pre-commit: PR184182 by zyw-bot · Pull Request #3528 · dtcxzyw/llvm-opt-benchmark

zyw-bot · 2026-03-02T17:59:48Z

Link: llvm/llvm-project#184182
Requested by: @andjo403

zyw-bot · 2026-03-02T18:19:16Z

Diff mode

runner: ariselab-64c-docker
baseline: llvm/llvm-project@b11a424
patch: llvm/llvm-project#184182
sha256: 8f3173abd0f5a57b9058daac5f447ab6aacf02def745ba2b3f22582eacfcb5ee
commit: 4dc2780

3364 files changed, 2123743 insertions(+), 2096266 deletions(-)

Improvements:
  instsimplify.NumExpand 291222 -> 298851 +2.62%
  correlated-value-propagation.NumNNeg 95896 -> 97485 +1.66%
  correlated-value-propagation.NumShlNSW 116433 -> 118126 +1.45%
  bdce.NumRemoved 378049 -> 382263 +1.11%
  correlated-value-propagation.NumShlNW 268712 -> 270411 +0.63%
  gvn.NumGVNLoad 1318293 -> 1325766 +0.57%
  instsimplify.NumReassoc 815152 -> 819522 +0.54%
  instcount.NumOrInst 1142633 -> 1148327 +0.50%
  instcount.NumTruncInst 2935508 -> 2949446 +0.47%
  instcombine.NumExpand 2699 -> 2708 +0.33%
Regressions:
  aggressive-instcombine.NumAnyOrAllBitsSet 70 -> 20 -71.43%
  aggressive-instcombine.NumInstrsReduced 45455 -> 43478 -4.35%
  tailcallelim.NumAccumAdded 151 -> 149 -1.32%
  aggressive-instcombine.NumExprsReduced 15651 -> 15553 -0.63%
  correlated-value-propagation.NumAnd 44210 -> 43967 -0.55%
  indvars.NumElimIdentity 1833 -> 1823 -0.55%
  correlated-value-propagation.NumCmps 109384 -> 108793 -0.54%
  correlated-value-propagation.NumSaturating 2573 -> 2562 -0.43%
  correlated-value-propagation.NumOverflows 4343 -> 4330 -0.30%
  constraint-elimination.NumCondsRemoved 2633915 -> 2629322 -0.17%

+26 grpc/ssl_transport_security.ll
+24 ocio/MatrixOp.ll
+16 rocksdb/clock_cache.ll
+12 sdl/SDL_gamepad.ll
+8 nori/widget.ll
+8 rocksdb/lru_cache.ll
+6 z3/dl_mk_interp_tail_simplifier.ll
+5 ffmpeg/vc1_block.ll
+5 grpc/tls_security_connector.ll
+5 hermes/DCE.ll
+5 hyperscan/ng_literal_component.ll
+5 nuttx/sched_removereadytorun.ll
+5 proxygen/HTTP1xCodec.ll
+4 hermes/BigIntSupport.ll
+4 openssl/cmp_genm.ll
+4 pola-rs/4k9mmppgjee0j2w63yqma6wuu.ll
+4 rust-analyzer-rs/2334ao9w0k9d7973.ll
+4 rust-analyzer-rs/rilullg9p294yp1.ll
+3 openjdk/cfgnode.ll
+3 postgres/ruleutils.ll
+3 rustfmt-rs/llbxf4pclolbp5s.ll
+2 cmake/zstd_compress_superblock.ll
+2 duckdb/serialize_parquet.ll
+2 duckdb/ub_duckdb_common.ll
+2 llvm/Scalarizer.ll
+2 pbrt-v4/integrator.ll
+2 slurm/hostlist.ll
+2 uv-rs/13uh81w4oy46nmpmfbcefagqn.ll
+2 verilator/V3Expand.ll
+2 wireshark/disabled_protos.ll
+1 coreutils-rs/2i3dvgzkmy2gn6v1.ll
+1 cpython/complexobject.ll
+1 cpython/pylifecycle.ll
+1 cpython/textio.ll
+1 delta-rs/11w0at10aiwuq3yr.ll
+1 foundations-rs/208outlkq19x3hykb3fof84u4.ll
+1 jemalloc/extent_dss.ll
+1 linux/offchannel.ll
+1 llvm/CodeMetrics.ll
+1 luau/Quantify.ll
+1 mitsuba3/path.ll
+1 opencv/matmul.dispatch.ll
+1 php/session.ll
+1 pola-rs/akny94jrhz4eylr1elklgkf62.ll
+1 protobuf/tokenizer.ll
+1 qemu/job.ll
+1 redis/extent_dss.ll
+1 rust-analyzer-rs/54fx8hdjlkc0fd42.ll
+1 tokenizers-rs/4hn9gefsll13qr1r.ll
+1 tree-sitter-rs/3akexam875pc2p1h.ll
+1 wasmtime-rs/35xpok2vrm65hidj.ll
+1 wireshark/packet-fcswils.ll
+0 abseil-cpp/gtest-all.ll
+0 brotli/decode.ll
+0 bullet3/btConvexHull.ll
-1 abc/bmcCexTools.ll
-1 cmake/archive_read_support_format_mtree.ll
-1 coreutils-rs/2pqvixtdp9wizsl2.ll
-1 cpython/preconfig.ll
-1 cvc5/SimpSolver.ll
-1 darktable/thumbtable.ll
-1 delta-rs/47qjbhol909h8zu7.ll
-1 ffmpeg/mss12.ll
-1 ffmpeg/remove_extradata.ll
-1 folly/AsyncSocket.ll
-1 git/config.ll
-1 git/remote-curl.ll
-1 grpc/alts_handshaker_client.ll
-1 just-rs/2sblcsgax6v4zfcc.ll
-1 libphonenumber/phonemetadata.pb.ll
-1 linux/buffer.ll
-1 linux/hda_auto_parser.ll
-1 luajit/lj_carith.ll
-1 minetest/CSkinnedMesh.ll
-1 minetest/nodedef.ll
-1 mitsuba3/scene.ll
-1 openspiel/go_board.ll
-1 openusd/changes.ll
-1 postgres/brin_minmax_multi.ll
-1 quiche-rs/5lgkl05cn1nxa92c04lryr5y4.ll
-1 quickjs/libbf.ll
-1 recastnavigation/catch_amalgamated.ll
-1 ruby/rational.ll
-1 ruby/re.ll
-1 rust-analyzer-rs/leba1wmgxgrzxkl.ll
-1 rustfmt-rs/2vbyym84o66crvo9.ll
-1 stb/stb_vorbis.ll
-1 wireshark/packet-bacapp.ll
-2 abc/abcBalance.ll
-2 boost/algorithm.ll
-2 coreutils-rs/4gs2z359bfnc1tys.ll
-2 freetype/cff.ll
-2 freetype/ftbase.ll
-2 image-rs/1clnprdgqfw2q9lq.ll
-2 libwebp/vp8l_dec.ll
-2 lief/ASN1Reader.ll
-2 oiio/argparse.ll
-2 openssl/threadstest.ll
-2 quest/QuEST_validation.ll
-2 verilator/V3Split.ll
-2 wasmtime-rs/4bsmuvpz9r22ks1w.ll
-3 abc/sbdCut2.ll
-3 hermes/RegAlloc.ll
-3 icu/xmlparser.ll
-3 openjdk/modRefBarrierSetC1.ll
-3 proj/geodsigntest.ll
-3 regex-rs/32jw1oy2yofrhudk.ll
-3 vcpkg/commands.set-installed.ll
-3 yosys/SimpSolver.ll
-4 abc/giaSupp.ll
-4 arrow/api_vector.ll
-4 cvc5/theory_arith_private.ll
-4 ffmpeg/format.ll
-4 git/notes-utils.ll
-4 linux/main.ll
-4 linux/trace_dynevent.ll
-4 linux/waitid.ll
-4 llvm/EvaluationResult.ll
-4 openusd/loopPatchBuilder.ll
-4 php/array.ll
-4 sundials/sunmatrix_sparse.ll
-5 graphviz/strmatch.ll
-5 libwebp/webpdec.ll
-5 llvm/DWARFDie.ll
-6 g2o/hyper_graph.ll
-8 ockam-rs/on09w5afel7x9qz.ll
-8 rust-analyzer-rs/2mbx5ptcpq6fo7sc.ll
-9 icu/udata.ll
-12 luau/Unifier2.ll
-20 ruby/range.ll
-29 openssl/x_pubkey.ll

github-actions · 2026-03-02T18:20:04Z

This patch contains numerous low-level LLVM IR optimizations across many benchmarks, primarily focused on improving boolean logic and control flow. The major changes are:

Boolean Logic Simplification: Replaced icmp ne/icmp eq comparisons followed by zext with direct trunc operations (e.g., %x = trunc nuw i32 %y to i1) where the integer is known to be 0 or 1. This eliminates unnecessary comparison instructions and reduces instruction count.
Conditional Logic Refactoring: Converted chains of icmp + select + and/or into more efficient patterns — often using select directly with boolean operands instead of integer-based branching logic, and simplifying boolean algebra (e.g., replacing and i1 %a, %b with select i1 %b, i1 %a, i1 false when appropriate).
Phi Node Type Corrections: Updated phi node types from i8/i32 to i1 in loops and conditionals where the values being merged are logically boolean (e.g., phi i1 [true, ...] instead of phi i8 [1, ...]), improving type safety and enabling downstream optimizations.
Control Flow Merging & Edge Correction: Fixed misaligned branch targets and critical edge handling (e.g., updating .crit_edge, .thread, and loop exit labels), corrected phi operand order in merged blocks, and removed redundant branches — especially around early-exit conditions and error-handling paths.
Cleanup of Redundant Instructions: Removed unused or dead instructions (e.g., unused phi inputs, redundant zext/icmp, stale br targets) and consolidated landing pads and exception handling logic to reduce IR size and improve correctness.

These changes collectively enhance code generation quality by tightening boolean semantics, reducing instruction overhead, and improving CFG structure — all without altering program behavior.

model: qwen-plus-latest
CompletionUsage(completion_tokens=409, prompt_tokens=107715, total_tokens=108124, completion_tokens_details=None, prompt_tokens_details=None)

andjo403 · 2026-03-02T18:32:41Z

bench/arrow/optimized/api_vector.ll

  %.val3.i5.i.i.i.i = load i32, ptr %12, align 4, !tbaa !696
  %13 = icmp eq i32 %.val.i4.i.i.i.i, %.val3.i5.i.i.i.i
-  %14 = and i1 %8, %13
+  %14 = select i1 %13, i1 %8, i1 false


unclear why https://github.com/llvm/llvm-project/blob/ae363d50ad292b4d8c33bf795ad16c79eb640b19/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp#L3635-L3638 is not triggered here

andjo403

from local run 860 files with less lines, 1668 same line count and 840 files that is more lines many due to more inlining

andjo403 · 2026-03-02T18:37:36Z

bench/brotli/optimized/decode.ll

 47:                                               ; preds = %38
-  %48 = icmp ne i32 %4, 0
-  %or.cond = or i1 %48, %46
+  %48 = trunc nuw i32 %4 to i1


shall isGuaranteedNotToBePoison check the range for %4 so the select can be folded to or?

andjo403 · 2026-03-02T18:50:11Z

bench/delta-rs/optimized/11w0at10aiwuq3yr.ll

+  %15 = trunc nuw i8 %14 to i1
+  %16 = xor i1 %15, true


shall xor (trunc nuw x to i1), true be folded to icmp eq x, 0 ?
will lose information about the range of %14

andjo403 · 2026-03-02T18:54:51Z

bench/duckdb/optimized/serialize_parquet.ll

+  %.not = xor i1 %68, true
+  %69 = trunc nuw i8 %.pre22 to i1
+  %or.cond = select i1 %.not, i1 %69, i1 false


missing demorgan when trunc is one of values

andjo403 · 2026-03-02T19:00:38Z

bench/foundations-rs/optimized/208outlkq19x3hykb3fof84u4.ll

-  %5 = icmp eq i8 %4, 0
-  br i1 %5, label %.lr.ph, label %"_ZN4itoa55_$LT$impl$u20$itoa..private..Sealed$u20$for$u20$u32$GT$5write17hb7b886c46e130987E.exit"
+  %5 = tail call i8 @llvm.usub.sat.i8(i8 1, i8 %4)
+  %6 = trunc nuw i8 %5 to i1


trunc usub.sat(1,X) to i1 -> icmp eq X, 0

llvm/llvm-project#185524

andjo403 · 2026-03-02T19:14:57Z

bench/llvm/optimized/CodeMetrics.ll

+  %276 = trunc nuw i8 %275 to i1
+  %277 = or i1 %273, %276
+  %278 = zext i1 %277 to i8
+  store i8 %278, ptr %274, align 2, !tbaa !129


zext (or X, (trunc nuw Y)) -> or (zext X), Y
maybe is llvm/llvm-project#98017
many times in this diff shall this have a special fold

andjo403 · 2026-03-02T19:23:27Z

bench/llvm/optimized/Scalarizer.ll

+  %or.cond142.i.i = trunc nuw i8 %.sink.i.i.i319 to i1
+  %or.cond143.i.i = select i1 %or.cond142.i.i, i1 %.sroa.14.0.i138.i.i, i1 false


llvm/llvm-project#186621

andjo403 · 2026-03-02T19:26:38Z

bench/nuttx/optimized/sched_removereadytorun.ll

+  %.1 = phi i8 [ %46, %42 ], [ %.034, %36 ]
+  %48 = trunc nuw i8 %.1 to i1


missing that the type can be change to i1 for the phi

andjo403 · 2026-03-02T19:35:11Z

bench/rust-analyzer-rs/optimized/2334ao9w0k9d7973.ll

+  %4 = icmp eq i64 %3, 2
+  %5 = trunc nuw i64 %3 to i1
+  %spec.select = select i1 %4, i1 true, i1 %5


as the range for %3 is (0,3]

X == 2 && X != 0 -> X != 0

as we know that the range for %x is not (0,2] shall trunc nuw %x to i1 be converted to icmp ne %x , 0 or maybe at least if it is a second operand in a logical and/or

pre-commit: PR184182

cb1bfe1

github-actions bot mentioned this pull request Mar 2, 2026

Task submission #1312

Open

github-actions bot added 2 commits March 2, 2026 18:18

pre-commit: Update

b4060ae

pre-commit: Remap

4dc2780

andjo403 reviewed Mar 2, 2026

View reviewed changes

		%or.cond142.i.i = trunc nuw i8 %.sink.i.i.i319 to i1
		%or.cond143.i.i = select i1 %or.cond142.i.i, i1 %.sroa.14.0.i138.i.i, i1 false

		%.1 = phi i8 [ %46, %42 ], [ %.034, %36 ]
		%48 = trunc nuw i8 %.1 to i1

Conversation

zyw-bot commented Mar 2, 2026

Uh oh!

zyw-bot commented Mar 2, 2026

Diff mode

Uh oh!

github-actions bot commented Mar 2, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andjo403 left a comment

Choose a reason for hiding this comment

Uh oh!

andjo403 Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andjo403 Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

andjo403 Mar 2, 2026 •

edited

Loading

andjo403 Mar 2, 2026 •

edited

Loading