65 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt < %s -mtriple=riscv64-unknown-linux -mattr=+v -passes=slp-vectorizer -S | FileCheck %s3 4@src = common global [8 x double] zeroinitializer, align 645@dst = common global [4 x double] zeroinitializer, align 646 7define void @test() {8; CHECK-LABEL: @test(9; CHECK-NEXT: [[TMP4:%.*]] = load <8 x double>, ptr @src, align 810; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <8 x double> [[TMP4]], <8 x double> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>11; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <8 x double> [[TMP4]], <8 x double> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>12; CHECK-NEXT: [[TMP3:%.*]] = fsub fast <4 x double> [[TMP1]], [[TMP2]]13; CHECK-NEXT: store <4 x double> [[TMP3]], ptr @dst, align 814; CHECK-NEXT: ret void15;16 %a0 = load double, ptr @src, align 817 %a1 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 1), align 818 %a2 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 2), align 819 %a3 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 3), align 820 %a4 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 4), align 821 %a5 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 5), align 822 %a6 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 6), align 823 %a7 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 7), align 824 %res1 = fsub fast double %a0, %a125 %res2 = fsub fast double %a2, %a326 %res3 = fsub fast double %a4, %a527 %res4 = fsub fast double %a6, %a728 store double %res1, ptr @dst, align 829 store double %res2, ptr getelementptr inbounds ([8 x double], ptr @dst, i32 0, i64 1), align 830 store double %res3, ptr getelementptr inbounds ([8 x double], ptr @dst, i32 0, i64 2), align 831 store double %res4, ptr getelementptr inbounds ([8 x double], ptr @dst, i32 0, i64 3), align 832 ret void33}34 35; Same as above, but %a7 is also used as a scalar and must be extracted from36; the wide load. (Or in this case, kept as a scalar load).37define double @test_with_extract() {38; CHECK-LABEL: @test_with_extract(39; CHECK-NEXT: [[TMP1:%.*]] = load <8 x double>, ptr @src, align 840; CHECK-NEXT: [[A7:%.*]] = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 7), align 841; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <8 x double> [[TMP1]], <8 x double> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>42; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <8 x double> [[TMP1]], <8 x double> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>43; CHECK-NEXT: [[TMP4:%.*]] = fsub fast <4 x double> [[TMP2]], [[TMP3]]44; CHECK-NEXT: store <4 x double> [[TMP4]], ptr @dst, align 845; CHECK-NEXT: ret double [[A7]]46;47 %a0 = load double, ptr @src, align 848 %a1 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 1), align 849 %a2 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 2), align 850 %a3 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 3), align 851 %a4 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 4), align 852 %a5 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 5), align 853 %a6 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 6), align 854 %a7 = load double, ptr getelementptr inbounds ([8 x double], ptr @src, i32 0, i64 7), align 855 %res1 = fsub fast double %a0, %a156 %res2 = fsub fast double %a2, %a357 %res3 = fsub fast double %a4, %a558 %res4 = fsub fast double %a6, %a759 store double %res1, ptr @dst, align 860 store double %res2, ptr getelementptr inbounds ([8 x double], ptr @dst, i32 0, i64 1), align 861 store double %res3, ptr getelementptr inbounds ([8 x double], ptr @dst, i32 0, i64 2), align 862 store double %res4, ptr getelementptr inbounds ([8 x double], ptr @dst, i32 0, i64 3), align 863 ret double %a764}65