Commits · 64404a3b2c62db01c3a4459a2c15c37496ed9f65 · Roger Ferrer / llvm-epi · GitLab

Nov 29, 2011
- Fix VINSERTF128/VEXTRACTF128 to be marked as FP instructions. Allow execution... · 12b72def
  Craig Topper authored Nov 29, 2011
  12b72def
- Correctly mark VPERM2F128 as being an FP instruction and add execution domain... · 897a7d4b
  Craig Topper authored Nov 29, 2011
  897a7d4b
Nov 28, 2011
- Revert r145273 and fix in SelectionDAG::InferPtrAlignment() instead. · 4a5b2040
  Evan Cheng authored Nov 28, 2011
  4a5b2040
- DAG combine should not increase alignment of loads / stores with alignment less · a4b6404c
  Evan Cheng authored Nov 28, 2011
  a4b6404c
- Add X86 instruction selection for VPERM2I128 when AVX2 is enabled. Merge... · 818a983e
  Craig Topper authored Nov 28, 2011
  818a983e
Nov 27, 2011
- Take two on rotating the block ordering of loops. My previous attempt · 03adbd46
  Chandler Carruth authored Nov 27, 2011
  03adbd46
- Rework a bit of the implementation of loop block rotation to not rely so · a0545809
  Chandler Carruth authored Nov 27, 2011
  a0545809
- remove autoupgrade support for old forms of llvm.prefetch and the old · ee471c48
  Chris Lattner authored Nov 27, 2011
  ee471c48
- Upgrade syntax of tests using volatile instructions to use 'load volatile'... · 6a144a22
  Chris Lattner authored Nov 27, 2011
  6a144a22
- remove some old autoupgrade logic · 6aa6c0c3
  Chris Lattner authored Nov 27, 2011
  6aa6c0c3
- Introduce a loop block rotation optimization to the new block placement · 9ffb97e6
  Chandler Carruth authored Nov 27, 2011
  9ffb97e6
Nov 26, 2011
- Fix APFloat::convert so that it handles narrowing conversions correctly; it · a84ad7d0
  Eli Friedman authored Nov 26, 2011
  a84ad7d0
Nov 25, 2011
- This patch contains support for encoding FMA4 instructions and · 0f9a1f5e
  Bruno Cardoso Lopes authored Nov 25, 2011
  0f9a1f5e
Nov 24, 2011
- Remove 256-bit specific node types for UNPCKHPS/D and instead use the 128-bit... · d65a4444
  Craig Topper authored Nov 24, 2011
  d65a4444
- Fix a silly use-after-free issue. A much earlier version of this code · 7adee1a0
  Chandler Carruth authored Nov 24, 2011
  7adee1a0
- When adding blocks to the list of those which no longer have any CFG · d394bafd
  Chandler Carruth authored Nov 24, 2011
  d394bafd
Nov 23, 2011
- X86: Use btq for bit tests if the immediate can't be encoded in 32 bits. · ebcb4518
  Benjamin Kramer authored Nov 23, 2011
  ebcb4518
- test/CodeGen/X86/block-placement.ll: Add explicit -mtriple=i686-linux. X86... · 0b3e9964
  NAKAMURA Takumi authored Nov 23, 2011
  0b3e9964
- Relax an invariant that block placement was trying to assert a bit · 99fe42fb
  Chandler Carruth authored Nov 23, 2011
  99fe42fb
- I added several lines in X86 code generator that allow to choose · 779ba6d7
  Elena Demikhovsky authored Nov 23, 2011
  779ba6d7
- Handle the case of a no-return invoke correctly. It actually still has · 8c68f1f3
  Chandler Carruth authored Nov 23, 2011
  8c68f1f3
- Enable stack protectors for all arrays, not just char arrays. rdar://5875909 · ebb44646
  Bob Wilson authored Nov 23, 2011
  ebb44646
- Fix PR11422. · 02845410
  Jakob Stoklund Olesen authored Nov 23, 2011
  02845410
- Fix a crash in block placement due to an inner loop that happened to be · 4a87aa0c
  Chandler Carruth authored Nov 23, 2011
  4a87aa0c
Nov 22, 2011
- add basic PPC register-pressure feedback; adjust the vaarg test to match the... · 6f0ae783
  Hal Finkel authored Nov 22, 2011
  6f0ae783
- Fix a devilish miscompile exposed by block placement. The · ee54feb6
  Chandler Carruth authored Nov 22, 2011
  ee54feb6
- Add triple to the test. · c55e1af1
  Rafael Espindola authored Nov 22, 2011
  c55e1af1
- If a register is both an early clobber and part of a tied use, handle the use · 2021f382
  Rafael Espindola authored Nov 22, 2011
  2021f382
Nov 21, 2011
- Lowering for v32i8 to VPUNPCKLBW/VPUNPCKHBW when AVX2 is enabled. · 6270d072
  Craig Topper authored Nov 21, 2011
  6270d072
- Test case for r145026 · d12d6f4b
  Craig Topper authored Nov 21, 2011
  d12d6f4b
- Make LowerSIGN_EXTEND_INREG split 256-bit vectors when AVX1 is enabled and use... · a065238c
  Craig Topper authored Nov 21, 2011
  a065238c
Nov 20, 2011
- test/CodeGen/X86/block-placement.ll: Relax expressions for Win32. · 76dfa038
  NAKAMURA Takumi authored Nov 20, 2011
  76dfa038
- The logic for breaking the CFG in the presence of hot successors didn't · 18dfac38
  Chandler Carruth authored Nov 20, 2011
  18dfac38
- Add some comments to the latest test case I added here to document what · 20df3953
  Chandler Carruth authored Nov 20, 2011
  20df3953
- Add code for lowering v32i8 shifts by a splat to AVX2 immediate shift... · e79761df
  Craig Topper authored Nov 20, 2011
  e79761df
Nov 19, 2011
- Use 256-bit vcmpeqd for creating an all ones vector when AVX2 is enabled. · a3a65836
  Craig Topper authored Nov 19, 2011
  a3a65836
- Move the handling of unanalyzable branches out of the loop-driven chain · f3dc9eff
  Chandler Carruth authored Nov 19, 2011
  f3dc9eff
- Test cases for SSSE3/AVX integer horizontal add/sub. · 6d77f4ae
  Craig Topper authored Nov 19, 2011
  6d77f4ae
- Extend VPBLENDVB and VPSIGN lowering to work for AVX2. · de6b73bb
  Craig Topper authored Nov 19, 2011
  de6b73bb
Nov 18, 2011
- Add AVX2 vpbroadcast support · 1ec141d0
  Nadav Rotem authored Nov 18, 2011
  1ec141d0