[Assignment Tracking][23/*] Account for assignment tracking in SLP Vectorizer

[Orlando]

Summary

The SLP-Vectorizer can merge a set of scalar stores into a single vectorized store. Merge DIAssignID intrinsics from the scalar stores onto the new vectorized store.

---

The test currently requires x86 - if anyone knows any command line options to tell opt to enable the vectorization in the test without the triple please let me know.

[jmorse]

LGTM with one question.

The test currently requires x86 - if anyone knows any command line options to tell opt to enable the vectorization in the test without the triple please let me know.

I suspect there are numerous target hooks that have to be called to determine whether or not the vectorization is profitable, chances of getting a fully generic test that works is low.

llvm/test/DebugInfo/Generic/assignment-tracking/slp-vectorizer/merge-scalars.ll

29–30

This raises the tantalising question of why the store goes in the middle of the dbg.assigns, and whether that's going to mean we generate a different location list.

(Not a big deal IMO).

[Orlando]

llvm/test/DebugInfo/Generic/assignment-tracking/slp-vectorizer/merge-scalars.ll

29–30

This is a good point. I think the importance of the changes to a loclist probably depend on the instructions linenos / relationship to stepping. I had similar thoughts and questions when working on SROA, which are explained in the large comment in D133296 that starts "We could use more precision".

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

for (InsertElementInst *IE : reverse(ShuffledInserts[I].InsertElements)) {
                             PoisonValue::get(IE->getOperand(0)->getType()));
       IE->replaceUsesOfWith(IE->getOperand(1),
                             PoisonValue::get(IE->getOperand(1)->getType()));
       eraseInstruction(IE);
     }
     CSEBlocks.insert(LastInsert->getParent());
   }
 
+  SmallVector<Instruction *> RemovedInsts;
   // For each vectorized value:
   for (auto &TEPtr : VectorizableTree) {
     TreeEntry *Entry = TEPtr.get();
 
     // No need to handle users of gathered values.
     if (Entry->State == TreeEntry::NeedToGather)
       continue;
 
#ifndef NDEBUG
                   (isa_and_nonnull<Instruction>(U) &&
                    isDeleted(cast<Instruction>(U)))) &&
                  "Deleting out-of-tree value");
         }
       }
 #endif
       LLVM_DEBUG(dbgs() << "SLP: \tErasing scalar:" << *Scalar << ".\n");
       eraseInstruction(cast<Instruction>(Scalar));
+      // Retain to-be-deleted instructions for some debug-info
+      // bookkeeping. NOTE: eraseInstruction only marks the instruction for
+      // deletion - instructions are not deleted until later.
+      RemovedInsts.push_back(cast<Instruction>(Scalar));
     }
   }
 
+  // Merge the DIAssignIDs from the about-to-be-deleted instructions into the
+  // new vector instruction.
+  if (auto *V = dyn_cast<Instruction>(VectorizableTree[0]->VectorizedValue))
+    V->mergeDIAssignID(RemovedInsts);
+
   Builder.ClearInsertionPoint();
   InstrElementSize.clear();
 
   return VectorizableTree[0]->VectorizedValue;
 }
 
 void BoUpSLP::optimizeGatherSequence() {
   LLVM_DEBUG(dbgs() << "SLP: Optimizing " << GatherShuffleExtractSeq.size()

llvm/test/DebugInfo/Generic/assignment-tracking/slp-vectorizer/merge-scalars.ll

+; REQUIRES: x86-registered-target
+; RUN: opt -passes=slp-vectorizer -S -o - %s -experimental-assignment-tracking \
+; RUN: | FileCheck %s
+
+;; $ cat test.cpp
+;; float get();
+;; float ext(float*);
+;; void fun(float k1, float k2, float k3, float k4) {
+;;   float quad[4];
+;;   float c1 = get();
+;;   float c2 = get();
+;;   quad[0] = c1 - k1 - k3;
+;;   quad[1] = c2 - k2 - k4;
+;;   quad[2] = c1 - k1 + k3;
+;;   quad[3] = c2 - k2 + k4;
+;;   ext(quad);
+;; }
+;;
+;; Generated by grabbingthe IR before SLP in:
+;; $ clang++ -O2 -g test.cpp -Xclang -fexperimental-assignment-tracking
+
+;; Test that dbg.assigns linked to the the scalar stores to quad get linked to
+;; the vector store that replaces them.
+
+; CHECK: call void @llvm.dbg.assign(metadata float undef, metadata ![[VAR:[0-9]+]], metadata !DIExpression(DW_OP_LLVM_fragment, 0, 32), metadata ![[ID:[0-9]+]], metadata ptr %arrayidx, metadata !DIExpression())
+; CHECK: call void @llvm.dbg.assign(metadata float undef, metadata ![[VAR]], metadata !DIExpression(DW_OP_LLVM_fragment, 32, 32), metadata ![[ID]], metadata ptr %quad, metadata !DIExpression(DW_OP_plus_uconst, 4))
+; CHECK: call void @llvm.dbg.assign(metadata float undef, metadata ![[VAR]], metadata !DIExpression(DW_OP_LLVM_fragment, 64, 32), metadata ![[ID]], metadata ptr %quad, metadata !DIExpression(DW_OP_plus_uconst, 8))
+; CHECK: store <4 x float> {{.*}} !DIAssignID ![[ID]]
+; CHECK: call void @llvm.dbg.assign(metadata float undef, metadata ![[VAR]], metadata !DIExpression(DW_OP_LLVM_fragment, 96, 32), metadata ![[ID]], metadata ptr %quad, metadata !DIExpression(DW_OP_plus_uconst, 12))
+
+target triple = "x86_64-unknown-unknown"
+
+define dso_local void @_Z3funffff(float %k1, float %k2, float %k3, float %k4) local_unnamed_addr #0 !dbg !7 {
+entry:
+  %quad = alloca [4 x float], align 16, !DIAssignID !27
+  call void @llvm.dbg.assign(metadata i1 undef, metadata !16, metadata !DIExpression(), metadata !27, metadata ptr %quad, metadata !DIExpression()), !dbg !23
+  call void @llvm.dbg.assign(metadata float %k1, metadata !12, metadata !DIExpression(), metadata !30, metadata ptr undef, metadata !DIExpression()), !dbg !23
+  call void @llvm.dbg.assign(metadata float %k2, metadata !13, metadata !DIExpression(), metadata !31, metadata ptr undef, metadata !DIExpression()), !dbg !23
+  call void @llvm.dbg.assign(metadata float %k3, metadata !14, metadata !DIExpression(), metadata !32, metadata ptr undef, metadata !DIExpression()), !dbg !23
+  call void @llvm.dbg.assign(metadata float %k4, metadata !15, metadata !DIExpression(), metadata !33, metadata ptr undef, metadata !DIExpression()), !dbg !23
+  %call = tail call float @_Z3getv(), !dbg !35
+  call void @llvm.dbg.assign(metadata float %call, metadata !20, metadata !DIExpression(), metadata !36, metadata ptr undef, metadata !DIExpression()), !dbg !23
+  %call1 = tail call float @_Z3getv(), !dbg !37
+  call void @llvm.dbg.assign(metadata float %call1, metadata !21, metadata !DIExpression(), metadata !38, metadata ptr undef, metadata !DIExpression()), !dbg !23
+  %sub = fsub float %call, %k1, !dbg !39
+  %sub2 = fsub float %sub, %k3, !dbg !40
+  %arrayidx = getelementptr inbounds [4 x float], ptr %quad, i64 0, i64 0, !dbg !41
+  store float %sub2, ptr %arrayidx, align 16, !dbg !42, !DIAssignID !47
+  call void @llvm.dbg.assign(metadata float %sub2, metadata !16, metadata !DIExpression(DW_OP_LLVM_fragment, 0, 32), metadata !47, metadata ptr %arrayidx, metadata !DIExpression()), !dbg !23
+  %sub3 = fsub float %call1, %k2, !dbg !48
+  %sub4 = fsub float %sub3, %k4, !dbg !49
+  %arrayidx5 = getelementptr inbounds [4 x float], ptr %quad, i64 0, i64 1, !dbg !50
+  store float %sub4, ptr %arrayidx5, align 4, !dbg !51, !DIAssignID !52
+  call void @llvm.dbg.assign(metadata float %sub4, metadata !16, metadata !DIExpression(DW_OP_LLVM_fragment, 32, 32), metadata !52, metadata ptr %arrayidx5, metadata !DIExpression()), !dbg !23
+  %add = fadd float %sub, %k3, !dbg !53
+  %arrayidx7 = getelementptr inbounds [4 x float], ptr %quad, i64 0, i64 2, !dbg !54
+  store float %add, ptr %arrayidx7, align 8, !dbg !55, !DIAssignID !56
+  call void @llvm.dbg.assign(metadata float %add, metadata !16, metadata !DIExpression(DW_OP_LLVM_fragment, 64, 32), metadata !56, metadata ptr %arrayidx7, metadata !DIExpression()), !dbg !23
+  %add9 = fadd float %sub3, %k4, !dbg !57
+  %arrayidx10 = getelementptr inbounds [4 x float], ptr %quad, i64 0, i64 3, !dbg !58
+  store float %add9, ptr %arrayidx10, align 4, !dbg !59, !DIAssignID !60
+  call void @llvm.dbg.assign(metadata float %add9, metadata !16, metadata !DIExpression(DW_OP_LLVM_fragment, 96, 32), metadata !60, metadata ptr %arrayidx10, metadata !DIExpression()), !dbg !23
+  %call11 = call float @_Z3extPf(ptr nonnull %arrayidx), !dbg !61
+  ret void, !dbg !62
+}
+
+declare void @llvm.lifetime.start.p0i8(i64 immarg, ptr nocapture) #1
+declare !dbg !63 dso_local float @_Z3getv() local_unnamed_addr #2
+declare !dbg !66 dso_local float @_Z3extPf(ptr) local_unnamed_addr #2
+declare void @llvm.lifetime.end.p0i8(i64 immarg, ptr nocapture) #1
+declare void @llvm.dbg.assign(metadata, metadata, metadata, metadata, metadata, metadata) #3
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !1, producer: "clang version 12.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "test.cpp", directory: "/")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 12.0.0"}
+!7 = distinct !DISubprogram(name: "fun", linkageName: "_Z3funffff", scope: !1, file: !1, line: 3, type: !8, scopeLine: 3, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !11)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10, !10, !10, !10}
+!10 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
+!11 = !{!12, !13, !14, !15, !16, !20, !21}
+!12 = !DILocalVariable(name: "k1", arg: 1, scope: !7, file: !1, line: 3, type: !10)
+!13 = !DILocalVariable(name: "k2", arg: 2, scope: !7, file: !1, line: 3, type: !10)
+!14 = !DILocalVariable(name: "k3", arg: 3, scope: !7, file: !1, line: 3, type: !10)
+!15 = !DILocalVariable(name: "k4", arg: 4, scope: !7, file: !1, line: 3, type: !10)
+!16 = !DILocalVariable(name: "quad", scope: !7, file: !1, line: 4, type: !17)
+!17 = !DICompositeType(tag: DW_TAG_array_type, baseType: !10, size: 128, elements: !18)
+!18 = !{!19}
+!19 = !DISubrange(count: 4)
+!20 = !DILocalVariable(name: "c1", scope: !7, file: !1, line: 5, type: !10)
+!21 = !DILocalVariable(name: "c2", scope: !7, file: !1, line: 6, type: !10)
+!22 = distinct !DIAssignID()
+!23 = !DILocation(line: 0, scope: !7)
+!24 = distinct !DIAssignID()
+!25 = distinct !DIAssignID()
+!26 = distinct !DIAssignID()
+!27 = distinct !DIAssignID()
+!28 = distinct !DIAssignID()
+!29 = distinct !DIAssignID()
+!30 = distinct !DIAssignID()
+!31 = distinct !DIAssignID()
+!32 = distinct !DIAssignID()
+!33 = distinct !DIAssignID()
+!34 = !DILocation(line: 4, column: 3, scope: !7)
+!35 = !DILocation(line: 5, column: 14, scope: !7)
+!36 = distinct !DIAssignID()
+!37 = !DILocation(line: 6, column: 14, scope: !7)
+!38 = distinct !DIAssignID()
+!39 = !DILocation(line: 7, column: 16, scope: !7)
+!40 = !DILocation(line: 7, column: 21, scope: !7)
+!41 = !DILocation(line: 7, column: 3, scope: !7)
+!42 = !DILocation(line: 7, column: 11, scope: !7)
+!47 = distinct !DIAssignID()
+!48 = !DILocation(line: 8, column: 16, scope: !7)
+!49 = !DILocation(line: 8, column: 21, scope: !7)
+!50 = !DILocation(line: 8, column: 3, scope: !7)
+!51 = !DILocation(line: 8, column: 11, scope: !7)
+!52 = distinct !DIAssignID()
+!53 = !DILocation(line: 9, column: 21, scope: !7)
+!54 = !DILocation(line: 9, column: 3, scope: !7)
+!55 = !DILocation(line: 9, column: 11, scope: !7)
+!56 = distinct !DIAssignID()
+!57 = !DILocation(line: 10, column: 21, scope: !7)
+!58 = !DILocation(line: 10, column: 3, scope: !7)
+!59 = !DILocation(line: 10, column: 11, scope: !7)
+!60 = distinct !DIAssignID()
+!61 = !DILocation(line: 11, column: 3, scope: !7)
+!62 = !DILocation(line: 12, column: 1, scope: !7)
+!63 = !DISubprogram(name: "get", linkageName: "_Z3getv", scope: !1, file: !1, line: 1, type: !64, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!64 = !DISubroutineType(types: !65)
+!65 = !{!10}
+!66 = !DISubprogram(name: "ext", linkageName: "_Z3extPf", scope: !1, file: !1, line: 2, type: !67, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!67 = !DISubroutineType(types: !68)
+!68 = !{!10, !69}
+!69 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !10, size: 64)