search

dtcxzyw / llvm-opt-benchmark

An LLVM IR dataset for data-driven compiler optimization research
https://dtcxzyw.github.io/llvm-opt-benchmark/
MIT License
19 stars 5 forks source link

Update diff September 20th 2024, 1:30:45 pm #1354

Open github-actions[bot] opened 1 month ago

github-actions[bot] commented 1 month ago

from: https://github.com/llvm/llvm-project/commit/766ec7d50f207b7e17f23b16bdf4fca1cdc8dd23 to: https://github.com/llvm/llvm-project/commit/2f50b280dc8e995ef67ad31a5f71adc4c270890d commit: 3609cb84d341697aca3f9c5be801b884f03c7052

Change Logs

from 766ec7d50f207b7e17f23b16bdf4fca1cdc8dd23 to 2f50b280dc8e995ef67ad31a5f71adc4c270890d

2f50b280dc8e995ef67ad31a5f71adc4c270890d [DebugInfo] Enable deprecation of iterator-insertion methods (#102608) 1808fc13c83c0bffb7e61adf1bb26a3097bc3c81 [AArch64][InstCombine] Bail from combining SRAD on +/-1 divisor (#109274) 73b8074e68e4b01bd7cb0dd2372af5adc2e89231 [AMDGPU] Do not use APInt for simple 64-bit arithmetic. NFC. (#109414) 3bcffe5ccc694d7c40f4bc84bd9ced1cc0e30d5c [X86] Fix MSVC implicit shift extension warning. cee0bf962648cf69a07e58dd8b02977d2a7e6007 [AMDGPU] Use Lo_32 and Hi_32 helpers (NFC) (#109413) 3127b659fad358b135721bd937fead49e5c73de5 [CVP] Infer range return attribute (#99620) efdb3ae23247850d3886e3708400f0d991ed59e1 Revert "[SystemZ][z/OS] Propagate IsText parameter to open text files as text (#107906)" d6f91200fe95dbf30a091af3ba039a817e7ddf14 Revert "[SystemZ][z/OS] Propagate IsText flag continuation"

Top 5 improvements:
  faiss/IndexIVFPQFastScan.cpp.ll 446219327 433346624 -2.88%
  lightgbm/dataset_loader.cpp.ll 7628411136 7446555019 -2.38%
  gromacs/gmx_hbond.cpp.ll 6233349886 6088838548 -2.32%
  lightgbm/gradient_discretizer.cpp.ll 1343095129 1313128007 -2.23%
  meshlab/filter_icp.cpp.ll 30377141944 29811853639 -1.86%
Top 5 regressions:
  lightgbm/dataset.cpp.ll 10823004657 11176617621 +3.27%
  lightgbm/gbdt_model_text.cpp.ll 3382692541 3492902227 +3.26%
  faiss/IndexHNSW.cpp.ll 2627170446 2691671450 +2.46%
  cpython/_codecs_hk.ll 517616730 529461503 +2.29%
  gromacs/manage_threading.cpp.ll 565316803 577782785 +2.21%

Overall: 0.08503901%
Improvements:
correlated-value-propagation.NumAddNUW 144501 -> 144943 +0.31%
correlated-value-propagation.NumAddNW 288190 -> 288672 +0.17%
correlated-value-propagation.NumNUW 317203 -> 317649 +0.14%
correlated-value-propagation.NumNW 664995 -> 665484 +0.07%
correlated-value-propagation.NumSExt 35692 -> 35704 +0.03%
correlated-value-propagation.NumAddNSW 143689 -> 143729 +0.03%
correlated-value-propagation.NumSICmps 43266 -> 43277 +0.03%
abstract-call-sites.NumInvalidAbstractCallSitesNoCallback 91753 -> 91774 +0.02%
abstract-call-sites.NumInvalidAbstractCallSitesUnknownUse 586236 -> 586369 +0.02%
correlated-value-propagation.NumNNeg 70410 -> 70423 +0.02%
Regressions:
function-attrs.NumNoUndefReturn 66275 -> 65887 -0.59%
bdce.NumSimplified 6022 -> 6020 -0.03%
ipt.NumInstScanned 279115516 -> 279110215 -0.00%
instcombine.NumReassoc 231033 -> 231030 -0.00%
bdce.NumRemoved 271135 -> 271133 -0.00%
simplifycfg.NumSpeculations 274608 -> 274606 -0.00%
assume-queries.NumAssumeQueries 12567889 -> 12567859 -0.00%
instcombine.NumPHICSEs 1283211 -> 1283210 -0.00%
capture-tracking.NumCapturedBefore 8427876 -> 8427870 -0.00%
instsimplify.NumSimplified 1799354 -> 1799353 -0.00%
dtcxzyw commented 1 month ago

See https://github.com/dtcxzyw/llvm-opt-benchmark/pull/1342#discussion_r1767033125