116 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 i16 %v7, ptr %out7"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 i16] 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 i16 %v7, ptr %out7, align 218; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 219; SSE2: LV: Found an estimated cost of 34 for VF 2 For instruction: store i16 %v7, ptr %out7, align 220; SSE2: LV: Found an estimated cost of 68 for VF 4 For instruction: store i16 %v7, ptr %out7, align 221; SSE2: LV: Found an estimated cost of 136 for VF 8 For instruction: store i16 %v7, ptr %out7, align 222; SSE2: LV: Found an estimated cost of 272 for VF 16 For instruction: store i16 %v7, ptr %out7, align 223;24; AVX1-LABEL: 'test'25; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 226; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 227; AVX1: LV: Found an estimated cost of 34 for VF 2 For instruction: store i16 %v7, ptr %out7, align 228; AVX1: LV: Found an estimated cost of 68 for VF 4 For instruction: store i16 %v7, ptr %out7, align 229; AVX1: LV: Found an estimated cost of 136 for VF 8 For instruction: store i16 %v7, ptr %out7, align 230; AVX1: LV: Found an estimated cost of 280 for VF 16 For instruction: store i16 %v7, ptr %out7, align 231; AVX1: LV: Found an estimated cost of 560 for VF 32 For instruction: store i16 %v7, ptr %out7, align 232;33; AVX2-LABEL: 'test'34; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 235; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 236; AVX2: LV: Found an estimated cost of 34 for VF 2 For instruction: store i16 %v7, ptr %out7, align 237; AVX2: LV: Found an estimated cost of 68 for VF 4 For instruction: store i16 %v7, ptr %out7, align 238; AVX2: LV: Found an estimated cost of 136 for VF 8 For instruction: store i16 %v7, ptr %out7, align 239; AVX2: LV: Found an estimated cost of 280 for VF 16 For instruction: store i16 %v7, ptr %out7, align 240; AVX2: LV: Found an estimated cost of 560 for VF 32 For instruction: store i16 %v7, ptr %out7, align 241;42; AVX512DQ-LABEL: 'test'43; AVX512DQ: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 244; AVX512DQ: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 245; AVX512DQ: LV: Found an estimated cost of 34 for VF 2 For instruction: store i16 %v7, ptr %out7, align 246; AVX512DQ: LV: Found an estimated cost of 68 for VF 4 For instruction: store i16 %v7, ptr %out7, align 247; AVX512DQ: LV: Found an estimated cost of 136 for VF 8 For instruction: store i16 %v7, ptr %out7, align 248; AVX512DQ: LV: Found an estimated cost of 280 for VF 16 For instruction: store i16 %v7, ptr %out7, align 249; AVX512DQ: LV: Found an estimated cost of 568 for VF 32 For instruction: store i16 %v7, ptr %out7, align 250; AVX512DQ: LV: Found an estimated cost of 1136 for VF 64 For instruction: store i16 %v7, ptr %out7, align 251;52; AVX512BW-LABEL: 'test'53; AVX512BW: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 254; AVX512BW: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %v7, ptr %out7, align 255; AVX512BW: LV: Found an estimated cost of 18 for VF 2 For instruction: store i16 %v7, ptr %out7, align 256; AVX512BW: LV: Found an estimated cost of 18 for VF 4 For instruction: store i16 %v7, ptr %out7, align 257; AVX512BW: LV: Found an estimated cost of 37 for VF 8 For instruction: store i16 %v7, ptr %out7, align 258; AVX512BW: LV: Found an estimated cost of 74 for VF 16 For instruction: store i16 %v7, ptr %out7, align 259; AVX512BW: LV: Found an estimated cost of 148 for VF 32 For instruction: store i16 %v7, ptr %out7, align 260; AVX512BW: LV: Found an estimated cost of 296 for VF 64 For instruction: store i16 %v7, ptr %out7, align 261;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 %iv.6 = add nuw nsw i64 %iv, 675 %iv.7 = add nuw nsw i64 %iv, 776 77 %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.078 %v.narrow = load i8, ptr %in79 80 %v = zext i8 %v.narrow to i1681 82 %v0 = add i16 %v, 083 %v1 = add i16 %v, 184 %v2 = add i16 %v, 285 %v3 = add i16 %v, 386 %v4 = add i16 %v, 487 %v5 = add i16 %v, 588 %v6 = add i16 %v, 689 %v7 = add i16 %v, 790 91 %out0 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.092 %out1 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.193 %out2 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.294 %out3 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.395 %out4 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.496 %out5 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.597 %out6 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.698 %out7 = getelementptr inbounds [1024 x i16], ptr @B, i64 0, i64 %iv.799 100 store i16 %v0, ptr %out0101 store i16 %v1, ptr %out1102 store i16 %v2, ptr %out2103 store i16 %v3, ptr %out3104 store i16 %v4, ptr %out4105 store i16 %v5, ptr %out5106 store i16 %v6, ptr %out6107 store i16 %v7, ptr %out7108 109 %iv.next = add nuw nsw i64 %iv.0, 8110 %cmp = icmp ult i64 %iv.next, 1024111 br i1 %cmp, label %for.body, label %for.cond.cleanup112 113for.cond.cleanup:114 ret void115}116