|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py |
| 2 | +; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=SSE |
| 3 | +; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx | FileCheck %s --check-prefixes=AVX |
| 4 | +; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512 |
| 5 | + |
| 6 | +declare i1 @llvm.isnan.f16(half) |
| 7 | +declare <4 x i1> @llvm.isnan.v4f16(<4 x half>) |
| 8 | +declare <8 x i1> @llvm.isnan.v8f16(<8 x half>) |
| 9 | +declare <16 x i1> @llvm.isnan.v16f16(<16 x half>) |
| 10 | +declare <32 x i1> @llvm.isnan.v32f16(<32 x half>) |
| 11 | + |
| 12 | +declare i1 @llvm.isnan.f32(float) |
| 13 | +declare <2 x i1> @llvm.isnan.v2f32(<2 x float>) |
| 14 | +declare <4 x i1> @llvm.isnan.v4f32(<4 x float>) |
| 15 | +declare <8 x i1> @llvm.isnan.v8f32(<8 x float>) |
| 16 | +declare <16 x i1> @llvm.isnan.v16f32(<16 x float>) |
| 17 | + |
| 18 | +declare i1 @llvm.isnan.f64(double) |
| 19 | +declare <2 x i1> @llvm.isnan.v2f64(<2 x double>) |
| 20 | +declare <4 x i1> @llvm.isnan.v4f64(<4 x double>) |
| 21 | +declare <8 x i1> @llvm.isnan.v8f64(<8 x double>) |
| 22 | +declare <16 x i1> @llvm.isnan.v16f64(<16 x double>) |
| 23 | + |
| 24 | +define i32 @isnan(i32 %arg) { |
| 25 | +; SSE-LABEL: 'isnan' |
| 26 | +; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F16 = call i1 @llvm.isnan.f16(half undef) |
| 27 | +; SSE-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V4F16 = call <4 x i1> @llvm.isnan.v4f16(<4 x half> undef) |
| 28 | +; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x i1> @llvm.isnan.v8f16(<8 x half> undef) |
| 29 | +; SSE-NEXT: Cost Model: Found an estimated cost of 47 for instruction: %V16F16 = call <16 x i1> @llvm.isnan.v16f16(<16 x half> undef) |
| 30 | +; SSE-NEXT: Cost Model: Found an estimated cost of 94 for instruction: %V3216 = call <32 x i1> @llvm.isnan.v32f16(<32 x half> undef) |
| 31 | +; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call i1 @llvm.isnan.f32(float undef) |
| 32 | +; SSE-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2F32 = call <2 x i1> @llvm.isnan.v2f32(<2 x float> undef) |
| 33 | +; SSE-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V4F32 = call <4 x i1> @llvm.isnan.v4f32(<4 x float> undef) |
| 34 | +; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F32 = call <8 x i1> @llvm.isnan.v8f32(<8 x float> undef) |
| 35 | +; SSE-NEXT: Cost Model: Found an estimated cost of 47 for instruction: %V16F32 = call <16 x i1> @llvm.isnan.v16f32(<16 x float> undef) |
| 36 | +; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call i1 @llvm.isnan.f64(double undef) |
| 37 | +; SSE-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V2F64 = call <2 x i1> @llvm.isnan.v2f64(<2 x double> undef) |
| 38 | +; SSE-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V4F64 = call <4 x i1> @llvm.isnan.v4f64(<4 x double> undef) |
| 39 | +; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x i1> @llvm.isnan.v8f64(<8 x double> undef) |
| 40 | +; SSE-NEXT: Cost Model: Found an estimated cost of 47 for instruction: %V16F64 = call <16 x i1> @llvm.isnan.v16f64(<16 x double> undef) |
| 41 | +; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef |
| 42 | +; |
| 43 | +; AVX-LABEL: 'isnan' |
| 44 | +; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F16 = call i1 @llvm.isnan.f16(half undef) |
| 45 | +; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = call <4 x i1> @llvm.isnan.v4f16(<4 x half> undef) |
| 46 | +; AVX-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x i1> @llvm.isnan.v8f16(<8 x half> undef) |
| 47 | +; AVX-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = call <16 x i1> @llvm.isnan.v16f16(<16 x half> undef) |
| 48 | +; AVX-NEXT: Cost Model: Found an estimated cost of 66 for instruction: %V3216 = call <32 x i1> @llvm.isnan.v32f16(<32 x half> undef) |
| 49 | +; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call i1 @llvm.isnan.f32(float undef) |
| 50 | +; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F32 = call <2 x i1> @llvm.isnan.v2f32(<2 x float> undef) |
| 51 | +; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F32 = call <4 x i1> @llvm.isnan.v4f32(<4 x float> undef) |
| 52 | +; AVX-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F32 = call <8 x i1> @llvm.isnan.v8f32(<8 x float> undef) |
| 53 | +; AVX-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16F32 = call <16 x i1> @llvm.isnan.v16f32(<16 x float> undef) |
| 54 | +; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call i1 @llvm.isnan.f64(double undef) |
| 55 | +; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = call <2 x i1> @llvm.isnan.v2f64(<2 x double> undef) |
| 56 | +; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = call <4 x i1> @llvm.isnan.v4f64(<4 x double> undef) |
| 57 | +; AVX-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x i1> @llvm.isnan.v8f64(<8 x double> undef) |
| 58 | +; AVX-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16F64 = call <16 x i1> @llvm.isnan.v16f64(<16 x double> undef) |
| 59 | +; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef |
| 60 | +; |
| 61 | +; AVX512-LABEL: 'isnan' |
| 62 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F16 = call i1 @llvm.isnan.f16(half undef) |
| 63 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = call <4 x i1> @llvm.isnan.v4f16(<4 x half> undef) |
| 64 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x i1> @llvm.isnan.v8f16(<8 x half> undef) |
| 65 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = call <16 x i1> @llvm.isnan.v16f16(<16 x half> undef) |
| 66 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V3216 = call <32 x i1> @llvm.isnan.v32f16(<32 x half> undef) |
| 67 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call i1 @llvm.isnan.f32(float undef) |
| 68 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F32 = call <2 x i1> @llvm.isnan.v2f32(<2 x float> undef) |
| 69 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F32 = call <4 x i1> @llvm.isnan.v4f32(<4 x float> undef) |
| 70 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F32 = call <8 x i1> @llvm.isnan.v8f32(<8 x float> undef) |
| 71 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16F32 = call <16 x i1> @llvm.isnan.v16f32(<16 x float> undef) |
| 72 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call i1 @llvm.isnan.f64(double undef) |
| 73 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = call <2 x i1> @llvm.isnan.v2f64(<2 x double> undef) |
| 74 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = call <4 x i1> @llvm.isnan.v4f64(<4 x double> undef) |
| 75 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x i1> @llvm.isnan.v8f64(<8 x double> undef) |
| 76 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16F64 = call <16 x i1> @llvm.isnan.v16f64(<16 x double> undef) |
| 77 | +; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef |
| 78 | +; |
| 79 | + %F16 = call i1 @llvm.isnan.f16(half undef) |
| 80 | + %V4F16 = call <4 x i1> @llvm.isnan.v4f16(<4 x half> undef) |
| 81 | + %V8F16 = call <8 x i1> @llvm.isnan.v8f16(<8 x half> undef) |
| 82 | + %V16F16 = call <16 x i1> @llvm.isnan.v16f16(<16 x half> undef) |
| 83 | + %V3216 = call <32 x i1> @llvm.isnan.v32f16(<32 x half> undef) |
| 84 | + |
| 85 | + %F32 = call i1 @llvm.isnan.f32(float undef) |
| 86 | + %V2F32 = call <2 x i1> @llvm.isnan.v2f32(<2 x float> undef) |
| 87 | + %V4F32 = call <4 x i1> @llvm.isnan.v4f32(<4 x float> undef) |
| 88 | + %V8F32 = call <8 x i1> @llvm.isnan.v8f32(<8 x float> undef) |
| 89 | + %V16F32 = call <16 x i1> @llvm.isnan.v16f32(<16 x float> undef) |
| 90 | + |
| 91 | + %F64 = call i1 @llvm.isnan.f64(double undef) |
| 92 | + %V2F64 = call <2 x i1> @llvm.isnan.v2f64(<2 x double> undef) |
| 93 | + %V4F64 = call <4 x i1> @llvm.isnan.v4f64(<4 x double> undef) |
| 94 | + %V8F64 = call <8 x i1> @llvm.isnan.v8f64(<8 x double> undef) |
| 95 | + %V16F64 = call <16 x i1> @llvm.isnan.v16f64(<16 x double> undef) |
| 96 | + |
| 97 | + ret i32 undef |
| 98 | +} |
0 commit comments