106 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 i8 %v5, ptr %out5"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=AVX512DQ6; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize --disable-output < %s 2>&1 | FileCheck %s --check-prefix=AVX512BW7; REQUIRES: asserts8 9target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"10target triple = "x86_64-unknown-linux-gnu"11 12@A = global [1024 x i8] zeroinitializer, align 12813@B = global [1024 x i8] zeroinitializer, align 12814 15define void @test() {16; SSE2-LABEL: 'test'17; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 118; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 119; SSE2: LV: Found an estimated cost of 49 for VF 2 For instruction: store i8 %v5, ptr %out5, align 120; SSE2: LV: Found an estimated cost of 98 for VF 4 For instruction: store i8 %v5, ptr %out5, align 121; SSE2: LV: Found an estimated cost of 201 for VF 8 For instruction: store i8 %v5, ptr %out5, align 122; SSE2: LV: Found an estimated cost of 408 for VF 16 For instruction: store i8 %v5, ptr %out5, align 123;24; AVX1-LABEL: 'test'25; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 126; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 127; AVX1: LV: Found an estimated cost of 27 for VF 2 For instruction: store i8 %v5, ptr %out5, align 128; AVX1: LV: Found an estimated cost of 53 for VF 4 For instruction: store i8 %v5, ptr %out5, align 129; AVX1: LV: Found an estimated cost of 100 for VF 8 For instruction: store i8 %v5, ptr %out5, align 130; AVX1: LV: Found an estimated cost of 198 for VF 16 For instruction: store i8 %v5, ptr %out5, align 131; AVX1: LV: Found an estimated cost of 402 for VF 32 For instruction: store i8 %v5, ptr %out5, align 132;33; AVX2-LABEL: 'test'34; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 135; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 136; AVX2: LV: Found an estimated cost of 10 for VF 2 For instruction: store i8 %v5, ptr %out5, align 137; AVX2: LV: Found an estimated cost of 12 for VF 4 For instruction: store i8 %v5, ptr %out5, align 138; AVX2: LV: Found an estimated cost of 18 for VF 8 For instruction: store i8 %v5, ptr %out5, align 139; AVX2: LV: Found an estimated cost of 30 for VF 16 For instruction: store i8 %v5, ptr %out5, align 140; AVX2: LV: Found an estimated cost of 96 for VF 32 For instruction: store i8 %v5, ptr %out5, align 141;42; AVX512DQ-LABEL: 'test'43; AVX512DQ: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 144; AVX512DQ: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 145; AVX512DQ: LV: Found an estimated cost of 10 for VF 2 For instruction: store i8 %v5, ptr %out5, align 146; AVX512DQ: LV: Found an estimated cost of 12 for VF 4 For instruction: store i8 %v5, ptr %out5, align 147; AVX512DQ: LV: Found an estimated cost of 19 for VF 8 For instruction: store i8 %v5, ptr %out5, align 148; AVX512DQ: LV: Found an estimated cost of 29 for VF 16 For instruction: store i8 %v5, ptr %out5, align 149; AVX512DQ: LV: Found an estimated cost of 93 for VF 32 For instruction: store i8 %v5, ptr %out5, align 150; AVX512DQ: LV: Found an estimated cost of 810 for VF 64 For instruction: store i8 %v5, ptr %out5, align 151;52; AVX512BW-LABEL: 'test'53; AVX512BW: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 154; AVX512BW: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, ptr %out5, align 155; AVX512BW: LV: Found an estimated cost of 18 for VF 2 For instruction: store i8 %v5, ptr %out5, align 156; AVX512BW: LV: Found an estimated cost of 38 for VF 4 For instruction: store i8 %v5, ptr %out5, align 157; AVX512BW: LV: Found an estimated cost of 98 for VF 8 For instruction: store i8 %v5, ptr %out5, align 158; AVX512BW: LV: Found an estimated cost of 197 for VF 16 For instruction: store i8 %v5, ptr %out5, align 159; AVX512BW: LV: Found an estimated cost of 295 for VF 32 For instruction: store i8 %v5, ptr %out5, align 160; AVX512BW: LV: Found an estimated cost of 591 for VF 64 For instruction: store i8 %v5, ptr %out5, align 161;62entry:63 br label %for.body64 65for.body:66 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]67 68 %iv.0 = add nuw nsw i64 %iv, 069 %iv.1 = add nuw nsw i64 %iv, 170 %iv.2 = add nuw nsw i64 %iv, 271 %iv.3 = add nuw nsw i64 %iv, 372 %iv.4 = add nuw nsw i64 %iv, 473 %iv.5 = add nuw nsw i64 %iv, 574 75 %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.076 %v = load i8, ptr %in77 78 %v0 = add i8 %v, 079 %v1 = add i8 %v, 180 %v2 = add i8 %v, 281 %v3 = add i8 %v, 382 %v4 = add i8 %v, 483 %v5 = add i8 %v, 584 85 %out0 = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.086 %out1 = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.187 %out2 = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.288 %out3 = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.389 %out4 = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.490 %out5 = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.591 92 store i8 %v0, ptr %out093 store i8 %v1, ptr %out194 store i8 %v2, ptr %out295 store i8 %v3, ptr %out396 store i8 %v4, ptr %out497 store i8 %v5, ptr %out598 99 %iv.next = add nuw nsw i64 %iv.0, 6100 %cmp = icmp ult i64 %iv.next, 1024101 br i1 %cmp, label %for.body, label %for.cond.cleanup102 103for.cond.cleanup:104 ret void105}106