/***************************************************************************** * I422_cyuv: planar YUV 4:2:2 to upside-down packed UYVY 4:2:2 *****************************************************************************/ VLC_TARGET static void I422_cyuv( filter_t *p_filter, picture_t *p_source, picture_t *p_dest ) { uint8_t *p_line = p_dest->p->p_pixels + p_dest->p->i_visible_lines * p_dest->p->i_pitch; uint8_t *p_y = p_source->Y_PIXELS; uint8_t *p_u = p_source->U_PIXELS; uint8_t *p_v = p_source->V_PIXELS; int i_x, i_y; const int i_source_margin = p_source->p[0].i_pitch - p_source->p[0].i_visible_pitch; const int i_source_margin_c = p_source->p[1].i_pitch - p_source->p[1].i_visible_pitch; const int i_dest_margin = p_dest->p->i_pitch - p_dest->p->i_visible_pitch; #if defined (MODULE_NAME_IS_i422_yuy2_sse2) if( 0 == (15 & (p_source->p[Y_PLANE].i_pitch|p_dest->p->i_pitch| ((intptr_t)p_line|(intptr_t)p_y))) ) { /* use faster SSE2 aligned fetch and store */ for( i_y = p_filter->fmt_in.video.i_height ; i_y-- ; ) { p_line -= 2 * p_dest->p->i_pitch; for( i_x = p_filter->fmt_in.video.i_width / 16 ; i_x-- ; ) { SSE2_CALL( SSE2_YUV422_UYVY_ALIGNED ); } for( i_x = ( p_filter->fmt_in.video.i_width % 16 ) / 2; i_x-- ; ) { C_YUV422_UYVY( p_line, p_y, p_u, p_v ); } p_y += i_source_margin; p_u += i_source_margin_c; p_v += i_source_margin_c; p_line += i_dest_margin; } } else { /* use slower SSE2 unaligned fetch and store */ for( i_y = p_filter->fmt_in.video.i_height ; i_y-- ; ) { p_line -= 2 * p_dest->p->i_pitch; for( i_x = p_filter->fmt_in.video.i_width / 16 ; i_x-- ; ) { SSE2_CALL( SSE2_YUV422_UYVY_UNALIGNED ); } for( i_x = ( p_filter->fmt_in.video.i_width % 16 ) / 2; i_x-- ; ) { C_YUV422_UYVY( p_line, p_y, p_u, p_v ); } p_y += i_source_margin; p_u += i_source_margin_c; p_v += i_source_margin_c; p_line += i_dest_margin; } } SSE2_END; #else for( i_y = p_filter->fmt_in.video.i_height ; i_y-- ; ) { for( i_x = p_filter->fmt_in.video.i_width / 8 ; i_x-- ; ) { p_line -= 2 * p_dest->p->i_pitch; #if defined (MODULE_NAME_IS_i422_yuy2) C_YUV422_UYVY( p_line, p_y, p_u, p_v ); C_YUV422_UYVY( p_line, p_y, p_u, p_v ); C_YUV422_UYVY( p_line, p_y, p_u, p_v ); C_YUV422_UYVY( p_line, p_y, p_u, p_v ); #elif defined (MODULE_NAME_IS_i422_yuy2_mmx) MMX_CALL( MMX_YUV422_UYVY ); #endif } p_y += i_source_margin; p_u += i_source_margin_c; p_v += i_source_margin_c; p_line += i_dest_margin; } #if defined (MODULE_NAME_IS_i422_yuy2_mmx) MMX_END; #elif defined (MODULE_NAME_IS_i422_yuy2_sse2) SSE2_END; #endif #endif }
void I420_R5G5B5_MMX(void * pHandle, uint8_t *p_y, uint8_t *p_u, uint8_t *p_v, uint16_t *p_d) { PSOFTCONV pconv = pHandle; if (!pHandle) return; int b_hscale = pconv->b_hscale; /* horizontal scaling type */ unsigned int i_vscale = pconv->i_vscale; /* vertical scaling type */ unsigned int i_x, i_y; /* horizontal and vertical indexes */ CC_RECT * dst_rect = &pconv->dst_dis_rect; CC_RECT * src_rect = &pconv->src_rect; int dst_stride = pconv->dst_pitch; uint16_t *p_pic = (uint16_t *)((uint8_t *)p_d + pconv->dst_offset); int i_right_margin = pconv->i_right_margin; int i_rewind; int i_scale_count; /* scale modulo counter */ int i_chroma_width = WIDTHOF(src_rect) / 2; /* chroma width */ uint16_t * p_pic_start; /* beginning of the current line for copy */ /* Conversion buffer pointer */ uint16_t * p_buffer_start = pconv->pConvBuffer; uint16_t * p_buffer; /* Offset array pointer */ int * p_offset_start = pconv->pOffsetBuffer; int * p_offset; const int i_source_margin = pconv->i_source_margin; const int i_source_margin_c = pconv->i_source_margin_c; /* * Perform conversion */ i_scale_count = ( i_vscale == 1 ) ? HEIGHTOF(dst_rect) : HEIGHTOF(src_rect); if( WIDTHOF(src_rect) & 7 ) { i_rewind = 8 - ( WIDTHOF(src_rect) & 7 ); } else { i_rewind = 0; } for( i_y = 0; i_y < HEIGHTOF(src_rect); i_y++ ) { p_pic_start = p_pic; p_buffer = b_hscale ? p_buffer_start : p_pic; for ( i_x = WIDTHOF(src_rect) / 8; i_x--; ) { MMX_CALL ( MMX_INIT_16 MMX_YUV_MUL MMX_YUV_ADD MMX_UNPACK_15 ); p_y += 8; p_u += 4; p_v += 4; p_buffer += 8; } /* Here we do some unaligned reads and duplicate conversions, but * at least we have all the pixels */ if( i_rewind ) { p_y -= i_rewind; p_u -= i_rewind >> 1; p_v -= i_rewind >> 1; p_buffer -= i_rewind; MMX_CALL ( MMX_INIT_16 MMX_YUV_MUL MMX_YUV_ADD MMX_UNPACK_15 ); p_y += 8; p_u += 4; p_v += 4; p_buffer += 8; } SCALE_WIDTH; SCALE_HEIGHT( 2 ); p_y += i_source_margin; if( i_y % 2 ) { p_u += i_source_margin_c; p_v += i_source_margin_c; } } /* re-enable FPU registers */ MMX_END; }