Commit Graph

5086 Commits

Author SHA1 Message Date
Geza Lore b51f887567
Perform VCD tracing in parallel when using --threads (#3449)
VCD tracing is now parallelized using the same thread pool as the model.
We achieve this by breaking the top level trace functions into multiple
top level functions (as many as --threads), and after emitting the time
stamp to the VCD file on the main thread, we execute the tracing
functions in parallel on the same thread pool as the model (which we
pass to the trace file during registration), tracing into a secondary
per thread buffer. The main thread will then stitch (memcpy) the buffers
together into the output file.

This makes the `--trace-threads` option redundant with `--trace`, which
now only affects `--trace-fst`. FST tracing uses the previous offloading
scheme.

This obviously helps a lot in VCD tracing performance, and I have seen
better than Amdahl speedup, namely I get 3.9x on XiangShan 4T (2.7x on
OpenTitan 4T).
2022-05-29 19:08:39 +01:00
Geza Lore c4b8675d77 Always inline some small, hot trace routines 2022-05-28 12:47:09 +01:00
Geza Lore a7cd7a1ed9 Initialize VerilatedTrace members in class 2022-05-28 12:47:07 +01:00
Geza Lore a48c779367 Rename verilated_trace_imp.cpp -> verilated_trace_imp.h
Also fix file header to describe purpose of this file.
2022-05-28 12:20:35 +01:00
Geza Lore cf1eccc24f Make local function 'static' in verilated_profiler.h 2022-05-28 12:17:39 +01:00
Geza Lore d45caca011 Remove legacy VCD tracing API
This has not been used by Verilator for a while, but was kept for
compatibility with some external code. Now removed.
2022-05-28 12:07:24 +01:00
Geza Lore 0722f47539
Improve V3MergeCond by reordering statements (#3125)
V3MergeCond merges consecutive conditional `_ = cond ? _ : _` and
`if (cond) ...` statements. This patch adds an analysis and ordering
phase that moves statements with identical conditions closer to each
other, in order to enable more merging opportunities. This in turn
eliminates a lot of repeated conditionals which reduced dynamic branch
count and branch misprediction rate. Observed 6.5% improvement on
multi-threaded large designs, at the cost of less than 2% increase in
Verilation speed.
2022-05-27 16:57:51 +01:00
github action a372e010bd Apply 'make format' 2022-05-25 04:51:51 +00:00
Wilson Snyder 530817191e Support non-ANSI interface port declarations (#3439). 2022-05-25 00:50:50 -04:00
Geza Lore 1282548a1c Only consider definitions in t_flag_csplit 2022-05-20 18:02:00 +01:00
Geza Lore c7610ed044 Fix FST tracing thread in CMake build 2022-05-20 17:04:46 +01:00
Geza Lore ffc1c51526 Commentary 2022-05-20 16:44:53 +01:00
Geza Lore b130a8cfeb Add -DVM_TRACE_VCD in model builds with Make with --trace 2022-05-20 16:44:38 +01:00
Geza Lore 551bd284dd Rename some internals related to multi-threaded tracing
Rename the implementation internals of current multi-threaded tracing to
be "offload mode". No functional change, nor user interface change
intended.
2022-05-20 16:44:35 +01:00
Krzysztof Bieganski 9edccfdffa
Initial support for intra-assignment timing controls, net delays (#3427)
This is a pre-PR to #3363.

Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-17 19:19:44 +01:00
Geza Lore 1a056f6db9 Fix invalid conditional merging when starting at 'c = c ? a : b'
Fixes #3409.
2022-05-17 18:36:40 +01:00
Krzysztof Bieganski e018eb7bac
Support AstClass::repairCache() after V3Class (#3431)
This is a pre-PR to #3363.

Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-17 09:22:43 -04:00
Krzysztof Bieganski 0a91ddf38a
Tests: Better grep check in t_foreach (#3435)
This is a pre-PR to #3363.

Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-17 09:20:59 -04:00
Krzysztof Bieganski 67bb2c640e
Tests: Rename t_timing_clkgen to t_timing_clkgen1 (#3430)
This is a pre-PR to #3363, which will introduce more clock gen tests.

Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-17 09:19:51 -04:00
Krzysztof Bieganski 561eaa311d
Tests: Enable CI testing with GCC 10 (#3432)
This is a pre-PR to #3363.

Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-17 09:19:06 -04:00
Krzysztof Bieganski 3f7a248ed4
Refactor some of the Begin handling to a separate function (#3426)
Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-16 20:45:33 +01:00
Krzysztof Bieganski ecaa07a72a
Rename AstTimingControl to AstEventControl (#3425)
Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-05-16 20:44:41 +01:00
Wilson Snyder 99bdc27be3 Internals: Cleanup some statics, trivial part towards (#3419) 2022-05-15 14:26:55 -04:00
Wilson Snyder c3c46967dc Tests: Appease sanitizer (#3121). 2022-05-15 11:50:52 -04:00
Wilson Snyder 7f1a9239ab Commentary, fix typo (#3121) 2022-05-15 11:14:07 -04:00
Geza Lore c8102c8ffe Fix typo 2022-05-15 16:01:35 +01:00
Wilson Snyder 5aa12e9b51 Add assert when VerilatedContext is mis-deleted (#3121). 2022-05-15 10:51:03 -04:00
Geza Lore 829437b20b Commentary - dependencies 2022-05-15 15:25:46 +01:00
Wilson Snyder 3c4131d45d Fix 'with' operator with type casting (#3387). 2022-05-15 09:53:48 -04:00
Wilson Snyder ae8d8ee1ac Fix crash with misuse of display. 2022-05-15 09:29:45 -04:00
Geza Lore 89ec3d16dc Allow const nodes in VNRef
No functional change.
2022-05-15 13:30:07 +01:00
Wilson Snyder 7602c983c5 Commentary 2022-05-14 18:16:31 -04:00
HungMingWu 560efb2c9e
Internals: Fix memory leak in V3FileLine (#3407) (#3408). No functional change intended. 2022-05-14 18:15:38 -04:00
Wilson Snyder c2328ef46a Spelling fixes. 2022-05-14 16:12:57 -04:00
Wilson Snyder f6035447ae Internals: Use mutable for mutexes. No functional change. 2022-05-13 07:21:39 -04:00
Wilson Snyder 38438b3373 Internals: Cleanup some defaults. No functional change. 2022-05-12 23:30:39 -04:00
Wilson Snyder 71dedccbbe Support compile time trace signal selection with tracing_on/off (#3323). 2022-05-12 22:28:08 -04:00
Martin Stadler cddbe4642f
Fix cmake rules to support higher-level targest (#3377) (#3386).
Don't add linker flags as generator expression to support linking `TARGET` to higher-level targets in a top-level CMakeLists.txt file.
2022-05-11 21:33:05 -04:00
Wilson Snyder bdfdc737a0 Internals: Cleanup V3Config. No functional change intended. 2022-05-11 00:47:52 -04:00
Wilson Snyder 3d045c3aee Internals: Cleanup some verilog.y formatting. No functional change. 2022-05-09 00:37:51 -04:00
HungMingWu 9583f152ee
Fix compile error when enable VL_LEAK_CHECKS (#3411).
Signed-off-by: HungMingWu <u9089000@gmail.com>
2022-05-08 20:49:13 -04:00
Wilson Snyder 5b2755d28d Untabify verilog.y (#3412). No functional change. 2022-05-08 20:46:18 -04:00
Kamil Rakoczy 9378259779
Fix UNOPTFLAT warning from initial static var (#3406)
Signed-off-by: Kamil Rakoczy <krakoczy@antmicro.com>
2022-05-06 10:24:03 +02:00
Wilson Snyder 3d762282b9 Fix hang with large case statement optimization (#3405). 2022-05-05 07:02:52 -04:00
Wilson Snyder 30783e6a79 devel release 2022-05-02 22:23:05 -04:00
Wilson Snyder aa86c777f4 Version bump 2022-05-02 22:17:20 -04:00
Geza Lore a2792785fe Add V3GraphVertex::dotRank to add GraphViz ranks to graph dumps
This is a simple debugging aid to allow constraining the graph layout
via GraphViz rank directives. Note this is not related in any way to the
vertex 'rank' attribute used by some of the graph algorithms.

No functional change.
2022-05-02 10:27:26 +01:00
Wilson Snyder 267315e7d4 Commentary: Update ChangeLog 2022-05-01 22:01:30 -04:00
Wilson Snyder 99d39cb6ff Fix autoconf warning 2022-05-01 10:43:04 -04:00
Wilson Snyder 915ceb2d04 Tests: Untabify tests. No functional change. 2022-05-01 10:10:00 -04:00