luke
/
verilator
mirror of https://github.com/verilator/verilator.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
9,909 Commits 3 Branches 170 Tags 224 MiB
Commit Graph

278 Commits

Author SHA1 Message Date
Thomas Santerre bd6b9161dc
Optimize bit-scan loops into $mostsetbitp1 / $countones (#7822) 
Recognize the common single-bit scan loop idioms in V3Unroll (before it
unrolls) and lower them to bit-reduction primitives, replacing a literal
W-iteration loop with one intrinsic-backed expression:

  target=0; for (i=0;i<W;i++) if (vec[i]) target = i + 1;      -> $mostsetbitp1(vec)
  target=0; for (i=0;i<W;i++) if (vec[i]) target = target + 1; -> $countones(vec)

The leading-one form lowers to a new AstMostSetBitP1 node, emitted as
VL_MOSTSETBITP1_{I,Q,W}; those runtime helpers now use __builtin_clz where
available (same pattern as VL_REDXOR's __builtin_parity), with the existing
bit scan as fallback.  The count-ones form reuses AstCountOnes ($countones,
popcount); as the DFG requires a 32-bit countones result it is built at 32
bits and narrowed to the accumulator width with a select.

Matching is structural to stay sound: the index must start at 0, increment
by exactly 1, and scan all W==width(vec) bits via a single 1-bit select of a
distinct vector, with the target pre-zeroed and no else branch.  The loop
bound is accepted as a strict ascending 'idx < W' written either way and
signed or unsigned (Gt/GtS/Lt/LtS).  Gated by -fbit-scan-loops (on at -O).

Adds t_bit_scan_loops (I/Q/W, count-ones and unsigned-index positives;
step-2, start-1, idx*2+1, vec[idx+1], target=idx and W!=width negatives, all
self-checked and asserted via --stats not to lower) plus t_bit_scan_loops_off
for the disable flag.

Motivated by a transformer inference design whose 80-bit leading-one detector
ran every cycle (~37% of runtime); the lowering is worth ~39% there.
 2026-06-24 10:43:05 +01:00
Geza Lore c76c94ef16
Optimize additional expression patterns in V3Const (#7824) 
- Masking that returns known zero
- More general Sel over Extend
 2026-06-23 04:22:35 +01:00
Geza Lore f11c4084d1 Internals: Minor cleanup of case statement lowering 
- Factor out match mask construction
- Rename V3Number method to reflect behaviour
 2026-06-17 14:39:06 +01:00
Igor Zaworski 25d4827bd5
Internals: Four state pre-pull (types) (#7520) 2026-04-30 16:56:15 -04:00
Wilson Snyder 31757df229
Internals: clangtidy cleanups. No functional change intended (#7343) 2026-03-27 23:14:18 -04:00
Wilson Snyder 602ee384de
Support $sformat with runtime format string (#7212). (#7257) 
Fixes #7212.
 2026-03-14 22:43:56 -04:00
Wilson Snyder 08c6d5bde5 Improve some display error handling (#7212 prep) 2026-03-12 07:49:21 -04:00
Wilson Snyder d87035d187 Commentary: Fix non-grammar 2026-03-11 19:53:23 -04:00
Todd Strader 6a5d3b0b72
Add --max-replication option (#7139) 2026-02-23 16:51:37 -05:00
Wilson Snyder 9049c93009 Internals: Call common V3NUmber function to zero data type. Cleanup chandle clearling. 2026-02-17 06:18:57 -05:00
Wilson Snyder 7c6c6a684b Add SPDX copyright identifiers, and get 'reuse' clean. No functional change. 2026-01-26 20:24:34 -05:00
Wilson Snyder 13327fa9c0 Copyright year update. 2026-01-01 07:22:09 -05:00
Wilson Snyder 5c0ad5bd1f
Internals: cppcheck fixes. No functional change. (#6687) 2025-11-12 18:54:22 -05:00
Wilson Snyder eae1e5bdf2 Internals/Tests: Cleanup some missing dev coverage items 2025-10-11 20:58:03 -04:00
Geza Lore 838b8a2bec Internals: Remove dead code, fix or sign off coverage 
Remove/fix/signoff uncontroversial code coverage holes. Also added a
couple TODOs that should be investigated at some point
 2025-10-08 08:24:06 +01:00
Wilson Snyder f09c30df35
Internals: Refector to create single V3Number::emitC. Fix some double emit issues (#6528). 2025-10-04 07:23:13 -04:00
Wilson Snyder caca1fcef0 Internals: Change order of V3Number constructor args to follow V3Const. No functional change intended. 2025-09-30 22:04:24 -04:00
Wilson Snyder 703f0d8c5d Commentary: spelling 2025-08-25 18:47:08 -04:00
Geza Lore 327d55d13d
Internals: Fix remaining cppcheck errors (#6319) 
Fixed the non const-related issue and added suppressions for the const
ones. With that `make cppcheck` should be clean.
 2025-08-21 09:43:37 +01:00
Rodrigo Batista de Moraes d0f0919830
Fix write of 0 in '%c' (#6248) (#6249) 2025-08-01 07:54:18 +02:00
Wilson Snyder 460bfbf181 Fix replicate of negative giving 'REPLICATE has no expected width' internal error (#6048). 2025-07-23 18:07:50 -04:00
Geza Lore 7401a8a43a
Refactor DFG IndependentBits analysis to work for any vertex. (#6209) 
This used to be restricted to variable vertices, but now can handle
arbitrary circular vertices that represent packed values. It also
converges faster than the earlier version. Prep for resolving loops
through arrays.
 2025-07-21 19:57:02 +01:00
Wilson Snyder 2f199f20cf Add ENUMITEMWIDTH error, and apply to X-extended and ranged values. 2025-07-12 14:14:17 -04:00
Peter Birch 75df36bc8f
Add hint of the signed rhsval in oversized replication error (#6098) 2025-06-17 16:59:18 -07:00
Wilson Snyder b553502d8a Fix number dumps to show 'null' 2025-06-10 08:17:45 -04:00
Wilson Snyder 46c7b69c64 Internals: UINFO now includes newline itself. No functional change. 2025-05-22 20:29:32 -04:00
Wilson Snyder 66667b6172
Support SARIF JSON diagnostic output with `--diagnostics-sarif`. (#6017) 2025-05-17 15:46:15 -04:00
Zhen Yan fc700538a5
Fix arithmetic left-shift by constants over 32 bits (#6007) (#6015) 2025-05-16 21:57:52 -04:00
Wilson Snyder 0f528d136d Fix arithmetic right-shift by constants over 32 bits (#5994). 2025-05-10 13:59:56 -04:00
Wilson Snyder 7191a0ba8b Internals: Favor preincrements. No functional change. 2025-05-09 21:20:35 -04:00
Wilson Snyder 27ad648c16 Commentary: Indicate V3Number width() criticality. 2025-05-04 16:24:36 -04:00
Wilson Snyder 0664cf407c Fix constant propagation making upper bits Xs (#5969). 2025-04-30 08:22:05 -04:00
Wilson Snyder 3658e5f0f1 Internals: Rename widthToFit, no functional change. 2025-04-28 21:54:58 -04:00
Wilson Snyder 8fbb725f34 Copyright year update. 2025-01-01 08:30:25 -05:00
Geza Lore f458951b17
Fix slow unsized number parsing (#5577) 
Try to avoid allocating and deallocating a full --max-num-width
buffer on parsing every single unsized number literal.
 2024-11-01 14:10:44 +00:00
Geza Lore 8f1ea09558
Add lsb index to V3Number::setMask (#5522) 2024-10-08 13:15:36 +01:00
Wilson Snyder e7de2c5a05 Remove warning on unsized numbers exceeding 32-bits. 2024-09-21 19:19:12 -04:00
Wilson Snyder 0ec5e02ce1 Improve hex too many digits error (#5419). 2024-08-30 07:48:00 -04:00
Wilson Snyder f0cd6dd95c Fix REALCVT warning on integral timescale conversions (#5378). 2024-08-24 08:01:28 -04:00
Wilson Snyder f9511e9755 Internals: Make V3Number::selfTest() 2024-08-24 05:41:22 -04:00
Wilson Snyder 3a659d460d Fix ==? and !=? with X values. 2024-07-28 14:40:55 -04:00
Wilson Snyder 8707c88787 Tests: Close misc internal code coverage holes 2024-07-28 14:18:24 -04:00
Wilson Snyder 50a5a1ff5b Tests: Fix some verilog.y parser coverage holes 2024-07-26 07:29:15 -04:00
Wilson Snyder 3f886f7c61 Fix select value too wide (#5148) (#5153). 
Co-authored-by: Dercury <dercury@qq.com>
 2024-06-07 08:27:03 -04:00
Wilson Snyder ea8f86dd30 Internals: Prefer '(void)' to avoid unused var. No functional change. 2024-03-27 18:07:14 -04:00
Wilson Snyder 0ff77fc352 Internals: Remove extra single-line {}. No functional change. 2024-03-27 17:57:49 -04:00
Kefa Chen 5f1dc73a1b
Support public packed struct / union (#860) (#4878) 2024-03-03 10:23:04 -05:00
Wilson Snyder 3786f59e03 Change to IEEE 1800-2023 warning mentions 2024-03-02 10:15:19 -05:00
Wilson Snyder 22687a6901 Internals: Use C++14 quoted literal std::string 2024-01-28 21:00:20 -05:00
Wilson Snyder a36a07c297 Internals: Favor UASSERT over v3fatalSrc. No functional change intended. 2024-01-05 18:00:06 -05:00