pre-commit: PR167771

[zyw-bot]

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

[zyw-bot]

Diff mode

runner: ariselab-64c-docker
baseline: llvm/llvm-project@863a4e4
patch: llvm/llvm-project#167771
sha256: 8125e24db6d247a746769a8c3348836ed89a1e887ce0fa36654ef34e0e064059
commit: 8baa98c

145 files changed, 93098 insertions(+), 93245 deletions(-)

Improvements:
  simplifycfg.NumBitMaps 2245 -> 2246 +0.04%
  simplifycfg.NumLookupTablesHoles 2394 -> 2395 +0.04%
  correlated-value-propagation.NumNonNull 10869930 -> 10874315 +0.04%
  mem2reg.NumDeadAlloca 1963420 -> 1964171 +0.04%
  sroa.NumVectorized 697857 -> 697970 +0.02%
  simple-loop-unswitch.NumCostMultiplierSkipped 17253 -> 17255 +0.01%
  correlated-value-propagation.NumNUW 571064 -> 571123 +0.01%
  correlated-value-propagation.NumAnd 44492 -> 44496 +0.01%
  instcombine.NumPHIsOfInsertValues 26429 -> 26431 +0.01%
  constmerge.NumIdenticalMerged 15653 -> 15654 +0.01%
Regressions:
  bdce.NumRemoved 384222 -> 383427 -0.21%
  vector-combine.NumVecLoad 4291 -> 4288 -0.07%
  gvn.NumGVNLoad 1312961 -> 1312122 -0.06%
  instcombine.NumConstProp 168345 -> 168244 -0.06%
  mem2reg.NumLocalPromoted 549701 -> 549536 -0.03%
  licm.NumBOAssociationsHoisted 3516 -> 3515 -0.03%
  sroa.NumLoadsPredicated 14459 -> 14456 -0.02%
  licm.NumLoadStorePromoted 64161 -> 64150 -0.02%
  local.NumPHICSEs 185818 -> 185787 -0.02%
  gvn.NumGVNSimpl 4418808 -> 4418075 -0.02%

+18 openvdb/Prune.ll
+2 open3d/Geometry3D.ll
+0 assimp/zip.ll
+0 ceres/block_sparse_matrix.ll
+0 cmake/cmComputeLinkDepends.ll
+0 faiss/Clustering.ll
+0 g2o/edge_project_xyz.ll
+0 g2o/optimizable_graph.ll
+0 g2o/optimization_algorithm_dogleg.ll
+0 g2o/slam2d_linear.ll
+0 hdf5/sio_perf.ll
+0 libigl/eigs.ll
+0 llvm/MsgPackDocument.ll
+0 meshlab/edit_align.ll
+0 meshlab/shell.ll
+0 nlohmann_json/diag_on.ll
+0 nori/accel.ll
+0 ockam-rs/1411u8drt798uxi8.ll
+0 opencv/gfluidbackend.ll
+0 pbrt-v4/integrator.ll
+0 pbrt-v4/integrators.ll
+0 php/zip.ll
+0 proj/adams.ll
+0 proj/airocean.ll
+0 proj/datum_set.ll
+0 proj/geod_set.ll
+0 proj/helmert.ll
+0 proj/horner.ll
+0 proj/info.ll
+0 proj/tmerc.ll
+0 quickjs/qjsc.ll
+0 quickjs/quickjs-libc.ll
+0 typst-rs/2d3c2n5y91mtl0x0.ll
+0 typst-rs/d6l9ieo9tcw33dn.ll
+0 wasmtime-rs/4ab4rlryc5h7bf6z.ll
-2 open3d/VoxelGrid.ll
-30 proj/param.ll
-31 typst-rs/2i78fvbm4wocuesi.ll

[github-actions]

This patch contains a series of targeted LLVM IR modifications across many benchmarks, primarily focused on improving correctness and consistency in control flow and dataflow. The major changes are:

1. Phi node predecessor updates: Multiple .phi instructions have their predecessor block lists updated to reflect accurate control flow edges—e.g., replacing %thread428 with %thread426, %72 with %73, or %59 with %58. These are critical for SSA form validity and ensure phi operands correspond to actual incoming edges.

2. Loop header and loop exit block corrections: Several loops have their loop headers (e.g., %.lr.ph75 → %.lr.ph76, %.lr.ph146 → %.lr.ph143) and loop exit blocks (e.g., %.loopexit513 → %.loopexit511) adjusted. These changes align with loop canonicalization and preserve correct loop-carried dependencies.

3. Store/load type and alignment refinements: In multiple places (e.g., ockam-rs, open3d, openvdb), <4 x i8> loads/stores are replaced with i32, and double bitcasts are reordered or eliminated. This improves type safety, enables better optimization, and fixes potential miscompilations from vector-to-scalar mismatches.

4. Vector arithmetic refactoring: In g2o, meshlab, and nori, scalar operations feeding vector computations (e.g., fadd, fmul) are now consistently built via insertelement + shufflevector instead of legacy bitcast + shufflevector. This improves readability, enables more aggressive vector optimizations, and avoids undefined behavior from improper bitcasts.

5. Exception handling and landing pad consistency: In cmake, typst-rs, and quickjs, landing pad phi nodes and unwind edges are updated to reference correct predecessor blocks (e.g., %75 → %74, %150 → %145). This ensures exception propagation remains semantically correct after transformations.

All changes are mechanical but essential for correctness: they fix stale CFG edges, update types to match actual memory layout, and modernize vector code patterns—all without altering program semantics.

model: qwen-plus-latest
CompletionUsage(completion_tokens=489, prompt_tokens=113217, total_tokens=113706, completion_tokens_details=None, prompt_tokens_details=None)