[SLP]Fix a check for multi-users for icmp user.

The compiler should not take into account the type of the cmp
instruction, otherwise it may treat the size incorrectly and it may lead
to incorrect codegen.

Author: alexey-bataev

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -14530,7 +14530,8 @@ bool BoUpSLP::collectValuesToDemote(
         return !all_of(V->users(), [=](User *U) {
           return getTreeEntry(U) ||
                  (UserIgnoreList && UserIgnoreList->contains(U)) ||
-                 (U->getType()->isSized() && !U->getType()->isScalableTy() &&
+                 (!isa<CmpInst>(U) && U->getType()->isSized() &&
+                  !U->getType()->isScalableTy() &&
                   DL->getTypeSizeInBits(U->getType()) <= BitWidth);
         }) && !IsPotentiallyTruncated(V, BitWidth);
       }))

llvm/test/Transforms/SLPVectorizer/AArch64/external-use-icmp.ll

@@ -7,23 +7,22 @@ define i16 @foo(i16 %in1, i16 %in2) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i16> poison, i16 [[IN1]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i16> [[TMP0]], <2 x i16> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP4:%.*]] = zext <2 x i16> [[TMP1]] to <2 x i64>
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i16> poison, i16 [[IN2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP4:%.*]] = mul <2 x i16> [[TMP3]], [[TMP1]]
-; CHECK-NEXT:    [[TMP5:%.*]] = and <2 x i16> [[TMP4]], <i16 -1, i16 -1>
-; CHECK-NEXT:    [[TMP6:%.*]] = zext <2 x i16> [[TMP5]] to <2 x i64>
+; CHECK-NEXT:    [[TMP5:%.*]] = zext <2 x i16> [[TMP3]] to <2 x i64>
+; CHECK-NEXT:    [[TMP9:%.*]] = mul nuw nsw <2 x i64> [[TMP5]], [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = and <2 x i64> [[TMP9]], <i64 65535, i64 65535>
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp ne <2 x i64> [[TMP6]], <i64 65533, i64 65533>
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP7]], i32 1
 ; CHECK-NEXT:    [[ZEXT3_1:%.*]] = zext i1 [[TMP8]] to i16
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i16> [[TMP4]], i32 1
-; CHECK-NEXT:    [[TMP10:%.*]] = zext i16 [[TMP9]] to i64
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i64> [[TMP9]], i32 1
 ; CHECK-NEXT:    [[CMP2_1:%.*]] = icmp ne i64 [[TMP10]], 196605
 ; CHECK-NEXT:    [[ZEXT4_1:%.*]] = zext i1 [[CMP2_1]] to i16
 ; CHECK-NEXT:    [[ADD1:%.*]] = add nuw nsw i16 [[ZEXT3_1]], [[ZEXT4_1]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <2 x i1> [[TMP7]], i32 0
 ; CHECK-NEXT:    [[ZEXT3_2:%.*]] = zext i1 [[TMP11]] to i16
-; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x i16> [[TMP4]], i32 0
-; CHECK-NEXT:    [[TMP13:%.*]] = zext i16 [[TMP12]] to i64
+; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <2 x i64> [[TMP9]], i32 0
 ; CHECK-NEXT:    [[CMP2_2:%.*]] = icmp ne i64 [[TMP13]], 196605
 ; CHECK-NEXT:    [[ZEXT4_2:%.*]] = zext i1 [[CMP2_2]] to i16
 ; CHECK-NEXT:    [[ADD2:%.*]] = add nuw nsw i16 [[ADD1]], [[ZEXT4_2]]