Commit ba5bcf96 authored by Derek Buitenhuis's avatar Derek Buitenhuis

Merge commit '2ade1cda'

* commit '2ade1cda':
  intrax8: K&R formatting cosmetics
Merged-by: 's avatarDerek Buitenhuis <derek.buitenhuis@gmail.com>
parents 7120bff0 2ade1cda
...@@ -31,7 +31,8 @@ ...@@ -31,7 +31,8 @@
#include "intrax8.h" #include "intrax8.h"
#include "intrax8dsp.h" #include "intrax8dsp.h"
#define MAX_TABLE_DEPTH(table_bits, max_bits) ((max_bits+table_bits-1)/table_bits) #define MAX_TABLE_DEPTH(table_bits, max_bits) \
((max_bits + table_bits - 1) / table_bits)
#define DC_VLC_BITS 9 #define DC_VLC_BITS 9
#define AC_VLC_BITS 9 #define AC_VLC_BITS 9
...@@ -41,15 +42,16 @@ ...@@ -41,15 +42,16 @@
#define AC_VLC_MTD MAX_TABLE_DEPTH(AC_VLC_BITS, MAX_AC_VLC_BITS) #define AC_VLC_MTD MAX_TABLE_DEPTH(AC_VLC_BITS, MAX_AC_VLC_BITS)
#define OR_VLC_MTD MAX_TABLE_DEPTH(OR_VLC_BITS, MAX_OR_VLC_BITS) #define OR_VLC_MTD MAX_TABLE_DEPTH(OR_VLC_BITS, MAX_OR_VLC_BITS)
static VLC j_ac_vlc[2][2][8]; //[quant<13],[intra/inter],[select] static VLC j_ac_vlc[2][2][8]; // [quant < 13], [intra / inter], [select]
static VLC j_dc_vlc[2][8]; //[quant], [select] static VLC j_dc_vlc[2][8]; // [quant], [select]
static VLC j_orient_vlc[2][4]; //[quant], [select] static VLC j_orient_vlc[2][4]; // [quant], [select]
static av_cold void x8_vlc_init(void){ static av_cold void x8_vlc_init(void)
{
int i; int i;
int offset = 0; int offset = 0;
int sizeidx = 0; int sizeidx = 0;
static const uint16_t sizes[8*4 + 8*2 + 2 + 4] = { static const uint16_t sizes[8 * 4 + 8 * 2 + 2 + 4] = {
576, 548, 582, 618, 546, 616, 560, 642, 576, 548, 582, 618, 546, 616, 560, 642,
584, 582, 704, 664, 512, 544, 656, 640, 584, 582, 704, 664, 512, 544, 656, 640,
512, 648, 582, 566, 532, 614, 596, 648, 512, 648, 582, 566, 532, 614, 596, 648,
...@@ -58,95 +60,99 @@ static av_cold void x8_vlc_init(void){ ...@@ -58,95 +60,99 @@ static av_cold void x8_vlc_init(void){
528, 528, 526, 528, 536, 528, 526, 544, 528, 528, 526, 528, 536, 528, 526, 544,
544, 512, 512, 528, 528, 544, 512, 544, 544, 512, 512, 528, 528, 544, 512, 544,
128, 128, 128, 128, 128, 128}; 128, 128, 128, 128, 128, 128,
};
static VLC_TYPE table[28150][2]; static VLC_TYPE table[28150][2];
#define init_ac_vlc(dst,src) do { \ // set ac tables
dst.table = &table[offset]; \ #define init_ac_vlc(dst, src) \
dst.table_allocated = sizes[sizeidx]; \ do { \
offset += sizes[sizeidx++]; \ dst.table = &table[offset]; \
init_vlc(&dst, \ dst.table_allocated = sizes[sizeidx]; \
AC_VLC_BITS,77, \ offset += sizes[sizeidx++]; \
&src[1],4,2, \ init_vlc(&dst, AC_VLC_BITS, 77, &src[1], 4, 2, &src[0], 4, 2, \
&src[0],4,2, \ INIT_VLC_USE_NEW_STATIC); \
INIT_VLC_USE_NEW_STATIC); \
} while(0) } while(0)
//set ac tables
for(i=0;i<8;i++){ for (i = 0; i < 8; i++) {
init_ac_vlc( j_ac_vlc[0][0][i], x8_ac0_highquant_table[i][0] ); init_ac_vlc(j_ac_vlc[0][0][i], x8_ac0_highquant_table[i][0]);
init_ac_vlc( j_ac_vlc[0][1][i], x8_ac1_highquant_table[i][0] ); init_ac_vlc(j_ac_vlc[0][1][i], x8_ac1_highquant_table[i][0]);
init_ac_vlc( j_ac_vlc[1][0][i], x8_ac0_lowquant_table [i][0] ); init_ac_vlc(j_ac_vlc[1][0][i], x8_ac0_lowquant_table[i][0]);
init_ac_vlc( j_ac_vlc[1][1][i], x8_ac1_lowquant_table [i][0] ); init_ac_vlc(j_ac_vlc[1][1][i], x8_ac1_lowquant_table[i][0]);
} }
#undef init_ac_vlc #undef init_ac_vlc
//set dc tables // set dc tables
#define init_dc_vlc(dst,src) do { \ #define init_dc_vlc(dst, src) \
dst.table = &table[offset]; \ do { \
dst.table_allocated = sizes[sizeidx]; \ dst.table = &table[offset]; \
offset += sizes[sizeidx++]; \ dst.table_allocated = sizes[sizeidx]; \
init_vlc(&dst, \ offset += sizes[sizeidx++]; \
DC_VLC_BITS,34, \ init_vlc(&dst, DC_VLC_BITS, 34, &src[1], 4, 2, &src[0], 4, 2, \
&src[1],4,2, \ INIT_VLC_USE_NEW_STATIC); \
&src[0],4,2, \
INIT_VLC_USE_NEW_STATIC); \
} while(0) } while(0)
for(i=0;i<8;i++){
init_dc_vlc( j_dc_vlc[0][i], x8_dc_highquant_table[i][0]); for (i = 0; i < 8; i++) {
init_dc_vlc( j_dc_vlc[1][i], x8_dc_lowquant_table [i][0]); init_dc_vlc(j_dc_vlc[0][i], x8_dc_highquant_table[i][0]);
init_dc_vlc(j_dc_vlc[1][i], x8_dc_lowquant_table[i][0]);
} }
#undef init_dc_vlc #undef init_dc_vlc
//set orient tables // set orient tables
#define init_or_vlc(dst,src) do { \ #define init_or_vlc(dst, src) \
dst.table = &table[offset]; \ do { \
dst.table_allocated = sizes[sizeidx]; \ dst.table = &table[offset]; \
offset += sizes[sizeidx++]; \ dst.table_allocated = sizes[sizeidx]; \
init_vlc(&dst, \ offset += sizes[sizeidx++]; \
OR_VLC_BITS,12, \ init_vlc(&dst, OR_VLC_BITS, 12, &src[1], 4, 2, &src[0], 4, 2, \
&src[1],4,2, \ INIT_VLC_USE_NEW_STATIC); \
&src[0],4,2, \
INIT_VLC_USE_NEW_STATIC); \
} while(0) } while(0)
for(i=0;i<2;i++){
init_or_vlc( j_orient_vlc[0][i], x8_orient_highquant_table[i][0]); for (i = 0; i < 2; i++)
} init_or_vlc(j_orient_vlc[0][i], x8_orient_highquant_table[i][0]);
for(i=0;i<4;i++){ for (i = 0; i < 4; i++)
init_or_vlc( j_orient_vlc[1][i], x8_orient_lowquant_table [i][0]); init_or_vlc(j_orient_vlc[1][i], x8_orient_lowquant_table[i][0]);
} #undef init_or_vlc
if (offset != sizeof(table)/sizeof(VLC_TYPE)/2) if (offset != sizeof(table)/sizeof(VLC_TYPE)/2)
av_log(NULL, AV_LOG_ERROR, "table size %zd does not match needed %i\n", av_log(NULL, AV_LOG_ERROR, "table size %zd does not match needed %i\n",
sizeof(table) / sizeof(VLC_TYPE) / 2, offset); sizeof(table) / sizeof(VLC_TYPE) / 2, offset);
} }
#undef init_or_vlc
static void x8_reset_vlc_tables(IntraX8Context * w){ static void x8_reset_vlc_tables(IntraX8Context *w)
memset(w->j_dc_vlc,0,sizeof(w->j_dc_vlc)); {
memset(w->j_ac_vlc,0,sizeof(w->j_ac_vlc)); memset(w->j_dc_vlc, 0, sizeof(w->j_dc_vlc));
w->j_orient_vlc=NULL; memset(w->j_ac_vlc, 0, sizeof(w->j_ac_vlc));
w->j_orient_vlc = NULL;
} }
static inline void x8_select_ac_table(IntraX8Context * const w , int mode){ static inline void x8_select_ac_table(IntraX8Context *const w, int mode)
MpegEncContext * const s= w->s; {
MpegEncContext *const s = w->s;
int table_index; int table_index;
av_assert2(mode<4); av_assert2(mode < 4);
if( w->j_ac_vlc[mode] ) return; if (w->j_ac_vlc[mode])
return;
table_index = get_bits(&s->gb, 3);
// 2 modes use same tables
w->j_ac_vlc[mode] = &j_ac_vlc[w->quant < 13][mode >> 1][table_index];
table_index = get_bits(&s->gb, 3); table_index = get_bits(&s->gb, 3);
w->j_ac_vlc[mode] = &j_ac_vlc[w->quant<13][mode>>1][table_index];//2 modes use same tables w->j_ac_vlc[mode] = &j_ac_vlc[w->quant<13][mode>>1][table_index];//2 modes use same tables
av_assert2(w->j_ac_vlc[mode]); av_assert2(w->j_ac_vlc[mode]);
} }
static inline int x8_get_orient_vlc(IntraX8Context * w){ static inline int x8_get_orient_vlc(IntraX8Context *w)
MpegEncContext * const s= w->s; {
int table_index; MpegEncContext *const s = w->s;
if(!w->j_orient_vlc ){ if (!w->j_orient_vlc) {
table_index = get_bits(&s->gb, 1+(w->quant<13) ); int table_index = get_bits(&s->gb, 1 + (w->quant < 13));
w->j_orient_vlc = &j_orient_vlc[w->quant<13][table_index]; w->j_orient_vlc = &j_orient_vlc[w->quant < 13][table_index];
} }
return get_vlc2(&s->gb, w->j_orient_vlc->table, OR_VLC_BITS, OR_VLC_MTD); return get_vlc2(&s->gb, w->j_orient_vlc->table, OR_VLC_BITS, OR_VLC_MTD);
...@@ -157,41 +163,41 @@ static inline int x8_get_orient_vlc(IntraX8Context * w){ ...@@ -157,41 +163,41 @@ static inline int x8_get_orient_vlc(IntraX8Context * w){
#define extra_level (0x00 << 8) // 1 bit #define extra_level (0x00 << 8) // 1 bit
#define run_offset(r) ((r) << 16) // 6 bits #define run_offset(r) ((r) << 16) // 6 bits
#define level_offset(l) ((l) << 24) // 5 bits #define level_offset(l) ((l) << 24) // 5 bits
static const uint32_t ac_decode_table[]={ static const uint32_t ac_decode_table[] = {
/*46*/ extra_bits(3) | extra_run | run_offset(16) | level_offset( 0), /* 46 */ extra_bits(3) | extra_run | run_offset(16) | level_offset(0),
/*47*/ extra_bits(3) | extra_run | run_offset(24) | level_offset( 0), /* 47 */ extra_bits(3) | extra_run | run_offset(24) | level_offset(0),
/*48*/ extra_bits(2) | extra_run | run_offset( 4) | level_offset( 1), /* 48 */ extra_bits(2) | extra_run | run_offset(4) | level_offset(1),
/*49*/ extra_bits(3) | extra_run | run_offset( 8) | level_offset( 1), /* 49 */ extra_bits(3) | extra_run | run_offset(8) | level_offset(1),
/*50*/ extra_bits(5) | extra_run | run_offset(32) | level_offset( 0), /* 50 */ extra_bits(5) | extra_run | run_offset(32) | level_offset(0),
/*51*/ extra_bits(4) | extra_run | run_offset(16) | level_offset( 1), /* 51 */ extra_bits(4) | extra_run | run_offset(16) | level_offset(1),
/*52*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 4), /* 52 */ extra_bits(2) | extra_level | run_offset(0) | level_offset(4),
/*53*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 8), /* 53 */ extra_bits(2) | extra_level | run_offset(0) | level_offset(8),
/*54*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset(12), /* 54 */ extra_bits(2) | extra_level | run_offset(0) | level_offset(12),
/*55*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset(16), /* 55 */ extra_bits(3) | extra_level | run_offset(0) | level_offset(16),
/*56*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset(24), /* 56 */ extra_bits(3) | extra_level | run_offset(0) | level_offset(24),
/*57*/ extra_bits(2) | extra_level | run_offset( 1) | level_offset( 3), /* 57 */ extra_bits(2) | extra_level | run_offset(1) | level_offset(3),
/*58*/ extra_bits(3) | extra_level | run_offset( 1) | level_offset( 7), /* 58 */ extra_bits(3) | extra_level | run_offset(1) | level_offset(7),
/*59*/ extra_bits(2) | extra_run | run_offset(16) | level_offset( 0), /* 59 */ extra_bits(2) | extra_run | run_offset(16) | level_offset(0),
/*60*/ extra_bits(2) | extra_run | run_offset(20) | level_offset( 0), /* 60 */ extra_bits(2) | extra_run | run_offset(20) | level_offset(0),
/*61*/ extra_bits(2) | extra_run | run_offset(24) | level_offset( 0), /* 61 */ extra_bits(2) | extra_run | run_offset(24) | level_offset(0),
/*62*/ extra_bits(2) | extra_run | run_offset(28) | level_offset( 0), /* 62 */ extra_bits(2) | extra_run | run_offset(28) | level_offset(0),
/*63*/ extra_bits(4) | extra_run | run_offset(32) | level_offset( 0), /* 63 */ extra_bits(4) | extra_run | run_offset(32) | level_offset(0),
/*64*/ extra_bits(4) | extra_run | run_offset(48) | level_offset( 0), /* 64 */ extra_bits(4) | extra_run | run_offset(48) | level_offset(0),
/*65*/ extra_bits(2) | extra_run | run_offset( 4) | level_offset( 1), /* 65 */ extra_bits(2) | extra_run | run_offset(4) | level_offset(1),
/*66*/ extra_bits(3) | extra_run | run_offset( 8) | level_offset( 1), /* 66 */ extra_bits(3) | extra_run | run_offset(8) | level_offset(1),
/*67*/ extra_bits(4) | extra_run | run_offset(16) | level_offset( 1), /* 67 */ extra_bits(4) | extra_run | run_offset(16) | level_offset(1),
/*68*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 4), /* 68 */ extra_bits(2) | extra_level | run_offset(0) | level_offset(4),
/*69*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset( 8), /* 69 */ extra_bits(3) | extra_level | run_offset(0) | level_offset(8),
/*70*/ extra_bits(4) | extra_level | run_offset( 0) | level_offset(16), /* 70 */ extra_bits(4) | extra_level | run_offset(0) | level_offset(16),
/*71*/ extra_bits(2) | extra_level | run_offset( 1) | level_offset( 3), /* 71 */ extra_bits(2) | extra_level | run_offset(1) | level_offset(3),
/*72*/ extra_bits(3) | extra_level | run_offset( 1) | level_offset( 7), /* 72 */ extra_bits(3) | extra_level | run_offset(1) | level_offset(7),
}; };
#undef extra_bits #undef extra_bits
#undef extra_run #undef extra_run
...@@ -199,306 +205,347 @@ static const uint32_t ac_decode_table[]={ ...@@ -199,306 +205,347 @@ static const uint32_t ac_decode_table[]={
#undef run_offset #undef run_offset
#undef level_offset #undef level_offset
static void x8_get_ac_rlf(IntraX8Context * const w, const int mode, static void x8_get_ac_rlf(IntraX8Context *const w, const int mode,
int * const run, int * const level, int * const final){ int *const run, int *const level, int *const final)
MpegEncContext * const s= w->s; {
int i,e; MpegEncContext *const s = w->s;
int i, e;
// x8_select_ac_table(w,mode); // x8_select_ac_table(w, mode);
i = get_vlc2(&s->gb, w->j_ac_vlc[mode]->table, AC_VLC_BITS, AC_VLC_MTD); i = get_vlc2(&s->gb, w->j_ac_vlc[mode]->table, AC_VLC_BITS, AC_VLC_MTD);
if(i<46){ //[0-45] if (i < 46) { // [0-45]
int t,l; int t, l;
if(i<0){ if (i < 0) {
(*level)=(*final)=//prevent 'may be used unilitialized' (*level) =
(*run)=64;//this would cause error exit in the ac loop (*final) = // prevent 'may be used unilitialized'
(*run) = 64; // this would cause error exit in the ac loop
return; return;
} }
(*final) = t = (i>22); /*
i-=23*t; * i == 0-15 r = 0-15 l = 0; r = i & %01111
/* * i == 16-19 r = 0-3 l = 1; r = i & %00011
i== 0-15 r=0-15 l=0 ;r=i& %01111 * i == 20-21 r = 0-1 l = 2; r = i & %00001
i==16-19 r=0-3 l=1 ;r=i& %00011 * i == 22 r = 0 l = 3; r = i & %00000
i==20-21 r=0-1 l=2 ;r=i& %00001 */
i==22 r=0 l=3 ;r=i& %00000
l=lut_l[i/2]={0,0,0,0,0,0,0,0,1,1,2,3}[i>>1];// 11 10'01 01'00 00'00 00'00 00'00 00 => 0xE50000 (*final) =
t=lut_mask[l]={0x0f,0x03,0x01,0x00}[l]; as i<256 the higher bits do not matter */ t = (i > 22);
l=(0xE50000>>(i&(0x1E)))&3;/*0x1E or (~1) or ((i>>1)<<1)*/ i -= 23 * t;
t=(0x01030F>>(l<<3));
/* l = lut_l[i / 2] = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 2, 3 }[i >> 1];
(*run) = i&t; * 11 10'01 01'00 00'00 00'00 00'00 00 => 0xE50000 */
l = (0xE50000 >> (i & (0x1E))) & 3; // 0x1E or (~1) or ((i >> 1) << 1)
/* t = lut_mask[l] = { 0x0f, 0x03, 0x01, 0x00 }[l];
* as i < 256 the higher bits do not matter */
t = (0x01030F >> (l << 3));
(*run) = i & t;
(*level) = l; (*level) = l;
}else if(i<73){//[46-72] } else if (i < 73) { // [46-72]
uint32_t sm; uint32_t sm;
uint32_t mask; uint32_t mask;
i-=46; i -= 46;
sm=ac_decode_table[i]; sm = ac_decode_table[i];
e=get_bits(&s->gb,sm&0xF);sm>>=8;//3bits e = get_bits(&s->gb, sm & 0xF);
mask=sm&0xff;sm>>=8; //1bit sm >>= 8; // 3bits
mask = sm & 0xff;
(*run) =(sm&0xff) + (e&( mask));//6bits sm >>= 8; // 1bit
(*level)=(sm>>8) + (e&(~mask));//5bits
(*final)=i>(58-46); (*run) = (sm & 0xff) + (e & (mask)); // 6bits
}else if(i<75){//[73-74] (*level) = (sm >> 8) + (e & (~mask)); // 5bits
static const uint8_t crazy_mix_runlevel[32]={ (*final) = i > (58 - 46);
0x22,0x32,0x33,0x53,0x23,0x42,0x43,0x63, } else if (i < 75) { // [73-74]
0x24,0x52,0x34,0x73,0x25,0x62,0x44,0x83, static const uint8_t crazy_mix_runlevel[32] = {
0x26,0x72,0x35,0x54,0x27,0x82,0x45,0x64, 0x22, 0x32, 0x33, 0x53, 0x23, 0x42, 0x43, 0x63,
0x28,0x92,0x36,0x74,0x29,0xa2,0x46,0x84}; 0x24, 0x52, 0x34, 0x73, 0x25, 0x62, 0x44, 0x83,
0x26, 0x72, 0x35, 0x54, 0x27, 0x82, 0x45, 0x64,
(*final)=!(i&1); 0x28, 0x92, 0x36, 0x74, 0x29, 0xa2, 0x46, 0x84,
e=get_bits(&s->gb,5);//get the extra bits };
(*run) =crazy_mix_runlevel[e]>>4;
(*level)=crazy_mix_runlevel[e]&0x0F; (*final) = !(i & 1);
}else{ e = get_bits(&s->gb, 5); // get the extra bits
(*level)=get_bits( &s->gb, 7-3*(i&1)); (*run) = crazy_mix_runlevel[e] >> 4;
(*run) =get_bits( &s->gb, 6); (*level) = crazy_mix_runlevel[e] & 0x0F;
(*final)=get_bits1(&s->gb); } else {
(*level) = get_bits(&s->gb, 7 - 3 * (i & 1));
(*run) = get_bits(&s->gb, 6);
(*final) = get_bits1(&s->gb);
} }
return; return;
} }
//static const uint8_t dc_extra_sbits[] ={0, 1,1, 1,1, 2,2, 3,3, 4,4, 5,5, 6,6, 7,7 }; /* static const uint8_t dc_extra_sbits[] = {
static const uint8_t dc_index_offset[] ={ 0, 1,2, 3,4, 5,7, 9,13, 17,25, 33,49, 65,97, 129,193}; * 0, 1, 1, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7,
* }; */
static int x8_get_dc_rlf(IntraX8Context * const w,int const mode, int * const level, int * const final){ static const uint8_t dc_index_offset[] = {
MpegEncContext * const s= w->s; 0, 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
int i,e,c; };
av_assert2(mode<3); static int x8_get_dc_rlf(IntraX8Context *const w,
if( !w->j_dc_vlc[mode] ) { int const mode, int *const level, int *const final)
int table_index; {
table_index = get_bits(&s->gb, 3); MpegEncContext *const s = w->s;
//4 modes, same table int i, e, c;
w->j_dc_vlc[mode]= &j_dc_vlc[w->quant<13][table_index];
av_assert2(mode < 3);
if (!w->j_dc_vlc[mode]) {
int table_index = get_bits(&s->gb, 3);
// 4 modes, same table
w->j_dc_vlc[mode] = &j_dc_vlc[w->quant < 13][table_index];
} }
i=get_vlc2(&s->gb, w->j_dc_vlc[mode]->table, DC_VLC_BITS, DC_VLC_MTD); i = get_vlc2(&s->gb, w->j_dc_vlc[mode]->table, DC_VLC_BITS, DC_VLC_MTD);
/*(i>=17) {i-=17;final=1;}*/ /* (i >= 17) { i -= 17; final =1; } */
c= i>16; c = i > 16;
(*final)=c; (*final) = c;
i-=17*c; i -= 17 * c;
if(i<=0){ if (i <= 0) {
(*level)=0; (*level) = 0;
return -i; return -i;
} }
c=(i+1)>>1;//hackish way to calculate dc_extra_sbits[] c = (i + 1) >> 1; // hackish way to calculate dc_extra_sbits[]
c-=c>1; c -= c > 1;
e=get_bits(&s->gb,c);//get the extra bits e = get_bits(&s->gb, c); // get the extra bits
i=dc_index_offset[i]+(e>>1); i = dc_index_offset[i] + (e >> 1);
e= -(e & 1);//0,0xffffff e = -(e & 1); // 0, 0xffffff
(*level)= (i ^ e) - e;// (i^0)-0 , (i^0xff)-(-1) (*level) = (i ^ e) - e; // (i ^ 0) -0 , (i ^ 0xff) - (-1)
return 0; return 0;
} }
//end of huffman
static int x8_setup_spatial_predictor(IntraX8Context * const w, const int chroma){ // end of huffman
MpegEncContext * const s= w->s;
static int x8_setup_spatial_predictor(IntraX8Context *const w, const int chroma)
{
MpegEncContext *const s = w->s;
int range; int range;
int sum; int sum;
int quant; int quant;
w->dsp.setup_spatial_compensation(s->dest[chroma], s->sc.edge_emu_buffer, w->dsp.setup_spatial_compensation(s->dest[chroma], s->sc.edge_emu_buffer,
s->current_picture.f->linesize[chroma>0], s->current_picture.f->linesize[chroma > 0],
&range, &sum, w->edges); &range, &sum, w->edges);
if(chroma){ if (chroma) {
w->orient=w->chroma_orient; w->orient = w->chroma_orient;
quant=w->quant_dc_chroma; quant = w->quant_dc_chroma;
}else{ } else {
quant=w->quant; quant = w->quant;
} }
w->flat_dc=0; w->flat_dc = 0;
if(range < quant || range < 3){ if (range < quant || range < 3) {
w->orient=0; w->orient = 0;
if(range < 3){//yep you read right, a +-1 idct error may break decoding!
w->flat_dc=1; // yep you read right, a +-1 idct error may break decoding!
sum+=9; if (range < 3) {
w->predicted_dc = (sum*6899)>>17;//((1<<17)+9)/(8+8+1+2)=6899 w->flat_dc = 1;
sum += 9;
// ((1 << 17) + 9) / (8 + 8 + 1 + 2) = 6899
w->predicted_dc = (sum * 6899) >> 17;
} }
} }
if(chroma) if (chroma)
return 0; return 0;
av_assert2(w->orient < 3); av_assert2(w->orient < 3);
if(range < 2*w->quant){ if (range < 2 * w->quant) {
if( (w->edges&3) == 0){ if ((w->edges & 3) == 0) {
if(w->orient==1) w->orient=11; if (w->orient == 1)
if(w->orient==2) w->orient=10; w->orient = 11;
}else{ if (w->orient == 2)
w->orient=0; w->orient = 10;
} else {
w->orient = 0;
} }
w->raw_orient=0; w->raw_orient = 0;
}else{ } else {
static const uint8_t prediction_table[3][12]={ static const uint8_t prediction_table[3][12] = {
{0,8,4, 10,11, 2,6,9,1,3,5,7}, { 0, 8, 4, 10, 11, 2, 6, 9, 1, 3, 5, 7 },
{4,0,8, 11,10, 3,5,2,6,9,1,7}, { 4, 0, 8, 11, 10, 3, 5, 2, 6, 9, 1, 7 },
{8,0,4, 10,11, 1,7,2,6,9,3,5} { 8, 0, 4, 10, 11, 1, 7, 2, 6, 9, 3, 5 },
}; };
w->raw_orient=x8_get_orient_vlc(w); w->raw_orient = x8_get_orient_vlc(w);
if(w->raw_orient<0) return -1; if (w->raw_orient < 0)
av_assert2(w->raw_orient < 12 ); return -1;
av_assert2(w->orient<3); av_assert2(w->raw_orient < 12);
av_assert2(w->orient < 3);
w->orient=prediction_table[w->orient][w->raw_orient]; w->orient=prediction_table[w->orient][w->raw_orient];
} }
return 0; return 0;
} }
static void x8_update_predictions(IntraX8Context * const w, const int orient, const int est_run ){ static void x8_update_predictions(IntraX8Context *const w, const int orient,
MpegEncContext * const s= w->s; const int est_run)
{
MpegEncContext *const s = w->s;
w->prediction_table[s->mb_x*2+(s->mb_y&1)] = (est_run<<2) + 1*(orient==4) + 2*(orient==8); w->prediction_table[s->mb_x * 2 + (s->mb_y & 1)] = (est_run << 2) + 1 * (orient == 4) + 2 * (orient == 8);
/* /*
y=2n+0 ->//0 2 4 * y = 2n + 0 -> // 0 2 4
y=2n+1 ->//1 3 5 * y = 2n + 1 -> // 1 3 5
*/ */
} }
static void x8_get_prediction_chroma(IntraX8Context * const w){
MpegEncContext * const s= w->s;
w->edges = 1*( !(s->mb_x>>1) ); static void x8_get_prediction_chroma(IntraX8Context *const w)
w->edges|= 2*( !(s->mb_y>>1) ); {
w->edges|= 4*( s->mb_x >= (2*s->mb_width-1) );//mb_x for chroma would always be odd MpegEncContext *const s = w->s;
w->raw_orient=0; w->edges = 1 * (!(s->mb_x >> 1));
if(w->edges&3){//lut_co[8]={inv,4,8,8, inv,4,8,8}<- =>{1,1,0,0;1,1,0,0} => 0xCC w->edges |= 2 * (!(s->mb_y >> 1));
w->chroma_orient=4<<((0xCC>>w->edges)&1); w->edges |= 4 * (s->mb_x >= (2 * s->mb_width - 1)); // mb_x for chroma would always be odd
w->raw_orient = 0;
// lut_co[8] = {inv,4,8,8, inv,4,8,8} <- => {1,1,0,0;1,1,0,0} => 0xCC
if (w->edges & 3) {
w->chroma_orient = 4 << ((0xCC >> w->edges) & 1);
return; return;
} }
w->chroma_orient = (w->prediction_table[2*s->mb_x-2] & 0x03)<<2;//block[x-1][y|1-1)] // block[x - 1][y | 1 - 1)]
w->chroma_orient = (w->prediction_table[2 * s->mb_x - 2] & 0x03) << 2;
} }
static void x8_get_prediction(IntraX8Context * const w){ static void x8_get_prediction(IntraX8Context *const w)
MpegEncContext * const s= w->s; {
int a,b,c,i; MpegEncContext *const s = w->s;
int a, b, c, i;
w->edges = 1*( !s->mb_x );
w->edges|= 2*( !s->mb_y ); w->edges = 1 * (!s->mb_x);
w->edges|= 4*( s->mb_x >= (2*s->mb_width-1) ); w->edges |= 2 * (!s->mb_y);
w->edges |= 4 * (s->mb_x >= (2 * s->mb_width - 1));
switch(w->edges&3){
case 0: switch (w->edges & 3) {
break; case 0:
case 1: break;
//take the one from the above block[0][y-1] case 1:
w->est_run = w->prediction_table[!(s->mb_y&1)]>>2; // take the one from the above block[0][y - 1]
w->orient = 1; w->est_run = w->prediction_table[!(s->mb_y & 1)] >> 2;
return; w->orient = 1;
case 2: return;
//take the one from the previous block[x-1][0] case 2:
w->est_run = w->prediction_table[2*s->mb_x-2]>>2; // take the one from the previous block[x - 1][0]
w->orient = 2; w->est_run = w->prediction_table[2 * s->mb_x - 2] >> 2;
return; w->orient = 2;
case 3: return;
w->est_run = 16; case 3:
w->orient = 0; w->est_run = 16;
return; w->orient = 0;
return;
} }
//no edge cases // no edge cases
b= w->prediction_table[2*s->mb_x + !(s->mb_y&1) ];//block[x ][y-1] b = w->prediction_table[2 * s->mb_x + !(s->mb_y & 1)]; // block[x ][y - 1]
a= w->prediction_table[2*s->mb_x-2 + (s->mb_y&1) ];//block[x-1][y ] a = w->prediction_table[2 * s->mb_x - 2 + (s->mb_y & 1)]; // block[x - 1][y ]
c= w->prediction_table[2*s->mb_x-2 + !(s->mb_y&1) ];//block[x-1][y-1] c = w->prediction_table[2 * s->mb_x - 2 + !(s->mb_y & 1)]; // block[x - 1][y - 1]
w->est_run = FFMIN(b,a); w->est_run = FFMIN(b, a);
/* This condition has nothing to do with w->edges, even if it looks /* This condition has nothing to do with w->edges, even if it looks
similar it would trigger if e.g. x=3;y=2; * similar it would trigger if e.g. x = 3; y = 2;
I guess somebody wrote something wrong and it became standard. */ * I guess somebody wrote something wrong and it became standard. */
if( (s->mb_x & s->mb_y) != 0 ) w->est_run=FFMIN(c,w->est_run); if ((s->mb_x & s->mb_y) != 0)
w->est_run>>=2; w->est_run = FFMIN(c, w->est_run);
w->est_run >>= 2;
a&=3;
b&=3; a &= 3;
c&=3; b &= 3;
c &= 3;
i=( 0xFFEAF4C4>>(2*b+8*a) )&3;
if(i!=3) w->orient=i; i = (0xFFEAF4C4 >> (2 * b + 8 * a)) & 3;
else w->orient=( 0xFFEAD8>>(2*c+8*(w->quant>12)) )&3; if (i != 3)
w->orient = i;
else
w->orient = (0xFFEAD8 >> (2 * c + 8 * (w->quant > 12))) & 3;
/* /*
lut1[b][a]={ * lut1[b][a] = {
->{0, 1, 0, pad}, * ->{ 0, 1, 0, pad },
{0, 1, X, pad}, * { 0, 1, X, pad },
{2, 2, 2, pad}} * { 2, 2, 2, pad }
pad 2 2 2; pad X 1 0; pad 0 1 0 <- * }
-> 11 10 '10 10 '11 11'01 00 '11 00'01 00=>0xEAF4C4 * pad 2 2 2;
* pad X 1 0;
lut2[q>12][c]={ * pad 0 1 0 <-
->{0,2,1,pad}, * -> 11 10 '10 10 '11 11'01 00 '11 00'01 00 => 0xEAF4C4
{2,2,2,pad}} *
pad 2 2 2; pad 1 2 0 <- * lut2[q>12][c] = {
-> 11 10'10 10 '11 01'10 00=>0xEAD8 * ->{ 0, 2, 1, pad},
*/ * { 2, 2, 2, pad}
* }
* pad 2 2 2;
* pad 1 2 0 <-
* -> 11 10'10 10 '11 01'10 00 => 0xEAD8
*/
} }
static void x8_ac_compensation(IntraX8Context *const w, int const direction,
static void x8_ac_compensation(IntraX8Context * const w, int const direction, int const dc_level){ int const dc_level)
MpegEncContext * const s= w->s; {
MpegEncContext *const s = w->s;
int t; int t;
#define B(x,y) s->block[0][w->idct_permutation[(x)+(y)*8]] #define B(x,y) s->block[0][w->idct_permutation[(x)+(y)*8]]
#define T(x) ((x) * dc_level + 0x8000) >> 16; #define T(x) ((x) * dc_level + 0x8000) >> 16;
switch(direction){ switch (direction) {
case 0: case 0:
t = T(3811);//h t = T(3811); // h
B(1,0) -= t; B(1, 0) -= t;
B(0,1) -= t; B(0, 1) -= t;
t = T(487);//e t = T(487); // e
B(2,0) -= t; B(2, 0) -= t;
B(0,2) -= t; B(0, 2) -= t;
t = T(506);//f t = T(506); // f
B(3,0) -= t; B(3, 0) -= t;
B(0,3) -= t; B(0, 3) -= t;
t = T(135);//c t = T(135); // c
B(4,0) -= t; B(4, 0) -= t;
B(0,4) -= t; B(0, 4) -= t;
B(2,1) += t; B(2, 1) += t;
B(1,2) += t; B(1, 2) += t;
B(3,1) += t; B(3, 1) += t;
B(1,3) += t; B(1, 3) += t;
t = T(173);//d t = T(173); // d
B(5,0) -= t; B(5, 0) -= t;
B(0,5) -= t; B(0, 5) -= t;
t = T(61);//b t = T(61); // b
B(6,0) -= t; B(6, 0) -= t;
B(0,6) -= t; B(0, 6) -= t;
B(5,1) += t; B(5, 1) += t;
B(1,5) += t; B(1, 5) += t;
t = T(42); //a t = T(42); // a
B(7,0) -= t; B(7, 0) -= t;
B(0,7) -= t; B(0, 7) -= t;
B(4,1) += t; B(4, 1) += t;
B(1,4) += t; B(1, 4) += t;
B(4,4) += t; B(4, 4) += t;
t = T(1084);//g t = T(1084); // g
B(1,1) += t; B(1, 1) += t;
s->block_last_index[0] = FFMAX(s->block_last_index[0], 7*8); s->block_last_index[0] = FFMAX(s->block_last_index[0], 7 * 8);
break; break;
case 1: case 1:
B(0,1) -= T(6269); B(0, 1) -= T(6269);
B(0,3) -= T( 708); B(0, 3) -= T(708);
B(0,5) -= T( 172); B(0, 5) -= T(172);
B(0,7) -= T( 73); B(0, 7) -= T(73);
s->block_last_index[0] = FFMAX(s->block_last_index[0], 7*8); s->block_last_index[0] = FFMAX(s->block_last_index[0], 7 * 8);
break; break;
case 2: case 2:
B(1,0) -= T(6269); B(1, 0) -= T(6269);
B(3,0) -= T( 708); B(3, 0) -= T(708);
B(5,0) -= T( 172); B(5, 0) -= T(172);
B(7,0) -= T( 73); B(7, 0) -= T(73);
s->block_last_index[0] = FFMAX(s->block_last_index[0], 7); s->block_last_index[0] = FFMAX(s->block_last_index[0], 7);
break; break;
...@@ -507,172 +554,177 @@ static void x8_ac_compensation(IntraX8Context * const w, int const direction, in ...@@ -507,172 +554,177 @@ static void x8_ac_compensation(IntraX8Context * const w, int const direction, in
#undef T #undef T
} }
static void dsp_x8_put_solidcolor(uint8_t const pix, uint8_t * dst, int const linesize){ static void dsp_x8_put_solidcolor(uint8_t const pix, uint8_t *dst,
int const linesize)
{
int k; int k;
for(k=0;k<8;k++){ for (k = 0; k < 8; k++) {
memset(dst,pix,8); memset(dst, pix, 8);
dst+=linesize; dst += linesize;
} }
} }
static const int16_t quant_table[64] = { static const int16_t quant_table[64] = {
256, 256, 256, 256, 256, 256, 259, 262, 256, 256, 256, 256, 256, 256, 259, 262,
265, 269, 272, 275, 278, 282, 285, 288, 265, 269, 272, 275, 278, 282, 285, 288,
292, 295, 299, 303, 306, 310, 314, 317, 292, 295, 299, 303, 306, 310, 314, 317,
321, 325, 329, 333, 337, 341, 345, 349, 321, 325, 329, 333, 337, 341, 345, 349,
353, 358, 362, 366, 371, 375, 379, 384, 353, 358, 362, 366, 371, 375, 379, 384,
389, 393, 398, 403, 408, 413, 417, 422, 389, 393, 398, 403, 408, 413, 417, 422,
428, 433, 438, 443, 448, 454, 459, 465, 428, 433, 438, 443, 448, 454, 459, 465,
470, 476, 482, 488, 493, 499, 505, 511 470, 476, 482, 488, 493, 499, 505, 511,
}; };
static int x8_decode_intra_mb(IntraX8Context* const w, const int chroma){ static int x8_decode_intra_mb(IntraX8Context *const w, const int chroma)
MpegEncContext * const s= w->s; {
MpegEncContext *const s = w->s;
uint8_t * scantable; uint8_t *scantable;
int final,run,level; int final, run, level;
int ac_mode,dc_mode,est_run,dc_level; int ac_mode, dc_mode, est_run, dc_level;
int pos,n; int pos, n;
int zeros_only; int zeros_only;
int use_quant_matrix; int use_quant_matrix;
int sign; int sign;
av_assert2(w->orient<12); av_assert2(w->orient < 12);
s->bdsp.clear_block(s->block[0]); s->bdsp.clear_block(s->block[0]);
if(chroma){ if (chroma)
dc_mode=2; dc_mode = 2;
}else{ else
dc_mode=!!w->est_run;//0,1 dc_mode = !!w->est_run; // 0, 1
}
if (x8_get_dc_rlf(w, dc_mode, &dc_level, &final))
if(x8_get_dc_rlf(w, dc_mode, &dc_level, &final)) return -1; return -1;
n=0; n = 0;
zeros_only=0; zeros_only = 0;
if(!final){//decode ac if (!final) { // decode ac
use_quant_matrix=w->use_quant_matrix; use_quant_matrix = w->use_quant_matrix;
if(chroma){ if (chroma) {
ac_mode = 1; ac_mode = 1;
est_run = 64;//not used est_run = 64; // not used
}else{ } else {
if (w->raw_orient < 3){ if (w->raw_orient < 3)
use_quant_matrix = 0; use_quant_matrix = 0;
}
if(w->raw_orient > 4){ if (w->raw_orient > 4) {
ac_mode = 0; ac_mode = 0;
est_run = 64; est_run = 64;
}else{ } else {
if(w->est_run > 1){ if (w->est_run > 1) {
ac_mode = 2; ac_mode = 2;
est_run=w->est_run; est_run = w->est_run;
}else{ } else {
ac_mode = 3; ac_mode = 3;
est_run = 64; est_run = 64;
} }
} }
} }
x8_select_ac_table(w,ac_mode); x8_select_ac_table(w, ac_mode);
/*scantable_selector[12]={0,2,0,1,1,1,0,2,2,0,1,2};<- /* scantable_selector[12] = { 0, 2, 0, 1, 1, 1, 0, 2, 2, 0, 1, 2 }; <-
-> 10'01' 00'10' 10'00' 01'01' 01'00' 10'00 =>0x928548 */ * -> 10'01' 00'10' 10'00' 01'01' 01'00' 10'00 => 0x928548 */
scantable = w->scantable[ (0x928548>>(2*w->orient))&3 ].permutated; scantable = w->scantable[(0x928548 >> (2 * w->orient)) & 3].permutated;
pos=0; pos = 0;
do { do {
n++; n++;
if( n >= est_run ){ if (n >= est_run) {
ac_mode=3; ac_mode = 3;
x8_select_ac_table(w,3); x8_select_ac_table(w, 3);
} }
x8_get_ac_rlf(w,ac_mode,&run,&level,&final); x8_get_ac_rlf(w, ac_mode, &run, &level, &final);
pos+=run+1; pos += run + 1;
if(pos>63){ if (pos > 63) {
//this also handles vlc error in x8_get_ac_rlf // this also handles vlc error in x8_get_ac_rlf
return -1; return -1;
} }
level= (level+1) * w->dquant; level = (level + 1) * w->dquant;
level+= w->qsum; level += w->qsum;
sign = - get_bits1(&s->gb); sign = -get_bits1(&s->gb);
level = (level ^ sign) - sign; level = (level ^ sign) - sign;
if(use_quant_matrix){ if (use_quant_matrix)
level = (level*quant_table[pos])>>8; level = (level * quant_table[pos]) >> 8;
}
s->block[0][ scantable[pos] ]=level;
}while(!final);
s->block_last_index[0]=pos; s->block[0][scantable[pos]] = level;
}else{//DC only } while (!final);
s->block_last_index[0]=0;
if(w->flat_dc && ((unsigned)(dc_level+1)) < 3){//[-1;1]
int32_t divide_quant= !chroma ? w->divide_quant_dc_luma:
w->divide_quant_dc_chroma;
int32_t dc_quant = !chroma ? w->quant:
w->quant_dc_chroma;
//original intent dc_level+=predicted_dc/quant; but it got lost somewhere in the rounding s->block_last_index[0] = pos;
dc_level+= (w->predicted_dc*divide_quant + (1<<12) )>>13; } else { // DC only
s->block_last_index[0] = 0;
if (w->flat_dc && ((unsigned) (dc_level + 1)) < 3) { // [-1; 1]
int32_t divide_quant = !chroma ? w->divide_quant_dc_luma
: w->divide_quant_dc_chroma;
int32_t dc_quant = !chroma ? w->quant
: w->quant_dc_chroma;
dsp_x8_put_solidcolor( av_clip_uint8((dc_level*dc_quant+4)>>3), // original intent dc_level += predicted_dc/quant;
s->dest[chroma], s->current_picture.f->linesize[!!chroma]); // but it got lost somewhere in the rounding
dc_level += (w->predicted_dc * divide_quant + (1 << 12)) >> 13;
dsp_x8_put_solidcolor(av_clip_uint8((dc_level * dc_quant + 4) >> 3),
s->dest[chroma],
s->current_picture.f->linesize[!!chroma]);
goto block_placed; goto block_placed;
} }
zeros_only = (dc_level == 0); zeros_only = (dc_level == 0);
} }
if(!chroma){ if (!chroma)
s->block[0][0] = dc_level*w->quant; s->block[0][0] = dc_level * w->quant;
}else{ else
s->block[0][0] = dc_level*w->quant_dc_chroma; s->block[0][0] = dc_level * w->quant_dc_chroma;
}
//there is !zero_only check in the original, but dc_level check is enough // there is !zero_only check in the original, but dc_level check is enough
if( (unsigned int)(dc_level+1) >= 3 && (w->edges&3) != 3 ){ if ((unsigned int) (dc_level + 1) >= 3 && (w->edges & 3) != 3) {
int direction; int direction;
/*ac_comp_direction[orient] = { 0, 3, 3, 1, 1, 0, 0, 0, 2, 2, 2, 1 };<- /* ac_comp_direction[orient] = { 0, 3, 3, 1, 1, 0, 0, 0, 2, 2, 2, 1 }; <-
-> 01'10' 10'10' 00'00' 00'01' 01'11' 11'00 =>0x6A017C */ * -> 01'10' 10'10' 00'00' 00'01' 01'11' 11'00 => 0x6A017C */
direction= (0x6A017C>>(w->orient*2))&3; direction = (0x6A017C >> (w->orient * 2)) & 3;
if (direction != 3){ if (direction != 3) {
x8_ac_compensation(w, direction, s->block[0][0]);//modify block_last[] // modify block_last[]
x8_ac_compensation(w, direction, s->block[0][0]);
} }
} }
if(w->flat_dc){ if (w->flat_dc) {
dsp_x8_put_solidcolor(w->predicted_dc, s->dest[chroma], s->current_picture.f->linesize[!!chroma]); dsp_x8_put_solidcolor(w->predicted_dc, s->dest[chroma],
}else{ s->current_picture.f->linesize[!!chroma]);
w->dsp.spatial_compensation[w->orient]( s->sc.edge_emu_buffer, } else {
s->dest[chroma], w->dsp.spatial_compensation[w->orient](s->sc.edge_emu_buffer,
s->current_picture.f->linesize[!!chroma] ); s->dest[chroma],
s->current_picture.f->linesize[!!chroma]);
} }
if(!zeros_only) if (!zeros_only)
w->wdsp.idct_add(s->dest[chroma], w->wdsp.idct_add(s->dest[chroma],
s->current_picture.f->linesize[!!chroma], s->current_picture.f->linesize[!!chroma],
s->block[0]); s->block[0]);
block_placed: block_placed:
if (!chroma)
x8_update_predictions(w, w->orient, n);
if(!chroma){ if (s->loop_filter) {
x8_update_predictions(w,w->orient,n); uint8_t *ptr = s->dest[chroma];
}
if(s->loop_filter){
uint8_t* ptr = s->dest[chroma];
int linesize = s->current_picture.f->linesize[!!chroma]; int linesize = s->current_picture.f->linesize[!!chroma];
if(!( (w->edges&2) || ( zeros_only && (w->orient|4)==4 ) )){ if (!((w->edges & 2) || (zeros_only && (w->orient | 4) == 4)))
w->dsp.h_loop_filter(ptr, linesize, w->quant); w->dsp.h_loop_filter(ptr, linesize, w->quant);
}
if(!( (w->edges&1) || ( zeros_only && (w->orient|8)==8 ) )){ if (!((w->edges & 1) || (zeros_only && (w->orient | 8) == 8)))
w->dsp.v_loop_filter(ptr, linesize, w->quant); w->dsp.v_loop_filter(ptr, linesize, w->quant);
}
} }
return 0; return 0;
} }
static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_* // FIXME maybe merge with ff_*
//not s->linesize as this would be wrong for field pics static void x8_init_block_index(MpegEncContext *s)
//not that IntraX8 has interlacing support ;) {
// not s->linesize as this would be wrong for field pics
// not that IntraX8 has interlacing support ;)
const int linesize = s->current_picture.f->linesize[0]; const int linesize = s->current_picture.f->linesize[0];
const int uvlinesize = s->current_picture.f->linesize[1]; const int uvlinesize = s->current_picture.f->linesize[1];
...@@ -680,9 +732,10 @@ static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_ ...@@ -680,9 +732,10 @@ static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_
s->dest[1] = s->current_picture.f->data[1]; s->dest[1] = s->current_picture.f->data[1];
s->dest[2] = s->current_picture.f->data[2]; s->dest[2] = s->current_picture.f->data[2];
s->dest[0] += s->mb_y * linesize << 3; s->dest[0] += s->mb_y * linesize << 3;
s->dest[1] += ( s->mb_y&(~1) ) * uvlinesize << 2;//chroma blocks are on add rows // chroma blocks are on add rows
s->dest[2] += ( s->mb_y&(~1) ) * uvlinesize << 2; s->dest[1] += (s->mb_y & (~1)) * uvlinesize << 2;
s->dest[2] += (s->mb_y & (~1)) * uvlinesize << 2;
} }
/** /**
...@@ -691,20 +744,26 @@ static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_ ...@@ -691,20 +744,26 @@ static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_
* @param w pointer to IntraX8Context * @param w pointer to IntraX8Context
* @param s pointer to MpegEncContext of the parent codec * @param s pointer to MpegEncContext of the parent codec
*/ */
av_cold void ff_intrax8_common_init(IntraX8Context * w, MpegEncContext * const s){ av_cold void ff_intrax8_common_init(IntraX8Context *w, MpegEncContext *const s)
{
w->s=s; w->s = s;
x8_vlc_init(); x8_vlc_init();
av_assert0(s->mb_width>0); av_assert0(s->mb_width > 0);
w->prediction_table=av_mallocz(s->mb_width*2*2);//two rows, 2 blocks per cannon mb
//two rows, 2 blocks per cannon mb
w->prediction_table = av_mallocz(s->mb_width * 2 * 2);
ff_wmv2dsp_init(&w->wdsp); ff_wmv2dsp_init(&w->wdsp);
ff_init_scantable_permutation(w->idct_permutation, ff_init_scantable_permutation(w->idct_permutation,
w->wdsp.idct_perm); w->wdsp.idct_perm);
ff_init_scantable(w->idct_permutation, &w->scantable[0], ff_wmv1_scantable[0]); ff_init_scantable(w->idct_permutation, &w->scantable[0],
ff_init_scantable(w->idct_permutation, &w->scantable[1], ff_wmv1_scantable[2]); ff_wmv1_scantable[0]);
ff_init_scantable(w->idct_permutation, &w->scantable[2], ff_wmv1_scantable[3]); ff_init_scantable(w->idct_permutation, &w->scantable[1],
ff_wmv1_scantable[2]);
ff_init_scantable(w->idct_permutation, &w->scantable[2],
ff_wmv1_scantable[3]);
ff_intrax8dsp_init(&w->dsp); ff_intrax8dsp_init(&w->dsp);
} }
...@@ -713,7 +772,7 @@ av_cold void ff_intrax8_common_init(IntraX8Context * w, MpegEncContext * const s ...@@ -713,7 +772,7 @@ av_cold void ff_intrax8_common_init(IntraX8Context * w, MpegEncContext * const s
* Destroy IntraX8 frame structure. * Destroy IntraX8 frame structure.
* @param w pointer to IntraX8Context * @param w pointer to IntraX8Context
*/ */
av_cold void ff_intrax8_common_end(IntraX8Context * w) av_cold void ff_intrax8_common_end(IntraX8Context *w)
{ {
av_freep(&w->prediction_table); av_freep(&w->prediction_table);
} }
...@@ -729,8 +788,10 @@ av_cold void ff_intrax8_common_end(IntraX8Context * w) ...@@ -729,8 +788,10 @@ av_cold void ff_intrax8_common_end(IntraX8Context * w)
* @param dquant doubled quantizer, it would be odd in case of VC-1 halfpq==1. * @param dquant doubled quantizer, it would be odd in case of VC-1 halfpq==1.
* @param quant_offset offset away from zero * @param quant_offset offset away from zero
*/ */
int ff_intrax8_decode_picture(IntraX8Context * const w, int dquant, int quant_offset){ int ff_intrax8_decode_picture(IntraX8Context *const w, int dquant,
MpegEncContext * const s= w->s; int quant_offset)
{
MpegEncContext *const s = w->s;
int mb_xy; int mb_xy;
w->use_quant_matrix = get_bits1(&s->gb); w->use_quant_matrix = get_bits1(&s->gb);
...@@ -738,51 +799,53 @@ int ff_intrax8_decode_picture(IntraX8Context * const w, int dquant, int quant_of ...@@ -738,51 +799,53 @@ int ff_intrax8_decode_picture(IntraX8Context * const w, int dquant, int quant_of
w->quant = dquant >> 1; w->quant = dquant >> 1;
w->qsum = quant_offset; w->qsum = quant_offset;
w->divide_quant_dc_luma = ((1<<16) + (w->quant>>1)) / w->quant; w->divide_quant_dc_luma = ((1 << 16) + (w->quant >> 1)) / w->quant;
if(w->quant < 5){ if (w->quant < 5) {
w->quant_dc_chroma = w->quant; w->quant_dc_chroma = w->quant;
w->divide_quant_dc_chroma = w->divide_quant_dc_luma; w->divide_quant_dc_chroma = w->divide_quant_dc_luma;
}else{ } else {
w->quant_dc_chroma = w->quant+((w->quant+3)>>3); w->quant_dc_chroma = w->quant + ((w->quant + 3) >> 3);
w->divide_quant_dc_chroma = ((1<<16) + (w->quant_dc_chroma>>1)) / w->quant_dc_chroma; w->divide_quant_dc_chroma = ((1 << 16) + (w->quant_dc_chroma >> 1)) / w->quant_dc_chroma;
} }
x8_reset_vlc_tables(w); x8_reset_vlc_tables(w);
for (s->mb_y = 0; s->mb_y < s->mb_height * 2; s->mb_y++) {
for(s->mb_y=0; s->mb_y < s->mb_height*2; s->mb_y++){
x8_init_block_index(s); x8_init_block_index(s);
mb_xy=(s->mb_y>>1)*s->mb_stride; mb_xy = (s->mb_y >> 1) * s->mb_stride;
for(s->mb_x=0; s->mb_x < s->mb_width*2; s->mb_x++){ for (s->mb_x = 0; s->mb_x < s->mb_width * 2; s->mb_x++) {
x8_get_prediction(w); x8_get_prediction(w);
if(x8_setup_spatial_predictor(w,0)) goto error; if (x8_setup_spatial_predictor(w, 0))
if(x8_decode_intra_mb(w,0)) goto error; goto error;
if (x8_decode_intra_mb(w, 0))
goto error;
if( s->mb_x & s->mb_y & 1 ){ if (s->mb_x & s->mb_y & 1) {
x8_get_prediction_chroma(w); x8_get_prediction_chroma(w);
/*when setting up chroma, no vlc is read, /* when setting up chroma, no vlc is read,
so no error condition can be reached*/ * so no error condition can be reached */
x8_setup_spatial_predictor(w,1); x8_setup_spatial_predictor(w, 1);
if(x8_decode_intra_mb(w,1)) goto error; if (x8_decode_intra_mb(w, 1))
goto error;
x8_setup_spatial_predictor(w,2); x8_setup_spatial_predictor(w, 2);
if(x8_decode_intra_mb(w,2)) goto error; if (x8_decode_intra_mb(w, 2))
goto error;
s->dest[1]+= 8; s->dest[1] += 8;
s->dest[2]+= 8; s->dest[2] += 8;
/*emulate MB info in the relevant tables*/ /* emulate MB info in the relevant tables */
s->mbskip_table [mb_xy]=0; s->mbskip_table[mb_xy] = 0;
s->mbintra_table[mb_xy]=1; s->mbintra_table[mb_xy] = 1;
s->current_picture.qscale_table[mb_xy] = w->quant; s->current_picture.qscale_table[mb_xy] = w->quant;
mb_xy++; mb_xy++;
} }
s->dest[0]+= 8; s->dest[0] += 8;
}
if(s->mb_y&1){
ff_mpeg_draw_horiz_band(s, (s->mb_y-1)*8, 16);
} }
if (s->mb_y & 1)
ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 8, 16);
} }
error: error:
......
...@@ -25,27 +25,31 @@ ...@@ -25,27 +25,31 @@
#include "wmv2dsp.h" #include "wmv2dsp.h"
typedef struct IntraX8Context { typedef struct IntraX8Context {
VLC * j_ac_vlc[4];//they point to the static j_mb_vlc VLC *j_ac_vlc[4]; // they point to the static j_mb_vlc
VLC * j_orient_vlc; VLC *j_orient_vlc;
VLC * j_dc_vlc[3]; VLC *j_dc_vlc[3];
int use_quant_matrix; int use_quant_matrix;
//set by ff_intrax8_common_init
uint8_t * prediction_table;//2*(mb_w*2) // set by ff_intrax8_common_init
uint8_t *prediction_table; // 2 * (mb_w * 2)
ScanTable scantable[3]; ScanTable scantable[3];
WMV2DSPContext wdsp; WMV2DSPContext wdsp;
uint8_t idct_permutation[64]; uint8_t idct_permutation[64];
//set by the caller codec
//set by the caller codec
MpegEncContext * s; MpegEncContext * s;
IntraX8DSPContext dsp; IntraX8DSPContext dsp;
int quant; int quant;
int dquant; int dquant;
int qsum; int qsum;
//calculated per frame
// calculated per frame
int quant_dc_chroma; int quant_dc_chroma;
int divide_quant_dc_luma; int divide_quant_dc_luma;
int divide_quant_dc_chroma; int divide_quant_dc_chroma;
//changed per block
// changed per block
int edges; int edges;
int flat_dc; int flat_dc;
int predicted_dc; int predicted_dc;
...@@ -55,8 +59,8 @@ typedef struct IntraX8Context { ...@@ -55,8 +59,8 @@ typedef struct IntraX8Context {
int est_run; int est_run;
} IntraX8Context; } IntraX8Context;
void ff_intrax8_common_init(IntraX8Context * w, MpegEncContext * const s); void ff_intrax8_common_init(IntraX8Context *w, MpegEncContext *const s);
void ff_intrax8_common_end(IntraX8Context * w); void ff_intrax8_common_end(IntraX8Context *w);
int ff_intrax8_decode_picture(IntraX8Context * w, int quant, int halfpq); int ff_intrax8_decode_picture(IntraX8Context *w, int quant, int halfpq);
#endif /* AVCODEC_INTRAX8_H */ #endif /* AVCODEC_INTRAX8_H */
...@@ -17,7 +17,7 @@ ...@@ -17,7 +17,7 @@
*/ */
/** /**
* @file * @file
*@brief IntraX8 frame subdecoder image manipulation routines *@brief IntraX8 frame subdecoder image manipulation routines
*/ */
...@@ -25,27 +25,27 @@ ...@@ -25,27 +25,27 @@
#include "libavutil/common.h" #include "libavutil/common.h"
/* /*
area positions, #3 is 1 pixel only, other are 8 pixels * area positions, #3 is 1 pixel only, other are 8 pixels
|66666666| * |66666666|
3|44444444|55555555| * 3|44444444|55555555|
- -+--------+--------+ * - -+--------+--------+
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
1 2|XXXXXXXX| * 1 2|XXXXXXXX|
^-start * ^-start
*/ */
#define area1 (0) #define area1 (0)
#define area2 (8) #define area2 (8)
#define area3 (8+8) #define area3 (8 + 8)
#define area4 (8+8+1) #define area4 (8 + 8 + 1)
#define area5 (8+8+1+8) #define area5 (8 + 8 + 1 + 8)
#define area6 (8+8+1+16) #define area6 (8 + 8 + 1 + 16)
/** /**
Collect statistics and prepare the edge pixels required by the other spatial compensation functions. Collect statistics and prepare the edge pixels required by the other spatial compensation functions.
...@@ -61,372 +61,404 @@ area positions, #3 is 1 pixel only, other are 8 pixels ...@@ -61,372 +61,404 @@ area positions, #3 is 1 pixel only, other are 8 pixels
2 - mb_y==0 - first row, interpolate area #3,#4,#5,#6; 2 - mb_y==0 - first row, interpolate area #3,#4,#5,#6;
note: 1|2 - mb_x==mb_y==0 - first block, use 0x80 value for all areas; note: 1|2 - mb_x==mb_y==0 - first block, use 0x80 value for all areas;
4 - mb_x>= (mb_width-1) last block in the row, interpolate area #5; 4 - mb_x>= (mb_width-1) last block in the row, interpolate area #5;
*/ -*/
static void x8_setup_spatial_compensation(uint8_t *src, uint8_t *dst, int linesize, static void x8_setup_spatial_compensation(uint8_t *src, uint8_t *dst,
int * range, int * psum, int edges){ int linesize, int *range, int *psum,
uint8_t * ptr; int edges)
{
uint8_t *ptr;
int sum; int sum;
int i; int i;
int min_pix,max_pix; int min_pix, max_pix;
uint8_t c; uint8_t c;
if((edges&3)==3){ if ((edges & 3) == 3) {
*psum=0x80*(8+1+8+2); *psum = 0x80 * (8 + 1 + 8 + 2);
*range=0; *range = 0;
memset(dst,0x80,16+1+16+8); memset(dst, 0x80, 16 + 1 + 16 + 8);
//this triggers flat_dc for sure. /* this triggers flat_dc for sure. flat_dc avoids all (other)
//flat_dc avoids all (other) prediction modes, but requires dc_level decoding. * prediction modes, but requires dc_level decoding. */
return; return;
} }
min_pix=256; min_pix = 256;
max_pix=-1; max_pix = -1;
sum=0; sum = 0;
if(!(edges&1)){//(mb_x!=0)//there is previous block on this row if (!(edges & 1)) { // (mb_x != 0) // there is previous block on this row
ptr=src-1;//left column, area 2 ptr = src - 1; // left column, area 2
for(i=7;i>=0;i--){ for (i = 7; i >= 0; i--) {
c=*(ptr-1);//area1, same mb as area2, no need to check c = *(ptr - 1); // area1, same mb as area2, no need to check
dst[area1+i]=c; dst[area1 + i] = c;
c=*(ptr); c = *(ptr);
sum+=c; sum += c;
min_pix=FFMIN(min_pix,c); min_pix = FFMIN(min_pix, c);
max_pix=FFMAX(max_pix,c); max_pix = FFMAX(max_pix, c);
dst[area2+i]=c; dst[area2 + i] = c;
ptr+=linesize; ptr += linesize;
} }
} }
if(!(edges&2)){ //(mb_y!=0)//there is row above if (!(edges & 2)) { // (mb_y != 0) // there is row above
ptr=src-linesize;//top line ptr = src - linesize; // top line
for(i=0;i<8;i++){ for (i = 0; i < 8; i++) {
c=*(ptr+i); c = *(ptr + i);
sum+=c; sum += c;
min_pix=FFMIN(min_pix, c); min_pix = FFMIN(min_pix, c);
max_pix=FFMAX(max_pix, c); max_pix = FFMAX(max_pix, c);
} }
if(edges&4){//last block on the row? if (edges & 4) { // last block on the row?
memset(dst+area5,c,8);//set with last pixel fr memset(dst + area5, c, 8); // set with last pixel fr
memcpy(dst+area4, ptr, 8); memcpy(dst + area4, ptr, 8);
}else{ } else {
memcpy(dst+area4, ptr, 16);//both area4 and 5 memcpy(dst + area4, ptr, 16); // both area4 and 5
} }
memcpy(dst+area6, ptr-linesize, 8);//area6 always present in the above block // area6 always present in the above block
memcpy(dst + area6, ptr - linesize, 8);
} }
//now calculate the stuff we need // now calculate the stuff we need
if(edges&3){//mb_x==0 || mb_y==0){ if (edges & 3) { // mb_x ==0 || mb_y == 0) {
int avg=(sum+4)>>3; int avg = (sum + 4) >> 3;
if(edges&1){ //(mb_x==0) {//implies mb_y!=0
memset(dst+area1,avg,8+8+1);//areas 1,2 and 3 are averaged if (edges & 1) // (mb_x == 0) { // implies mb_y !=0
}else{//implies y==0 x!=0 memset(dst + area1, avg, 8 + 8 + 1); // areas 1, 2, 3 are averaged
memset(dst+area3,avg, 1+16+8);//areas 3, 4,5,6 else // implies y == 0 x != 0
} memset(dst + area3, avg, 1 + 16 + 8); // areas 3, 4, 5, 6
sum+=avg*9;
}else{ sum += avg * 9;
uint8_t c=*(src-1-linesize);//the edge pixel, in the top line and left column } else {
dst[area3]=c; // the edge pixel, in the top line and left column
sum+=c; uint8_t c = *(src - 1 - linesize);
//edge pixel is not part of min/max dst[area3] = c;
sum += c;
// edge pixel is not part of min/max
} }
(*range) = max_pix - min_pix; (*range) = max_pix - min_pix;
sum += *(dst+area5) + *(dst+area5+1); sum += *(dst + area5) + *(dst + area5 + 1);
*psum = sum; *psum = sum;
} }
static const uint16_t zero_prediction_weights[64 * 2] = {
static const uint16_t zero_prediction_weights[64*2] = { 640, 640, 669, 480, 708, 354, 748, 257,
640, 640, 669, 480, 708, 354, 748, 257, 792, 198, 760, 143, 808, 101, 772, 72, 792, 198, 760, 143, 808, 101, 772, 72,
480, 669, 537, 537, 598, 416, 661, 316, 719, 250, 707, 185, 768, 134, 745, 97, 480, 669, 537, 537, 598, 416, 661, 316,
354, 708, 416, 598, 488, 488, 564, 388, 634, 317, 642, 241, 716, 179, 706, 132, 719, 250, 707, 185, 768, 134, 745, 97,
257, 748, 316, 661, 388, 564, 469, 469, 543, 395, 571, 311, 655, 238, 660, 180, 354, 708, 416, 598, 488, 488, 564, 388,
198, 792, 250, 719, 317, 634, 395, 543, 469, 469, 507, 380, 597, 299, 616, 231, 634, 317, 642, 241, 716, 179, 706, 132,
161, 855, 206, 788, 266, 710, 340, 623, 411, 548, 455, 455, 548, 366, 576, 288, 257, 748, 316, 661, 388, 564, 469, 469,
122, 972, 159, 914, 211, 842, 276, 758, 341, 682, 389, 584, 483, 483, 520, 390, 543, 395, 571, 311, 655, 238, 660, 180,
110, 1172, 144, 1107, 193, 1028, 254, 932, 317, 846, 366, 731, 458, 611, 499, 499 198, 792, 250, 719, 317, 634, 395, 543,
469, 469, 507, 380, 597, 299, 616, 231,
161, 855, 206, 788, 266, 710, 340, 623,
411, 548, 455, 455, 548, 366, 576, 288,
122, 972, 159, 914, 211, 842, 276, 758,
341, 682, 389, 584, 483, 483, 520, 390,
110, 1172, 144, 1107, 193, 1028, 254, 932,
317, 846, 366, 731, 458, 611, 499, 499,
}; };
static void spatial_compensation_0(uint8_t *src , uint8_t *dst, int linesize){ static void spatial_compensation_0(uint8_t *src, uint8_t *dst, int linesize)
int i,j; {
int x,y; int i, j;
unsigned int p;//power divided by 2 int x, y;
unsigned int p; // power divided by 2
int a; int a;
uint16_t left_sum[2][8] = { { 0 } }; uint16_t left_sum[2][8] = { { 0 } };
uint16_t top_sum[2][8] = { { 0 } }; uint16_t top_sum[2][8] = { { 0 } };
for(i=0;i<8;i++){ for (i = 0; i < 8; i++) {
a=src[area2+7-i]<<4; a = src[area2 + 7 - i] << 4;
for(j=0;j<8;j++){ for (j = 0; j < 8; j++) {
p=abs(i-j); p = abs(i - j);
left_sum[p&1][j]+= a>>(p>>1); left_sum[p & 1][j] += a >> (p >> 1);
} }
} }
for(i=0;i<8;i++){ for (i = 0; i < 8; i++) {
a=src[area4+i]<<4; a = src[area4 + i] << 4;
for(j=0;j<8;j++){ for (j = 0; j < 8; j++) {
p=abs(i-j); p = abs(i - j);
top_sum[p&1][j]+= a>>(p>>1); top_sum[p & 1][j] += a >> (p >> 1);
} }
} }
for(;i<10;i++){ for (; i < 10; i++) {
a=src[area4+i]<<4; a = src[area4 + i] << 4;
for(j=5;j<8;j++){ for (j = 5; j < 8; j++) {
p=abs(i-j); p = abs(i - j);
top_sum[p&1][j]+= a>>(p>>1); top_sum[p & 1][j] += a >> (p >> 1);
} }
} }
for(;i<12;i++){ for (; i < 12; i++) {
a=src[area4+i]<<4; a = src[area4 + i] << 4;
for(j=7;j<8;j++){ for (j = 7; j < 8; j++) {
p=abs(i-j); p = abs(i - j);
top_sum[p&1][j]+= a>>(p>>1); top_sum[p & 1][j] += a >> (p >> 1);
} }
} }
for(i=0;i<8;i++){ for (i = 0; i < 8; i++) {
top_sum [0][i]+=(top_sum [1][i]*181 + 128 )>>8;//181 is sqrt(2)/2 top_sum[0][i] += (top_sum[1][i] * 181 + 128) >> 8; // 181 is sqrt(2)/2
left_sum[0][i]+=(left_sum[1][i]*181 + 128 )>>8; left_sum[0][i] += (left_sum[1][i] * 181 + 128) >> 8;
} }
for(y=0;y<8;y++){ for (y = 0; y < 8; y++) {
for(x=0;x<8;x++){ for (x = 0; x < 8; x++)
dst[x] = ( dst[x] = ((uint32_t) top_sum[0][x] * zero_prediction_weights[y * 16 + x * 2 + 0] +
(uint32_t)top_sum [0][x]*zero_prediction_weights[y*16+x*2+0] + (uint32_t) left_sum[0][y] * zero_prediction_weights[y * 16 + x * 2 + 1] +
(uint32_t)left_sum[0][y]*zero_prediction_weights[y*16+x*2+1] + 0x8000) >> 16;
0x8000 dst += linesize;
)>>16;
}
dst+=linesize;
} }
} }
static void spatial_compensation_1(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_1(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=src[area4 + FFMIN(2*y+x+2, 15) ]; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = src[area4 + FFMIN(2 * y + x + 2, 15)];
dst += linesize;
} }
} }
static void spatial_compensation_2(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_2(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=src[area4 +1+y+x]; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = src[area4 + 1 + y + x];
dst += linesize;
} }
} }
static void spatial_compensation_3(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_3(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=src[area4 +((y+1)>>1)+x]; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = src[area4 + ((y + 1) >> 1) + x];
dst += linesize;
} }
} }
static void spatial_compensation_4(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_4(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=( src[area4+x] + src[area6+x] + 1 )>>1; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = (src[area4 + x] + src[area6 + x] + 1) >> 1;
dst += linesize;
} }
} }
static void spatial_compensation_5(uint8_t *src , uint8_t *dst, int linesize){
int x,y; static void spatial_compensation_5(uint8_t *src, uint8_t *dst, int linesize)
{
for(y=0;y<8;y++){ int x, y;
for(x=0;x<8;x++){
if(2*x-y<0){ for (y = 0; y < 8; y++) {
dst[x]=src[area2+9+2*x-y]; for (x = 0; x < 8; x++) {
}else{ if (2 * x - y < 0)
dst[x]=src[area4 +x-((y+1)>>1)]; dst[x] = src[area2 + 9 + 2 * x - y];
} else
dst[x] = src[area4 + x - ((y + 1) >> 1)];
} }
dst+=linesize; dst += linesize;
} }
} }
static void spatial_compensation_6(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_6(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=src[area3+x-y]; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = src[area3 + x - y];
dst += linesize;
} }
} }
static void spatial_compensation_7(uint8_t *src , uint8_t *dst, int linesize){
int x,y; static void spatial_compensation_7(uint8_t *src, uint8_t *dst, int linesize)
{
for(y=0;y<8;y++){ int x, y;
for(x=0;x<8;x++){
if(x-2*y>0){ for (y = 0; y < 8; y++) {
dst[x]=( src[area3-1+x-2*y] + src[area3+x-2*y] + 1)>>1; for (x = 0; x < 8; x++) {
}else{ if (x - 2 * y > 0)
dst[x]=src[area2+8-y +(x>>1)]; dst[x] = (src[area3 - 1 + x - 2 * y] + src[area3 + x - 2 * y] + 1) >> 1;
} else
dst[x] = src[area2 + 8 - y + (x >> 1)];
} }
dst+=linesize; dst += linesize;
} }
} }
static void spatial_compensation_8(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_8(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=( src[area1+7-y] + src[area2+7-y] + 1 )>>1; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = (src[area1 + 7 - y] + src[area2 + 7 - y] + 1) >> 1;
dst += linesize;
} }
} }
static void spatial_compensation_9(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_9(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=src[area2+6-FFMIN(x+y,6)]; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = src[area2 + 6 - FFMIN(x + y, 6)];
dst += linesize;
} }
} }
static void spatial_compensation_10(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_10(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=(src[area2+7-y]*(8-x)+src[area4+x]*x+4)>>3; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = (src[area2 + 7 - y] * (8 - x) + src[area4 + x] * x + 4) >> 3;
dst += linesize;
} }
} }
static void spatial_compensation_11(uint8_t *src , uint8_t *dst, int linesize){
int x,y;
for(y=0;y<8;y++){ static void spatial_compensation_11(uint8_t *src, uint8_t *dst, int linesize)
for(x=0;x<8;x++){ {
dst[x]=(src[area2+7-y]*y+src[area4+x]*(8-y)+4)>>3; int x, y;
}
dst+=linesize; for (y = 0; y < 8; y++) {
for (x = 0; x < 8; x++)
dst[x] = (src[area2 + 7 - y] * y + src[area4 + x] * (8 - y) + 4) >> 3;
dst += linesize;
} }
} }
static void x8_loop_filter(uint8_t * ptr, const int a_stride, const int b_stride, int quant){ static void x8_loop_filter(uint8_t *ptr, const int a_stride, const int b_stride, int quant)
int i,t; {
int p0,p1,p2,p3,p4,p5,p6,p7,p8,p9; int i, t;
int ql=(quant+10)>>3; int p0, p1, p2, p3, p4, p5, p6, p7, p8, p9;
int ql = (quant + 10) >> 3;
for(i=0; i<8; i++,ptr+=b_stride){
p0=ptr[-5*a_stride]; for (i = 0; i < 8; i++, ptr += b_stride) {
p1=ptr[-4*a_stride]; p0 = ptr[-5 * a_stride];
p2=ptr[-3*a_stride]; p1 = ptr[-4 * a_stride];
p3=ptr[-2*a_stride]; p2 = ptr[-3 * a_stride];
p4=ptr[-1*a_stride]; p3 = ptr[-2 * a_stride];
p5=ptr[ 0 ]; p4 = ptr[-1 * a_stride];
p6=ptr[ 1*a_stride]; p5 = ptr[0];
p7=ptr[ 2*a_stride]; p6 = ptr[1 * a_stride];
p8=ptr[ 3*a_stride]; p7 = ptr[2 * a_stride];
p9=ptr[ 4*a_stride]; p8 = ptr[3 * a_stride];
p9 = ptr[4 * a_stride];
t=
(FFABS(p1-p2) <= ql) + t = (FFABS(p1 - p2) <= ql) +
(FFABS(p2-p3) <= ql) + (FFABS(p2 - p3) <= ql) +
(FFABS(p3-p4) <= ql) + (FFABS(p3 - p4) <= ql) +
(FFABS(p4-p5) <= ql); (FFABS(p4 - p5) <= ql);
if(t>0){//You need at least 1 to be able to reach a total score of 6.
t+= // You need at least 1 to be able to reach a total score of 6.
(FFABS(p5-p6) <= ql) + if (t > 0) {
(FFABS(p6-p7) <= ql) + t += (FFABS(p5 - p6) <= ql) +
(FFABS(p7-p8) <= ql) + (FFABS(p6 - p7) <= ql) +
(FFABS(p8-p9) <= ql) + (FFABS(p7 - p8) <= ql) +
(FFABS(p0-p1) <= ql); (FFABS(p8 - p9) <= ql) +
if(t>=6){ (FFABS(p0 - p1) <= ql);
int min,max; if (t >= 6) {
int min, max;
min=max=p1;
min=FFMIN(min,p3); max=FFMAX(max,p3); min = max = p1;
min=FFMIN(min,p5); max=FFMAX(max,p5); min = FFMIN(min, p3);
min=FFMIN(min,p8); max=FFMAX(max,p8); max = FFMAX(max, p3);
if(max-min<2*quant){//early stop min = FFMIN(min, p5);
min=FFMIN(min,p2); max=FFMAX(max,p2); max = FFMAX(max, p5);
min=FFMIN(min,p4); max=FFMAX(max,p4); min = FFMIN(min, p8);
min=FFMIN(min,p6); max=FFMAX(max,p6); max = FFMAX(max, p8);
min=FFMIN(min,p7); max=FFMAX(max,p7); if (max - min < 2 * quant) { // early stop
if(max-min<2*quant){ min = FFMIN(min, p2);
ptr[-2*a_stride]=(4*p2 + 3*p3 + 1*p7 + 4)>>3; max = FFMAX(max, p2);
ptr[-1*a_stride]=(3*p2 + 3*p4 + 2*p7 + 4)>>3; min = FFMIN(min, p4);
ptr[ 0 ]=(2*p2 + 3*p5 + 3*p7 + 4)>>3; max = FFMAX(max, p4);
ptr[ 1*a_stride]=(1*p2 + 3*p6 + 4*p7 + 4)>>3; min = FFMIN(min, p6);
max = FFMAX(max, p6);
min = FFMIN(min, p7);
max = FFMAX(max, p7);
if (max - min < 2 * quant) {
ptr[-2 * a_stride] = (4 * p2 + 3 * p3 + 1 * p7 + 4) >> 3;
ptr[-1 * a_stride] = (3 * p2 + 3 * p4 + 2 * p7 + 4) >> 3;
ptr[0] = (2 * p2 + 3 * p5 + 3 * p7 + 4) >> 3;
ptr[1 * a_stride] = (1 * p2 + 3 * p6 + 4 * p7 + 4) >> 3;
continue; continue;
}; }
} }
} }
} }
{ {
int x,x0,x1,x2; int x, x0, x1, x2;
int m; int m;
x0 = (2*p3 - 5*p4 + 5*p5 - 2*p6 + 4)>>3; x0 = (2 * p3 - 5 * p4 + 5 * p5 - 2 * p6 + 4) >> 3;
if(FFABS(x0) < quant){ if (FFABS(x0) < quant) {
x1=(2*p1 - 5*p2 + 5*p3 - 2*p4 + 4)>>3; x1 = (2 * p1 - 5 * p2 + 5 * p3 - 2 * p4 + 4) >> 3;
x2=(2*p5 - 5*p6 + 5*p7 - 2*p8 + 4)>>3; x2 = (2 * p5 - 5 * p6 + 5 * p7 - 2 * p8 + 4) >> 3;
x=FFABS(x0) - FFMIN( FFABS(x1), FFABS(x2) ); x = FFABS(x0) - FFMIN(FFABS(x1), FFABS(x2));
m=p4-p5; m = p4 - p5;
if( x > 0 && (m^x0) <0){ if (x > 0 && (m ^ x0) < 0) {
int32_t sign; int32_t sign;
sign=m>>31; sign = m >> 31;
m=(m^sign)-sign;//abs(m) m = (m ^ sign) - sign; // abs(m)
m>>=1; m >>= 1;
x=(5*x)>>3; x = (5 * x) >> 3;
if(x>m) x=m; if (x > m)
x = m;
x=(x^sign)-sign; x = (x ^ sign) - sign;
ptr[-1*a_stride] -= x; ptr[-1 * a_stride] -= x;
ptr[ 0] += x; ptr[0] += x;
} }
} }
} }
} }
} }
static void x8_h_loop_filter(uint8_t *src, int stride, int qscale){ static void x8_h_loop_filter(uint8_t *src, int stride, int qscale)
{
x8_loop_filter(src, stride, 1, qscale); x8_loop_filter(src, stride, 1, qscale);
} }
static void x8_v_loop_filter(uint8_t *src, int stride, int qscale){ static void x8_v_loop_filter(uint8_t *src, int stride, int qscale)
{
x8_loop_filter(src, 1, stride, qscale); x8_loop_filter(src, 1, stride, qscale);
} }
av_cold void ff_intrax8dsp_init(IntraX8DSPContext *dsp) av_cold void ff_intrax8dsp_init(IntraX8DSPContext *dsp)
{ {
dsp->h_loop_filter=x8_h_loop_filter; dsp->h_loop_filter = x8_h_loop_filter;
dsp->v_loop_filter=x8_v_loop_filter; dsp->v_loop_filter = x8_v_loop_filter;
dsp->setup_spatial_compensation=x8_setup_spatial_compensation; dsp->setup_spatial_compensation = x8_setup_spatial_compensation;
dsp->spatial_compensation[0]=spatial_compensation_0; dsp->spatial_compensation[0] = spatial_compensation_0;
dsp->spatial_compensation[1]=spatial_compensation_1; dsp->spatial_compensation[1] = spatial_compensation_1;
dsp->spatial_compensation[2]=spatial_compensation_2; dsp->spatial_compensation[2] = spatial_compensation_2;
dsp->spatial_compensation[3]=spatial_compensation_3; dsp->spatial_compensation[3] = spatial_compensation_3;
dsp->spatial_compensation[4]=spatial_compensation_4; dsp->spatial_compensation[4] = spatial_compensation_4;
dsp->spatial_compensation[5]=spatial_compensation_5; dsp->spatial_compensation[5] = spatial_compensation_5;
dsp->spatial_compensation[6]=spatial_compensation_6; dsp->spatial_compensation[6] = spatial_compensation_6;
dsp->spatial_compensation[7]=spatial_compensation_7; dsp->spatial_compensation[7] = spatial_compensation_7;
dsp->spatial_compensation[8]=spatial_compensation_8; dsp->spatial_compensation[8] = spatial_compensation_8;
dsp->spatial_compensation[9]=spatial_compensation_9; dsp->spatial_compensation[9] = spatial_compensation_9;
dsp->spatial_compensation[10]=spatial_compensation_10; dsp->spatial_compensation[10] = spatial_compensation_10;
dsp->spatial_compensation[11]=spatial_compensation_11; dsp->spatial_compensation[11] = spatial_compensation_11;
} }
...@@ -25,9 +25,9 @@ typedef struct IntraX8DSPContext { ...@@ -25,9 +25,9 @@ typedef struct IntraX8DSPContext {
void (*v_loop_filter)(uint8_t *src, int stride, int qscale); void (*v_loop_filter)(uint8_t *src, int stride, int qscale);
void (*h_loop_filter)(uint8_t *src, int stride, int qscale); void (*h_loop_filter)(uint8_t *src, int stride, int qscale);
void (*spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize); void (*spatial_compensation[12])(uint8_t *src, uint8_t *dst, int linesize);
void (*setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize, void (*setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
int *range, int *sum, int edges); int *range, int *sum, int edges);
} IntraX8DSPContext; } IntraX8DSPContext;
void ff_intrax8dsp_init(IntraX8DSPContext *dsp); void ff_intrax8dsp_init(IntraX8DSPContext *dsp);
......
This source diff could not be displayed because it is too large. You can view the blob instead.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment