pre-commit: PRantoniofrighetto/llvm-project/compare/main...feature/ic-intrinsics-into-phis

[zyw-bot]

Link: antoniofrighetto/llvm-project@main...feature/ic-intrinsics-into-phis
Requested by: @antoniofrighetto

[zyw-bot]

Diff mode

runner: ariselab-64c-docker
baseline: llvm/llvm-project@46526f8
patch: antoniofrighetto/llvm-project@main...feature/ic-intrinsics-into-phis
sha256: 158a0434a8afe6ce3cbb16c61232e3b2b22dd1cd16d1ac46b669fb215fefd07c
commit: 798918c

1493 files changed, 860508 insertions(+), 859435 deletions(-)

Improvements:
  tailcallelim.NumAccumAdded 152 -> 156 +2.63%
  instcombine.NegatorNumNegationsFoundInCache 4467 -> 4472 +0.11%
  simple-loop-unswitch.NumSwitches 1851 -> 1852 +0.05%
  instcombine.NegatorMaxDepthVisited 20668 -> 20678 +0.05%
  tailcallelim.NumEliminated 9298 -> 9302 +0.04%
  tailcallelim.NumRetDuped 16581 -> 16585 +0.02%
  instcombine.NegatorMaxTotalValuesVisited 64691 -> 64706 +0.02%
  instcombine.NegatorNumTreesNegated 51893 -> 51903 +0.02%
  instcombine.NegatorNumInstructionsNegatedSuccess 63736 -> 63748 +0.02%
  correlated-value-propagation.NumNNeg 105077 -> 105096 +0.02%
Regressions:
  correlated-value-propagation.NumMinMax 12466 -> 12446 -0.16%
  move-auto-init.NumMoved 2634 -> 2632 -0.08%
  simplifycfg.NumLookupTablesHoles 2552 -> 2551 -0.04%
  loop-simplifycfg.NumLoopExitsDeleted 3905 -> 3904 -0.03%
  memdep.NumUncacheNonLocal 27302 -> 27296 -0.02%
  correlated-value-propagation.NumShlNUW 163239 -> 163204 -0.02%
  simplifycfg.NumSpeculations 412794 -> 412710 -0.02%
  licm.NumMovedCalls 34912 -> 34905 -0.02%
  simplifycfg.NumLookupTables 22897 -> 22893 -0.02%
  loop-instsimplify.NumSimplified 182221 -> 182190 -0.02%

7 4 bench/abc/optimized/abcUtil.ll
6 10 bench/abc/optimized/cbaNtk.ll
77 74 bench/abseil-cpp/optimized/crc.ll
5 9 bench/abseil-cpp/optimized/nonsecure_base_test.ll
28 24 bench/arrow/optimized/arena.ll
15 9 bench/assimp/optimized/Exporter.ll
2 3 bench/assimp/optimized/IFCOpenings.ll
11 8 bench/assimp/optimized/SpatialSort.ll
8 5 bench/boost/optimized/alloc_lib.ll
16 18 bench/boost/optimized/async.ll
7 3 bench/boost/optimized/lock_pool.ll
64 61 bench/box2d/optimized/shape.ll
9 6 bench/bullet3/optimized/btSoftBodyHelpers.ll
10 12 bench/c3c/optimized/float.ll
5 9 bench/clap-rs/optimized/3b4nqkxyl1xqdcre.ll
12 20 bench/clap-rs/optimized/ms9r2ilwmkwzn0d.ll
32 31 bench/cmake/optimized/cmCTestMultiProcessHandler.ll
23 21 bench/cmake/optimized/cmQtAutoGen.ll
8 4 bench/cmake/optimized/io.ll
61 73 bench/coreutils-rs/optimized/5e48q1wpjedhlmp9.ll
33 48 bench/coreutils-rs/optimized/kym4n2c7483sba2.ll
12 9 bench/cpython/optimized/dtoa.ll
4 8 bench/csmith/optimized/FunctionInvocation.ll
13 10 bench/darktable/optimized/TableLookUp.ll
10 7 bench/darktable/optimized/curves.ll
14 8 bench/darktable/optimized/introspection_atrous.ll
10 16 bench/diesel-rs/optimized/4l2dlnns3xjar0un.ll
21 37 bench/diesel-rs/optimized/6cb7zukciz78xvw.ll
4 8 bench/duckdb/optimized/pac.ll
10 18 bench/eastl/optimized/EADateTime.ll
5 6 bench/ffmpeg/optimized/tx_double.ll
5 7 bench/ffmpeg/optimized/tx_float.ll
3 2 bench/fish-rs/optimized/6luvh0o6z0xeo16dcenatpsob.ll
15 17 bench/folly/optimized/EventBase.ll
16 8 bench/freetype/optimized/ftbase.ll
22 19 bench/git/optimized/gc.ll
15 7 bench/git/optimized/trailer.ll
8 14 bench/git/optimized/xmerge.ll
2 6 bench/git/optimized/xutils.ll
44 32 bench/harfbuzz/optimized/hb-subset-cff1.ll
22 16 bench/harfbuzz/optimized/hb-subset-cff2.ll
4 5 bench/hdf5/optimized/H5Shyper.ll
45 36 bench/hdf5/optimized/h5diff_array.ll
10 11 bench/hwloc/optimized/lstopo-draw.ll
30 26 bench/hyperscan/optimized/ng_limex_accel.ll
13 14 bench/hyperscan/optimized/rose_build_infix.ll
53 49 bench/icu/optimized/decNumber.ll
15 17 bench/icu/optimized/unames.ll
2 4 bench/icu/optimized/utext.ll
9 17 bench/ipopt/optimized/SensAlgorithm.ll
7 12 bench/ipopt/optimized/SensMetadataMeasurement.ll
67 55 bench/jq/optimized/regcomp.ll
16 13 bench/lean4/optimized/static.ll
11 19 bench/libigl/optimized/assign_scalar.ll
10 12 bench/libigl/optimized/nchoosek.ll
44 37 bench/libpng/optimized/pngread.ll
32 34 bench/libquic/optimized/v3_lib.ll
3 7 bench/lief/optimized/bignum_core.ll
6 2 bench/linux/optimized/e820.ll
6 3 bench/linux/optimized/hid-core.ll
21 18 bench/llama.cpp/optimized/common.ll
3 8 bench/llvm/optimized/ARMTargetParserCommon.ll
10 8 bench/llvm/optimized/DWARFAcceleratorTable.ll
4 5 bench/llvm/optimized/InlineFunction.ll
30 28 bench/luajit/optimized/lj_profile.ll
17 28 bench/lvgl/optimized/lv_text.ll
6 7 bench/meshlab/optimized/filter_isoparametrization.ll
14 11 bench/meshlab/optimized/mainwindow_RunTime.ll
35 31 bench/mimalloc/optimized/arena.ll
9 7 bench/minetest/optimized/CGUITTFont.ll
4 5 bench/minetest/optimized/content_cso.ll
3 4 bench/minetest/optimized/l_env.ll
40 36 bench/mitsuba3/optimized/x86rapass.ll
6 10 bench/oiio/optimized/imagebufalgo_yee.ll
29 30 bench/oiio/optimized/imageinput.ll
51 42 bench/opencv/optimized/dwt.ll
34 35 bench/opencv/optimized/reduce_layer.ll
7 4 bench/opencv/optimized/upcean_decoder.ll
78 72 bench/openexr/optimized/internal_b44.ll
28 35 bench/openjdk/optimized/ad_x86.ll
8 16 bench/openspiel/optimized/oos.ll
14 16 bench/openssl/optimized/cipher_aes_gcm_siv_polyval.ll
26 46 bench/openssl/optimized/provider.ll
22 25 bench/openssl/optimized/quic_txpim.ll
8 12 bench/openssl/optimized/x509_cmp.ll
53 54 bench/openusd/optimized/decodemv.ll
11 15 bench/openusd/optimized/stageCache.ll
42 36 bench/openusd/optimized/testArResolverContext.ll
29 24 bench/php/optimized/browscap.ll
32 37 bench/pola-rs/optimized/1nb2gyhm0tinqn6ey8ai4djfd.ll
8 20 bench/pola-rs/optimized/2x1snfrtze7jig38mzpup27sc.ll
24 17 bench/pola-rs/optimized/6iuyrj4ccijpoxxv1fx09rxds.ll
4 5 bench/postgres/optimized/mbprint.ll
14 21 bench/postgres/optimized/vacuumparallel.ll
4 11 bench/protobuf/optimized/lexer.ll
24 12 bench/protobuf/optimized/ruby_generator.ll
7 11 bench/quantlib/optimized/abcd.ll
16 10 bench/quantlib/optimized/analyticdoublebarrierbinaryengine.ll
11 8 bench/quantlib/optimized/pseudosqrt.ll
17 21 bench/quiche-rs/optimized/evfqynx3m984ucpmhnvcicaft.ll
24 32 bench/quinn-rs/optimized/2o4hl0ivtqvy0gs3hj842e52l.ll
33 32 bench/raylib/optimized/raudio.ll
7 3 bench/redis/optimized/dict.ll
7 10 bench/redis/optimized/module.ll
28 25 bench/redis/optimized/server.ll
13 10 bench/rocksdb/optimized/db_impl_open.ll
25 26 bench/ruby/optimized/process.ll
13 11 bench/ruff-rs/optimized/9kdq8p9vbvp469l9m5kewm2r5.ll
25 26 bench/rust-analyzer-rs/optimized/4fo6e09jeifrbt47.ll
21 19 bench/sqlite/optimized/shell.ll
28 33 bench/sqlite/optimized/sqlite3.ll
13 21 bench/turborepo-rs/optimized/74s3x9u7hld928rwmttmetb1y.ll
16 8 bench/typst-rs/optimized/4khbogid70pr8yfn.ll
6 8 bench/velox/optimized/AsyncDataCache.ll
9 8 bench/velox/optimized/Counters.ll
21 14 bench/wasmtime-rs/optimized/1h2b9qnacbe9cbqw.ll
18 11 bench/wasmtime-rs/optimized/4u85yh8sn1llpfha.ll
11 15 bench/wireshark/optimized/packet-dcerpc-lsa.ll
15 19 bench/wireshark/optimized/packet-dcerpc-samr.ll
13 10 bench/wireshark/optimized/packet-rlc-lte.ll
5 11 bench/wireshark/optimized/wslog.ll
5 9 bench/z3/optimized/arith_solver.ll
11 14 bench/z3/optimized/pb2bv_rewriter.ll
9 13 bench/z3/optimized/used_vars.ll
14 18 bench/zed-rs/optimized/59wnsznecs6we2kopjyje48jo.ll
8 12 bench/zxing/optimized/HRI.ll

[github-actions]

The provided diff contains numerous changes across multiple LLVM IR files. After filtering out non-interesting changes (e.g., formatting, renaming, reordering), the major changes are:

1. Optimization of phi nodes and control flow in Abc_NtkGetCubePairNum: The function replaces a phi node in the .critedge block with a direct call to @llvm.umin.i64, simplifying the control flow and reducing the number of blocks by merging .critedge labels.

2. Simplification of conditional branches and phi nodes in Cba_FonRangeSize: The function streamlines the control flow by reducing the number of phi nodes and merging exit blocks, directly computing the result using @llvm.abs.i32 and add operations.

3. Refactoring of loop and branch logic in Cba_NtkRangeSize.exit: The function simplifies the phi node usage by combining multiple predecessors into a single phi node, reducing complexity and improving clarity.

4. Optimization of bit manipulation in Crc32::UnextendByZeroesImpl: The function replaces a sequence of bit manipulation operations with a direct call to @llvm.bitreverse.i32, enhancing performance and reducing instruction count.

5. Streamlining of control flow in float_from_string: The function optimizes the control flow by reducing the number of phi nodes and merging exit blocks, improving efficiency and readability.

These changes collectively aim to optimize performance by simplifying control flow, reducing the number of instructions, and leveraging LLVM intrinsics for common operations.

model: qwen-plus-latest
CompletionUsage(completion_tokens=329, prompt_tokens=112645, total_tokens=112974, completion_tokens_details=None, prompt_tokens_details=None)

[antoniofrighetto]

/add-label reviewed
/close