yading@10: /*
yading@10:  * MMX optimized discrete wavelet transform
yading@10:  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
yading@10:  * Copyright (c) 2010 David Conrad
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/x86/asm.h"
yading@10: #include "dsputil_mmx.h"
yading@10: #include "dirac_dwt.h"
yading@10: 
yading@10: #define COMPOSE_VERTICAL(ext, align) \
yading@10: void ff_vertical_compose53iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width); \
yading@10: void ff_vertical_compose_dirac53iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width); \
yading@10: void ff_vertical_compose_dd137iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, int width); \
yading@10: void ff_vertical_compose_dd97iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, int width); \
yading@10: void ff_vertical_compose_haar##ext(IDWTELEM *b0, IDWTELEM *b1, int width); \
yading@10: void ff_horizontal_compose_haar0i##ext(IDWTELEM *b, IDWTELEM *tmp, int w);\
yading@10: void ff_horizontal_compose_haar1i##ext(IDWTELEM *b, IDWTELEM *tmp, int w);\
yading@10: \
yading@10: static void vertical_compose53iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width) \
yading@10: { \
yading@10:     int i, width_align = width&~(align-1); \
yading@10: \
yading@10:     for(i=width_align; i<width; i++) \
yading@10:         b1[i] = COMPOSE_53iL0(b0[i], b1[i], b2[i]); \
yading@10: \
yading@10:     ff_vertical_compose53iL0##ext(b0, b1, b2, width_align); \
yading@10: } \
yading@10: \
yading@10: static void vertical_compose_dirac53iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width) \
yading@10: { \
yading@10:     int i, width_align = width&~(align-1); \
yading@10: \
yading@10:     for(i=width_align; i<width; i++) \
yading@10:         b1[i] = COMPOSE_DIRAC53iH0(b0[i], b1[i], b2[i]); \
yading@10: \
yading@10:     ff_vertical_compose_dirac53iH0##ext(b0, b1, b2, width_align); \
yading@10: } \
yading@10: \
yading@10: static void vertical_compose_dd137iL0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, \
yading@10:                                            IDWTELEM *b3, IDWTELEM *b4, int width) \
yading@10: { \
yading@10:     int i, width_align = width&~(align-1); \
yading@10: \
yading@10:     for(i=width_align; i<width; i++) \
yading@10:         b2[i] = COMPOSE_DD137iL0(b0[i], b1[i], b2[i], b3[i], b4[i]); \
yading@10: \
yading@10:     ff_vertical_compose_dd137iL0##ext(b0, b1, b2, b3, b4, width_align); \
yading@10: } \
yading@10: \
yading@10: static void vertical_compose_dd97iH0##ext(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, \
yading@10:                                           IDWTELEM *b3, IDWTELEM *b4, int width) \
yading@10: { \
yading@10:     int i, width_align = width&~(align-1); \
yading@10: \
yading@10:     for(i=width_align; i<width; i++) \
yading@10:         b2[i] = COMPOSE_DD97iH0(b0[i], b1[i], b2[i], b3[i], b4[i]); \
yading@10: \
yading@10:     ff_vertical_compose_dd97iH0##ext(b0, b1, b2, b3, b4, width_align); \
yading@10: } \
yading@10: static void vertical_compose_haar##ext(IDWTELEM *b0, IDWTELEM *b1, int width) \
yading@10: { \
yading@10:     int i, width_align = width&~(align-1); \
yading@10: \
yading@10:     for(i=width_align; i<width; i++) { \
yading@10:         b0[i] = COMPOSE_HAARiL0(b0[i], b1[i]); \
yading@10:         b1[i] = COMPOSE_HAARiH0(b1[i], b0[i]); \
yading@10:     } \
yading@10: \
yading@10:     ff_vertical_compose_haar##ext(b0, b1, width_align); \
yading@10: } \
yading@10: static void horizontal_compose_haar0i##ext(IDWTELEM *b, IDWTELEM *tmp, int w)\
yading@10: {\
yading@10:     int w2= w>>1;\
yading@10:     int x= w2 - (w2&(align-1));\
yading@10:     ff_horizontal_compose_haar0i##ext(b, tmp, w);\
yading@10: \
yading@10:     for (; x < w2; x++) {\
yading@10:         b[2*x  ] = tmp[x];\
yading@10:         b[2*x+1] = COMPOSE_HAARiH0(b[x+w2], tmp[x]);\
yading@10:     }\
yading@10: }\
yading@10: static void horizontal_compose_haar1i##ext(IDWTELEM *b, IDWTELEM *tmp, int w)\
yading@10: {\
yading@10:     int w2= w>>1;\
yading@10:     int x= w2 - (w2&(align-1));\
yading@10:     ff_horizontal_compose_haar1i##ext(b, tmp, w);\
yading@10: \
yading@10:     for (; x < w2; x++) {\
yading@10:         b[2*x  ] = (tmp[x] + 1)>>1;\
yading@10:         b[2*x+1] = (COMPOSE_HAARiH0(b[x+w2], tmp[x]) + 1)>>1;\
yading@10:     }\
yading@10: }\
yading@10: \
yading@10: 
yading@10: #if HAVE_YASM
yading@10: #if !ARCH_X86_64
yading@10: COMPOSE_VERTICAL(_mmx, 4)
yading@10: #endif
yading@10: COMPOSE_VERTICAL(_sse2, 8)
yading@10: 
yading@10: 
yading@10: void ff_horizontal_compose_dd97i_ssse3(IDWTELEM *b, IDWTELEM *tmp, int w);
yading@10: 
yading@10: static void horizontal_compose_dd97i_ssse3(IDWTELEM *b, IDWTELEM *tmp, int w)
yading@10: {
yading@10:     int w2= w>>1;
yading@10:     int x= w2 - (w2&7);
yading@10:     ff_horizontal_compose_dd97i_ssse3(b, tmp, w);
yading@10: 
yading@10:     for (; x < w2; x++) {
yading@10:         b[2*x  ] = (tmp[x] + 1)>>1;
yading@10:         b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
yading@10:     }
yading@10: }
yading@10: #endif
yading@10: 
yading@10: void ff_spatial_idwt_init_mmx(DWTContext *d, enum dwt_type type)
yading@10: {
yading@10: #if HAVE_YASM
yading@10:   int mm_flags = av_get_cpu_flags();
yading@10: 
yading@10: #if !ARCH_X86_64
yading@10:     if (!(mm_flags & AV_CPU_FLAG_MMX))
yading@10:         return;
yading@10: 
yading@10:     switch (type) {
yading@10:     case DWT_DIRAC_DD9_7:
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose53iL0_mmx;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0_mmx;
yading@10:         break;
yading@10:     case DWT_DIRAC_LEGALL5_3:
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose53iL0_mmx;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dirac53iH0_mmx;
yading@10:         break;
yading@10:     case DWT_DIRAC_DD13_7:
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose_dd137iL0_mmx;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0_mmx;
yading@10:         break;
yading@10:     case DWT_DIRAC_HAAR0:
yading@10:         d->vertical_compose   = (void*)vertical_compose_haar_mmx;
yading@10:         d->horizontal_compose = horizontal_compose_haar0i_mmx;
yading@10:         break;
yading@10:     case DWT_DIRAC_HAAR1:
yading@10:         d->vertical_compose   = (void*)vertical_compose_haar_mmx;
yading@10:         d->horizontal_compose = horizontal_compose_haar1i_mmx;
yading@10:         break;
yading@10:     }
yading@10: #endif
yading@10: 
yading@10:     if (!(mm_flags & AV_CPU_FLAG_SSE2))
yading@10:         return;
yading@10: 
yading@10:     switch (type) {
yading@10:     case DWT_DIRAC_DD9_7:
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose53iL0_sse2;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0_sse2;
yading@10:         break;
yading@10:     case DWT_DIRAC_LEGALL5_3:
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose53iL0_sse2;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dirac53iH0_sse2;
yading@10:         break;
yading@10:     case DWT_DIRAC_DD13_7:
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose_dd137iL0_sse2;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0_sse2;
yading@10:         break;
yading@10:     case DWT_DIRAC_HAAR0:
yading@10:         d->vertical_compose   = (void*)vertical_compose_haar_sse2;
yading@10:         d->horizontal_compose = horizontal_compose_haar0i_sse2;
yading@10:         break;
yading@10:     case DWT_DIRAC_HAAR1:
yading@10:         d->vertical_compose   = (void*)vertical_compose_haar_sse2;
yading@10:         d->horizontal_compose = horizontal_compose_haar1i_sse2;
yading@10:         break;
yading@10:     }
yading@10: 
yading@10:     if (!(mm_flags & AV_CPU_FLAG_SSSE3))
yading@10:         return;
yading@10: 
yading@10:     switch (type) {
yading@10:     case DWT_DIRAC_DD9_7:
yading@10:         d->horizontal_compose = horizontal_compose_dd97i_ssse3;
yading@10:         break;
yading@10:     }
yading@10: #endif // HAVE_YASM
yading@10: }