projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 598d9dd) | patch
[X86] Correct the implementation of the avx512 masked fmsubadd autoupgrade code to...
authorCraig Topper <craig.topper@intel.com>
Fri, 7 Feb 2020 23:04:22 +0000 (15:04 -0800)
committerCraig Topper <craig.topper@intel.com>
Fri, 7 Feb 2020 23:27:05 +0000 (15:27 -0800)
commitbb717d3f469b997ce6b13f2b87ffd78d2d2254b7
treea29cba1b3cd3b34ae3150aad1579c70218573bf6tree | snapshot
parent598d9dd846f1086abfc1c5055df0d25e8a485fc9commit | diff
[X86] Correct the implementation of the avx512 masked fmsubadd autoupgrade code to not leave the negate unconnected.

This was causing us to generate fmaddsub instead of fmsubadd if
rounding control is not 4.
llvm/lib/IR/AutoUpgrade.cpp diff | blob | history
llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll diff | blob | history
Domain: System / Toolchain;