1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/os/ossrv/genericopenlibs/liboil/src/i386/trans8x8_i386.c Fri Jun 15 03:10:57 2012 +0200
1.3 @@ -0,0 +1,267 @@
1.4 +/*
1.5 + * LIBOIL - Library of Optimized Inner Loops
1.6 + * Copyright (c) 2003,2004 David A. Schleef <ds@schleef.org>
1.7 + * All rights reserved.
1.8 + *
1.9 + * Redistribution and use in source and binary forms, with or without
1.10 + * modification, are permitted provided that the following conditions
1.11 + * are met:
1.12 + * 1. Redistributions of source code must retain the above copyright
1.13 + * notice, this list of conditions and the following disclaimer.
1.14 + * 2. Redistributions in binary form must reproduce the above copyright
1.15 + * notice, this list of conditions and the following disclaimer in the
1.16 + * documentation and/or other materials provided with the distribution.
1.17 + *
1.18 + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
1.19 + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1.20 + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
1.21 + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
1.22 + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1.23 + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1.24 + * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1.25 + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1.26 + * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1.27 + * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1.28 + * POSSIBILITY OF SUCH DAMAGE.
1.29 + */
1.30 +//Portions Copyright (c) 2008-2009 Nokia Corporation and/or its subsidiary(-ies). All rights reserved.
1.31 +
1.32 +#ifdef HAVE_CONFIG_H
1.33 +#include "config.h"
1.34 +#endif
1.35 +
1.36 +#include <liboil/liboilfunction.h>
1.37 +#include <math.h>
1.38 +
1.39 +OIL_DECLARE_CLASS(trans8x8_u16);
1.40 +
1.41 +/* this could use additional work. */
1.42 +static void
1.43 +trans8x8_u16_mmx (uint16_t *dest, int dstr, uint16_t *src, int sstr)
1.44 +{
1.45 +#if !defined(__WINSCW__) && !defined(__WINS__)
1.46 + asm volatile (
1.47 + " leal (%3,%3,2),%%eax \n" // UBER 0:
1.48 + " movq (%1), %%mm0 \n" // UBER 1:
1.49 + " movq (%1,%3,2), %%mm2 \n" // UBER 2:
1.50 + " movq %%mm0, %%mm4 \n" // UBER 3: 1
1.51 + " movq %%mm2, %%mm5 \n" // UBER 4: 2
1.52 + " punpcklwd (%1,%3), %%mm0 \n" // UBER 5: 1
1.53 + " punpcklwd (%1,%%eax), %%mm2 \n" // UBER 6: 0 2
1.54 + " punpckhwd (%1,%3), %%mm4 \n" // UBER 7: 3
1.55 + " punpckhwd (%1,%%eax), %%mm5 \n" // UBER 8: 4
1.56 + " movq %%mm0, %%mm1 \n" // UBER 9: 5
1.57 + " movq %%mm4, %%mm3 \n" // UBER 10: 7
1.58 + " punpckldq %%mm2, %%mm0 \n" // UBER 11: 5 6
1.59 + " punpckldq %%mm5, %%mm4 \n" // UBER 12: 7 8
1.60 + " punpckhdq %%mm2, %%mm1 \n" // UBER 13: 6 9
1.61 + " punpckhdq %%mm5, %%mm3 \n" // UBER 14: 9 10
1.62 + " leal (%2,%2,2),%%eax \n" // UBER 15: 8
1.63 + " movq %%mm0, 0(%0) \n" // UBER 16: 11
1.64 + " movq %%mm1, (%0,%2) \n" // UBER 17: 13
1.65 + " movq %%mm4, (%0,%2,2) \n" // UBER 18: 12
1.66 + " movq %%mm3, (%0,%%eax) \n" // UBER 19: 14 15
1.67 +
1.68 + " leal (%3,%3,2),%%eax \n"
1.69 + " movq 8(%1), %%mm0 \n"
1.70 + " movq 8(%1,%3,2), %%mm2 \n"
1.71 + " movq %%mm0, %%mm4 \n"
1.72 + " movq %%mm2, %%mm5 \n"
1.73 + " punpcklwd 8(%1,%3), %%mm0 \n"
1.74 + " punpcklwd 8(%1,%%eax), %%mm2 \n"
1.75 + " punpckhwd 8(%1,%3), %%mm4 \n"
1.76 + " punpckhwd 8(%1,%%eax), %%mm5 \n"
1.77 + " movq %%mm0, %%mm1 \n"
1.78 + " movq %%mm4, %%mm3 \n"
1.79 + " punpckldq %%mm2, %%mm0 \n"
1.80 + " punpckldq %%mm5, %%mm4 \n"
1.81 + " punpckhdq %%mm2, %%mm1 \n"
1.82 + " punpckhdq %%mm5, %%mm3 \n"
1.83 + " leal (%2,%2,2),%%eax \n"
1.84 + " leal (%0,%2,4),%0 \n"
1.85 + " movq %%mm0, 0(%0) \n"
1.86 + " movq %%mm1, (%0,%2) \n"
1.87 + " movq %%mm4, (%0,%2,2) \n"
1.88 + " movq %%mm3, (%0,%%eax) \n"
1.89 +
1.90 + " leal (%1,%3,4),%1 \n"
1.91 + " leal (%3,%3,2),%%eax \n"
1.92 + " movq 0(%1), %%mm0 \n"
1.93 + " movq 0(%1,%3,2), %%mm2 \n"
1.94 + " movq %%mm0, %%mm4 \n"
1.95 + " movq %%mm2, %%mm5 \n"
1.96 + " punpcklwd 0(%1,%3), %%mm0 \n"
1.97 + " punpcklwd 0(%1,%%eax), %%mm2 \n"
1.98 + " punpckhwd 0(%1,%3), %%mm4 \n"
1.99 + " punpckhwd 0(%1,%%eax), %%mm5 \n"
1.100 + " movq %%mm0, %%mm1 \n"
1.101 + " movq %%mm4, %%mm3 \n"
1.102 + " punpckldq %%mm2, %%mm0 \n"
1.103 + " punpckldq %%mm5, %%mm4 \n"
1.104 + " punpckhdq %%mm2, %%mm1 \n"
1.105 + " punpckhdq %%mm5, %%mm3 \n"
1.106 + " leal (%2,%2,2),%%eax \n"
1.107 + " neg %2 \n"
1.108 + " leal (%0,%2,4),%0 \n"
1.109 + " neg %2 \n"
1.110 + " movq %%mm0, 8(%0) \n"
1.111 + " movq %%mm1, 8(%0,%2) \n"
1.112 + " movq %%mm4, 8(%0,%2,2) \n"
1.113 + " movq %%mm3, 8(%0,%%eax) \n"
1.114 +
1.115 + " leal (%3,%3,2),%%eax \n"
1.116 + " movq 8(%1), %%mm0 \n"
1.117 + " movq 8(%1,%3,2), %%mm2 \n"
1.118 + " movq %%mm0, %%mm4 \n"
1.119 + " movq %%mm2, %%mm5 \n"
1.120 + " punpcklwd 8(%1,%3), %%mm0 \n"
1.121 + " punpcklwd 8(%1,%%eax), %%mm2 \n"
1.122 + " punpckhwd 8(%1,%3), %%mm4 \n"
1.123 + " punpckhwd 8(%1,%%eax), %%mm5 \n"
1.124 + " movq %%mm0, %%mm1 \n"
1.125 + " movq %%mm4, %%mm3 \n"
1.126 + " punpckldq %%mm2, %%mm0 \n"
1.127 + " punpckldq %%mm5, %%mm4 \n"
1.128 + " punpckhdq %%mm2, %%mm1 \n"
1.129 + " punpckhdq %%mm5, %%mm3 \n"
1.130 + " leal (%2,%2,2),%%eax \n"
1.131 + " leal (%0,%2,4),%0 \n"
1.132 + " movq %%mm0, 8(%0) \n"
1.133 + " movq %%mm1, 8(%0,%2) \n"
1.134 + " movq %%mm4, 8(%0,%2,2) \n"
1.135 + " movq %%mm3, 8(%0,%%eax) \n"
1.136 + " emms \n"
1.137 + : "+r" (dest), "+r" (src), "+r" (dstr), "+r" (sstr)
1.138 + :
1.139 + : "eax");
1.140 +#endif
1.141 +}
1.142 +OIL_DEFINE_IMPL_FULL (trans8x8_u16_mmx, trans8x8_u16, OIL_IMPL_FLAG_MMX);
1.143 +
1.144 +static void
1.145 +trans8x8_u16_asm1 (uint16_t *dest, int dstr, uint16_t *src, int sstr)
1.146 +{
1.147 +#if !defined(__WINSCW__) && !defined(__WINS__)
1.148 + int saved_ebx = 0;
1.149 + asm (
1.150 + " movl %%ebx, %4 \n"
1.151 + " movl %0, %%ecx \n"
1.152 + " movl %2, %%ebx \n"
1.153 + " movl %1, %%edx \n"
1.154 + " lea (%%ecx,%%edx,8), %%esi \n"
1.155 + " sub %%edx, %%esi\n "
1.156 + " movl $7, %%edi \n"
1.157 + "1: \n"
1.158 +
1.159 + " mov (%%ebx), %%ax \n"
1.160 + " mov %%ax,(%%ecx) \n"
1.161 + " mov 2(%%ebx), %%ax \n"
1.162 + " mov %%ax,(%%ecx,%%edx,1) \n"
1.163 + " mov 4(%%ebx), %%ax \n"
1.164 + " mov %%ax,(%%ecx,%%edx,2) \n"
1.165 + " mov 8(%%ebx), %%ax \n"
1.166 + " mov %%ax,(%%ecx,%%edx,4) \n"
1.167 +
1.168 + " neg %%edx \n"
1.169 +
1.170 + " mov 6(%%ebx), %%ax \n"
1.171 + " mov %%ax,(%%esi,%%edx,4) \n"
1.172 + " mov 10(%%ebx), %%ax \n"
1.173 + " mov %%ax,(%%esi,%%edx,2) \n"
1.174 + " mov 12(%%ebx), %%ax \n"
1.175 + " mov %%ax,(%%esi,%%edx,1) \n"
1.176 + " mov 14(%%ebx), %%ax \n"
1.177 + " mov %%ax,(%%esi) \n"
1.178 +
1.179 + " neg %%edx \n"
1.180 + " add %3, %%ebx \n"
1.181 + " add $2, %%ecx \n"
1.182 + " add $2, %%esi \n"
1.183 +
1.184 + " dec %%edi \n"
1.185 + " jge 1b \n"
1.186 + " movl %4, %%ebx \n"
1.187 + :
1.188 + : "m" (dest), "m" (dstr), "m" (src), "m" (sstr), "m" (saved_ebx)
1.189 + : "eax", "ecx", "edx", "esi", "edi");
1.190 +#endif
1.191 +}
1.192 +OIL_DEFINE_IMPL (trans8x8_u16_asm1, trans8x8_u16);
1.193 +
1.194 +static void
1.195 +trans8x8_u16_asm2 (uint16_t *dest, int dstr, uint16_t *src, int sstr)
1.196 +{
1.197 +#if !defined(__WINSCW__) && !defined(__WINS__)
1.198 + int i;
1.199 + int saved_ebx = 0;
1.200 + asm (
1.201 + " movl %%ebx, %5 \n"
1.202 + " movl %0, %%ecx \n"
1.203 + " movl %2, %%ebx \n"
1.204 + " movl %1, %%edx \n"
1.205 + " lea (%%ecx,%%edx,8), %%esi \n"
1.206 + " sub %%edx, %%esi\n "
1.207 + " movl $7, %4 \n"
1.208 + " movl %%edx, %%edi \n"
1.209 + " negl %%edi \n"
1.210 + "1: \n"
1.211 +
1.212 + " movl (%%ebx), %%eax \n"
1.213 + " mov %%ax,(%%ecx) \n"
1.214 + " shr $16, %%eax \n"
1.215 + " mov %%ax,(%%ecx,%%edx,1) \n"
1.216 +
1.217 + " movl 4(%%ebx), %%eax \n"
1.218 + " mov %%ax,(%%ecx,%%edx,2) \n"
1.219 + " shr $16, %%eax \n"
1.220 + " mov %%ax,(%%esi,%%edi,4) \n"
1.221 +
1.222 + " movl 8(%%ebx), %%eax \n"
1.223 + " mov %%ax,(%%ecx,%%edx,4) \n"
1.224 + " shr $16, %%eax \n"
1.225 + " mov %%ax,(%%esi,%%edi,2) \n"
1.226 +
1.227 + " movl 12(%%ebx), %%eax \n"
1.228 + " mov %%ax,(%%esi,%%edi,1) \n"
1.229 + " shr $16, %%eax \n"
1.230 + " mov %%ax,(%%esi) \n"
1.231 +
1.232 + " add %3, %%ebx \n"
1.233 + " add $2, %%ecx \n"
1.234 + " add $2, %%esi \n"
1.235 +
1.236 + " decl %4 \n"
1.237 + " jge 1b \n"
1.238 + " movl %5, %%ebx \n"
1.239 + :
1.240 + : "m" (dest), "m" (dstr), "m" (src), "m" (sstr), "m" (i), "m" (saved_ebx)
1.241 + : "eax", "ecx", "edx", "esi", "edi");
1.242 +#endif
1.243 +}
1.244 +OIL_DEFINE_IMPL (trans8x8_u16_asm2, trans8x8_u16);
1.245 +
1.246 +
1.247 +
1.248 +#ifdef __SYMBIAN32__
1.249 +
1.250 +OilFunctionImpl* __oil_function_impl_trans8x8_u16_mmx, trans8x8_u16() {
1.251 + return &_oil_function_impl_trans8x8_u16_mmx, trans8x8_u16;
1.252 +}
1.253 +#endif
1.254 +
1.255 +
1.256 +
1.257 +#ifdef __SYMBIAN32__
1.258 +
1.259 +OilFunctionImpl* __oil_function_impl_trans8x8_u16_asm1() {
1.260 + return &_oil_function_impl_trans8x8_u16_asm1;
1.261 +}
1.262 +#endif
1.263 +
1.264 +#ifdef __SYMBIAN32__
1.265 +
1.266 +OilFunctionImpl* __oil_function_impl_trans8x8_u16_asm2() {
1.267 + return &_oil_function_impl_trans8x8_u16_asm2;
1.268 +}
1.269 +#endif
1.270 +