Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / e612ab0c80116b5ca9eb1710429cb570f62439dd / . / llvm / test / Analysis / CostModel / X86 / reduce-add.ll
blob: 046aaf04e338d3bc9f03bc07e76de58235d8037f [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
2; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
3; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+ssse3 | FileCheck %s --check-prefixes=CHECK,SSE,SSSE3
4; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
5; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
6; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
7; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
8; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
9; RUN: opt < %s -cost-model -costmodel-reduxcost=true -mtriple=x86_64-apple-darwin -analyze -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512DQ
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)
14; SSE2-NEXT: Cost Model: Found an estimated cost of 4 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 7 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 12 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 21 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
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)
22; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 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 7 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 12 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 21 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
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)
48; AVX512-NEXT: Cost Model: Found an estimated cost of 14 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 17 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64> undef)
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'
62; SSE2-NEXT: Cost Model: Found an estimated cost of 4 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 8 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 11 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 16 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 25 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
67; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
68;
69; SSSE3-LABEL: 'reduce_i32'
70; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 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 8 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 11 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 16 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 25 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
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)
97; AVX512-NEXT: Cost Model: Found an estimated cost of 24 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 27 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
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'
111; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
112; SSE2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
113; SSE2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
114; SSE2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
115; SSE2-NEXT: Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
116; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
117;
118; SSSE3-LABEL: 'reduce_i16'
119; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
120; SSSE3-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
121; SSSE3-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
122; SSSE3-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
123; SSSE3-NEXT: Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
124; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
125;
126; SSE42-LABEL: 'reduce_i16'
127; 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)
128; 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)
129; 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)
130; 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)
131; 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)
132; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
133;
134; AVX1-LABEL: 'reduce_i16'
135; 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)
136; 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)
137; AVX1-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
138; AVX1-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
139; AVX1-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
140; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
141;
142; AVX2-LABEL: 'reduce_i16'
143; 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)
144; 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)
145; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
146; AVX2-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
147; AVX2-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)
148; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
149;
150; AVX512F-LABEL: 'reduce_i16'
151; 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)
152; 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)
153; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
154; AVX512F-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
155; AVX512F-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)
156; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
157;
158; AVX512BW-LABEL: 'reduce_i16'
159; 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)
160; 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)
161; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
162; AVX512BW-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
163; AVX512BW-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
164; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
165;
166; AVX512DQ-LABEL: 'reduce_i16'
167; 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)
168; 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)
169; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
170; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
171; AVX512DQ-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)
172; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
173;
174 %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
175 %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
176 %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
177 %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
178 %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
179 ret i32 undef
180}
181
182define i32 @reduce_i8(i32 %arg) {
183; SSE2-LABEL: 'reduce_i8'
184; SSE2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
185; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
186; SSE2-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
187; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
188; SSE2-NEXT: Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
189; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
190;
191; SSSE3-LABEL: 'reduce_i8'
192; SSSE3-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
193; SSSE3-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
194; SSSE3-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
195; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
196; SSSE3-NEXT: Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
197; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
198;
199; SSE42-LABEL: 'reduce_i8'
200; 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)
201; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
202; SSE42-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
203; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
204; SSE42-NEXT: Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
205; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
206;
207; AVX1-LABEL: 'reduce_i8'
208; 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)
209; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
210; AVX1-NEXT: Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
211; AVX1-NEXT: Cost Model: Found an estimated cost of 66 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
212; AVX1-NEXT: Cost Model: Found an estimated cost of 83 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
213; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
214;
215; AVX2-LABEL: 'reduce_i8'
216; 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)
217; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
218; AVX2-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
219; AVX2-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
220; AVX2-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
221; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
222;
223; AVX512F-LABEL: 'reduce_i8'
224; 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)
225; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
226; AVX512F-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
227; AVX512F-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
228; AVX512F-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
229; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
230;
231; AVX512BW-LABEL: 'reduce_i8'
232; 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)
233; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
234; AVX512BW-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
235; AVX512BW-NEXT: Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
236; AVX512BW-NEXT: Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
237; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
238;
239; AVX512DQ-LABEL: 'reduce_i8'
240; 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)
241; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
242; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
243; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
244; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
245; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
246;
247 %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
248 %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
249 %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
250 %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
251 %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
252 ret i32 undef
253}
254
255declare i64 @llvm.experimental.vector.reduce.add.i64.v1i64(<1 x i64>)
256declare i64 @llvm.experimental.vector.reduce.add.i64.v2i64(<2 x i64>)
257declare i64 @llvm.experimental.vector.reduce.add.i64.v4i64(<4 x i64>)
258declare i64 @llvm.experimental.vector.reduce.add.i64.v8i64(<8 x i64>)
259declare i64 @llvm.experimental.vector.reduce.add.i64.v16i64(<16 x i64>)
260
261declare i32 @llvm.experimental.vector.reduce.add.i32.v2i32(<2 x i32>)
262declare i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32>)
263declare i32 @llvm.experimental.vector.reduce.add.i32.v8i32(<8 x i32>)
264declare i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32>)
265declare i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32>)
266
267declare i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16>)
268declare i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16>)
269declare i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16>)
270declare i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16>)
271declare i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16>)
272
273declare i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8>)
274declare i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8>)
275declare i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8>)
276declare i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8>)
277declare i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8>)