pre-commit: PR167771

[zyw-bot]

Link: llvm/llvm-project#167771
Requested by: @dtcxzyw

[zyw-bot]

Diff mode

runner: ariselab-64c-docker
baseline: llvm/llvm-project@e9bda49
patch: llvm/llvm-project#167771
sha256: 2cf77b54801fba4baf8ead60b96bc1500c69e6c5ddec59c25430f42e31f0f9d8
commit: 074cab7

938 files changed, 673194 insertions(+), 686156 deletions(-)

Improvements:
  bdce.NumSimplified 5795 -> 5816 +0.36%
  memcpyopt.NumCpyToSet 10940 -> 10953 +0.12%
  simplifycfg.NumFoldValueComparisonIntoPredecessors 512831 -> 513089 +0.05%
  simplifycfg.NumFoldBranchToCommonDest 748129 -> 748468 +0.05%
  simplifycfg.NumBitMaps 2245 -> 2246 +0.04%
  simplifycfg.NumLookupTablesHoles 2393 -> 2394 +0.04%
  mem2reg.NumDeadAlloca 1963403 -> 1964174 +0.04%
  instcombine.NumFactor 45557 -> 45570 +0.03%
  correlated-value-propagation.NumNNeg 96003 -> 96027 +0.02%
  sroa.NumVectorized 697857 -> 697970 +0.02%
Regressions:
  instcombine.NumGlobalCopies 184204 -> 183372 -0.45%
  bdce.NumRemoved 384284 -> 383487 -0.21%
  globalsmodref-aa.NumNonAddrTakenGlobalVars 416312 -> 415531 -0.19%
  vector-combine.NumVecLoad 4291 -> 4288 -0.07%
  correlated-value-propagation.NumNonNull 10870047 -> 10862737 -0.07%
  gvn.NumGVNLoad 1320135 -> 1319335 -0.06%
  sroa.MaxPartitionsPerAlloca 206825 -> 206703 -0.06%
  memcpyopt.NumStackMove 80837 -> 80790 -0.06%
  instcombine.NumConstProp 155480 -> 155393 -0.06%
  aggressive-instcombine.NumInstrsReduced 44996 -> 44976 -0.04%

+12 just-rs/xt6bvrk4jdo7mpy.ll
+10 wasmtime-rs/24tijvi11z3k9odl.ll
+10 wasmtime-rs/cxg2cx8gn6v5trt.ll
+8 ockam-rs/4gv6ak02caawu3g.ll
+8 rust-analyzer-rs/10elsj6wik4dx3zk.ll
+6 ockam-rs/48m1civl6b0v2j4c.ll
+6 ockam-rs/4i4les6ijtr4jgtl.ll
+4 ockam-rs/2tmak98bqpx7zrgd.ll
+4 ockam-rs/3spcfx0kj7tua4em.ll
+3 llvm/AttributorAttributes.ll
+2 ockam-rs/7hitgxreqt236i.ll
+2 open3d/Geometry3D.ll
+2 syn/3e28wvtnspuupgei.ll
+2 wasmtime-rs/1zz7jsxv168dc7km.ll
+2 wasmtime-rs/2wry4odhn7m84bj2.ll
+2 wasmtime-rs/3hwarn3zl2ncla2z.ll
+1 ffmpeg/avformat.ll
+1 tokenizers-rs/2mot01sr7ebui81b.ll
+0 actix-rs/2pgtlhl4qd060m9c.ll
+0 actix-rs/3tnlvci61w0nl5yv.ll
+0 assimp/COBLoader.ll
+0 bullet3/b3ConvexHullContact.ll
+0 bullet3/b3GpuGridBroadphase.ll
+0 bullet3/b3GpuPgsContactSolver.ll
+0 bullet3/btUnionFind.ll
+0 clamav/qtmd.ll
+0 clap-rs/48fdpr2dy8inq5cq.ll
+0 cmake/zstd_opt.ll
+0 coreutils-rs/2vd9rhgrl6pfdvv0.ll
+0 coreutils-rs/3wh0yla9idangd55.ll
+0 diesel-rs/3btvyztn2iiarpvf.ll
+0 ffmpeg/bsf.ll
+0 folly/ObserverManager.ll
+0 freetype/psaux.ll
+0 grpc/hpack_encoder.ll
+0 hdf5/sio_perf.ll
+0 image-rs/244uszkx0e8t5ie1.ll
+0 libcxx/time_zone.ll
+0 luau/CodeGenX64.ll
+0 nlohmann_json/diag_on.ll
+0 pbrt-v4/integrator.ll
+0 proj/helmert.ll
+0 proj/horner.ll
+0 protobuf/descriptor.ll
+0 quickjs/qjsc.ll
+0 wireshark/proto_tree.ll
-1 syn/3r5osr990qsnm7hf.ll
-2 delta-rs/31i6j0ayu6bevt9n.ll
-2 influxdb-rs/26y592k8de9dg2n1.ll
-2 just-rs/2sotx4bdu006z307.ll
-2 mini-lsm-rs/4vkix0mjdn2idd4b.ll
-2 ockam-rs/111eznecj9pn2lkp.ll
-2 regex-rs/3ixfkxlmcuecmmus.ll
-2 typst-rs/4sdr2z2pqpqxs72u.ll
-4 duckdb/fastcover.ll
-4 hermes/DebugInfo.ll
-4 influxdb-rs/1zobaqtttt01ywwe.ll
-4 open3d/Tensor.ll
-8 tokio-rs/4zaq7yghb5wawieu.ll
-12 clap-rs/1rbie63mhpvpjak.ll
-12 tokio-rs/2i86qkpybymk1snv.ll
-14 tokio-rs/46apu3hqx0w07xoc.ll
-24 diesel-rs/3nhirlswgfgsoryk.ll
-24 syn/2khi0xu1ufmhwoo.ll
-27 tokio-rs/1rl1r5ea6bzd5c9z.ll
-30 proj/param.ll

[dtcxzyw]

cc @YonahGoldberg

[github-actions]

The provided patch contains several changes to LLVM IR code across multiple benchmark files. The main changes can be summarized as follows:

1. Phi Node Updates: Several phi instructions have their predecessor basic blocks updated. For example, in actix-rs, assimp, and bullet3 benchmarks, the predecessor labels in phi nodes are changed (e.g., from %thread83.i.i.i.i to %thread84.i.i.i.i). This indicates a shift in control flow or block ordering.

2. Store Instruction Modifications: In the clap-rs and just-rs benchmarks, store operations that previously stored 8-bit values (i8) now store 32-bit integers (i32). For instance, store i8 3 becomes store i32 3. This suggests a change in data layout or type representation.

3. Function Call and Exception Handling Adjustments: Multiple files show updates to exception handling paths. Labels in invoke/unwind sequences are modified (e.g., .noexc33.i → .noexc35.i), indicating adjustments in landing pad or cleanup logic.

4. Struct Initialization Changes: In clap-rs, the initializes attribute on function parameters is updated to reflect broader initialization ranges (e.g., (0,1), (4,5) → (0,1), (4,15)), suggesting more fields are being initialized by the function.

5. Memory Initialization Optimizations: In ockam-rs and influxdb-rs, calls to llvm.memset.p0.i64 are expanded to cover larger memory regions, and redundant stores are replaced with a single memset, improving initialization efficiency.

These changes generally reflect optimizations in control flow, memory initialization, and data layout, likely resulting from improved static analysis or restructuring of high-level constructs.

model: qwen-plus-latest
CompletionUsage(completion_tokens=402, prompt_tokens=115461, total_tokens=115863, completion_tokens_details=None, prompt_tokens_details=None)