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_mad_avx(len, vec, vec_i, mul_array, src, dest); 32;;; 33 34%include "reg_sizes.asm" 35 36%define PS 8 37 38%ifidn __OUTPUT_FORMAT__, win64 39 %define arg0 rcx 40 %define arg0.w ecx 41 %define arg1 rdx 42 %define arg2 r8 43 %define arg3 r9 44 %define arg4 r12 45 %define arg5 r15 46 %define tmp r11 47 %define tmp2 r10 48 %define tmp3 r13 49 %define tmp4 r14 50 %define tmp5 rdi 51 %define return rax 52 %define return.w eax 53 %define stack_size 16*10 + 5*8 54 %define arg(x) [rsp + stack_size + PS + PS*x] 55 %define func(x) proc_frame x 56 57%macro FUNC_SAVE 0 58 sub rsp, stack_size 59 movdqa [rsp+16*0],xmm6 60 movdqa [rsp+16*1],xmm7 61 movdqa [rsp+16*2],xmm8 62 movdqa [rsp+16*3],xmm9 63 movdqa [rsp+16*4],xmm10 64 movdqa [rsp+16*5],xmm11 65 movdqa [rsp+16*6],xmm12 66 movdqa [rsp+16*7],xmm13 67 movdqa [rsp+16*8],xmm14 68 movdqa [rsp+16*9],xmm15 69 save_reg r12, 10*16 + 0*8 70 save_reg r13, 10*16 + 1*8 71 save_reg r14, 10*16 + 2*8 72 save_reg r15, 10*16 + 3*8 73 save_reg rdi, 10*16 + 4*8 74 end_prolog 75 mov arg4, arg(4) 76 mov arg5, arg(5) 77%endmacro 78 79%macro FUNC_RESTORE 0 80 movdqa xmm6, [rsp+16*0] 81 movdqa xmm7, [rsp+16*1] 82 movdqa xmm8, [rsp+16*2] 83 movdqa xmm9, [rsp+16*3] 84 movdqa xmm10, [rsp+16*4] 85 movdqa xmm11, [rsp+16*5] 86 movdqa xmm12, [rsp+16*6] 87 movdqa xmm13, [rsp+16*7] 88 movdqa xmm14, [rsp+16*8] 89 movdqa xmm15, [rsp+16*9] 90 mov r12, [rsp + 10*16 + 0*8] 91 mov r13, [rsp + 10*16 + 1*8] 92 mov r14, [rsp + 10*16 + 2*8] 93 mov r15, [rsp + 10*16 + 3*8] 94 mov rdi, [rsp + 10*16 + 4*8] 95 add rsp, stack_size 96%endmacro 97 98%elifidn __OUTPUT_FORMAT__, elf64 99 %define arg0 rdi 100 %define arg0.w edi 101 %define arg1 rsi 102 %define arg2 rdx 103 %define arg3 rcx 104 %define arg4 r8 105 %define arg5 r9 106 %define tmp r11 107 %define tmp2 r10 108 %define tmp3 r12 109 %define tmp4 r13 110 %define tmp5 r14 111 %define return rax 112 %define return.w eax 113 114 %define func(x) x: 115 %macro FUNC_SAVE 0 116 push r12 117 push r13 118 push r14 119 %endmacro 120 %macro FUNC_RESTORE 0 121 pop r14 122 pop r13 123 pop r12 124 %endmacro 125%endif 126 127;;; gf_6vect_mad_avx(len, vec, vec_i, mul_array, src, dest) 128%define len arg0 129%define len.w arg0.w 130%define vec arg1 131%define vec_i arg2 132%define mul_array arg3 133%define src arg4 134%define dest1 arg5 135%define pos return 136%define pos.w return.w 137 138%define dest2 tmp4 139%define dest3 tmp2 140%define dest4 mul_array 141%define dest5 tmp5 142%define dest6 vec_i 143 144%ifndef EC_ALIGNED_ADDR 145;;; Use Un-aligned load/store 146 %define XLDR vmovdqu 147 %define XSTR vmovdqu 148%else 149;;; Use Non-temporal load/stor 150 %ifdef NO_NT_LDST 151 %define XLDR vmovdqa 152 %define XSTR vmovdqa 153 %else 154 %define XLDR vmovntdqa 155 %define XSTR vmovntdq 156 %endif 157%endif 158 159 160default rel 161 162[bits 64] 163section .text 164 165%define xmask0f xmm15 166%define xgft4_lo xmm14 167%define xgft4_hi xmm13 168%define xgft5_lo xmm12 169%define xgft5_hi xmm11 170%define xgft6_lo xmm10 171%define xgft6_hi xmm9 172 173%define x0 xmm0 174%define xtmpa xmm1 175%define xtmph1 xmm2 176%define xtmpl1 xmm3 177%define xtmph2 xmm4 178%define xtmpl2 xmm5 179%define xtmph3 xmm6 180%define xtmpl3 xmm7 181%define xd1 xmm8 182%define xd2 xtmpl1 183%define xd3 xtmph1 184 185 186align 16 187global gf_6vect_mad_avx:function 188func(gf_6vect_mad_avx) 189 FUNC_SAVE 190 sub len, 16 191 jl .return_fail 192 xor pos, pos 193 vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte 194 mov tmp, vec 195 sal vec_i, 5 ;Multiply by 32 196 lea tmp3, [mul_array + vec_i] 197 sal tmp, 6 ;Multiply by 64 198 199 sal vec, 5 ;Multiply by 32 200 lea vec_i, [tmp + vec] ;vec_i = vec*96 201 lea mul_array, [tmp + vec_i] ;mul_array = vec*160 202 203 vmovdqu xgft5_lo, [tmp3+2*tmp] ;Load array Ex{00}, Ex{01}, ..., Ex{0f} 204 vmovdqu xgft5_hi, [tmp3+2*tmp+16] ; " Ex{00}, Ex{10}, ..., Ex{f0} 205 vmovdqu xgft4_lo, [tmp3+vec_i] ;Load array Dx{00}, Dx{01}, Dx{02}, ... 206 vmovdqu xgft4_hi, [tmp3+vec_i+16] ; " Dx{00}, Dx{10}, Dx{20}, ... , Dx{f0} 207 vmovdqu xgft6_lo, [tmp3+mul_array] ;Load array Fx{00}, Fx{01}, ..., Fx{0f} 208 vmovdqu xgft6_hi, [tmp3+mul_array+16] ; " Fx{00}, Fx{10}, ..., Fx{f0} 209 210 mov dest2, [dest1+PS] 211 mov dest3, [dest1+2*PS] 212 mov dest4, [dest1+3*PS] ; reuse mul_array 213 mov dest5, [dest1+4*PS] 214 mov dest6, [dest1+5*PS] ; reuse vec_i 215 mov dest1, [dest1] 216 217.loop16: 218 XLDR x0, [src+pos] ;Get next source vector 219 220 vmovdqu xtmpl1, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ... 221 vmovdqu xtmph1, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0} 222 vmovdqu xtmpl2, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ... 223 vmovdqu xtmph2, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0} 224 vmovdqu xtmpl3, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ... 225 vmovdqu xtmph3, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0} 226 XLDR xd1, [dest1+pos] ;Get next dest vector 227 228 vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0 229 vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0 230 vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0 231 232 233 ;dest1 234 vpshufb xtmph1, x0 ;Lookup mul table of high nibble 235 vpshufb xtmpl1, xtmpa ;Lookup mul table of low nibble 236 vpxor xtmph1, xtmpl1 ;GF add high and low partials 237 vpxor xd1, xtmph1 238 239 XLDR xd2, [dest2+pos] ;reuse xtmpl1. Get next dest vector 240 XLDR xd3, [dest3+pos] ;reuse xtmph1. Get next dest vector 241 242 ;dest2 243 vpshufb xtmph2, x0 ;Lookup mul table of high nibble 244 vpshufb xtmpl2, xtmpa ;Lookup mul table of low nibble 245 vpxor xtmph2, xtmpl2 ;GF add high and low partials 246 vpxor xd2, xtmph2 247 248 ;dest3 249 vpshufb xtmph3, x0 ;Lookup mul table of high nibble 250 vpshufb xtmpl3, xtmpa ;Lookup mul table of low nibble 251 vpxor xtmph3, xtmpl3 ;GF add high and low partials 252 vpxor xd3, xtmph3 253 254 XSTR [dest1+pos], xd1 ;Store result into dest1 255 XSTR [dest2+pos], xd2 ;Store result into dest2 256 XSTR [dest3+pos], xd3 ;Store result into dest3 257 258 ;dest4 259 XLDR xd1, [dest4+pos] ;Get next dest vector 260 vpshufb xtmph1, xgft4_hi, x0 ;Lookup mul table of high nibble 261 vpshufb xtmpl1, xgft4_lo, xtmpa ;Lookup mul table of low nibble 262 vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials 263 vpxor xd1, xd1, xtmph1 264 265 XLDR xd2, [dest5+pos] ;reuse xtmpl1. Get next dest vector 266 XLDR xd3, [dest6+pos] ;reuse xtmph1. Get next dest vector 267 268 ;dest5 269 vpshufb xtmph2, xgft5_hi, x0 ;Lookup mul table of high nibble 270 vpshufb xtmpl2, xgft5_lo, xtmpa ;Lookup mul table of low nibble 271 vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials 272 vpxor xd2, xd2, xtmph2 273 274 ;dest6 275 vpshufb xtmph3, xgft6_hi, x0 ;Lookup mul table of high nibble 276 vpshufb xtmpl3, xgft6_lo, xtmpa ;Lookup mul table of low nibble 277 vpxor xtmph3, xtmph3, xtmpl3 ;GF add high and low partials 278 vpxor xd3, xd3, xtmph3 279 280 XSTR [dest4+pos], xd1 ;Store result into dest4 281 XSTR [dest5+pos], xd2 ;Store result into dest5 282 XSTR [dest6+pos], xd3 ;Store result into dest6 283 284 add pos, 16 ;Loop on 16 bytes at a time 285 cmp pos, len 286 jle .loop16 287 288 lea tmp, [len + 16] 289 cmp pos, tmp 290 je .return_pass 291 292.lessthan16: 293 ;; Tail len 294 ;; Do one more overlap pass 295 ;; Overlapped offset length-16 296 mov tmp, len ;Backup len as len=rdi 297 298 XLDR x0, [src+tmp] ;Get next source vector 299 XLDR xd1, [dest4+tmp] ;Get next dest vector 300 XLDR xd2, [dest5+tmp] ;reuse xtmpl1. Get next dest vector 301 XLDR xd3, [dest6+tmp] ;reuse xtmph1. Get next dest vector 302 303 sub len, pos 304 305 vmovdqa xtmph3, [constip16] ;Load const of i + 16 306 vpinsrb xtmpl3, len.w, 15 307 vpshufb xtmpl3, xmask0f ;Broadcast len to all bytes 308 vpcmpgtb xtmpl3, xtmpl3, xtmph3 309 310 vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0 311 vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0 312 vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0 313 314 ;dest4 315 vpshufb xgft4_hi, xgft4_hi, x0 ;Lookup mul table of high nibble 316 vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble 317 vpxor xgft4_hi, xgft4_hi, xgft4_lo ;GF add high and low partials 318 vpand xgft4_hi, xgft4_hi, xtmpl3 319 vpxor xd1, xd1, xgft4_hi 320 321 ;dest5 322 vpshufb xgft5_hi, xgft5_hi, x0 ;Lookup mul table of high nibble 323 vpshufb xgft5_lo, xgft5_lo, xtmpa ;Lookup mul table of low nibble 324 vpxor xgft5_hi, xgft5_hi, xgft5_lo ;GF add high and low partials 325 vpand xgft5_hi, xgft5_hi, xtmpl3 326 vpxor xd2, xd2, xgft5_hi 327 328 ;dest6 329 vpshufb xgft6_hi, xgft6_hi, x0 ;Lookup mul table of high nibble 330 vpshufb xgft6_lo, xgft6_lo, xtmpa ;Lookup mul table of low nibble 331 vpxor xgft6_hi, xgft6_hi, xgft6_lo ;GF add high and low partials 332 vpand xgft6_hi, xgft6_hi, xtmpl3 333 vpxor xd3, xd3, xgft6_hi 334 335 XSTR [dest4+tmp], xd1 ;Store result into dest4 336 XSTR [dest5+tmp], xd2 ;Store result into dest5 337 XSTR [dest6+tmp], xd3 ;Store result into dest6 338 339 vmovdqu xgft4_lo, [tmp3] ;Load array Ax{00}, Ax{01}, Ax{02}, ... 340 vmovdqu xgft4_hi, [tmp3+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0} 341 vmovdqu xgft5_lo, [tmp3+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ... 342 vmovdqu xgft5_hi, [tmp3+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0} 343 vmovdqu xgft6_lo, [tmp3+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ... 344 vmovdqu xgft6_hi, [tmp3+2*vec+16] ; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0} 345 XLDR xd1, [dest1+tmp] ;Get next dest vector 346 XLDR xd2, [dest2+tmp] ;reuse xtmpl1. Get next dest vector 347 XLDR xd3, [dest3+tmp] ;reuse xtmph1. Get next dest3 vector 348 349 ;dest1 350 vpshufb xgft4_hi, xgft4_hi, x0 ;Lookup mul table of high nibble 351 vpshufb xgft4_lo, xgft4_lo, xtmpa ;Lookup mul table of low nibble 352 vpxor xgft4_hi, xgft4_hi, xgft4_lo ;GF add high and low partials 353 vpand xgft4_hi, xgft4_hi, xtmpl3 354 vpxor xd1, xd1, xgft4_hi 355 356 ;dest2 357 vpshufb xgft5_hi, xgft5_hi, x0 ;Lookup mul table of high nibble 358 vpshufb xgft5_lo, xgft5_lo, xtmpa ;Lookup mul table of low nibble 359 vpxor xgft5_hi, xgft5_hi, xgft5_lo ;GF add high and low partials 360 vpand xgft5_hi, xgft5_hi, xtmpl3 361 vpxor xd2, xd2, xgft5_hi 362 363 ;dest3 364 vpshufb xgft6_hi, xgft6_hi, x0 ;Lookup mul table of high nibble 365 vpshufb xgft6_lo, xgft6_lo, xtmpa ;Lookup mul table of low nibble 366 vpxor xgft6_hi, xgft6_hi, xgft6_lo ;GF add high and low partials 367 vpand xgft6_hi, xgft6_hi, xtmpl3 368 vpxor xd3, xd3, xgft6_hi 369 370 XSTR [dest1+tmp], xd1 ;Store result into dest1 371 XSTR [dest2+tmp], xd2 ;Store result into dest2 372 XSTR [dest3+tmp], xd3 ;Store result into dest3 373 374.return_pass: 375 FUNC_RESTORE 376 mov return, 0 377 ret 378 379.return_fail: 380 FUNC_RESTORE 381 mov return, 1 382 ret 383 384endproc_frame 385 386section .data 387 388align 16 389mask0f: dq 0x0f0f0f0f0f0f0f0f, 0x0f0f0f0f0f0f0f0f 390constip16: 391 dq 0xf8f9fafbfcfdfeff, 0xf0f1f2f3f4f5f6f7 392 393;;; func core, ver, snum 394slversion gf_6vect_mad_avx, 02, 01, 0210 395