]>
Commit | Line | Data |
---|---|---|
9f95a23c TL |
1 | ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; |
2 | ; Copyright(c) 2011-2015 Intel Corporation All rights reserved. | |
3 | ; | |
4 | ; Redistribution and use in source and binary forms, with or without | |
5 | ; modification, are permitted provided that the following conditions | |
6 | ; are met: | |
7 | ; * Redistributions of source code must retain the above copyright | |
8 | ; notice, this list of conditions and the following disclaimer. | |
9 | ; * Redistributions in binary form must reproduce the above copyright | |
10 | ; notice, this list of conditions and the following disclaimer in | |
11 | ; the documentation and/or other materials provided with the | |
12 | ; distribution. | |
13 | ; * Neither the name of Intel Corporation nor the names of its | |
14 | ; contributors may be used to endorse or promote products derived | |
15 | ; from this software without specific prior written permission. | |
16 | ; | |
17 | ; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
18 | ; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
19 | ; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
20 | ; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
21 | ; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
22 | ; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
23 | ; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
24 | ; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
25 | ; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
26 | ; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
27 | ; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
28 | ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
29 | ||
30 | ;;; | |
31 | ;;; gf_6vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests); | |
32 | ;;; | |
33 | ||
34 | %include "reg_sizes.asm" | |
35 | ||
36 | %ifidn __OUTPUT_FORMAT__, elf64 | |
37 | %define arg0 rdi | |
38 | %define arg1 rsi | |
39 | %define arg2 rdx | |
40 | %define arg3 rcx | |
41 | %define arg4 r8 | |
42 | %define arg5 r9 | |
43 | ||
44 | %define tmp r11 | |
45 | %define tmp2 r10 | |
46 | %define tmp3 r13 ; must be saved and restored | |
47 | %define tmp4 r12 ; must be saved and restored | |
48 | %define tmp5 r14 ; must be saved and restored | |
49 | %define tmp6 r15 ; must be saved and restored | |
50 | %define return rax | |
51 | %define PS 8 | |
52 | %define LOG_PS 3 | |
53 | ||
54 | %define func(x) x: | |
55 | %macro FUNC_SAVE 0 | |
56 | push r12 | |
57 | push r13 | |
58 | push r14 | |
59 | push r15 | |
60 | %endmacro | |
61 | %macro FUNC_RESTORE 0 | |
62 | pop r15 | |
63 | pop r14 | |
64 | pop r13 | |
65 | pop r12 | |
66 | %endmacro | |
67 | %endif | |
68 | ||
69 | %ifidn __OUTPUT_FORMAT__, win64 | |
70 | %define arg0 rcx | |
71 | %define arg1 rdx | |
72 | %define arg2 r8 | |
73 | %define arg3 r9 | |
74 | ||
75 | %define arg4 r12 ; must be saved, loaded and restored | |
76 | %define arg5 r15 ; must be saved and restored | |
77 | %define tmp r11 | |
78 | %define tmp2 r10 | |
79 | %define tmp3 r13 ; must be saved and restored | |
80 | %define tmp4 r14 ; must be saved and restored | |
81 | %define tmp5 rdi ; must be saved and restored | |
82 | %define tmp6 rsi ; must be saved and restored | |
83 | %define return rax | |
84 | %define PS 8 | |
85 | %define LOG_PS 3 | |
86 | %define stack_size 10*16 + 7*8 ; must be an odd multiple of 8 | |
87 | %define arg(x) [rsp + stack_size + PS + PS*x] | |
88 | ||
89 | %define func(x) proc_frame x | |
90 | %macro FUNC_SAVE 0 | |
91 | alloc_stack stack_size | |
92 | save_xmm128 xmm6, 0*16 | |
93 | save_xmm128 xmm7, 1*16 | |
94 | save_xmm128 xmm8, 2*16 | |
95 | save_xmm128 xmm9, 3*16 | |
96 | save_xmm128 xmm10, 4*16 | |
97 | save_xmm128 xmm11, 5*16 | |
98 | save_xmm128 xmm12, 6*16 | |
99 | save_xmm128 xmm13, 7*16 | |
100 | save_xmm128 xmm14, 8*16 | |
101 | save_xmm128 xmm15, 9*16 | |
102 | save_reg r12, 10*16 + 0*8 | |
103 | save_reg r13, 10*16 + 1*8 | |
104 | save_reg r14, 10*16 + 2*8 | |
105 | save_reg r15, 10*16 + 3*8 | |
106 | save_reg rdi, 10*16 + 4*8 | |
107 | save_reg rsi, 10*16 + 5*8 | |
108 | end_prolog | |
109 | mov arg4, arg(4) | |
110 | %endmacro | |
111 | ||
112 | %macro FUNC_RESTORE 0 | |
113 | vmovdqa xmm6, [rsp + 0*16] | |
114 | vmovdqa xmm7, [rsp + 1*16] | |
115 | vmovdqa xmm8, [rsp + 2*16] | |
116 | vmovdqa xmm9, [rsp + 3*16] | |
117 | vmovdqa xmm10, [rsp + 4*16] | |
118 | vmovdqa xmm11, [rsp + 5*16] | |
119 | vmovdqa xmm12, [rsp + 6*16] | |
120 | vmovdqa xmm13, [rsp + 7*16] | |
121 | vmovdqa xmm14, [rsp + 8*16] | |
122 | vmovdqa xmm15, [rsp + 9*16] | |
123 | mov r12, [rsp + 10*16 + 0*8] | |
124 | mov r13, [rsp + 10*16 + 1*8] | |
125 | mov r14, [rsp + 10*16 + 2*8] | |
126 | mov r15, [rsp + 10*16 + 3*8] | |
127 | mov rdi, [rsp + 10*16 + 4*8] | |
128 | mov rsi, [rsp + 10*16 + 5*8] | |
129 | add rsp, stack_size | |
130 | %endmacro | |
131 | %endif | |
132 | ||
133 | %define len arg0 | |
134 | %define vec arg1 | |
135 | %define mul_array arg2 | |
136 | %define src arg3 | |
137 | %define dest arg4 | |
138 | %define ptr arg5 | |
139 | %define vec_i tmp2 | |
140 | %define dest1 tmp3 | |
141 | %define dest2 tmp4 | |
142 | %define vskip1 tmp5 | |
143 | %define vskip3 tmp6 | |
144 | %define pos return | |
145 | ||
146 | ||
147 | %ifndef EC_ALIGNED_ADDR | |
148 | ;;; Use Un-aligned load/store | |
149 | %define XLDR vmovdqu | |
150 | %define XSTR vmovdqu | |
151 | %else | |
152 | ;;; Use Non-temporal load/stor | |
153 | %ifdef NO_NT_LDST | |
154 | %define XLDR vmovdqa | |
155 | %define XSTR vmovdqa | |
156 | %else | |
157 | %define XLDR vmovntdqa | |
158 | %define XSTR vmovntdq | |
159 | %endif | |
160 | %endif | |
161 | ||
162 | ||
163 | default rel | |
164 | ||
165 | [bits 64] | |
166 | section .text | |
167 | ||
168 | %define xmask0f xmm15 | |
169 | %define xgft1_lo xmm14 | |
170 | %define xgft1_hi xmm13 | |
171 | %define xgft2_lo xmm12 | |
172 | %define xgft2_hi xmm11 | |
173 | %define xgft3_lo xmm10 | |
174 | %define xgft3_hi xmm9 | |
175 | %define x0 xmm0 | |
176 | %define xtmpa xmm1 | |
177 | %define xp1 xmm2 | |
178 | %define xp2 xmm3 | |
179 | %define xp3 xmm4 | |
180 | %define xp4 xmm5 | |
181 | %define xp5 xmm6 | |
182 | %define xp6 xmm7 | |
183 | ||
184 | align 16 | |
f67539c2 | 185 | global gf_6vect_dot_prod_avx:ISAL_SYM_TYPE_FUNCTION |
9f95a23c TL |
186 | func(gf_6vect_dot_prod_avx) |
187 | FUNC_SAVE | |
188 | sub len, 16 | |
189 | jl .return_fail | |
190 | xor pos, pos | |
191 | vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte | |
192 | mov vskip1, vec | |
193 | imul vskip1, 32 | |
194 | mov vskip3, vec | |
195 | imul vskip3, 96 | |
196 | sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS | |
197 | mov dest1, [dest] | |
198 | mov dest2, [dest+PS] | |
199 | ||
200 | ||
201 | .loop16: | |
202 | mov tmp, mul_array | |
203 | xor vec_i, vec_i | |
204 | vpxor xp1, xp1 | |
205 | vpxor xp2, xp2 | |
206 | vpxor xp3, xp3 | |
207 | vpxor xp4, xp4 | |
208 | vpxor xp5, xp5 | |
209 | vpxor xp6, xp6 | |
210 | ||
211 | .next_vect: | |
212 | mov ptr, [src+vec_i] | |
213 | add vec_i, PS | |
214 | XLDR x0, [ptr+pos] ;Get next source vector | |
215 | ||
216 | vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f} | |
217 | vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0} | |
218 | vmovdqu xgft2_lo, [tmp+vskip1*1] ;Load array Bx{00}, Bx{01}, ..., Bx{0f} | |
219 | vmovdqu xgft2_hi, [tmp+vskip1*1+16] ; " Bx{00}, Bx{10}, ..., Bx{f0} | |
220 | vmovdqu xgft3_lo, [tmp+vskip1*2] ;Load array Cx{00}, Cx{01}, ..., Cx{0f} | |
221 | vmovdqu xgft3_hi, [tmp+vskip1*2+16] ; " Cx{00}, Cx{10}, ..., Cx{f0} | |
222 | lea ptr, [vskip1 + vskip1*4] ;ptr = vskip5 | |
223 | ||
224 | vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0 | |
225 | vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0 | |
226 | vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0 | |
227 | ||
228 | ||
229 | vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble | |
230 | vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble | |
231 | vpxor xgft1_hi, xgft1_lo ;GF add high and low partials | |
232 | vpxor xp1, xgft1_hi ;xp1 += partial | |
233 | ||
234 | vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble | |
235 | vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble | |
236 | vpxor xgft2_hi, xgft2_lo ;GF add high and low partials | |
237 | vpxor xp2, xgft2_hi ;xp2 += partial | |
238 | ||
239 | vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble | |
240 | vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble | |
241 | vpxor xgft3_hi, xgft3_lo ;GF add high and low partials | |
242 | vpxor xp3, xgft3_hi ;xp3 += partial | |
243 | ||
244 | ||
245 | vmovdqu xgft1_lo, [tmp+vskip3] ;Load array Dx{00}, Dx{01}, ..., Dx{0f} | |
246 | vmovdqu xgft1_hi, [tmp+vskip3+16] ; " Dx{00}, Dx{10}, ..., Dx{f0} | |
247 | vmovdqu xgft2_lo, [tmp+vskip1*4] ;Load array Ex{00}, Ex{01}, ..., Ex{0f} | |
248 | vmovdqu xgft2_hi, [tmp+vskip1*4+16] ; " Ex{00}, Ex{10}, ..., Ex{f0} | |
249 | vmovdqu xgft3_lo, [tmp+ptr] ;Load array Fx{00}, Fx{01}, ..., Fx{0f} | |
250 | vmovdqu xgft3_hi, [tmp+ptr+16] ; " Fx{00}, Fx{10}, ..., Fx{f0} | |
251 | add tmp, 32 | |
252 | ||
253 | ||
254 | vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble | |
255 | vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble | |
256 | vpxor xgft1_hi, xgft1_lo ;GF add high and low partials | |
257 | vpxor xp4, xgft1_hi ;xp4 += partial | |
258 | ||
259 | vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble | |
260 | vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble | |
261 | vpxor xgft2_hi, xgft2_lo ;GF add high and low partials | |
262 | vpxor xp5, xgft2_hi ;xp5 += partial | |
263 | ||
264 | vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble | |
265 | vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble | |
266 | vpxor xgft3_hi, xgft3_lo ;GF add high and low partials | |
267 | vpxor xp6, xgft3_hi ;xp6 += partial | |
268 | ||
269 | cmp vec_i, vec | |
270 | jl .next_vect | |
271 | ||
272 | ||
273 | mov tmp, [dest+2*PS] | |
274 | mov ptr, [dest+3*PS] | |
275 | mov vec_i, [dest+4*PS] | |
276 | ||
277 | XSTR [dest1+pos], xp1 | |
278 | XSTR [dest2+pos], xp2 | |
279 | XSTR [tmp+pos], xp3 | |
280 | mov tmp, [dest+5*PS] | |
281 | XSTR [ptr+pos], xp4 | |
282 | XSTR [vec_i+pos], xp5 | |
283 | XSTR [tmp+pos], xp6 | |
284 | ||
285 | add pos, 16 ;Loop on 16 bytes at a time | |
286 | cmp pos, len | |
287 | jle .loop16 | |
288 | ||
289 | lea tmp, [len + 16] | |
290 | cmp pos, tmp | |
291 | je .return_pass | |
292 | ||
293 | ;; Tail len | |
294 | mov pos, len ;Overlapped offset length-16 | |
295 | jmp .loop16 ;Do one more overlap pass | |
296 | ||
297 | .return_pass: | |
298 | FUNC_RESTORE | |
299 | mov return, 0 | |
300 | ret | |
301 | ||
302 | .return_fail: | |
303 | FUNC_RESTORE | |
304 | mov return, 1 | |
305 | ret | |
306 | ||
307 | endproc_frame | |
308 | ||
309 | section .data | |
310 | ||
311 | align 16 | |
312 | mask0f: dq 0x0f0f0f0f0f0f0f0f, 0x0f0f0f0f0f0f0f0f | |
313 | ||
314 | ;;; func core, ver, snum | |
315 | slversion gf_6vect_dot_prod_avx, 02, 04, 0195 |