[DebugInfo][LoopVectorize] Preserve DL in generated phi instruction

[gramanas]

Summary

Similar to rL335904.

When creating phi instructions to resume at the scalar part of the loop,
copy the DebugLoc from the original phi over to the new one.

Based on Greg's report:
https://bugs.llvm.org/show_bug.cgi?id=37955

[vsk]

test/Transforms/LoopVectorize/calloc.ll

2 ↗

(On Diff #153483)

The two RUN lines here are testing substantially different things. If there isn't a simpler file to repurpose as a debug info test, it would be simpler to create a (perhaps more reduced?) test.

10 ↗

(On Diff #153483)

Please add a 'CHECK-LABEL: define {{.*}} <function name>' to ensure that subsequent check lines are testing the correct function.

52 ↗

(On Diff #153483)

This simply tests that the phi is assigned a debug location, but it doesn't test that it's assigned a correct location. Please tighten the test by checking the line number.

[gramanas]

Address inline comments.

[gramanas]

If there isn't a simpler file to repurpose as a debug info test, it would be simpler to create a (perhaps more reduced?) test.

Since this particular fix is about the phi that resumes the scalar part of the loop, I don't think it would be appropriate to create a new reduced test just for this.
It seems like it's a quite common thing to occur in loop-vectorize.

[gramanas]

Formatting

[vsk]

It'd be a good idea rename the test so that readers know what to expect if it fails (say, 'preserve-dbg-loc-and-loop-metadata.ll').

Otherwise, LGTM, thanks!

llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp

BasicBlock *InnerLoopVectorizer::createVectorizedLoopSkeleton() {
   LoopVectorizationLegality::InductionList *List = Legal->getInductionVars();
   for (auto &InductionEntry : *List) {
     PHINode *OrigPhi = InductionEntry.first;
     InductionDescriptor II = InductionEntry.second;
 
     // Create phi nodes to merge from the  backedge-taken check block.
     PHINode *BCResumeVal = PHINode::Create(
         OrigPhi->getType(), 3, "bc.resume.val", ScalarPH->getTerminator());
+    // Copy original phi DL over to the new one.
+    BCResumeVal->setDebugLoc(OrigPhi->getDebugLoc());
     Value *&EndValue = IVEndValues[OrigPhi];
     if (OrigPhi == OldInduction) {
       // We know what the end value is.
       EndValue = CountRoundDown;
     } else {
       IRBuilder<> B(Lp->getLoopPreheader()->getTerminator());
       Type *StepType = II.getStep()->getType();
       Instruction::CastOps CastOp =

llvm/trunk/test/Transforms/LoopVectorize/duplicated-metadata.ll

-; RUN: opt < %s -loop-vectorize -S 2>&1 | FileCheck %s
-target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
-
-; This test makes sure we don't duplicate the loop vectorizer's metadata
-; while marking them as already vectorized (by setting width = 1), even
-; at lower optimization levels, where no extra cleanup is done
-
-define void @_Z3fooPf(float* %a) {
-entry:
-  br label %for.body
-
-for.body:                                         ; preds = %for.body, %entry
-  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
-  %arrayidx = getelementptr inbounds float, float* %a, i64 %indvars.iv
-  %p = load float, float* %arrayidx, align 4
-  %mul = fmul float %p, 2.000000e+00
-  store float %mul, float* %arrayidx, align 4
-  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
-  %exitcond = icmp eq i64 %indvars.iv.next, 1024
-  br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0
-
-for.end:                                          ; preds = %for.body
-  ret void
-}
-
-!0 = !{!0, !1}
-!1 = !{!"llvm.loop.vectorize.width", i32 4}
-; CHECK-NOT: !{metadata !"llvm.loop.vectorize.width", i32 4}
-; CHECK: !{!"llvm.loop.isvectorized", i32 1}

llvm/trunk/test/Transforms/LoopVectorize/preserve-dbg-loc-and-loop-metadata.ll

+; RUN: opt < %s -loop-vectorize -S 2>&1 | FileCheck %s
+; RUN: opt < %s -debugify -loop-vectorize -S | FileCheck %s -check-prefix DEBUGLOC
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+; This test makes sure we don't duplicate the loop vectorizer's metadata
+; while marking them as already vectorized (by setting width = 1), even
+; at lower optimization levels, where no extra cleanup is done
+
+; DEBUGLOC-LABEL: define void @_Z3fooPf(
+; Check that the phi to resume the scalar part of the loop
+; has Debug Location.
+define void @_Z3fooPf(float* %a) {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds float, float* %a, i64 %indvars.iv
+  %p = load float, float* %arrayidx, align 4
+  %mul = fmul float %p, 2.000000e+00
+  store float %mul, float* %arrayidx, align 4
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 1024
+  br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0
+
+; DEBUGLOC: scalar.ph:
+; DEBUGLOC-NEXT:    %bc.resume.val = phi {{.*}} !dbg ![[DbgLoc:[0-9]+]]
+;
+; DEBUGLOC: ![[DbgLoc]] = !DILocation(line: 2
+
+for.end:                                          ; preds = %for.body
+  ret void
+}
+
+!0 = !{!0, !1}
+!1 = !{!"llvm.loop.vectorize.width", i32 4}
+; CHECK-NOT: !{metadata !"llvm.loop.vectorize.width", i32 4}
+; CHECK: !{!"llvm.loop.isvectorized", i32 1}