yading@10: /*
yading@10:  * Copyright (C) 2004 the ffmpeg project
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg 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:  * FFmpeg 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 FFmpeg; 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:  * Standard C DSP-oriented functions cribbed from the original VP3
yading@10:  * source code.
yading@10:  */
yading@10: 
yading@10: #include "libavutil/attributes.h"
yading@10: #include "libavutil/common.h"
yading@10: #include "avcodec.h"
yading@10: #include "dsputil.h"
yading@10: #include "rnd_avg.h"
yading@10: #include "vp3dsp.h"
yading@10: 
yading@10: #define IdctAdjustBeforeShift 8
yading@10: #define xC1S7 64277
yading@10: #define xC2S6 60547
yading@10: #define xC3S5 54491
yading@10: #define xC4S4 46341
yading@10: #define xC5S3 36410
yading@10: #define xC6S2 25080
yading@10: #define xC7S1 12785
yading@10: 
yading@10: #define M(a,b) (((a) * (b))>>16)
yading@10: 
yading@10: static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
yading@10: {
yading@10:     int16_t *ip = input;
yading@10: 
yading@10:     int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
yading@10:     int Ed, Gd, Add, Bdd, Fd, Hd;
yading@10: 
yading@10:     int i;
yading@10: 
yading@10:     /* Inverse DCT on the rows now */
yading@10:     for (i = 0; i < 8; i++) {
yading@10:         /* Check for non-zero values */
yading@10:         if ( ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
yading@10:              ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
yading@10:             A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
yading@10:             B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
yading@10:             C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
yading@10:             D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
yading@10: 
yading@10:             Ad = M(xC4S4, (A - C));
yading@10:             Bd = M(xC4S4, (B - D));
yading@10: 
yading@10:             Cd = A + C;
yading@10:             Dd = B + D;
yading@10: 
yading@10:             E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
yading@10:             F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
yading@10: 
yading@10:             G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
yading@10:             H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
yading@10: 
yading@10:             Ed = E - G;
yading@10:             Gd = E + G;
yading@10: 
yading@10:             Add = F + Ad;
yading@10:             Bdd = Bd - H;
yading@10: 
yading@10:             Fd = F - Ad;
yading@10:             Hd = Bd + H;
yading@10: 
yading@10:             /*  Final sequence of operations over-write original inputs. */
yading@10:             ip[0 * 8] = Gd + Cd ;
yading@10:             ip[7 * 8] = Gd - Cd ;
yading@10: 
yading@10:             ip[1 * 8] = Add + Hd;
yading@10:             ip[2 * 8] = Add - Hd;
yading@10: 
yading@10:             ip[3 * 8] = Ed + Dd ;
yading@10:             ip[4 * 8] = Ed - Dd ;
yading@10: 
yading@10:             ip[5 * 8] = Fd + Bdd;
yading@10:             ip[6 * 8] = Fd - Bdd;
yading@10:         }
yading@10: 
yading@10:         ip += 1;            /* next row */
yading@10:     }
yading@10: 
yading@10:     ip = input;
yading@10: 
yading@10:     for ( i = 0; i < 8; i++) {
yading@10:         /* Check for non-zero values (bitwise or faster than ||) */
yading@10:         if ( ip[1] | ip[2] | ip[3] |
yading@10:              ip[4] | ip[5] | ip[6] | ip[7] ) {
yading@10: 
yading@10:             A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
yading@10:             B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
yading@10:             C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
yading@10:             D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
yading@10: 
yading@10:             Ad = M(xC4S4, (A - C));
yading@10:             Bd = M(xC4S4, (B - D));
yading@10: 
yading@10:             Cd = A + C;
yading@10:             Dd = B + D;
yading@10: 
yading@10:             E = M(xC4S4, (ip[0] + ip[4])) + 8;
yading@10:             F = M(xC4S4, (ip[0] - ip[4])) + 8;
yading@10: 
yading@10:             if(type==1){  //HACK
yading@10:                 E += 16*128;
yading@10:                 F += 16*128;
yading@10:             }
yading@10: 
yading@10:             G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
yading@10:             H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
yading@10: 
yading@10:             Ed = E - G;
yading@10:             Gd = E + G;
yading@10: 
yading@10:             Add = F + Ad;
yading@10:             Bdd = Bd - H;
yading@10: 
yading@10:             Fd = F - Ad;
yading@10:             Hd = Bd + H;
yading@10: 
yading@10:             /* Final sequence of operations over-write original inputs. */
yading@10:             if (type == 1) {
yading@10:                 dst[0*stride] = av_clip_uint8((Gd + Cd )  >> 4);
yading@10:                 dst[7*stride] = av_clip_uint8((Gd - Cd )  >> 4);
yading@10: 
yading@10:                 dst[1*stride] = av_clip_uint8((Add + Hd ) >> 4);
yading@10:                 dst[2*stride] = av_clip_uint8((Add - Hd ) >> 4);
yading@10: 
yading@10:                 dst[3*stride] = av_clip_uint8((Ed + Dd )  >> 4);
yading@10:                 dst[4*stride] = av_clip_uint8((Ed - Dd )  >> 4);
yading@10: 
yading@10:                 dst[5*stride] = av_clip_uint8((Fd + Bdd ) >> 4);
yading@10:                 dst[6*stride] = av_clip_uint8((Fd - Bdd ) >> 4);
yading@10:             }else{
yading@10:                 dst[0*stride] = av_clip_uint8(dst[0*stride] + ((Gd + Cd )  >> 4));
yading@10:                 dst[7*stride] = av_clip_uint8(dst[7*stride] + ((Gd - Cd )  >> 4));
yading@10: 
yading@10:                 dst[1*stride] = av_clip_uint8(dst[1*stride] + ((Add + Hd ) >> 4));
yading@10:                 dst[2*stride] = av_clip_uint8(dst[2*stride] + ((Add - Hd ) >> 4));
yading@10: 
yading@10:                 dst[3*stride] = av_clip_uint8(dst[3*stride] + ((Ed + Dd )  >> 4));
yading@10:                 dst[4*stride] = av_clip_uint8(dst[4*stride] + ((Ed - Dd )  >> 4));
yading@10: 
yading@10:                 dst[5*stride] = av_clip_uint8(dst[5*stride] + ((Fd + Bdd ) >> 4));
yading@10:                 dst[6*stride] = av_clip_uint8(dst[6*stride] + ((Fd - Bdd ) >> 4));
yading@10:             }
yading@10: 
yading@10:         } else {
yading@10:             if (type == 1) {
yading@10:                 dst[0*stride]=
yading@10:                 dst[1*stride]=
yading@10:                 dst[2*stride]=
yading@10:                 dst[3*stride]=
yading@10:                 dst[4*stride]=
yading@10:                 dst[5*stride]=
yading@10:                 dst[6*stride]=
yading@10:                 dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20));
yading@10:             }else{
yading@10:                 if(ip[0]){
yading@10:                     int v= ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20);
yading@10:                     dst[0*stride] = av_clip_uint8(dst[0*stride] + v);
yading@10:                     dst[1*stride] = av_clip_uint8(dst[1*stride] + v);
yading@10:                     dst[2*stride] = av_clip_uint8(dst[2*stride] + v);
yading@10:                     dst[3*stride] = av_clip_uint8(dst[3*stride] + v);
yading@10:                     dst[4*stride] = av_clip_uint8(dst[4*stride] + v);
yading@10:                     dst[5*stride] = av_clip_uint8(dst[5*stride] + v);
yading@10:                     dst[6*stride] = av_clip_uint8(dst[6*stride] + v);
yading@10:                     dst[7*stride] = av_clip_uint8(dst[7*stride] + v);
yading@10:                 }
yading@10:             }
yading@10:         }
yading@10: 
yading@10:         ip += 8;            /* next column */
yading@10:         dst++;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size,
yading@10:                            int16_t *block/*align 16*/)
yading@10: {
yading@10:     idct(dest, line_size, block, 1);
yading@10:     memset(block, 0, sizeof(*block) * 64);
yading@10: }
yading@10: 
yading@10: static void vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size,
yading@10:                            int16_t *block/*align 16*/)
yading@10: {
yading@10:     idct(dest, line_size, block, 2);
yading@10:     memset(block, 0, sizeof(*block) * 64);
yading@10: }
yading@10: 
yading@10: static void vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size,
yading@10:                               int16_t *block/*align 16*/)
yading@10: {
yading@10:     int i, dc = (block[0] + 15) >> 5;
yading@10: 
yading@10:     for(i = 0; i < 8; i++){
yading@10:         dest[0] = av_clip_uint8(dest[0] + dc);
yading@10:         dest[1] = av_clip_uint8(dest[1] + dc);
yading@10:         dest[2] = av_clip_uint8(dest[2] + dc);
yading@10:         dest[3] = av_clip_uint8(dest[3] + dc);
yading@10:         dest[4] = av_clip_uint8(dest[4] + dc);
yading@10:         dest[5] = av_clip_uint8(dest[5] + dc);
yading@10:         dest[6] = av_clip_uint8(dest[6] + dc);
yading@10:         dest[7] = av_clip_uint8(dest[7] + dc);
yading@10:         dest += line_size;
yading@10:     }
yading@10:     block[0] = 0;
yading@10: }
yading@10: 
yading@10: static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
yading@10:                                 int *bounding_values)
yading@10: {
yading@10:     unsigned char *end;
yading@10:     int filter_value;
yading@10:     const int nstride= -stride;
yading@10: 
yading@10:     for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
yading@10:         filter_value =
yading@10:             (first_pixel[2 * nstride] - first_pixel[ stride])
yading@10:          +3*(first_pixel[0          ] - first_pixel[nstride]);
yading@10:         filter_value = bounding_values[(filter_value + 4) >> 3];
yading@10:         first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
yading@10:         first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
yading@10:                                 int *bounding_values)
yading@10: {
yading@10:     unsigned char *end;
yading@10:     int filter_value;
yading@10: 
yading@10:     for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
yading@10:         filter_value =
yading@10:             (first_pixel[-2] - first_pixel[ 1])
yading@10:          +3*(first_pixel[ 0] - first_pixel[-1]);
yading@10:         filter_value = bounding_values[(filter_value + 4) >> 3];
yading@10:         first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
yading@10:         first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
yading@10:                                  const uint8_t *src2, ptrdiff_t stride, int h)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < h; i++) {
yading@10:         uint32_t a, b;
yading@10: 
yading@10:         a = AV_RN32(&src1[i * stride]);
yading@10:         b = AV_RN32(&src2[i * stride]);
yading@10:         AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
yading@10:         a = AV_RN32(&src1[i * stride + 4]);
yading@10:         b = AV_RN32(&src2[i * stride + 4]);
yading@10:         AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
yading@10:     }
yading@10: }
yading@10: 
yading@10: av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
yading@10: {
yading@10:     c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
yading@10: 
yading@10:     c->idct_put      = vp3_idct_put_c;
yading@10:     c->idct_add      = vp3_idct_add_c;
yading@10:     c->idct_dc_add   = vp3_idct_dc_add_c;
yading@10:     c->v_loop_filter = vp3_v_loop_filter_c;
yading@10:     c->h_loop_filter = vp3_h_loop_filter_c;
yading@10: 
yading@10:     if (ARCH_ARM)
yading@10:         ff_vp3dsp_init_arm(c, flags);
yading@10:     if (ARCH_BFIN)
yading@10:         ff_vp3dsp_init_bfin(c, flags);
yading@10:     if (ARCH_PPC)
yading@10:         ff_vp3dsp_init_ppc(c, flags);
yading@10:     if (ARCH_X86)
yading@10:         ff_vp3dsp_init_x86(c, flags);
yading@10: }