[llvm-branch-commits] [llvm] [AMDGPU] Apply occupancy-aware register allocation anti-hints (PR #218074)

[Lukas Sommer via llvm-branch-commits]

================
@@ -4173,6 +4177,126 @@ bool SIRegisterInfo::getRegAllocationHints(Register VirtReg,
   }
 }
 
+bool SIRegisterInfo::shouldApplyAntiHints(
+    const MachineFunction &MF, unsigned NumAllocatedVGPRs,
+    unsigned &MaxVGPRsForCurrentOccupancy) const {
+
+  const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
+  unsigned DynamicVGPRBlockSize = MFI->getDynamicVGPRBlockSize();
+  unsigned TargetOccupancy = MFI->getOccupancy();
+  unsigned CurrentOccupancy =
+      ST.getOccupancyWithNumVGPRs(NumAllocatedVGPRs, DynamicVGPRBlockSize);
+  MaxVGPRsForCurrentOccupancy =
+      ST.getMaxNumVGPRs(CurrentOccupancy, DynamicVGPRBlockSize);
+
+  LLVM_DEBUG(dbgs() << "anti-hints: " << NumAllocatedVGPRs
+                    << " VGPRs allocated, target occupancy " << TargetOccupancy
+                    << ", current occupancy " << CurrentOccupancy << '\n');
+
+  // If we are already at lowest occupancy, then there is no need to protect
+  // against occupancy regression.
+  if (CurrentOccupancy == 1)
+    return true;
+
+  // Do not apply anti-hints if we are reaching close to the VGPR budget. For
+  // target occupancy, the 80% cutoff is a conservative: anti-hints are disabled
+  // early enough that later registers still have headroom to stay at target
+  // occupancy. For current occupancy, the 95% cutoff margin is used to apply
----------------
sommerlukas wrote:

"to apply" or "to not apply"?

https://github.com/llvm/llvm-project/pull/218074