[llvm] 69effe0 - [SLP]Check for potential safety of the truncation for vectorized scalars with multi uses

[Alexey Bataev via llvm-commits]

Author: Alexey Bataev
Date: 2025-02-27T08:41:46-08:00
New Revision: 69effe054c136defda8766688ac0de4626a0eb05

URL: https://github.com/llvm/llvm-project/commit/69effe054c136defda8766688ac0de4626a0eb05
DIFF: https://github.com/llvm/llvm-project/commit/69effe054c136defda8766688ac0de4626a0eb05.diff

LOG: [SLP]Check for potential safety of the truncation for vectorized scalars with multi uses

If the vectorized scalars has multiple uses, need to check if it is safe
to truncate the vectorized value, before actually trying doing it.
Otherwise, the compiler may loose some important bits, which may lead to
a miscompilation.

Fixes #129057

Added: 
    

Modified: 
    llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-different-bitwidth.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index e8c91ebd508ce..2d343ca31f07b 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -18071,6 +18071,12 @@ bool BoUpSLP::collectValuesToDemote(
           (void)for_each(E.Scalars, std::bind(IsPotentiallyTruncated, _1,
                                               std::ref(BitWidth)));
         } else {
+          // Several vectorized uses? Check if we can truncate it, otherwise -
+          // exit.
+          if (any_of(E.Scalars, [&](Value *V) {
+                return !V->hasOneUse() && !IsPotentiallyTruncated(V, BitWidth);
+              }))
+            return false;
           bool NeedToExit = false;
           if (Checker && !AttemptCheckBitwidth(Checker, NeedToExit))
             return false;

diff  --git a/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-
diff erent-bitwidth.ll b/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-
diff erent-bitwidth.ll
index 227c237826c6c..55fe7d6ed52e5 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-
diff erent-bitwidth.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-
diff erent-bitwidth.ll
@@ -12,15 +12,14 @@ define i32 @test() {
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i32> <i32 poison, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080>, i32 [[ADD_I_I]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <8 x i32> <i32 83886080, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>, [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = ashr <8 x i32> [[TMP1]], splat (i32 24)
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc <8 x i32> [[TMP2]] to <8 x i8>
-; CHECK-NEXT:    [[TMP4:%.*]] = and <8 x i8> [[TMP3]], <i8 -65, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[TMP5:%.*]] = zext <8 x i8> [[TMP4]] to <8 x i32>
+; CHECK-NEXT:    [[TMP5:%.*]] = and <8 x i32> [[TMP2]], <i32 66440127, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> poison, <2 x i32> <i32 0, i32 poison>
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <2 x i32> [[TMP6]], <2 x i32> <i32 poison, i32 0>, <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp ugt <2 x i32> [[TMP7]], <i32 33554431, i32 0>
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <8 x i1> @llvm.vector.insert.v8i1.v2i1(<8 x i1> <i1 poison, i1 poison, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false>, <2 x i1> [[TMP8]], i64 0)
-; CHECK-NEXT:    [[TMP10:%.*]] = select <8 x i1> [[TMP9]], <8 x i8> zeroinitializer, <8 x i8> <i8 6, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
-; CHECK-NEXT:    [[TMP11:%.*]] = shl <8 x i8> [[TMP4]], [[TMP10]]
+; CHECK-NEXT:    [[TMP10:%.*]] = select <8 x i1> [[TMP9]], <8 x i32> zeroinitializer, <8 x i32> <i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+; CHECK-NEXT:    [[TMP13:%.*]] = shl <8 x i32> [[TMP5]], [[TMP10]]
+; CHECK-NEXT:    [[TMP11:%.*]] = trunc <8 x i32> [[TMP13]] to <8 x i8>
 ; CHECK-NEXT:    [[TMP12:%.*]] = call i8 @llvm.vector.reduce.and.v8i8(<8 x i8> [[TMP11]])
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP12]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]