-
Notifications
You must be signed in to change notification settings - Fork 1
/
kernmulrsymm2dxy_wrapper.go
174 lines (152 loc) · 4.99 KB
/
kernmulrsymm2dxy_wrapper.go
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
package cuda
/*
THIS FILE IS AUTO-GENERATED BY CUDA2GO.
EDITING IS FUTILE.
*/
import (
"github.com/kuchkin/mumax3-gneb/cuda/cu"
"github.com/kuchkin/mumax3-gneb/timer"
"sync"
"unsafe"
)
// CUDA handle for kernmulRSymm2Dxy kernel
var kernmulRSymm2Dxy_code cu.Function
// Stores the arguments for kernmulRSymm2Dxy kernel invocation
type kernmulRSymm2Dxy_args_t struct {
arg_fftMx unsafe.Pointer
arg_fftMy unsafe.Pointer
arg_fftKxx unsafe.Pointer
arg_fftKyy unsafe.Pointer
arg_fftKxy unsafe.Pointer
arg_Nx int
arg_Ny int
argptr [7]unsafe.Pointer
sync.Mutex
}
// Stores the arguments for kernmulRSymm2Dxy kernel invocation
var kernmulRSymm2Dxy_args kernmulRSymm2Dxy_args_t
func init() {
// CUDA driver kernel call wants pointers to arguments, set them up once.
kernmulRSymm2Dxy_args.argptr[0] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_fftMx)
kernmulRSymm2Dxy_args.argptr[1] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_fftMy)
kernmulRSymm2Dxy_args.argptr[2] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_fftKxx)
kernmulRSymm2Dxy_args.argptr[3] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_fftKyy)
kernmulRSymm2Dxy_args.argptr[4] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_fftKxy)
kernmulRSymm2Dxy_args.argptr[5] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_Nx)
kernmulRSymm2Dxy_args.argptr[6] = unsafe.Pointer(&kernmulRSymm2Dxy_args.arg_Ny)
}
// Wrapper for kernmulRSymm2Dxy CUDA kernel, asynchronous.
func k_kernmulRSymm2Dxy_async(fftMx unsafe.Pointer, fftMy unsafe.Pointer, fftKxx unsafe.Pointer, fftKyy unsafe.Pointer, fftKxy unsafe.Pointer, Nx int, Ny int, cfg *config) {
if Synchronous { // debug
Sync()
timer.Start("kernmulRSymm2Dxy")
}
kernmulRSymm2Dxy_args.Lock()
defer kernmulRSymm2Dxy_args.Unlock()
if kernmulRSymm2Dxy_code == 0 {
kernmulRSymm2Dxy_code = fatbinLoad(kernmulRSymm2Dxy_map, "kernmulRSymm2Dxy")
}
kernmulRSymm2Dxy_args.arg_fftMx = fftMx
kernmulRSymm2Dxy_args.arg_fftMy = fftMy
kernmulRSymm2Dxy_args.arg_fftKxx = fftKxx
kernmulRSymm2Dxy_args.arg_fftKyy = fftKyy
kernmulRSymm2Dxy_args.arg_fftKxy = fftKxy
kernmulRSymm2Dxy_args.arg_Nx = Nx
kernmulRSymm2Dxy_args.arg_Ny = Ny
args := kernmulRSymm2Dxy_args.argptr[:]
cu.LaunchKernel(kernmulRSymm2Dxy_code, cfg.Grid.X, cfg.Grid.Y, cfg.Grid.Z, cfg.Block.X, cfg.Block.Y, cfg.Block.Z, 0, stream0, args)
if Synchronous { // debug
Sync()
timer.Stop("kernmulRSymm2Dxy")
}
}
// maps compute capability on PTX code for kernmulRSymm2Dxy kernel.
var kernmulRSymm2Dxy_map = map[int]string{0: "",
50: kernmulRSymm2Dxy_ptx_50}
// kernmulRSymm2Dxy PTX code for various compute capabilities.
const (
kernmulRSymm2Dxy_ptx_50 = `
.version 7.5
.target sm_50
.address_size 64
// .globl kernmulRSymm2Dxy
.visible .entry kernmulRSymm2Dxy(
.param .u64 kernmulRSymm2Dxy_param_0,
.param .u64 kernmulRSymm2Dxy_param_1,
.param .u64 kernmulRSymm2Dxy_param_2,
.param .u64 kernmulRSymm2Dxy_param_3,
.param .u64 kernmulRSymm2Dxy_param_4,
.param .u32 kernmulRSymm2Dxy_param_5,
.param .u32 kernmulRSymm2Dxy_param_6
)
{
.reg .pred %p<5>;
.reg .f32 %f<18>;
.reg .b32 %r<19>;
.reg .b64 %rd<18>;
ld.param.u64 %rd1, [kernmulRSymm2Dxy_param_0];
ld.param.u64 %rd2, [kernmulRSymm2Dxy_param_1];
ld.param.u64 %rd3, [kernmulRSymm2Dxy_param_2];
ld.param.u64 %rd4, [kernmulRSymm2Dxy_param_3];
ld.param.u64 %rd5, [kernmulRSymm2Dxy_param_4];
ld.param.u32 %r3, [kernmulRSymm2Dxy_param_5];
ld.param.u32 %r4, [kernmulRSymm2Dxy_param_6];
mov.u32 %r5, %ntid.x;
mov.u32 %r6, %ctaid.x;
mov.u32 %r7, %tid.x;
mad.lo.s32 %r1, %r6, %r5, %r7;
mov.u32 %r8, %ntid.y;
mov.u32 %r9, %ctaid.y;
mov.u32 %r10, %tid.y;
mad.lo.s32 %r2, %r9, %r8, %r10;
setp.ge.s32 %p1, %r1, %r3;
setp.ge.s32 %p2, %r2, %r4;
or.pred %p3, %p1, %p2;
@%p3 bra $L__BB0_2;
cvta.to.global.u64 %rd6, %rd3;
mad.lo.s32 %r11, %r2, %r3, %r1;
shl.b32 %r12, %r11, 1;
cvta.to.global.u64 %rd7, %rd1;
mul.wide.s32 %rd8, %r12, 4;
add.s64 %rd9, %rd7, %rd8;
ld.global.f32 %f1, [%rd9+4];
cvta.to.global.u64 %rd10, %rd2;
add.s64 %rd11, %rd10, %rd8;
ld.global.f32 %f2, [%rd11+4];
shr.u32 %r13, %r4, 31;
add.s32 %r14, %r4, %r13;
shr.s32 %r15, %r14, 1;
setp.gt.s32 %p4, %r2, %r15;
sub.s32 %r16, %r4, %r2;
selp.b32 %r17, %r16, %r2, %p4;
selp.f32 %f3, 0fBF800000, 0f3F800000, %p4;
mad.lo.s32 %r18, %r17, %r3, %r1;
mul.wide.s32 %rd12, %r18, 4;
add.s64 %rd13, %rd6, %rd12;
cvta.to.global.u64 %rd14, %rd4;
add.s64 %rd15, %rd14, %rd12;
ld.global.nc.f32 %f4, [%rd15];
cvta.to.global.u64 %rd16, %rd5;
add.s64 %rd17, %rd16, %rd12;
ld.global.nc.f32 %f5, [%rd17];
mul.f32 %f6, %f3, %f5;
ld.global.nc.f32 %f7, [%rd13];
ld.global.f32 %f8, [%rd9];
ld.global.f32 %f9, [%rd11];
mul.f32 %f10, %f9, %f6;
fma.rn.f32 %f11, %f8, %f7, %f10;
st.global.f32 [%rd9], %f11;
mul.f32 %f12, %f2, %f6;
fma.rn.f32 %f13, %f1, %f7, %f12;
st.global.f32 [%rd9+4], %f13;
mul.f32 %f14, %f8, %f6;
fma.rn.f32 %f15, %f9, %f4, %f14;
st.global.f32 [%rd11], %f15;
mul.f32 %f16, %f1, %f6;
fma.rn.f32 %f17, %f2, %f4, %f16;
st.global.f32 [%rd11+4], %f17;
$L__BB0_2:
ret;
}
`
)