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
234
235
236
237
238
239
240
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_vect_dot_prod_avx512(len, vec, *g_tbls, **buffs, *dest);
;;;
%include "reg_sizes.asm"
%ifdef HAVE_AS_KNOWS_AVX512
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x: endbranch
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp2 r10
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 0*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_reg r12, 0*8
save_reg r15, 1*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
mov r12, [rsp + 0*8]
mov r15, [rsp + 1*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu8
%define XSTR vmovdqu8
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa64
%define XSTR vmovdqa64
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%define xmask0f zmm5
%define xgft1_lo zmm4
%define xgft1_loy ymm4
%define xgft1_hi zmm3
%define x0 zmm0
%define xgft1_loy ymm4
%define x0y ymm0
%define xtmpa zmm1
%define xp1 zmm2
%define xp1y ymm2
default rel
[bits 64]
section .text
align 16
global gf_vect_dot_prod_avx512, function
func(gf_vect_dot_prod_avx512)
FUNC_SAVE
xor pos, pos
mov tmp, 0x0f
vpbroadcastb xmask0f, tmp ;Construct mask 0x0f0f0f...
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
sub len, 64
jl .len_lt_64
.loop64:
vpxorq xp1, xp1, xp1
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
mov ptr, [src+vec_i]
XLDR x0, [ptr+pos] ;Get next source vector
add vec_i, PS
vpandq xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpandq x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vmovdqu8 xgft1_loy, [tmp] ;Load array Ax{00}..{0f}, Ax{00}..{f0}
add tmp, 32
vshufi64x2 xgft1_hi, xgft1_lo, xgft1_lo, 0x55
vshufi64x2 xgft1_lo, xgft1_lo, xgft1_lo, 0x00
vpshufb xgft1_hi, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxorq xgft1_hi, xgft1_hi, xgft1_lo ;GF add high and low partials
vpxorq xp1, xp1, xgft1_hi ;xp1 += partial
cmp vec_i, vec
jl .next_vect
XSTR [dest1+pos], xp1
add pos, 64 ;Loop on 64 bytes at a time
cmp pos, len
jle .loop64
lea tmp, [len + 64]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-64
jmp .loop64 ;Do one more overlap pass
.len_lt_64: ; 32-byte version
add len, 32
jl .return_fail
.loop32:
vpxorq xp1, xp1, xp1
mov tmp, mul_array
xor vec_i, vec_i
.next_vect2:
mov ptr, [src+vec_i]
XLDR x0y, [ptr+pos] ;Get next source vector 32B
add vec_i, PS
vpsraw xtmpa, x0, 4 ;Shift to put high nibble into bits 4-0
vshufi64x2 x0, x0, xtmpa, 0x44 ;put x0 = xl:xh
vpandq x0, x0, xmask0f ;Mask bits 4-0
vmovdqu8 xgft1_loy, [tmp] ;Load array Ax{00}..{0f}, Ax{00}..{f0}
add tmp, 32
vshufi64x2 xgft1_lo, xgft1_lo, xgft1_lo, 0x50 ;=AlAh:AlAh
vpshufb xgft1_lo, xgft1_lo, x0 ;Lookup mul table
vshufi64x2 xgft1_hi, xgft1_lo, xgft1_lo, 0x0e ;=xh:
vpxorq xgft1_hi, xgft1_hi, xgft1_lo ;GF add high and low partials
vpxorq xp1, xp1, xgft1_hi ;xp1 += partial
cmp vec_i, vec
jl .next_vect2
XSTR [dest1+pos], xp1y
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-32
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
%else
%ifidn __OUTPUT_FORMAT__, win64
global no_gf_vect_dot_prod_avx512
no_gf_vect_dot_prod_avx512:
%endif
%endif ; ifdef HAVE_AS_KNOWS_AVX512
|
