[llvm] [AMDGPU] Add mark last scratch load pass (PR #75512)

[Jay Foad via llvm-commits]

================
@@ -0,0 +1,143 @@
+//===-- AMDGPUMarkLastScratchLoad.cpp -------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Mark scratch load/spill instructions which are guaranteed to be the last time
+// this scratch slot is used so it can be evicted from caches.
+//
+//===----------------------------------------------------------------------===//
+
+#include "AMDGPU.h"
+#include "GCNSubtarget.h"
+#include "llvm/CodeGen/LiveStacks.h"
+#include "llvm/CodeGen/MachineOperand.h"
+
+using namespace llvm;
+
+#define DEBUG_TYPE "amdgpu-mark-last-scratch-load"
+
+namespace {
+
+class AMDGPUMarkLastScratchLoad : public MachineFunctionPass {
+private:
+  LiveStacks *LS = nullptr;
+  SlotIndexes *SI = nullptr;
+  const SIInstrInfo *SII = nullptr;
+
+public:
+  static char ID;
+
+  AMDGPUMarkLastScratchLoad() : MachineFunctionPass(ID) {
+    initializeAMDGPUMarkLastScratchLoadPass(*PassRegistry::getPassRegistry());
+  }
+
+  bool runOnMachineFunction(MachineFunction &MF) override;
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.addRequired<SlotIndexes>();
+    AU.addRequired<LiveStacks>();
+    AU.setPreservesAll();
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+
+  StringRef getPassName() const override {
+    return "AMDGPU Mark Last Scratch Load";
+  }
+};
+
+} // end anonymous namespace
+
+bool AMDGPUMarkLastScratchLoad::runOnMachineFunction(MachineFunction &MF) {
+  LLVM_DEBUG({
+    dbgs() << "********** Mark Last Scratch Load **********\n"
+           << "********** Function: " << MF.getName() << '\n';
+  });
+
+  if (skipFunction(MF.getFunction()))
+    return false;
+
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  if (ST.getGeneration() < AMDGPUSubtarget::GFX12)
+    return false;
+
+  LS = &getAnalysis<LiveStacks>();
+  SI = &getAnalysis<SlotIndexes>();
+  SII = ST.getInstrInfo();
+
+  const unsigned NumSlots = LS->getNumIntervals();
+  if (NumSlots == 0) {
+    LLVM_DEBUG(dbgs() << "No live slots, skipping\n");
+    return false;
+  }
+
+  LLVM_DEBUG(dbgs() << LS->getNumIntervals() << " intervals\n");
+
+  bool Changed = false;
+
+  for (auto &[SS, LI] : *LS) {
+    LLVM_DEBUG(dbgs() << "Checking interval: " << LI << "\n");
+
+    for (const LiveRange::Segment &Segment : LI.segments) {
+      LLVM_DEBUG(dbgs() << "  Checking segment: " << Segment << "\n");
+
+      // Ignore segments that run to the end of basic block because in this case
+      // slot is still live at the end of it.
+      if (Segment.end.isBlock())
+        continue;
+
+      const int FrameIndex = Register::stackSlot2Index(LI.reg());
+      MachineInstr *LastLoad = nullptr;
+
+      MachineInstr *MISegmentStart = SI->getInstructionFromIndex(Segment.start);
+      MachineInstr *MISegmentEnd = SI->getInstructionFromIndex(Segment.end);
+      if (!MISegmentEnd) {
+        // FIXME: The start and end can refer to deleted instructions. We should
+        // be able to handle this more gracefully by finding the closest real
+        // instructions.
+        continue;
+      }
+      MachineBasicBlock *BB = MISegmentEnd->getParent();
+
+      // Start iteration backwards from segment end until the start of basic
+      // block or start of segment if it is in the same basic block.
+      auto End = BB->instr_rend();
+      if (MISegmentStart && MISegmentStart->getParent() == BB)
+        End = MISegmentStart->getReverseIterator();
+
+      for (auto MI = MISegmentEnd->getReverseIterator(); MI != End; ++MI) {
+        int LoadFI = 0;
+
+        if (SII->isLoadFromStackSlot(*MI, LoadFI) && LoadFI == FrameIndex) {
+          LastLoad = &*MI;
+          break;
+        }
+      }
+
+      if (LastLoad) {
+        MachineOperand *LastUse =
+            SII->getNamedOperand(*LastLoad, AMDGPU::OpName::last_use);
+        assert(LastUse && "This instruction must have a last_use operand");
+        LastUse->setImm(1);
----------------
jayfoad wrote:

Nowehere yet. There needs to be a follow up patch that sets the temporal hint on the load instructions to LU (last use) when this is set. How would you suggest propagating this information, without modifying the pseudo?

https://github.com/llvm/llvm-project/pull/75512