2 * Copyright (c) 2012 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavutil/x86/asm.h"
22 #include "libavutil/cpu.h"
23 #include "libswresample/swresample_internal.h"
25 int swri_resample_int16_mmx2 (struct ResampleContext *c, int16_t *dst, const int16_t *src, int *consumed, int src_size, int dst_size, int update_ctx);
26 int swri_resample_int16_sse2 (struct ResampleContext *c, int16_t *dst, const int16_t *src, int *consumed, int src_size, int dst_size, int update_ctx);
27 int swri_resample_float_sse (struct ResampleContext *c, float *dst, const float *src, int *consumed, int src_size, int dst_size, int update_ctx);
28 int swri_resample_double_sse2(struct ResampleContext *c, double *dst, const double *src, int *consumed, int src_size, int dst_size, int update_ctx);
30 DECLARE_ALIGNED(16, const uint64_t, ff_resample_int16_rounder)[2] = { 0x0000000000004000ULL, 0x0000000000000000ULL};
32 #define COMMON_CORE_INT16_MMX2 \
33 x86_reg len= -2*c->filter_length;\
35 "movq "MANGLE(ff_resample_int16_rounder)", %%mm0 \n\t"\
37 "movq (%1, %0), %%mm1 \n\t"\
38 "pmaddwd (%2, %0), %%mm1 \n\t"\
39 "paddd %%mm1, %%mm0 \n\t"\
42 "pshufw $0x0E, %%mm0, %%mm1 \n\t"\
43 "paddd %%mm1, %%mm0 \n\t"\
44 "psrad $15, %%mm0 \n\t"\
45 "packssdw %%mm0, %%mm0 \n\t"\
46 "movd %%mm0, (%3) \n\t"\
48 : "r" (((uint8_t*)(src+sample_index))-len),\
49 "r" (((uint8_t*)filter)-len),\
51 NAMED_CONSTRAINTS_ADD(ff_resample_int16_rounder)\
54 #define LINEAR_CORE_INT16_MMX2 \
55 x86_reg len= -2*c->filter_length;\
57 "pxor %%mm0, %%mm0 \n\t"\
58 "pxor %%mm2, %%mm2 \n\t"\
60 "movq (%3, %0), %%mm1 \n\t"\
61 "movq %%mm1, %%mm3 \n\t"\
62 "pmaddwd (%4, %0), %%mm1 \n\t"\
63 "pmaddwd (%5, %0), %%mm3 \n\t"\
64 "paddd %%mm1, %%mm0 \n\t"\
65 "paddd %%mm3, %%mm2 \n\t"\
68 "pshufw $0x0E, %%mm0, %%mm1 \n\t"\
69 "pshufw $0x0E, %%mm2, %%mm3 \n\t"\
70 "paddd %%mm1, %%mm0 \n\t"\
71 "paddd %%mm3, %%mm2 \n\t"\
72 "movd %%mm0, %1 \n\t"\
73 "movd %%mm2, %2 \n\t"\
77 : "r" (((uint8_t*)(src+sample_index))-len),\
78 "r" (((uint8_t*)filter)-len),\
79 "r" (((uint8_t*)(filter+c->filter_alloc))-len)\
82 #define COMMON_CORE_INT16_SSE2 \
83 x86_reg len= -2*c->filter_length;\
85 "movdqa "MANGLE(ff_resample_int16_rounder)", %%xmm0 \n\t"\
87 "movdqu (%1, %0), %%xmm1 \n\t"\
88 "pmaddwd (%2, %0), %%xmm1 \n\t"\
89 "paddd %%xmm1, %%xmm0 \n\t"\
92 "pshufd $0x0E, %%xmm0, %%xmm1 \n\t"\
93 "paddd %%xmm1, %%xmm0 \n\t"\
94 "pshufd $0x01, %%xmm0, %%xmm1 \n\t"\
95 "paddd %%xmm1, %%xmm0 \n\t"\
96 "psrad $15, %%xmm0 \n\t"\
97 "packssdw %%xmm0, %%xmm0 \n\t"\
98 "movd %%xmm0, (%3) \n\t"\
100 : "r" (((uint8_t*)(src+sample_index))-len),\
101 "r" (((uint8_t*)filter)-len),\
103 NAMED_CONSTRAINTS_ADD(ff_resample_int16_rounder)\
106 #define LINEAR_CORE_INT16_SSE2 \
107 x86_reg len= -2*c->filter_length;\
109 "pxor %%xmm0, %%xmm0 \n\t"\
110 "pxor %%xmm2, %%xmm2 \n\t"\
112 "movdqu (%3, %0), %%xmm1 \n\t"\
113 "movdqa %%xmm1, %%xmm3 \n\t"\
114 "pmaddwd (%4, %0), %%xmm1 \n\t"\
115 "pmaddwd (%5, %0), %%xmm3 \n\t"\
116 "paddd %%xmm1, %%xmm0 \n\t"\
117 "paddd %%xmm3, %%xmm2 \n\t"\
120 "pshufd $0x0E, %%xmm0, %%xmm1 \n\t"\
121 "pshufd $0x0E, %%xmm2, %%xmm3 \n\t"\
122 "paddd %%xmm1, %%xmm0 \n\t"\
123 "paddd %%xmm3, %%xmm2 \n\t"\
124 "pshufd $0x01, %%xmm0, %%xmm1 \n\t"\
125 "pshufd $0x01, %%xmm2, %%xmm3 \n\t"\
126 "paddd %%xmm1, %%xmm0 \n\t"\
127 "paddd %%xmm3, %%xmm2 \n\t"\
128 "movd %%xmm0, %1 \n\t"\
129 "movd %%xmm2, %2 \n\t"\
133 : "r" (((uint8_t*)(src+sample_index))-len),\
134 "r" (((uint8_t*)filter)-len),\
135 "r" (((uint8_t*)(filter+c->filter_alloc))-len)\
136 XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3")\
139 #define COMMON_CORE_FLT_SSE \
140 x86_reg len= -4*c->filter_length;\
142 "xorps %%xmm0, %%xmm0 \n\t"\
144 "movups (%1, %0), %%xmm1 \n\t"\
145 "mulps (%2, %0), %%xmm1 \n\t"\
146 "addps %%xmm1, %%xmm0 \n\t"\
149 "movhlps %%xmm0, %%xmm1 \n\t"\
150 "addps %%xmm1, %%xmm0 \n\t"\
151 "movss %%xmm0, %%xmm1 \n\t"\
152 "shufps $1, %%xmm0, %%xmm0 \n\t"\
153 "addps %%xmm1, %%xmm0 \n\t"\
154 "movss %%xmm0, (%3) \n\t"\
156 : "r" (((uint8_t*)(src+sample_index))-len),\
157 "r" (((uint8_t*)filter)-len),\
161 #define LINEAR_CORE_FLT_SSE \
162 x86_reg len= -4*c->filter_length;\
164 "xorps %%xmm0, %%xmm0 \n\t"\
165 "xorps %%xmm2, %%xmm2 \n\t"\
167 "movups (%3, %0), %%xmm1 \n\t"\
168 "movaps %%xmm1, %%xmm3 \n\t"\
169 "mulps (%4, %0), %%xmm1 \n\t"\
170 "mulps (%5, %0), %%xmm3 \n\t"\
171 "addps %%xmm1, %%xmm0 \n\t"\
172 "addps %%xmm3, %%xmm2 \n\t"\
175 "movhlps %%xmm0, %%xmm1 \n\t"\
176 "movhlps %%xmm2, %%xmm3 \n\t"\
177 "addps %%xmm1, %%xmm0 \n\t"\
178 "addps %%xmm3, %%xmm2 \n\t"\
179 "movss %%xmm0, %%xmm1 \n\t"\
180 "movss %%xmm2, %%xmm3 \n\t"\
181 "shufps $1, %%xmm0, %%xmm0 \n\t"\
182 "shufps $1, %%xmm2, %%xmm2 \n\t"\
183 "addps %%xmm1, %%xmm0 \n\t"\
184 "addps %%xmm3, %%xmm2 \n\t"\
185 "movss %%xmm0, %1 \n\t"\
186 "movss %%xmm2, %2 \n\t"\
190 : "r" (((uint8_t*)(src+sample_index))-len),\
191 "r" (((uint8_t*)filter)-len),\
192 "r" (((uint8_t*)(filter+c->filter_alloc))-len)\
193 XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3")\
196 #define COMMON_CORE_DBL_SSE2 \
197 x86_reg len= -8*c->filter_length;\
199 "xorpd %%xmm0, %%xmm0 \n\t"\
201 "movupd (%1, %0), %%xmm1 \n\t"\
202 "mulpd (%2, %0), %%xmm1 \n\t"\
203 "addpd %%xmm1, %%xmm0 \n\t"\
206 "movhlps %%xmm0, %%xmm1 \n\t"\
207 "addpd %%xmm1, %%xmm0 \n\t"\
208 "movsd %%xmm0, (%3) \n\t"\
210 : "r" (((uint8_t*)(src+sample_index))-len),\
211 "r" (((uint8_t*)filter)-len),\
215 #define LINEAR_CORE_DBL_SSE2 \
216 x86_reg len= -8*c->filter_length;\
218 "xorpd %%xmm0, %%xmm0 \n\t"\
219 "xorpd %%xmm2, %%xmm2 \n\t"\
221 "movupd (%3, %0), %%xmm1 \n\t"\
222 "movapd %%xmm1, %%xmm3 \n\t"\
223 "mulpd (%4, %0), %%xmm1 \n\t"\
224 "mulpd (%5, %0), %%xmm3 \n\t"\
225 "addpd %%xmm1, %%xmm0 \n\t"\
226 "addpd %%xmm3, %%xmm2 \n\t"\
229 "movhlps %%xmm0, %%xmm1 \n\t"\
230 "movhlps %%xmm2, %%xmm3 \n\t"\
231 "addpd %%xmm1, %%xmm0 \n\t"\
232 "addpd %%xmm3, %%xmm2 \n\t"\
233 "movsd %%xmm0, %1 \n\t"\
234 "movsd %%xmm2, %2 \n\t"\
238 : "r" (((uint8_t*)(src+sample_index))-len),\
239 "r" (((uint8_t*)filter)-len),\
240 "r" (((uint8_t*)(filter+c->filter_alloc))-len)\
241 XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3")\