1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
2; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse2 | FileCheck %s -check-prefixes=CHECK,SSE,SSE2
3; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+sse4.2 | FileCheck %s -check-prefixes=CHECK,SSE,SSE42
4; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx | FileCheck %s -check-prefixes=CHECK,AVX,AVX1
5; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+avx2 | FileCheck %s -check-prefixes=CHECK,AVX,AVX2
6; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx | FileCheck %s -check-prefixes=CHECK,AVX,AVX1
7; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+xop,+avx2 | FileCheck %s -check-prefixes=CHECK,AVX,AVX2
8
9; Verify the cost of vector bswap instructions.
10
11declare <2 x i64> @llvm.bswap.v2i64(<2 x i64>)
12declare <4 x i32> @llvm.bswap.v4i32(<4 x i32>)
13declare <8 x i16> @llvm.bswap.v8i16(<8 x i16>)
14
15declare <4 x i64> @llvm.bswap.v4i64(<4 x i64>)
16declare <8 x i32> @llvm.bswap.v8i32(<8 x i32>)
17declare <16 x i16> @llvm.bswap.v16i16(<16 x i16>)
18
19define <2 x i64> @var_bswap_v2i64(<2 x i64> %a) {
20; SSE2-LABEL: 'var_bswap_v2i64'
21; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bswap = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a)
22; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bswap
23;
24; SSE42-LABEL: 'var_bswap_v2i64'
25; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a)
26; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bswap
27;
28; AVX-LABEL: 'var_bswap_v2i64'
29; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a)
30; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bswap
31;
32  %bswap = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %a)
33  ret <2 x i64> %bswap
34}
35
36define <4 x i64> @var_bswap_v4i64(<4 x i64> %a) {
37; SSE2-LABEL: 'var_bswap_v4i64'
38; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bswap = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a)
39; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bswap
40;
41; SSE42-LABEL: 'var_bswap_v4i64'
42; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bswap = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a)
43; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bswap
44;
45; AVX1-LABEL: 'var_bswap_v4i64'
46; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bswap = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a)
47; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bswap
48;
49; AVX2-LABEL: 'var_bswap_v4i64'
50; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a)
51; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bswap
52;
53  %bswap = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %a)
54  ret <4 x i64> %bswap
55}
56
57define <4 x i32> @var_bswap_v4i32(<4 x i32> %a) {
58; SSE2-LABEL: 'var_bswap_v4i32'
59; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bswap = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a)
60; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bswap
61;
62; SSE42-LABEL: 'var_bswap_v4i32'
63; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a)
64; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bswap
65;
66; AVX-LABEL: 'var_bswap_v4i32'
67; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a)
68; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bswap
69;
70  %bswap = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a)
71  ret <4 x i32> %bswap
72}
73
74define <8 x i32> @var_bswap_v8i32(<8 x i32> %a) {
75; SSE2-LABEL: 'var_bswap_v8i32'
76; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bswap = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a)
77; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bswap
78;
79; SSE42-LABEL: 'var_bswap_v8i32'
80; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bswap = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a)
81; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bswap
82;
83; AVX1-LABEL: 'var_bswap_v8i32'
84; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bswap = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a)
85; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bswap
86;
87; AVX2-LABEL: 'var_bswap_v8i32'
88; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a)
89; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bswap
90;
91  %bswap = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %a)
92  ret <8 x i32> %bswap
93}
94
95define <8 x i16> @var_bswap_v8i16(<8 x i16> %a) {
96; SSE2-LABEL: 'var_bswap_v8i16'
97; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bswap = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a)
98; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bswap
99;
100; SSE42-LABEL: 'var_bswap_v8i16'
101; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a)
102; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bswap
103;
104; AVX-LABEL: 'var_bswap_v8i16'
105; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a)
106; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bswap
107;
108  %bswap = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %a)
109  ret <8 x i16> %bswap
110}
111
112define <16 x i16> @var_bswap_v16i16(<16 x i16> %a) {
113; SSE2-LABEL: 'var_bswap_v16i16'
114; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bswap = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a)
115; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bswap
116;
117; SSE42-LABEL: 'var_bswap_v16i16'
118; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bswap = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a)
119; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bswap
120;
121; AVX1-LABEL: 'var_bswap_v16i16'
122; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bswap = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a)
123; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bswap
124;
125; AVX2-LABEL: 'var_bswap_v16i16'
126; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bswap = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a)
127; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bswap
128;
129  %bswap = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %a)
130  ret <16 x i16> %bswap
131}
132