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_2vect_mad_sse(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 return rax 49 %define return.w eax 50 %define stack_size 16*9 + 3*8 51 %define arg(x) [rsp + stack_size + PS + PS*x] 52 %define func(x) proc_frame x 53 54%macro FUNC_SAVE 0 55 sub rsp, stack_size 56 movdqa [rsp+16*0],xmm6 57 movdqa [rsp+16*1],xmm7 58 movdqa [rsp+16*2],xmm8 59 movdqa [rsp+16*3],xmm9 60 movdqa [rsp+16*4],xmm10 61 movdqa [rsp+16*5],xmm11 62 movdqa [rsp+16*6],xmm12 63 movdqa [rsp+16*7],xmm13 64 movdqa [rsp+16*8],xmm14 65 save_reg r12, 9*16 + 0*8 66 save_reg r15, 9*16 + 1*8 67 end_prolog 68 mov arg4, arg(4) 69 mov arg5, arg(5) 70%endmacro 71 72%macro FUNC_RESTORE 0 73 movdqa xmm6, [rsp+16*0] 74 movdqa xmm7, [rsp+16*1] 75 movdqa xmm8, [rsp+16*2] 76 movdqa xmm9, [rsp+16*3] 77 movdqa xmm10, [rsp+16*4] 78 movdqa xmm11, [rsp+16*5] 79 movdqa xmm12, [rsp+16*6] 80 movdqa xmm13, [rsp+16*7] 81 movdqa xmm14, [rsp+16*8] 82 mov r12, [rsp + 9*16 + 0*8] 83 mov r15, [rsp + 9*16 + 1*8] 84 add rsp, stack_size 85%endmacro 86 87%elifidn __OUTPUT_FORMAT__, elf64 88 %define arg0 rdi 89 %define arg0.w edi 90 %define arg1 rsi 91 %define arg2 rdx 92 %define arg3 rcx 93 %define arg4 r8 94 %define arg5 r9 95 %define tmp r11 96 %define tmp2 r10 97 %define return rax 98 %define return.w eax 99 100 %define func(x) x: 101 %define FUNC_SAVE 102 %define FUNC_RESTORE 103%endif 104 105;;; gf_2vect_mad_sse(len, vec, vec_i, mul_array, src, dest) 106%define len arg0 107%define len.w arg0.w 108%define vec arg1 109%define vec_i arg2 110%define mul_array arg3 111%define src arg4 112%define dest1 arg5 113%define pos return 114%define pos.w return.w 115 116%define dest2 tmp2 117 118%ifndef EC_ALIGNED_ADDR 119;;; Use Un-aligned load/store 120 %define XLDR movdqu 121 %define XSTR movdqu 122%else 123;;; Use Non-temporal load/stor 124 %ifdef NO_NT_LDST 125 %define XLDR movdqa 126 %define XSTR movdqa 127 %else 128 %define XLDR movntdqa 129 %define XSTR movntdq 130 %endif 131%endif 132 133default rel 134 135[bits 64] 136section .text 137 138%define xmask0f xmm14 139%define xgft1_lo xmm13 140%define xgft1_hi xmm12 141%define xgft2_lo xmm11 142%define xgft2_hi xmm10 143 144%define x0 xmm0 145%define xtmpa xmm1 146%define xtmph1 xmm2 147%define xtmpl1 xmm3 148%define xtmph2 xmm4 149%define xtmpl2 xmm5 150%define xd1 xmm6 151%define xd2 xmm7 152%define xtmpd1 xmm8 153%define xtmpd2 xmm9 154 155 156align 16 157global gf_2vect_mad_sse:ISAL_SYM_TYPE_FUNCTION 158func(gf_2vect_mad_sse) 159 FUNC_SAVE 160 sub len, 16 161 jl .return_fail 162 163 xor pos, pos 164 movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte 165 sal vec_i, 5 ;Multiply by 32 166 sal vec, 5 167 lea tmp, [mul_array + vec_i] 168 movdqu xgft1_lo,[tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ... 169 movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0} 170 movdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ... 171 movdqu xgft2_hi, [tmp+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0} 172 mov dest2, [dest1+PS] 173 mov dest1, [dest1] 174 175 XLDR xtmpd1, [dest1+len] ;backup the last 16 bytes in dest 176 XLDR xtmpd2, [dest2+len] ;backup the last 16 bytes in dest 177 178.loop16: 179 XLDR xd1, [dest1+pos] ;Get next dest vector 180 XLDR xd2, [dest2+pos] ;Get next dest vector 181.loop16_overlap: 182 XLDR x0, [src+pos] ;Get next source vector 183 movdqa xtmph1, xgft1_hi ;Reload const array registers 184 movdqa xtmpl1, xgft1_lo 185 movdqa xtmph2, xgft2_hi ;Reload const array registers 186 movdqa xtmpl2, xgft2_lo 187 movdqa xtmpa, x0 ;Keep unshifted copy of src 188 psraw x0, 4 ;Shift to put high nibble into bits 4-0 189 pand x0, xmask0f ;Mask high src nibble in bits 4-0 190 pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0 191 192 pshufb xtmph1, x0 ;Lookup mul table of high nibble 193 pshufb xtmpl1, xtmpa ;Lookup mul table of low nibble 194 pxor xtmph1, xtmpl1 ;GF add high and low partials 195 pxor xd1, xtmph1 196 197 pshufb xtmph2, x0 ;Lookup mul table of high nibble 198 pshufb xtmpl2, xtmpa ;Lookup mul table of low nibble 199 pxor xtmph2, xtmpl2 ;GF add high and low partials 200 pxor xd2, xtmph2 201 202 XSTR [dest1+pos], xd1 ;Store result 203 XSTR [dest2+pos], xd2 ;Store result 204 205 add pos, 16 ;Loop on 16 bytes at a time 206 cmp pos, len 207 jle .loop16 208 209 lea tmp, [len + 16] 210 cmp pos, tmp 211 je .return_pass 212 213 ;; Tail len 214 mov pos, len ;Overlapped offset length-16 215 movdqa xd1, xtmpd1 ;Restore xd1 216 movdqa xd2, xtmpd2 ;Restore xd2 217 jmp .loop16_overlap ;Do one more overlap pass 218 219.return_pass: 220 FUNC_RESTORE 221 mov return, 0 222 ret 223 224.return_fail: 225 FUNC_RESTORE 226 mov return, 1 227 ret 228 229endproc_frame 230 231section .data 232 233align 16 234 235mask0f: 236 dq 0x0f0f0f0f0f0f0f0f, 0x0f0f0f0f0f0f0f0f 237 238;;; func core, ver, snum 239slversion gf_2vect_mad_sse, 00, 01, 0203 240