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
|
#include "regs.h"
#include "sse.h"
#include "avx.h"
DATA one(SB)/8,$1.0
GLOBL one(SB), $8
TEXT dotvec2_sse4(SB), 1, $0
MOVQ SP, AX
MOVDQU_mr(8, rAX, rX0) /* MOVDQU a+0(FP), X0 */
MOVDQU_mr(32, rAX, rX1) /* MOVDQU b+24(FP), X1 */
DPPD(rX1, rX0) /* DPPD $0x31, X1, X0 */
RET
TEXT dotvec2_avx(SB), 1, $0
MOVQ SP, AX
VMOVUPD_128mr(8, rAX, rX0) /* VMOVUPD a+0(FP), X0 */
VMOVUPD_128mr(32, rAX, rX1) /* VMOVUPD b+24(FP), X1 */
VDPPD(rX1, rX0, rX0) /* VDPPD $0x31, X1, X0, X0 */
RET
TEXT dotvec3_sse4(SB), 1, $0
MOVQ SP, AX
MOVLPD(8, rAX, rX0) /* MOVLPD a+0(FP), X0 */
MOVHPD(16, rAX, rX0) /* MOVHPD a+8(FP), X0 */
MOVLPD(40, rAX, rX1) /* MOVLPD b+32(FP), X1 */
MOVHPD(48, rAX, rX1) /* MOVHPD b+40(FP), X1 */
DPPD(rX1, rX0) /* DPPD $0x31, X1, X0 */
MOVSD one(SB), X1
MOVHPD(24, rAX, rX0) /* MOVHPD a+16(FP), X0 */
MOVHPD(56, rAX, rX1) /* MOVHPD b+48(FP), X1 */
DPPD(rX1, rX0) /* DPPD $0x31, X1, X0 */
RET
TEXT dotvec3_avx(SB), 1, $0
MOVQ SP, AX
VMOVUPD_128mr(8, rAX, rX0) /* VMOVUPD a+0(FP), X0 */
VMOVUPD_128mr(40, rAX, rX1) /* VMOVUPD b+32(FP), X1 */
VDPPD(rX1, rX0, rX0) /* VDPPD $0x31, X1, X0, X0 */
MOVSD a+16(FP), X1
MOVSD b+48(FP), X2
VFMADD231SD(rX1, rX2, rX0)
RET
TEXT Pt2b(SB), 1, $0
MOVQ BP, DI
MOVSD x+8(FP), X0
MOVSD X0, 0(DI)
MOVSD y+16(FP), X0
MOVSD X0, 8(DI)
MOVSD w+24(FP), X0
MOVSD X0, 16(DI)
RET
TEXT hsubpd(SB), 1, $0
MOVQ SP, AX
MOVLPD(8, rAX, rX0) /* MOVLPD a+0(FP), X0 */
MOVHPD(16, rAX, rX0) /* MOVHPD b+8(FP), X0 */
HSUBPD(rX0, rX0) /* HSUBPD X0, X0 */
RET
TEXT crossvec3_sse(SB), 1, $0
MOVQ SP, AX
ADDQ $8, AX
MOVLPD(40, rAX, rX0) /* MOVLPD b+32(FP), X0 */
MOVHPD(8, rAX, rX0) /* MOVHPD a+0(FP), X0 */
MOVLPD(16, rAX, rX1) /* MOVLPD a+8(FP), X1 */
MOVHPD(48, rAX, rX1) /* MOVHPD b+40(FP), X1 */
MOVLPD(56, rAX, rX2) /* MOVLPD b+48(FP), X2 */
MOVHPD(24, rAX, rX2) /* MOVHPD a+16(FP), X2 */
MOVAPD X1, X3
MULPD X2, X3
HSUBPD(rX3, rX3) /* x */
MOVAPD X2, X4
SHUFPD $0x1, X4, X4
MULPD X0, X4
HSUBPD(rX4, rX4) /* y */
MOVAPD X0, X5
MULPD X1, X5
SHUFPD $0x1, X5, X5
HSUBPD(rX5, rX5) /* z */
MOVQ BP, DI
MOVSD X3, 0(DI)
MOVSD X4, 8(DI)
MOVSD X5, 16(DI)
XORPD X0, X0
MOVSD X0, 24(DI)
RET
TEXT crossvec3_avx(SB), 1, $0
MOVQ SP, AX
ADDQ $8, AX
TEXT fma(SB), 1, $0
MOVSD a+0(FP), X0
MOVSD b+8(FP), X1
MOVSD c+16(FP), X2
VFMADD231SD(rX1, rX2, rX0)
RET
TEXT addpt2_avx(SB), 1, $0
MOVQ SP, AX
ADDQ $8, AX
VMOVDQU_256mr(8, rAX, rX0)
VMOVDQU_256mr(32, rAX, rX1)
VADDPD_256rr(rX1, rX0, rX0)
VMOVDQU_256rm(rX0, rAX)
RET
|
