[PATCH] D40098: [AMDGPU] Improve hazard checks for inline asm statements
Mark Searles via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu Nov 30 16:03:38 PST 2017
msearles marked 2 inline comments as done.
msearles added inline comments.
================
Comment at: lib/Target/AMDGPU/GCNHazardRecognizer.cpp:519
+int GCNHazardRecognizer::checkVALUHazardsHelper(const MachineOperand &Def, const MachineRegisterInfo &MRI) {
+ // Helper to check for the hazard where VMEM instructions that store more than
----------------
arsenm wrote:
> This looks like it goes over 80 lines
Fixed
================
Comment at: lib/Target/AMDGPU/GCNHazardRecognizer.cpp:548
- const SIRegisterInfo *TRI = ST.getRegisterInfo();
const MachineRegisterInfo &MRI = VALU->getParent()->getParent()->getRegInfo();
int WaitStatesNeeded = 0;
----------------
arsenm wrote:
> MF.getRegInfo()
Fixed this instance as well as a few others.
https://reviews.llvm.org/D40098
More information about the llvm-commits
mailing list