yading@10: /*
yading@10:  * RV30/40 decoder common dsp functions
yading@10:  * Copyright (c) 2007 Mike Melanson, Konstantin Shishkov
yading@10:  * Copyright (c) 2011 Janne Grunau
yading@10:  *
yading@10:  * This file is part of Libav.
yading@10:  *
yading@10:  * Libav is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * Libav is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with Libav; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * RV30/40 decoder common dsp functions
yading@10:  */
yading@10: 
yading@10: #include "rv34dsp.h"
yading@10: #include "libavutil/common.h"
yading@10: 
yading@10: /**
yading@10:  * @name RV30/40 inverse transform functions
yading@10:  * @{
yading@10:  */
yading@10: 
yading@10: static av_always_inline void rv34_row_transform(int temp[16], int16_t *block)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i = 0; i < 4; i++){
yading@10:         const int z0 = 13*(block[i+4*0] +    block[i+4*2]);
yading@10:         const int z1 = 13*(block[i+4*0] -    block[i+4*2]);
yading@10:         const int z2 =  7* block[i+4*1] - 17*block[i+4*3];
yading@10:         const int z3 = 17* block[i+4*1] +  7*block[i+4*3];
yading@10: 
yading@10:         temp[4*i+0] = z0 + z3;
yading@10:         temp[4*i+1] = z1 + z2;
yading@10:         temp[4*i+2] = z1 - z2;
yading@10:         temp[4*i+3] = z0 - z3;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Real Video 3.0/4.0 inverse transform + sample reconstruction
yading@10:  * Code is almost the same as in SVQ3, only scaling is different.
yading@10:  */
yading@10: static void rv34_idct_add_c(uint8_t *dst, ptrdiff_t stride, int16_t *block){
yading@10:     int      temp[16];
yading@10:     int      i;
yading@10: 
yading@10:     rv34_row_transform(temp, block);
yading@10:     memset(block, 0, 16*sizeof(int16_t));
yading@10: 
yading@10:     for(i = 0; i < 4; i++){
yading@10:         const int z0 = 13*(temp[4*0+i] +    temp[4*2+i]) + 0x200;
yading@10:         const int z1 = 13*(temp[4*0+i] -    temp[4*2+i]) + 0x200;
yading@10:         const int z2 =  7* temp[4*1+i] - 17*temp[4*3+i];
yading@10:         const int z3 = 17* temp[4*1+i] +  7*temp[4*3+i];
yading@10: 
yading@10:         dst[0] = av_clip_uint8( dst[0] + ( (z0 + z3) >> 10 ) );
yading@10:         dst[1] = av_clip_uint8( dst[1] + ( (z1 + z2) >> 10 ) );
yading@10:         dst[2] = av_clip_uint8( dst[2] + ( (z1 - z2) >> 10 ) );
yading@10:         dst[3] = av_clip_uint8( dst[3] + ( (z0 - z3) >> 10 ) );
yading@10: 
yading@10:         dst  += stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * RealVideo 3.0/4.0 inverse transform for DC block
yading@10:  *
yading@10:  * Code is almost the same as rv34_inv_transform()
yading@10:  * but final coefficients are multiplied by 1.5 and have no rounding.
yading@10:  */
yading@10: static void rv34_inv_transform_noround_c(int16_t *block){
yading@10:     int temp[16];
yading@10:     int i;
yading@10: 
yading@10:     rv34_row_transform(temp, block);
yading@10: 
yading@10:     for(i = 0; i < 4; i++){
yading@10:         const int z0 = 39*(temp[4*0+i] +    temp[4*2+i]);
yading@10:         const int z1 = 39*(temp[4*0+i] -    temp[4*2+i]);
yading@10:         const int z2 = 21* temp[4*1+i] - 51*temp[4*3+i];
yading@10:         const int z3 = 51* temp[4*1+i] + 21*temp[4*3+i];
yading@10: 
yading@10:         block[i*4+0] = (z0 + z3) >> 11;
yading@10:         block[i*4+1] = (z1 + z2) >> 11;
yading@10:         block[i*4+2] = (z1 - z2) >> 11;
yading@10:         block[i*4+3] = (z0 - z3) >> 11;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void rv34_idct_dc_add_c(uint8_t *dst, ptrdiff_t stride, int dc)
yading@10: {
yading@10:     int i, j;
yading@10: 
yading@10:     dc = (13*13*dc + 0x200) >> 10;
yading@10:     for (i = 0; i < 4; i++)
yading@10:     {
yading@10:         for (j = 0; j < 4; j++)
yading@10:             dst[j] = av_clip_uint8( dst[j] + dc );
yading@10: 
yading@10:         dst += stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void rv34_inv_transform_dc_noround_c(int16_t *block)
yading@10: {
yading@10:     int16_t dc = (13 * 13 * 3 * block[0]) >> 11;
yading@10:     int i, j;
yading@10: 
yading@10:     for (i = 0; i < 4; i++, block += 4)
yading@10:         for (j = 0; j < 4; j++)
yading@10:             block[j] = dc;
yading@10: }
yading@10: 
yading@10: /** @} */ // transform
yading@10: 
yading@10: 
yading@10: av_cold void ff_rv34dsp_init(RV34DSPContext *c)
yading@10: {
yading@10:     c->rv34_inv_transform    = rv34_inv_transform_noround_c;
yading@10:     c->rv34_inv_transform_dc = rv34_inv_transform_dc_noround_c;
yading@10: 
yading@10:     c->rv34_idct_add    = rv34_idct_add_c;
yading@10:     c->rv34_idct_dc_add = rv34_idct_dc_add_c;
yading@10: 
yading@10:     if (ARCH_ARM)
yading@10:         ff_rv34dsp_init_arm(c);
yading@10:     if (ARCH_X86)
yading@10:         ff_rv34dsp_init_x86(c);
yading@10: }