[llvm] [AMDGPU] Track LDS DMA VGPR source reads in expert scheduling (PR #203292)

[Jay Foad via llvm-commits]

================
@@ -2858,7 +2858,7 @@ bool SIInsertWaitcnts::generateWaitcnt(AMDGPU::Waitcnt Wait,
 
 std::optional<WaitEventType>
 SIInsertWaitcnts::getExpertSchedulingEventType(const MachineInstr &Inst) const {
-  if (TII.isVALU(Inst)) {
+  if (TII.isVALU(Inst) && !SIInstrInfo::isLDSDMA(Inst)) {
----------------
jayfoad wrote:

Could alternatively reorder the cases so the VMEM ones are checked before VALU. But this is fine too and makes the diff smaller :)

https://github.com/llvm/llvm-project/pull/203292