* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
*****************************************************************************/
-#include <stdio.h>
-#include <string.h>
-
#include "common.h"
-#include "clip1.h"
-#ifdef HAVE_MMXEXT
-#include "i386/mc.h"
+#ifdef HAVE_MMX
+#include "x86/mc.h"
#endif
#ifdef ARCH_PPC
#include "ppc/mc.h"
#endif
-static inline int x264_tapfilter( uint8_t *pix, int i_pix_next )
-{
- return pix[-2*i_pix_next] - 5*pix[-1*i_pix_next] + 20*(pix[0] + pix[1*i_pix_next]) - 5*pix[ 2*i_pix_next] + pix[ 3*i_pix_next];
-}
-static inline int x264_tapfilter1( uint8_t *pix )
-{
- return pix[-2] - 5*pix[-1] + 20*(pix[0] + pix[1]) - 5*pix[ 2] + pix[ 3];
-}
-
static inline void pixel_avg( uint8_t *dst, int i_dst_stride,
uint8_t *src1, int i_src1_stride,
uint8_t *src2, int i_src2_stride,
#undef op_scale2
#undef PIXEL_AVG_WEIGHT_C
-typedef void (*pf_mc_t)(uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_width, int i_height );
-
static void mc_copy( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_width, int i_height )
{
int y;
dst += i_dst_stride;
}
}
-static inline void mc_hh( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_width, int i_height )
-{
- int x, y;
- for( y = 0; y < i_height; y++ )
- {
- for( x = 0; x < i_width; x++ )
- {
- dst[x] = x264_mc_clip1( ( x264_tapfilter1( &src[x] ) + 16 ) >> 5 );
- }
- src += i_src_stride;
- dst += i_dst_stride;
- }
-}
-static inline void mc_hv( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_width, int i_height )
+#define TAPFILTER(pix, d) ((pix)[x-2*d] + (pix)[x+3*d] - 5*((pix)[x-d] + (pix)[x+2*d]) + 20*((pix)[x] + (pix)[x+d]))
+static void hpel_filter( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src,
+ int stride, int width, int height )
{
+ int16_t *buf = x264_malloc((width+5)*sizeof(int16_t));
int x, y;
-
- for( y = 0; y < i_height; y++ )
+ for( y=0; y<height; y++ )
{
- for( x = 0; x < i_width; x++ )
+ for( x=-2; x<width+3; x++ )
{
- dst[x] = x264_mc_clip1( ( x264_tapfilter( &src[x], i_src_stride ) + 16 ) >> 5 );
- }
- src += i_src_stride;
- dst += i_dst_stride;
- }
-}
-static inline void mc_hc( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_width, int i_height )
-{
- uint8_t *out;
- uint8_t *pix;
- int x, y;
-
- for( x = 0; x < i_width; x++ )
- {
- int tap[6];
-
- pix = &src[x];
- out = &dst[x];
-
- tap[0] = x264_tapfilter1( &pix[-2*i_src_stride] );
- tap[1] = x264_tapfilter1( &pix[-1*i_src_stride] );
- tap[2] = x264_tapfilter1( &pix[ 0*i_src_stride] );
- tap[3] = x264_tapfilter1( &pix[ 1*i_src_stride] );
- tap[4] = x264_tapfilter1( &pix[ 2*i_src_stride] );
-
- for( y = 0; y < i_height; y++ )
- {
- tap[5] = x264_tapfilter1( &pix[ 3*i_src_stride] );
-
- *out = x264_mc_clip1( ( tap[0] - 5*tap[1] + 20 * tap[2] + 20 * tap[3] -5*tap[4] + tap[5] + 512 ) >> 10 );
-
- /* Next line */
- pix += i_src_stride;
- out += i_dst_stride;
- tap[0] = tap[1];
- tap[1] = tap[2];
- tap[2] = tap[3];
- tap[3] = tap[4];
- tap[4] = tap[5];
+ int v = TAPFILTER(src,stride);
+ dstv[x] = x264_clip_uint8((v + 16) >> 5);
+ buf[x+2] = v;
}
+ for( x=0; x<width; x++ )
+ dstc[x] = x264_clip_uint8((TAPFILTER(buf+2,1) + 512) >> 10);
+ for( x=0; x<width; x++ )
+ dsth[x] = x264_clip_uint8((TAPFILTER(src,1) + 16) >> 5);
+ dsth += stride;
+ dstv += stride;
+ dstc += stride;
+ src += stride;
}
+ x264_free(buf);
}
static const int hpel_ref0[16] = {0,1,1,1,0,1,1,1,2,3,3,3,0,1,1,1};
static const int hpel_ref1[16] = {0,0,0,0,2,2,3,2,2,2,3,2,2,2,3,2};
-static void mc_luma( uint8_t *src[4], int i_src_stride,
- uint8_t *dst, int i_dst_stride,
- int mvx,int mvy,
+static void mc_luma( uint8_t *dst, int i_dst_stride,
+ uint8_t *src[4], int i_src_stride,
+ int mvx, int mvy,
int i_width, int i_height )
{
int qpel_idx = ((mvy&3)<<2) + (mvx&3);
if( qpel_idx & 5 ) /* qpel interpolation needed */
{
uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
-
pixel_avg( dst, i_dst_stride, src1, i_src_stride,
src2, i_src_stride, i_width, i_height );
}
}
}
-static uint8_t *get_ref( uint8_t *src[4], int i_src_stride,
- uint8_t *dst, int * i_dst_stride,
- int mvx,int mvy,
+static uint8_t *get_ref( uint8_t *dst, int *i_dst_stride,
+ uint8_t *src[4], int i_src_stride,
+ int mvx, int mvy,
int i_width, int i_height )
{
int qpel_idx = ((mvy&3)<<2) + (mvx&3);
if( qpel_idx & 5 ) /* qpel interpolation needed */
{
uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
-
pixel_avg( dst, *i_dst_stride, src1, i_src_stride,
src2, i_src_stride, i_width, i_height );
-
return dst;
}
else
}
/* full chroma mc (ie until 1/8 pixel)*/
-static void motion_compensation_chroma( uint8_t *src, int i_src_stride,
- uint8_t *dst, int i_dst_stride,
- int mvx, int mvy,
- int i_width, int i_height )
+static void mc_chroma( uint8_t *dst, int i_dst_stride,
+ uint8_t *src, int i_src_stride,
+ int mvx, int mvy,
+ int i_width, int i_height )
{
uint8_t *srcp;
int x, y;
}
}
+void prefetch_fenc_null( uint8_t *pix_y, int stride_y,
+ uint8_t *pix_uv, int stride_uv, int mb_x )
+{}
+
+void prefetch_ref_null( uint8_t *pix, int stride, int parity )
+{}
+
void x264_mc_init( int cpu, x264_mc_functions_t *pf )
{
pf->mc_luma = mc_luma;
pf->get_ref = get_ref;
- pf->mc_chroma = motion_compensation_chroma;
+ pf->mc_chroma = mc_chroma;
pf->avg[PIXEL_16x16]= pixel_avg_16x16;
pf->avg[PIXEL_16x8] = pixel_avg_16x8;
pf->copy[PIXEL_4x4] = mc_copy_w4;
pf->plane_copy = plane_copy;
+ pf->hpel_filter = hpel_filter;
-#ifdef HAVE_MMXEXT
- if( cpu&X264_CPU_MMXEXT ) {
- x264_mc_mmxext_init( pf );
- pf->mc_chroma = x264_mc_chroma_mmxext;
- }
-#endif
-#ifdef HAVE_SSE2
- if( cpu&X264_CPU_SSE2 )
- x264_mc_sse2_init( pf );
+ pf->prefetch_fenc = prefetch_fenc_null;
+ pf->prefetch_ref = prefetch_ref_null;
+ pf->memcpy_aligned = memcpy;
+
+#ifdef HAVE_MMX
+ x264_mc_init_mmx( cpu, pf );
#endif
#ifdef ARCH_PPC
if( cpu&X264_CPU_ALTIVEC )
#endif
}
-extern void x264_horizontal_filter_mmxext( uint8_t *dst, int i_dst_stride,
- uint8_t *src, int i_src_stride,
- int i_width, int i_height );
-extern void x264_center_filter_mmxext( uint8_t *dst1, int i_dst1_stride,
- uint8_t *dst2, int i_dst2_stride,
- uint8_t *src, int i_src_stride,
- int i_width, int i_height );
-
-void x264_frame_filter( int cpu, x264_frame_t *frame, int b_interlaced )
+void x264_frame_filter( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
{
- const int x_inc = 16, y_inc = 16;
+ const int b_interlaced = h->sh.b_mbaff;
const int stride = frame->i_stride[0] << b_interlaced;
- const int height = frame->i_lines[0] >> b_interlaced;
+ const int width = frame->i_width[0];
+ int start = (mb_y*16 >> b_interlaced) - 8; // buffer = 4 for deblock + 3 for 6tap, rounded to 8
+ int height = ((b_end ? frame->i_lines[0] : mb_y*16) >> b_interlaced) + 8;
+ int offs = start*stride - 8; // buffer = 3 for 6tap, aligned to 8 for simd
int x, y;
- pf_mc_t int_h = mc_hh;
- pf_mc_t int_v = mc_hv;
- pf_mc_t int_hv = mc_hc;
+ if( mb_y & b_interlaced )
+ return;
-#ifdef HAVE_MMXEXT
- if ( cpu & X264_CPU_MMXEXT )
+ for( y=0; y<=b_interlaced; y++, offs+=frame->i_stride[0] )
{
- x264_horizontal_filter_mmxext(frame->filtered[1] - 8 * stride - 8, stride,
- frame->plane[0] - 8 * stride - 8, stride,
- stride - 48, height + 16);
- x264_center_filter_mmxext(frame->filtered[2] - 8 * stride - 8, stride,
- frame->filtered[3] - 8 * stride - 8, stride,
- frame->plane[0] - 8 * stride - 8, stride,
- stride - 48, height + 16);
- }
- else
-#endif
- {
- for( y = -8; y < height + 8; y += y_inc )
- {
- uint8_t *p_in = frame->plane[0] + y * stride - 8;
- uint8_t *p_h = frame->filtered[1] + y * stride - 8;
- uint8_t *p_v = frame->filtered[2] + y * stride - 8;
- uint8_t *p_hv = frame->filtered[3] + y * stride - 8;
- for( x = -8; x < stride - 64 + 8; x += x_inc )
- {
- int_h( p_in, stride, p_h, stride, x_inc, y_inc );
- int_v( p_in, stride, p_v, stride, x_inc, y_inc );
- int_hv( p_in, stride, p_hv, stride, x_inc, y_inc );
-
- p_h += x_inc;
- p_v += x_inc;
- p_hv += x_inc;
- p_in += x_inc;
- }
- }
+ h->mc.hpel_filter(
+ frame->filtered[1] + offs,
+ frame->filtered[2] + offs,
+ frame->filtered[3] + offs,
+ frame->plane[0] + offs,
+ stride, width + 16, height - start );
}
/* generate integral image:
* frame->integral contains 2 planes. in the upper plane, each element is
* the sum of an 8x8 pixel region with top-left corner on that point.
- * in the lower plane, 4x4 sums (needed only with --analyse p4x4). */
+ * in the lower plane, 4x4 sums (needed only with --partitions p4x4). */
if( frame->integral )
{
- memset( frame->integral - 32 * stride - 32, 0, stride * sizeof(uint16_t) );
- for( y = -32; y < frame->i_lines[0] + 31; y++ )
+ if( start < 0 )
{
- uint8_t *ref = frame->plane[0] + y * stride - 32;
- uint16_t *line = frame->integral + (y+1) * stride - 31;
- uint16_t v = line[0] = 0;
- for( x = 0; x < stride-1; x++ )
- line[x] = v += ref[x] + line[x-stride] - line[x-stride-1];
+ memset( frame->integral - PADV * stride - PADH, 0, stride * sizeof(uint16_t) );
+ start = -PADV;
}
- for( y = -31; y < frame->i_lines[0] + 24; y++ )
+ if( b_end )
+ height += PADV-8;
+ for( y = start; y < height; y++ )
{
- uint16_t *line = frame->integral + y * stride - 31;
- uint16_t *sum4 = line + frame->i_stride[0] * (frame->i_lines[0] + 64);
- for( x = -31; x < stride - 40; x++, line++, sum4++ )
+ uint8_t *ref = frame->plane[0] + y * stride - PADH;
+ uint16_t *line = frame->integral + (y+1) * stride - PADH + 1;
+ uint16_t v = line[0] = 0;
+ for( x = 1; x < stride-1; x++ )
+ line[x] = v += ref[x] + line[x-stride] - line[x-stride-1];
+ line -= 8*stride;
+ if( y >= 9-PADV )
{
- sum4[0] = line[4+4*stride] - line[4] - line[4*stride] + line[0];
- line[0] += line[8+8*stride] - line[8] - line[8*stride];
+ uint16_t *sum4 = line + stride * (frame->i_lines[0] + PADV*2);
+ for( x = 1; x < stride-8; x++, line++, sum4++ )
+ {
+ sum4[0] = line[4+4*stride] - line[4] - line[4*stride] + line[0];
+ line[0] += line[8+8*stride] - line[8] - line[8*stride];
+ }
}
}
}
}
-void x264_frame_init_lowres( int cpu, x264_frame_t *frame )
+void x264_frame_init_lowres( x264_t *h, x264_frame_t *frame )
{
// FIXME: tapfilter?
const int i_stride = frame->i_stride[0];
const int i_stride2 = frame->i_stride_lowres;
- const int i_width2 = i_stride2 - 64;
+ const int i_width2 = frame->i_width_lowres;
int x, y, i;
for( y = 0; y < frame->i_lines_lowres - 1; y++ )
{