nengel@2: /*
nengel@2:  * MMX optimized DSP utils
nengel@2:  * Copyright (c) 2007  Aurelien Jacobs <aurel@gnuage.org>
nengel@2:  *
nengel@2:  * This file is part of FFmpeg.
nengel@2:  *
nengel@2:  * FFmpeg is free software; you can redistribute it and/or
nengel@2:  * modify it under the terms of the GNU Lesser General Public
nengel@2:  * License as published by the Free Software Foundation; either
nengel@2:  * version 2.1 of the License, or (at your option) any later version.
nengel@2:  *
nengel@2:  * FFmpeg is distributed in the hope that it will be useful,
nengel@2:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
nengel@2:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
nengel@2:  * Lesser General Public License for more details.
nengel@2:  *
nengel@2:  * You should have received a copy of the GNU Lesser General Public
nengel@2:  * License along with FFmpeg; if not, write to the Free Software
nengel@2:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
nengel@2:  */
nengel@2: 
nengel@2: #ifndef AVCODEC_X86_DSPUTIL_MMX_H
nengel@2: #define AVCODEC_X86_DSPUTIL_MMX_H
nengel@2: 
nengel@2: #include <stdint.h>
nengel@2: #include "libavcodec/dsputil.h"
nengel@2: 
nengel@2: typedef struct { uint64_t a, b; } xmm_reg;
nengel@2: 
nengel@2: extern const uint64_t ff_bone;
nengel@2: extern const uint64_t ff_wtwo;
nengel@2: 
nengel@2: extern const uint64_t ff_pdw_80000000[2];
nengel@2: 
nengel@2: extern const uint64_t ff_pw_3;
nengel@2: extern const uint64_t ff_pw_4;
nengel@2: extern const xmm_reg  ff_pw_5;
nengel@2: extern const xmm_reg  ff_pw_8;
nengel@2: extern const uint64_t ff_pw_15;
nengel@2: extern const xmm_reg  ff_pw_16;
nengel@2: extern const uint64_t ff_pw_20;
nengel@2: extern const xmm_reg  ff_pw_28;
nengel@2: extern const xmm_reg  ff_pw_32;
nengel@2: extern const uint64_t ff_pw_42;
nengel@2: extern const xmm_reg  ff_pw_64;
nengel@2: extern const uint64_t ff_pw_96;
nengel@2: extern const uint64_t ff_pw_128;
nengel@2: extern const uint64_t ff_pw_255;
nengel@2: 
nengel@2: extern const uint64_t ff_pb_1;
nengel@2: extern const uint64_t ff_pb_3;
nengel@2: extern const uint64_t ff_pb_7;
nengel@2: extern const uint64_t ff_pb_1F;
nengel@2: extern const uint64_t ff_pb_3F;
nengel@2: extern const uint64_t ff_pb_81;
nengel@2: extern const uint64_t ff_pb_A1;
nengel@2: extern const uint64_t ff_pb_FC;
nengel@2: 
nengel@2: extern const double ff_pd_1[2];
nengel@2: extern const double ff_pd_2[2];
nengel@2: 
nengel@2: #define LOAD4(stride,in,a,b,c,d)\
nengel@2:     "movq 0*"#stride"+"#in", "#a"\n\t"\
nengel@2:     "movq 1*"#stride"+"#in", "#b"\n\t"\
nengel@2:     "movq 2*"#stride"+"#in", "#c"\n\t"\
nengel@2:     "movq 3*"#stride"+"#in", "#d"\n\t"
nengel@2: 
nengel@2: #define STORE4(stride,out,a,b,c,d)\
nengel@2:     "movq "#a", 0*"#stride"+"#out"\n\t"\
nengel@2:     "movq "#b", 1*"#stride"+"#out"\n\t"\
nengel@2:     "movq "#c", 2*"#stride"+"#out"\n\t"\
nengel@2:     "movq "#d", 3*"#stride"+"#out"\n\t"
nengel@2: 
nengel@2: /* in/out: mma=mma+mmb, mmb=mmb-mma */
nengel@2: #define SUMSUB_BA( a, b ) \
nengel@2:     "paddw "#b", "#a" \n\t"\
nengel@2:     "paddw "#b", "#b" \n\t"\
nengel@2:     "psubw "#a", "#b" \n\t"
nengel@2: 
nengel@2: #define SBUTTERFLY(a,b,t,n,m)\
nengel@2:     "mov" #m " " #a ", " #t "         \n\t" /* abcd */\
nengel@2:     "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
nengel@2:     "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
nengel@2: 
nengel@2: #define TRANSPOSE4(a,b,c,d,t)\
nengel@2:     SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
nengel@2:     SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
nengel@2:     SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
nengel@2:     SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
nengel@2: 
nengel@2: // e,f,g,h can be memory
nengel@2: // out: a,d,t,c
nengel@2: #define TRANSPOSE8x4(a,b,c,d,e,f,g,h,t)\
nengel@2:     "punpcklbw " #e ", " #a " \n\t" /* a0 e0 a1 e1 a2 e2 a3 e3 */\
nengel@2:     "punpcklbw " #f ", " #b " \n\t" /* b0 f0 b1 f1 b2 f2 b3 f3 */\
nengel@2:     "punpcklbw " #g ", " #c " \n\t" /* c0 g0 c1 g1 c2 g2 d3 g3 */\
nengel@2:     "punpcklbw " #h ", " #d " \n\t" /* d0 h0 d1 h1 d2 h2 d3 h3 */\
nengel@2:     SBUTTERFLY(a, b, t, bw, q)   /* a= a0 b0 e0 f0 a1 b1 e1 f1 */\
nengel@2:                                  /* t= a2 b2 e2 f2 a3 b3 e3 f3 */\
nengel@2:     SBUTTERFLY(c, d, b, bw, q)   /* c= c0 d0 g0 h0 c1 d1 g1 h1 */\
nengel@2:                                  /* b= c2 d2 g2 h2 c3 d3 g3 h3 */\
nengel@2:     SBUTTERFLY(a, c, d, wd, q)   /* a= a0 b0 c0 d0 e0 f0 g0 h0 */\
nengel@2:                                  /* d= a1 b1 c1 d1 e1 f1 g1 h1 */\
nengel@2:     SBUTTERFLY(t, b, c, wd, q)   /* t= a2 b2 c2 d2 e2 f2 g2 h2 */\
nengel@2:                                  /* c= a3 b3 c3 d3 e3 f3 g3 h3 */
nengel@2: 
nengel@2: #if ARCH_X86_64
nengel@2: // permutes 01234567 -> 05736421
nengel@2: #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
nengel@2:     SBUTTERFLY(a,b,%%xmm8,wd,dqa)\
nengel@2:     SBUTTERFLY(c,d,b,wd,dqa)\
nengel@2:     SBUTTERFLY(e,f,d,wd,dqa)\
nengel@2:     SBUTTERFLY(g,h,f,wd,dqa)\
nengel@2:     SBUTTERFLY(a,c,h,dq,dqa)\
nengel@2:     SBUTTERFLY(%%xmm8,b,c,dq,dqa)\
nengel@2:     SBUTTERFLY(e,g,b,dq,dqa)\
nengel@2:     SBUTTERFLY(d,f,g,dq,dqa)\
nengel@2:     SBUTTERFLY(a,e,f,qdq,dqa)\
nengel@2:     SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\
nengel@2:     SBUTTERFLY(h,b,d,qdq,dqa)\
nengel@2:     SBUTTERFLY(c,g,b,qdq,dqa)\
nengel@2:     "movdqa %%xmm8, "#g"              \n\t"
nengel@2: #else
nengel@2: #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
nengel@2:     "movdqa "#h", "#t"                \n\t"\
nengel@2:     SBUTTERFLY(a,b,h,wd,dqa)\
nengel@2:     "movdqa "#h", 16"#t"              \n\t"\
nengel@2:     "movdqa "#t", "#h"                \n\t"\
nengel@2:     SBUTTERFLY(c,d,b,wd,dqa)\
nengel@2:     SBUTTERFLY(e,f,d,wd,dqa)\
nengel@2:     SBUTTERFLY(g,h,f,wd,dqa)\
nengel@2:     SBUTTERFLY(a,c,h,dq,dqa)\
nengel@2:     "movdqa "#h", "#t"                \n\t"\
nengel@2:     "movdqa 16"#t", "#h"              \n\t"\
nengel@2:     SBUTTERFLY(h,b,c,dq,dqa)\
nengel@2:     SBUTTERFLY(e,g,b,dq,dqa)\
nengel@2:     SBUTTERFLY(d,f,g,dq,dqa)\
nengel@2:     SBUTTERFLY(a,e,f,qdq,dqa)\
nengel@2:     SBUTTERFLY(h,d,e,qdq,dqa)\
nengel@2:     "movdqa "#h", 16"#t"              \n\t"\
nengel@2:     "movdqa "#t", "#h"                \n\t"\
nengel@2:     SBUTTERFLY(h,b,d,qdq,dqa)\
nengel@2:     SBUTTERFLY(c,g,b,qdq,dqa)\
nengel@2:     "movdqa 16"#t", "#g"              \n\t"
nengel@2: #endif
nengel@2: 
nengel@2: #define MOVQ_WONE(regd) \
nengel@2:     __asm__ volatile ( \
nengel@2:     "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
nengel@2:     "psrlw $15, %%" #regd ::)
nengel@2: 
nengel@2: void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
nengel@2: void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
nengel@2: void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
nengel@2: 
nengel@2: void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
nengel@2: void ff_avg_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
nengel@2: void ff_put_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
nengel@2: void ff_avg_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
nengel@2: 
nengel@2: void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd);
nengel@2: void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src, int stride, int rnd);
nengel@2: 
nengel@2: void ff_lpc_compute_autocorr_sse2(const int32_t *data, int len, int lag,
nengel@2:                                    double *autoc);
nengel@2: 
nengel@2: void ff_mmx_idct(DCTELEM *block);
nengel@2: void ff_mmxext_idct(DCTELEM *block);
nengel@2: 
nengel@2: #endif /* AVCODEC_X86_DSPUTIL_MMX_H */