIZE - 1];
-    bufferCheck();
+
+#ifdef VL_TRACE_PARALLEL
+    // Double the size of the buffer if necessary
+    if (VL_UNLIKELY(m_writep >= m_growp)) {
+        // Compute occupied size of current buffer
+        const size_t usedSize = m_writep - m_bufp;
+        // We are always doubling the size
+        m_size *= 2;
+        // Allocate the new buffer
+        char* const newBufp = new char[m_size];
+        // Copy from current buffer to new buffer
+        std::memcpy(newBufp, m_bufp, usedSize);
+        // Delete current buffer
+        delete[] m_bufp;
+        // Make new buffer the current buffer
+        m_bufp = newBufp;
+        // Adjust write pointer
+        m_writep = m_bufp + usedSize;
+        // Adjust resize limit
+        adjustGrowp();
+    }
+#else
+    // Flush the write buffer if there's not enough space left for new information
+    // We only call this once per vector, so we need enough slop for a very wide "b###" line
+    if (VL_UNLIKELY(m_writep > m_wrFlushp)) {
+        m_owner.m_writep = m_writep;
+        m_owner.bufferFlush();
+        m_writep = m_owner.m_writep;
+    }
+#endif
 }

 //=============================================================================
 // emit* trace routines

 // Note: emit* are only ever called from one place (full* in
-// verilated_trace_imp.cpp, which is included in this file at the top),
+// verilated_trace_imp.h, which is included in this file at the top),
 // so always inline them.

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitBit(vluint32_t code, CData newval) {
+void VerilatedVcdBuffer::emitBit(uint32_t code, CData newval) {
    // Don't prefetch suffix as it's a bit too late;
    char* wp = m_writep;
    *wp++ = '0' | static_cast<char>(newval);
@@ -632,7 +726,7 @@ void VerilatedVcd::emitBit(vluint32_t code, CData newval) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitCData(vluint32_t code, CData newval, int bits) {
+void VerilatedVcdBuffer::emitCData(uint32_t code, CData newval, int bits) {
    char* wp = m_writep;
    *wp++ = 'b';
    cvtCDataToStr(wp, newval << (VL_BYTESIZE - bits));
@@ -640,7 +734,7 @@ void VerilatedVcd::emitCData(vluint32_t code, CData newval, int bits) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitSData(vluint32_t code, SData newval, int bits) {
+void VerilatedVcdBuffer::emitSData(uint32_t code, SData newval, int bits) {
    char* wp = m_writep;
    *wp++ = 'b';
    cvtSDataToStr(wp, newval << (VL_SHORTSIZE - bits));
@@ -648,7 +742,7 @@ void VerilatedVcd::emitSData(vluint32_t code, SData newval, int bits) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitIData(vluint32_t code, IData newval, int bits) {
+void VerilatedVcdBuffer::emitIData(uint32_t code, IData newval, int bits) {
    char* wp = m_writep;
    *wp++ = 'b';
    cvtIDataToStr(wp, newval << (VL_IDATASIZE - bits));
@@ -656,7 +750,7 @@ void VerilatedVcd::emitIData(vluint32_t code, IData newval, int bits) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitQData(vluint32_t code, QData newval, int bits) {
+void VerilatedVcdBuffer::emitQData(uint32_t code, QData newval, int bits) {
    char* wp = m_writep;
    *wp++ = 'b';
    cvtQDataToStr(wp, newval << (VL_QUADSIZE - bits));
@@ -664,7 +758,7 @@ void VerilatedVcd::emitQData(vluint32_t code, QData newval, int bits) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitWData(vluint32_t code, const WData* newvalp, int bits) {
+void VerilatedVcdBuffer::emitWData(uint32_t code, const WData* newvalp, int bits) {
    int words = VL_WORDS_I(bits);
    char* wp = m_writep;
    *wp++ = 'b';
@@ -681,268 +775,10 @@ void VerilatedVcd::emitWData(vluint32_t code, const WData* newvalp, int bits) {
 }

 VL_ATTR_ALWINLINE
-void VerilatedVcd::emitDouble(vluint32_t code, double newval) {
+void VerilatedVcdBuffer::emitDouble(uint32_t code, double newval) {
    char* wp = m_writep;
    // Buffer can't overflow before VL_SNPRINTF; we sized during declaration
-    VL_SNPRINTF(wp, m_wrChunkSize, "r%.16g", newval);
+    VL_SNPRINTF(wp, m_maxSignalBytes, "r%.16g", newval);
    wp += std::strlen(wp);
    finishLine(code, wp);
 }
-
-#ifdef VL_TRACE_VCD_OLD_API
-
-void VerilatedVcd::fullBit(vluint32_t code, const vluint32_t newval) {
-    // Note the &1, so we don't require clean input -- makes more common no change case faster
-    *oldp(code) = newval;
-    *m_writep++ = ('0' + static_cast<char>(newval & 1));
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullBus(vluint32_t code, const vluint32_t newval, int bits) {
-    *oldp(code) = newval;
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        *m_writep++ = ((newval & (1L << bit)) ? '1' : '0');
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullQuad(vluint32_t code, const vluint64_t newval, int bits) {
-    (*(reinterpret_cast<vluint64_t*>(oldp(code)))) = newval;
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        *m_writep++ = ((newval & (1ULL << bit)) ? '1' : '0');
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullArray(vluint32_t code, const vluint32_t* newval, int bits) {
-    for (int word = 0; word < (((bits - 1) / 32) + 1); ++word) { oldp(code)[word] = newval[word]; }
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        *m_writep++ = ((newval[(bit / 32)] & (1L << (bit & 0x1f))) ? '1' : '0');
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullArray(vluint32_t code, const vluint64_t* newval, int bits) {
-    for (int word = 0; word < (((bits - 1) / 64) + 1); ++word) { oldp(code)[word] = newval[word]; }
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        *m_writep++ = ((newval[(bit / 64)] & (1ULL << (bit & 0x3f))) ? '1' : '0');
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullTriBit(vluint32_t code, const vluint32_t newval, const vluint32_t newtri) {
-    oldp(code)[0] = newval;
-    oldp(code)[1] = newtri;
-    *m_writep++ = "01zz"[newval | (newtri << 1)];
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullTriBus(vluint32_t code, const vluint32_t newval, const vluint32_t newtri,
-                              int bits) {
-    oldp(code)[0] = newval;
-    oldp(code)[1] = newtri;
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        *m_writep++ = "01zz"[((newval >> bit) & 1) | (((newtri >> bit) & 1) << 1)];
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullTriQuad(vluint32_t code, const vluint64_t newval, const vluint64_t newtri,
-                               int bits) {
-    (*(reinterpret_cast<vluint64_t*>(oldp(code)))) = newval;
-    (*(reinterpret_cast<vluint64_t*>(oldp(code + 1)))) = newtri;
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        *m_writep++ = "01zz"[((newval >> bit) & 1ULL) | (((newtri >> bit) & 1ULL) << 1ULL)];
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullTriArray(vluint32_t code, const vluint32_t* newvalp,
-                                const vluint32_t* newtrip, int bits) {
-    for (int word = 0; word < (((bits - 1) / 32) + 1); ++word) {
-        oldp(code)[word * 2] = newvalp[word];
-        oldp(code)[word * 2 + 1] = newtrip[word];
-    }
-    *m_writep++ = 'b';
-    for (int bit = bits - 1; bit >= 0; --bit) {
-        vluint32_t valbit = (newvalp[(bit / 32)] >> (bit & 0x1f)) & 1;
-        vluint32_t tribit = (newtrip[(bit / 32)] >> (bit & 0x1f)) & 1;
-        *m_writep++ = "01zz"[valbit | (tribit << 1)];
-    }
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-void VerilatedVcd::fullDouble(vluint32_t code, const double newval) {
-    // cppcheck-suppress invalidPointerCast
-    (*(reinterpret_cast<double*>(oldp(code)))) = newval;
-    // Buffer can't overflow before VL_SNPRINTF; we sized during declaration
-    VL_SNPRINTF(m_writep, m_wrChunkSize, "r%.16g", newval);
-    m_writep += std::strlen(m_writep);
-    *m_writep++ = ' ';
-    m_writep = writeCode(m_writep, code);
-    *m_writep++ = '\n';
-    bufferCheck();
-}
-
-#endif  // VL_TRACE_VCD_OLD_API
-
-//======================================================================
-//======================================================================
-//======================================================================
-
-#ifdef VERILATED_VCD_TEST
-#include <iostream>
-
-extern void verilated_trace_imp_selftest();
-
-vluint32_t v1, v2, s1, s2[3];
-vluint32_t tri96[3];
-vluint32_t tri96__tri[3];
-vluint64_t quad96[2];
-vluint64_t tquad;
-vluint64_t tquad__tri;
-vluint8_t ch;
-vluint64_t timestamp = 1;
-double doub = 0.0;
-float flo = 0.0f;
-
-void vcdInit(void*, VerilatedVcd* vcdp, vluint32_t) {
-    vcdp->scopeEscape('.');
-    vcdp->module("top");
-    /**/ vcdp->declBus(0x2, "v1", -1, 0, 5, 1);
-    /**/ vcdp->declBus(0x3, "v2", -1, 0, 6, 1);
-    /**/ vcdp->module("top.sub1");
-    /***/ vcdp->declBit(0x4, "s1", -1, 0);
-    /***/ vcdp->declBit(0x5, "ch", -1, 0);
-    /**/ vcdp->module("top.sub2");
-    /***/ vcdp->declArray(0x6, "s2", -1, 0, 40, 3);
-    // Note need to add 3 for next code.
-    vcdp->module("top2");
-    /**/ vcdp->declBus(0x2, "t2v1", -1, 0, 4, 1);
-    /**/ vcdp->declTriBit(0x10, "io1", -1, 0);
-    /**/ vcdp->declTriBus(0x12, "io5", -1, 0, 4, 0);
-    /**/ vcdp->declTriArray(0x16, "io96", -1, 0, 95, 0);
-    /**/  // Note need to add 6 for next code.
-    /**/ vcdp->declDouble(0x1c, "doub", -1, 0);
-    /**/  // Note need to add 2 for next code.
-    /**/ vcdp->declArray(0x20, "q2", -1, 0, 95, 0);
-    /**/  // Note need to add 4 for next code.
-    /**/ vcdp->declTriQuad(0x24, "tq", -1, 0, 63, 0);
-    /**/  // Note need to add 4 for next code.
-}
-
-void vcdFull(void*, VerilatedVcd* vcdp) {
-    vcdp->fullBus(0x2, v1, 5);
-    vcdp->fullBus(0x3, v2, 7);
-    vcdp->fullBit(0x4, s1);
-    vcdp->fullBus(0x5, ch, 2);
-    vcdp->fullArray(0x6, &s2[0], 38);
-    vcdp->fullTriBit(0x10, tri96[0] & 1, tri96__tri[0] & 1);
-    vcdp->fullTriBus(0x12, tri96[0] & 0x1f, tri96__tri[0] & 0x1f, 5);
-    vcdp->fullTriArray(0x16, tri96, tri96__tri, 96);
-    vcdp->fullDouble(0x1c, doub);
-    vcdp->fullArray(0x20, &quad96[0], 96);
-    vcdp->fullTriQuad(0x24, tquad, tquad__tri, 64);
-}
-
-void vcdChange(void*, VerilatedVcd* vcdp) {
-    vcdp->chgBus(0x2, v1, 5);
-    vcdp->chgBus(0x3, v2, 7);
-    vcdp->chgBit(0x4, s1);
-    vcdp->chgBus(0x5, ch, 2);
-    vcdp->chgArray(0x6, &s2[0], 38);
-    vcdp->chgTriBit(0x10, tri96[0] & 1, tri96__tri[0] & 1);
-    vcdp->chgTriBus(0x12, tri96[0] & 0x1f, tri96__tri[0] & 0x1f, 5);
-    vcdp->chgTriArray(0x16, tri96, tri96__tri, 96);
-    vcdp->chgDouble(0x1c, doub);
-    vcdp->chgArray(0x20, &quad96[0], 96);
-    vcdp->chgTriQuad(0x24, tquad, tquad__tri, 64);
-}
-
-// clang-format off
-void vcdTestMain(const char* filenamep) {
-    verilated_trace_imp_selftest();
-
-    v1 = v2 = s1 = 0;
-    s2[0] = s2[1] = s2[2] = 0;
-    tri96[2] = tri96[1] = tri96[0] = 0;
-    tri96__tri[2] = tri96__tri[1] = tri96__tri[0] = ~0;
-    quad96[1] = quad96[0] = 0;
-    ch = 0;
-    doub = 0;
-    tquad = tquad__tri = 0;
-    {
-        VerilatedVcdC* vcdp = new VerilatedVcdC;
-        vcdp->evcd(true);
-        vcdp->set_time_unit("1ms");
-        vcdp->set_time_unit(std::string{"1ms"});
-        vcdp->set_time_resolution("1ns");
-        vcdp->set_time_resolution(std::string{"1ns"});
-        vcdp->spTrace()->addInitCb(&vcdInit, 0);
-        vcdp->spTrace()->addFullCb(&vcdFull, 0);
-        vcdp->spTrace()->addChgCb(&vcdChange, 0);
-        vcdp->open(filenamep);
-        // Dumping
-        vcdp->dump(++timestamp);
-        v1 = 0xfff;
-        tri96[2] = 4; tri96[1] = 2; tri96[0] = 1;
-        tri96__tri[2] = tri96__tri[1] = tri96__tri[0] = ~0;  // Still tri
-        quad96[1] = 0xffffffff; quad96[0] = 0;
-        doub = 1.5;
-        flo = 1.4f;
-        vcdp->dump(++timestamp);
-        v2 = 0x1;
-        s2[1] = 2;
-        tri96__tri[2] = tri96__tri[1] = tri96__tri[0] = 0;  // enable w/o data change
-        quad96[1] = 0; quad96[0] = ~0;
-        doub = -1.66e13;
-        flo = 0.123f;
-        tquad = 0x00ff00ff00ff00ffULL;
-        tquad__tri = 0x0000fffff0000ffffULL;
-        vcdp->dump(++timestamp);
-        ch = 2;
-        tri96[2] = ~4; tri96[1] = ~2; tri96[0] = ~1;
-        doub = -3.33e-13;
-        vcdp->dump(++timestamp);
-        vcdp->dump(++timestamp);
-# ifdef VERILATED_VCD_TEST_64BIT
-        const vluint64_t bytesPerDump = 15ULL;
-        for (vluint64_t i = 0; i < ((1ULL << 32) / bytesPerDump); i++) {
-            v1 = i;
-            vcdp->dump(++timestamp);
-        }
-# endif
-        vcdp->close();
-        VL_DO_CLEAR(delete vcdp, vcdp = nullptr);
-    }
-}
-#endif
-// clang-format on
-
-//********************************************************************
-// ;compile-command: "v4make test_regress/t/t_trace_c_api.pl"
-//
-// Local Variables:
-// End:
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -28,7 +28,189 @@
 #include <string>
 #include <vector>

-class VerilatedVcd;
+class VerilatedVcdBuffer;
+class VerilatedVcdFile;
+
+//=============================================================================
+// VerilatedVcd
+// Base class to create a Verilator VCD dump
+// This is an internally used class - see VerilatedVcdC for what to call from applications
+
+class VerilatedVcd VL_NOT_FINAL : public VerilatedTrace<VerilatedVcd, VerilatedVcdBuffer> {
+public:
+    using Super = VerilatedTrace<VerilatedVcd, VerilatedVcdBuffer>;
+
+private:
+    friend Buffer;  // Give the buffer access to the private bits
+
+    //=========================================================================
+    // VCD specific internals
+
+    VerilatedVcdFile* m_filep;  // File we're writing to
+    bool m_fileNewed;  // m_filep needs destruction
+    bool m_isOpen = false;  // True indicates open file
+    bool m_evcd = false;  // True for evcd format
+    std::string m_filename;  // Filename we're writing to (if open)
+    uint64_t m_rolloverMB = 0;  // MB of file size to rollover at
+    int m_modDepth = 0;  // Depth of module hierarchy
+
+    char* m_wrBufp;  // Output buffer
+    char* m_wrFlushp;  // Output buffer flush trigger location
+    char* m_writep;  // Write pointer into output buffer
+    size_t m_wrChunkSize;  // Output buffer size
+    size_t m_maxSignalBytes = 0;  // Upper bound on number of bytes a single signal can generate
+    uint64_t m_wroteBytes = 0;  // Number of bytes written to this file
+
+    std::vector<char> m_suffixes;  // VCD line end string codes + metadata
+
+    using NameMap = std::map<const std::string, const std::string>;
+    NameMap* m_namemapp = nullptr;  // List of names for the header
+
+#ifdef VL_TRACE_PARALLEL
+    // Vector of free trace buffers as (pointer, size) pairs.
+    std::vector<std::pair<char*, size_t>> m_freeBuffers;
+    size_t m_numBuffers = 0;  // Number of trace buffers allocated
+#endif
+
+    void bufferResize(size_t minsize);
+    void bufferFlush() VL_MT_UNSAFE_ONE;
+    inline void bufferCheck() {
+        // Flush the write buffer if there's not enough space left for new information
+        // We only call this once per vector, so we need enough slop for a very wide "b###" line
+        if (VL_UNLIKELY(m_writep > m_wrFlushp)) bufferFlush();
+    }
+    void openNextImp(bool incFilename);
+    void closePrev();
+    void closeErr();
+    void makeNameMap();
+    void deleteNameMap();
+    void printIndent(int level_change);
+    void printStr(const char* str);
+    void printQuad(uint64_t n);
+    void printTime(uint64_t timeui);
+    void declare(uint32_t code, const char* name, const char* wirep, bool array, int arraynum,
+                 bool tri, bool bussed, int msb, int lsb);
+
+    void dumpHeader();
+
+    static char* writeCode(char* writep, uint32_t code);
+
+    // CONSTRUCTORS
+    VL_UNCOPYABLE(VerilatedVcd);
+
+protected:
+    //=========================================================================
+    // Implementation of VerilatedTrace interface
+
+    // Called when the trace moves forward to a new time point
+    virtual void emitTimeChange(uint64_t timeui) override;
+
+    // Hooks called from VerilatedTrace
+    virtual bool preFullDump() override { return isOpen(); }
+    virtual bool preChangeDump() override;
+
+    // Trace buffer management
+    virtual VerilatedVcdBuffer* getTraceBuffer() override;
+    virtual void commitTraceBuffer(VerilatedVcdBuffer*) override;
+
+public:
+    //=========================================================================
+    // External interface to client code
+
+    // CONSTRUCTOR
+    explicit VerilatedVcd(VerilatedVcdFile* filep = nullptr);
+    ~VerilatedVcd();
+
+    // ACCESSORS
+    // Set size in megabytes after which new file should be created
+    void rolloverMB(uint64_t rolloverMB) { m_rolloverMB = rolloverMB; }
+
+    // METHODS - All must be thread safe
+    // Open the file; call isOpen() to see if errors
+    void open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex);
+    // Open next data-only file
+    void openNext(bool incFilename) VL_MT_SAFE_EXCLUDES(m_mutex);
+    // Close the file
+    void close() VL_MT_SAFE_EXCLUDES(m_mutex);
+    // Flush any remaining data to this file
+    void flush() VL_MT_SAFE_EXCLUDES(m_mutex);
+    // Return if file is open
+    bool isOpen() const VL_MT_SAFE { return m_isOpen; }
+
+    //=========================================================================
+    // Internal interface to Verilator generated code
+
+    void declBit(uint32_t code, const char* name, bool array, int arraynum);
+    void declBus(uint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
+    void declQuad(uint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
+    void declArray(uint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
+    void declDouble(uint32_t code, const char* name, bool array, int arraynum);
+};
+
+#ifndef DOXYGEN
+// Declare specialization here as it's used in VerilatedFstC just below
+template <> void VerilatedVcd::Super::dump(uint64_t time);
+template <> void VerilatedVcd::Super::set_time_unit(const char* unitp);
+template <> void VerilatedVcd::Super::set_time_unit(const std::string& unit);
+template <> void VerilatedVcd::Super::set_time_resolution(const char* unitp);
+template <> void VerilatedVcd::Super::set_time_resolution(const std::string& unit);
+template <> void VerilatedVcd::Super::dumpvars(int level, const std::string& hier);
+#endif  // DOXYGEN
+
+//=============================================================================
+// VerilatedVcdBuffer
+
+class VerilatedVcdBuffer final : public VerilatedTraceBuffer<VerilatedVcd, VerilatedVcdBuffer> {
+    // Give the trace file access to the private bits
+    friend VerilatedVcd;
+    friend VerilatedVcd::Super;
+
+#ifdef VL_TRACE_PARALLEL
+    char* m_writep;  // Write pointer into m_bufp
+    char* m_bufp;  // The beginning of the trace buffer
+    size_t m_size;  // The size of the buffer at m_bufp
+    char* m_growp;  // Resize limit pointer
+#else
+    char* m_writep = m_owner.m_writep;  // Write pointer into output buffer
+    char* const m_wrFlushp = m_owner.m_wrFlushp;  // Output buffer flush trigger location
+#endif
+
+    // VCD line end string codes + metadata
+    const char* const m_suffixes = m_owner.m_suffixes.data();
+    // The maximum number of bytes a single signal can emit
+    const size_t m_maxSignalBytes = m_owner.m_maxSignalBytes;
+
+    void finishLine(uint32_t code, char* writep);
+
+#ifdef VL_TRACE_PARALLEL
+    void adjustGrowp() {
+        m_growp = (m_bufp + m_size) - (2 * m_maxSignalBytes);
+        assert(m_growp >= m_bufp + m_maxSignalBytes);
+    }
+#endif
+
+public:
+    // CONSTRUCTOR
+#ifdef VL_TRACE_PARALLEL
+    explicit VerilatedVcdBuffer(VerilatedVcd& owner, char* bufp, size_t size);
+#else
+    explicit VerilatedVcdBuffer(VerilatedVcd& owner);
+#endif
+    ~VerilatedVcdBuffer() = default;
+
+    //=========================================================================
+    // Implementation of VerilatedTraceBuffer interface
+
+    // Implementations of duck-typed methods for VerilatedTraceBuffer. These are
+    // called from only one place (the full* methods), so always inline them.
+    VL_ATTR_ALWINLINE inline void emitBit(uint32_t code, CData newval);
+    VL_ATTR_ALWINLINE inline void emitCData(uint32_t code, CData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitSData(uint32_t code, SData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitIData(uint32_t code, IData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitQData(uint32_t code, QData newval, int bits);
+    VL_ATTR_ALWINLINE inline void emitWData(uint32_t code, const WData* newvalp, int bits);
+    VL_ATTR_ALWINLINE inline void emitDouble(uint32_t code, double newval);
+};

 //=============================================================================
 // VerilatedFile
@@ -52,284 +234,6 @@ public:
    virtual ssize_t write(const char* bufp, ssize_t len) VL_MT_UNSAFE;
 };

-//=============================================================================
-// VerilatedVcd
-// Base class to create a Verilator VCD dump
-// This is an internally used class - see VerilatedVcdC for what to call from applications
-
-class VerilatedVcd VL_NOT_FINAL : public VerilatedTrace<VerilatedVcd> {
-private:
-    // Give the superclass access to private bits (to avoid virtual functions)
-    friend class VerilatedTrace<VerilatedVcd>;
-
-    //=========================================================================
-    // VCD specific internals
-
-    VerilatedVcdFile* m_filep;  // File we're writing to
-    bool m_fileNewed;  // m_filep needs destruction
-    bool m_isOpen = false;  // True indicates open file
-    bool m_evcd = false;  // True for evcd format
-    std::string m_filename;  // Filename we're writing to (if open)
-    vluint64_t m_rolloverMB = 0;  // MB of file size to rollover at
-    int m_modDepth = 0;  // Depth of module hierarchy
-
-    char* m_wrBufp;  // Output buffer
-    const char* m_wrFlushp;  // Output buffer flush trigger location
-    char* m_writep;  // Write pointer into output buffer
-    vluint64_t m_wrChunkSize;  // Output buffer size
-    vluint64_t m_wroteBytes = 0;  // Number of bytes written to this file
-
-    std::vector<char> m_suffixes;  // VCD line end string codes + metadata
-
-    using NameMap = std::map<const std::string, const std::string>;
-    NameMap* m_namemapp = nullptr;  // List of names for the header
-
-    void bufferResize(vluint64_t minsize);
-    void bufferFlush() VL_MT_UNSAFE_ONE;
-    inline void bufferCheck() {
-        // Flush the write buffer if there's not enough space left for new information
-        // We only call this once per vector, so we need enough slop for a very wide "b###" line
-        if (VL_UNLIKELY(m_writep > m_wrFlushp)) bufferFlush();
-    }
-    void openNextImp(bool incFilename);
-    void closePrev();
-    void closeErr();
-    void makeNameMap();
-    void deleteNameMap();
-    void printIndent(int level_change);
-    void printStr(const char* str);
-    void printQuad(vluint64_t n);
-    void printTime(vluint64_t timeui);
-    void declare(vluint32_t code, const char* name, const char* wirep, bool array, int arraynum,
-                 bool tri, bool bussed, int msb, int lsb);
-
-    void dumpHeader();
-
-    static char* writeCode(char* writep, vluint32_t code);
-
-    void finishLine(vluint32_t code, char* writep);
-
-    // CONSTRUCTORS
-    VL_UNCOPYABLE(VerilatedVcd);
-
-protected:
-    //=========================================================================
-    // Implementation of VerilatedTrace interface
-
-    // Implementations of protected virtual methods for VerilatedTrace
-    virtual void emitTimeChange(vluint64_t timeui) override;
-
-    // Hooks called from VerilatedTrace
-    virtual bool preFullDump() override { return isOpen(); }
-    virtual bool preChangeDump() override;
-
-    // Implementations of duck-typed methods for VerilatedTrace. These are
-    // called from only one place (namely full*) so always inline them.
-    inline void emitBit(vluint32_t code, CData newval);
-    inline void emitCData(vluint32_t code, CData newval, int bits);
-    inline void emitSData(vluint32_t code, SData newval, int bits);
-    inline void emitIData(vluint32_t code, IData newval, int bits);
-    inline void emitQData(vluint32_t code, QData newval, int bits);
-    inline void emitWData(vluint32_t code, const WData* newvalp, int bits);
-    inline void emitDouble(vluint32_t code, double newval);
-
-public:
-    //=========================================================================
-    // External interface to client code
-
-    explicit VerilatedVcd(VerilatedVcdFile* filep = nullptr);
-    ~VerilatedVcd();
-
-    // ACCESSORS
-    // Set size in megabytes after which new file should be created
-    void rolloverMB(vluint64_t rolloverMB) { m_rolloverMB = rolloverMB; }
-
-    // METHODS
-    // Open the file; call isOpen() to see if errors
-    void open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex);
-    // Open next data-only file
-    void openNext(bool incFilename) VL_MT_SAFE_EXCLUDES(m_mutex);
-    // Close the file
-    void close() VL_MT_SAFE_EXCLUDES(m_mutex);
-    // Flush any remaining data to this file
-    void flush() VL_MT_SAFE_EXCLUDES(m_mutex);
-    // Return if file is open
-    bool isOpen() const VL_MT_SAFE { return m_isOpen; }
-
-    //=========================================================================
-    // Internal interface to Verilator generated code
-
-    void declBit(vluint32_t code, const char* name, bool array, int arraynum);
-    void declBus(vluint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
-    void declQuad(vluint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
-    void declArray(vluint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
-    void declDouble(vluint32_t code, const char* name, bool array, int arraynum);
-
-#ifdef VL_TRACE_VCD_OLD_API
-    //=========================================================================
-    // Note: These are only for testing for backward compatibility with foreign
-    // code and is not used by Verilator. Do not use these as there is no
-    // guarantee of functionality.
-
-    void declTriBit(vluint32_t code, const char* name, bool array, int arraynum);
-    void declTriBus(vluint32_t code, const char* name, bool array, int arraynum, int msb, int lsb);
-    void declTriQuad(vluint32_t code, const char* name, bool array, int arraynum, int msb,
-                     int lsb);
-    void declTriArray(vluint32_t code, const char* name, bool array, int arraynum, int msb,
-                      int lsb);
-
-    void fullBit(vluint32_t* oldp, CData newval) { fullBit(oldp - this->oldp(0), newval); }
-    void fullCData(vluint32_t* oldp, CData newval, int bits) {
-        fullBus(oldp - this->oldp(0), newval, bits);
-    }
-    void fullSData(vluint32_t* oldp, SData newval, int bits) {
-        fullBus(oldp - this->oldp(0), newval, bits);
-    }
-    void fullIData(vluint32_t* oldp, IData newval, int bits) {
-        fullBus(oldp - this->oldp(0), newval, bits);
-    }
-    void fullQData(vluint32_t* oldp, QData newval, int bits) {
-        fullQuad(oldp - this->oldp(0), newval, bits);
-    }
-    void fullWData(vluint32_t* oldp, const WData* newvalp, int bits) {
-        fullArray(oldp - this->oldp(0), newvalp, bits);
-    }
-    void fullDouble(vluint32_t* oldp, double newval) { fullDouble(oldp - this->oldp(0), newval); }
-
-    inline void chgBit(vluint32_t* oldp, CData newval) { chgBit(oldp - this->oldp(0), newval); }
-    inline void chgCData(vluint32_t* oldp, CData newval, int bits) {
-        chgBus(oldp - this->oldp(0), newval, bits);
-    }
-    inline void chgSData(vluint32_t* oldp, SData newval, int bits) {
-        chgBus(oldp - this->oldp(0), newval, bits);
-    }
-    inline void chgIData(vluint32_t* oldp, IData newval, int bits) {
-        chgBus(oldp - this->oldp(0), newval, bits);
-    }
-    inline void chgQData(vluint32_t* oldp, QData newval, int bits) {
-        chgQuad(oldp - this->oldp(0), newval, bits);
-    }
-    inline void chgWData(vluint32_t* oldp, const WData* newvalp, int bits) {
-        chgArray(oldp - this->oldp(0), newvalp, bits);
-    }
-    inline void chgDouble(vluint32_t* oldp, double newval) {
-        chgDouble(oldp - this->oldp(0), newval);
-    }
-
-    // Inside dumping routines, dump one signal, faster when not inlined
-    // due to code size reduction.
-    void fullBit(vluint32_t code, const vluint32_t newval);
-    void fullBus(vluint32_t code, const vluint32_t newval, int bits);
-    void fullQuad(vluint32_t code, const vluint64_t newval, int bits);
-    void fullArray(vluint32_t code, const vluint32_t* newvalp, int bits);
-    void fullArray(vluint32_t code, const vluint64_t* newvalp, int bits);
-    void fullTriBit(vluint32_t code, const vluint32_t newval, const vluint32_t newtri);
-    void fullTriBus(vluint32_t code, const vluint32_t newval, const vluint32_t newtri, int bits);
-    void fullTriQuad(vluint32_t code, const vluint64_t newval, const vluint64_t newtri, int bits);
-    void fullTriArray(vluint32_t code, const vluint32_t* newvalp, const vluint32_t* newtrip,
-                      int bits);
-    void fullDouble(vluint32_t code, const double newval);
-
-    // Inside dumping routines, dump one signal if it has changed.
-    // We do want to inline these to avoid calls when the value did not change.
-    inline void chgBit(vluint32_t code, const vluint32_t newval) {
-        const vluint32_t diff = oldp(code)[0] ^ newval;
-        if (VL_UNLIKELY(diff)) fullBit(code, newval);
-    }
-    inline void chgBus(vluint32_t code, const vluint32_t newval, int bits) {
-        const vluint32_t diff = oldp(code)[0] ^ newval;
-        if (VL_UNLIKELY(diff)) {
-            if (VL_UNLIKELY(bits == 32 || (diff & ((1U << bits) - 1)))) {
-                fullBus(code, newval, bits);
-            }
-        }
-    }
-    inline void chgQuad(vluint32_t code, const vluint64_t newval, int bits) {
-        const vluint64_t diff = (*(reinterpret_cast<vluint64_t*>(oldp(code)))) ^ newval;
-        if (VL_UNLIKELY(diff)) {
-            if (VL_UNLIKELY(bits == 64 || (diff & ((1ULL << bits) - 1)))) {
-                fullQuad(code, newval, bits);
-            }
-        }
-    }
-    inline void chgArray(vluint32_t code, const vluint32_t* newvalp, int bits) {
-        for (int word = 0; word < (((bits - 1) / 32) + 1); ++word) {
-            if (VL_UNLIKELY(oldp(code)[word] ^ newvalp[word])) {
-                fullArray(code, newvalp, bits);
-                return;
-            }
-        }
-    }
-    inline void chgArray(vluint32_t code, const vluint64_t* newvalp, int bits) {
-        for (int word = 0; word < (((bits - 1) / 64) + 1); ++word) {
-            if (VL_UNLIKELY(*(reinterpret_cast<vluint64_t*>(oldp(code + 2 * word)))
-                            ^ newvalp[word])) {
-                fullArray(code, newvalp, bits);
-                return;
-            }
-        }
-    }
-    inline void chgTriBit(vluint32_t code, const vluint32_t newval, const vluint32_t newtri) {
-        const vluint32_t diff = ((oldp(code)[0] ^ newval) | (oldp(code)[1] ^ newtri));
-        if (VL_UNLIKELY(diff)) {
-            // Verilator 3.510 and newer provide clean input, so the below
-            // is only for back compatibility
-            if (VL_UNLIKELY(diff & 1)) {  // Change after clean?
-                fullTriBit(code, newval, newtri);
-            }
-        }
-    }
-    inline void chgTriBus(vluint32_t code, const vluint32_t newval, const vluint32_t newtri,
-                          int bits) {
-        const vluint32_t diff = ((oldp(code)[0] ^ newval) | (oldp(code)[1] ^ newtri));
-        if (VL_UNLIKELY(diff)) {
-            if (VL_UNLIKELY(bits == 32 || (diff & ((1U << bits) - 1)))) {
-                fullTriBus(code, newval, newtri, bits);
-            }
-        }
-    }
-    inline void chgTriQuad(vluint32_t code, const vluint64_t newval, const vluint64_t newtri,
-                           int bits) {
-        const vluint64_t diff = (((*(reinterpret_cast<vluint64_t*>(oldp(code)))) ^ newval)
-                                 | ((*(reinterpret_cast<vluint64_t*>(oldp(code + 1)))) ^ newtri));
-        if (VL_UNLIKELY(diff)) {
-            if (VL_UNLIKELY(bits == 64 || (diff & ((1ULL << bits) - 1)))) {
-                fullTriQuad(code, newval, newtri, bits);
-            }
-        }
-    }
-    inline void chgTriArray(vluint32_t code, const vluint32_t* newvalp, const vluint32_t* newtrip,
-                            int bits) {
-        for (int word = 0; word < (((bits - 1) / 32) + 1); ++word) {
-            if (VL_UNLIKELY((oldp(code)[word * 2] ^ newvalp[word])
-                            | (oldp(code)[word * 2 + 1] ^ newtrip[word]))) {
-                fullTriArray(code, newvalp, newtrip, bits);
-                return;
-            }
-        }
-    }
-    inline void chgDouble(vluint32_t code, const double newval) {
-        // cppcheck-suppress invalidPointerCast
-        if (VL_UNLIKELY((*(reinterpret_cast<double*>(oldp(code)))) != newval)) {
-            fullDouble(code, newval);
-        }
-    }
-
-    // METHODS
-    // Old/standalone API only
-    void evcd(bool flag) { m_evcd = flag; }
-#endif  // VL_TRACE_VCD_OLD_API
-};
-
-#ifndef DOXYGEN
-// Declare specializations here they are used in VerilatedVcdC just below
-template <> void VerilatedTrace<VerilatedVcd>::dump(vluint64_t timeui);
-template <> void VerilatedTrace<VerilatedVcd>::set_time_unit(const char* unitp);
-template <> void VerilatedTrace<VerilatedVcd>::set_time_unit(const std::string& unit);
-template <> void VerilatedTrace<VerilatedVcd>::set_time_resolution(const char* unitp);
-template <> void VerilatedTrace<VerilatedVcd>::set_time_resolution(const std::string& unit);
-#endif  // DOXYGEN
-
 //=============================================================================
 // VerilatedVcdC
 /// Class representing a VCD dump file in C standalone (no SystemC)
@@ -346,7 +250,7 @@ public:
    explicit VerilatedVcdC(VerilatedVcdFile* filep = nullptr)
        : m_sptrace{filep} {}
    /// Destruct, flush, and close the dump
-    ~VerilatedVcdC() { close(); }
+    virtual ~VerilatedVcdC() { close(); }

 public:
    // METHODS - User called
@@ -356,7 +260,7 @@ public:
    /// Open a new VCD file
    /// This includes a complete header dump each time it is called,
    /// just as if this object was deleted and reconstructed.
-    void open(const char* filename) VL_MT_SAFE { m_sptrace.open(filename); }
+    virtual void open(const char* filename) VL_MT_SAFE { m_sptrace.open(filename); }
    /// Continue a VCD dump by rotating to a new file name
    /// The header is only in the first file created, this allows
    /// "cat" to be used to combine the header plus any number of data files.
@@ -370,12 +274,12 @@ public:
    /// Write one cycle of dump data
    /// Call with the current context's time just after eval'ed,
    /// e.g. ->dump(contextp->time())
-    void dump(vluint64_t timeui) VL_MT_SAFE { m_sptrace.dump(timeui); }
+    void dump(uint64_t timeui) VL_MT_SAFE { m_sptrace.dump(timeui); }
    /// Write one cycle of dump data - backward compatible and to reduce
-    /// conversion warnings.  It's better to use a vluint64_t time instead.
-    void dump(double timestamp) { dump(static_cast<vluint64_t>(timestamp)); }
-    void dump(vluint32_t timestamp) { dump(static_cast<vluint64_t>(timestamp)); }
-    void dump(int timestamp) { dump(static_cast<vluint64_t>(timestamp)); }
+    /// conversion warnings.  It's better to use a uint64_t time instead.
+    void dump(double timestamp) { dump(static_cast<uint64_t>(timestamp)); }
+    void dump(uint32_t timestamp) { dump(static_cast<uint64_t>(timestamp)); }
+    void dump(int timestamp) { dump(static_cast<uint64_t>(timestamp)); }

    // METHODS - Internal/backward compatible
    // \protectedsection
@@ -392,19 +296,14 @@ public:
    void set_time_resolution(const std::string& unit) VL_MT_SAFE {
        m_sptrace.set_time_resolution(unit);
    }
+    // Set variables to dump, using $dumpvars format
+    // If level = 0, dump everything and hier is then ignored
+    void dumpvars(int level, const std::string& hier) VL_MT_SAFE {
+        m_sptrace.dumpvars(level, hier);
+    }

    // Internal class access
    inline VerilatedVcd* spTrace() { return &m_sptrace; }
-
-#ifdef VL_TRACE_VCD_OLD_API
-    //=========================================================================
-    // Note: These are only for testing for backward compatibility with foreign
-    // code and is not used by Verilator. Do not use these as there is no
-    // guarantee of functionality.
-
-    // Use evcd format
-    void evcd(bool flag) VL_MT_UNSAFE_ONE { m_sptrace.evcd(flag); }
-#endif
 };

 #endif  // guard
@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you
 // can redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -27,6 +27,17 @@
 //======================================================================
 //======================================================================

+void VerilatedVcdSc::open(const char* filename) {
+    if (!sc_core::sc_get_curr_simcontext()->elaboration_done()) {
+        vl_fatal(__FILE__, __LINE__, "VerilatedVcdSc",
+                 ("%Error: VerilatedVcdSc::open(\"" + std::string(filename)
+                  + "\") is called before sc_core::sc_start(). "
+                    "Run sc_core::sc_start(sc_core::SC_ZERO_TIME) before opening a wave file.")
+                     .c_str());
+    }
+    VerilatedVcdC::open(filename);
+}
+
 //--------------------------------------------------
 // SystemC 2.1.v1
 // cppcheck-suppress unusedFunction
@@ -1,7 +1,7 @@
 // -*- mode: C++; c-file-style: "cc-mode" -*-
 //=============================================================================
 //
-// Copyright 2001-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2001-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -26,6 +26,8 @@
 #include "verilated_sc.h"
 #include "verilated_vcd_c.h"

+#include <string>
+
 //=============================================================================
 // VerilatedVcdSc
 ///
@@ -60,6 +62,9 @@ public:
        if (!delta_cycle) this->dump(sc_time_stamp().to_double());
    }

+    // Override VerilatedVcdC. Must be called after starting simulation.
+    virtual void open(const char* filename) /*override*/ VL_MT_SAFE;
+
 private:
    // METHODS - Fake outs for linker

@@ -3,7 +3,7 @@
 //
 // Code available from: https://verilator.org
 //
-// Copyright 2009-2021 by Wilson Snyder. This program is free software; you can
+// Copyright 2009-2022 by Wilson Snyder. This program is free software; you can
 // redistribute it and/or modify it under the terms of either the GNU
 // Lesser General Public License Version 3 or the Perl Artistic License
 // Version 2.0.
@@ -33,6 +33,10 @@
 #include <list>
 #include <map>
 #include <set>
+#include <string>
+#include <unordered_set>
+#include <utility>
+#include <vector>

 //======================================================================
 // Internal constants
@@ -62,11 +66,11 @@ class VerilatedVpio VL_NOT_FINAL {
    // CONSTANTS
    // Magic value stored in front of object to detect double free etc
    // Must be odd, as aligned pointer can never be odd
-    static constexpr vluint32_t activeMagic() { return 0xfeed100f; }
+    static constexpr uint32_t activeMagic() { return 0xfeed100f; }

    // MEM MANGLEMENT
    // Internal note: Globals may multi-construct, see verilated.cpp top.
-    static VL_THREAD_LOCAL vluint8_t* t_freeHead;
+    static VL_THREAD_LOCAL uint8_t* t_freeHead;

 public:
    // CONSTRUCTORS
@@ -77,22 +81,23 @@ public:
        // To simplify our free list, we use a size large enough for all derived types
        // We reserve word zero for the next pointer, as that's safer in case a
        // dangling reference to the original remains around.
-        static const size_t chunk = 96;
-        if (VL_UNCOVERABLE(size > chunk)) VL_FATAL_MT(__FILE__, __LINE__, "", "increase chunk");
+        static constexpr size_t CHUNK_SIZE = 96;
+        if (VL_UNCOVERABLE(size > CHUNK_SIZE))
+            VL_FATAL_MT(__FILE__, __LINE__, "", "increase CHUNK_SIZE");
        if (VL_LIKELY(t_freeHead)) {
-            vluint8_t* const newp = t_freeHead;
-            t_freeHead = *(reinterpret_cast<vluint8_t**>(newp));
-            *(reinterpret_cast<vluint32_t*>(newp)) = activeMagic();
+            uint8_t* const newp = t_freeHead;
+            t_freeHead = *(reinterpret_cast<uint8_t**>(newp));
+            *(reinterpret_cast<uint32_t*>(newp)) = activeMagic();
            return newp + 8;
        }
        // +8: 8 bytes for next
-        vluint8_t* newp = reinterpret_cast<vluint8_t*>(::operator new(chunk + 8));
-        *(reinterpret_cast<vluint32_t*>(newp)) = activeMagic();
+        uint8_t* newp = reinterpret_cast<uint8_t*>(::operator new(CHUNK_SIZE + 8));
+        *(reinterpret_cast<uint32_t*>(newp)) = activeMagic();
        return newp + 8;
    }
    static void operator delete(void* obj, size_t /*size*/)VL_MT_SAFE {
-        vluint8_t* const oldp = (static_cast<vluint8_t*>(obj)) - 8;
-        if (VL_UNLIKELY(*(reinterpret_cast<vluint32_t*>(oldp)) != activeMagic())) {
+        uint8_t* const oldp = (static_cast<uint8_t*>(obj)) - 8;
+        if (VL_UNLIKELY(*(reinterpret_cast<uint32_t*>(oldp)) != activeMagic())) {
            VL_FATAL_MT(__FILE__, __LINE__, "",
                        "vpi_release_handle() called on same object twice, or on non-Verilator "
                        "VPI object");
@@ -113,8 +118,8 @@ public:
    virtual const char* name() const { return "<null>"; }
    virtual const char* fullname() const { return "<null>"; }
    virtual const char* defname() const { return "<null>"; }
-    virtual vluint32_t type() const { return 0; }
-    virtual vluint32_t size() const { return 0; }
+    virtual uint32_t type() const { return 0; }
+    virtual uint32_t size() const { return 0; }
    virtual const VerilatedRange* rangep() const { return nullptr; }
    virtual vpiHandle dovpi_scan() { return nullptr; }
    virtual PLI_INT32 dovpi_remove_cb() { return 0; }
@@ -123,52 +128,52 @@ public:
 class VerilatedVpioTimedCb final : public VerilatedVpio {
    // A handle to a timed callback created with vpi_register_cb
    // User can call vpi_remove_cb or vpi_release_handle on it
-    const vluint64_t m_id;  // Unique id/sequence number to find schedule's event
+    const uint64_t m_id;  // Unique id/sequence number to find schedule's event
    const QData m_time;

 public:
-    VerilatedVpioTimedCb(vluint64_t id, QData time)
+    VerilatedVpioTimedCb(uint64_t id, QData time)
        : m_id{id}
        , m_time{time} {}
    virtual ~VerilatedVpioTimedCb() override = default;
    static VerilatedVpioTimedCb* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioTimedCb*>(reinterpret_cast<VerilatedVpioTimedCb*>(h));
    }
-    virtual vluint32_t type() const override { return vpiCallback; }
+    virtual uint32_t type() const override { return vpiCallback; }
    virtual PLI_INT32 dovpi_remove_cb() override;
 };

 class VerilatedVpioReasonCb final : public VerilatedVpio {
    // A handle to a non-timed callback created with vpi_register_cb
    // User can call vpi_remove_cb or vpi_release_handle on it
-    const vluint64_t m_id;  // Unique id/sequence number to find schedule's event
+    const uint64_t m_id;  // Unique id/sequence number to find schedule's event
    const PLI_INT32 m_reason;  // VPI callback reason code

 public:
    // cppcheck-suppress uninitVar  // m_value
-    VerilatedVpioReasonCb(vluint64_t id, PLI_INT32 reason)
+    VerilatedVpioReasonCb(uint64_t id, PLI_INT32 reason)
        : m_id{id}
        , m_reason{reason} {}
    virtual ~VerilatedVpioReasonCb() override = default;
    static VerilatedVpioReasonCb* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioReasonCb*>(reinterpret_cast<VerilatedVpioReasonCb*>(h));
    }
-    virtual vluint32_t type() const override { return vpiCallback; }
+    virtual uint32_t type() const override { return vpiCallback; }
    virtual PLI_INT32 dovpi_remove_cb() override;
 };

 class VerilatedVpioConst final : public VerilatedVpio {
-    const vlsint32_t m_num;
+    const int32_t m_num;

 public:
-    explicit VerilatedVpioConst(vlsint32_t num)
+    explicit VerilatedVpioConst(int32_t num)
        : m_num{num} {}
    virtual ~VerilatedVpioConst() override = default;
    static VerilatedVpioConst* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioConst*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiConstant; }
-    vlsint32_t num() const { return m_num; }
+    virtual uint32_t type() const override { return vpiConstant; }
+    int32_t num() const { return m_num; }
 };

 class VerilatedVpioVarBase VL_NOT_FINAL : public VerilatedVpio {
@@ -195,7 +200,7 @@ public:
    }
    const VerilatedVar* varp() const { return m_varp; }
    const VerilatedScope* scopep() const { return m_scopep; }
-    virtual vluint32_t size() const override { return get_range().elements(); }
+    virtual uint32_t size() const override { return get_range().elements(); }
    virtual const VerilatedRange* rangep() const override { return &get_range(); }
    virtual const char* name() const override { return m_varp->name(); }
    virtual const char* fullname() const override {
@@ -214,7 +219,7 @@ public:
    static VerilatedVpioParam* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioParam*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiParameter; }
+    virtual uint32_t type() const override { return vpiParameter; }
    void* varDatap() const { return m_varp->datap(); }
 };

@@ -228,8 +233,8 @@ public:
    static VerilatedVpioRange* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioRange*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiRange; }
-    virtual vluint32_t size() const override { return m_range->elements(); }
+    virtual uint32_t type() const override { return vpiRange; }
+    virtual uint32_t size() const override { return m_range->elements(); }
    virtual const VerilatedRange* rangep() const override { return m_range; }
 };

@@ -245,7 +250,7 @@ public:
    static VerilatedVpioRangeIter* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioRangeIter*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiIterator; }
+    virtual uint32_t type() const override { return vpiIterator; }
    virtual vpiHandle dovpi_scan() override {
        if (VL_UNLIKELY(m_done)) {
            delete this;  // IEEE 37.2.2 vpi_scan at end does a vpi_release_handle
@@ -267,22 +272,22 @@ public:
    static VerilatedVpioScope* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioScope*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiScope; }
+    virtual uint32_t type() const override { return vpiScope; }
    const VerilatedScope* scopep() const { return m_scopep; }
    virtual const char* name() const override { return m_scopep->name(); }
    virtual const char* fullname() const override { return m_scopep->name(); }
 };

 class VerilatedVpioVar VL_NOT_FINAL : public VerilatedVpioVarBase {
-    vluint8_t* m_prevDatap = nullptr;  // Previous value of data, for cbValueChange
+    uint8_t* m_prevDatap = nullptr;  // Previous value of data, for cbValueChange
    union {
-        vluint8_t u8[4];
-        vluint32_t u32;
+        uint8_t u8[4];
+        uint32_t u32;
    } m_mask;  // memoized variable mask
-    vluint32_t m_entSize = 0;  // memoized variable size
+    uint32_t m_entSize = 0;  // memoized variable size
 protected:
    void* m_varDatap = nullptr;  // varp()->datap() adjusted for array entries
-    vlsint32_t m_index = 0;
+    int32_t m_index = 0;

 public:
    VerilatedVpioVar(const VerilatedVar* varp, const VerilatedScope* scopep)
@@ -309,18 +314,18 @@ public:
    static VerilatedVpioVar* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioVar*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    vluint32_t mask() const { return m_mask.u32; }
-    vluint8_t mask_byte(int idx) const { return m_mask.u8[idx & 3]; }
-    vluint32_t entSize() const { return m_entSize; }
-    vluint32_t index() const { return m_index; }
-    virtual vluint32_t type() const override {
+    uint32_t mask() const { return m_mask.u32; }
+    uint8_t mask_byte(int idx) const { return m_mask.u8[idx & 3]; }
+    uint32_t entSize() const { return m_entSize; }
+    uint32_t index() const { return m_index; }
+    virtual uint32_t type() const override {
        return (varp()->dims() > 1) ? vpiMemory : vpiReg;  // but might be wire, logic
    }
    void* prevDatap() const { return m_prevDatap; }
    void* varDatap() const { return m_varDatap; }
    void createPrevDatap() {
        if (VL_UNLIKELY(!m_prevDatap)) {
-            m_prevDatap = new vluint8_t[entSize()];
+            m_prevDatap = new uint8_t[entSize()];
            std::memcpy(prevDatap(), varp()->datap(), entSize());
        }
    }
@@ -328,18 +333,18 @@ public:

 class VerilatedVpioMemoryWord final : public VerilatedVpioVar {
 public:
-    VerilatedVpioMemoryWord(const VerilatedVar* varp, const VerilatedScope* scopep,
-                            vlsint32_t index, int offset)
+    VerilatedVpioMemoryWord(const VerilatedVar* varp, const VerilatedScope* scopep, int32_t index,
+                            int offset)
        : VerilatedVpioVar{varp, scopep} {
        m_index = index;
-        m_varDatap = (static_cast<vluint8_t*>(varp->datap())) + entSize() * offset;
+        m_varDatap = (static_cast<uint8_t*>(varp->datap())) + entSize() * offset;
    }
    virtual ~VerilatedVpioMemoryWord() override = default;
    static VerilatedVpioMemoryWord* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioMemoryWord*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiMemoryWord; }
-    virtual vluint32_t size() const override { return varp()->packed().elements(); }
+    virtual uint32_t type() const override { return vpiMemoryWord; }
+    virtual uint32_t size() const override { return varp()->packed().elements(); }
    virtual const VerilatedRange* rangep() const override { return &(varp()->packed()); }
    virtual const char* fullname() const override {
        static VL_THREAD_LOCAL std::string t_out;
@@ -363,7 +368,7 @@ public:
    static VerilatedVpioVarIter* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioVarIter*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiIterator; }
+    virtual uint32_t type() const override { return vpiIterator; }
    virtual vpiHandle dovpi_scan() override {
        if (VL_LIKELY(m_scopep->varsp())) {
            const VerilatedVarNameMap* const varsp = m_scopep->varsp();
@@ -390,8 +395,8 @@ public:
 class VerilatedVpioMemoryWordIter final : public VerilatedVpio {
    const vpiHandle m_handle;
    const VerilatedVar* const m_varp;
-    vlsint32_t m_iteration;
-    const vlsint32_t m_direction;
+    int32_t m_iteration;
+    const int32_t m_direction;
    bool m_done = false;

 public:
@@ -404,7 +409,7 @@ public:
    static VerilatedVpioMemoryWordIter* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioMemoryWordIter*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiIterator; }
+    virtual uint32_t type() const override { return vpiIterator; }
    void iterationInc() {
        if (!(m_done = (m_iteration == m_varp->unpacked().left()))) m_iteration += m_direction;
    }
@@ -433,7 +438,7 @@ public:
    static VerilatedVpioModule* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioModule*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiModule; }
+    virtual uint32_t type() const override { return vpiModule; }
    virtual const char* name() const override { return m_name; }
    virtual const char* fullname() const override { return m_fullname; }
 };
@@ -451,7 +456,7 @@ public:
    static VerilatedVpioModuleIter* castp(vpiHandle h) {
        return dynamic_cast<VerilatedVpioModuleIter*>(reinterpret_cast<VerilatedVpio*>(h));
    }
-    virtual vluint32_t type() const override { return vpiIterator; }
+    virtual uint32_t type() const override { return vpiIterator; }
    virtual vpiHandle dovpi_scan() override {
        if (m_it == m_vec->end()) {
            delete this;  // IEEE 37.2.2 vpi_scan at end does a vpi_release_handle
@@ -468,14 +473,14 @@ using VerilatedPliCb = PLI_INT32 (*)(struct t_cb_data*);

 class VerilatedVpiCbHolder final {
    // Holds information needed to call a callback
-    vluint64_t m_id;
+    uint64_t m_id;
    s_cb_data m_cbData;
    s_vpi_value m_value;
    VerilatedVpioVar m_varo;  // If a cbValueChange callback, the object we will return

 public:
    // cppcheck-suppress uninitVar  // m_value
-    VerilatedVpiCbHolder(vluint64_t id, const s_cb_data* cbDatap, const VerilatedVpioVar* varop)
+    VerilatedVpiCbHolder(uint64_t id, const s_cb_data* cbDatap, const VerilatedVpioVar* varop)
        : m_id{id}
        , m_cbData{*cbDatap}
        , m_varo{varop} {
@@ -491,15 +496,15 @@ public:
    ~VerilatedVpiCbHolder() = default;
    VerilatedPliCb cb_rtnp() const { return m_cbData.cb_rtn; }
    s_cb_data* cb_datap() { return &m_cbData; }
-    vluint64_t id() const { return m_id; }
+    uint64_t id() const { return m_id; }
    bool invalid() const { return !m_id; }
    void invalidate() { m_id = 0; }
 };

 struct VerilatedVpiTimedCbsCmp {
    // Ordering sets keyed by time, then callback unique id
-    bool operator()(const std::pair<QData, vluint64_t>& a,
-                    const std::pair<QData, vluint64_t>& b) const {
+    bool operator()(const std::pair<QData, uint64_t>& a,
+                    const std::pair<QData, uint64_t>& b) const {
        if (a.first < b.first) return true;
        if (a.first > b.first) return false;
        return a.second < b.second;
@@ -511,14 +516,14 @@ class VerilatedVpiError;
 class VerilatedVpiImp final {
    enum { CB_ENUM_MAX_VALUE = cbAtEndOfSimTime + 1 };  // Maxium callback reason
    using VpioCbList = std::list<VerilatedVpiCbHolder>;
-    using VpioTimedCbs = std::map<std::pair<QData, vluint64_t>, VerilatedVpiCbHolder>;
+    using VpioTimedCbs = std::map<std::pair<QData, uint64_t>, VerilatedVpiCbHolder>;

    // All only medium-speed, so use singleton function
    VpioCbList m_cbObjLists[CB_ENUM_MAX_VALUE];  // Callbacks for each supported reason
    VpioTimedCbs m_timedCbs;  // Time based callbacks
    VerilatedVpiError* m_errorInfop = nullptr;  // Container for vpi error info
    VerilatedAssertOneThread m_assertOne;  // Assert only called from single thread
-    vluint64_t m_nextCallbackId = 1;  // Id to identify callback
+    uint64_t m_nextCallbackId = 1;  // Id to identify callback

    static VerilatedVpiImp& s() {  // Singleton
        static VerilatedVpiImp s_s;
@@ -527,29 +532,29 @@ class VerilatedVpiImp final {

 public:
    static void assertOneCheck() { s().m_assertOne.check(); }
-    static vluint64_t nextCallbackId() { return ++s().m_nextCallbackId; }
+    static uint64_t nextCallbackId() { return ++s().m_nextCallbackId; }

-    static void cbReasonAdd(vluint64_t id, const s_cb_data* cb_data_p) {
+    static void cbReasonAdd(uint64_t id, const s_cb_data* cb_data_p) {
        // The passed cb_data_p was property of the user, so need to recreate
        if (VL_UNCOVERABLE(cb_data_p->reason >= CB_ENUM_MAX_VALUE)) {
            VL_FATAL_MT(__FILE__, __LINE__, "", "vpi bb reason too large");
        }
-        VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: vpi_register_cb reason=%d id=%" VL_PRI64 "d obj=%p\n",
+        VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: vpi_register_cb reason=%d id=%" PRId64 " obj=%p\n",
                                    cb_data_p->reason, id, cb_data_p->obj););
        VerilatedVpioVar* varop = nullptr;
        if (cb_data_p->reason == cbValueChange) varop = VerilatedVpioVar::castp(cb_data_p->obj);
        s().m_cbObjLists[cb_data_p->reason].emplace_back(id, cb_data_p, varop);
    }
-    static void cbTimedAdd(vluint64_t id, const s_cb_data* cb_data_p, QData time) {
+    static void cbTimedAdd(uint64_t id, const s_cb_data* cb_data_p, QData time) {
        // The passed cb_data_p was property of the user, so need to recreate
-        VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: vpi_register_cb reason=%d id=%" VL_PRI64
-                                    "d delay=%" VL_PRI64 "u\n",
+        VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: vpi_register_cb reason=%d id=%" PRId64
+                                    " delay=%" PRIu64 "\n",
                                    cb_data_p->reason, id, time););
        s().m_timedCbs.emplace(std::piecewise_construct,
                               std::forward_as_tuple(std::make_pair(time, id)),
                               std::forward_as_tuple(id, cb_data_p, nullptr));
    }
-    static void cbReasonRemove(vluint64_t id, vluint32_t reason) {
+    static void cbReasonRemove(uint64_t id, uint32_t reason) {
        // Id might no longer exist, if already removed due to call after event, or teardown
        VpioCbList& cbObjList = s().m_cbObjLists[reason];
        // We do not remove it now as we may be iterating the list,
@@ -558,7 +563,7 @@ public:
            if (ir.id() == id) ir.invalidate();
        }
    }
-    static void cbTimedRemove(vluint64_t id, QData time) {
+    static void cbTimedRemove(uint64_t id, QData time) {
        // Id might no longer exist, if already removed due to call after event, or teardown
        const auto it = s().m_timedCbs.find(std::make_pair(time, id));
        if (VL_LIKELY(it != s().m_timedCbs.end())) it->second.invalidate();
@@ -573,7 +578,7 @@ public:
                ++it;
                if (VL_UNLIKELY(!ho.invalid())) {
                    VL_DEBUG_IF_PLI(
-                        VL_DBG_MSGF("- vpi: timed_callback id=%" VL_PRI64 "d\n", ho.id()););
+                        VL_DBG_MSGF("- vpi: timed_callback id=%" PRId64 "\n", ho.id()););
                    ho.invalidate();  // Timed callbacks are one-shot
                    (ho.cb_rtnp())(ho.cb_datap());
                }
@@ -588,7 +593,7 @@ public:
        if (VL_LIKELY(it != s().m_timedCbs.cend())) return it->first.first;
        return ~0ULL;  // maxquad
    }
-    static bool callCbs(const vluint32_t reason) VL_MT_UNSAFE_ONE {
+    static bool callCbs(const uint32_t reason) VL_MT_UNSAFE_ONE {
        VpioCbList& cbObjList = s().m_cbObjLists[reason];
        bool called = false;
        if (cbObjList.empty()) return called;
@@ -602,7 +607,7 @@ public:
                continue;
            }
            VerilatedVpiCbHolder& ho = *it;
-            VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: reason_callback reason=%d id=%" VL_PRI64 "d\n",
+            VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: reason_callback reason=%d id=%" PRId64 "\n",
                                        reason, ho.id()););
            (ho.cb_rtnp())(ho.cb_datap());
            called = true;
@@ -636,9 +641,9 @@ public:
                                            *(static_cast<CData*>(prevDatap)), newDatap,
                                            prevDatap););
                if (std::memcmp(prevDatap, newDatap, varop->entSize()) != 0) {
-                    VL_DEBUG_IF_PLI(
-                        VL_DBG_MSGF("- vpi: value_callback %" VL_PRI64 "d %s v[0]=%d\n", ho.id(),
-                                    varop->fullname(), *(static_cast<CData*>(newDatap))););
+                    VL_DEBUG_IF_PLI(VL_DBG_MSGF("- vpi: value_callback %" PRId64 " %s v[0]=%d\n",
+                                                ho.id(), varop->fullname(),
+                                                *(static_cast<CData*>(newDatap))););
                    update.insert(varop);
                    vpi_get_value(ho.cb_datap()->obj, ho.cb_datap()->value);
                    (ho.cb_rtnp())(ho.cb_datap());
@@ -660,7 +665,7 @@ public:
 // Statics
 // Internal note: Globals may multi-construct, see verilated.cpp top.

-VL_THREAD_LOCAL vluint8_t* VerilatedVpio::t_freeHead = nullptr;
+VL_THREAD_LOCAL uint8_t* VerilatedVpio::t_freeHead = nullptr;

 //======================================================================
 // VerilatedVpiError
@@ -740,7 +745,7 @@ void VerilatedVpi::callTimedCbs() VL_MT_UNSAFE_ONE { VerilatedVpiImp::callTimedC

 bool VerilatedVpi::callValueCbs() VL_MT_UNSAFE_ONE { return VerilatedVpiImp::callValueCbs(); }

-bool VerilatedVpi::callCbs(vluint32_t reason) VL_MT_UNSAFE_ONE {
+bool VerilatedVpi::callCbs(uint32_t reason) VL_MT_UNSAFE_ONE {
    return VerilatedVpiImp::callCbs(reason);
 }

@@ -1305,7 +1310,7 @@ vpiHandle vpi_register_cb(p_cb_data cb_data_p) {
        QData time = 0;
        if (cb_data_p->time) time = VL_SET_QII(cb_data_p->time->high, cb_data_p->time->low);
        const QData abstime = VL_TIME_Q() + time;
-        const vluint64_t id = VerilatedVpiImp::nextCallbackId();
+        const uint64_t id = VerilatedVpiImp::nextCallbackId();
        VerilatedVpioTimedCb* const vop = new VerilatedVpioTimedCb{id, abstime};
        VerilatedVpiImp::cbTimedAdd(id, cb_data_p, abstime);
        return vop->castVpiHandle();
@@ -1320,7 +1325,7 @@ vpiHandle vpi_register_cb(p_cb_data cb_data_p) {
    case cbEnterInteractive:  // FALLTHRU // NOP, but need to return handle, so make object
    case cbExitInteractive:  // FALLTHRU // NOP, but need to return handle, so make object
    case cbInteractiveScopeChange: {  // FALLTHRU // NOP, but need to return handle, so make object
-        const vluint64_t id = VerilatedVpiImp::nextCallbackId();
+        const uint64_t id = VerilatedVpiImp::nextCallbackId();
        VerilatedVpioReasonCb* const vop = new VerilatedVpioReasonCb{id, cb_data_p->reason};
        VerilatedVpiImp::cbReasonAdd(id, cb_data_p);
        return vop->castVpiHandle();
@@ -1470,7 +1475,7 @@ vpiHandle vpi_handle(PLI_INT32 type, vpiHandle object) {
    case vpiIndex: {
        const VerilatedVpioVar* const vop = VerilatedVpioVar::castp(object);
        if (VL_UNLIKELY(!vop)) return nullptr;
-        const vlsint32_t val = vop->index();
+        const int32_t val = vop->index();
        return (new VerilatedVpioConst{val})->castVpiHandle();
    }
    case vpiScope: {
@@ -1697,7 +1702,7 @@ void vl_get_value(const VerilatedVar* varp, void* varDatap, p_vpi_value valuep,
    // Maximum required size is for binary string, one byte per bit plus null termination
    static VL_THREAD_LOCAL char t_outStr[VL_VALUE_STRING_MAX_WORDS * VL_EDATASIZE + 1];
    // cppcheck-suppress variableScope
-    const static VL_THREAD_LOCAL int t_outStrSz = sizeof(t_outStr) - 1;
+    static const VL_THREAD_LOCAL int t_outStrSz = sizeof(t_outStr) - 1;
    // We used to presume vpiValue.format = vpiIntVal or if single bit vpiScalarVal
    // This may cause backward compatibility issues with older code.
    if (valuep->format == vpiVectorVal) {
@@ -1992,7 +1997,7 @@ vpiHandle vpi_put_value(vpiHandle object, p_vpi_value valuep, p_vpi_time /*time_
            for (int i = 0; i < chars; ++i) {
                union {
                    char byte[2];
-                    vluint16_t half;
+                    uint16_t half;
                } val;
                idx = div(i * 3, 8);
                if (i < len) {
--- a/Show More
+++ b/Show More