94 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --filter "LV: Found an estimated cost of [0-9]+ for VF [0-9]+ For instruction:\s*store double %v6, ptr %out6"2; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize --disable-output < %s 2>&1 | FileCheck %s --check-prefix=SSE23; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize --disable-output < %s 2>&1 | FileCheck %s --check-prefix=AVX14; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize --disable-output < %s 2>&1 | FileCheck %s --check-prefix=AVX25; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize --disable-output < %s 2>&1 | FileCheck %s --check-prefix=AVX5126; REQUIRES: asserts7 8target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"9target triple = "x86_64-unknown-linux-gnu"10 11@A = global [1024 x i8] zeroinitializer, align 12812@B = global [1024 x double] zeroinitializer, align 12813 14define void @test() {15; SSE2-LABEL: 'test'16; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 817; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 818; SSE2: LV: Found an estimated cost of 22 for VF 2 For instruction: store double %v6, ptr %out6, align 819; SSE2: LV: Found an estimated cost of 44 for VF 4 For instruction: store double %v6, ptr %out6, align 820;21; AVX1-LABEL: 'test'22; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 823; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 824; AVX1: LV: Found an estimated cost of 23 for VF 2 For instruction: store double %v6, ptr %out6, align 825; AVX1: LV: Found an estimated cost of 52 for VF 4 For instruction: store double %v6, ptr %out6, align 826; AVX1: LV: Found an estimated cost of 104 for VF 8 For instruction: store double %v6, ptr %out6, align 827;28; AVX2-LABEL: 'test'29; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 830; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 831; AVX2: LV: Found an estimated cost of 23 for VF 2 For instruction: store double %v6, ptr %out6, align 832; AVX2: LV: Found an estimated cost of 52 for VF 4 For instruction: store double %v6, ptr %out6, align 833; AVX2: LV: Found an estimated cost of 104 for VF 8 For instruction: store double %v6, ptr %out6, align 834;35; AVX512-LABEL: 'test'36; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 837; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v6, ptr %out6, align 838; AVX512: LV: Found an estimated cost of 20 for VF 2 For instruction: store double %v6, ptr %out6, align 839; AVX512: LV: Found an estimated cost of 40 for VF 4 For instruction: store double %v6, ptr %out6, align 840; AVX512: LV: Found an estimated cost of 70 for VF 8 For instruction: store double %v6, ptr %out6, align 841; AVX512: LV: Found an estimated cost of 140 for VF 16 For instruction: store double %v6, ptr %out6, align 842; AVX512: LV: Found an estimated cost of 280 for VF 32 For instruction: store double %v6, ptr %out6, align 843;44entry:45 br label %for.body46 47for.body:48 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]49 50 %iv.0 = add nuw nsw i64 %iv, 051 %iv.1 = add nuw nsw i64 %iv, 152 %iv.2 = add nuw nsw i64 %iv, 253 %iv.3 = add nuw nsw i64 %iv, 354 %iv.4 = add nuw nsw i64 %iv, 455 %iv.5 = add nuw nsw i64 %iv, 556 %iv.6 = add nuw nsw i64 %iv, 657 58 %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.059 %v.narrow = load i8, ptr %in60 61 %v = uitofp i8 %v.narrow to double62 63 %v0 = fadd double %v, 0.064 %v1 = fadd double %v, 1.065 %v2 = fadd double %v, 2.066 %v3 = fadd double %v, 3.067 %v4 = fadd double %v, 4.068 %v5 = fadd double %v, 5.069 %v6 = fadd double %v, 6.070 71 %out0 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.072 %out1 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.173 %out2 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.274 %out3 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.375 %out4 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.476 %out5 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.577 %out6 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.678 79 store double %v0, ptr %out080 store double %v1, ptr %out181 store double %v2, ptr %out282 store double %v3, ptr %out383 store double %v4, ptr %out484 store double %v5, ptr %out585 store double %v6, ptr %out686 87 %iv.next = add nuw nsw i64 %iv.0, 788 %cmp = icmp ult i64 %iv.next, 102489 br i1 %cmp, label %for.body, label %for.cond.cleanup90 91for.cond.cleanup:92 ret void93}94