Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / 381b4fb0abe02ffa795529b1b88237b79edd5447 / . / llvm / test / Analysis / CostModel / X86 / reduce-add.ll
blob: fedb0b31c3c0da4417ef83b458373ab8b8daa404 [file] [log] [blame]
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
Craig Topper381b4fb2018-12-05 07:56:50 +0000[diff] [blame^]2; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
3; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+ssse3 | FileCheck %s --check-prefixes=CHECK,SSE,SSSE3
4; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
5; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
6; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
7; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
8; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
9; RUN: opt < %s -cost-model -mtriple=x86_64-apple-darwin -analyze -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512DQ
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]10
11define i32 @reduce_i64(i32 %arg) {
12; SSE2-LABEL: 'reduce_i64'
13; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]14; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64> undef)
15; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64> undef)
16; SSE2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64> undef)
17; SSE2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]18; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
19;
20; SSSE3-LABEL: 'reduce_i64'
21; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]22; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64> undef)
23; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64> undef)
24; SSSE3-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64> undef)
25; SSSE3-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]26; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
27;
28; SSE42-LABEL: 'reduce_i64'
29; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64> undef)
30; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64> undef)
31; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64> undef)
32; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64> undef)
33; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
34; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
35;
36; AVX-LABEL: 'reduce_i64'
37; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64> undef)
38; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64> undef)
39; AVX-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64> undef)
40; AVX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64> undef)
41; AVX-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
42; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
43;
44; AVX512-LABEL: 'reduce_i64'
45; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64> undef)
46; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64> undef)
47; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]48; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64> undef)
49; AVX512-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]50; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
51;
52 %V1 = call i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64> undef)
53 %V2 = call i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64> undef)
54 %V4 = call i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64> undef)
55 %V8 = call i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64> undef)
56 %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
57 ret i32 undef
58}
59
60define i32 @reduce_i32(i32 %arg) {
61; SSE2-LABEL: 'reduce_i32'
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]62; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32> undef)
63; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> undef)
64; SSE2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32> undef)
65; SSE2-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
66; SSE2-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]67; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
68;
69; SSSE3-LABEL: 'reduce_i32'
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]70; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32> undef)
71; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> undef)
72; SSSE3-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32> undef)
73; SSSE3-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
74; SSSE3-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]75; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
76;
77; SSE42-LABEL: 'reduce_i32'
78; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32> undef)
79; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> undef)
80; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32> undef)
81; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
82; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
83; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
84;
85; AVX-LABEL: 'reduce_i32'
86; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32> undef)
87; AVX-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> undef)
88; AVX-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32> undef)
89; AVX-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
90; AVX-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
91; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
92;
93; AVX512-LABEL: 'reduce_i32'
94; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32> undef)
95; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> undef)
96; AVX512-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]97; AVX512-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
98; AVX512-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]99; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
100;
101 %V2 = call i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32> undef)
102 %V4 = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> undef)
103 %V8 = call i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32> undef)
104 %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
105 %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
106 ret i32 undef
107}
108
109define i32 @reduce_i16(i32 %arg) {
110; SSE2-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]111; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]112; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
113; SSE2-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
114; SSE2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
115; SSE2-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
116; SSE2-NEXT: Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]117; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
118;
119; SSSE3-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]120; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]121; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
122; SSSE3-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
123; SSSE3-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
124; SSSE3-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
125; SSSE3-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]126; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
127;
128; SSE42-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]129; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]130; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
131; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
132; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
133; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
134; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
135; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
136;
137; AVX1-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]138; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]139; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
140; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]141; AVX1-NEXT: Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
142; AVX1-NEXT: Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
143; AVX1-NEXT: Cost Model: Found an estimated cost of 73 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]144; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
145;
146; AVX2-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]147; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]148; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
149; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]150; AVX2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
151; AVX2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
152; AVX2-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]153; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
154;
155; AVX512F-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]156; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]157; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
158; AVX512F-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]159; AVX512F-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
160; AVX512F-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
161; AVX512F-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]162; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
163;
164; AVX512BW-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]165; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]166; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
167; AVX512BW-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]168; AVX512BW-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
169; AVX512BW-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
170; AVX512BW-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]171; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
172;
173; AVX512DQ-LABEL: 'reduce_i16'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]174; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]175; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
176; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]177; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
178; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
179; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]180; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
181;
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]182 %V2 = call i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]183 %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
184 %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
185 %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
186 %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
187 %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
188 ret i32 undef
189}
190
191define i32 @reduce_i8(i32 %arg) {
192; SSE2-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]193; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
194; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]195; SSE2-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
196; SSE2-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
197; SSE2-NEXT: Cost Model: Found an estimated cost of 47 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
198; SSE2-NEXT: Cost Model: Found an estimated cost of 51 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
199; SSE2-NEXT: Cost Model: Found an estimated cost of 59 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]200; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
201;
202; SSSE3-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]203; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
204; SSSE3-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]205; SSSE3-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
206; SSSE3-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
207; SSSE3-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
208; SSSE3-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
209; SSSE3-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]210; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
211;
212; SSE42-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]213; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
214; SSE42-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]215; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]216; SSE42-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
217; SSE42-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
218; SSE42-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
219; SSE42-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]220; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
221;
222; AVX1-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]223; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
224; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]225; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]226; AVX1-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
227; AVX1-NEXT: Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
228; AVX1-NEXT: Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
229; AVX1-NEXT: Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]230; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
231;
232; AVX2-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]233; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
234; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]235; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]236; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
237; AVX2-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
238; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
239; AVX2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]240; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
241;
242; AVX512F-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]243; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
244; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]245; AVX512F-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]246; AVX512F-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
247; AVX512F-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
248; AVX512F-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
249; AVX512F-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]250; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
251;
252; AVX512BW-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]253; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
254; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]255; AVX512BW-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]256; AVX512BW-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
257; AVX512BW-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
258; AVX512BW-NEXT: Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
259; AVX512BW-NEXT: Cost Model: Found an estimated cost of 57 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]260; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
261;
262; AVX512DQ-LABEL: 'reduce_i8'
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]263; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
264; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]265; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
Simon Pilgrim102854f2018-12-01 14:18:31 +0000[diff] [blame]266; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
267; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
268; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
269; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]270; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
271;
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]272 %V2 = call i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8> undef)
273 %V4 = call i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8> undef)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]274 %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
275 %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
276 %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
277 %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
278 %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
279 ret i32 undef
280}
281
282declare i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64>)
283declare i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64>)
284declare i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64>)
285declare i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64>)
286declare i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64>)
287
288declare i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32>)
289declare i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32>)
290declare i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32>)
291declare i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32>)
292declare i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32>)
293
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]294declare i16 @llvm.experimental.vector.reduce.add.i16.v2i16(<2 x i16>)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]295declare i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16>)
296declare i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16>)
297declare i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16>)
298declare i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16>)
299declare i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16>)
300
Craig Topperb4719e52018-12-05 07:26:57 +0000[diff] [blame]301declare i8 @llvm.experimental.vector.reduce.add.i8.v2i8(<2 x i8>)
302declare i8 @llvm.experimental.vector.reduce.add.i8.v4i8(<4 x i8>)
Simon Pilgrime612ab02018-10-20 14:29:59 +0000[diff] [blame]303declare i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8>)
304declare i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8>)
305declare i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8>)
306declare i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8>)
307declare i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8>)