llvm · vporpo · Feb 28, 2025
diff --git a/llvm/include/llvm/SandboxIR/Pass.h b/llvm/include/llvm/SandboxIR/Pass.h
@@ -17,6 +17,7 @@ namespace llvm {
 class AAResults;
 class ScalarEvolution;
 class TargetTransformInfo;
+class OptimizationRemarkEmitter;
 
 namespace sandboxir {
 
@@ -27,17 +28,20 @@ class Analyses {
   AAResults *AA = nullptr;
   ScalarEvolution *SE = nullptr;
   TargetTransformInfo *TTI = nullptr;
+  OptimizationRemarkEmitter *ORE = nullptr;
 
   Analyses() = default;
 
 public:
-  Analyses(AAResults &AA, ScalarEvolution &SE, TargetTransformInfo &TTI)
-      : AA(&AA), SE(&SE), TTI(&TTI) {}
+  Analyses(AAResults &AA, ScalarEvolution &SE, TargetTransformInfo &TTI,
+           OptimizationRemarkEmitter &ORE)
+      : AA(&AA), SE(&SE), TTI(&TTI), ORE(&ORE) {}
 
 public:
   AAResults &getAA() const { return *AA; }
   ScalarEvolution &getScalarEvolution() const { return *SE; }
   TargetTransformInfo &getTTI() const { return *TTI; }
+  OptimizationRemarkEmitter &getORE() const { return *ORE; }
   /// For use by unit tests.
   static Analyses emptyForTesting() { return Analyses(); }
 };

diff --git a/llvm/include/llvm/SandboxIR/Utils.h b/llvm/include/llvm/SandboxIR/Utils.h
@@ -17,6 +17,7 @@
 #include "llvm/Analysis/MemoryLocation.h"
 #include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/Analysis/ValueTracking.h"
+#include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/Verifier.h"
 #include "llvm/SandboxIR/Function.h"
 #include "llvm/SandboxIR/Instruction.h"
@@ -131,6 +132,15 @@ class Utils {
     const auto &LLVMF = *cast<llvm::Function>(F->Val);
     return llvm::verifyFunction(LLVMF, &OS);
   }
+
+  /// Returns an optimization remark for \p I's location.
+  // TODO: OptimizationRemark can leak llvm IR through getRegion().
+  static OptimizationRemark getOptimizationRemark(const char *PassName,
+                                                  StringRef RemarkName,
+                                                  Instruction *I) {
+    return OptimizationRemark(PassName, RemarkName,
+                              cast<llvm::Instruction>(I->Val));
+  }
 };
 
 } // namespace llvm::sandboxir

diff --git a/llvm/include/llvm/Transforms/Vectorize/SandboxVectorizer/Debug.h b/llvm/include/llvm/Transforms/Vectorize/SandboxVectorizer/Debug.h
@@ -15,7 +15,8 @@
 
 #include "llvm/Support/Debug.h"
 
-#define DEBUG_TYPE "sandbox-vectorizer"
+#define PASS_NAME "sandbox-vectorizer"
+#define DEBUG_TYPE PASS_NAME
 #define DEBUG_PREFIX "SBVec: "
 
 #endif // LLVM_TRANSFORMS_VECTORIZE_SANDBOXVECTORIZER_DEBUG_H
diff --git a/llvm/include/llvm/Transforms/Vectorize/SandboxVectorizer/SandboxVectorizer.h b/llvm/include/llvm/Transforms/Vectorize/SandboxVectorizer/SandboxVectorizer.h
@@ -11,6 +11,7 @@
 #include <memory>
 
 #include "llvm/Analysis/AliasAnalysis.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/IR/PassManager.h"
 #include "llvm/SandboxIR/Context.h"
@@ -24,6 +25,7 @@ class SandboxVectorizerPass : public PassInfoMixin<SandboxVectorizerPass> {
   TargetTransformInfo *TTI = nullptr;
   AAResults *AA = nullptr;
   ScalarEvolution *SE = nullptr;
+  OptimizationRemarkEmitter *ORE = nullptr;
   // NOTE: We define the Context as a pass-scope object instead of local object
   // in runOnFunction() because the passes defined in the pass-manager need
   // access to it for registering/deregistering callbacks during construction

diff --git a/llvm/lib/Transforms/Vectorize/SandboxVectorizer/Passes/TransactionAcceptOrRevert.cpp b/llvm/lib/Transforms/Vectorize/SandboxVectorizer/Passes/TransactionAcceptOrRevert.cpp
@@ -7,6 +7,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Transforms/Vectorize/SandboxVectorizer/Passes/TransactionAcceptOrRevert.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
+#include "llvm/SandboxIR/Utils.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/InstructionCost.h"
 #include "llvm/Transforms/Vectorize/SandboxVectorizer/Debug.h"
@@ -29,9 +31,13 @@ bool TransactionAcceptOrRevert::runOnRegion(Region &Rgn, const Analyses &A) {
                     << CostAfter << "/" << CostThreshold << ")\n");
   // TODO: Print costs / write to remarks.
   auto &Tracker = Rgn.getContext().getTracker();
-  if (CostAfterMinusBefore < -CostThreshold) {
+  if (!Rgn.empty() && CostAfterMinusBefore < -CostThreshold) {
     bool HasChanges = !Tracker.empty();
     Tracker.accept();
+    // TODO: Use a more accurate debug location than Rgn.begin().
+    A.getORE().emit(
+        Utils::getOptimizationRemark(PASS_NAME, "Vectorized", *Rgn.begin())
+        << "Vectorized with cost " << ore::NV("Cost", CostAfterMinusBefore));
     LLVM_DEBUG(dbgs() << DEBUG_PREFIX << "*** Transaction Accept ***\n");
     return HasChanges;
   }

diff --git a/llvm/lib/Transforms/Vectorize/SandboxVectorizer/SandboxVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SandboxVectorizer/SandboxVectorizer.cpp
@@ -72,6 +72,7 @@ PreservedAnalyses SandboxVectorizerPass::run(Function &F,
   TTI = &AM.getResult<TargetIRAnalysis>(F);
   AA = &AM.getResult<AAManager>(F);
   SE = &AM.getResult<ScalarEvolutionAnalysis>(F);
+  ORE = &AM.getResult<OptimizationRemarkEmitterAnalysis>(F);
 
   bool Changed = runImpl(F);
   if (!Changed)
@@ -132,7 +133,7 @@ bool SandboxVectorizerPass::runImpl(Function &LLVMF) {
 
   // Create SandboxIR for LLVMF and run BottomUpVec on it.
   sandboxir::Function &F = *Ctx->createFunction(&LLVMF);
-  sandboxir::Analyses A(*AA, *SE, *TTI);
+  sandboxir::Analyses A(*AA, *SE, *TTI, *ORE);
   bool Change = FPM.runOnFunction(F, A);
   // Given that sandboxir::Context `Ctx` is defined at a pass-level scope, the
   // maps from LLVM IR to Sandbox IR may go stale as later passes remove LLVM IR

diff --git a/llvm/test/Transforms/SandboxVectorizer/remarks.ll b/llvm/test/Transforms/SandboxVectorizer/remarks.ll
@@ -0,0 +1,31 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -passes=sandbox-vectorizer -sbvec-vec-reg-bits=1024 -pass-remarks-output=%t < %s | FileCheck %s
+; RUN: FileCheck --input-file=%t --check-prefix=REMARKS %s
+
+; REMARKS-LABEL: --- !Passed
+; REMARKS-LABEL: Pass:            sandbox-vectorizer
+; REMARKS-LABEL: Name:            Vectorized
+; REMARKS-LABEL: Function:        remarks1
+; REMARKS-LABEL: Args:
+; REMARKS-LABEL:   - String:          'Vectorized with cost '
+; REMARKS-LABEL:   - Cost:            '-3'
+
+define void @remarks1(ptr %ptr) {
+; CHECK-LABEL: define void @remarks1(
+; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-NEXT:    [[PTR0:%.*]] = getelementptr float, ptr [[PTR]], i32 0
+; CHECK-NEXT:    [[VECL:%.*]] = load <2 x float>, ptr [[PTR0]], align 4, !sandboxvec [[META0:![0-9]+]]
+; CHECK-NEXT:    store <2 x float> [[VECL]], ptr [[PTR0]], align 4, !sandboxvec [[META0]]
+; CHECK-NEXT:    ret void
+;
+  %ptr0 = getelementptr float, ptr %ptr, i32 0
+  %ptr1 = getelementptr float, ptr %ptr, i32 1
+  %ld0 = load float, ptr %ptr0
+  %ld1 = load float, ptr %ptr1
+  store float %ld0, ptr %ptr0
+  store float %ld1, ptr %ptr1
+  ret void
+}
+;.
+; CHECK: [[META0]] = distinct !{!"sandboxregion"}
+;.
diff --git a/llvm/test/Transforms/SandboxVectorizer/remarks_not_vectorized.ll b/llvm/test/Transforms/SandboxVectorizer/remarks_not_vectorized.ll
@@ -0,0 +1,50 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -passes=sandbox-vectorizer -sbvec-cost-threshold=-9999 -sbvec-vec-reg-bits=1024 -pass-remarks-output=%t < %s | FileCheck %s
+; RUN: FileCheck --allow-empty --input-file=%t --check-prefix=REMARKS %s
+
+; REMARKS-NOT: .*
+
+; Check that we don't get any remarks when the code isn't vectorized
+; but the threshold allows it.
+define void @no_remarks1(ptr %ptr) {
+; CHECK-LABEL: define void @no_remarks1(
+; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-NEXT:    [[PTR0:%.*]] = getelementptr float, ptr [[PTR]], i32 0
+; CHECK-NEXT:    [[PTR1:%.*]] = getelementptr float, ptr [[PTR]], i32 42
+; CHECK-NEXT:    [[LD0:%.*]] = load float, ptr [[PTR0]], align 4
+; CHECK-NEXT:    [[LD1:%.*]] = load float, ptr [[PTR1]], align 4
+; CHECK-NEXT:    store float [[LD0]], ptr [[PTR0]], align 4
+; CHECK-NEXT:    store float [[LD1]], ptr [[PTR1]], align 4
+; CHECK-NEXT:    ret void
+;
+  %ptr0 = getelementptr float, ptr %ptr, i32 0
+  %ptr1 = getelementptr float, ptr %ptr, i32 42
+  %ld0 = load float, ptr %ptr0
+  %ld1 = load float, ptr %ptr1
+  store float %ld0, ptr %ptr0
+  store float %ld1, ptr %ptr1
+  ret void
+}
+
+; Same but doesn't get vectorized because of scheduling.
+define void @no_remarks2(ptr %ptr, ptr %ptrX) {
+; CHECK-LABEL: define void @no_remarks2(
+; CHECK-SAME: ptr [[PTR:%.*]], ptr [[PTRX:%.*]]) {
+; CHECK-NEXT:    [[PTR0:%.*]] = getelementptr float, ptr [[PTR]], i32 0
+; CHECK-NEXT:    [[PTR1:%.*]] = getelementptr float, ptr [[PTR]], i32 1
+; CHECK-NEXT:    [[LD0:%.*]] = load float, ptr [[PTR0]], align 4
+; CHECK-NEXT:    [[LD1:%.*]] = load float, ptr [[PTR1]], align 4
+; CHECK-NEXT:    store float [[LD0]], ptr [[PTR0]], align 4
+; CHECK-NEXT:    store float 0.000000e+00, ptr [[PTRX]], align 4
+; CHECK-NEXT:    store float [[LD1]], ptr [[PTR1]], align 4
+; CHECK-NEXT:    ret void
+;
+  %ptr0 = getelementptr float, ptr %ptr, i32 0
+  %ptr1 = getelementptr float, ptr %ptr, i32 1
+  %ld0 = load float, ptr %ptr0
+  %ld1 = load float, ptr %ptr1
+  store float %ld0, ptr %ptr0
+  store float 0.0, ptr %ptrX
+  store float %ld1, ptr %ptr1
+  ret void
+}
diff --git a/llvm/unittests/Transforms/Vectorize/SandboxVectorizer/SandboxVectorizerTest.cpp b/llvm/unittests/Transforms/Vectorize/SandboxVectorizer/SandboxVectorizerTest.cpp
@@ -57,6 +57,7 @@ define void @foo() {
   AM.registerPass([] { return AssumptionAnalysis(); });
   AM.registerPass([] { return DominatorTreeAnalysis(); });
   AM.registerPass([] { return LoopAnalysis(); });
+  AM.registerPass([] { return OptimizationRemarkEmitterAnalysis(); });
   SVecPass.run(LLVMF, AM);
   // This shouldn't crash.
   SVecPass.run(LLVMF, AM);