37f4581b9ca1af002c1c6ab1dd051491261e68c2
[libav.git] / libavcodec / x86 / dsputil_mmx.h
1 /*
2 * MMX optimized DSP utils
3 * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
4 *
5 * This file is part of Libav.
6 *
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #ifndef AVCODEC_X86_DSPUTIL_MMX_H
23 #define AVCODEC_X86_DSPUTIL_MMX_H
24
25 #include <stdint.h>
26 #include "libavcodec/dsputil.h"
27 #include "libavutil/x86_cpu.h"
28
29 typedef struct { uint64_t a, b; } xmm_reg;
30
31 extern const uint64_t ff_bone;
32 extern const uint64_t ff_wtwo;
33
34 extern const uint64_t ff_pdw_80000000[2];
35
36 extern const xmm_reg ff_pw_3;
37 extern const xmm_reg ff_pw_4;
38 extern const xmm_reg ff_pw_5;
39 extern const xmm_reg ff_pw_8;
40 extern const uint64_t ff_pw_15;
41 extern const xmm_reg ff_pw_16;
42 extern const xmm_reg ff_pw_18;
43 extern const uint64_t ff_pw_20;
44 extern const xmm_reg ff_pw_27;
45 extern const xmm_reg ff_pw_28;
46 extern const xmm_reg ff_pw_32;
47 extern const uint64_t ff_pw_42;
48 extern const uint64_t ff_pw_53;
49 extern const xmm_reg ff_pw_63;
50 extern const xmm_reg ff_pw_64;
51 extern const uint64_t ff_pw_96;
52 extern const uint64_t ff_pw_128;
53 extern const uint64_t ff_pw_255;
54
55 extern const xmm_reg ff_pb_1;
56 extern const xmm_reg ff_pb_3;
57 extern const uint64_t ff_pb_7;
58 extern const uint64_t ff_pb_1F;
59 extern const uint64_t ff_pb_3F;
60 extern const uint64_t ff_pb_81;
61 extern const xmm_reg ff_pb_A1;
62 extern const xmm_reg ff_pb_F8;
63 extern const uint64_t ff_pb_FC;
64 extern const xmm_reg ff_pb_FE;
65
66 extern const double ff_pd_1[2];
67 extern const double ff_pd_2[2];
68
69 #define LOAD4(stride,in,a,b,c,d)\
70 "movq 0*"#stride"+"#in", "#a"\n\t"\
71 "movq 1*"#stride"+"#in", "#b"\n\t"\
72 "movq 2*"#stride"+"#in", "#c"\n\t"\
73 "movq 3*"#stride"+"#in", "#d"\n\t"
74
75 #define STORE4(stride,out,a,b,c,d)\
76 "movq "#a", 0*"#stride"+"#out"\n\t"\
77 "movq "#b", 1*"#stride"+"#out"\n\t"\
78 "movq "#c", 2*"#stride"+"#out"\n\t"\
79 "movq "#d", 3*"#stride"+"#out"\n\t"
80
81 /* in/out: mma=mma+mmb, mmb=mmb-mma */
82 #define SUMSUB_BA( a, b ) \
83 "paddw "#b", "#a" \n\t"\
84 "paddw "#b", "#b" \n\t"\
85 "psubw "#a", "#b" \n\t"
86
87 #define SBUTTERFLY(a,b,t,n,m)\
88 "mov" #m " " #a ", " #t " \n\t" /* abcd */\
89 "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\
90 "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\
91
92 #define TRANSPOSE4(a,b,c,d,t)\
93 SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
94 SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
95 SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
96 SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
97
98 static inline void transpose4x4(uint8_t *dst, uint8_t *src, x86_reg dst_stride, x86_reg src_stride){
99 __asm__ volatile( //FIXME could save 1 instruction if done as 8x4 ...
100 "movd (%1), %%mm0 \n\t"
101 "add %3, %1 \n\t"
102 "movd (%1), %%mm1 \n\t"
103 "movd (%1,%3,1), %%mm2 \n\t"
104 "movd (%1,%3,2), %%mm3 \n\t"
105 "punpcklbw %%mm1, %%mm0 \n\t"
106 "punpcklbw %%mm3, %%mm2 \n\t"
107 "movq %%mm0, %%mm1 \n\t"
108 "punpcklwd %%mm2, %%mm0 \n\t"
109 "punpckhwd %%mm2, %%mm1 \n\t"
110 "movd %%mm0, (%0) \n\t"
111 "add %2, %0 \n\t"
112 "punpckhdq %%mm0, %%mm0 \n\t"
113 "movd %%mm0, (%0) \n\t"
114 "movd %%mm1, (%0,%2,1) \n\t"
115 "punpckhdq %%mm1, %%mm1 \n\t"
116 "movd %%mm1, (%0,%2,2) \n\t"
117
118 : "+&r" (dst),
119 "+&r" (src)
120 : "r" (dst_stride),
121 "r" (src_stride)
122 : "memory"
123 );
124 }
125
126 // e,f,g,h can be memory
127 // out: a,d,t,c
128 #define TRANSPOSE8x4(a,b,c,d,e,f,g,h,t)\
129 "punpcklbw " #e ", " #a " \n\t" /* a0 e0 a1 e1 a2 e2 a3 e3 */\
130 "punpcklbw " #f ", " #b " \n\t" /* b0 f0 b1 f1 b2 f2 b3 f3 */\
131 "punpcklbw " #g ", " #c " \n\t" /* c0 g0 c1 g1 c2 g2 d3 g3 */\
132 "punpcklbw " #h ", " #d " \n\t" /* d0 h0 d1 h1 d2 h2 d3 h3 */\
133 SBUTTERFLY(a, b, t, bw, q) /* a= a0 b0 e0 f0 a1 b1 e1 f1 */\
134 /* t= a2 b2 e2 f2 a3 b3 e3 f3 */\
135 SBUTTERFLY(c, d, b, bw, q) /* c= c0 d0 g0 h0 c1 d1 g1 h1 */\
136 /* b= c2 d2 g2 h2 c3 d3 g3 h3 */\
137 SBUTTERFLY(a, c, d, wd, q) /* a= a0 b0 c0 d0 e0 f0 g0 h0 */\
138 /* d= a1 b1 c1 d1 e1 f1 g1 h1 */\
139 SBUTTERFLY(t, b, c, wd, q) /* t= a2 b2 c2 d2 e2 f2 g2 h2 */\
140 /* c= a3 b3 c3 d3 e3 f3 g3 h3 */
141
142 #if ARCH_X86_64
143 // permutes 01234567 -> 05736421
144 #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
145 SBUTTERFLY(a,b,%%xmm8,wd,dqa)\
146 SBUTTERFLY(c,d,b,wd,dqa)\
147 SBUTTERFLY(e,f,d,wd,dqa)\
148 SBUTTERFLY(g,h,f,wd,dqa)\
149 SBUTTERFLY(a,c,h,dq,dqa)\
150 SBUTTERFLY(%%xmm8,b,c,dq,dqa)\
151 SBUTTERFLY(e,g,b,dq,dqa)\
152 SBUTTERFLY(d,f,g,dq,dqa)\
153 SBUTTERFLY(a,e,f,qdq,dqa)\
154 SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\
155 SBUTTERFLY(h,b,d,qdq,dqa)\
156 SBUTTERFLY(c,g,b,qdq,dqa)\
157 "movdqa %%xmm8, "#g" \n\t"
158 #else
159 #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
160 "movdqa "#h", "#t" \n\t"\
161 SBUTTERFLY(a,b,h,wd,dqa)\
162 "movdqa "#h", 16"#t" \n\t"\
163 "movdqa "#t", "#h" \n\t"\
164 SBUTTERFLY(c,d,b,wd,dqa)\
165 SBUTTERFLY(e,f,d,wd,dqa)\
166 SBUTTERFLY(g,h,f,wd,dqa)\
167 SBUTTERFLY(a,c,h,dq,dqa)\
168 "movdqa "#h", "#t" \n\t"\
169 "movdqa 16"#t", "#h" \n\t"\
170 SBUTTERFLY(h,b,c,dq,dqa)\
171 SBUTTERFLY(e,g,b,dq,dqa)\
172 SBUTTERFLY(d,f,g,dq,dqa)\
173 SBUTTERFLY(a,e,f,qdq,dqa)\
174 SBUTTERFLY(h,d,e,qdq,dqa)\
175 "movdqa "#h", 16"#t" \n\t"\
176 "movdqa "#t", "#h" \n\t"\
177 SBUTTERFLY(h,b,d,qdq,dqa)\
178 SBUTTERFLY(c,g,b,qdq,dqa)\
179 "movdqa 16"#t", "#g" \n\t"
180 #endif
181
182 #define MOVQ_WONE(regd) \
183 __asm__ volatile ( \
184 "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
185 "psrlw $15, %%" #regd ::)
186
187 void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx);
188 void ff_dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
189
190 void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
191 void ff_put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
192 void ff_put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
193
194 void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
195 void ff_avg_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
196 void ff_put_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
197 void ff_avg_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
198
199 void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd);
200 void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src, int stride, int rnd);
201
202 void ff_put_rv40_qpel8_mc33_mmx(uint8_t *block, uint8_t *pixels, int line_size);
203 void ff_put_rv40_qpel16_mc33_mmx(uint8_t *block, uint8_t *pixels, int line_size);
204 void ff_avg_rv40_qpel8_mc33_mmx(uint8_t *block, uint8_t *pixels, int line_size);
205 void ff_avg_rv40_qpel16_mc33_mmx(uint8_t *block, uint8_t *pixels, int line_size);
206
207 void ff_mmx_idct(DCTELEM *block);
208 void ff_mmxext_idct(DCTELEM *block);
209
210
211 void ff_deinterlace_line_mmx(uint8_t *dst,
212 const uint8_t *lum_m4, const uint8_t *lum_m3,
213 const uint8_t *lum_m2, const uint8_t *lum_m1,
214 const uint8_t *lum,
215 int size);
216
217 void ff_deinterlace_line_inplace_mmx(const uint8_t *lum_m4,
218 const uint8_t *lum_m3,
219 const uint8_t *lum_m2,
220 const uint8_t *lum_m1,
221 const uint8_t *lum, int size);
222
223 #endif /* AVCODEC_X86_DSPUTIL_MMX_H */