Mercurial > hg > CbC > CbC_llvm

comparison test/CodeGen/X86/fold-load-binops.ll @ 95:afa8332a0e37 LLVM3.8

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
LLVM 3.8
author Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp>
date Tue, 13 Oct 2015 17:48:58 +0900
parents
children 7d135dc70f03
comparison
equal deleted inserted replaced
84:f3e34b893a5f 95:afa8332a0e37
1 ; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+sse2 < %s | FileCheck %s --check-prefix=SSE
2 ; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+avx < %s | FileCheck %s --check-prefix=AVX
3
4 ; Verify that we're folding the load into the math instruction.
5 ; This pattern is generated out of the simplest intrinsics usage:
6 ; _mm_add_ss(a, _mm_load_ss(b));
7
8 define <4 x float> @addss(<4 x float> %va, float* %pb) {
9 ; SSE-LABEL: addss:
10 ; SSE: # BB#0:
11 ; SSE-NEXT: addss (%rdi), %xmm0
12 ; SSE-NEXT: retq
13 ;
14 ; AVX-LABEL: addss:
15 ; AVX: # BB#0:
16 ; AVX-NEXT: vaddss (%rdi), %xmm0, %xmm0
17 ; AVX-NEXT: retq
18 %a = extractelement <4 x float> %va, i32 0
19 %b = load float, float* %pb
20 %r = fadd float %a, %b
21 %vr = insertelement <4 x float> %va, float %r, i32 0
22 ret <4 x float> %vr
23 }
24
25 define <2 x double> @addsd(<2 x double> %va, double* %pb) {
26 ; SSE-LABEL: addsd:
27 ; SSE: # BB#0:
28 ; SSE-NEXT: addsd (%rdi), %xmm0
29 ; SSE-NEXT: retq
30 ;
31 ; AVX-LABEL: addsd:
32 ; AVX: # BB#0:
33 ; AVX-NEXT: vaddsd (%rdi), %xmm0, %xmm0
34 ; AVX-NEXT: retq
35 %a = extractelement <2 x double> %va, i32 0
36 %b = load double, double* %pb
37 %r = fadd double %a, %b
38 %vr = insertelement <2 x double> %va, double %r, i32 0
39 ret <2 x double> %vr
40 }
41
42 define <4 x float> @subss(<4 x float> %va, float* %pb) {
43 ; SSE-LABEL: subss:
44 ; SSE: # BB#0:
45 ; SSE-NEXT: subss (%rdi), %xmm0
46 ; SSE-NEXT: retq
47 ;
48 ; AVX-LABEL: subss:
49 ; AVX: # BB#0:
50 ; AVX-NEXT: vsubss (%rdi), %xmm0, %xmm0
51 ; AVX-NEXT: retq
52 %a = extractelement <4 x float> %va, i32 0
53 %b = load float, float* %pb
54 %r = fsub float %a, %b
55 %vr = insertelement <4 x float> %va, float %r, i32 0
56 ret <4 x float> %vr
57 }
58
59 define <2 x double> @subsd(<2 x double> %va, double* %pb) {
60 ; SSE-LABEL: subsd:
61 ; SSE: # BB#0:
62 ; SSE-NEXT: subsd (%rdi), %xmm0
63 ; SSE-NEXT: retq
64 ;
65 ; AVX-LABEL: subsd:
66 ; AVX: # BB#0:
67 ; AVX-NEXT: vsubsd (%rdi), %xmm0, %xmm0
68 ; AVX-NEXT: retq
69 %a = extractelement <2 x double> %va, i32 0
70 %b = load double, double* %pb
71 %r = fsub double %a, %b
72 %vr = insertelement <2 x double> %va, double %r, i32 0
73 ret <2 x double> %vr
74 }
75
76 define <4 x float> @mulss(<4 x float> %va, float* %pb) {
77 ; SSE-LABEL: mulss:
78 ; SSE: # BB#0:
79 ; SSE-NEXT: mulss (%rdi), %xmm0
80 ; SSE-NEXT: retq
81 ;
82 ; AVX-LABEL: mulss:
83 ; AVX: # BB#0:
84 ; AVX-NEXT: vmulss (%rdi), %xmm0, %xmm0
85 ; AVX-NEXT: retq
86 %a = extractelement <4 x float> %va, i32 0
87 %b = load float, float* %pb
88 %r = fmul float %a, %b
89 %vr = insertelement <4 x float> %va, float %r, i32 0
90 ret <4 x float> %vr
91 }
92
93 define <2 x double> @mulsd(<2 x double> %va, double* %pb) {
94 ; SSE-LABEL: mulsd:
95 ; SSE: # BB#0:
96 ; SSE-NEXT: mulsd (%rdi), %xmm0
97 ; SSE-NEXT: retq
98 ;
99 ; AVX-LABEL: mulsd:
100 ; AVX: # BB#0:
101 ; AVX-NEXT: vmulsd (%rdi), %xmm0, %xmm0
102 ; AVX-NEXT: retq
103 %a = extractelement <2 x double> %va, i32 0
104 %b = load double, double* %pb
105 %r = fmul double %a, %b
106 %vr = insertelement <2 x double> %va, double %r, i32 0
107 ret <2 x double> %vr
108 }
109
110 define <4 x float> @divss(<4 x float> %va, float* %pb) {
111 ; SSE-LABEL: divss:
112 ; SSE: # BB#0:
113 ; SSE-NEXT: divss (%rdi), %xmm0
114 ; SSE-NEXT: retq
115 ;
116 ; AVX-LABEL: divss:
117 ; AVX: # BB#0:
118 ; AVX-NEXT: vdivss (%rdi), %xmm0, %xmm0
119 ; AVX-NEXT: retq
120 %a = extractelement <4 x float> %va, i32 0
121 %b = load float, float* %pb
122 %r = fdiv float %a, %b
123 %vr = insertelement <4 x float> %va, float %r, i32 0
124 ret <4 x float> %vr
125 }
126
127 define <2 x double> @divsd(<2 x double> %va, double* %pb) {
128 ; SSE-LABEL: divsd:
129 ; SSE: # BB#0:
130 ; SSE-NEXT: divsd (%rdi), %xmm0
131 ; SSE-NEXT: retq
132 ;
133 ; AVX-LABEL: divsd:
134 ; AVX: # BB#0:
135 ; AVX-NEXT: vdivsd (%rdi), %xmm0, %xmm0
136 ; AVX-NEXT: retq
137 %a = extractelement <2 x double> %va, i32 0
138 %b = load double, double* %pb
139 %r = fdiv double %a, %b
140 %vr = insertelement <2 x double> %va, double %r, i32 0
141 ret <2 x double> %vr
142 }