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
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;; Optimized xor of N source vectors using AVX
;;; int xor_gen_avx(int vects, int len, void **array)
;;; Generates xor parity vector from N (vects-1) sources in array of pointers
;;; (**array). Last pointer is the dest.
;;; Vectors must be aligned to 32 bytes. Length can be any value.
%include "reg_sizes.asm"
%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 tmp3 arg4
%define func(x) x: endbranch
%define return rax
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define tmp r11
%define tmp3 r10
%define func(x) proc_frame x
%define return rax
%define stack_size 2*32 + 8 ;must be an odd multiple of 8
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqu [rsp + 0*32], ymm6
vmovdqu [rsp + 1*32], ymm7
end_prolog
%endmacro
%macro FUNC_RESTORE 0
vmovdqu ymm6, [rsp + 0*32]
vmovdqu ymm7, [rsp + 1*32]
add rsp, stack_size
%endmacro
%endif ;output formats
%define vec arg0
%define len arg1
%define ptr arg3
%define tmp2 rax
%define tmp2.b al
%define pos tmp3
%define PS 8
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovdqa
%define XSTR vmovntdq
%endif
default rel
[bits 64]
section .text
align 16
mk_global xor_gen_avx, function
func(xor_gen_avx)
FUNC_SAVE
sub vec, 2 ;Keep as offset to last source
jng return_fail ;Must have at least 2 sources
cmp len, 0
je return_pass
test len, (128-1) ;Check alignment of length
jnz len_not_aligned
len_aligned_128bytes:
sub len, 128
mov pos, 0
loop128:
mov tmp, vec ;Back to last vector
mov tmp2, [arg2+vec*PS] ;Fetch last pointer in array
sub tmp, 1 ;Next vect
XLDR ymm0, [tmp2+pos] ;Start with end of array in last vector
XLDR ymm1, [tmp2+pos+32] ;Keep xor parity in xmm0-7
XLDR ymm2, [tmp2+pos+(2*32)]
XLDR ymm3, [tmp2+pos+(3*32)]
next_vect:
mov ptr, [arg2+tmp*PS]
sub tmp, 1
XLDR ymm4, [ptr+pos] ;Get next vector (source)
XLDR ymm5, [ptr+pos+32]
XLDR ymm6, [ptr+pos+(2*32)]
XLDR ymm7, [ptr+pos+(3*32)]
vxorpd ymm0, ymm0, ymm4 ;Add to xor parity
vxorpd ymm1, ymm1, ymm5
vxorpd ymm2, ymm2, ymm6
vxorpd ymm3, ymm3, ymm7
jge next_vect ;Loop for each source
mov ptr, [arg2+PS+vec*PS] ;Address of parity vector
XSTR [ptr+pos], ymm0 ;Write parity xor vector
XSTR [ptr+pos+(1*32)], ymm1
XSTR [ptr+pos+(2*32)], ymm2
XSTR [ptr+pos+(3*32)], ymm3
add pos, 128
cmp pos, len
jle loop128
return_pass:
FUNC_RESTORE
mov return, 0
ret
;;; Do one byte at a time for no alignment case
loop_1byte:
mov tmp, vec ;Back to last vector
mov ptr, [arg2+vec*PS] ;Fetch last pointer in array
mov tmp2.b, [ptr+len-1] ;Get array n
sub tmp, 1
nextvect_1byte:
mov ptr, [arg2+tmp*PS]
xor tmp2.b, [ptr+len-1]
sub tmp, 1
jge nextvect_1byte
mov tmp, vec
add tmp, 1 ;Add back to point to last vec
mov ptr, [arg2+tmp*PS]
mov [ptr+len-1], tmp2.b ;Write parity
sub len, 1
test len, (PS-1)
jnz loop_1byte
cmp len, 0
je return_pass
test len, (128-1) ;If not 0 and 128bit aligned
jz len_aligned_128bytes ; then do aligned case. len = y * 128
;; else we are 8-byte aligned so fall through to recheck
;; Unaligned length cases
len_not_aligned:
test len, (PS-1)
jne loop_1byte
mov tmp3, len
and tmp3, (128-1) ;Do the unaligned bytes 8 at a time
;; Run backwards 8 bytes at a time for (tmp3) bytes
loop8_bytes:
mov tmp, vec ;Back to last vector
mov ptr, [arg2+vec*PS] ;Fetch last pointer in array
mov tmp2, [ptr+len-PS] ;Get array n
sub tmp, 1
nextvect_8bytes:
mov ptr, [arg2+tmp*PS] ;Get pointer to next vector
xor tmp2, [ptr+len-PS]
sub tmp, 1
jge nextvect_8bytes ;Loop for each source
mov tmp, vec
add tmp, 1 ;Add back to point to last vec
mov ptr, [arg2+tmp*PS]
mov [ptr+len-PS], tmp2 ;Write parity
sub len, PS
sub tmp3, PS
jg loop8_bytes
cmp len, 128 ;Now len is aligned to 128B
jge len_aligned_128bytes ;We can do the rest aligned
cmp len, 0
je return_pass
return_fail:
FUNC_RESTORE
mov return, 1
ret
endproc_frame
section .data
;;; func core, ver, snum
slversion xor_gen_avx, 02, 05, 0037
|
