FFmpeg: libswscale/x86/swscale

00001 /*
00002  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
00003  *
00004  * This file is part of FFmpeg.
00005  *
00006  * FFmpeg is free software; you can redistribute it and/or
00007  * modify it under the terms of the GNU Lesser General Public
00008  * License as published by the Free Software Foundation; either
00009  * version 2.1 of the License, or (at your option) any later version.
00010  *
00011  * FFmpeg is distributed in the hope that it will be useful,
00012  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00013  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00014  * Lesser General Public License for more details.
00015  *
00016  * You should have received a copy of the GNU Lesser General Public
00017  * License along with FFmpeg; if not, write to the Free Software
00018  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00019  */
00020 
00021 #include <inttypes.h>
00022 #include "config.h"
00023 #include "libswscale/swscale.h"
00024 #include "libswscale/swscale_internal.h"
00025 #include "libavutil/intreadwrite.h"
00026 #include "libavutil/x86_cpu.h"
00027 #include "libavutil/cpu.h"
00028 #include "libavutil/pixdesc.h"
00029 
00030 DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
00031 DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
00032 DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
00033 DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
00034 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
00035 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
00036 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
00037 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
00038 
00039 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
00040     0x0103010301030103LL,
00041     0x0200020002000200LL,};
00042 
00043 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
00044     0x0602060206020602LL,
00045     0x0004000400040004LL,};
00046 
00047 DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
00048 DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
00049 DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
00050 DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
00051 DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
00052 DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
00053 
00054 DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
00055 DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
00056 DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
00057 
00058 #ifdef FAST_BGR2YV12
00059 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
00060 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
00061 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
00062 #else
00063 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
00064 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
00065 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
00066 #endif /* FAST_BGR2YV12 */
00067 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
00068 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
00069 DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
00070 
00071 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
00072 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
00073 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
00074 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
00075 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008010000080100ULL;
00076 
00077 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
00078     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
00079     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
00080 };
00081 
00082 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040010000400100ULL;
00083 
00084 //MMX versions
00085 #if HAVE_MMX
00086 #undef RENAME
00087 #define COMPILE_TEMPLATE_MMX2 0
00088 #define RENAME(a) a ## _MMX
00089 #include "swscale_template.c"
00090 #endif
00091 
00092 //MMX2 versions
00093 #if HAVE_MMX2
00094 #undef RENAME
00095 #undef COMPILE_TEMPLATE_MMX2
00096 #define COMPILE_TEMPLATE_MMX2 1
00097 #define RENAME(a) a ## _MMX2
00098 #include "swscale_template.c"
00099 #endif
00100 
00101 void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex,
00102                            int lastInLumBuf, int lastInChrBuf)
00103 {
00104     const int dstH= c->dstH;
00105     const int flags= c->flags;
00106     int16_t **lumPixBuf= c->lumPixBuf;
00107     int16_t **chrUPixBuf= c->chrUPixBuf;
00108     int16_t **alpPixBuf= c->alpPixBuf;
00109     const int vLumBufSize= c->vLumBufSize;
00110     const int vChrBufSize= c->vChrBufSize;
00111     int16_t *vLumFilterPos= c->vLumFilterPos;
00112     int16_t *vChrFilterPos= c->vChrFilterPos;
00113     int16_t *vLumFilter= c->vLumFilter;
00114     int16_t *vChrFilter= c->vChrFilter;
00115     int32_t *lumMmxFilter= c->lumMmxFilter;
00116     int32_t *chrMmxFilter= c->chrMmxFilter;
00117     int32_t av_unused *alpMmxFilter= c->alpMmxFilter;
00118     const int vLumFilterSize= c->vLumFilterSize;
00119     const int vChrFilterSize= c->vChrFilterSize;
00120     const int chrDstY= dstY>>c->chrDstVSubSample;
00121     const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input
00122     const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input
00123 
00124     c->blueDither= ff_dither8[dstY&1];
00125     if (c->dstFormat == PIX_FMT_RGB555 || c->dstFormat == PIX_FMT_BGR555)
00126         c->greenDither= ff_dither8[dstY&1];
00127     else
00128         c->greenDither= ff_dither4[dstY&1];
00129     c->redDither= ff_dither8[(dstY+1)&1];
00130     if (dstY < dstH - 2) {
00131         const int16_t **lumSrcPtr= (const int16_t **) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
00132         const int16_t **chrUSrcPtr= (const int16_t **) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
00133         const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
00134         int i;
00135 
00136         if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->srcH) {
00137             const int16_t **tmpY = (const int16_t **) lumPixBuf + 2 * vLumBufSize;
00138             int neg = -firstLumSrcY, i, end = FFMIN(c->srcH - firstLumSrcY, vLumFilterSize);
00139             for (i = 0; i < neg;            i++)
00140                 tmpY[i] = lumSrcPtr[neg];
00141             for (     ; i < end;            i++)
00142                 tmpY[i] = lumSrcPtr[i];
00143             for (     ; i < vLumFilterSize; i++)
00144                 tmpY[i] = tmpY[i-1];
00145             lumSrcPtr = tmpY;
00146 
00147             if (alpSrcPtr) {
00148                 const int16_t **tmpA = (const int16_t **) alpPixBuf + 2 * vLumBufSize;
00149                 for (i = 0; i < neg;            i++)
00150                     tmpA[i] = alpSrcPtr[neg];
00151                 for (     ; i < end;            i++)
00152                     tmpA[i] = alpSrcPtr[i];
00153                 for (     ; i < vLumFilterSize; i++)
00154                     tmpA[i] = tmpA[i - 1];
00155                 alpSrcPtr = tmpA;
00156             }
00157         }
00158         if (firstChrSrcY < 0 || firstChrSrcY + vChrFilterSize > c->chrSrcH) {
00159             const int16_t **tmpU = (const int16_t **) chrUPixBuf + 2 * vChrBufSize;
00160             int neg = -firstChrSrcY, i, end = FFMIN(c->chrSrcH - firstChrSrcY, vChrFilterSize);
00161             for (i = 0; i < neg;            i++) {
00162                 tmpU[i] = chrUSrcPtr[neg];
00163             }
00164             for (     ; i < end;            i++) {
00165                 tmpU[i] = chrUSrcPtr[i];
00166             }
00167             for (     ; i < vChrFilterSize; i++) {
00168                 tmpU[i] = tmpU[i - 1];
00169             }
00170             chrUSrcPtr = tmpU;
00171         }
00172 
00173         if (flags & SWS_ACCURATE_RND) {
00174             int s= APCK_SIZE / 8;
00175             for (i=0; i<vLumFilterSize; i+=2) {
00176                 *(const void**)&lumMmxFilter[s*i              ]= lumSrcPtr[i  ];
00177                 *(const void**)&lumMmxFilter[s*i+APCK_PTR2/4  ]= lumSrcPtr[i+(vLumFilterSize>1)];
00178                 lumMmxFilter[s*i+APCK_COEF/4  ]=
00179                 lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i    ]
00180                 + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);
00181                 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
00182                     *(const void**)&alpMmxFilter[s*i              ]= alpSrcPtr[i  ];
00183                     *(const void**)&alpMmxFilter[s*i+APCK_PTR2/4  ]= alpSrcPtr[i+(vLumFilterSize>1)];
00184                     alpMmxFilter[s*i+APCK_COEF/4  ]=
00185                     alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4  ];
00186                 }
00187             }
00188             for (i=0; i<vChrFilterSize; i+=2) {
00189                 *(const void**)&chrMmxFilter[s*i              ]= chrUSrcPtr[i  ];
00190                 *(const void**)&chrMmxFilter[s*i+APCK_PTR2/4  ]= chrUSrcPtr[i+(vChrFilterSize>1)];
00191                 chrMmxFilter[s*i+APCK_COEF/4  ]=
00192                 chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i    ]
00193                 + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);
00194             }
00195         } else {
00196             for (i=0; i<vLumFilterSize; i++) {
00197                 *(const void**)&lumMmxFilter[4*i+0]= lumSrcPtr[i];
00198                 lumMmxFilter[4*i+2]=
00199                 lumMmxFilter[4*i+3]=
00200                 ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;
00201                 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
00202                     *(const void**)&alpMmxFilter[4*i+0]= alpSrcPtr[i];
00203                     alpMmxFilter[4*i+2]=
00204                     alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2];
00205                 }
00206             }
00207             for (i=0; i<vChrFilterSize; i++) {
00208                 *(const void**)&chrMmxFilter[4*i+0]= chrUSrcPtr[i];
00209                 chrMmxFilter[4*i+2]=
00210                 chrMmxFilter[4*i+3]=
00211                 ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;
00212             }
00213         }
00214     }
00215 }
00216 
00217 void ff_sws_init_swScale_mmx(SwsContext *c)
00218 {
00219     int cpu_flags = av_get_cpu_flags();
00220 
00221     if (cpu_flags & AV_CPU_FLAG_MMX)
00222         sws_init_swScale_MMX(c);
00223 #if HAVE_MMX2
00224     if (cpu_flags & AV_CPU_FLAG_MMX2)
00225         sws_init_swScale_MMX2(c);
00226 #endif
00227 }
libswscale/x86/swscale_mmx.c
