yading@10: /*
yading@10:  * BlackFin DSPUTILS
yading@10:  *
yading@10:  * Copyright (C) 2007 Marc Hoffman <marc.hoffman@analog.com>
yading@10:  * Copyright (c) 2006 Michael Benjamin <michael.benjamin@analog.com>
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: #include "libavutil/attributes.h"
yading@10: #include "libavcodec/avcodec.h"
yading@10: #include "libavcodec/dsputil.h"
yading@10: #include "dsputil_bfin.h"
yading@10: 
yading@10: int off;
yading@10: 
yading@10: static void bfin_idct_add (uint8_t *dest, int line_size, int16_t *block)
yading@10: {
yading@10:     ff_bfin_idct (block);
yading@10:     ff_bfin_add_pixels_clamped (block, dest, line_size);
yading@10: }
yading@10: 
yading@10: static void bfin_idct_put (uint8_t *dest, int line_size, int16_t *block)
yading@10: {
yading@10:     ff_bfin_idct (block);
yading@10:     ff_bfin_put_pixels_clamped (block, dest, line_size);
yading@10: }
yading@10: 
yading@10: 
yading@10: static void bfin_clear_blocks (int16_t *blocks)
yading@10: {
yading@10:     // This is just a simple memset.
yading@10:     //
yading@10:     __asm__("P0=192; "
yading@10:         "I0=%0;  "
yading@10:         "R0=0;   "
yading@10:         "LSETUP(clear_blocks_blkfn_lab,clear_blocks_blkfn_lab)LC0=P0;"
yading@10:         "clear_blocks_blkfn_lab:"
yading@10:         "[I0++]=R0;"
yading@10:         ::"a" (blocks):"P0","I0","R0");
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs16 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     return ff_bfin_z_sad16x16 (blk1,blk2,line_size,line_size,h);
yading@10: }
yading@10: 
yading@10: static int bfin_vsad_intra16 (void *c, uint8_t *blk1, uint8_t *dummy, int stride, int h) {
yading@10:     return ff_bfin_z_sad16x16 (blk1,blk1+stride,stride<<1,stride<<1,h);
yading@10: }
yading@10: 
yading@10: static int bfin_vsad (void *c, uint8_t *blk1, uint8_t *blk2, int stride, int h) {
yading@10:     return ff_bfin_z_sad16x16 (blk1,blk1+stride,stride<<1,stride<<1,h)
yading@10:         + ff_bfin_z_sad16x16 (blk2,blk2+stride,stride<<1,stride<<1,h);
yading@10: }
yading@10: 
yading@10: static uint8_t vtmp_blk[256] attribute_l1_data_b;
yading@10: 
yading@10: static int bfin_pix_abs16_x2 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     ff_bfin_put_pixels16uc (vtmp_blk, blk2, blk2+1, 16, line_size, h);
yading@10:     return ff_bfin_z_sad16x16 (blk1, vtmp_blk, line_size, 16, h);
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs16_y2 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     ff_bfin_put_pixels16uc (vtmp_blk, blk2, blk2+line_size, 16, line_size, h);
yading@10:     return ff_bfin_z_sad16x16 (blk1, vtmp_blk, line_size, 16, h);
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs16_xy2 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     ff_bfin_z_put_pixels16_xy2 (vtmp_blk, blk2, 16, line_size, h);
yading@10:     return ff_bfin_z_sad16x16 (blk1, vtmp_blk, line_size, 16, h);
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs8 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     return ff_bfin_z_sad8x8 (blk1,blk2,line_size,line_size, h);
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs8_x2 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     ff_bfin_put_pixels8uc (vtmp_blk, blk2, blk2+1, 8, line_size, h);
yading@10:     return ff_bfin_z_sad8x8 (blk1, vtmp_blk, line_size, 8, h);
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs8_y2 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     ff_bfin_put_pixels8uc (vtmp_blk, blk2, blk2+line_size, 8, line_size, h);
yading@10:     return ff_bfin_z_sad8x8 (blk1, vtmp_blk, line_size, 8, h);
yading@10: }
yading@10: 
yading@10: static int bfin_pix_abs8_xy2 (void *c, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
yading@10: {
yading@10:     ff_bfin_z_put_pixels8_xy2 (vtmp_blk, blk2, 8, line_size, h);
yading@10:     return ff_bfin_z_sad8x8 (blk1, vtmp_blk, line_size, 8, h);
yading@10: }
yading@10: 
yading@10: 
yading@10: /*
yading@10:   decoder optimization
yading@10:   start on 2/11 100 frames of 352x240@25 compiled with no optimization -g debugging
yading@10:   9.824s ~ 2.44x off
yading@10:   6.360s ~ 1.58x off with -O2
yading@10:   5.740s ~ 1.43x off with idcts
yading@10: 
yading@10:   2.64s    2/20 same sman.mp4 decode only
yading@10: 
yading@10: */
yading@10: 
yading@10: av_cold void ff_dsputil_init_bfin(DSPContext *c, AVCodecContext *avctx)
yading@10: {
yading@10:     const int high_bit_depth = avctx->bits_per_raw_sample > 8;
yading@10: 
yading@10:     c->diff_pixels        = ff_bfin_diff_pixels;
yading@10:     c->put_pixels_clamped = ff_bfin_put_pixels_clamped;
yading@10:     c->add_pixels_clamped = ff_bfin_add_pixels_clamped;
yading@10: 
yading@10:     if (!high_bit_depth)
yading@10:         c->get_pixels     = ff_bfin_get_pixels;
yading@10:     c->clear_blocks       = bfin_clear_blocks;
yading@10:     c->pix_sum            = ff_bfin_pix_sum;
yading@10:     c->pix_norm1          = ff_bfin_pix_norm1;
yading@10: 
yading@10:     c->sad[0]             = bfin_pix_abs16;
yading@10:     c->sad[1]             = bfin_pix_abs8;
yading@10: 
yading@10: /*     c->vsad[0]            = bfin_vsad; */
yading@10: /*     c->vsad[4]            = bfin_vsad_intra16; */
yading@10: 
yading@10:     /* TODO [0] 16  [1] 8 */
yading@10:     c->pix_abs[0][0] = bfin_pix_abs16;
yading@10:     c->pix_abs[0][1] = bfin_pix_abs16_x2;
yading@10:     c->pix_abs[0][2] = bfin_pix_abs16_y2;
yading@10:     c->pix_abs[0][3] = bfin_pix_abs16_xy2;
yading@10: 
yading@10:     c->pix_abs[1][0] = bfin_pix_abs8;
yading@10:     c->pix_abs[1][1] = bfin_pix_abs8_x2;
yading@10:     c->pix_abs[1][2] = bfin_pix_abs8_y2;
yading@10:     c->pix_abs[1][3] = bfin_pix_abs8_xy2;
yading@10: 
yading@10: 
yading@10:     c->sse[0] = ff_bfin_sse16;
yading@10:     c->sse[1] = ff_bfin_sse8;
yading@10:     c->sse[2] = ff_bfin_sse4;
yading@10: 
yading@10:     if (avctx->bits_per_raw_sample <= 8) {
yading@10:         if (avctx->dct_algo == FF_DCT_AUTO)
yading@10:             c->fdct                  = ff_bfin_fdct;
yading@10: 
yading@10:         if (avctx->idct_algo == FF_IDCT_AUTO) {
yading@10:             c->idct_permutation_type = FF_NO_IDCT_PERM;
yading@10:             c->idct                  = ff_bfin_idct;
yading@10:             c->idct_add              = bfin_idct_add;
yading@10:             c->idct_put              = bfin_idct_put;
yading@10:         }
yading@10:     }
yading@10: }