1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
|
; RUN: llc < %s -mtriple=armv7-eabi -mattr=+neon,+vfp4 -fp-contract=fast | FileCheck %s
; RUN: llc < %s -mtriple=arm-arm-eabi -mcpu=cortex-m7 -fp-contract=fast | FileCheck %s
; RUN: llc < %s -mtriple=arm-arm-eabi -mcpu=cortex-m4 -fp-contract=fast | FileCheck %s -check-prefix=DONT-FUSE
; RUN: llc < %s -mtriple=arm-arm-eabi -mcpu=cortex-m33 -fp-contract=fast | FileCheck %s -check-prefix=DONT-FUSE
; Check generated fused MAC and MLS.
define arm_aapcs_vfpcc double @fusedMACTest1(double %d1, double %d2, double %d3) {
;CHECK-LABEL: fusedMACTest1:
;CHECK: vfma.f64
%1 = fmul double %d1, %d2
%2 = fadd double %1, %d3
ret double %2
}
define arm_aapcs_vfpcc float @fusedMACTest2(float %f1, float %f2, float %f3) {
;CHECK-LABEL: fusedMACTest2:
;CHECK: vfma.f32
;DONT-FUSE-LABEL: fusedMACTest2:
;DONT-FUSE: vmul.f32
;DONT-FUSE-NEXT: vadd.f32
%1 = fmul float %f1, %f2
%2 = fadd float %1, %f3
ret float %2
}
define arm_aapcs_vfpcc double @fusedMACTest3(double %d1, double %d2, double %d3) {
;CHECK-LABEL: fusedMACTest3:
;CHECK: vfms.f64
%1 = fmul double %d2, %d3
%2 = fsub double %d1, %1
ret double %2
}
define arm_aapcs_vfpcc float @fusedMACTest4(float %f1, float %f2, float %f3) {
;CHECK-LABEL: fusedMACTest4:
;CHECK: vfms.f32
%1 = fmul float %f2, %f3
%2 = fsub float %f1, %1
ret float %2
}
define arm_aapcs_vfpcc double @fusedMACTest5(double %d1, double %d2, double %d3) {
;CHECK-LABEL: fusedMACTest5:
;CHECK: vfnma.f64
%1 = fmul double %d1, %d2
%2 = fsub double -0.0, %1
%3 = fsub double %2, %d3
ret double %3
}
define arm_aapcs_vfpcc float @fusedMACTest6(float %f1, float %f2, float %f3) {
;CHECK-LABEL: fusedMACTest6:
;CHECK: vfnma.f32
%1 = fmul float %f1, %f2
%2 = fsub float -0.0, %1
%3 = fsub float %2, %f3
ret float %3
}
define arm_aapcs_vfpcc double @fusedMACTest7(double %d1, double %d2, double %d3) {
;CHECK-LABEL: fusedMACTest7:
;CHECK: vfnms.f64
%1 = fmul double %d1, %d2
%2 = fsub double %1, %d3
ret double %2
}
define arm_aapcs_vfpcc float @fusedMACTest8(float %f1, float %f2, float %f3) {
;CHECK-LABEL: fusedMACTest8:
;CHECK: vfnms.f32
%1 = fmul float %f1, %f2
%2 = fsub float %1, %f3
ret float %2
}
define arm_aapcs_vfpcc <2 x float> @fusedMACTest9(<2 x float> %a, <2 x float> %b) {
;CHECK-LABEL: fusedMACTest9:
;CHECK: vfma.f32
%mul = fmul <2 x float> %a, %b
%add = fadd <2 x float> %mul, %a
ret <2 x float> %add
}
define arm_aapcs_vfpcc <2 x float> @fusedMACTest10(<2 x float> %a, <2 x float> %b) {
;CHECK-LABEL: fusedMACTest10:
;CHECK: vfms.f32
%mul = fmul <2 x float> %a, %b
%sub = fsub <2 x float> %a, %mul
ret <2 x float> %sub
}
define arm_aapcs_vfpcc <4 x float> @fusedMACTest11(<4 x float> %a, <4 x float> %b) {
;CHECK-LABEL: fusedMACTest11:
;CHECK: vfma.f32
%mul = fmul <4 x float> %a, %b
%add = fadd <4 x float> %mul, %a
ret <4 x float> %add
}
define arm_aapcs_vfpcc <4 x float> @fusedMACTest12(<4 x float> %a, <4 x float> %b) {
;CHECK-LABEL: fusedMACTest12:
;CHECK: vfms.f32
%mul = fmul <4 x float> %a, %b
%sub = fsub <4 x float> %a, %mul
ret <4 x float> %sub
}
define arm_aapcs_vfpcc float @test_fma_f32(float %a, float %b, float %c) nounwind readnone ssp {
entry:
; CHECK: test_fma_f32
; CHECK: vfma.f32
%tmp1 = tail call float @llvm.fma.f32(float %a, float %b, float %c) nounwind readnone
ret float %tmp1
}
define arm_aapcs_vfpcc double @test_fma_f64(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fma_f64
; CHECK: vfma.f64
%tmp1 = tail call double @llvm.fma.f64(double %a, double %b, double %c) nounwind readnone
ret double %tmp1
}
define arm_aapcs_vfpcc <2 x float> @test_fma_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind readnone ssp {
entry:
; CHECK: test_fma_v2f32
; CHECK: vfma.f32
%tmp1 = tail call <2 x float> @llvm.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind
ret <2 x float> %tmp1
}
define arm_aapcs_vfpcc double @test_fms_f64(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fms_f64
; CHECK: vfms.f64
%tmp1 = fsub double -0.0, %a
%tmp2 = tail call double @llvm.fma.f64(double %tmp1, double %b, double %c) nounwind readnone
ret double %tmp2
}
define arm_aapcs_vfpcc double @test_fms_f64_2(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fms_f64_2
; CHECK: vfms.f64
%tmp1 = fsub double -0.0, %b
%tmp2 = tail call double @llvm.fma.f64(double %a, double %tmp1, double %c) nounwind readnone
ret double %tmp2
}
define arm_aapcs_vfpcc float @test_fnms_f32(float %a, float %b, ptr %c) nounwind readnone ssp {
; CHECK: test_fnms_f32
; CHECK: vfnms.f32
%tmp1 = load float, ptr %c, align 4
%tmp2 = fsub float -0.0, %tmp1
%tmp3 = tail call float @llvm.fma.f32(float %a, float %b, float %tmp2) nounwind readnone
ret float %tmp3
}
define arm_aapcs_vfpcc double @test_fnms_f64(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fnms_f64
; CHECK: vfnms.f64
%tmp1 = fsub double -0.0, %a
%tmp2 = tail call double @llvm.fma.f64(double %tmp1, double %b, double %c) nounwind readnone
%tmp3 = fsub double -0.0, %tmp2
ret double %tmp3
}
define arm_aapcs_vfpcc double @test_fnms_f64_2(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fnms_f64_2
; CHECK: vfnms.f64
%tmp1 = fsub double -0.0, %b
%tmp2 = tail call double @llvm.fma.f64(double %a, double %tmp1, double %c) nounwind readnone
%tmp3 = fsub double -0.0, %tmp2
ret double %tmp3
}
define arm_aapcs_vfpcc double @test_fnma_f64(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fnma_f64
; CHECK: vfnma.f64
%tmp1 = tail call double @llvm.fma.f64(double %a, double %b, double %c) nounwind readnone
%tmp2 = fsub double -0.0, %tmp1
ret double %tmp2
}
define arm_aapcs_vfpcc double @test_fnma_f64_2(double %a, double %b, double %c) nounwind readnone ssp {
entry:
; CHECK: test_fnma_f64_2
; CHECK: vfnma.f64
%tmp1 = fsub double -0.0, %a
%tmp2 = fsub double -0.0, %c
%tmp3 = tail call double @llvm.fma.f64(double %tmp1, double %b, double %tmp2) nounwind readnone
ret double %tmp3
}
define arm_aapcs_vfpcc float @test_fma_const_fold(float %a, float %b) nounwind {
; CHECK: test_fma_const_fold
; CHECK-NOT: vfma
; CHECK-NOT: vmul
; CHECK: vadd
%ret = call float @llvm.fma.f32(float %a, float 1.0, float %b)
ret float %ret
}
define arm_aapcs_vfpcc float @test_fma_canonicalize(float %a, float %b) nounwind {
; CHECK: test_fma_canonicalize
; CHECK: vmov.f32 [[R1:s[0-9]+]], #2.000000e+00
; CHECK: vfma.f32 {{s[0-9]+}}, {{s[0-9]+}}, [[R1]]
%ret = call float @llvm.fma.f32(float 2.0, float %a, float %b)
ret float %ret
}
; Check that very wide vector fma's can be split into legal fma's.
define arm_aapcs_vfpcc void @test_fma_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, ptr %p) nounwind readnone ssp {
; CHECK: test_fma_v8f32
; CHECK: vfma.f32
; CHECK: vfma.f32
entry:
%call = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) nounwind readnone
store <8 x float> %call, ptr %p, align 16
ret void
}
declare float @llvm.fma.f32(float, float, float) nounwind readnone
declare double @llvm.fma.f64(double, double, double) nounwind readnone
declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>) nounwind readnone
declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>) nounwind readnone
|
