Commits · 58142dd8a1c9d682c7ce14bba3671f6d73dd3083 · Lorenzo Albano / LLVM bpEVL · GitLab

May 15, 2010
- "trap" pseudo-op turned out to be apple-local. · 2b7aace2
  Anton Korobeynikov authored May 15, 2010
  2b7aace2
- improve portability to systems that don't have round, patch by · 9e01b615
  Chris Lattner authored May 15, 2010
  9e01b615
- improve portability to systems that don't have powf/modf (e.g. solaris 9) · 93cd0f1c
  Chris Lattner authored May 15, 2010
  93cd0f1c
- Fix an GCC warning that seems to have actually caught a bug (!!!) in · 75142e6b
  Chandler Carruth authored May 15, 2010
  75142e6b
- Model 128-bit vld lane with REG_SEQUENCE. · 9e688cbc
  Evan Cheng authored May 15, 2010
  9e688cbc
- Calculate liveness on the fly for local registers. · 84ce2908
  Jakob Stoklund Olesen authored May 15, 2010
  84ce2908
- Teach the always inliner to release its inline cost estimates, like the basic · b35818eb
  Nick Lewycky authored May 15, 2010
  b35818eb
- Clean up, no functional change. · 002a45eb
  Nick Lewycky authored May 15, 2010
  002a45eb
- Add reg_nodbg_iterator · 3ced97c0
  Jakob Stoklund Olesen authored May 15, 2010
  3ced97c0
- v4i64 and v8i64 are only synthesizable when NEON is available. · 3d214cdf
  Evan Cheng authored May 15, 2010
  3d214cdf
- Allow TargetLowering::getRegClassFor() to be called on illegal types. Also · 4cad68eb
  Evan Cheng authored May 15, 2010
  4cad68eb
- Model 64-bit lane vld with REG_SEQUENCE. · 0cbd11df
  Evan Cheng authored May 15, 2010
  0cbd11df
- A partial re-def instruction may be a copy. · e26e56e7
  Evan Cheng authored May 15, 2010
  e26e56e7
- Teach two-address pass to do some coalescing while eliminating REG_SEQUENCE · 8c2d062e
  Evan Cheng authored May 14, 2010
  8c2d062e
- Add some comments about undef. · 62c70559
  Dan Gohman authored May 14, 2010
  62c70559
- Model VST*_UPD and VST*oddUPD pair with REG_SEQUENCE. · cb78e555
  Evan Cheng authored May 14, 2010
  cb78e555
- Fast ISel trivially coalesces away no-op casts, so check for this when · 88fb2535
  Dan Gohman authored May 14, 2010
  88fb2535
- Don't bother spilling before a return · 089e9421
  Jakob Stoklund Olesen authored May 14, 2010
  089e9421
- RegAllocLocal can count copies too · cdef6bc8
  Jakob Stoklund Olesen authored May 14, 2010
  cdef6bc8
- SystemZ really does mean "has calls" and not just "adjusts stack." Go ahead and · 0160e558
  Bill Wendling authored May 14, 2010
  0160e558
- Track allocatable instead of reserved regs, and never take an unallocatable hint. · b1601393
  Jakob Stoklund Olesen authored May 14, 2010
  b1601393
- Don't set kill flags for instructions which the scheduler has cloned. · 2f277c86
  Dan Gohman authored May 14, 2010
  2f277c86
- BR is a barrier. · 062a97f0
  Dan Gohman authored May 14, 2010
  062a97f0
May 14, 2010
- Several tail call tests apparently rely upon this being "adjusts stack" instead · 1713d958
  Bill Wendling authored May 14, 2010
  1713d958
- Avoid scanning the long tail of physreg operands on calls · e68b814c
  Jakob Stoklund Olesen authored May 14, 2010
  e68b814c
- Do not forget to mark prcessed arguments. · 36debf80
  Devang Patel authored May 14, 2010
  36debf80
- Count coalesced copies · 6c038e33
  Jakob Stoklund Olesen authored May 14, 2010
  6c038e33
- Allow virtreg redefines when verifying for RegAllocFast · 33af4fcd
  Jakob Stoklund Olesen authored May 14, 2010
  33af4fcd
- This should happen if there are no calls, not if it just doesn't adjust the · e9ac7ad6
  Bill Wendling authored May 14, 2010
  e9ac7ad6
- Revert r103804. The comment is correct. · db4bc54c
  Bill Wendling authored May 14, 2010
  db4bc54c
- Remove trailing whitespace · 866b74ba
  Jim Grosbach authored May 14, 2010
  866b74ba
- 80 column and trailing whitespace cleanup · d772bdeb
  Jim Grosbach authored May 14, 2010
  d772bdeb
- add cmd line option to leave dbgvalues in during post-RA sceduling. Useful · 25749ad5
  Jim Grosbach authored May 14, 2010
  25749ad5
- Fix comment. · 7bf4bae5
  Bill Wendling authored May 14, 2010
  7bf4bae5
- Rename "HasCalls" in MachineFrameInfo to "AdjustsStack" to better describe what · 95f6ebcb
  Bill Wendling authored May 14, 2010
  95f6ebcb
- Test case for r103800. · c87e8671
  Devang Patel authored May 14, 2010
  c87e8671
- Lowering of atomic instructions can result in operands being · 35dd005d
  Dan Gohman authored May 14, 2010
  35dd005d
- Add support to preserve type info for the variables that are removed by the optimizer. · e0a94bfe
  Devang Patel authored May 14, 2010
  e0a94bfe
- When verifying two-address instructions, check the following: · 670492c8
  Jakob Stoklund Olesen authored May 14, 2010
  670492c8
- Fix so "int3" is correctly accepted, added "into" and fixed "int" with an · 7bc111f5
  Kevin Enderby authored May 14, 2010
  7bc111f5