yading@10: /* yading@10: * DSP functions for Indeo Video Interactive codecs (Indeo4 and Indeo5) yading@10: * yading@10: * Copyright (c) 2009-2011 Maxim Poliakovski 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: * DSP functions (inverse transforms, motion compensation, wavelet recompostions) yading@10: * for Indeo Video Interactive codecs. yading@10: */ yading@10: yading@10: #include "avcodec.h" yading@10: #include "ivi_common.h" yading@10: #include "ivi_dsp.h" yading@10: yading@10: void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst, yading@10: const int dst_pitch) yading@10: { yading@10: int x, y, indx; yading@10: int32_t p0, p1, p2, p3, tmp0, tmp1, tmp2; yading@10: int32_t b0_1, b0_2, b1_1, b1_2, b1_3, b2_1, b2_2, b2_3, b2_4, b2_5, b2_6; yading@10: int32_t b3_1, b3_2, b3_3, b3_4, b3_5, b3_6, b3_7, b3_8, b3_9; yading@10: int32_t pitch, back_pitch; yading@10: const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr; yading@10: const int num_bands = 4; yading@10: yading@10: /* all bands should have the same pitch */ yading@10: pitch = plane->bands[0].pitch; yading@10: yading@10: /* pixels at the position "y-1" will be set to pixels at the "y" for the 1st iteration */ yading@10: back_pitch = 0; yading@10: yading@10: /* get pointers to the wavelet bands */ yading@10: b0_ptr = plane->bands[0].buf; yading@10: b1_ptr = plane->bands[1].buf; yading@10: b2_ptr = plane->bands[2].buf; yading@10: b3_ptr = plane->bands[3].buf; yading@10: yading@10: for (y = 0; y < plane->height; y += 2) { yading@10: yading@10: if (y+2 >= plane->height) yading@10: pitch= 0; yading@10: /* load storage variables with values */ yading@10: if (num_bands > 0) { yading@10: b0_1 = b0_ptr[0]; yading@10: b0_2 = b0_ptr[pitch]; yading@10: } yading@10: yading@10: if (num_bands > 1) { yading@10: b1_1 = b1_ptr[back_pitch]; yading@10: b1_2 = b1_ptr[0]; yading@10: b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch]; yading@10: } yading@10: yading@10: if (num_bands > 2) { yading@10: b2_2 = b2_ptr[0]; // b2[x, y ] yading@10: b2_3 = b2_2; // b2[x+1,y ] = b2[x,y] yading@10: b2_5 = b2_ptr[pitch]; // b2[x ,y+1] yading@10: b2_6 = b2_5; // b2[x+1,y+1] = b2[x,y+1] yading@10: } yading@10: yading@10: if (num_bands > 3) { yading@10: b3_2 = b3_ptr[back_pitch]; // b3[x ,y-1] yading@10: b3_3 = b3_2; // b3[x+1,y-1] = b3[x ,y-1] yading@10: b3_5 = b3_ptr[0]; // b3[x ,y ] yading@10: b3_6 = b3_5; // b3[x+1,y ] = b3[x ,y ] yading@10: b3_8 = b3_2 - b3_5*6 + b3_ptr[pitch]; yading@10: b3_9 = b3_8; yading@10: } yading@10: yading@10: for (x = 0, indx = 0; x < plane->width; x+=2, indx++) { yading@10: if (x+2 >= plane->width) { yading@10: b0_ptr --; yading@10: b1_ptr --; yading@10: b2_ptr --; yading@10: b3_ptr --; yading@10: } yading@10: yading@10: /* some values calculated in the previous iterations can */ yading@10: /* be reused in the next ones, so do appropriate copying */ yading@10: b2_1 = b2_2; // b2[x-1,y ] = b2[x, y ] yading@10: b2_2 = b2_3; // b2[x ,y ] = b2[x+1,y ] yading@10: b2_4 = b2_5; // b2[x-1,y+1] = b2[x ,y+1] yading@10: b2_5 = b2_6; // b2[x ,y+1] = b2[x+1,y+1] yading@10: b3_1 = b3_2; // b3[x-1,y-1] = b3[x ,y-1] yading@10: b3_2 = b3_3; // b3[x ,y-1] = b3[x+1,y-1] yading@10: b3_4 = b3_5; // b3[x-1,y ] = b3[x ,y ] yading@10: b3_5 = b3_6; // b3[x ,y ] = b3[x+1,y ] yading@10: b3_7 = b3_8; // vert_HPF(x-1) yading@10: b3_8 = b3_9; // vert_HPF(x ) yading@10: yading@10: p0 = p1 = p2 = p3 = 0; yading@10: yading@10: /* process the LL-band by applying LPF both vertically and horizontally */ yading@10: if (num_bands > 0) { yading@10: tmp0 = b0_1; yading@10: tmp2 = b0_2; yading@10: b0_1 = b0_ptr[indx+1]; yading@10: b0_2 = b0_ptr[pitch+indx+1]; yading@10: tmp1 = tmp0 + b0_1; yading@10: yading@10: p0 = tmp0 << 4; yading@10: p1 = tmp1 << 3; yading@10: p2 = (tmp0 + tmp2) << 3; yading@10: p3 = (tmp1 + tmp2 + b0_2) << 2; yading@10: } yading@10: yading@10: /* process the HL-band by applying HPF vertically and LPF horizontally */ yading@10: if (num_bands > 1) { yading@10: tmp0 = b1_2; yading@10: tmp1 = b1_1; yading@10: b1_2 = b1_ptr[indx+1]; yading@10: b1_1 = b1_ptr[back_pitch+indx+1]; yading@10: yading@10: tmp2 = tmp1 - tmp0*6 + b1_3; yading@10: b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch+indx+1]; yading@10: yading@10: p0 += (tmp0 + tmp1) << 3; yading@10: p1 += (tmp0 + tmp1 + b1_1 + b1_2) << 2; yading@10: p2 += tmp2 << 2; yading@10: p3 += (tmp2 + b1_3) << 1; yading@10: } yading@10: yading@10: /* process the LH-band by applying LPF vertically and HPF horizontally */ yading@10: if (num_bands > 2) { yading@10: b2_3 = b2_ptr[indx+1]; yading@10: b2_6 = b2_ptr[pitch+indx+1]; yading@10: yading@10: tmp0 = b2_1 + b2_2; yading@10: tmp1 = b2_1 - b2_2*6 + b2_3; yading@10: yading@10: p0 += tmp0 << 3; yading@10: p1 += tmp1 << 2; yading@10: p2 += (tmp0 + b2_4 + b2_5) << 2; yading@10: p3 += (tmp1 + b2_4 - b2_5*6 + b2_6) << 1; yading@10: } yading@10: yading@10: /* process the HH-band by applying HPF both vertically and horizontally */ yading@10: if (num_bands > 3) { yading@10: b3_6 = b3_ptr[indx+1]; // b3[x+1,y ] yading@10: b3_3 = b3_ptr[back_pitch+indx+1]; // b3[x+1,y-1] yading@10: yading@10: tmp0 = b3_1 + b3_4; yading@10: tmp1 = b3_2 + b3_5; yading@10: tmp2 = b3_3 + b3_6; yading@10: yading@10: b3_9 = b3_3 - b3_6*6 + b3_ptr[pitch+indx+1]; yading@10: yading@10: p0 += (tmp0 + tmp1) << 2; yading@10: p1 += (tmp0 - tmp1*6 + tmp2) << 1; yading@10: p2 += (b3_7 + b3_8) << 1; yading@10: p3 += b3_7 - b3_8*6 + b3_9; yading@10: } yading@10: yading@10: /* output four pixels */ yading@10: dst[x] = av_clip_uint8((p0 >> 6) + 128); yading@10: dst[x+1] = av_clip_uint8((p1 >> 6) + 128); yading@10: dst[dst_pitch+x] = av_clip_uint8((p2 >> 6) + 128); yading@10: dst[dst_pitch+x+1] = av_clip_uint8((p3 >> 6) + 128); yading@10: }// for x yading@10: yading@10: dst += dst_pitch << 1; yading@10: yading@10: back_pitch = -pitch; yading@10: yading@10: b0_ptr += pitch + 1; yading@10: b1_ptr += pitch + 1; yading@10: b2_ptr += pitch + 1; yading@10: b3_ptr += pitch + 1; yading@10: } yading@10: } yading@10: yading@10: void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst, yading@10: const int dst_pitch) yading@10: { yading@10: int x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3; yading@10: const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr; yading@10: int32_t pitch; yading@10: yading@10: /* all bands should have the same pitch */ yading@10: pitch = plane->bands[0].pitch; yading@10: yading@10: /* get pointers to the wavelet bands */ yading@10: b0_ptr = plane->bands[0].buf; yading@10: b1_ptr = plane->bands[1].buf; yading@10: b2_ptr = plane->bands[2].buf; yading@10: b3_ptr = plane->bands[3].buf; yading@10: yading@10: for (y = 0; y < plane->height; y += 2) { yading@10: for (x = 0, indx = 0; x < plane->width; x += 2, indx++) { yading@10: /* load coefficients */ yading@10: b0 = b0_ptr[indx]; //should be: b0 = (num_bands > 0) ? b0_ptr[indx] : 0; yading@10: b1 = b1_ptr[indx]; //should be: b1 = (num_bands > 1) ? b1_ptr[indx] : 0; yading@10: b2 = b2_ptr[indx]; //should be: b2 = (num_bands > 2) ? b2_ptr[indx] : 0; yading@10: b3 = b3_ptr[indx]; //should be: b3 = (num_bands > 3) ? b3_ptr[indx] : 0; yading@10: yading@10: /* haar wavelet recomposition */ yading@10: p0 = (b0 + b1 + b2 + b3 + 2) >> 2; yading@10: p1 = (b0 + b1 - b2 - b3 + 2) >> 2; yading@10: p2 = (b0 - b1 + b2 - b3 + 2) >> 2; yading@10: p3 = (b0 - b1 - b2 + b3 + 2) >> 2; yading@10: yading@10: /* bias, convert and output four pixels */ yading@10: dst[x] = av_clip_uint8(p0 + 128); yading@10: dst[x + 1] = av_clip_uint8(p1 + 128); yading@10: dst[dst_pitch + x] = av_clip_uint8(p2 + 128); yading@10: dst[dst_pitch + x + 1] = av_clip_uint8(p3 + 128); yading@10: }// for x yading@10: yading@10: dst += dst_pitch << 1; yading@10: yading@10: b0_ptr += pitch; yading@10: b1_ptr += pitch; yading@10: b2_ptr += pitch; yading@10: b3_ptr += pitch; yading@10: }// for y yading@10: } yading@10: yading@10: /** butterfly operation for the inverse Haar transform */ yading@10: #define IVI_HAAR_BFLY(s1, s2, o1, o2, t) \ yading@10: t = (s1 - s2) >> 1;\ yading@10: o1 = (s1 + s2) >> 1;\ yading@10: o2 = t;\ yading@10: yading@10: /** inverse 8-point Haar transform */ yading@10: #define INV_HAAR8(s1, s5, s3, s7, s2, s4, s6, s8,\ yading@10: d1, d2, d3, d4, d5, d6, d7, d8,\ yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8) {\ yading@10: t1 = s1 << 1; t5 = s5 << 1;\ yading@10: IVI_HAAR_BFLY(t1, t5, t1, t5, t0); IVI_HAAR_BFLY(t1, s3, t1, t3, t0);\ yading@10: IVI_HAAR_BFLY(t5, s7, t5, t7, t0); IVI_HAAR_BFLY(t1, s2, t1, t2, t0);\ yading@10: IVI_HAAR_BFLY(t3, s4, t3, t4, t0); IVI_HAAR_BFLY(t5, s6, t5, t6, t0);\ yading@10: IVI_HAAR_BFLY(t7, s8, t7, t8, t0);\ yading@10: d1 = COMPENSATE(t1);\ yading@10: d2 = COMPENSATE(t2);\ yading@10: d3 = COMPENSATE(t3);\ yading@10: d4 = COMPENSATE(t4);\ yading@10: d5 = COMPENSATE(t5);\ yading@10: d6 = COMPENSATE(t6);\ yading@10: d7 = COMPENSATE(t7);\ yading@10: d8 = COMPENSATE(t8); } yading@10: yading@10: /** inverse 4-point Haar transform */ yading@10: #define INV_HAAR4(s1, s3, s5, s7) {\ yading@10: HAAR_BFLY(s1, s5); HAAR_BFLY(s1, s3); HAAR_BFLY(s5, s7);\ yading@10: s1 = COMPENSATE(s1);\ yading@10: s3 = COMPENSATE(s3);\ yading@10: s5 = COMPENSATE(s5);\ yading@10: s7 = COMPENSATE(s7); } yading@10: yading@10: void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch, yading@10: const uint8_t *flags) yading@10: { yading@10: int i, shift, sp1, sp2, sp3, sp4; yading@10: const int32_t *src; yading@10: int32_t *dst; yading@10: int tmp[64]; yading@10: int t0, t1, t2, t3, t4, t5, t6, t7, t8; yading@10: yading@10: /* apply the InvHaar8 to all columns */ yading@10: #define COMPENSATE(x) (x) yading@10: src = in; yading@10: dst = tmp; yading@10: for (i = 0; i < 8; i++) { yading@10: if (flags[i]) { yading@10: /* pre-scaling */ yading@10: shift = !(i & 4); yading@10: sp1 = src[ 0] << shift; yading@10: sp2 = src[ 8] << shift; yading@10: sp3 = src[16] << shift; yading@10: sp4 = src[24] << shift; yading@10: INV_HAAR8( sp1, sp2, sp3, sp4, yading@10: src[32], src[40], src[48], src[56], yading@10: dst[ 0], dst[ 8], dst[16], dst[24], yading@10: dst[32], dst[40], dst[48], dst[56], yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8); yading@10: } else yading@10: dst[ 0] = dst[ 8] = dst[16] = dst[24] = yading@10: dst[32] = dst[40] = dst[48] = dst[56] = 0; yading@10: yading@10: src++; yading@10: dst++; yading@10: } yading@10: #undef COMPENSATE yading@10: yading@10: /* apply the InvHaar8 to all rows */ yading@10: #define COMPENSATE(x) (x) yading@10: src = tmp; yading@10: for (i = 0; i < 8; i++) { yading@10: if ( !src[0] && !src[1] && !src[2] && !src[3] yading@10: && !src[4] && !src[5] && !src[6] && !src[7]) { yading@10: memset(out, 0, 8 * sizeof(out[0])); yading@10: } else { yading@10: INV_HAAR8(src[0], src[1], src[2], src[3], yading@10: src[4], src[5], src[6], src[7], yading@10: out[0], out[1], out[2], out[3], yading@10: out[4], out[5], out[6], out[7], yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8); yading@10: } yading@10: src += 8; yading@10: out += pitch; yading@10: } yading@10: #undef COMPENSATE yading@10: } yading@10: yading@10: void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch, yading@10: int blk_size) yading@10: { yading@10: int x, y; yading@10: int16_t dc_coeff; yading@10: yading@10: dc_coeff = (*in + 0) >> 3; yading@10: yading@10: for (y = 0; y < blk_size; out += pitch, y++) { yading@10: for (x = 0; x < blk_size; x++) yading@10: out[x] = dc_coeff; yading@10: } yading@10: } yading@10: yading@10: /** butterfly operation for the inverse slant transform */ yading@10: #define IVI_SLANT_BFLY(s1, s2, o1, o2, t) \ yading@10: t = s1 - s2;\ yading@10: o1 = s1 + s2;\ yading@10: o2 = t;\ yading@10: yading@10: /** This is a reflection a,b = 1/2, 5/4 for the inverse slant transform */ yading@10: #define IVI_IREFLECT(s1, s2, o1, o2, t) \ yading@10: t = ((s1 + s2*2 + 2) >> 2) + s1;\ yading@10: o2 = ((s1*2 - s2 + 2) >> 2) - s2;\ yading@10: o1 = t;\ yading@10: yading@10: /** This is a reflection a,b = 1/2, 7/8 for the inverse slant transform */ yading@10: #define IVI_SLANT_PART4(s1, s2, o1, o2, t) \ yading@10: t = s2 + ((s1*4 - s2 + 4) >> 3);\ yading@10: o2 = s1 + ((-s1 - s2*4 + 4) >> 3);\ yading@10: o1 = t;\ yading@10: yading@10: /** inverse slant8 transform */ yading@10: #define IVI_INV_SLANT8(s1, s4, s8, s5, s2, s6, s3, s7,\ yading@10: d1, d2, d3, d4, d5, d6, d7, d8,\ yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8) {\ yading@10: IVI_SLANT_PART4(s4, s5, t4, t5, t0);\ yading@10: \ yading@10: IVI_SLANT_BFLY(s1, t5, t1, t5, t0); IVI_SLANT_BFLY(s2, s6, t2, t6, t0);\ yading@10: IVI_SLANT_BFLY(s7, s3, t7, t3, t0); IVI_SLANT_BFLY(t4, s8, t4, t8, t0);\ yading@10: \ yading@10: IVI_SLANT_BFLY(t1, t2, t1, t2, t0); IVI_IREFLECT (t4, t3, t4, t3, t0);\ yading@10: IVI_SLANT_BFLY(t5, t6, t5, t6, t0); IVI_IREFLECT (t8, t7, t8, t7, t0);\ yading@10: IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\ yading@10: IVI_SLANT_BFLY(t5, t8, t5, t8, t0); IVI_SLANT_BFLY(t6, t7, t6, t7, t0);\ yading@10: d1 = COMPENSATE(t1);\ yading@10: d2 = COMPENSATE(t2);\ yading@10: d3 = COMPENSATE(t3);\ yading@10: d4 = COMPENSATE(t4);\ yading@10: d5 = COMPENSATE(t5);\ yading@10: d6 = COMPENSATE(t6);\ yading@10: d7 = COMPENSATE(t7);\ yading@10: d8 = COMPENSATE(t8);} yading@10: yading@10: /** inverse slant4 transform */ yading@10: #define IVI_INV_SLANT4(s1, s4, s2, s3, d1, d2, d3, d4, t0, t1, t2, t3, t4) {\ yading@10: IVI_SLANT_BFLY(s1, s2, t1, t2, t0); IVI_IREFLECT (s4, s3, t4, t3, t0);\ yading@10: \ yading@10: IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\ yading@10: d1 = COMPENSATE(t1);\ yading@10: d2 = COMPENSATE(t2);\ yading@10: d3 = COMPENSATE(t3);\ yading@10: d4 = COMPENSATE(t4);} yading@10: yading@10: void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags) yading@10: { yading@10: int i; yading@10: const int32_t *src; yading@10: int32_t *dst; yading@10: int tmp[64]; yading@10: int t0, t1, t2, t3, t4, t5, t6, t7, t8; yading@10: yading@10: #define COMPENSATE(x) (x) yading@10: src = in; yading@10: dst = tmp; yading@10: for (i = 0; i < 8; i++) { yading@10: if (flags[i]) { yading@10: IVI_INV_SLANT8(src[0], src[8], src[16], src[24], src[32], src[40], src[48], src[56], yading@10: dst[0], dst[8], dst[16], dst[24], dst[32], dst[40], dst[48], dst[56], yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8); yading@10: } else yading@10: dst[0] = dst[8] = dst[16] = dst[24] = dst[32] = dst[40] = dst[48] = dst[56] = 0; yading@10: yading@10: src++; yading@10: dst++; yading@10: } yading@10: #undef COMPENSATE yading@10: yading@10: #define COMPENSATE(x) ((x + 1)>>1) yading@10: src = tmp; yading@10: for (i = 0; i < 8; i++) { yading@10: if (!src[0] && !src[1] && !src[2] && !src[3] && !src[4] && !src[5] && !src[6] && !src[7]) { yading@10: memset(out, 0, 8*sizeof(out[0])); yading@10: } else { yading@10: IVI_INV_SLANT8(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7], yading@10: out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7], yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8); yading@10: } yading@10: src += 8; yading@10: out += pitch; yading@10: } yading@10: #undef COMPENSATE yading@10: } yading@10: yading@10: void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags) yading@10: { yading@10: int i; yading@10: const int32_t *src; yading@10: int32_t *dst; yading@10: int tmp[16]; yading@10: int t0, t1, t2, t3, t4; yading@10: yading@10: #define COMPENSATE(x) (x) yading@10: src = in; yading@10: dst = tmp; yading@10: for (i = 0; i < 4; i++) { yading@10: if (flags[i]) { yading@10: IVI_INV_SLANT4(src[0], src[4], src[8], src[12], yading@10: dst[0], dst[4], dst[8], dst[12], yading@10: t0, t1, t2, t3, t4); yading@10: } else yading@10: dst[0] = dst[4] = dst[8] = dst[12] = 0; yading@10: yading@10: src++; yading@10: dst++; yading@10: } yading@10: #undef COMPENSATE yading@10: yading@10: #define COMPENSATE(x) ((x + 1)>>1) yading@10: src = tmp; yading@10: for (i = 0; i < 4; i++) { yading@10: if (!src[0] && !src[1] && !src[2] && !src[3]) { yading@10: out[0] = out[1] = out[2] = out[3] = 0; yading@10: } else { yading@10: IVI_INV_SLANT4(src[0], src[1], src[2], src[3], yading@10: out[0], out[1], out[2], out[3], yading@10: t0, t1, t2, t3, t4); yading@10: } yading@10: src += 4; yading@10: out += pitch; yading@10: } yading@10: #undef COMPENSATE yading@10: } yading@10: yading@10: void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size) yading@10: { yading@10: int x, y; yading@10: int16_t dc_coeff; yading@10: yading@10: dc_coeff = (*in + 1) >> 1; yading@10: yading@10: for (y = 0; y < blk_size; out += pitch, y++) { yading@10: for (x = 0; x < blk_size; x++) yading@10: out[x] = dc_coeff; yading@10: } yading@10: } yading@10: yading@10: void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags) yading@10: { yading@10: int i; yading@10: int t0, t1, t2, t3, t4, t5, t6, t7, t8; yading@10: yading@10: #define COMPENSATE(x) ((x + 1)>>1) yading@10: for (i = 0; i < 8; i++) { yading@10: if (!in[0] && !in[1] && !in[2] && !in[3] && !in[4] && !in[5] && !in[6] && !in[7]) { yading@10: memset(out, 0, 8*sizeof(out[0])); yading@10: } else { yading@10: IVI_INV_SLANT8( in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7], yading@10: out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7], yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8); yading@10: } yading@10: in += 8; yading@10: out += pitch; yading@10: } yading@10: #undef COMPENSATE yading@10: } yading@10: yading@10: void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size) yading@10: { yading@10: int x, y; yading@10: int16_t dc_coeff; yading@10: yading@10: dc_coeff = (*in + 1) >> 1; yading@10: yading@10: for (x = 0; x < blk_size; x++) yading@10: out[x] = dc_coeff; yading@10: yading@10: out += pitch; yading@10: yading@10: for (y = 1; y < blk_size; out += pitch, y++) { yading@10: for (x = 0; x < blk_size; x++) yading@10: out[x] = 0; yading@10: } yading@10: } yading@10: yading@10: void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags) yading@10: { yading@10: int i, row2, row4, row8; yading@10: int t0, t1, t2, t3, t4, t5, t6, t7, t8; yading@10: yading@10: row2 = pitch << 1; yading@10: row4 = pitch << 2; yading@10: row8 = pitch << 3; yading@10: yading@10: #define COMPENSATE(x) ((x + 1)>>1) yading@10: for (i = 0; i < 8; i++) { yading@10: if (flags[i]) { yading@10: IVI_INV_SLANT8(in[0], in[8], in[16], in[24], in[32], in[40], in[48], in[56], yading@10: out[0], out[pitch], out[row2], out[row2 + pitch], out[row4], yading@10: out[row4 + pitch], out[row4 + row2], out[row8 - pitch], yading@10: t0, t1, t2, t3, t4, t5, t6, t7, t8); yading@10: } else { yading@10: out[0] = out[pitch] = out[row2] = out[row2 + pitch] = out[row4] = yading@10: out[row4 + pitch] = out[row4 + row2] = out[row8 - pitch] = 0; yading@10: } yading@10: yading@10: in++; yading@10: out++; yading@10: } yading@10: #undef COMPENSATE yading@10: } yading@10: yading@10: void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size) yading@10: { yading@10: int x, y; yading@10: int16_t dc_coeff; yading@10: yading@10: dc_coeff = (*in + 1) >> 1; yading@10: yading@10: for (y = 0; y < blk_size; out += pitch, y++) { yading@10: out[0] = dc_coeff; yading@10: for (x = 1; x < blk_size; x++) yading@10: out[x] = 0; yading@10: } yading@10: } yading@10: yading@10: void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch, yading@10: const uint8_t *flags) yading@10: { yading@10: int x, y; yading@10: yading@10: for (y = 0; y < 8; out += pitch, in += 8, y++) yading@10: for (x = 0; x < 8; x++) yading@10: out[x] = in[x]; yading@10: } yading@10: yading@10: void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch, yading@10: int blk_size) yading@10: { yading@10: int y; yading@10: yading@10: out[0] = in[0]; yading@10: memset(out + 1, 0, 7*sizeof(out[0])); yading@10: out += pitch; yading@10: yading@10: for (y = 1; y < 8; out += pitch, y++) yading@10: memset(out, 0, 8*sizeof(out[0])); yading@10: } yading@10: yading@10: #define IVI_MC_TEMPLATE(size, suffix, OP) \ yading@10: void ff_ivi_mc_ ## size ##x## size ## suffix (int16_t *buf, const int16_t *ref_buf, \ yading@10: uint32_t pitch, int mc_type) \ yading@10: { \ yading@10: int i, j; \ yading@10: const int16_t *wptr; \ yading@10: \ yading@10: switch (mc_type) { \ yading@10: case 0: /* fullpel (no interpolation) */ \ yading@10: for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) { \ yading@10: for (j = 0; j < size; j++) {\ yading@10: OP(buf[j], ref_buf[j]); \ yading@10: } \ yading@10: } \ yading@10: break; \ yading@10: case 1: /* horizontal halfpel interpolation */ \ yading@10: for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) \ yading@10: for (j = 0; j < size; j++) \ yading@10: OP(buf[j], (ref_buf[j] + ref_buf[j+1]) >> 1); \ yading@10: break; \ yading@10: case 2: /* vertical halfpel interpolation */ \ yading@10: wptr = ref_buf + pitch; \ yading@10: for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \ yading@10: for (j = 0; j < size; j++) \ yading@10: OP(buf[j], (ref_buf[j] + wptr[j]) >> 1); \ yading@10: break; \ yading@10: case 3: /* vertical and horizontal halfpel interpolation */ \ yading@10: wptr = ref_buf + pitch; \ yading@10: for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \ yading@10: for (j = 0; j < size; j++) \ yading@10: OP(buf[j], (ref_buf[j] + ref_buf[j+1] + wptr[j] + wptr[j+1]) >> 2); \ yading@10: break; \ yading@10: } \ yading@10: } \ yading@10: yading@10: #define OP_PUT(a, b) (a) = (b) yading@10: #define OP_ADD(a, b) (a) += (b) yading@10: yading@10: IVI_MC_TEMPLATE(8, _no_delta, OP_PUT) yading@10: IVI_MC_TEMPLATE(8, _delta, OP_ADD) yading@10: IVI_MC_TEMPLATE(4, _no_delta, OP_PUT) yading@10: IVI_MC_TEMPLATE(4, _delta, OP_ADD)