diff options
Diffstat (limited to 'contrib/ffmpeg/libpostproc')
| -rw-r--r-- | contrib/ffmpeg/libpostproc/Makefile | 14 | ||||
| -rw-r--r-- | contrib/ffmpeg/libpostproc/mangle.h | 47 | ||||
| -rw-r--r-- | contrib/ffmpeg/libpostproc/postprocess.c | 96 | ||||
| -rw-r--r-- | contrib/ffmpeg/libpostproc/postprocess.h | 42 | ||||
| -rw-r--r-- | contrib/ffmpeg/libpostproc/postprocess_altivec_template.c | 14 | ||||
| -rw-r--r-- | contrib/ffmpeg/libpostproc/postprocess_internal.h | 29 | ||||
| -rw-r--r-- | contrib/ffmpeg/libpostproc/postprocess_template.c | 41 |
7 files changed, 105 insertions, 178 deletions
diff --git a/contrib/ffmpeg/libpostproc/Makefile b/contrib/ffmpeg/libpostproc/Makefile index d30b4a18d..8a285c6b0 100644 --- a/contrib/ffmpeg/libpostproc/Makefile +++ b/contrib/ffmpeg/libpostproc/Makefile @@ -1,24 +1,14 @@ include ../config.mak -# Overload incdir, postproc include files go in a different directory. -incdir=$(prefix)/include/postproc - EXTRALIBS := -L$(BUILD_ROOT)/libavutil -lavutil$(BUILDSUF) $(EXTRALIBS) NAME=postproc LIBVERSION=$(SPPVERSION) LIBMAJOR=$(SPPMAJOR) -STATIC_OBJS=postprocess.o -SHARED_OBJS=postprocess_pic.o - HEADERS = postprocess.h -include ../common.mak - -depend dep: postprocess.c - -postprocess_pic.o: postprocess.c - $(CC) -c $(CFLAGS) -fomit-frame-pointer -fPIC -DPIC -o $@ $< +OBJS = postprocess.o +include ../common.mak diff --git a/contrib/ffmpeg/libpostproc/mangle.h b/contrib/ffmpeg/libpostproc/mangle.h deleted file mode 100644 index 3521fa9bb..000000000 --- a/contrib/ffmpeg/libpostproc/mangle.h +++ /dev/null @@ -1,47 +0,0 @@ -/* - * mangle.h - This file has some CPP macros to deal with different symbol - * mangling across binary formats. - * - * (c)2002 by Felix Buenemann <atmosfear at users.sourceforge.net> - * - * This file is part of FFmpeg. - * - * FFmpeg is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * FFmpeg is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with FFmpeg; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - */ - -#ifndef __MANGLE_H -#define __MANGLE_H - -/* Feel free to add more to the list, eg. a.out IMO */ -/* Use rip-relative addressing if compiling PIC code on x86-64. */ -#if defined(__CYGWIN__) || defined(__MINGW32__) || defined(__OS2__) || \ - (defined(__OpenBSD__) && !defined(__ELF__)) -#if defined(ARCH_X86_64) && defined(PIC) -#define MANGLE(a) "_" #a"(%%rip)" -#else -#define MANGLE(a) "_" #a -#endif -#else -#if defined(ARCH_X86_64) && defined(PIC) -#define MANGLE(a) #a"(%%rip)" -#elif defined(CONFIG_DARWIN) -#define MANGLE(a) "_" #a -#else -#define MANGLE(a) #a -#endif -#endif - -#endif /* !__MANGLE_H */ - diff --git a/contrib/ffmpeg/libpostproc/postprocess.c b/contrib/ffmpeg/libpostproc/postprocess.c index f68c8a723..b274bb47c 100644 --- a/contrib/ffmpeg/libpostproc/postprocess.c +++ b/contrib/ffmpeg/libpostproc/postprocess.c @@ -47,8 +47,8 @@ LinBlendDeinterlace e E E* MedianDeinterlace# E Ec Ec TempDeNoiser# E e e Ec -* i dont have a 3dnow CPU -> its untested, but noone said it doesnt work so it seems to work -# more or less selfinvented filters so the exactness isnt too meaningfull +* i do not have a 3DNow! CPU -> it is untested, but no one said it does not work so it seems to work +# more or less selfinvented filters so the exactness is not too meaningful E = Exact implementation e = allmost exact implementation (slightly different rounding,...) a = alternative / approximate impl @@ -87,14 +87,9 @@ try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks //#undef HAVE_MMX //#undef ARCH_X86 //#define DEBUG_BRIGHTNESS -#ifdef USE_FASTMEMCPY -#include "libvo/fastmemcpy.h" -#endif #include "postprocess.h" #include "postprocess_internal.h" -#include "mangle.h" //FIXME should be supressed - #ifdef HAVE_ALTIVEC_H #include <altivec.h> #endif @@ -106,20 +101,17 @@ try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks //#define NUM_BLOCKS_AT_ONCE 16 //not used yet #if defined(ARCH_X86) -static uint64_t __attribute__((aligned(8))) attribute_used w05= 0x0005000500050005LL; -static uint64_t __attribute__((aligned(8))) attribute_used w04= 0x0004000400040004LL; -static uint64_t __attribute__((aligned(8))) attribute_used w20= 0x0020002000200020LL; -static uint64_t __attribute__((aligned(8))) attribute_used b00= 0x0000000000000000LL; -static uint64_t __attribute__((aligned(8))) attribute_used b01= 0x0101010101010101LL; -static uint64_t __attribute__((aligned(8))) attribute_used b02= 0x0202020202020202LL; -static uint64_t __attribute__((aligned(8))) attribute_used b08= 0x0808080808080808LL; -static uint64_t __attribute__((aligned(8))) attribute_used b80= 0x8080808080808080LL; +DECLARE_ASM_CONST(8, uint64_t, w05)= 0x0005000500050005LL; +DECLARE_ASM_CONST(8, uint64_t, w04)= 0x0004000400040004LL; +DECLARE_ASM_CONST(8, uint64_t, w20)= 0x0020002000200020LL; +DECLARE_ASM_CONST(8, uint64_t, b00)= 0x0000000000000000LL; +DECLARE_ASM_CONST(8, uint64_t, b01)= 0x0101010101010101LL; +DECLARE_ASM_CONST(8, uint64_t, b02)= 0x0202020202020202LL; +DECLARE_ASM_CONST(8, uint64_t, b08)= 0x0808080808080808LL; +DECLARE_ASM_CONST(8, uint64_t, b80)= 0x8080808080808080LL; #endif -static uint8_t clip_table[3*256]; -static uint8_t * const clip_tab= clip_table + 256; - -static const int attribute_used deringThreshold= 20; +DECLARE_ASM_CONST(8, int, deringThreshold)= 20; static struct PPFilter filters[]= @@ -147,11 +139,11 @@ static struct PPFilter filters[]= static const char *replaceTable[]= { - "default", "hdeblock:a,vdeblock:a,dering:a", - "de", "hdeblock:a,vdeblock:a,dering:a", - "fast", "x1hdeblock:a,x1vdeblock:a,dering:a", - "fa", "x1hdeblock:a,x1vdeblock:a,dering:a", - "ac", "ha:a:128:7,va:a,dering:a", + "default", "hb:a,vb:a,dr:a", + "de", "hb:a,vb:a,dr:a", + "fast", "h1:a,v1:a,dr:a", + "fa", "h1:a,v1:a,dr:a", + "ac", "ha:a:128:7,va:a,dr:a", NULL //End Marker }; @@ -401,8 +393,8 @@ static inline void doHorizLowPass_C(uint8_t dst[], int stride, PPContext *c) * Experimental Filter 1 (Horizontal) * will not damage linear gradients * Flat blocks should look like they where passed through the (1,1,2,2,4,2,2,1,1) 9-Tap filter - * can only smooth blocks at the expected locations (it cant smooth them if they did move) - * MMX2 version does correct clipping C version doesnt + * can only smooth blocks at the expected locations (it cannot smooth them if they did move) + * MMX2 version does correct clipping C version does not * not identical with the vertical one */ static inline void horizX1Filter(uint8_t *src, int stride, int QP) @@ -649,18 +641,18 @@ static av_always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, #include "postprocess_template.c" #endif -// minor note: the HAVE_xyz is messed up after that line so dont use it +// minor note: the HAVE_xyz is messed up after that line so do not use it. -static inline void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, - QP_STORE_T QPs[], int QPStride, int isColor, pp_mode_t *vm, pp_context_t *vc) +static inline void postProcess(const uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, + const QP_STORE_T QPs[], int QPStride, int isColor, pp_mode_t *vm, pp_context_t *vc) { PPContext *c= (PPContext *)vc; PPMode *ppMode= (PPMode *)vm; c->ppMode= *ppMode; //FIXME - // useing ifs here as they are faster than function pointers allthough the - // difference wouldnt be messureable here but its much better because - // someone might exchange the cpu whithout restarting mplayer ;) + // Using ifs here as they are faster than function pointers although the + // difference would not be measureable here but it is much better because + // someone might exchange the CPU whithout restarting MPlayer ;) #ifdef RUNTIME_CPUDETECT #if defined(ARCH_X86) // ordered per speed fasterst first @@ -702,7 +694,11 @@ static inline void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int /* -pp Command line Help */ -char *pp_help= +#if LIBPOSTPROC_VERSION_INT < (52<<16) +const char *const pp_help= +#else +const char pp_help[] = +#endif "Available postprocessing filters:\n" "Filters Options\n" "short long name short long option Description\n" @@ -746,12 +742,12 @@ char *pp_help= "\n" ; -pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality) +pp_mode_t *pp_get_mode_by_name_and_quality(const char *name, int quality) { char temp[GET_MODE_BUFFER_SIZE]; char *p= temp; - const char *filterDelimiters= ",/"; - const char *optionDelimiters= ":"; + static const char filterDelimiters[] = ",/"; + static const char optionDelimiters[] = ":"; struct PPMode *ppMode; char *filterToken; @@ -964,7 +960,7 @@ static void reallocBuffers(PPContext *c, int width, int height, int stride, int for(i=0; i<3; i++) { - //Note:the +17*1024 is just there so i dont have to worry about r/w over te end + //Note: The +17*1024 is just there so i do not have to worry about r/w over the end. reallocAlign((void **)&c->tempBlured[i], 8, stride*mbHeight*16 + 17*1024); reallocAlign((void **)&c->tempBluredPast[i], 8, 256*((height+7)&(~7))/2 + 17*1024);//FIXME size } @@ -975,27 +971,17 @@ static void reallocBuffers(PPContext *c, int width, int height, int stride, int reallocAlign((void **)&c->forcedQPTable, 8, mbWidth*sizeof(QP_STORE_T)); } -static void global_init(void){ - int i; - memset(clip_table, 0, 256); - for(i=256; i<512; i++) - clip_table[i]= i; - memset(clip_table+512, 0, 256); -} - static const char * context_to_name(void * ptr) { return "postproc"; } -static AVClass av_codec_context_class = { "Postproc", context_to_name, NULL }; +static const AVClass av_codec_context_class = { "Postproc", context_to_name, NULL }; pp_context_t *pp_get_context(int width, int height, int cpuCaps){ PPContext *c= av_malloc(sizeof(PPContext)); int stride= (width+15)&(~15); //assumed / will realloc if needed int qpStride= (width+15)/16 + 2; //assumed / will realloc if needed - global_init(); - memset(c, 0, sizeof(PPContext)); c->av_class = &av_codec_context_class; c->cpuCaps= cpuCaps; @@ -1035,10 +1021,10 @@ void pp_free_context(void *vc){ av_free(c); } -void pp_postprocess(uint8_t * src[3], int srcStride[3], - uint8_t * dst[3], int dstStride[3], +void pp_postprocess(const uint8_t * src[3], const int srcStride[3], + uint8_t * dst[3], const int dstStride[3], int width, int height, - QP_STORE_T *QP_store, int QPStride, + const QP_STORE_T *QP_store, int QPStride, pp_mode_t *vm, void *vc, int pict_type) { int mbWidth = (width+15)>>4; @@ -1060,16 +1046,16 @@ void pp_postprocess(uint8_t * src[3], int srcStride[3], QP_store= c->forcedQPTable; absQPStride = QPStride = 0; if(mode->lumMode & FORCE_QUANT) - for(i=0; i<mbWidth; i++) QP_store[i]= mode->forcedQuant; + for(i=0; i<mbWidth; i++) c->forcedQPTable[i]= mode->forcedQuant; else - for(i=0; i<mbWidth; i++) QP_store[i]= 1; + for(i=0; i<mbWidth; i++) c->forcedQPTable[i]= 1; } if(pict_type & PP_PICT_TYPE_QP2){ int i; const int count= mbHeight * absQPStride; for(i=0; i<(count>>2); i++){ - ((uint32_t*)c->stdQPTable)[i] = (((uint32_t*)QP_store)[i]>>1) & 0x7F7F7F7F; + ((uint32_t*)c->stdQPTable)[i] = (((const uint32_t*)QP_store)[i]>>1) & 0x7F7F7F7F; } for(i<<=2; i<count; i++){ c->stdQPTable[i] = QP_store[i]>>1; @@ -1095,7 +1081,7 @@ for(y=0; y<mbHeight; y++){ int i; const int count= mbHeight * QPStride; for(i=0; i<(count>>2); i++){ - ((uint32_t*)c->nonBQPTable)[i] = ((uint32_t*)QP_store)[i] & 0x3F3F3F3F; + ((uint32_t*)c->nonBQPTable)[i] = ((const uint32_t*)QP_store)[i] & 0x3F3F3F3F; } for(i<<=2; i<count; i++){ c->nonBQPTable[i] = QP_store[i] & 0x3F; diff --git a/contrib/ffmpeg/libpostproc/postprocess.h b/contrib/ffmpeg/libpostproc/postprocess.h index eed92ba6d..818fb41ba 100644 --- a/contrib/ffmpeg/libpostproc/postprocess.h +++ b/contrib/ffmpeg/libpostproc/postprocess.h @@ -18,8 +18,8 @@ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ -#ifndef NEWPOSTPROCESS_H -#define NEWPOSTPROCESS_H +#ifndef FFMPEG_POSTPROCESS_H +#define FFMPEG_POSTPROCESS_H /** * @file postprocess.h @@ -27,12 +27,18 @@ * external api for the pp stuff */ -#ifdef __cplusplus -extern "C" { -#endif +#include "libavutil/avutil.h" + +#define LIBPOSTPROC_VERSION_MAJOR 51 +#define LIBPOSTPROC_VERSION_MINOR 1 +#define LIBPOSTPROC_VERSION_MICRO 0 -#define LIBPOSTPROC_VERSION_INT ((51<<16)+(1<<8)+0) -#define LIBPOSTPROC_VERSION 51.1.0 +#define LIBPOSTPROC_VERSION_INT AV_VERSION_INT(LIBPOSTPROC_VERSION_MAJOR, \ + LIBPOSTPROC_VERSION_MINOR, \ + LIBPOSTPROC_VERSION_MICRO) +#define LIBPOSTPROC_VERSION AV_VERSION(LIBPOSTPROC_VERSION_MAJOR, \ + LIBPOSTPROC_VERSION_MINOR, \ + LIBPOSTPROC_VERSION_MICRO) #define LIBPOSTPROC_BUILD LIBPOSTPROC_VERSION_INT #define LIBPOSTPROC_IDENT "postproc" AV_STRINGIFY(LIBPOSTPROC_VERSION) @@ -41,15 +47,21 @@ extern "C" { #define QP_STORE_T int8_t +#include <inttypes.h> + typedef void pp_context_t; typedef void pp_mode_t; -extern char *pp_help; ///< a simple help text +#if LIBPOSTPROC_VERSION_INT < (52<<16) +extern const char *const pp_help; ///< a simple help text +#else +extern const char pp_help[]; ///< a simple help text +#endif -void pp_postprocess(uint8_t * src[3], int srcStride[3], - uint8_t * dst[3], int dstStride[3], +void pp_postprocess(const uint8_t * src[3], const int srcStride[3], + uint8_t * dst[3], const int dstStride[3], int horizontalSize, int verticalSize, - QP_STORE_T *QP_store, int QP_stride, + const QP_STORE_T *QP_store, int QP_stride, pp_mode_t *mode, pp_context_t *ppContext, int pict_type); @@ -58,7 +70,7 @@ void pp_postprocess(uint8_t * src[3], int srcStride[3], * name is the string after "-pp" on the command line * quality is a number from 0 to PP_QUALITY_MAX */ -pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality); +pp_mode_t *pp_get_mode_by_name_and_quality(const char *name, int quality); void pp_free_mode(pp_mode_t *mode); pp_context_t *pp_get_context(int width, int height, int flags); @@ -77,8 +89,4 @@ void pp_free_context(pp_context_t *ppContext); #define PP_PICT_TYPE_QP2 0x00000010 ///< MPEG2 style QScale -#ifdef __cplusplus -} -#endif - -#endif +#endif /* FFMPEG_POSTPROCESS_H */ diff --git a/contrib/ffmpeg/libpostproc/postprocess_altivec_template.c b/contrib/ffmpeg/libpostproc/postprocess_altivec_template.c index 6a76c0eb7..e93ce7ef8 100644 --- a/contrib/ffmpeg/libpostproc/postprocess_altivec_template.c +++ b/contrib/ffmpeg/libpostproc/postprocess_altivec_template.c @@ -22,12 +22,6 @@ #include "avutil.h" -#ifdef CONFIG_DARWIN -#define AVV(x...) (x) -#else -#define AVV(x...) {x} -#endif - #define ALTIVEC_TRANSPOSE_8x8_SHORT(src_a,src_b,src_c,src_d,src_e,src_f,src_g,src_h) \ do { \ __typeof__(src_a) tempA1, tempB1, tempC1, tempD1; \ @@ -200,7 +194,7 @@ static inline void doVertLowPass_altivec(uint8_t *src, int stride, PPContext *c) One could remove the recomputation of the perm vector by assuming (stride % 16) == 0, unfortunately this is not always true. Quite a lot of load/stores - can be removed by assuming proper alignement of + can be removed by assuming proper alignment of src & stride :-( */ uint8_t *src2 = src; @@ -388,7 +382,7 @@ static inline void doVertDefFilter_altivec(uint8_t src[], int stride, PPContext One could remove the recomputation of the perm vector by assuming (stride % 16) == 0, unfortunately this is not always true. Quite a lot of load/stores - can be removed by assuming proper alignement of + can be removed by assuming proper alignment of src & stride :-( */ uint8_t *src2 = src; @@ -475,7 +469,7 @@ static inline void doVertDefFilter_altivec(uint8_t src[], int stride, PPContext const vector signed short dornotd = vec_sel((vector signed short)zero, dclampedfinal, vec_cmplt(absmE, vqp)); - /* add/substract to l4 and l5 */ + /* add/subtract to l4 and l5 */ const vector signed short vb4minusd = vec_sub(vb4, dornotd); const vector signed short vb5plusd = vec_add(vb5, dornotd); /* finally, stores */ @@ -512,7 +506,7 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) { One could remove the recomputation of the perm vector by assuming (stride % 16) == 0, unfortunately this is not always true. Quite a lot of load/stores - can be removed by assuming proper alignement of + can be removed by assuming proper alignment of src & stride :-( */ uint8_t *srcCopy = src; diff --git a/contrib/ffmpeg/libpostproc/postprocess_internal.h b/contrib/ffmpeg/libpostproc/postprocess_internal.h index 537d728c0..f2d40e1d5 100644 --- a/contrib/ffmpeg/libpostproc/postprocess_internal.h +++ b/contrib/ffmpeg/libpostproc/postprocess_internal.h @@ -23,7 +23,11 @@ * internal api header. */ +#ifndef FFMPEG_POSTPROCESS_INTERNAL_H +#define FFMPEG_POSTPROCESS_INTERNAL_H + #include "avutil.h" +#include "postprocess.h" #define V_DEBLOCK 0x01 #define H_DEBLOCK 0x02 @@ -62,27 +66,16 @@ #define TEMP_NOISE_FILTER 0x100000 #define FORCE_QUANT 0x200000 -#if ( defined(__PIC__) || defined(__pic__) ) && ! defined(PIC) -# define PIC -#endif - -//use if u want a faster postprocessing code -//cant differentiate between chroma & luma filters (both on or both off) -//obviosly the -pp option at the commandline has no effect except turning the here selected +//use if you want a faster postprocessing code +//cannot differentiate between chroma & luma filters (both on or both off) +//obviously the -pp option on the command line has no effect except turning the here selected //filters on //#define COMPILE_TIME_MODE 0x77 -#if 1 static inline int CLIP(int a){ if(a&256) return ((a)>>31)^(-1); else return a; } -//#define CLIP(a) (((a)&256) ? ((a)>>31)^(-1) : (a)) -#elif 0 -#define CLIP(a) clip_tab[a] -#else -#define CLIP(a) (a) -#endif /** * Postprocessng filter. */ @@ -122,7 +115,7 @@ typedef struct PPContext{ /** * info on struct for av_log */ - AVClass *av_class; + const AVClass *av_class; uint8_t *tempBlocks; ///<used for the horizontal code @@ -173,11 +166,13 @@ typedef struct PPContext{ } PPContext; -static inline void linecpy(void *dest, void *src, int lines, int stride) +static inline void linecpy(void *dest, const void *src, int lines, int stride) { if (stride > 0) { memcpy(dest, src, lines*stride); } else { - memcpy(dest+(lines-1)*stride, src+(lines-1)*stride, -lines*stride); + memcpy((uint8_t*)dest+(lines-1)*stride, (const uint8_t*)src+(lines-1)*stride, -lines*stride); } } + +#endif /* FFMPEG_POSTPROCESS_INTERNAL_H */ diff --git a/contrib/ffmpeg/libpostproc/postprocess_template.c b/contrib/ffmpeg/libpostproc/postprocess_template.c index f084130a6..bdc23250a 100644 --- a/contrib/ffmpeg/libpostproc/postprocess_template.c +++ b/contrib/ffmpeg/libpostproc/postprocess_template.c @@ -16,7 +16,7 @@ * You should have received a copy of the GNU General Public License * along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA -*/ + */ /** * @file postprocess_template.c @@ -73,7 +73,7 @@ "paddb " #a ", " #b " \n\t" #endif -//FIXME? |255-0| = 1 (shouldnt be a problem ...) +//FIXME? |255-0| = 1 (should not be a problem ...) #ifdef HAVE_MMX /** * Check if the middle 8x8 Block in the given 8x16 block is flat @@ -373,7 +373,8 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c) * Experimental implementation of the filter (Algorithm 1) described in a paper from Ramkishor & Karandikar * values are correctly clipped (MMX2) * values are wraparound (C) - * conclusion: its fast, but introduces ugly horizontal patterns if there is a continious gradient + * Conclusion: It is fast, but introduces ugly horizontal patterns + * if there is a continuous gradient. 0 8 16 24 x = 8 x/2 = 4 @@ -478,8 +479,8 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP) * Experimental Filter 1 * will not damage linear gradients * Flat blocks should look like they where passed through the (1,1,2,2,4,2,2,1,1) 9-Tap filter - * can only smooth blocks at the expected locations (it cant smooth them if they did move) - * MMX2 version does correct clipping C version doesnt + * can only smooth blocks at the expected locations (it cannot smooth them if they did move) + * MMX2 version does correct clipping C version does not */ static inline void RENAME(vertX1Filter)(uint8_t *src, int stride, PPContext *co) { @@ -3179,16 +3180,16 @@ asm volatile( } #endif //HAVE_MMX -static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, - QP_STORE_T QPs[], int QPStride, int isColor, PPContext *c); +static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, + const QP_STORE_T QPs[], int QPStride, int isColor, PPContext *c); /** - * Copies a block from src to dst and fixes the blacklevel - * levelFix == 0 -> dont touch the brighness & contrast + * Copies a block from src to dst and fixes the blacklevel. + * levelFix == 0 -> do not touch the brighness & contrast */ #undef SCALED_CPY -static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[], int srcStride, +static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, const uint8_t src[], int srcStride, int levelFix, int64_t *packedOffsetAndScale) { #ifndef HAVE_MMX @@ -3345,10 +3346,10 @@ static inline void RENAME(duplicate)(uint8_t src[], int stride) /** * Filters array of bytes (Y or U or V values) */ -static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, - QP_STORE_T QPs[], int QPStride, int isColor, PPContext *c2) +static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, + const QP_STORE_T QPs[], int QPStride, int isColor, PPContext *c2) { - PPContext __attribute__((aligned(8))) c= *c2; //copy to stack for faster access + DECLARE_ALIGNED(8, PPContext, c)= *c2; //copy to stack for faster access int x,y; #ifdef COMPILE_TIME_MODE const int mode= COMPILE_TIME_MODE; @@ -3415,7 +3416,7 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int sum+= yHistogram[i]; } - /* we allways get a completly black picture first */ + /* We always get a completely black picture first. */ maxClipped= (uint64_t)(sum * c.ppMode.maxClippedThreshold); clipped= sum; @@ -3461,7 +3462,7 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int /* copy & deinterlace first row of blocks */ y=-BLOCK_SIZE; { - uint8_t *srcBlock= &(src[y*srcStride]); + const uint8_t *srcBlock= &(src[y*srcStride]); uint8_t *dstBlock= tempDst + dstStride; // From this point on it is guaranteed that we can read and write 16 lines downward @@ -3498,7 +3499,7 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int ); #elif defined(HAVE_3DNOW) -//FIXME check if this is faster on an 3dnow chip or if its faster without the prefetch or ... +//FIXME check if this is faster on an 3dnow chip or if it is faster without the prefetch or ... /* prefetch(srcBlock + (((x>>3)&3) + 5)*srcStride + 32); prefetch(srcBlock + (((x>>3)&3) + 9)*srcStride + 32); prefetchw(dstBlock + (((x>>3)&3) + 5)*dstStride + 32); @@ -3544,13 +3545,13 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int for(y=0; y<height; y+=BLOCK_SIZE) { //1% speedup if these are here instead of the inner loop - uint8_t *srcBlock= &(src[y*srcStride]); + const uint8_t *srcBlock= &(src[y*srcStride]); uint8_t *dstBlock= &(dst[y*dstStride]); #ifdef HAVE_MMX uint8_t *tempBlock1= c.tempBlocks; uint8_t *tempBlock2= c.tempBlocks + 8; #endif - int8_t *QPptr= &QPs[(y>>qpVShift)*QPStride]; + const int8_t *QPptr= &QPs[(y>>qpVShift)*QPStride]; int8_t *nonBQPptr= &c.nonBQPTable[(y>>qpVShift)*FFABS(QPStride)]; int QP=0; /* can we mess with a 8x16 block from srcBlock/dstBlock downwards and 1 line upwards @@ -3642,7 +3643,7 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int ); #elif defined(HAVE_3DNOW) -//FIXME check if this is faster on an 3dnow chip or if its faster without the prefetch or ... +//FIXME check if this is faster on an 3dnow chip or if it is faster without the prefetch or ... /* prefetch(srcBlock + (((x>>3)&3) + 5)*srcStride + 32); prefetch(srcBlock + (((x>>3)&3) + 9)*srcStride + 32); prefetchw(dstBlock + (((x>>3)&3) + 5)*dstStride + 32); @@ -3717,7 +3718,7 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int else if(mode & H_DEBLOCK) { #ifdef HAVE_ALTIVEC - unsigned char __attribute__ ((aligned(16))) tempBlock[272]; + DECLARE_ALIGNED(16, unsigned char, tempBlock[272]); transpose_16x8_char_toPackedAlign_altivec(tempBlock, dstBlock - (4 + 1), stride); const int t=vertClassify_altivec(tempBlock-48, 16, &c); |