[llvm] 96cded5 - [SLP] Incorrect handling of external scalar values

[Evgeniy Brevnov via llvm-commits]

Author: Evgeniy Brevnov
Date: 2021-06-16T13:27:36+07:00
New Revision: 96cded5b796f405d91c4b403078fbb13f13a056b

URL: https://github.com/llvm/llvm-project/commit/96cded5b796f405d91c4b403078fbb13f13a056b
DIFF: https://github.com/llvm/llvm-project/commit/96cded5b796f405d91c4b403078fbb13f13a056b.diff

LOG: [SLP] Incorrect handling of external scalar values

Reviewed By: ABataev

Differential Revision: https://reviews.llvm.org/D103954

Added: 
    llvm/test/Transforms/SLPVectorizer/slp-hr-with-reuse.ll

Modified: 
    llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 46a1b00e3e01f..41b88f1caa4b9 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -5713,8 +5713,11 @@ BoUpSLP::vectorizeTree(ExtraValueToDebugLocsMap &ExternallyUsedValues) {
       }
       Value *NewInst = ExtractAndExtendIfNeeded(Vec);
       CSEBlocks.insert(cast<Instruction>(Scalar)->getParent());
-      auto &Locs = ExternallyUsedValues[Scalar];
-      ExternallyUsedValues.insert({NewInst, Locs});
+      auto &NewInstLocs = ExternallyUsedValues[NewInst];
+      auto It = ExternallyUsedValues.find(Scalar);
+      assert(It != ExternallyUsedValues.end() &&
+             "Externally used scalar is not found in ExternallyUsedValues");
+      NewInstLocs.append(It->second);
       ExternallyUsedValues.erase(Scalar);
       // Required to update internally referenced instructions.
       Scalar->replaceAllUsesWith(NewInst);

diff  --git a/llvm/test/Transforms/SLPVectorizer/slp-hr-with-reuse.ll b/llvm/test/Transforms/SLPVectorizer/slp-hr-with-reuse.ll
new file mode 100644
index 0000000000000..e19c75f367b22
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/slp-hr-with-reuse.ll
@@ -0,0 +1,44 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt --passes=slp-vectorizer,instcombine -slp-threshold=-1000000 -S < %s | FileCheck %s
+
+define i32 @foo() {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    ret i32 -162
+;
+entry:
+  %i = xor i32 4, -9
+  %i1 = xor i32 5, %i
+  %i2 = xor i32 6, %i1
+  %i3 = add i32 %i2, 0
+  %i4 = add i32 -9, %i3
+  %i5 = xor i32 8, -9
+  %i6 = add i32 %i5, %i4
+  %i7 = xor i32 9, %i5
+  %i8 = add i32 %i7, %i6
+  %i9 = xor i32 10, %i7
+  %i10 = add i32 %i9, %i8
+  %i11 = add i32 -9, %i10
+  %i12 = xor i32 12, -9
+  %i13 = add i32 %i12, %i11
+  %i14 = xor i32 13, %i12
+  %i15 = add i32 %i14, %i13
+  %i16 = xor i32 14, %i14
+  %i17 = add i32 %i16, %i15
+  %i18 = add i32 -9, %i17
+  %i19 = xor i32 16, -9
+  %i20 = add i32 %i19, %i18
+  %i21 = xor i32 17, %i19
+  %i22 = add i32 %i21, %i20
+  %i23 = xor i32 18, %i21
+  %i24 = add i32 %i23, %i22
+  %i25 = add i32 -9, %i24
+  %i26 = add i32 0, %i25
+  %i27 = add i32 0, %i26
+  %i28 = add i32 0, %i27
+  %i29 = add i32 -9, %i28
+  %i30 = add i32 0, %i29
+  %i31 = add i32 0, %i30
+  %i32 = add i32 0, %i31
+  ret i32 %i32
+}