Commits · eeec35ae1cce76d8f0d4fc35d231deebf2fa5f38 · Roger Ferrer / llvm-epi · GitLab

Oct 20, 2014
- Teach the load analysis driving core instcombine logic and other bits of · eeec35ae
  Chandler Carruth authored Oct 20, 2014
  eeec35ae
- Add a datalayout string to this test so that it exercises the full gamut · b5f4c328
  Chandler Carruth authored Oct 20, 2014
  b5f4c328
Oct 19, 2014
- Do a better and more complete job of preserving metadata when combining · bc6378de
  Chandler Carruth authored Oct 19, 2014
  bc6378de
- Move previously dead code to handle computing the known bits of an alias · 5b8cd2f7
  Chandler Carruth authored Oct 19, 2014
  5b8cd2f7
- InstCombine: (sub (or A B) (xor A B)) --> (and A B) · 312c3e5f
  David Majnemer authored Oct 19, 2014
  312c3e5f
- InstCombine: Optimize icmp eq/ne (shl Const2, A), Const1 · 59939acd
  David Majnemer authored Oct 19, 2014
  59939acd
- Fix a long-standing miscompile in the load analysis that was uncovered · a801dd57
  Chandler Carruth authored Oct 19, 2014
  a801dd57
Oct 18, 2014
- Preserve AA metadata when combining (cast (load (...))) -> (load (cast · be9dccd6
  Chandler Carruth authored Oct 18, 2014
  be9dccd6
- [InstCombine] Do an about-face on how LLVM canonicalizes (cast (load · 2f75fcfe
  Chandler Carruth authored Oct 18, 2014
  2f75fcfe
- Remove a test that was ported from the old llvm-gcc frontend test suite. · 71009cad
  Chandler Carruth authored Oct 18, 2014
  71009cad
- [SROA] Change how SROA does vector-based promotion of allocas to handle · 2dc9682e
  Chandler Carruth authored Oct 18, 2014
  2dc9682e
Oct 17, 2014
- Revert "TRE: make TRE a bit more aggressive" · 7da1ea83
  Rafael Espindola authored Oct 17, 2014
  7da1ea83
- [DSE] Remove no-data-layout-only type-based overlap checking · dd38c0b8
  Hal Finkel authored Oct 17, 2014
  dd38c0b8
Oct 16, 2014
- Delete -std-compile-opts. · 11aaaeeb
  Rafael Espindola authored Oct 16, 2014
  11aaaeeb
- Allow call-slop optzn for destinations with a suitable dereferenceable attribute · d20816fd
  Bjorn Steinbrink authored Oct 16, 2014
  d20816fd
- fold: sqrt(x * x * y) -> fabs(x) * sqrt(y) · c699a611
  Sanjay Patel authored Oct 16, 2014
  c699a611
- Reapply r219832 - InstCombine: Narrow switch instructions using known bits. · 5c221ef9
  Akira Hatanaka authored Oct 16, 2014
  5c221ef9
- TRE: make TRE a bit more aggressive · 7f529219
  Saleem Abdulrasool authored Oct 16, 2014
  7f529219
- Revert r219832. · 40c2cf4a
  Akira Hatanaka authored Oct 16, 2014
  40c2cf4a
- Revert "r219834 - Teach ScalarEvolution to sharpen range information" · 360b1ed5
  Sanjoy Das authored Oct 15, 2014
  360b1ed5
- Preserve non-byval pointer alignment attributes using @llvm.assume when inlining · 68dc3c7a
  Hal Finkel authored Oct 15, 2014
  68dc3c7a
Oct 15, 2014
- Teach ScalarEvolution to sharpen range information. · 90c2f145
  Sanjoy Das authored Oct 15, 2014
  90c2f145
- InstCombine: Narrow switch instructions using known bits. · 5bb9346a
  Akira Hatanaka authored Oct 15, 2014
  5bb9346a
- [SLPVectorize] Basic ephemeral-value awareness · 3b7fc866
  Hal Finkel authored Oct 15, 2014
  3b7fc866
- [LoopVectorize] Ignore @llvm.assume for cost estimates and legality · 1a600fab
  Hal Finkel authored Oct 14, 2014
  1a600fab
Oct 14, 2014
- Optimize away fabs() calls when input is squared (known positive). · 0ca42bb5
  Sanjay Patel authored Oct 14, 2014
  0ca42bb5
- InstCombine: Don't miscompile X % ((Pow2 << A) >>u B) · dad21038
  David Majnemer authored Oct 14, 2014
  dad21038
- Revert "r216914 - Revert: [APFloat] Fixed a bug in method 'fusedMultiplyAdd'" · 171c2ec0
  Hal Finkel authored Oct 14, 2014
  171c2ec0
- [LVI] Check for @llvm.assume dominating the edge branch · a3f23e37
  Hal Finkel authored Oct 14, 2014
  a3f23e37
- Switch to select optimization for two-case switches · 5bbe3df6
  Marcello Maggioni authored Oct 14, 2014
  5bbe3df6
- InstCombine: Fix miscompile in X % -Y -> X % Y transform · db077308
  David Majnemer authored Oct 13, 2014
  db077308
Oct 13, 2014
- InstCombine: Don't miscompile (x lshr C1) udiv C2 · a2521389
  David Majnemer authored Oct 13, 2014
  a2521389
Oct 12, 2014
- Revert r219223, it creates invalid PHI nodes. · 5ca10d0e
  Joerg Sonnenberger authored Oct 12, 2014
  5ca10d0e
- InstCombine: Turn (x != 0 & x <u C) into the canonical range check form (x-1 <u C-1) · 240b85ee
  Benjamin Kramer authored Oct 12, 2014
  240b85ee
Oct 11, 2014
- InstCombine: Don't fold (X <<s log(INT_MIN)) /s INT_MIN to X · fe7fccff
  David Majnemer authored Oct 11, 2014
  fe7fccff
- InstCombine, InstSimplify: (%X /s C1) /s C2 isn't always 0 when C1 * C2 overflow · cb9d5966
  David Majnemer authored Oct 11, 2014
  cb9d5966
- InstCombine: mul to shl shouldn't preserve nsw · 3cac85e0
  David Majnemer authored Oct 11, 2014
  3cac85e0
- Return undef on FP <-> Int conversions that overflow (PR21330). · ad8b6666
  Sanjay Patel authored Oct 10, 2014
  ad8b6666
Oct 10, 2014
- This patch teaches ScalarEvolution to pick and use !range metadata. · 1f05c51e
  Sanjoy Das authored Oct 10, 2014
  1f05c51e
- This patch de-pessimizes the calculation of loop trip counts in · 2beab5f0
  Mark Heffernan authored Oct 10, 2014
  2beab5f0