1 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2 ; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
4 ; Redistribution and use in source and binary forms, with or without
5 ; modification, are permitted provided that the following conditions
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
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.
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 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
31 ;;; gf_5vect_mad_sse(len, vec, vec_i, mul_array, src, dest);
34 %include "reg_sizes.asm"
38 %ifidn __OUTPUT_FORMAT__, win64
52 %define stack_size 16*10 + 5*8
53 %define arg(x) [rsp + stack_size + PS + PS*x]
54 %define func(x) proc_frame x
58 movdqa [rsp+16*0],xmm6
59 movdqa [rsp+16*1],xmm7
60 movdqa [rsp+16*2],xmm8
61 movdqa [rsp+16*3],xmm9
62 movdqa [rsp+16*4],xmm10
63 movdqa [rsp+16*5],xmm11
64 movdqa [rsp+16*6],xmm12
65 movdqa [rsp+16*7],xmm13
66 movdqa [rsp+16*8],xmm14
67 movdqa [rsp+16*9],xmm15
68 save_reg r12, 10*16 + 0*8
69 save_reg r13, 10*16 + 1*8
70 save_reg r14, 10*16 + 2*8
71 save_reg r15, 10*16 + 3*8
78 movdqa xmm6, [rsp+16*0]
79 movdqa xmm7, [rsp+16*1]
80 movdqa xmm8, [rsp+16*2]
81 movdqa xmm9, [rsp+16*3]
82 movdqa xmm10, [rsp+16*4]
83 movdqa xmm11, [rsp+16*5]
84 movdqa xmm12, [rsp+16*6]
85 movdqa xmm13, [rsp+16*7]
86 movdqa xmm14, [rsp+16*8]
87 movdqa xmm15, [rsp+16*9]
88 mov r12, [rsp + 10*16 + 0*8]
89 mov r13, [rsp + 10*16 + 1*8]
90 mov r14, [rsp + 10*16 + 2*8]
91 mov r15, [rsp + 10*16 + 3*8]
95 %elifidn __OUTPUT_FORMAT__, elf64
115 %macro FUNC_RESTORE 0
121 ;;; gf_5vect_mad_sse(len, vec, vec_i, mul_array, src, dest)
126 %define mul_array arg3
130 %define pos.w return.w
133 %define dest3 mul_array
137 %ifndef EC_ALIGNED_ADDR
138 ;;; Use Un-aligned load/store
142 ;;; Use Non-temporal load/stor
147 %define XLDR movntdqa
157 %define xmask0f xmm15
158 %define xgft5_hi xmm14
159 %define xgft4_lo xmm13
160 %define xgft4_hi xmm12
180 global gf_5vect_mad_sse:function
181 func(gf_5vect_mad_sse)
187 movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
189 sal vec_i, 5 ;Multiply by 32
190 lea tmp3, [mul_array + vec_i]
191 sal tmp, 6 ;Multiply by 64
192 movdqu xgft5_hi, [tmp3+2*tmp+16] ; " Ex{00}, Ex{10}, ..., Ex{f0}
193 sal vec, 5 ;Multiply by 32
195 movdqu xgft4_hi, [tmp3+tmp+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0}
196 movdqu xgft4_lo, [tmp3+tmp] ;Load array Dx{00}, Dx{01}, Dx{02}, ...
198 mov dest3, [dest1+2*PS] ; reuse mul_array
199 mov dest4, [dest1+3*PS]
200 mov dest5, [dest1+4*PS] ; reuse vec_i
201 mov dest2, [dest1+PS]
205 XLDR x0, [src+pos] ;Get next source vector
207 movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
208 movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
209 movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
210 movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
211 movdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
212 movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
213 movdqu xtmpl5, [tmp3+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
214 movdqa xtmph5, xgft5_hi ;Reload const array registers
216 XLDR xd1, [dest1+pos] ;Get next dest vector
217 XLDR xd2, [dest2+pos] ;Get next dest vector
219 movdqa xtmpa, x0 ;Keep unshifted copy of src
220 psraw x0, 4 ;Shift to put high nibble into bits 4-0
221 pand x0, xmask0f ;Mask high src nibble in bits 4-0
222 pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
225 pshufb xtmph1, x0 ;Lookup mul table of high nibble
226 pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
227 pxor xtmph1, xtmpl1 ;GF add high and low partials
230 XLDR xd3, [dest3+pos] ;Reuse xtmpl1, Get next dest vector
231 XLDR xd4, [dest4+pos] ;Reuse xtmph1. Get next dest vector
234 pshufb xtmph2, x0 ;Lookup mul table of high nibble
235 pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
236 pxor xtmph2, xtmpl2 ;GF add high and low partials
239 XLDR xd5, [dest5+pos] ;Reuse xtmpl2. Get next dest vector
242 pshufb xtmph3, x0 ;Lookup mul table of high nibble
243 pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
244 pxor xtmph3, xtmpl3 ;GF add high and low partials
247 movdqa xtmph2, xgft4_hi ;Reload const array registers
248 movdqa xtmpl3, xgft4_lo ;Reload const array registers
251 pshufb xtmph5, x0 ;Lookup mul table of high nibble
252 pshufb xtmpl5, xtmpa ;Lookup mul table of low nibble
253 pxor xtmph5, xtmpl5 ;GF add high and low partials
257 pshufb xtmph2, x0 ;Lookup mul table of high nibble
258 pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
259 pxor xtmph2, xtmpl3 ;GF add high and low partials
262 XSTR [dest1+pos], xd1 ;Store result into dest1
263 XSTR [dest2+pos], xd2 ;Store result into dest2
264 XSTR [dest3+pos], xd3 ;Store result into dest3
265 XSTR [dest4+pos], xd4 ;Store result into dest4
266 XSTR [dest5+pos], xd5 ;Store result into dest5
268 add pos, 16 ;Loop on 16 bytes at a time
278 ;; Do one more overlap pass
279 mov tmp, len ;Overlapped offset length-16
280 XLDR x0, [src+tmp] ;Get next source vector
284 movdqa xtmpl1, [constip16] ;Load const of i + 16
285 pinsrb xtmph5, len.w, 15
286 pshufb xtmph5, xmask0f ;Broadcast len to all bytes
287 pcmpgtb xtmph5, xtmpl1
289 movdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
290 movdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
291 movdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
292 movdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
293 movdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
294 movdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
295 movdqu xtmpl5, [tmp3+4*vec] ;Load array Ex{00}, Ex{01}, ..., Ex{0f}
297 XLDR xd1, [dest1+tmp] ;Get next dest vector
298 XLDR xd2, [dest2+tmp] ;Get next dest vector
300 movdqa xtmpa, x0 ;Keep unshifted copy of src
301 psraw x0, 4 ;Shift to put high nibble into bits 4-0
302 pand x0, xmask0f ;Mask high src nibble in bits 4-0
303 pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
306 pshufb xtmph1, x0 ;Lookup mul table of high nibble
307 pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble
308 pxor xtmph1, xtmpl1 ;GF add high and low partials
312 XLDR xd3, [dest3+tmp] ;Reuse xtmpl1, Get next dest vector
313 XLDR xd4, [dest4+tmp] ;Reuse xtmph1. Get next dest vector
316 pshufb xtmph2, x0 ;Lookup mul table of high nibble
317 pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
318 pxor xtmph2, xtmpl2 ;GF add high and low partials
322 XLDR xd5, [dest5+tmp] ;Reuse xtmpl2. Get next dest vector
325 pshufb xtmph3, x0 ;Lookup mul table of high nibble
326 pshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
327 pxor xtmph3, xtmpl3 ;GF add high and low partials
332 pshufb xgft4_hi, x0 ;Lookup mul table of high nibble
333 pshufb xgft4_lo, xtmpa ;Lookup mul table of low nibble
334 pxor xgft4_hi, xgft4_lo ;GF add high and low partials
335 pand xgft4_hi, xtmph5
339 pshufb xgft5_hi, x0 ;Lookup mul table of high nibble
340 pshufb xtmpl5, xtmpa ;Lookup mul table of low nibble
341 pxor xgft5_hi, xtmpl5 ;GF add high and low partials
342 pand xgft5_hi, xtmph5
345 XSTR [dest1+tmp], xd1 ;Store result into dest1
346 XSTR [dest2+tmp], xd2 ;Store result into dest2
347 XSTR [dest3+tmp], xd3 ;Store result into dest3
348 XSTR [dest4+tmp], xd4 ;Store result into dest4
349 XSTR [dest5+tmp], xd5 ;Store result into dest5
368 ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
370 ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
372 ;;; func core, ver, snum
373 slversion gf_5vect_mad_sse, 00, 01, 020c