-
Notifications
You must be signed in to change notification settings - Fork 1.5k
/
Copy pathdscal_lasx.S
194 lines (180 loc) · 3.68 KB
/
dscal_lasx.S
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
#define ASSEMBLER
#include "common.h"
#define N $r4
#define ALPHA $f0
#define X $r7
#define INCX $r8
#define I $r12
#define TEMP $r13
#define t1 $r14
#define t2 $r18
#define t3 $r15
#define t4 $r17
#define XX $r16
#define VX0 $xr12
#define VX1 $xr13
#define VT0 $xr14
#define VT1 $xr15
#define VALPHA $xr19
#define a1 $f8
#define a2 $f23
PROLOGUE
bge $r0, N, .L999
bge $r0, INCX, .L999
li.d TEMP, 1
movgr2fr.d a1, $r0
ffint.d.l a1, a1
movgr2fr.d a2, TEMP
ffint.d.l a2, a2
slli.d TEMP, TEMP, BASE_SHIFT
slli.d INCX, INCX, BASE_SHIFT
fcmp.ceq.d $fcc0, ALPHA, a1
bcnez $fcc0, .L20 //ALPHA==0
fcmp.ceq.d $fcc0, ALPHA, a2
bcnez $fcc0, .L999 //ALPHA==1 return
srai.d I, N, 3
beq INCX, TEMP, .L30 //ALPHA！=0|1 and INCX==1
movfr2gr.d TEMP, ALPHA
xvreplgr2vr.d VALPHA, TEMP
move XX, X
.align 3
.L10:
bge $r0, I, .L32
.align 3
.L11:
ld.d t1, X, 0 * SIZE
add.d X, X, INCX
ld.d t2, X, 0 * SIZE
add.d X, X, INCX
ld.d t3, X, 0 * SIZE
add.d X, X, INCX
ld.d t4, X, 0 * SIZE
add.d X, X, INCX
xvinsgr2vr.d VX0, t1, 0
xvinsgr2vr.d VX0, t2, 1
xvinsgr2vr.d VX0, t3, 2
xvinsgr2vr.d VX0, t4, 3
ld.d t1, X, 0 * SIZE
add.d X, X, INCX
ld.d t2, X, 0 * SIZE
add.d X, X, INCX
xvfmul.d VT0, VX0, VALPHA
ld.d t3, X, 0 * SIZE
add.d X, X, INCX
ld.d t4, X, 0 * SIZE
add.d X, X, INCX
xvinsgr2vr.d VX1, t1, 0
xvinsgr2vr.d VX1, t2, 1
xvinsgr2vr.d VX1, t3, 2
xvinsgr2vr.d VX1, t4, 3
xvstelm.d VT0, XX, 0, 0
add.d XX, XX, INCX
xvstelm.d VT0, XX, 0, 1
add.d XX, XX, INCX
xvstelm.d VT0, XX, 0, 2
add.d XX, XX, INCX
xvstelm.d VT0, XX, 0, 3
add.d XX, XX, INCX
xvfmul.d VT1, VX1, VALPHA
xvstelm.d VT1, XX, 0, 0
add.d XX, XX, INCX
xvstelm.d VT1, XX, 0, 1
add.d XX, XX, INCX
xvstelm.d VT1, XX, 0, 2
add.d XX, XX, INCX
xvstelm.d VT1, XX, 0, 3
add.d XX, XX, INCX
addi.d I, I, -1
blt $r0, I, .L11
b .L32
.align 3
.L20:
srai.d I, N, 3
beq INCX, TEMP, .L24
bge $r0, I, .L22
.align 3
.L21:
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
fst.d a1, X, 0
add.d X, X, INCX
addi.d I, I, -1
blt $r0, I, .L21
.align 3
.L22:
andi I, N, 7
bge $r0, I, .L999
.align 3
.L23:
fst.d a1, X, 0 * SIZE
addi.d I, I, -1
add.d X, X, INCX
blt $r0, I, .L23
jirl $r0, $r1, 0
.align 3
.L24:
bge $r0, I, .L26 /*N<8 INCX==1*/
.align 3
.L25:
xvxor.v VX0, VX0, VX0
xvst VX0, X, 0 * SIZE
xvst VX0, X, 4 * SIZE
addi.d I, I, -1
addi.d X, X, 8 * SIZE
blt $r0, I, .L25
.align 3
.L26:
andi I, N, 7
bge $r0, I, .L999
.align 3
.L27:
fst.d a1, X, 0 * SIZE
addi.d I, I, -1
addi.d X, X, SIZE
blt $r0, I, .L27
jirl $r0, $r1, 0
.align 3
.L30:
bge $r0, I, .L32/*N<8 INCX==1*/
movfr2gr.d TEMP, ALPHA
xvreplgr2vr.d VALPHA , TEMP
.align 3
.L31:
xvld VX0, X, 0 * SIZE
xvld VX1, X, 4 * SIZE
xvfmul.d VT0, VX0, VALPHA
xvfmul.d VT1, VX1, VALPHA
addi.d I, I, -1
xvst VT0, X, 0 * SIZE
xvst VT1, X, 4 * SIZE
addi.d X, X, 8 * SIZE
blt $r0, I, .L31
.align 3
.L32:
andi I, N, 7
bge $r0, I, .L999
.align 3
.L33:
fld.d a1, X, 0 * SIZE
addi.d I, I, -1
fmul.d a1, ALPHA, a1
fst.d a1, X, 0 * SIZE
add.d X, X, INCX
blt $r0, I, .L33
jirl $r0, $r1, 0
.align 3
.L999:
jirl $r0, $r1, 0x0
EPILOGUE