projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: a5fbf24) | patch
R600/SI: Optimize SI_KILL for constant operands
authorMichel Danzer <michel.daenzer@amd.com>
Thu, 27 Feb 2014 01:47:09 +0000 (01:47 +0000)
committerMichel Danzer <michel.daenzer@amd.com>
Thu, 27 Feb 2014 01:47:09 +0000 (01:47 +0000)
commit644aecfc973e6b899230a3c7eb7541c948e69522
tree3e540e03f1a4a7d908d296b4e6fb08607ae592e3tree | snapshot
parenta5fbf2471688af979966da0d540875ec43308b07commit | diff
R600/SI: Optimize SI_KILL for constant operands

If the SI_KILL operand is constant, we can either clear the exec mask if
the operand is negative, or do nothing otherwise.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202337 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/R600/SIInstructions.td diff | blob | history
lib/Target/R600/SILowerControlFlow.cpp diff | blob | history
test/CodeGen/R600/llvm.AMDGPU.kill.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors