#ifdef ARCH_PPC
# include "ppc/quant.h"
#endif
+#ifdef ARCH_ARM
+# include "arm/quant.h"
+#endif
#define QUANT_ONE( coef, mf, f ) \
{ \
(coef) = (f + (coef)) * (mf) >> 16; \
else \
(coef) = - ((f - (coef)) * (mf) >> 16); \
+ nz |= (coef); \
}
-static void quant_8x8( int16_t dct[8][8], uint16_t mf[64], uint16_t bias[64] )
+static int quant_8x8( int16_t dct[64], uint16_t mf[64], uint16_t bias[64] )
{
- int i;
- for( i = 0; i < 64; i++ )
- QUANT_ONE( dct[0][i], mf[i], bias[i] );
+ int nz = 0;
+ for( int i = 0; i < 64; i++ )
+ QUANT_ONE( dct[i], mf[i], bias[i] );
+ return !!nz;
}
-static void quant_4x4( int16_t dct[4][4], uint16_t mf[16], uint16_t bias[16] )
+static int quant_4x4( int16_t dct[16], uint16_t mf[16], uint16_t bias[16] )
{
- int i;
- for( i = 0; i < 16; i++ )
- QUANT_ONE( dct[0][i], mf[i], bias[i] );
+ int nz = 0;
+ for( int i = 0; i < 16; i++ )
+ QUANT_ONE( dct[i], mf[i], bias[i] );
+ return !!nz;
}
-static void quant_4x4_dc( int16_t dct[4][4], int mf, int bias )
+static int quant_4x4_dc( int16_t dct[16], int mf, int bias )
{
- int i;
- for( i = 0; i < 16; i++ )
- QUANT_ONE( dct[0][i], mf, bias );
+ int nz = 0;
+ for( int i = 0; i < 16; i++ )
+ QUANT_ONE( dct[i], mf, bias );
+ return !!nz;
}
-static void quant_2x2_dc( int16_t dct[2][2], int mf, int bias )
+static int quant_2x2_dc( int16_t dct[4], int mf, int bias )
{
- QUANT_ONE( dct[0][0], mf, bias );
- QUANT_ONE( dct[0][1], mf, bias );
- QUANT_ONE( dct[0][2], mf, bias );
- QUANT_ONE( dct[0][3], mf, bias );
+ int nz = 0;
+ QUANT_ONE( dct[0], mf, bias );
+ QUANT_ONE( dct[1], mf, bias );
+ QUANT_ONE( dct[2], mf, bias );
+ QUANT_ONE( dct[3], mf, bias );
+ return !!nz;
}
#define DEQUANT_SHL( x ) \
- dct[y][x] = ( dct[y][x] * dequant_mf[i_mf][y][x] ) << i_qbits
+ dct[x] = ( dct[x] * dequant_mf[i_mf][x] ) << i_qbits
#define DEQUANT_SHR( x ) \
- dct[y][x] = ( dct[y][x] * dequant_mf[i_mf][y][x] + f ) >> (-i_qbits)
+ dct[x] = ( dct[x] * dequant_mf[i_mf][x] + f ) >> (-i_qbits)
-static void dequant_4x4( int16_t dct[4][4], int dequant_mf[6][4][4], int i_qp )
+static void dequant_4x4( int16_t dct[16], int dequant_mf[6][16], int i_qp )
{
const int i_mf = i_qp%6;
const int i_qbits = i_qp/6 - 4;
- int y;
if( i_qbits >= 0 )
{
- for( y = 0; y < 4; y++ )
- {
- DEQUANT_SHL( 0 );
- DEQUANT_SHL( 1 );
- DEQUANT_SHL( 2 );
- DEQUANT_SHL( 3 );
- }
+ for( int i = 0; i < 16; i++ )
+ DEQUANT_SHL( i );
}
else
{
const int f = 1 << (-i_qbits-1);
- for( y = 0; y < 4; y++ )
- {
- DEQUANT_SHR( 0 );
- DEQUANT_SHR( 1 );
- DEQUANT_SHR( 2 );
- DEQUANT_SHR( 3 );
- }
+ for( int i = 0; i < 16; i++ )
+ DEQUANT_SHR( i );
}
}
-static void dequant_8x8( int16_t dct[8][8], int dequant_mf[6][8][8], int i_qp )
+static void dequant_8x8( int16_t dct[64], int dequant_mf[6][64], int i_qp )
{
const int i_mf = i_qp%6;
const int i_qbits = i_qp/6 - 6;
- int y;
if( i_qbits >= 0 )
{
- for( y = 0; y < 8; y++ )
- {
- DEQUANT_SHL( 0 );
- DEQUANT_SHL( 1 );
- DEQUANT_SHL( 2 );
- DEQUANT_SHL( 3 );
- DEQUANT_SHL( 4 );
- DEQUANT_SHL( 5 );
- DEQUANT_SHL( 6 );
- DEQUANT_SHL( 7 );
- }
+ for( int i = 0; i < 64; i++ )
+ DEQUANT_SHL( i );
}
else
{
const int f = 1 << (-i_qbits-1);
- for( y = 0; y < 8; y++ )
- {
- DEQUANT_SHR( 0 );
- DEQUANT_SHR( 1 );
- DEQUANT_SHR( 2 );
- DEQUANT_SHR( 3 );
- DEQUANT_SHR( 4 );
- DEQUANT_SHR( 5 );
- DEQUANT_SHR( 6 );
- DEQUANT_SHR( 7 );
- }
+ for( int i = 0; i < 64; i++ )
+ DEQUANT_SHR( i );
}
}
-static void dequant_4x4_dc( int16_t dct[4][4], int dequant_mf[6][4][4], int i_qp )
+static void dequant_4x4_dc( int16_t dct[16], int dequant_mf[6][16], int i_qp )
{
const int i_qbits = i_qp/6 - 6;
- int y;
if( i_qbits >= 0 )
{
- const int i_dmf = dequant_mf[i_qp%6][0][0] << i_qbits;
-
- for( y = 0; y < 4; y++ )
- {
- dct[y][0] *= i_dmf;
- dct[y][1] *= i_dmf;
- dct[y][2] *= i_dmf;
- dct[y][3] *= i_dmf;
- }
+ const int i_dmf = dequant_mf[i_qp%6][0] << i_qbits;
+ for( int i = 0; i < 16; i++ )
+ dct[i] *= i_dmf;
}
else
{
- const int i_dmf = dequant_mf[i_qp%6][0][0];
+ const int i_dmf = dequant_mf[i_qp%6][0];
const int f = 1 << (-i_qbits-1);
-
- for( y = 0; y < 4; y++ )
- {
- dct[y][0] = ( dct[y][0] * i_dmf + f ) >> (-i_qbits);
- dct[y][1] = ( dct[y][1] * i_dmf + f ) >> (-i_qbits);
- dct[y][2] = ( dct[y][2] * i_dmf + f ) >> (-i_qbits);
- dct[y][3] = ( dct[y][3] * i_dmf + f ) >> (-i_qbits);
- }
+ for( int i = 0; i < 16; i++ )
+ dct[i] = ( dct[i] * i_dmf + f ) >> (-i_qbits);
}
}
static void x264_denoise_dct( int16_t *dct, uint32_t *sum, uint16_t *offset, int size )
{
- int i;
- for( i=1; i<size; i++ )
+ for( int i = 1; i < size; i++ )
{
int level = dct[i];
int sign = level>>15;
* chroma: for the complete mb: if score < 7 -> null
*/
-const uint8_t x264_decimate_table4[16] = {
- 3,2,2,1,1,1,0,0,0,0,0,0,0,0,0,0 };
-const uint8_t x264_decimate_table8[64] = {
+const uint8_t x264_decimate_table4[16] =
+{
+ 3,2,2,1,1,1,0,0,0,0,0,0,0,0,0,0
+};
+const uint8_t x264_decimate_table8[64] =
+{
3,3,3,3,2,2,2,2,2,2,2,2,1,1,1,1,
1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 };
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+};
static int ALWAYS_INLINE x264_decimate_score_internal( int16_t *dct, int i_max )
{
int idx = i_max - 1;
/* Yes, dct[idx-1] is guaranteed to be 32-bit aligned. idx>=0 instead of 1 works correctly for the same reason */
- while( idx >= 0 && *(uint32_t*)&dct[idx-1] == 0 )
+ while( idx >= 0 && M32( &dct[idx-1] ) == 0 )
idx -= 2;
if( idx >= 0 && dct[idx] == 0 )
idx--;
{
int i_last;
for( i_last = i_count-1; i_last >= 3; i_last -= 4 )
- if( *(uint64_t*)(l+i_last-3) )
+ if( M64( l+i_last-3 ) )
break;
while( i_last >= 0 && l[i_last] == 0 )
i_last--;
pf->decimate_score16 = x264_decimate_score16_ssse3;
pf->decimate_score64 = x264_decimate_score64_ssse3;
}
+
+ if( cpu&X264_CPU_SSE4 )
+ {
+ pf->quant_4x4_dc = x264_quant_4x4_dc_sse4;
+ pf->quant_4x4 = x264_quant_4x4_sse4;
+ pf->quant_8x8 = x264_quant_8x8_sse4;
+ }
#endif // HAVE_MMX
-#ifdef ARCH_PPC
+#ifdef HAVE_ALTIVEC
if( cpu&X264_CPU_ALTIVEC ) {
pf->quant_2x2_dc = x264_quant_2x2_dc_altivec;
pf->quant_4x4_dc = x264_quant_4x4_dc_altivec;
pf->dequant_8x8 = x264_dequant_8x8_altivec;
}
#endif
+
+#ifdef HAVE_ARMV6
+ if( cpu&X264_CPU_ARMV6 )
+ pf->coeff_last[DCT_CHROMA_DC] = x264_coeff_last4_arm;
+
+ if( cpu&X264_CPU_NEON )
+ {
+ pf->quant_2x2_dc = x264_quant_2x2_dc_neon;
+ pf->quant_4x4 = x264_quant_4x4_neon;
+ pf->quant_4x4_dc = x264_quant_4x4_dc_neon;
+ pf->quant_8x8 = x264_quant_8x8_neon;
+ pf->dequant_4x4 = x264_dequant_4x4_neon;
+ pf->dequant_4x4_dc = x264_dequant_4x4_dc_neon;
+ pf->dequant_8x8 = x264_dequant_8x8_neon;
+ pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_neon;
+ pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_neon;
+ pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_neon;
+ }
+#endif
pf->coeff_last[ DCT_LUMA_DC] = pf->coeff_last[DCT_LUMA_4x4];
pf->coeff_last[DCT_CHROMA_AC] = pf->coeff_last[ DCT_LUMA_AC];
pf->coeff_level_run[ DCT_LUMA_DC] = pf->coeff_level_run[DCT_LUMA_4x4];