Version bump

Commentary: Changes update
Fix --hierarchical with order-based pin connections (#3585 ).
2026-10-04 09:03:31 +02:00 · 2022-08-31 19:46:45 -04:00 · 2022-08-31 19:27:43 -04:00 · 2022-08-31 18:12:21 -04:00 · 2022-08-30 06:33:42 -04:00 · 2022-08-30 01:02:39 -04:00
1449 changed files with 88769 additions and 42475 deletions
@@ -17,7 +17,7 @@ AllowShortLoopsOnASingleLine: true
 AlwaysBreakAfterDefinitionReturnType: None
 AlwaysBreakAfterReturnType: None
 AlwaysBreakBeforeMultilineStrings: false
-AlwaysBreakTemplateDeclarations: false
+AlwaysBreakTemplateDeclarations: true
 BinPackArguments: true
 BinPackParameters: true
 BraceWrapping:
@@ -59,15 +59,27 @@ ForEachMacros:
  - foreach
  - Q_FOREACH
  - BOOST_FOREACH
-IncludeBlocks:   Preserve
+
+# Include grouping/sorting
+SortIncludes:    true
+IncludeBlocks:   Regroup
 IncludeCategories:
-  - Regex:           '^"(llvm|llvm-c|clang|clang-c)/'
-    Priority:        2
-  - Regex:           '^(<|"(gtest|gmock|isl|json)/)'
-    Priority:        3
-  - Regex:           '.*'
-    Priority:        1
-IncludeIsMainRegex: '(Test)?$'
+  - Regex:           '"(config_build|verilated_config|verilatedos)\.h"'
+    Priority:        -1 # Sepecials before main header
+  - Regex:           '(<|")verilated.*'
+    Priority:        1 # Runtime headers
+  - Regex:           '"V3.*__gen.*\.h"'
+    Priority:        3 # Generated internal headers separately
+  - Regex:           '"V3.*"'
+    Priority:        2 # Internal header
+  - Regex:           '".*"'
+    Priority:        4 # Other non-system headers
+  - Regex:           '<[[:alnum:]_.]+>'
+    Priority:        5 # Simple system headers next
+  - Regex:           '<.*>'
+    Priority:        6 # Other system headers next
+IncludeIsMainRegex: '$'
+
 IndentCaseLabels: false
 IndentPPDirectives: None
 IndentWidth:     4
@@ -91,7 +103,6 @@ PenaltyExcessCharacter: 1000000
 PenaltyReturnTypeOnItsOwnLine: 60
 PointerAlignment: Left
 ReflowComments:  true
-SortIncludes:    false
 SortUsingDeclarations: true
 SpaceAfterCStyleCast: false
 SpaceAfterTemplateKeyword: true
@@ -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: ''
@@ -13,4 +13,6 @@ Can you attach an example that shows the issue?  (Must be openly licensed, ideal

 What 'verilator --version' are you using?  Did you try it with the git master version?

-Would you be willing to try to fix Verilator yourself with assistance?
+What OS and distribution are you using?
+
+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:
@@ -39,3 +39,4 @@ verilator-config-version.cmake
 **/__pycache__/*
 **/_build/*
 **/obj_dir/*
+/.vscode/
@@ -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
@@ -8,10 +8,198 @@ The changes in each Verilator version are described below.  The
 contributors that suggested a given feature are shown in []. Thanks!


+Verilator 4.226 2022-08-31
+==========================
+
+**Minor:**
+
+* Add --future0 and --future1 options.
+* Support class parameters (#2231) (#3541). [Arkadiusz Kozdra/Antmicro]
+* Support wildcard index associative arrays (#3501). [Arkadiusz Kozdra/Antmicro]
+* Support negated properties (#3572). [Aleksander Kiryk]
+* Support $test$plusargs(expr) (#3489).
+* Rename trace rolloverSize() (#3570).
+* Improve Verilation speed with --threads on large designs. [Geza Lore]
+* Fix struct pattern assignment (#2328) (#3517). [Mostafa Gamal]
+* Fix public combo propagation issues (#2905). [Todd Strader]
+* Fix incorrect tristate logic (#3399) [shareefj, Vighnesh Iyer]
+* Fix incorrect bit op tree optimization (#3470). [algrobman]
+* Fix bisonpre for MSYS2 (#3471).
+* Fix max memory usage (#3483). [Kamil Rakoczy]
+* Fix empty string arguments to display (#3484). [Grulfen]
+* Fix table misoptimizing away display (#3488). [Stefan Post]
+* Fix unique_ptr memory header for MinGW64 (#3493).
+* Fix $dump systemtask with --output-split-cfuncs (#3495) (#3497). [Varun Koyyalagunta]
+* Fix wrong bit op tree optimization (#3509). [Nathan Graybeal]
+* Fix nested default assignment for struct pattern (#3511) (#3524). [Mostafa Gamal]
+* Fix sformat string incorrectly cleared (#3515) (#3519). [Gustav Svensk]
+* Fix segfault exporting non-existant package (#3535).
+* Fix void-cast queue pop_front or pop_back (#3542) (#3364). [Drew Ranck]
+* Fix case statement comparing string literal (#3544). [Gustav Svensk]
+* Fix === with some tristate constants (#3551). [Ryszard Rozak]
+* Fix converting subclasses to string (#3552). [Arkadiusz Kozdra/Antmicro]
+* Fix --hierarchical with order-based pin connections (#3583) (#3585). [Kelin9298]
+
+
+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/Antmicro]
+* 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
+==========================
+
+**Major:**
+
+* Add --lib-create, similar to --protect-lib but without protections.
+* Support tracing through --hierarchical/--lib-create libraries (#3200).
+
+**Minor:**
+
+* Internal code cleanups and improvements. [Geza Lore]
+* 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).
+* Fix array method names with parenthesis (#3181) (#3183). [Teng Huang]
+* Fix split_var assign merging (#3177) (#3179). [Yutetsu TAKATSUKASA]
+* Fix wrong bit op tree optimization (#3185). [Yutetsu TAKATSUKASA]
+* Fix some SliceSels not being constants (#3186) (#3218). [Michaël Lefebvre]
+* Fix nested generate if genblk naming (#3189). [yanx21]
+* Fix hang on recursive definition error (#3199). [Jonathan Kimmitt]
+* Fix display of signed without format (#3204). [Julie Schwartz]
+* Fix display of empty string constant (#3207) (#3215). [Julie Schwartz]
+* Fix incorrect width after and-or optimization (#3208). [Julie Schwartz]
+* Fix $fopen etc on integer arrays (#3214). [adrienlemasle]
+* Fix $size on dynamic strings (#3216).
+* Fix %0 format on $value$plusargs (#3217).
+* Fix timescale portability on Arm64 (#3222).
+
+
 Verilator 4.214 2021-10-17
 ==========================

-* Add profile-guided optmization of mtasks (#3150).
+**Major:**
+
+* Add profile-guided optimization of mtasks (#3150).
+
+**Minor:**
+
 * Verilator_gantt has removed the ASCII graphics, use the VCD output instead.
 * Verilator_gantt now shows the predicted mtask times, eval times, and additional statistics.
 * Verilator_gantt data files now include processor information, to allow later processing.
@@ -38,7 +226,7 @@ Verilator 4.212 2021-09-01
 * Fix re-evaluation of logic dependent on state set in DPI exports (#3091). [Geza Lore]
 * Support unpacked array localparams in tasks/functions (#3078). [Geza Lore]
 * Support timeunit/timeprecision in $unit.
-* Support assignment patterns as children of pins (#3041). [Krzysztof Bieganski]
+* Support assignment patterns as children of pins (#3041). [Krzysztof Bieganski/Antmicro]
 * Add --instr-count-dpi to tune assumed DPI import cost for multithreaded
  model scheduling. Default value changed to 200 (#3068). [Yinan Xu]
 * Output files are split based on the set of headers required
@@ -74,7 +262,7 @@ 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]
@@ -103,7 +291,7 @@ Verilator 4.204 2021-06-12
 * Prep work towards better ccache hashing/performance. [Geza Lore]
 * Fix assertion failure in bitOpTree optimization (#2891) (#2899). [Raynard Qiao]
 * Fix DPI functions not seen as vpiModule (#2893). [Todd Strader]
-* Fix bounds check in VL_SEL_IWII (#2910). [Krzysztof Bieganski]
+* Fix bounds check in VL_SEL_IWII (#2910). [Krzysztof Bieganski/Antmicro]
 * Fix slowdown in elaboration (#2911). [Nathan Graybeal]
 * Fix initialization of assoc in assoc array (#2914). [myftptoyman]
 * Fix make support for gmake 3.x (#2920) (#2921). [Philipp Wagner]
@@ -113,10 +301,10 @@ 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]


@@ -154,7 +342,7 @@ 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]
@@ -218,7 +406,7 @@ Verilator 4.108 2021-01-10
 **Major:**

 * Many VPI changes for IEEE compatibility, which may alter behavior from previous releases.
-* Support randomize() class method and rand (#2607). [Krzysztof Bieganski]
+* Support randomize() class method and rand (#2607). [Krzysztof Bieganski/Antmicro]

 **Minor:**

@@ -278,7 +466,7 @@ Verilator 4.104 2020-11-14
 * Support queue and associative array 'with' statements (#2616).
 * Support queue slicing (#2326).
 * Support associative array pattern assignments and defaults.
-* Support static methods and typedefs in classes (#2615). [Krzysztof Bieganski]
+* Support static methods and typedefs in classes (#2615). [Krzysztof Bieganski/Antmicro]
 * Add error on typedef referencing self (#2539). [Cody Piersall]
 * With --debug, turn off address space layout randomization.
 * Fix iteration over mutating list bug in VPI (#2588). [Kaleb Barrett]
@@ -301,7 +489,7 @@ 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]
@@ -327,7 +515,7 @@ 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.
@@ -3727,7 +3915,7 @@ 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.
@@ -322,7 +322,10 @@ cppcheck: $(CPPCHECK_DEP)
 	$(CPPCHECK) $(CPPCHECK_FLAGS) -DVL_DEBUG=1 -DVL_CPPCHECK=1 -DVL_THREADED=1 $(CPPCHECK_INC) $<

 CLANGTIDY = clang-tidy
-CLANGTIDY_FLAGS = -config='' -checks='-fuchsia-*,-cppcoreguidelines-avoid-c-arrays,-cppcoreguidelines-init-variables'
+CLANGTIDY_FLAGS = -config='' \
+	-header-filter='.*' \
+	-checks='-fuchsia-*,-cppcoreguidelines-avoid-c-arrays,-cppcoreguidelines-init-variables,-cppcoreguidelines-avoid-goto,-modernize-avoid-c-arrays,-readability-magic-numbers,-readability-simplify-boolean-expr,-cppcoreguidelines-macro-usage' \
+
 CLANGTIDY_DEP = $(subst .h,.h.tidy,$(CPPCHECK_H)) \
 	$(subst .cpp,.cpp.tidy,$(CPPCHECK_CPP))
 CLANGTIDY_DEFS = -DVL_DEBUG=1 -DVL_THREADED=1 -DVL_CPPCHECK=1
@@ -331,7 +334,7 @@ clang-tidy: $(CLANGTIDY_DEP)
 %.cpp.tidy: %.cpp
 	$(CLANGTIDY) $(CLANGTIDY_FLAGS) $< -- $(CLANGTIDY_DEFS) $(CPPCHECK_INC) | 2>&1 tee $@
 %.h.tidy: %.h
-	$(CLANGTIDY) $(CLANGTIDY_FLAGS) $< -- $(CLANGTIDY_DEFS) $(CPPCHECK_INC) | 2>&1 tee $@
+	$(CLANGTIDY) $(CLANGTIDY_FLAGS) $< -- $(CLANGTIDY_DEFS) $(CPPCHECK_INC) -x c++-header | 2>&1 tee $@

 analyzer-src:
 	-rm -rf src/obj_dbg
@@ -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.
@@ -284,33 +284,34 @@ detailed descriptions of these arguments.
    --bbox-unsup                Blackbox unsupported language features
    --bin <filename>            Override Verilator binary
    --build                     Build model executable/library after Verilation
-     -CFLAGS <flags>            C++ compiler arguments for makefile
    --cc                        Create C++ output
    --cdc                       Clock domain crossing analysis
+     -CFLAGS <flags>            C++ compiler arguments for makefile
    --clk <signal-name>         Mark specified signal as clock
-    --make <build-tool>         Generate scripts for specified build tool
+    --no-clk <signal-name>      Prevent marking specified signal as clock
    --compiler <compiler-name>  Tune for specified C++ compiler
    --converge-limit <loops>    Tune convergence settle time
    --coverage                  Enable all coverage
    --coverage-line             Enable line coverage
    --coverage-max-width <width>   Maximum array depth for coverage
    --coverage-toggle           Enable toggle coverage
-    --coverage-user             Enable SVL user coverage
    --coverage-underscore       Enable coverage of _signals
+    --coverage-user             Enable SVL user coverage
     -D<var>[=<value>]          Set preprocessor define
    --debug                     Enable debugging
    --debug-check               Enable debugging assertions
    --no-debug-leak             Disable leaking memory in --debug mode
    --debugi <level>            Enable debugging at a specified level
    --debugi-<srcfile> <level>  Enable debugging a source file at a level
+    --no-decoration             Disable comments and symbol decorations
    --default-language <lang>   Default language to parse
     +define+<var>=<value>      Set preprocessor define
    --dpi-hdr-only              Only produce the DPI header file
    --dump-defines              Show preprocessor defines with -E
    --dump-tree                 Enable dumping .tree files
+    --dump-tree-addrids         Use short identifiers instead of addresses
    --dump-treei <level>        Enable dumping .tree files at a level
    --dump-treei-<srcfile> <level>  Enable dumping .tree file at a source file at a level
-    --dump-tree-addrids         Use short identifiers instead of addresses
     -E                         Preprocess, but do not compile
    --error-limit <value>       Abort after this number of errors
    --exe                       Link to create executable
@@ -319,7 +320,9 @@ 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
+    --gate-stmts <value>        Tune gate optimizer depth
    --gdb                       Run Verilator under GDB interactively
    --gdbbt                     Run Verilator under GDB for backtrace
    --generate-key              Create random key for --protect-key
@@ -327,53 +330,50 @@ detailed descriptions of these arguments.
    --help                      Display this help
    --hierarchical              Enable hierarchical Verilation
     -I<dir>                    Directory to search for includes
-     -j <jobs>                  Parallelism for --build
-    --gate-stmts <value>        Tune gate optimizer depth
    --if-depth <value>          Tune IFDEPTH warning
     +incdir+<dir>              Directory to search for includes
    --inline-mult <value>       Tune module inlining
    --instr-count-dpi <value>   Assumed dynamic instruction count of DPI imports
-     -LDFLAGS <flags>           Linker pre-object arguments for makefile
+     -j <jobs>                  Parallelism for --build
    --l2-name <value>           Verilog scope name of the top module
    --language <lang>           Default language standard to parse
+     -LDFLAGS <flags>           Linker pre-object arguments for makefile
+    --lib-create <name>         Create a DPI library
     +libext+<ext>+[ext]...     Extensions for finding modules
    --lint-only                 Lint, but do not make output
+    --make <build-tool>         Generate scripts for specified build tool
     -MAKEFLAGS <flags>         Arguments to pass to make during --build
    --max-num-width <value>     Maximum number width (default: 64K)
-    --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
+    --MMD                       Create .d dependency files
    --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-skip-identical         Disable skipping identical output
+    --MP                        Create phony dependency targets
     +notimingchecks            Ignored
     -O0                        Disable optimizations
     -O3                        High performance optimizations
     -O<optimization-letter>    Selectable optimizations
     -o <executable>            Name of final executable
    --no-order-clock-delay      Disable ordering clock enable assignments
-    --no-verilate               Skip verilation and just compile previously Verilated code.
    --output-split <statements>          Split .cpp files into pieces
    --output-split-cfuncs <statements>   Split model functions
    --output-split-ctrace <statements>   Split tracing functions
     -P                         Disable line numbers and blanks with -E
    --pins-bv <bits>            Specify types for top level ports
-    --pins-sc-uint              Specify types for top level ports
    --pins-sc-biguint           Specify types for top level ports
+    --pins-sc-uint              Specify types for top level ports
    --pins-uint8                Specify types for top level ports
+    --no-pins64                 Don't use uint64_t's for 33-64 bit sigs
    --pipe-filter <command>     Filter all input through a script
    --pp-comments               Show preprocessor comments with -E
    --prefix <topname>          Name of top level class
+    --private                   Debugging; see docs
    --prof-c                    Compile C++ code with profiling
    --prof-cfuncs               Name functions for profiling
-    --prof-threads              Enable generating gantt chart data for threads
-    --protect-key <key>         Key for symbol protection
+    --prof-exec                 Enable generating execution profile for gantt chart
+    --prof-pgo                  Enable generating profiling data for PGO
    --protect-ids               Hash identifier names for obscurity
+    --protect-key <key>         Key for symbol protection
    --protect-lib <name>        Create a DPI protected library
-    --private                   Debugging; see docs
    --public                    Debugging; see docs
    --public-flat-rw            Mark all variables, etc as public_flat_rw
     -pvalue+<name>=<value>     Overwrite toplevel parameter
@@ -384,6 +384,7 @@ detailed descriptions of these arguments.
    --rr                        Run Verilator and record with rr
    --savable                   Enable model save-restore
    --sc                        Create SystemC output
+    --no-skip-identical         Disable skipping identical output
    --stats                     Create statistics file
    --stats-vars                Provide statistics on variables
     -sv                        Enable SystemVerilog parsing
@@ -403,7 +404,7 @@ 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
@@ -411,6 +412,7 @@ detailed descriptions of these arguments.
    --unused-regexp <regexp>    Tune UNUSED lint signals
     -V                         Verbose version and config
     -v <filename>              Verilog library
+    --no-verilate               Skip verilation and just compile previously Verilated code.
     +verilog1995ext+<ext>      Synonym for +1364-1995ext+<ext>
     +verilog2001ext+<ext>      Synonym for +1364-2001ext+<ext>
    --version                   Displays program version and exits
@@ -425,6 +427,9 @@ detailed descriptions of these arguments.
     -Wno-lint                  Disable all lint warnings
     -Wno-style                 Disable all style warnings
     -Wpedantic                 Warn on compliance-test issues
+     -Wwarn-<message>           Enable specified warning message
+     -Wwarn-lint                Enable lint warning message
+     -Wwarn-style               Enable style warning message
    --x-assign <mode>           Assign non-initial Xs to this value
    --x-initial <mode>          Assign initial Xs to this value
    --x-initial-edge            Enable initial X->0 and X->1 edge triggers
@@ -444,10 +449,10 @@ 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
@@ -458,7 +463,7 @@ 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],[4.214 2021-10-17],
-        [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_INIT([Verilator],[#.### YYYY-MM-DD])
+#AC_INIT([Verilator],[#.### devel])
+AC_INIT([Verilator],[4.226 2022-08-31],
+        [https://verilator.org],
+        [verilator],[https://verilator.org])

-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
+*Arch*Linux* | *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 +3,17 @@ 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
+Aleksander Kiryk
 Alex Chadwick
 Àlex Torregrosa
+Aliaksei Chapyzhenka
 Ameya Vikram Singh
 Andreas Kuster
+Arkadiusz Kozdra
 Chris Randall
+Chuxuan Wang
 Conor McCullough
 Dan Petrisko
 Daniel Bates
@@ -16,6 +21,7 @@ David Horton
 David Metz
 David Stanford
 David Turner
+Drew Ranck
 Drew Taussig
 Driss Hafdi
 Edgar E. Iglesias
@@ -27,9 +33,13 @@ Geza Lore
 Gianfranco Costamagna
 Glen Gibb
 Graham Rushton
+Guokai Chen
+Gustav Svensk
 Harald Heckmann
 Howard Su
 Huang Rui
+Huanghuang Zhou
+HungMingWu
 HyungKi Jeong
 Iru Cai
 Ivan Vnučec
@@ -38,6 +48,7 @@ James Hanlon
 James Hutchinson
 James Pallister
 Jamey Hicks
+Jamie Iles
 Jan Van Winkel
 Jean Berniolles
 Jeremy Bennett
@@ -45,27 +56,34 @@ John Coiner
 John Demme
 Jonathan Drolet
 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
 Maciej Sobkowski
 Marco Widmer
+Mariusz Glebocki
 Markus Krause
 Marlon James
 Marshal Qiao
 Martin Schmidt
+Martin Stadler
 Matthew Ballance
 Michael Killough
+Michaël Lefebvre
 Mike Popoloski
 Miodrag Milanović
 Morten Borup Petersen
+Mostafa Gamal
 Nandu Raj
 Nathan Kohagen
 Nathan Myers
@@ -79,8 +97,10 @@ Pieter Kapsenberg
 Piotr Binkowski
 Qingyao Sun
 Rafal Kapuscik
+Raynard Qiao
 Richard Myers
 Rupert Swarbrick
+Ryszard Rozak
 Samuel Riedel
 Sean Cross
 Sebastien Van Cauwenberghe
@@ -88,6 +108,7 @@ Sergi Granell
 Stefan Wallentowitz
 Stephen Henry
 Steven Hugg
+Teng Huang
 Tim Snyder
 Tobias Rosenkranz
 Tobias Wölfel
@@ -96,10 +117,14 @@ Tomasz Gorochowik
 Tymoteusz Blazejczyk
 Udi Finkelstein
 Unai Martinez-Corral
+Varun Koyyalagunta
 Vassilis Papaefstathiou
 Veripool API Bot
 Victor Besyakov
+William D. Jones
 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:
@@ -86,58 +86,17 @@ Connecting to C++
 In C++ output mode (:vlopt:`--cc`), the Verilator generated model class is a
 simple C++ class.  The user must write a C++ wrapper and main loop for the
 simulation, which instantiates the model class, and link with the Verilated
-model.  Here is a simple example:
+model.

-.. code-block:: C++
+Refer to ``examples/make_tracing_c`` in the distribution for a detailed
+commented example.

-         #include <verilated.h>          // Defines common routines
-         #include <iostream>             // Need std::cout
-         #include "Vtop.h"               // From Verilating "top.v"
+Top level IO signals are read and written as members of the model.  You
+call the model's :code:`eval()` method to evaluate the model.  When the
+simulation is complete call the model's :code:`final()` method to execute
+any SystemVerilog final blocks, and complete any assertions. See
+:ref:`Evaluation Loop`.

-         Vtop *top;                      // Instantiation of model
-
-         vluint64_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)
-
-         double sc_time_stamp() {        // Called by $time in Verilog
-             return main_time;           // converts to double, to match
-                                         // what SystemC does
-         }
-
-         int main(int argc, char** argv) {
-             Verilated::commandArgs(argc, argv);   // Remember args
-
-             top = new Vtop;             // Create model
-
-             top->reset_l = 0;           // Set some inputs
-
-             while (!Verilated::gotFinish()) {
-                 if (main_time > 10) {
-                     top->reset_l = 1;   // Deassert reset
-                 }
-                 if ((main_time % 10) == 1) {
-                     top->clk = 1;       // Toggle clock
-                 }
-                 if ((main_time % 10) == 6) {
-                     top->clk = 0;
-                 }
-                 top->eval();            // Evaluate model
-                 cout << top->out << endl;       // Read a output
-                 main_time++;            // Time passes...
-             }
-
-             top->final();               // Done simulating
-             //    // (Though this example doesn't get here)
-             delete top;
-         }
-
-
-Note top level IO signals are read and written as members of the model.  You
-call the :code:`eval()` method to evaluate the model.  When the simulation is
-complete call the :code:`final()` method to execute any SystemVerilog final
-blocks, and complete any assertions. See :ref:`Evaluation Loop`.


 Connecting to SystemC
@@ -150,7 +109,7 @@ 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 +159,7 @@ 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 +216,9 @@ 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 +383,12 @@ 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() {
@@ -444,14 +405,15 @@ accesses the above signal "readme" would be:

       int main(int argc, char** argv, char** env) {
           Verilated::commandArgs(argc, argv);
-           Vour* top = new Vour;
-           Verilated::internalsDump();  // See scopes to help debug
-           while (!Verilated::gotFinish()) {
+           const std::unique_ptr<VerilatedContext> contextp{new VerilatedContext};
+           const std::unique_ptr<Vour> top{new Vour{contextp.get()}};
+
+           contextp->internalsDump();  // See scopes to help debug
+           while (!contextp->gotFinish()) {
               top->eval();
               VerilatedVpi::callValueCbs();  // For signal callbacks
               read_and_check();
           }
-           delete top;
           return 0;
       }
     EOF
@@ -506,7 +468,7 @@ 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
@@ -9,17 +9,29 @@ The following deprecated items are scheduled for future removal:
 C++11 compiler support
  Verilator currently requires C++11 or newer compilers.  Verilator will
  require C++14 or newer compilers for both compiling Verilator and
-  compiling Verilated models no sooner than January 2022.
+  compiling Verilated models no sooner than January 2023.

 Verilated_heavy.h
  The legacy "verilated_heavy.h" include was replaced with just including
  "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:
@@ -8,7 +8,7 @@ Simulation Runtime Arguments

 The following are the arguments that may be passed to a Verilated
 executable, provided that executable calls
-:code:`Verilated::commandArgs()`.
+:code:`VerilatedContext*->commandArgs(argc, argv)`.

 All simulation runtime arguments begin with "+verilator", so that the
 user's executable may skip over all "+verilator" arguments when parsing its
@@ -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`.

@@ -84,7 +96,7 @@ Summary:
 .. option:: +verilator+noassert

   Disable assert checking per runtime argument. This is the same as
-   calling :code:`Verilated::assertOn(false)` in the model.
+   calling :code:`VerilatedContext*->assertOn(false)` in the model.

 .. option:: +verilator+V

@@ -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.

@@ -129,16 +129,6 @@ Summary:
   is also used). Verilator manages the build itself, and for this --build
   requires GNU Make to be available on the platform.

-.. option:: -CFLAGS <flags>
-
-   Add specified C compiler argument to the generated makefiles. For
-   multiple flags either pass them as a single argument with space
-   separators quoted in the shell (:command:`-CFLAGS "-a -b"`), or use
-   multiple -CFLAGS options (:command:`-CFLAGS -a -CFLAGS -b`).
-
-   When make is run on the generated makefile these will be passed to the
-   C++ compiler (g++/clang++/msvc++).
-
 .. option:: --cc

   Specifies C++ without SystemC output mode; see also :vlopt:`--sc`
@@ -156,43 +146,43 @@ Summary:
   have interest in adding more traditional CDC checks, please contact the
   authors.

+.. option:: -CFLAGS <flags>
+
+   Add specified C compiler argument to the generated makefiles. For
+   multiple flags either pass them as a single argument with space
+   separators quoted in the shell (:command:`-CFLAGS "-a -b"`), or use
+   multiple -CFLAGS options (:command:`-CFLAGS -a -CFLAGS -b`).
+
+   When make is run on the generated makefile these will be passed to the
+   C++ compiler (g++/clang++/msvc++).
+
 .. option:: --clk <signal-name>

-   Sometimes it is quite difficult for Verilator to distinguish clock
-   signals from other data signals. Occasionally the clock signals can end
-   up in the checking list of signals which determines if further
-   evaluation is needed. This will heavily degrade the performance of a
-   Verilated model.
-
   With :vlopt:`--clk`, the specified signal-name is taken as a root clock
-   into the model, then Verilator will mark the signal as clocker and
-   propagate the clocker attribute automatically to other signals derived
-   from that. In this way, Verilator will try to avoid taking the clocker
-   signal into checking list.
+   into the model; Verilator will mark the signal as clocker and
+   propagate the clocker attribute automatically to other signals downstream in
+   that clock tree.

-   Note signal-name is specified by the RTL hierarchy path. For example,
-   v.foo.bar.  If the signal is the input to top-module, the directly the
-   signal name. If you find it difficult to find the exact name, try to use
-   a :option:`/*verilator&32;clocker*/` metacomment in RTL file to mark the
+   The provided signal-name is specified using a RTL hierarchy path. For
+   example, v.foo.bar.  If the signal is the input to top-module, then
+   directly provide the signal name. Alternatively, use a
+   :option:`/*verilator&32;clocker*/` metacomment in RTL file to mark the
   signal directly.

-   If clock signals are assigned to vectors and then later used
-   individually, Verilator will attempt to decompose the vector and connect
-   the single-bit clock signals directly.  This should be transparent to
-   the user.
+   If clock signals are assigned to vectors and then later used as
+   individual bits, Verilator will attempt to decompose the vector and
+   connect the single-bit clock signals.

-.. option:: --make <build-tool>
+   The clocker attribute is useful in cases where Verilator does not
+   properly distinguish clock signals from other data signals. Using
+   clocker will cause the signal indicated to be considered a clock, and
+   remove it from the combinatorial logic reevaluation checking code. This
+   may greatly improve performance.

-   Generates a script for the specified build tool.
+.. option:: --no-clk <signal-name>

-   Supported values are ``gmake`` for GNU Make and ``cmake`` for CMake.
-   Both can be specified together.  If no build tool is specified, gmake is
-   assumed.  The executable of gmake can be configured via environment
-   variable "MAKE".
-
-   When using :vlopt:`--build` Verilator takes over the responsibility of
-   building the model library/executable.  For this reason :option:`--make`
-   cannot be specified when using :vlopt:`--build`.
+   Prevent the specified signal from being marked as clock. See
+   :vlopt:`--clk`.

 .. option:: --compiler <compiler-name>

@@ -303,6 +293,13 @@ Summary:
   <--debugi>`). Higher levels produce more detailed messages.  See
   :vlopt:`--debug` for other implications of enabling debug.

+.. option:: --no-decoration
+
+   When creating output Verilated code, minimize comments, white space,
+   symbol names and other decorative items, at the cost of greatly reduced
+   readability. This may assist C++ compile times. This will not typically
+   change the ultimate model's performance, but may in some cases.
+
 .. option:: --default-language <value>

   Select the language to be used by default when first processing each
@@ -446,6 +443,78 @@ 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:: -future0 <option>
+
+   Rarely needed.  Suppress an unknown Verilator option for an option that
+   takes no additional arguments.  This is used to allow scripts written
+   with pragmas for a later version of Verilator to run under a older
+   version.  e.g. :code:`-future0 option --option` would on older versions
+   that do not understand :code:`--option` or :code:`+option` suppress what
+   would otherwise be an invalid option error, and on newer versions that
+   implement :code:`--option`, :code:`-future0 option --option` would have
+   the :code:`-future0 option` ignored and the :code:`--option` would
+   function appropriately.
+
+.. option:: -future1 <option>
+
+   Rarely needed.  Suppress an unknown Verilator option for an option that
+   takes an additional argument.  This is used to allow scripts written
+   with pragmas for a later version of Verilator to run under a older
+   version.  e.g. :code:`-future1 option --option arg` would on older
+   versions that do not understand :code:`--option arg` or :code:`+option
+   arg` suppress what would otherwise be an invalid option error, and on
+   newer versions that implement :code:`--option arg`, :code:`-future1
+   option --option arg` would have the :code:`-future1 option` ignored and
+   the :code:`--option arg` would function appropriately.
+
 .. option:: -G<name>=<value>

   Overwrites the given parameter of the toplevel module. The value is
@@ -552,21 +621,6 @@ Summary:
   to limit the number of parallel build jobs but attempt to execute all
   independent build steps in parallel.

-.. option:: -LDFLAGS <flags>
-
-   Add specified C linker arguments to the generated makefiles.  For multiple
-   flags either pass them as a single argument with space separators quoted
-   in the shell (``-LDFLAGS "-a -b"``), or use multiple -LDFLAGS arguments
-   (``-LDFLAGS -a -LDFLAGS -b``).
-
-   When make is run on the generated makefile these will be passed to the
-   C++ linker (ld) **after** the primary file being linked.  This flag is
-   called :vlopt:`-LDFLAGS` as that's the traditional name in simulators;
-   it's would have been better called LDLIBS as that's the Makefile
-   variable it controls.  (In Make, LDFLAGS is before the first object,
-   LDLIBS after.  -L libraries need to be in the Make variable LDLIBS, not
-   LDFLAGS.)
-
 .. option:: --l2-name <value>

   Instead of using the module name when showing Verilog scope, use the
@@ -583,6 +637,38 @@ Summary:
   A synonym for :vlopt:`--default-language`, for compatibility with other
   tools and earlier versions of Verilator.

+.. option:: -LDFLAGS <flags>
+
+   Add specified C linker arguments to the generated makefiles.  For multiple
+   flags either pass them as a single argument with space separators quoted
+   in the shell (``-LDFLAGS "-a -b"``), or use multiple -LDFLAGS arguments
+   (``-LDFLAGS -a -LDFLAGS -b``).
+
+   When make is run on the generated makefile these will be passed to the
+   C++ linker (ld) **after** the primary file being linked.  This flag is
+   called :vlopt:`-LDFLAGS` as that's the traditional name in simulators;
+   it's would have been better called LDLIBS as that's the Makefile
+   variable it controls.  (In Make, LDFLAGS is before the first object,
+   LDLIBS after.  -L libraries need to be in the Make variable LDLIBS, not
+   LDFLAGS.)
+
+.. option:: --lib-create <name>
+
+   Produces C++, Verilog wrappers and a Makefile which can in turn produce
+   a DPI library which can be used by Verilator or other simulators along
+   with the corresponding Verilog wrapper.  The Makefile will build both a
+   static and dynamic version of the library named :file:`lib<name>.a` and
+   :file:`lib<name>.so` respectively.  This is done because some simulators
+   require a dynamic library, but the static library is arguably easier to
+   use if possible.  :vlopt:`--protect-lib` implies :vlopt:`--protect-ids`.
+
+   When using :vlopt:`--lib-create` it is advised to also use
+   :vlopt:`--timescale-override /1fs <--timescale-override>` to ensure the
+   model has a time resolution that is always compatible with the time
+   precision of the upper instantiating module.
+
+   See also :vlopt:`--protect-lib`.
+
 .. option:: +libext+<ext>[+<ext>][...]

   Specify the extensions that should be used for finding modules.  If for
@@ -601,6 +687,19 @@ Summary:
   If the design is not to be completely Verilated see also the
   :vlopt:`--bbox-sys` and :vlopt:`--bbox-unsup` options.

+.. option:: --make <build-tool>
+
+   Generates a script for the specified build tool.
+
+   Supported values are ``gmake`` for GNU Make and ``cmake`` for CMake.
+   Both can be specified together.  If no build tool is specified, gmake is
+   assumed.  The executable of gmake can be configured via environment
+   variable "MAKE".
+
+   When using :vlopt:`--build` Verilator takes over the responsibility of
+   building the model library/executable.  For this reason :option:`--make`
+   cannot be specified when using :vlopt:`--build`.
+
 .. option:: -MAKEFLAGS <string>

   When using :vlopt:`--build`, add the specified argument to the invoked
@@ -615,17 +714,6 @@ Summary:
   Set the maximum number literal width (e.g. in 1024'd22 this it the
   1024).  Defaults to 64K.

-.. option:: --MMD =item --no-MMD
-
-   Enable/disable creation of .d dependency files, used for make dependency
-   detection, similar to gcc -MMD option.  By default this option is
-   enabled for :vlopt:`--cc` or :vlopt:`--sc` modes.
-
-.. option:: --MP
-
-   When creating .d dependency files with :vlopt:`--MMD` option, make phony
-   targets.  Similar to :command:`gcc -MP` option.
-
 .. option:: --Mdir <directory>

   Specifies the name of the Make object directory.  All generated files
@@ -633,40 +721,23 @@ 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
+.. option:: --MMD

-   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:: --no-MMD
+
+   Enable/disable creation of .d dependency files, used for make dependency
+   detection, similar to gcc -MMD option.  By default this option is
+   enabled for :vlopt:`--cc` or :vlopt:`--sc` modes.

 .. option:: --mod-prefix <topname>

   Specifies the name to prepend to all lower level classes.  Defaults to
   the same as :vlopt:`--prefix`.

-.. option:: --no-clk <signal-name>
+.. option:: --MP

-   Prevent the specified signal from being marked as clock. See
-   :vlopt:`--clk`.
-
-.. option:: --no-decoration
-
-   When creating output Verilated code, minimize comments, white space,
-   symbol names and other decorative items, at the cost of greatly reduced
-   readability. This may assist C++ compile times. This will not typically
-   change the ultimate model's performance, but may in some cases.
-
-.. option:: --no-pins64
-
-   Backward compatible alias for :vlopt:`--pins-bv 33 <--pins-bv>`.
-
-.. option:: --no-skip-identical =item --skip-identical
-
-   Rarely needed.  Disables or enables skipping execution of Verilator if
-   all source files are identical, and all output files exist with newer
-   dates.  By default this option is enabled for :vlopt:`--cc` or
-   :vlopt:`--sc` modes only.
+   When creating .d dependency files with :vlopt:`--MMD` option, make phony
+   targets.  Similar to :command:`gcc -MP` option.

 .. option:: +notimingchecks

@@ -688,9 +759,9 @@ 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>

@@ -745,15 +816,10 @@ Summary:
   With :vlopt:`-E`, disable generation of :code:`&96;line` markers and
   blank lines, similar to :command:`gcc -P`.

-.. option:: --pins64
-
-   Backward compatible alias for :vlopt:`--pins-bv 65 <--pins-bv>`.  Note
-   that's a 65, not a 64.
-
 .. 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
@@ -782,6 +848,15 @@ Summary:
   of uint32_t.  Likewise pins of width 9-16 will use uint16_t instead of
   uint32_t.

+.. option:: --pins64
+
+   Backward compatible alias for :vlopt:`--pins-bv 65 <--pins-bv>`.  Note
+   that's a 65, not a 64.
+
+.. option:: --no-pins64
+
+   Backward compatible alias for :vlopt:`--pins-bv 33 <--pins-bv>`.
+
 .. option:: --pipe-filter <command>

   Rarely needed.  Verilator will spawn the specified command as a
@@ -811,6 +886,11 @@ Summary:
   prepended to the name of the :vlopt:`--top` option, or V prepended to
   the first Verilog filename passed on the command line.

+.. option:: --private
+
+   Opposite of :vlopt:`--public`.  Is the default; this option exists for
+   backwards compatibility.
+
 .. option:: --prof-c

   When compiling the C++ code, enable the compiler's profiling flag
@@ -830,27 +910,19 @@ 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`.
-
-.. option:: --protect-key <key>
-
-   Specifies the private key for :vlopt:`--protect-ids`. For best security
-   this key should be 16 or more random bytes, a reasonable secure choice
-   is the output of :command:`verilator --generate-key` . Typically, a key
-   would be created by the user once for a given protected design library,
-   then every Verilator run for subsequent versions of that library would
-   be passed the same :vlopt:`--protect-key`. Thus, if the input Verilog is
-   similar between library versions (Verilator runs), the Verilated code
-   will likewise be mostly similar.
-
-   If :vlopt:`--protect-key` is not specified and a key is needed,
-   Verilator will generate a new key for every Verilator run. As the key is
-   not saved, this is best for security, but means every Verilator run will
-   give vastly different output even for identical input, perhaps harming
-   compile times (and certainly thrashing any "ccache").
+   Deprecated. Same as --prof-exec and --prof-pgo together.

 .. option:: --protect-ids

@@ -869,34 +941,37 @@ 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-key <key>
+
+   Specifies the private key for :vlopt:`--protect-ids`. For best security
+   this key should be 16 or more random bytes, a reasonable secure choice
+   is the output of :command:`verilator --generate-key` . Typically, a key
+   would be created by the user once for a given protected design library,
+   then every Verilator run for subsequent versions of that library would
+   be passed the same :vlopt:`--protect-key`. Thus, if the input Verilog is
+   similar between library versions (Verilator runs), the Verilated code
+   will likewise be mostly similar.
+
+   If :vlopt:`--protect-key` is not specified and a key is needed,
+   Verilator will generate a new key for every Verilator run. As the key is
+   not saved, this is best for security, but means every Verilator run will
+   give vastly different output even for identical input, perhaps harming
+   compile times (and certainly thrashing any "ccache").
+
 .. option:: --protect-lib <name>

-   Produces C++, Verilog wrappers and a Makefile which can in turn produce
-   a DPI library which can be used by Verilator or other simulators along
-   with the corresponding Verilog wrapper.  The Makefile will build both a
-   static and dynamic version of the library named :file:`lib<name>.a` and
-   :file:`lib<name>.so` respectively.  This is done because some simulators
-   require a dynamic library, but the static library is arguably easier to
-   use if possible.  :vlopt:`--protect-lib` implies :vlopt:`--protect-ids`.
+   Produces a DPI library similar to :vlopt:`--lib-create`, but hides
+   internal design details.  :vlopt:`--protect-lib` implies
+   :vlopt:`--protect-ids`, and :vlopt:`--lib-create`.

   This allows for the secure delivery of sensitive IP without the need for
   encrypted RTL (i.e. IEEE P1735).  See :file:`examples/make_protect_lib`
   in the distribution for a demonstration of how to build and use the DPI
   library.

-   When using :vlopt:`--protect-lib` it is advised to also use
-   :vlopt:`--timescale-override /1fs <--timescale-override>` to ensure the
-   model has a time resolution that is always compatible with the time
-   precision of the upper instantiating module.
-
-.. option:: --private
-
-   Opposite of :vlopt:`--public`.  Is the default; this option exists for
-   backwards compatibility.
-
 .. option:: --public

   This is only for historical debug use.  Using it may result in
@@ -989,6 +1064,15 @@ Summary:

   Specifies SystemC output mode; see also :vlopt:`--cc` option.

+.. option:: --skip-identical
+
+.. option:: --no-skip-identical
+
+   Rarely needed.  Disables or enables skipping execution of Verilator if
+   all source files are identical, and all output files exist with newer
+   dates.  By default this option is enabled for :vlopt:`--cc` or
+   :vlopt:`--sc` modes only.
+
 .. option:: --stats

   Creates a dump file with statistics on the design in
@@ -1026,7 +1110,8 @@ 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

@@ -1104,7 +1189,8 @@ 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

@@ -1158,18 +1244,18 @@ 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

-   Enable tracing of signals that start with an underscore. Normally, these
-   signals are not output during tracing.  See also
-   :vlopt:`--coverage-underscore` option.
+   Enable tracing of signals or modules that start with an
+   underscore. Normally, these signals are not output during tracing.  See
+   also :vlopt:`--coverage-underscore` option.

 .. option:: -U<var>

@@ -1358,9 +1444,10 @@ 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

@@ -1506,9 +1593,9 @@ The grammar of configuration commands is as follows:

 .. option:: no_clocker -module "<modulename>" [-function "<funcname>"] -var "<signame>"

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

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

@@ -1522,6 +1609,14 @@ 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>
@@ -1623,7 +1718,7 @@ 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>"
@@ -1652,9 +1747,28 @@ 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

 *******************
@@ -48,6 +48,14 @@ or "`ifdef`"'s may break other tools.
   Verilog functions where the variable is flattened out, and also enable
   other optimizations.

+   Verilator does not use any text inside the quotes for
+   ordering/scheduling.  If you need the $c to be called at a specific
+   time, e.g. when a variable changes, then the $c must be under an
+   appropriate sensitivity statement, e.g. :code:`always @(posedge clk)
+   $c("func()")` to call it on every edge, or e.g. :code:`always @*
+   c("func(",a,")")` to call it when :code:`a` changes (the latter working
+   because :code:`a` is outside the quotes).
+
   If you will be reading or writing any Verilog variables inside the C++
   functions, the Verilog signals must be declared with
   :option:`/*verilator&32;public*/` metacomments.
@@ -176,10 +184,9 @@ or "`ifdef`"'s may break other tools.

 .. option:: /*verilator&32;no_clocker*/

-   Used after a signal declaration to indicate the signal is used as clock
-   or not. This information is used by Verilator to mark the signal as
-   clocker and propagate the clocker attribute automatically to derived
-   signals. See :vlopt:`--clk`.
+   Specifies that the signal is used as clock or not. This information is
+   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
   files.
@@ -202,6 +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:

@@ -424,7 +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
@@ -482,6 +503,13 @@ or "`ifdef`"'s may break other tools.
   Verilator) text that should be passed through to the XML output as a tag,
   for use by downstream applications.

+.. option:: /*verilator&32;trace_init_task*/
+
+   Attached to a DPI import to indicate that function should be called when
+   initializing tracing. This attribute is indented only to be used
+   internally in code that Verilator generates when :vlopt:`--lib-create`
+   or :vlopt:`--hierarchical` is used along with :vlopt:`--trace`.
+
 .. option:: /*verilator&32;tracing_off*/

   Disable waveform tracing for all future signals that are declared in
@@ -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

 ******************************
@@ -35,15 +35,17 @@ simulators.
 Does Verilator run under Windows?
 """""""""""""""""""""""""""""""""

-Yes, using Cygwin.  Verilated output also compiles under Microsoft Visual
-C++, but this is not tested every release.
+Yes, ideally run Ubuntu under Windows Subsystem for Linux (WSL2).
+Alternatively use Cygwin, though this tends to be slower and is not
+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
@@ -70,23 +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
@@ -132,7 +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()) {
@@ -351,8 +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.
@@ -442,7 +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.
@@ -479,7 +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

 *****
@@ -50,7 +50,7 @@ For --cc/--sc, it creates:
     - Make include file with class names (from --make gmake)
   * - *{prefix}*\ _hier.mk
     - Make file for hierarchy blocks (from --make gmake)
-   * - *{prefix|*\ _hierMkArgs.f
+   * - *{prefix}*\ _hierMkArgs.f
     - Arguments for hierarchical Verilation (from --make gmake)
   * - *{prefix}*\ _hierCMakeArgs.f
     - Arguments for hierarchical Verilation (from --make cmake)
@@ -62,13 +62,17 @@ For --cc/--sc, it creates:
     - Top level (SystemVerilog $root) internal header file
   * - *{prefix}*\ ___024root.cpp
     - Top level (SystemVerilog $root) internal C++ file
-   * - *{prefix}*___024root*{__n}*\ .cpp
-     - Additional top level internal C++ files (from --output-split)
+   * - *{prefix}*\ ___024root\ *{__n}*\ .cpp
+     - Additional top level internal C++ files
+   * - *{prefix}*\ ___024root\ *{__DepSet_hash__n}*\ .cpp
+     - Additional top level internal C++ files (hashed to reduce build times)
   * - *{prefix}*\ ___024root__Slow\ *{__n}*\ .cpp
     - Infrequent cold routines
-   * - *{prefix}*\ ___024root__Trace{__n}*\ .cpp
+   * - *{prefix}*\ ___024root\ *{__DepSet_hash__n}*\ .cpp
+     - Infrequent cold routines (hashed to reduce build times)
+   * - *{prefix}*\ ___024root__Trace\ *{__n}*\ .cpp
     - Wave file generation code (from --trace)
-   * - *{prefix}*\ ___024root__Trace__Slow{__n}*\ .cpp
+   * - *{prefix}*\ ___024root__Trace__Slow\ *{__n}*\ .cpp
     - Wave file generation code (from --trace)
   * - *{prefix}*\ __Dpi.h
     - DPI import and export declarations (from --dpi)
@@ -87,7 +91,9 @@ For --cc/--sc, it creates:
   * - *{prefix}{each_verilog_module}*\ .cpp
     - Lower level internal C++ files
   * - *{prefix}{each_verilog_module}{__n}*\ .cpp
-     - Additional lower C++ files (from --output-split)
+     - Additional lower C++ files
+   * - *{prefix}{each_verilog_module}{__DepSet_hash__n}*\ .cpp
+     - Additional lower C++ files (hased to reduce build times)

 For --hierarchy mode, it creates:

@@ -120,7 +126,7 @@ 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 +161,13 @@ 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:
@@ -77,17 +77,17 @@ 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
-Linux, HPUX and Solaris. It should run with minor porting on any
-GNU/Linux-ish platform. Verilator also works on Windows under Cygwin, and
-Windows under MinGW (gcc -mno-cygwin). Verilated output (not Verilator
-itself) compiles under all the options above, plus using MSVC++.
+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)
+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
 -----------------

@@ -471,7 +489,7 @@ $test$plusargs, $value$plusargs

  .. code-block:: C++

-        Verilated::commandArgs(argc, argv);
+        {VerilatedContext*} ->commandArgs(argc, argv);

  to register the command line before calling $test$plusargs or
  $value$plusargs.
@@ -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

 **********
@@ -68,12 +68,11 @@ 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 protect-lib. Each hierarchy
-block is Verilated to a protect-lib. User modules of the hierarchy blocks
-will see a tiny wrapper generated by protect-lib instead of the actual
-design.
+The current hierarchical Verilation is based on :vlopt:`--lib-create`. Each
+hierarchy block is Verilated into a library. User modules of the hierarchy
+blocks will see a tiny wrapper generated by :vlopt:`--lib-create`.


 Usage
@@ -222,9 +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.

@@ -232,12 +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
@@ -266,7 +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
@@ -442,7 +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.

@@ -99,7 +99,7 @@ run:

 	@echo "-- DONE --------------------"
 	@echo "Note: Once this example is understood, see examples/cmake_tracing_sc."
-	@echo "Note: Also see the EXAMPLE section in the verilator manpage/document."
+	@echo "Note: See also https://verilator.org/guide/latest/examples.html"

 clean mostlyclean distclean maintainer-clean:
 	@rm -rf build logs
@@ -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)
@@ -40,7 +40,7 @@ default:
 	obj_dir/Vtop
 	@echo "-- DONE --------------------"
 	@echo "Note: Once this example is understood, see examples/make_tracing_c."
-	@echo "Note: Also see the EXAMPLE section in the verilator manpage/document."
+	@echo "Note: See also https://verilator.org/guide/latest/examples.html"

 ######################################################################

@@ -21,11 +21,14 @@ int main(int argc, char** argv, char** env) {
    // Prevent unused variable warnings
    if (false && argc && argv && env) {}

+    // Construct a VerilatedContext to hold simulation time, etc.
+    VerilatedContext* contextp = new VerilatedContext;
+
    // Construct the Verilated model, from Vtop.h generated from Verilating "top.v"
-    Vtop* top = new Vtop;
+    Vtop* top = new Vtop{contextp};

    // Simulate until $finish
-    while (!Verilated::gotFinish()) {
+    while (!contextp->gotFinish()) {

        // Evaluate model
        top->eval();
@@ -4,7 +4,8 @@
 // any use, without warranty, 2017 by Wilson Snyder.
 // SPDX-License-Identifier: CC0-1.0

-// See also the EXAMPLE section in the verilator manpage/document.
+// See also https://verilator.org/guide/latest/examples.html"
+
 module top;
   initial begin
      $display("Hello World!");
@@ -49,7 +49,7 @@ run:
 	obj_dir/Vtop
 	@echo "-- DONE --------------------"
 	@echo "Note: Once this example is understood, see examples/make_tracing_sc."
-	@echo "Note: Also see the EXAMPLE section in the verilator manpage/document."
+	@echo "Note: See also https://verilator.org/guide/latest/examples.html"

 ######################################################################

@@ -4,7 +4,8 @@
 // any use, without warranty, 2017 by Wilson Snyder.
 // SPDX-License-Identifier: CC0-1.0

-// See also the EXAMPLE section in the verilator manpage/document.
+// See also https://verilator.org/guide/latest/examples.html"
+
 module top;
   initial begin
      $display("Hello World!");
@@ -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
@@ -4,7 +4,8 @@
 // any use, without warranty, 2019 by Todd Strader.
 // SPDX-License-Identifier: CC0-1.0

-// See also the EXAMPLE section in the verilator manpage/document.
+// See also https://verilator.org/guide/latest/examples.html"
+
 module top (input clk);

   integer cyc = 0;
@@ -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.
@@ -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
@@ -80,14 +81,19 @@
 #endif
 // clang-format on

+class VerilatedContext;
 class VerilatedContextImp;
 class VerilatedContextImpData;
 class VerilatedCovContext;
 class VerilatedEvalMsgQueue;
 class VerilatedFst;
 class VerilatedFstC;
+class VerilatedFstSc;
 class VerilatedScope;
 class VerilatedScopeNameMap;
+template <class, class>
+class VerilatedTrace;
+class VerilatedTraceConfig;
 class VerilatedVar;
 class VerilatedVarNameMap;
 class VerilatedVcd;
@@ -99,11 +105,11 @@ 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 +119,7 @@ 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 +147,11 @@ 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 +230,7 @@ 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.
@@ -250,6 +256,40 @@ public:
 #endif
 };

+//=========================================================================
+/// Base class of a Verilator generated (Verilated) model.
+///
+/// VerilatedModel is a base class of the user facing primary class generated
+/// by Verilator.
+
+class VerilatedModel VL_NOT_FINAL {
+    VL_UNCOPYABLE(VerilatedModel);
+
+    VerilatedContext& m_context;  // The VerilatedContext this model is instantiated under
+
+protected:
+    explicit VerilatedModel(VerilatedContext& context);
+    virtual ~VerilatedModel() = default;
+
+public:
+    /// Returns the VerilatedContext this model is instantiated under
+    /// Used to get to e.g. simulation time via contextp()->time()
+    inline VerilatedContext* contextp() const { return &m_context; }
+    /// Returns the hierarchical name of this module instance.
+    virtual const char* hierName() const = 0;
+    /// Returns the name of this model (the name of the generated model class).
+    virtual const char* modelName() const = 0;
+    /// Returns the thread level parallelism, this model was Verilated with. Always 1 or higher.
+    virtual unsigned threads() const = 0;
+
+private:
+    // The following are for use by Verilator internals only
+    template <class, class>
+    friend class VerilatedTrace;
+    // Run-time trace configuration requested by this model
+    virtual std::unique_ptr<VerilatedTraceConfig> traceConfig() const;
+};
+
 //=========================================================================
 /// Base class for all Verilated module classes.

@@ -264,10 +304,6 @@ public:
    const char* name() const { return m_namep; }  ///< Return name of module
 };

-/// Declare a module, ala SC_MODULE
-#define VL_MODULE(modname) class modname VL_NOT_FINAL : public VerilatedModule
-// Not class final in VL_MODULE, as users might be abstracting our models (--hierarchical)
-
 //=========================================================================
 // Functions overridable by user defines
 // (Internals however must use VL_PRINTF_MT, which calls these.)
@@ -318,10 +354,10 @@ 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 +378,10 @@ 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;

@@ -359,7 +395,17 @@ protected:
    } m_args VL_GUARDED_BY(m_argMutex);

    // Implementation details
-    std::unique_ptr<VerilatedContextImpData> m_impdatap;
+    const std::unique_ptr<VerilatedContextImpData> m_impdatap;
+    // Number of threads to use for simulation (size of m_threadPool + 1 for main thread)
+#ifdef VL_THREADED
+    unsigned m_threads = std::thread::hardware_concurrency();
+#else
+    const unsigned m_threads = 1;
+#endif
+    // The thread pool shared by all models added to this context
+    std::unique_ptr<VerilatedVirtualBase> m_threadPool;
+    // The execution profiler shared by all models added to this context
+    std::unique_ptr<VerilatedVirtualBase> m_executionProfiler;
    // Coverage access
    std::unique_ptr<VerilatedVirtualBase> m_coveragep;  // Pointer for coveragep()

@@ -372,6 +418,10 @@ 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 +516,16 @@ 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
@@ -489,6 +539,12 @@ public:
    /// Get time precision as IEEE-standard text
    const char* timeprecisionString() const VL_MT_SAFE;

+    /// Get number of threads used for simulation (including the main thread)
+    unsigned threads() const { return m_threads; }
+    /// Set number of threads used for simulation (including the main thread)
+    /// Can only be called before the thread pool is created (before first model is added).
+    void threads(unsigned n);
+
    /// Allow traces to at some point be enabled (disables some optimizations)
    void traceEverOn(bool flag) VL_MT_SAFE {
        if (flag) calcUnusedSigs(true);
@@ -504,25 +560,32 @@ public:
    /// releases - contact the authors before production use.
    void scopesDump() const VL_MT_SAFE;

-public:  // But for internal use only
+    // METHODS - public but for internal use only
+
    // Internal: access to implementation class
    VerilatedContextImp* impp() { return reinterpret_cast<VerilatedContextImp*>(this); }
    const VerilatedContextImp* impp() const {
        return reinterpret_cast<const VerilatedContextImp*>(this);
    }

+    void addModel(VerilatedModel*);
+
+    VerilatedVirtualBase* threadPoolp();
+    VerilatedVirtualBase*
+    enableExecutionProfiler(VerilatedVirtualBase* (*construct)(VerilatedContext&));
+
    // Internal: $dumpfile
    void dumpfile(const std::string& flag) VL_MT_SAFE_EXCLUDES(m_timeDumpMutex);
    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 +596,10 @@ 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 +624,7 @@ 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 +637,21 @@ 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 +701,9 @@ 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 +825,11 @@ 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(); }
@@ -820,11 +887,10 @@ public:
    }
 #endif

-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 +914,8 @@ 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 ($(CFG_WITH_CCWARN),yes)	# Local... Else don't burden users
@@ -141,26 +142,6 @@ ifneq ($(VM_THREADS),0)
 endif
 endif

-ifneq ($(VM_TRACE_THREADS),0)
- ifneq ($(VM_TRACE_THREADS),)
-  ifeq ($(findstring -DVL_THREADED,$(CPPFLAGS)),)
-   $(error VM_TRACE_THREADS requires VM_THREADS)
-  endif
-  CPPFLAGS += -DVL_TRACE_THREADED
-  VK_C11=1
-  VK_LIBS_THREADED=1
- endif
-endif
-
-
-ifneq ($(VM_TRACE_FST_WRITER_THREAD),0)
- ifneq ($(VM_TRACE_FST_WRITER_THREAD),)
-  CPPFLAGS += -DVL_TRACE_FST_WRITER_THREAD
-  VK_C11=1
-  VK_LIBS_THREADED=1
- endif
-endif
-
 ifneq ($(VK_C11),0)
 ifneq ($(VK_C11),)
  # Need C++11 at least, so always default to newest
@@ -186,13 +167,17 @@ VM_SLOW += $(VM_CLASSES_SLOW) $(VM_SUPPORT_SLOW)
 VK_FAST_OBJS = $(addsuffix .o, $(VM_FAST))
 VK_SLOW_OBJS = $(addsuffix .o, $(VM_SLOW))

-VK_USER_OBJS   = $(addsuffix .o, $(VM_USER_CLASSES))
+VK_USER_OBJS = $(addsuffix .o, $(VM_USER_CLASSES))

 # Note VM_GLOBAL_FAST and VM_GLOBAL_SLOW holds the files required from the
 # run-time library. In practice everything is actually in VM_GLOBAL_FAST,
 # 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 +208,9 @@ 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 +219,9 @@ 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.
@@ -22,13 +22,16 @@
 //=============================================================================

 #include "verilatedos.h"
-#include "verilated.h"
+
 #include "verilated_cov.h"
+
+#include "verilated.h"
 #include "verilated_cov_key.h"

 #include <deque>
 #include <fstream>
 #include <map>
+#include <utility>

 //=============================================================================
 // VerilatedCovConst
@@ -58,7 +61,7 @@ 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;
 };

@@ -68,22 +71,23 @@ public:  // But only local to this file
 // This isn't in the header file for auto-magic conversion because it
 // inlines to too much code and makes compilation too slow.

-template <class T> class VerilatedCoverItemSpec final : public VerilatedCovImpItem {
+template <class T>
+class VerilatedCoverItemSpec final : public VerilatedCovImpItem {
 private:
    // MEMBERS
    T* m_countp;  // Count value
 public:
    // METHODS
    // cppcheck-suppress truncLongCastReturn
-    virtual vluint64_t count() const override { return *m_countp; }
-    virtual void zero() const override { *m_countp = 0; }
+    uint64_t count() const override { return *m_countp; }
+    void zero() const override { *m_countp = 0; }
    // CONSTRUCTORS
    // cppcheck-suppress noExplicitConstructor
    explicit VerilatedCoverItemSpec(T* countp)
        : m_countp{countp} {
        *m_countp = 0;
    }
-    virtual ~VerilatedCoverItemSpec() override = default;
+    ~VerilatedCoverItemSpec() override = default;
 };

 //=============================================================================
@@ -102,7 +106,7 @@ 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
@@ -121,11 +125,7 @@ public:

 protected:
    friend class VerilatedCovContext;
-    virtual ~VerilatedCovImp() override { clearGuts(); }
-    static VerilatedCovImp& imp() VL_MT_SAFE {
-        static VerilatedCovImp s_singleton;
-        return s_singleton;
-    }
+    ~VerilatedCovImp() override { clearGuts(); }

 private:
    // PRIVATE METHODS
@@ -206,7 +206,7 @@ private:
        }

        // Forward to . so we have a whole word
-        std::string suffix = *bpost ? std::string{bpost + 1} : "";
+        const std::string suffix = *bpost ? std::string{bpost + 1} : "";

        std::string out = prefix + "*" + suffix;

@@ -219,7 +219,7 @@ private:
        for (int i = 0; i < VerilatedCovConst::MAX_KEYS; ++i) {
            if (itemp->m_keys[i] != VerilatedCovConst::KEY_UNDEF) {
                // We don't compare keys, only values
-                std::string val = m_indexValues[itemp->m_vals[i]];
+                const std::string val = m_indexValues[itemp->m_vals[i]];
                if (std::string::npos != val.find(match)) {  // Found
                    return true;
                }
@@ -231,7 +231,7 @@ 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*");
@@ -302,7 +302,7 @@ public:
        // First two key/vals are filename
        ckeyps[0] = "filename";
        valps[0] = m_insertFilenamep;
-        std::string linestr = vlCovCvtToStr(m_insertLineno);
+        const std::string linestr = vlCovCvtToStr(m_insertLineno);
        ckeyps[1] = "lineno";
        valps[1] = linestr.c_str();
        // Default page if not specified
@@ -342,7 +342,7 @@ public:
                m_insertp->m_vals[addKeynum] = valueIndex(val);
                ++addKeynum;
                if (VL_UNCOVERABLE(!legalKey(key))) {
-                    std::string msg
+                    const std::string msg
                        = ("%Error: Coverage keys of one character, or letter+digit are illegal: "
                           + key);  // LCOV_EXCL_LINE
                    VL_FATAL_MT("", 0, "", msg.c_str());
@@ -358,7 +358,7 @@ 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 +371,7 @@ 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;
@@ -380,8 +380,9 @@ public:

            for (int i = 0; i < VerilatedCovConst::MAX_KEYS; ++i) {
                if (itemp->m_keys[i] != VerilatedCovConst::KEY_UNDEF) {
-                    std::string key = VerilatedCovKey::shortKey(m_indexValues[itemp->m_keys[i]]);
-                    std::string val = m_indexValues[itemp->m_vals[i]];
+                    const std::string key
+                        = VerilatedCovKey::shortKey(m_indexValues[itemp->m_keys[i]]);
+                    const std::string val = m_indexValues[itemp->m_vals[i]];
                    if (key == VL_CIK_PER_INSTANCE) {
                        if (val != "0") per_instance = true;
                    }
@@ -437,11 +438,11 @@ 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);
@@ -513,8 +514,10 @@ VerilatedCovContext* VerilatedCov::threadCovp() VL_MT_SAFE {

 VerilatedCovContext* VerilatedContext::coveragep() VL_MT_SAFE {
    static VerilatedMutex s_mutex;
+    // cppcheck-suppress identicalInnerCondition
    if (VL_UNLIKELY(!m_coveragep)) {
        const VerilatedLockGuard lock{s_mutex};
+        // cppcheck-suppress identicalInnerCondition
        if (VL_LIKELY(!m_coveragep)) {  // Not redundant, prevents race
            m_coveragep.reset(new VerilatedCovImp);
        }
@@ -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.
@@ -26,6 +26,7 @@
 #define VERILATOR_VERILATED_COV_H_

 #include "verilatedos.h"
+
 #include "verilated.h"

 #include <iostream>
@@ -72,7 +73,7 @@ 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;
@@ -88,7 +89,8 @@ class VerilatedCovImp;
 //=============================================================================
 // Convert VL_COVER_INSERT value arguments to strings, is \internal

-template <class T> std::string vlCovCvtToStr(const T& t) VL_PURE {
+template <class T>
+std::string vlCovCvtToStr(const T& t) VL_PURE {
    std::ostringstream os;
    os << t;
    return os.str();
@@ -120,14 +122,15 @@ public:
    /// Zero coverage points
    void zero() VL_MT_SAFE;

-public:  // But Internal use only
+    // METHODS - public but Internal use only
+
    // Insert a coverage item
    // We accept from 1-30 key/value pairs, all as strings.
    // 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 +160,8 @@ protected:
    friend class VerilatedCovImp;
    // CONSTRUCTORS
    // Internal: Only made as part of VerilatedCovImp
-    VerilatedCovContext() {}
-    virtual ~VerilatedCovContext() {}
+    VerilatedCovContext() = default;
+    ~VerilatedCovContext() override = 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.
@@ -28,7 +28,9 @@
 #define VERILATOR_VERILATED_DPI_CPP_

 #include "verilatedos.h"
+
 #include "verilated_dpi.h"
+
 #include "verilated_imp.h"

 // On MSVC++ we need svdpi.h to declare exports, not imports
@@ -68,7 +70,7 @@ svLogic svGetBitselLogic(const svLogicVecVal* sp, int bit) {
            | (((sp[VL_BITWORD_I(bit)].bval >> VL_BITBIT_I(bit)) & 1) << 1));
 }

-void svPutBitselBit(svBitVecVal* dp, int bit, svBit s) { VL_ASSIGNBIT_WI(32, bit, dp, s); }
+void svPutBitselBit(svBitVecVal* dp, int bit, svBit s) { VL_ASSIGNBIT_WI(bit, dp, s); }
 void svPutBitselLogic(svLogicVecVal* dp, int bit, svLogic s) {
    // Verilator doesn't support X/Z so only aval
    dp[VL_BITWORD_I(bit)].aval = ((dp[VL_BITWORD_I(bit)].aval & ~(VL_UL(1) << VL_BITBIT_I(bit)))
@@ -417,7 +419,7 @@ 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);
@@ -428,13 +430,13 @@ void* svGetArrElemPtr(const svOpenArrayHandle h, int indx1, ...) {
    switch (varp->udims()) {
    case 1: datap = _vl_svGetArrElemPtr(h, 1, indx1, 0, 0); break;
    case 2: {
-        int indx2 = va_arg(ap, int);
+        const int indx2 = va_arg(ap, int);
        datap = _vl_svGetArrElemPtr(h, 2, indx1, indx2, 0);
        break;
    }
    case 3: {
-        int indx2 = va_arg(ap, int);
-        int indx3 = va_arg(ap, int);
+        const int indx2 = va_arg(ap, int);
+        const int indx3 = va_arg(ap, int);
        datap = _vl_svGetArrElemPtr(h, 3, indx1, indx2, indx3);
        break;
    }
@@ -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.
@@ -27,6 +27,7 @@
 #define VERILATOR_VERILATED_DPI_H_

 #include "verilatedos.h"
+
 #include "verilated.h"  // Also presumably included by caller
 #include "verilated_sym_props.h"

@@ -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 +28,7 @@
 #include "verilated_fst_c.h"

 // GTKWave configuration
-#ifdef VL_TRACE_FST_WRITER_THREAD
+#ifdef VL_THREADED
 # define HAVE_LIBPTHREAD
 # define FST_WRITER_PARALLEL
 #endif
@@ -83,15 +83,16 @@ 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

-VerilatedFst::VerilatedFst(void* fst)
-    : m_fst{fst} {}
+VerilatedFst::VerilatedFst(void* /*fst*/) {}

 VerilatedFst::~VerilatedFst() {
    if (m_fst) fstWriterClose(m_fst);
@@ -104,14 +105,12 @@ void VerilatedFst::open(const char* filename) VL_MT_SAFE_EXCLUDES(m_mutex) {
    m_fst = fstWriterCreate(filename, 1);
    fstWriterSetPackType(m_fst, FST_WR_PT_LZ4);
    fstWriterSetTimescaleFromString(m_fst, timeResStr().c_str());  // lintok-begin-on-ref
-#ifdef VL_TRACE_FST_WRITER_THREAD
-    fstWriterSetParallelMode(m_fst, 1);
-#endif
+    if (m_useFstWriterThread) fstWriterSetParallelMode(m_fst, 1);
    fullDump(true);  // First dump must be full for fst

    m_curScope.clear();

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

    // Clear the scope stack
    auto it = m_curScope.begin();
@@ -122,7 +121,7 @@ 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 +132,23 @@ 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 +156,21 @@ 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 +196,9 @@ 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 +206,7 @@ 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 +223,108 @@ 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
+
+VerilatedFst::Buffer* VerilatedFst::getTraceBuffer() {
+#ifdef VL_THREADED
+    if (offload()) return new OffloadBuffer{*this};
+#endif
+    return new Buffer{*this};
+}
+
+void VerilatedFst::commitTraceBuffer(VerilatedFst::Buffer* bufp) {
+#ifdef VL_THREADED
+    if (offload()) {
+        OffloadBuffer* const offloadBufferp = static_cast<OffloadBuffer*>(bufp);
+        if (offloadBufferp->m_offloadBufferWritep) {
+            m_offloadBufferWritep = offloadBufferp->m_offloadBufferWritep;
+            return;  // Buffer will be deleted by the offload thread
+        }
+    }
+#endif
+    delete bufp;
+}
+
+//=============================================================================
+// Configure
+
+void VerilatedFst::configure(const VerilatedTraceConfig& config) {
+    // If at least one model requests the FST writer thread, then use it
+    m_useFstWriterThread |= config.m_useFstWriterThread;
+}
+
+//=============================================================================
+// VerilatedFstBuffer implementation
+
+//=============================================================================
+// 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 +340,6 @@ 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.
@@ -31,60 +31,64 @@
 #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 VerilatedFstBuffer;  // Give the buffer access to the private bits

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

-    void* m_fst;
-    std::map<vluint32_t, fstHandle> m_code2symbol;
+    void* m_fst = nullptr;
+    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
    char* m_strbuf = nullptr;  // String buffer long enough to hold maxBits() chars

+    bool m_useFstWriterThread = false;  // Whether to use the separate FST writer thread
+
    // 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
+    void emitTimeChange(uint64_t timeui) override;

    // Hooks called from VerilatedTrace
-    virtual bool preFullDump() override { return isOpen(); }
-    virtual bool preChangeDump() override { return isOpen(); }
+    bool preFullDump() override { return isOpen(); }
+    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
+    Buffer* getTraceBuffer() override;
+    void commitTraceBuffer(Buffer*) override;
+
+    // Configure sub-class
+    void configure(const VerilatedTraceConfig&) 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 +101,75 @@ 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 VL_NOT_FINAL {
+    // Give the trace file access to the private bits
+    friend VerilatedFst;
+    friend VerilatedFst::Super;
+    friend VerilatedFst::Buffer;
+    friend VerilatedFst::OffloadBuffer;
+
+    VerilatedFst& m_owner;  // Trace file owning this buffer. Required by subclasses.
+
+    // 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;
+
+    // CONSTRUCTOR
+    explicit VerilatedFstBuffer(VerilatedFst& owner)
+        : m_owner{owner} {}
+    virtual ~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 +186,7 @@ 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 +201,12 @@ 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 +225,14 @@ 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.
@@ -22,11 +22,23 @@
 //=============================================================================

 #include "verilatedos.h"
+
 #include "verilated_fst_sc.h"

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

+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
@@ -52,9 +64,6 @@ void VerilatedFstSc::trace(const unsigned int&, const std::string&, const char**
    DECL_TRACE_METHOD_B( unsigned short )
    DECL_TRACE_METHOD_B( unsigned int )
    DECL_TRACE_METHOD_B( unsigned long )
-#ifdef SYSTEMC_64BIT_PATCHES
-    DECL_TRACE_METHOD_B( unsigned long long)
-#endif
    DECL_TRACE_METHOD_B( char )
    DECL_TRACE_METHOD_B( short )
    DECL_TRACE_METHOD_B( int )
@@ -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.
@@ -23,8 +23,11 @@
 #define _VERILATED_FST_SC_H_ 1

 #include "verilatedos.h"
-#include "verilated_sc.h"
+
 #include "verilated_fst_c.h"
+#include "verilated_sc.h"
+
+#include <string>

 //=============================================================================
 // VerilatedFstSc
@@ -49,7 +52,7 @@ public:
        }
        spTrace()->set_time_resolution(sc_get_time_resolution().to_string());
    }
-    virtual ~VerilatedFstSc() { close(); }
+    virtual ~VerilatedFstSc() /*override*/ { close(); }

    // METHODS
    /// Called by SystemC simulate()
@@ -57,6 +60,10 @@ public:
        if (!delta_cycle) { this->dump(sc_time_stamp().to_double()); }
    }

+    // Override VerilatedFstC. Must be called after starting simulation.
+    // cppcheck-suppress missingOverride  // GCC won't accept override
+    virtual void open(const char* filename) /*override*/ VL_MT_SAFE;
+
 private:
    /// Fake outs for linker

@@ -90,9 +97,6 @@ private:
    DECL_TRACE_METHOD_B( unsigned short )
    DECL_TRACE_METHOD_B( unsigned int )
    DECL_TRACE_METHOD_B( unsigned long )
-#ifdef SYSTEMC_64BIT_PATCHES
-    DECL_TRACE_METHOD_B( unsigned long long)
-#endif
    DECL_TRACE_METHOD_B( char )
    DECL_TRACE_METHOD_B( short )
    DECL_TRACE_METHOD_B( int )
@@ -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
@@ -138,11 +143,10 @@ class VerilatedThreadMsgQueue final {

 public:
    // CONSTRUCTORS
-    VerilatedThreadMsgQueue() {}
-    ~VerilatedThreadMsgQueue() {
-        // The only call of this with a non-empty queue is a fatal error.
-        // So this does not flush the queue, as the destination queue is not known to this class.
-    }
+    VerilatedThreadMsgQueue() = default;
+    ~VerilatedThreadMsgQueue() = default;
+    // The only call of destructor with a non-empty queue is a fatal error.
+    // So this does not flush the queue, as the destination queue is not known to this class.

 private:
    VL_UNCOPYABLE(VerilatedThreadMsgQueue);
@@ -183,11 +187,11 @@ class VerilatedFpList final {

 public:
    using const_iterator = FILE* const*;
-    explicit VerilatedFpList() {}
+    explicit VerilatedFpList() = default;
    const_iterator begin() const { return m_fp; }
    const_iterator end() const { return m_fp + m_sz; }
    std::size_t size() const { return m_sz; }
-    std::size_t capacity() const { return 31; }
+    static std::size_t capacity() { return 31; }
    void push_back(FILE* fd) {
        if (VL_LIKELY(size() < capacity())) m_fp[m_sz++] = fd;
    }
@@ -231,17 +235,16 @@ class VerilatedContextImp final : VerilatedContext {
        return s_s;
    }

-private:
+public:  // But only for verilated*.cpp
    // CONSTRUCTORS - no data can live here, use only VerilatedContext
    VerilatedContextImp() = delete;
    ~VerilatedContextImp() = delete;

-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 {
@@ -267,17 +270,16 @@ public:  // But only for verilated*.cpp
    std::string argPlusMatch(const char* prefixp) VL_MT_SAFE_EXCLUDES(m_argMutex);
    std::pair<int, char**> argc_argv() VL_MT_SAFE_EXCLUDES(m_argMutex);

-public:  // But only for verilated*.cpp
-    // METHODS - scope name
+    // METHODS - scope name - INTERNAL only for verilated*.cpp
    void scopeInsert(const VerilatedScope* scopep) VL_MT_SAFE;
    void scopeErase(const VerilatedScope* scopep) VL_MT_SAFE;

-public:  // But only for verilated*.cpp
-    // METHODS - file IO
+    // METHODS - file IO - INTERNAL only for verilated*.cpp
+
    IData fdNewMcd(const char* filenamep) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
        const VerilatedLockGuard lock{m_fdMutex};
        if (m_fdFreeMct.empty()) return 0;
-        IData idx = m_fdFreeMct.back();
+        const IData idx = m_fdFreeMct.back();
        m_fdFreeMct.pop_back();
        m_fdps[idx] = std::fopen(filenamep, "w");
        if (VL_UNLIKELY(!m_fdps[idx])) return 0;
@@ -293,13 +295,13 @@ public:  // But only for verilated*.cpp
            const std::size_t start = std::max<std::size_t>(31UL + 1UL + 3UL, m_fdps.size());
            const std::size_t excess = 10;
            m_fdps.resize(start + excess);
-            std::fill(m_fdps.begin() + start, m_fdps.end(), (FILE*)0);
+            std::fill(m_fdps.begin() + start, m_fdps.end(), static_cast<FILE*>(nullptr));
            m_fdFree.resize(excess);
            for (std::size_t i = 0, id = start; i < m_fdFree.size(); ++i, ++id) {
                m_fdFree[i] = id;
            }
        }
-        IData idx = m_fdFree.back();
+        const IData idx = m_fdFree.back();
        m_fdFree.pop_back();
        m_fdps[idx] = fp;
        return (idx | (1UL << 31));  // bit 31 indicates not MCD
@@ -312,14 +314,14 @@ 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) {
@@ -334,12 +336,12 @@ public:  // But only for verilated*.cpp
        const VerilatedLockGuard lock{m_fdMutex};
        if (VL_BITISSET_I(fdi, 31)) {
            // Non-MCD case
-            IData idx = VL_MASK_I(31) & fdi;
+            const IData idx = VL_MASK_I(31) & fdi;
            if (VL_UNLIKELY(idx >= m_fdps.size())) return;
            if (VL_UNLIKELY(idx <= 2)) return;  // stdout/stdin/stderr
            if (VL_UNLIKELY(!m_fdps[idx])) return;  // Already free
            std::fclose(m_fdps[idx]);
-            m_fdps[idx] = (FILE*)0;
+            m_fdps[idx] = nullptr;
            m_fdFree.push_back(idx);
        } else {
            // MCD case
@@ -390,7 +392,11 @@ 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);
 };

@@ -455,7 +461,6 @@ public:
    // METHODS - debug
    static void versionDump() VL_MT_SAFE;

-public:
    // METHODS - user scope tracking
    // We implement this as a single large map instead of one map per scope.
    // There's often many more scopes than userdata's and thus having a ~48byte
@@ -477,10 +482,11 @@ public:
        return it->second;
    }

-public:  // But only for verilated.cpp
+    // METHODS - 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) {
@@ -503,8 +509,7 @@ public:  // But only for verilated.cpp
        }
    }

-public:  // But only for verilated*.cpp
-    // METHODS - hierarchy
+    // METHODS - hierarchy - only for verilated*.cpp
    static void hierarchyAdd(const VerilatedScope* fromp, const VerilatedScope* top) VL_MT_SAFE {
        // Slow ok - called at construction for VPI accessible elements
        const VerilatedLockGuard lock{s().m_hierMapMutex};
@@ -525,8 +530,7 @@ public:  // But only for verilated*.cpp
        return &s().m_hierMap;
    }

-public:  // But only for verilated*.cpp
-    // METHODS - export names
+    // METHODS - export names - only for verilated*.cpp

    // Each function prototype is converted to a function number which we
    // then use to index a 2D table also indexed by scope number, because we
@@ -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 +1,225 @@
+// -*- 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(VerilatedContext& context)
+    : m_context{context} {
+    // Setup profiling on main thread
+    setupThread(0);
+}
+
+void VlExecutionProfiler::configure() {
+
+    if (VL_UNLIKELY(m_enabled)) {
+        --m_windowCount;
+        if (VL_UNLIKELY(m_windowCount == m_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 = m_context.profExecFilename();
+            dump(fileName.c_str(), tickEnd);
+            m_enabled = false;
+        }
+        return;
+    }
+
+    const uint64_t startReq = m_context.profExecStart() + 1;  // + 1, so we can start at time 0
+
+    if (VL_UNLIKELY(m_lastStartReq < startReq && VL_TIME_Q() >= m_context.profExecStart())) {
+        VL_DEBUG_IF(VL_DBG_MSGF("+ profile start warmup\n"););
+        VL_DEBUG_IF(assert(m_windowCount == 0););
+        m_enabled = true;
+        m_windowCount = m_context.profExecWindow() * 2;
+        m_lastStartReq = startReq;
+    }
+}
+
+VerilatedVirtualBase* VlExecutionProfiler::construct(VerilatedContext& context) {
+    VlExecutionProfiler* const selfp = new VlExecutionProfiler{context};
+#if VL_THREADED
+    if (VlThreadPool* const threadPoolp = static_cast<VlThreadPool*>(context.threadPoolp())) {
+        for (int i = 0; i < threadPoolp->numThreads(); ++i) {
+            // Data to pass to worker thread initialization
+            struct Data {
+                VlExecutionProfiler* const selfp;
+                const uint32_t threadId;
+            } data{selfp, static_cast<uint32_t>(i + 1)};
+
+            // Initialize worker thread
+            threadPoolp->workerp(i)->addTask(
+                [](void* userp, bool) {
+                    Data* const datap = static_cast<Data*>(userp);
+                    datap->selfp->setupThread(datap->threadId);
+                },
+                &data);
+
+            // Wait until initializationis complete
+            threadPoolp->workerp(i)->wait();
+        }
+    }
+#endif
+    return selfp;
+}
+
+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
+    bool exists;
+    {
+        const VerilatedLockGuard lock{m_mutex};
+        exists = !m_traceps.emplace(threadId, &t_trace).second;
+    }
+    if (VL_UNLIKELY(exists)) {
+        VL_FATAL_MT(__FILE__, __LINE__, "", "multiple initialization of profiler on some thread");
+    }
+}
+
+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 +23,207 @@
 #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 {
-    std::string m_name;  // Hashed name of mtask/etc
-    size_t m_counterNumber = 0;  // Which counter has data
+#include "verilated.h"
+
+#include <array>
+#include <atomic>
+#include <cassert>
+#include <string>
+#include <type_traits>
+#include <vector>
+
+class VlExecutionProfiler;
+class VlThreadPool;
+
+//=============================================================================
+// Macros to simplify generated code
+
+#define VL_EXEC_TRACE_ADD_RECORD(vlSymsp) \
+    if (VL_UNLIKELY((vlSymsp)->__Vm_executionProfilerp->enabled())) \
+    (vlSymsp)->__Vm_executionProfilerp->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 : public VerilatedVirtualBase {
+    // 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
+    VerilatedContext& m_context;  // The context this profiler is under
+    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
+    explicit VlExecutionProfiler(VerilatedContext& context);
+    ~VlExecutionProfiler() override = default;
+
+    // METHODS
+
+    // Is profiling enabled
+    inline bool enabled() const { return m_enabled; }
+    // Append a trace record to the trace buffer of the current thread
+    static inline VlExecutionRecord& addRecord() {
+        t_trace.emplace_back();
+        return t_trace.back();
+    }
+    // Configure profiler (called in beginning of 'eval')
+    void configure();
+    // 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);
+
+    // Passed to VerilatedContext to create the VlExecutionProfiler profiler instance
+    static VerilatedVirtualBase* construct(VerilatedContext& context);
+};
+
+//=============================================================================
+// 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 +231,15 @@ 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 +248,9 @@ 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.
@@ -24,8 +24,10 @@
 #define VERILATOR_VERILATED_SAVE_CPP_

 #include "verilatedos.h"
-#include "verilated.h"
+
 #include "verilated_save.h"
+
+#include "verilated.h"
 #include "verilated_imp.h"

 #include <cerrno>
@@ -38,13 +40,13 @@
 # include <unistd.h>
 #endif

-#ifndef O_LARGEFILE  // For example on WIN32
+#ifndef O_LARGEFILE  // WIN32 headers omit this
 # define O_LARGEFILE 0
 #endif
-#ifndef O_NONBLOCK
+#ifndef O_NONBLOCK  // WIN32 headers omit this
 # define O_NONBLOCK 0
 #endif
-#ifndef O_CLOEXEC
+#ifndef O_CLOEXEC  // WIN32 headers omit this
 # define O_CLOEXEC 0
 #endif
 // clang-format on
@@ -63,8 +65,8 @@ 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 +194,7 @@ void VerilatedRestore::close() VL_MT_UNSAFE_ONE {
 void VerilatedSave::flush() VL_MT_UNSAFE_ONE {
    m_assertOne.check();
    if (VL_UNLIKELY(!isOpen())) return;
-    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 +221,8 @@ 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.
@@ -23,6 +23,7 @@
 #define VERILATOR_VERILATED_SAVE_C_H_

 #include "verilatedos.h"
+
 #include "verilated.h"

 #include <string>
@@ -40,8 +41,8 @@ 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 +59,7 @@ 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 +78,12 @@ 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 = static_cast<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 +112,9 @@ 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 +132,7 @@ protected:
 public:
    /// Construct
    VerilatedDeserialize() {
-        m_bufp = new vluint8_t[bufferSize()];
+        m_bufp = new uint8_t[bufferSize()];
        m_cp = m_bufp;
    }
    /// Destruct
@@ -150,12 +151,12 @@ 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 +166,7 @@ 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));
    }

@@ -194,16 +195,16 @@ public:
    /// Construct new object
    VerilatedSave() = default;
    /// Flush, close and destruct
-    virtual ~VerilatedSave() override { close(); }
+    ~VerilatedSave() override { close(); }
    // METHODS
    /// Open the file; call isOpen() to see if errors
    void open(const char* filenamep) VL_MT_UNSAFE_ONE;
    /// Open the file; call isOpen() to see if errors
    void open(const std::string& filename) VL_MT_UNSAFE_ONE { open(filename.c_str()); }
    /// Flush and close the file
-    virtual void close() override VL_MT_UNSAFE_ONE;
+    void close() override VL_MT_UNSAFE_ONE;
    /// Flush data to file
-    virtual void flush() override VL_MT_UNSAFE_ONE;
+    void flush() override VL_MT_UNSAFE_ONE;
 };

 //=============================================================================
@@ -221,7 +222,7 @@ public:
    /// Construct new object
    VerilatedRestore() = default;
    /// Flush, close and destruct
-    virtual ~VerilatedRestore() override { close(); }
+    ~VerilatedRestore() override { close(); }

    // METHODS
    /// Open the file; call isOpen() to see if errors
@@ -229,35 +230,35 @@ public:
    /// Open the file; call isOpen() to see if errors
    void open(const std::string& filename) VL_MT_UNSAFE_ONE { open(filename.c_str()); }
    /// Close the file
-    virtual void close() override VL_MT_UNSAFE_ONE;
-    virtual void flush() override VL_MT_UNSAFE_ONE {}
-    virtual void fill() override VL_MT_UNSAFE_ONE;
+    void close() override VL_MT_UNSAFE_ONE;
+    void flush() override VL_MT_UNSAFE_ONE {}
+    void fill() override VL_MT_UNSAFE_ONE;
 };

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

-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) {
@@ -279,15 +280,16 @@ inline VerilatedDeserialize& operator>>(VerilatedDeserialize& os, float& rhs) {
    return os.read(&rhs, sizeof(rhs));
 }
 inline VerilatedSerialize& operator<<(VerilatedSerialize& os, const std::string& rhs) {
-    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);
+    // C cast is required below
+    return os.read((void*)(rhs.data()), len);
 }
 VerilatedSerialize& operator<<(VerilatedSerialize& os, VerilatedContext* rhsp);
 VerilatedDeserialize& operator>>(VerilatedDeserialize& os, VerilatedContext* rhsp);
@@ -295,7 +297,7 @@ 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 +309,10 @@ 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
@@ -134,7 +134,6 @@ public:
        initUnpacked(ulims);
    }

-public:
    ~VerilatedVarProps() = default;
    // METHODS
    bool magicOk() const { return m_magic == MAGIC; }
@@ -142,7 +141,7 @@ 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); }
@@ -192,8 +191,8 @@ public:

 class VerilatedDpiOpenVar final {
    // MEMBERS
-    const VerilatedVarProps* m_propsp;  // Variable properties
-    void* m_datap;  // Location of data (local to thread always, so safe)
+    const VerilatedVarProps* const m_propsp;  // Variable properties
+    void* const m_datap;  // Location of data (local to thread always, so safe)
 public:
    // CONSTRUCTORS
    VerilatedDpiOpenVar(const VerilatedVarProps* propsp, void* datap)
@@ -230,10 +229,10 @@ public:

 class VerilatedVar final : public VerilatedVarProps {
    // MEMBERS
-    void* m_datap;  // Location of data
-    const char* m_namep;  // Name - slowpath
+    void* const m_datap;  // Location of data
+    const char* const m_namep;  // Name - slowpath
 protected:
-    bool m_isParam;
+    const bool m_isParam;
    friend class VerilatedScope;
    // CONSTRUCTORS
    VerilatedVar(const char* namep, void* datap, VerilatedVarType vltype,
@@ -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.
@@ -30,6 +30,7 @@
 #define VERILATOR_VERILATED_SYMS_H_

 #include "verilatedos.h"
+
 #include "verilated.h"
 #include "verilated_sym_props.h"

@@ -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.
@@ -22,24 +22,24 @@
 //=============================================================================

 #include "verilatedos.h"
+
 #include "verilated_threads.h"

 #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));
@@ -48,176 +48,62 @@ VlMTaskVertex::VlMTaskVertex(vluint32_t upstreamDepCount)
 //=============================================================================
 // VlWorkerThread

-VlWorkerThread::VlWorkerThread(VlThreadPool* poolp, VerilatedContext* contextp, bool profiling)
+VlWorkerThread::VlWorkerThread(VerilatedContext* contextp)
    : 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_contextp{contextp} {}
+    , m_cthread{startWorker, this, contextp} {}

 VlWorkerThread::~VlWorkerThread() {
-    m_exiting.store(true, std::memory_order_release);
-    wakeUp();
+    shutdown();
    // The thread should exit; join it.
    m_cthread.join();
 }

-void VlWorkerThread::workerLoop() {
-    if (VL_UNLIKELY(m_profiling)) m_poolp->setupProfilingClientThread();
-
-    ExecRec work;
-    work.m_fnp = nullptr;
-
-    while (true) {
-        if (VL_LIKELY(!work.m_fnp)) dequeWork(&work);
-
-        // Do this here, not above, to avoid a race with the destructor.
-        if (VL_UNLIKELY(m_exiting.load(std::memory_order_acquire))) break;
-
-        if (VL_LIKELY(work.m_fnp)) {
-            work.m_fnp(work.m_selfp, work.m_evenCycle);
-            work.m_fnp = nullptr;
-        }
-    }
-
-    if (VL_UNLIKELY(m_profiling)) m_poolp->tearDownProfilingClientThread();
+static void shutdownTask(void*, bool) {
+    // Deliberately empty, we use the address of this function as a magic number
 }

-void VlWorkerThread::startWorker(VlWorkerThread* workerp) {
-    Verilated::threadContextp(workerp->m_contextp);
+void VlWorkerThread::shutdown() { addTask(shutdownTask, nullptr); }
+
+void VlWorkerThread::wait() {
+    // Enqueue a task that sets this flag. Execution is in-order so this ensures completion.
+    std::atomic<bool> flag{false};
+    addTask([](void* flagp, bool) { static_cast<std::atomic<bool>*>(flagp)->store(true); }, &flag);
+    // Spin wait
+    for (unsigned i = 0; i < VL_LOCK_SPINS; ++i) {
+        if (flag.load()) return;
+        VL_CPU_RELAX();
+    }
+    // Yield wait
+    while (!flag.load()) std::this_thread::yield();
+}
+
+void VlWorkerThread::workerLoop() {
+    ExecRec work;
+
+    // Wait for the first task without spinning, in case the thread is never actually used.
+    dequeWork</* SpinWait: */ false>(&work);
+
+    while (true) {
+        if (VL_UNLIKELY(work.m_fnp == shutdownTask)) break;
+        work.m_fnp(work.m_selfp, work.m_evenCycle);
+        // Wait for next task with spinning.
+        dequeWork</* SpinWait: */ true>(&work);
+    }
+}
+
+void VlWorkerThread::startWorker(VlWorkerThread* workerp, VerilatedContext* contextp) {
+    Verilated::threadContextp(contextp);
    workerp->workerLoop();
 }

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

-VlThreadPool::VlThreadPool(VerilatedContext* contextp, int nThreads, bool profiling)
-    : m_profiling{profiling} {
-    // --threads N passes nThreads=N-1, as the "main" threads counts as 1
-    const unsigned cpus = std::thread::hardware_concurrency();
-    if (cpus < nThreads + 1) {
-        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);
-        }
-    }
-    // Create'em
-    for (int i = 0; i < nThreads; ++i) {
-        m_workers.push_back(new VlWorkerThread{this, contextp, profiling});
-    }
-    // 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(VerilatedContext* contextp, unsigned nThreads) {
+    for (unsigned i = 0; i < nThreads; ++i) m_workers.push_back(new VlWorkerThread{contextp});
 }

 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.
@@ -24,6 +24,7 @@
 #define VERILATOR_VERILATED_THREADS_H_

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

 #ifndef VL_THREADED
@@ -35,19 +36,24 @@
 #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

+class VlExecutionProfiler;
+class VlThreadPool;
+
 // VlMTaskVertex and VlThreadpool will work with multiple model class types.
 // Since the type is opaque to VlMTaskVertex and VlThreadPool, represent it
 // as a void* here.
@@ -58,7 +64,7 @@ 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 +82,8 @@ 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 +91,10 @@ 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 +105,19 @@ 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 {
@@ -127,77 +133,14 @@ 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 VlThreadPool;
-
 class VlWorkerThread final {
 private:
    // TYPES
    struct ExecRec {
-        VlExecFnp m_fnp;  // Function to execute
-        VlSelfP m_selfp;  // Symbol table to execute
-        bool m_evenCycle;  // Even/odd for flag alternation
-        ExecRec()
-            : m_fnp{nullptr}
-            , m_selfp{nullptr}
-            , m_evenCycle{false} {}
+        VlExecFnp m_fnp = nullptr;  // Function to execute
+        VlSelfP m_selfp = nullptr;  // Symbol table to execute
+        bool m_evenCycle = false;  // Even/odd for flag alternation
+        ExecRec() = default;
        ExecRec(VlExecFnp fnp, VlSelfP selfp, bool evenCycle)
            : m_fnp{fnp}
            , m_selfp{selfp}
@@ -205,7 +148,7 @@ 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;
@@ -217,28 +160,24 @@ private:
    // Store the size atomically, so we can spin wait
    std::atomic<size_t> m_ready_size;

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

    VL_UNCOPYABLE(VlWorkerThread);

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

    // METHODS
+    template <bool SpinWait>
    inline void dequeWork(ExecRec* workp) VL_MT_SAFE_EXCLUDES(m_mutex) {
        // Spin for a while, waiting for new data
-        for (int i = 0; i < VL_LOCK_SPINS; ++i) {
-            if (VL_LIKELY(m_ready_size.load(std::memory_order_relaxed))) {  //
-                break;
+        if VL_CONSTEXPR_CXX17 (SpinWait) {
+            for (unsigned i = 0; i < VL_LOCK_SPINS; ++i) {
+                if (VL_LIKELY(m_ready_size.load(std::memory_order_relaxed))) break;
+                VL_CPU_RELAX();
            }
-            VL_CPU_RELAX();
        }
        VerilatedLockGuard lock{m_mutex};
        while (m_ready.empty()) {
@@ -252,8 +191,7 @@ public:
        m_ready.erase(m_ready.begin());
        m_ready_size.fetch_sub(1, std::memory_order_relaxed);
    }
-    inline void wakeUp() { addTask(nullptr, nullptr, false); }
-    inline void addTask(VlExecFnp fnp, VlSelfP selfp, bool evenCycle)
+    inline void addTask(VlExecFnp fnp, VlSelfP selfp, bool evenCycle = false)
        VL_MT_SAFE_EXCLUDES(m_mutex) {
        bool notify;
        {
@@ -264,36 +202,25 @@ public:
        }
        if (notify) m_cv.notify_one();
    }
+
+    void shutdown();  // Finish current tasks, then terminate thread
+    void wait();  // Blocks calling thread until all tasks complete in this thread
+
    void workerLoop();
-    static void startWorker(VlWorkerThread* workerp);
+    static void startWorker(VlWorkerThread* workerp, VerilatedContext* contextp);
 };

-class VlThreadPool final {
-    // TYPES
-    using ProfileTrace = std::vector<VlProfileRec>;
-
+class VlThreadPool final : public VerilatedVirtualBase {
    // MEMBERS
    std::vector<VlWorkerThread*> m_workers;  // our workers
-    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();
+    VlThreadPool(VerilatedContext* contextp, unsigned nThreads);
+    ~VlThreadPool() override;

    // METHODS
    inline int numThreads() const { return m_workers.size(); }
@@ -302,17 +229,6 @@ 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.
@@ -27,25 +27,36 @@
 #include "verilated.h"
 #include "verilated_trace_defs.h"

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

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

 // clang-format on

-#ifdef VL_TRACE_THREADED
+class VlThreadPool;
+template <class T_Buffer>
+class VerilatedTraceBuffer;
+template <class T_Buffer>
+class VerilatedTraceOffloadBuffer;
+
+#ifdef VL_THREADED
 //=============================================================================
-// Threaded tracing
+// Offloaded tracing

 // A simple synchronized first in first out queue
-template <class T> class VerilatedThreadQueue final {  // LCOV_EXCL_LINE  // lcov bug
+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 +97,11 @@ 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,
@@ -100,36 +111,65 @@ 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
    };
 };
 #endif

+//=============================================================================
+// VerilatedTraceConfig
+
+// Simple data representing trace configuration required by generated models.
+class VerilatedTraceConfig final {
+public:
+    const bool m_useParallel;  // Use parallel tracing
+    const bool m_useOffloading;  // Offloading trace rendering
+    const bool m_useFstWriterThread;  // Use the separate FST writer thread
+
+    VerilatedTraceConfig(bool useParallel, bool useOffloading, bool useFstWriterThread)
+        : m_useParallel{useParallel}
+        , m_useOffloading{useOffloading}
+        , m_useFstWriterThread{useFstWriterThread} {}
+};
+
 //=============================================================================
 // 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 base class of VerilatedTraceBuffer.
+template <class T_Trace, class T_Buffer>
+class VerilatedTrace VL_NOT_FINAL {
 public:
+    using Buffer = VerilatedTraceBuffer<T_Buffer>;
+    using OffloadBuffer = VerilatedTraceOffloadBuffer<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 dumpOffloadCb_t = void (*)(void*, OffloadBuffer*);  // Type of offload dump callbacks
+    using cleanupCb_t = void (*)(void*, T_Trace*);  // Type of cleanup callbacks

 private:
+    // Give the buffer (both base and derived) access to the private bits
+    friend T_Buffer;
+    friend Buffer;
+    friend OffloadBuffer;
+
    struct CallbackRecord {
        // Note: would make these fields const, but some old STL implementations
        // (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;
+            dumpOffloadCb_t m_dumpOffloadCb;
+            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 +178,110 @@ private:
        CallbackRecord(dumpCb_t cb, void* userp)
            : m_dumpCb{cb}
            , m_userp{userp} {}
+        CallbackRecord(dumpOffloadCb_t cb, void* userp)
+            : m_dumpOffloadCb{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
-    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)
+    bool m_offload = false;  // Use the offload thread (ignored if !VL_THREADED)
+    bool m_parallel = false;  // Use parallel tracing (ignored if !VL_THREADED)

-    void addCallbackRecord(std::vector<CallbackRecord>& cbVec, CallbackRecord& cbRec)
+#ifdef VL_THREADED
+    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
+
+protected:
+    uint32_t* m_sigs_oldvalp = nullptr;  // Previous value store
+    EData* m_sigs_enabledp = nullptr;  // Bit vector of enabled codes (nullptr = all on)
+private:
+    std::vector<bool> m_sigs_enabledVec;  // Staging for m_sigs_enabledp
+    std::vector<CallbackRecord> m_initCbs;  // Routines to initialize tracing
+    std::vector<CallbackRecord> m_fullCbs;  // Routines to perform full dump
+    std::vector<CallbackRecord> m_fullOffloadCbs;  // Routines to perform offloaded full dump
+    std::vector<CallbackRecord> m_chgCbs;  // Routines to perform incremental dump
+    std::vector<CallbackRecord> m_chgOffloadCbs;  // Routines to perform offloaded incremental dump
+    std::vector<CallbackRecord> m_cleanupCbs;  // Routines to call at the end of dump
+    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)
+    uint64_t m_timeLastDump = 0;  // Last time we did a dump
+    bool m_didSomeDump = false;  // Did at least one dump (i.e.: m_timeLastDump is valid)
+    VerilatedContext* m_contextp = nullptr;  // The context used by the traced models
+    std::unordered_set<const VerilatedModel*> m_models;  // The collection of models being traced
+
+    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 runCallbacks(const std::vector<CallbackRecord>& cbVec);
+    void runOffloadedCallbacks(const std::vector<CallbackRecord>& cbVec);

    // 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_THREADED
+    // 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 +291,12 @@ 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; }

    double timeRes() const { return m_timeRes; }
    double timeUnit() const { return m_timeUnit; }
@@ -223,27 +304,45 @@ 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();

+#ifdef VL_THREADED
+    inline bool offload() const { return m_offload; }
+    inline bool parallel() const { return m_parallel; }
+#else
+    static constexpr bool offload() { return false; }
+    static constexpr bool parallel() { return false; }
+#endif
+
    //=========================================================================
    // 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;
+
+    // Configure sub-class
+    virtual void configure(const VerilatedTraceConfig&) = 0;
+
 public:
    //=========================================================================
    // External interface to client code
@@ -257,21 +356,58 @@ 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 addModel(VerilatedModel*) VL_MT_SAFE_EXCLUDES(m_mutex);
    void addInitCb(initCb_t cb, void* userp) VL_MT_SAFE;
    void addFullCb(dumpCb_t cb, void* userp) VL_MT_SAFE;
+    void addFullCb(dumpOffloadCb_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 addChgCb(dumpOffloadCb_t cb, void* userp) 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_Buffer is the format specific base class of VerilatedTraceBuffer.
+// The format-specific hot-path methods use duck-typing via T_Buffer for performance.
+template <class T_Buffer>
+class VerilatedTraceBuffer VL_NOT_FINAL : public T_Buffer {
+protected:
+    // Type of the owner trace file
+    using Trace = typename std::remove_cv<
+        typename std::remove_reference<decltype(T_Buffer::m_owner)>::type>::type;
+
+    static_assert(std::has_virtual_destructor<T_Buffer>::value, "");
+    static_assert(std::is_base_of<VerilatedTrace<Trace, T_Buffer>, Trace>::value, "");
+
+    friend Trace;  // Give the trace file access to the private bits
+    friend std::default_delete<VerilatedTraceBuffer<T_Buffer>>;
+
+    uint32_t* const m_sigs_oldvalp;  // Previous value store
+    EData* const m_sigs_enabledp;  // Bit vector of enabled codes (nullptr = all on)
+
+    explicit VerilatedTraceBuffer(Trace& owner);
+    ~VerilatedTraceBuffer() override = default;
+
+public:
    //=========================================================================
    // Hot path internal interface to Verilator generated code

@@ -281,108 +417,51 @@ 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););
-    }
-    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 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 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 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 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 chgDouble(vluint32_t code, double newval) {
-        m_traceBufferWritep[0] = VerilatedTraceCommand::CHG_DOUBLE;
-        m_traceBufferWritep[1] = code;
-        // cppcheck-suppress invalidPointerCast
-        *reinterpret_cast<double*>(m_traceBufferWritep + 2) = newval;
-        m_traceBufferWritep += 4;
-        VL_DEBUG_IF(assert(m_traceBufferWritep <= m_traceBufferEndp););
-    }
-
-#define CHG(name) chg##name##Impl
-#else
-#define CHG(name) chg##name
-#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 +469,85 @@ 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_THREADED
+//=============================================================================
+// VerilatedTraceOffloadBuffer
+
+// T_Buffer is the format specific base class of VerilatedTraceBuffer.
+// The format-specific hot-path methods use duck-typing via T_Buffer for performance.
+template <class T_Buffer>
+class VerilatedTraceOffloadBuffer final : public VerilatedTraceBuffer<T_Buffer> {
+    using typename VerilatedTraceBuffer<T_Buffer>::Trace;
+
+    friend Trace;  // Give the trace file access to the private bits
+
+    uint32_t* m_offloadBufferWritep;  // Write pointer into current buffer
+    uint32_t* const m_offloadBufferEndp;  // End of offload buffer
+
+    explicit VerilatedTraceOffloadBuffer(Trace& owner);
+    ~VerilatedTraceOffloadBuffer() override = default;
+
+public:
+    //=========================================================================
+    // Hot path internal interface to Verilator generated code
+
+    // 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(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(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(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(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(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(uint32_t code, double newval) {
+        m_offloadBufferWritep[0] = VerilatedTraceOffloadCommand::CHG_DOUBLE;
+        m_offloadBufferWritep[1] = code;
+        // cppcheck-suppress invalidPointerCast
+        *reinterpret_cast<double*>(m_offloadBufferWritep + 2) = newval;
+        m_offloadBufferWritep += 4;
+        VL_DEBUG_IF(assert(m_offloadBufferWritep <= m_offloadBufferEndp););
+    }
+};
+#endif
+
 #endif  // guard
--- a/Show More
+++ b/Show More