1 /* 2 * MMX optimized DSP utils 3 * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org> 4 * 5 * This file is part of FFmpeg. 6 * 7 * FFmpeg is free software; you can redistribute it and/or 8 * modify it under the terms of the GNU Lesser General Public 9 * License as published by the Free Software Foundation; either 10 * version 2.1 of the License, or (at your option) any later version. 11 * 12 * FFmpeg is distributed in the hope that it will be useful, 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of 14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 15 * Lesser General Public License for more details. 16 * 17 * You should have received a copy of the GNU Lesser General Public 18 * License along with FFmpeg; if not, write to the Free Software 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 20 */ 21 22 #ifndef FFMPEG_DSPUTIL_MMX_H 23 #define FFMPEG_DSPUTIL_MMX_H 24 25 #include <stdint.h> 26 #include "dsputil.h" 27 28 typedef struct { uint64_t a, b; } xmm_t; 29 30 extern const uint64_t ff_bone; 31 extern const uint64_t ff_wtwo; 32 33 extern const uint64_t ff_pdw_80000000[2]; 34 35 extern const uint64_t ff_pw_3; 36 extern const uint64_t ff_pw_4; 37 extern const xmm_t ff_pw_5; 38 extern const uint64_t ff_pw_8; 39 extern const uint64_t ff_pw_15; 40 extern const xmm_t ff_pw_16; 41 extern const uint64_t ff_pw_20; 42 extern const xmm_t ff_pw_28; 43 extern const xmm_t ff_pw_32; 44 extern const uint64_t ff_pw_42; 45 extern const uint64_t ff_pw_64; 46 extern const uint64_t ff_pw_96; 47 extern const uint64_t ff_pw_128; 48 extern const uint64_t ff_pw_255; 49 50 extern const uint64_t ff_pb_1; 51 extern const uint64_t ff_pb_3; 52 extern const uint64_t ff_pb_7; 53 extern const uint64_t ff_pb_3F; 54 extern const uint64_t ff_pb_A1; 55 extern const uint64_t ff_pb_FC; 56 57 extern const double ff_pd_1[2]; 58 extern const double ff_pd_2[2]; 59 60 /* in/out: mma=mma+mmb, mmb=mmb-mma */ 61 #define SUMSUB_BA( a, b ) \ 62 "paddw "#b", "#a" \n\t"\ 63 "paddw "#b", "#b" \n\t"\ 64 "psubw "#a", "#b" \n\t" 65 66 #define SBUTTERFLY(a,b,t,n,m)\ 67 "mov" #m " " #a ", " #t " \n\t" /* abcd */\ 68 "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\ 69 "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\ 70 71 #define TRANSPOSE4(a,b,c,d,t)\ 72 SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\ 73 SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\ 74 SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\ 75 SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */ 76 77 #ifdef ARCH_X86_64 78 // permutes 01234567 -> 05736421 79 #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\ 80 SBUTTERFLY(a,b,%%xmm8,wd,dqa)\ 81 SBUTTERFLY(c,d,b,wd,dqa)\ 82 SBUTTERFLY(e,f,d,wd,dqa)\ 83 SBUTTERFLY(g,h,f,wd,dqa)\ 84 SBUTTERFLY(a,c,h,dq,dqa)\ 85 SBUTTERFLY(%%xmm8,b,c,dq,dqa)\ 86 SBUTTERFLY(e,g,b,dq,dqa)\ 87 SBUTTERFLY(d,f,g,dq,dqa)\ 88 SBUTTERFLY(a,e,f,qdq,dqa)\ 89 SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\ 90 SBUTTERFLY(h,b,d,qdq,dqa)\ 91 SBUTTERFLY(c,g,b,qdq,dqa)\ 92 "movdqa %%xmm8, "#g" \n\t" 93 #else 94 #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\ 95 "movdqa "#h", "#t" \n\t"\ 96 SBUTTERFLY(a,b,h,wd,dqa)\ 97 "movdqa "#h", 16"#t" \n\t"\ 98 "movdqa "#t", "#h" \n\t"\ 99 SBUTTERFLY(c,d,b,wd,dqa)\ 100 SBUTTERFLY(e,f,d,wd,dqa)\ 101 SBUTTERFLY(g,h,f,wd,dqa)\ 102 SBUTTERFLY(a,c,h,dq,dqa)\ 103 "movdqa "#h", "#t" \n\t"\ 104 "movdqa 16"#t", "#h" \n\t"\ 105 SBUTTERFLY(h,b,c,dq,dqa)\ 106 SBUTTERFLY(e,g,b,dq,dqa)\ 107 SBUTTERFLY(d,f,g,dq,dqa)\ 108 SBUTTERFLY(a,e,f,qdq,dqa)\ 109 SBUTTERFLY(h,d,e,qdq,dqa)\ 110 "movdqa "#h", 16"#t" \n\t"\ 111 "movdqa "#t", "#h" \n\t"\ 112 SBUTTERFLY(h,b,d,qdq,dqa)\ 113 SBUTTERFLY(c,g,b,qdq,dqa)\ 114 "movdqa 16"#t", "#g" \n\t" 115 #endif 116 117 #define MOVQ_WONE(regd) \ 118 asm volatile ( \ 119 "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ 120 "psrlw $15, %%" #regd ::) 121 122 void dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx); 123 124 #endif /* FFMPEG_DSPUTIL_MMX_H */ 125