[SLP][NFC]Add a test with @llvm.abs nodes, which can be analyzed for

better bitwidth.

Author: alexey-bataev

llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll

@@ -0,0 +1,70 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -S -mtriple=x86_64-unknown -mattr=+avx512vl -passes=slp-vectorizer -slp-threshold=-3 | FileCheck %s
+
+
+define i32 @test(ptr noalias %in, ptr noalias %inn, ptr %out) {
+; CHECK-LABEL: @test(
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i8>, ptr [[IN:%.*]], align 1
+; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr inbounds i8, ptr [[IN]], i64 2
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i8>, ptr [[GEP_2]], align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i8>, ptr [[INN:%.*]], align 1
+; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds i8, ptr [[INN]], i64 2
+; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i8>, ptr [[GEP_5]], align 1
+; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i8> [[TMP3]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i8> [[TMP2]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i8> [[TMP5]], <4 x i8> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP8:%.*]] = sext <4 x i8> [[TMP7]] to <4 x i32>
+; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x i8> [[TMP1]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x i8> [[TMP4]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i8> [[TMP9]], <4 x i8> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP12:%.*]] = sext <4 x i8> [[TMP11]] to <4 x i32>
+; CHECK-NEXT:    [[TMP13:%.*]] = sub <4 x i32> [[TMP12]], [[TMP8]]
+; CHECK-NEXT:    [[TMP14:%.*]] = call <4 x i32> @llvm.abs.v4i32(<4 x i32> [[TMP13]], i1 true)
+; CHECK-NEXT:    [[TMP15:%.*]] = trunc <4 x i32> [[TMP14]] to <4 x i16>
+; CHECK-NEXT:    store <4 x i16> [[TMP15]], ptr [[OUT:%.*]], align 2
+; CHECK-NEXT:    ret i32 undef
+;
+  %load.1 = load i8, ptr %in, align 1
+  %gep.1 = getelementptr inbounds i8, ptr %in, i64 1
+  %load.2 = load i8, ptr %gep.1, align 1
+  %gep.2 = getelementptr inbounds i8, ptr %in, i64 2
+  %load.3 = load i8, ptr %gep.2, align 1
+  %gep.3 = getelementptr inbounds i8, ptr %in, i64 3
+  %load.4 = load i8, ptr %gep.3, align 1
+  %load.5 = load i8, ptr %inn, align 1
+  %gep.4 = getelementptr inbounds i8, ptr %inn, i64 1
+  %load.6 = load i8, ptr %gep.4, align 1
+  %gep.5 = getelementptr inbounds i8, ptr %inn, i64 2
+  %load.7 = load i8, ptr %gep.5, align 1
+  %gep.6 = getelementptr inbounds i8, ptr %inn, i64 3
+  %load.8 = load i8, ptr %gep.6, align 1
+  %sext1 = sext i8 %load.1 to i32
+  %sext2 = sext i8 %load.2 to i32
+  %sext3 = sext i8 %load.3 to i32
+  %sext4 = sext i8 %load.4 to i32
+  %sext5 = sext i8 %load.5 to i32
+  %sext6 = sext i8 %load.6 to i32
+  %sext7 = sext i8 %load.7 to i32
+  %sext8 = sext i8 %load.8 to i32
+  %sub1 = sub i32 %sext1, %sext5
+  %sub2 = sub i32 %sext2, %sext6
+  %sub3 = sub i32 %sext7, %sext3
+  %sub4 = sub i32 %sext8, %sext4
+  %call1 = call i32 @llvm.abs(i32 %sub1, i1 true)
+  %call2 = call i32 @llvm.abs(i32 %sub2, i1 true)
+  %call3 = call i32 @llvm.abs(i32 %sub3, i1 true)
+  %call4 = call i32 @llvm.abs(i32 %sub4, i1 true)
+  %t1 = trunc i32 %call1 to i16
+  %t2 = trunc i32 %call2 to i16
+  %t3 = trunc i32 %call3 to i16
+  %t4 = trunc i32 %call4 to i16
+  %gep.8 = getelementptr inbounds i16, ptr %out, i64 1
+  %gep.9 = getelementptr inbounds i16, ptr %out, i64 2
+  %gep.10 = getelementptr inbounds i16, ptr %out, i64 3
+  store i16 %t1, ptr %out, align 2
+  store i16 %t2, ptr %gep.8, align 2
+  store i16 %t3, ptr %gep.9, align 2
+  store i16 %t4, ptr %gep.10, align 2
+
+  ret i32 undef
+}