1. ffmpeg
  2. Commit 7e22514d98ecd56cc4e55dac7c813551e2152b80
Browse code

Merge remote-tracking branch 'qatar/master'

* qatar/master:
float_dsp: ppc: add a separate header for Altivec function prototypes
ARM: fix float_dsp breakage from d5a7229
Add a float DSP framework to libavutil
PPC: Move types_altivec.h and util_altivec.h from libavcodec to libavutil
ARM: Move asm.S from libavcodec to libavutil
vc1dsp: mark put/avg_vc1_mspel_mc() always_inline

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Michael Niedermayer authored on 2012/06/09 06:02:54
Showing 95 changed files
libavcodec/aac.h
History View file @ 7e22514
... ... 
@@ -30,6 +30,7 @@
30 30 
 #ifndef AVCODEC_AAC_H
31 31 
 #define AVCODEC_AAC_H
32 32
33 
+#include "libavutil/float_dsp.h"
33 34 
 #include "avcodec.h"
34 35 
 #include "dsputil.h"
35 36 
 #include "fft.h"
... ... 
@@ -292,6 +293,7 @@ typedef struct {
292 292 
     FFTContext mdct_ltp;
293 293 
     DSPContext dsp;
294 294 
     FmtConvertContext fmt_conv;
295 
+    AVFloatDSPContext fdsp;
295 296 
     int random_state;
296 297 
     /** @} */
297 298
libavcodec/aacdec.c
History View file @ 7e22514
... ... 
@@ -79,7 +79,7 @@
79 79 
            Parametric Stereo.
80 80 
  */
81 81
82 
-
82 
+#include "libavutil/float_dsp.h"
83 83 
 #include "avcodec.h"
84 84 
 #include "internal.h"
85 85 
 #include "get_bits.h"
... ... 
@@ -901,6 +901,7 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
901 901
902 902 
     ff_dsputil_init(&ac->dsp, avctx);
903 903 
     ff_fmt_convert_init(&ac->fmt_conv, avctx);
904 
+    avpriv_float_dsp_init(&ac->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
904 905
905 906 
     ac->random_state = 0x1f2e3d4c;
906 907
... ... 
@@ -2069,10 +2070,10 @@ static void windowing_and_mdct_ltp(AACContext *ac, float *out,
2069 2069 
     const float *swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
2070 2070
2071 2071 
     if (ics->window_sequence[0] != LONG_STOP_SEQUENCE) {
2072 
-        ac->dsp.vector_fmul(in, in, lwindow_prev, 1024);
2072 
+        ac->fdsp.vector_fmul(in, in, lwindow_prev, 1024);
2073 2073 
     } else {
2074 2074 
         memset(in, 0, 448 * sizeof(float));
2075 
-        ac->dsp.vector_fmul(in + 448, in + 448, swindow_prev, 128);
2075 
+        ac->fdsp.vector_fmul(in + 448, in + 448, swindow_prev, 128);
2076 2076 
     }
2077 2077 
     if (ics->window_sequence[0] != LONG_START_SEQUENCE) {
2078 2078 
         ac->dsp.vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024);
libavcodec/aacenc.c
History View file @ 7e22514
... ... 
@@ -30,6 +30,7 @@
30 30 
  * add temporal noise shaping
31 31 
  ***********************************/
32 32
33 
+#include "libavutil/float_dsp.h"
33 34 
 #include "libavutil/opt.h"
34 35 
 #include "avcodec.h"
35 36 
 #include "put_bits.h"
... ... 
@@ -182,7 +183,9 @@ static void put_audio_specific_config(AVCodecContext *avctx)
182 182 
 }
183 183
184 184 
 #define WINDOW_FUNC(type) \
185 
-static void apply_ ##type ##_window(DSPContext *dsp, SingleChannelElement *sce, const float *audio)
185 
+static void apply_ ##type ##_window(DSPContext *dsp, AVFloatDSPContext *fdsp, \
186 
+                                    SingleChannelElement *sce, \
187 
+                                    const float *audio)
186 188
187 189 
 WINDOW_FUNC(only_long)
188 190 
 {
... ... 
@@ -190,7 +193,7 @@ WINDOW_FUNC(only_long)
190 190 
     const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
191 191 
     float *out = sce->ret;
192 192
193 
-    dsp->vector_fmul        (out,        audio,        lwindow, 1024);
193 
+    fdsp->vector_fmul       (out,        audio,        lwindow, 1024);
194 194 
     dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
195 195 
 }
196 196
... ... 
@@ -200,7 +203,7 @@ WINDOW_FUNC(long_start)
200 200 
     const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
201 201 
     float *out = sce->ret;
202 202
203 
-    dsp->vector_fmul(out, audio, lwindow, 1024);
203 
+    fdsp->vector_fmul(out, audio, lwindow, 1024);
204 204 
     memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448);
205 205 
     dsp->vector_fmul_reverse(out + 1024 + 448, audio + 1024 + 448, swindow, 128);
206 206 
     memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
... ... 
@@ -213,7 +216,7 @@ WINDOW_FUNC(long_stop)
213 213 
     float *out = sce->ret;
214 214
215 215 
     memset(out, 0, sizeof(out[0]) * 448);
216 
-    dsp->vector_fmul(out + 448, audio + 448, swindow, 128);
216 
+    fdsp->vector_fmul(out + 448, audio + 448, swindow, 128);
217 217 
     memcpy(out + 576, audio + 576, sizeof(out[0]) * 448);
218 218 
     dsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
219 219 
 }
... ... 
@@ -227,7 +230,7 @@ WINDOW_FUNC(eight_short)
227 227 
     int w;
228 228
229 229 
     for (w = 0; w < 8; w++) {
230 
-        dsp->vector_fmul        (out, in, w ? pwindow : swindow, 128);
230 
+        fdsp->vector_fmul       (out, in, w ? pwindow : swindow, 128);
231 231 
         out += 128;
232 232 
         in  += 128;
233 233 
         dsp->vector_fmul_reverse(out, in, swindow, 128);
... ... 
@@ -235,7 +238,9 @@ WINDOW_FUNC(eight_short)
235 235 
     }
236 236 
 }
237 237
238 
-static void (*const apply_window[4])(DSPContext *dsp, SingleChannelElement *sce, const float *audio) = {
238 
+static void (*const apply_window[4])(DSPContext *dsp, AVFloatDSPContext *fdsp,
239 
+                                     SingleChannelElement *sce,
240 
+                                     const float *audio) = {
239 241 
     [ONLY_LONG_SEQUENCE]   = apply_only_long_window,
240 242 
     [LONG_START_SEQUENCE]  = apply_long_start_window,
241 243 
     [EIGHT_SHORT_SEQUENCE] = apply_eight_short_window,
... ... 
@@ -248,7 +253,7 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
248 248 
     int i;
249 249 
     float *output = sce->ret;
250 250
251 
-    apply_window[sce->ics.window_sequence[0]](&s->dsp, sce, audio);
251 
+    apply_window[sce->ics.window_sequence[0]](&s->dsp, &s->fdsp, sce, audio);
252 252
253 253 
     if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE)
254 254 
         s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
... ... 
@@ -693,6 +698,7 @@ static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
693 693 
     int ret = 0;
694 694
695 695 
     ff_dsputil_init(&s->dsp, avctx);
696 
+    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
696 697
697 698 
     // window init
698 699 
     ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
libavcodec/aacenc.h
History View file @ 7e22514
... ... 
@@ -22,6 +22,7 @@
22 22 
 #ifndef AVCODEC_AACENC_H
23 23 
 #define AVCODEC_AACENC_H
24 24
25 
+#include "libavutil/float_dsp.h"
25 26 
 #include "avcodec.h"
26 27 
 #include "put_bits.h"
27 28 
 #include "dsputil.h"
... ... 
@@ -61,6 +62,7 @@ typedef struct AACEncContext {
61 61 
     FFTContext mdct1024;                         ///< long (1024 samples) frame transform context
62 62 
     FFTContext mdct128;                          ///< short (128 samples) frame transform context
63 63 
     DSPContext  dsp;
64 
+    AVFloatDSPContext fdsp;
64 65 
     float *planar_samples[6];                    ///< saved preprocessed input
65 66
66 67 
     int samplerate_index;                        ///< MPEG-4 samplerate index
libavcodec/ac3enc.c
History View file @ 7e22514
... ... 
@@ -2491,6 +2491,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx)
2491 2491 
 #endif
2492 2492
2493 2493 
     ff_dsputil_init(&s->dsp, avctx);
2494 
+    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2494 2495 
     ff_ac3dsp_init(&s->ac3dsp, avctx->flags & CODEC_FLAG_BITEXACT);
2495 2496
2496 2497 
     dprint_options(s);
libavcodec/ac3enc.h
History View file @ 7e22514
... ... 
@@ -29,6 +29,8 @@
29 29 
 #define AVCODEC_AC3ENC_H
30 30
31 31 
 #include <stdint.h>
32 
+
33 
+#include "libavutil/float_dsp.h"
32 34 
 #include "ac3.h"
33 35 
 #include "ac3dsp.h"
34 36 
 #include "avcodec.h"
... ... 
@@ -158,6 +160,7 @@ typedef struct AC3EncodeContext {
158 158 
     AVCodecContext *avctx;                  ///< parent AVCodecContext
159 159 
     PutBitContext pb;                       ///< bitstream writer context
160 160 
     DSPContext dsp;
161 
+    AVFloatDSPContext fdsp;
161 162 
     AC3DSPContext ac3dsp;                   ///< AC-3 optimized functions
162 163 
     FFTContext mdct;                        ///< FFT context for MDCT calculation
163 164 
     const SampleType *mdct_window;          ///< MDCT window function array
libavcodec/ac3enc_fixed.c
History View file @ 7e22514
... ... 
@@ -68,10 +68,11 @@ av_cold int AC3_NAME(mdct_init)(AC3EncodeContext *s)
68 68 
 /*
69 69 
  * Apply KBD window to input samples prior to MDCT.
70 70 
  */
71 
-static void apply_window(DSPContext *dsp, int16_t *output, const int16_t *input,
71 
+static void apply_window(void *dsp, int16_t *output, const int16_t *input,
72 72 
                          const int16_t *window, unsigned int len)
73 73 
 {
74 
-    dsp->apply_window_int16(output, input, window, len);
74 
+    DSPContext *dsp0 = dsp;
75 
+    dsp0->apply_window_int16(output, input, window, len);
75 76 
 }
76 77
77 78
libavcodec/ac3enc_float.c
History View file @ 7e22514
... ... 
@@ -86,10 +86,12 @@ av_cold int ff_ac3_float_mdct_init(AC3EncodeContext *s)
86 86 
 /*
87 87 
  * Apply KBD window to input samples prior to MDCT.
88 88 
  */
89 
-static void apply_window(DSPContext *dsp, float *output, const float *input,
90 
-                         const float *window, unsigned int len)
89 
+static void apply_window(void *dsp, float *output,
90 
+                         const float *input, const float *window,
91 
+                         unsigned int len)
91 92 
 {
92 
-    dsp->vector_fmul(output, input, window, len);
93 
+    AVFloatDSPContext *fdsp = dsp;
94 
+    fdsp->vector_fmul(output, input, window, len);
93 95 
 }
94 96
95 97
libavcodec/ac3enc_template.c
History View file @ 7e22514
... ... 
@@ -33,7 +33,7 @@
33 33
34 34 
 static void scale_coefficients(AC3EncodeContext *s);
35 35
36 
-static void apply_window(DSPContext *dsp, SampleType *output,
36 
+static void apply_window(void *dsp, SampleType *output,
37 37 
                          const SampleType *input, const SampleType *window,
38 38 
                          unsigned int len);
39 39
... ... 
@@ -110,8 +110,13 @@ static void apply_mdct(AC3EncodeContext *s)
110 110 
             AC3Block *block = &s->blocks[blk];
111 111 
             const SampleType *input_samples = &s->planar_samples[ch][blk * AC3_BLOCK_SIZE];
112 112
113 
+#if CONFIG_AC3ENC_FLOAT
114 
+            apply_window(&s->fdsp, s->windowed_samples, input_samples,
115 
+                         s->mdct_window, AC3_WINDOW_SIZE);
116 
+#else
113 117 
             apply_window(&s->dsp, s->windowed_samples, input_samples,
114 118 
                          s->mdct_window, AC3_WINDOW_SIZE);
119 
+#endif
115 120
116 121 
             if (s->fixed_point)
117 122 
                 block->coeff_shift[ch+1] = normalize_samples(s);
libavcodec/arm/aacpsdsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 function ff_ps_add_squares_neon, export=1
24 24 
         mov             r3,  r0
libavcodec/arm/ac3dsp_arm.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 function ff_ac3_update_bap_counts_arm, export=1
24 24 
         push            {lr}
libavcodec/arm/ac3dsp_armv6.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 function ff_ac3_bit_alloc_calc_bap_armv6, export=1
24 24 
         ldr             r12, [sp]
libavcodec/arm/ac3dsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 function ff_ac3_max_msb_abs_int16_neon, export=1
24 24 
         vmov.i16        q0,  #0
libavcodec/arm/asm.S
History View file @ 7e22514
25 25 
deleted file mode 100644
... ... 
@@ -1,231 +0,0 @@
1 
-/*
2 
- * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
3 
- *
4 
- * This file is part of FFmpeg.
5 
- *
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.
10 
- *
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.
15 
- *
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
19 
- */
20 
-
21 
-#include "config.h"
22 
-
23 
-#ifdef __ELF__
24 
-#   define ELF
25 
-#else
26 
-#   define ELF @
27 
-#endif
28 
-
29 
-#if CONFIG_THUMB
30 
-#   define A @
31 
-#   define T
32 
-#else
33 
-#   define A
34 
-#   define T @
35 
-#endif
36 
-
37 
-#if   HAVE_NEON
38 
-        .arch           armv7-a
39 
-#elif HAVE_ARMV6T2
40 
-        .arch           armv6t2
41 
-#elif HAVE_ARMV6
42 
-        .arch           armv6
43 
-#elif HAVE_ARMV5TE
44 
-        .arch           armv5te
45 
-#endif
46 
-
47 
-#if   HAVE_NEON
48 
-        .fpu            neon
49 
-#elif HAVE_ARMVFP
50 
-        .fpu            vfp
51 
-#endif
52 
-
53 
-        .syntax unified
54 
-T       .thumb
55 
-
56 
-.macro  require8 val=1
57 
-ELF     .eabi_attribute 24, \val
58 
-.endm
59 
-
60 
-.macro  preserve8 val=1
61 
-ELF     .eabi_attribute 25, \val
62 
-.endm
63 
-
64 
-.macro  function name, export=0
65 
-    .macro endfunc
66 
-ELF     .size   \name, . - \name
67 
-        .endfunc
68 
-        .purgem endfunc
69 
-    .endm
70 
-        .text
71 
-        .align          2
72 
-    .if \export
73 
-        .global EXTERN_ASM\name
74 
-EXTERN_ASM\name:
75 
-    .endif
76 
-ELF     .type   \name, %function
77 
-        .func   \name
78 
-\name:
79 
-.endm
80 
-
81 
-.macro  const   name, align=2
82 
-    .macro endconst
83 
-ELF     .size   \name, . - \name
84 
-        .purgem endconst
85 
-    .endm
86 
-        .section        .rodata
87 
-        .align          \align
88 
-\name:
89 
-.endm
90 
-
91 
-#if !HAVE_ARMV6T2
92 
-.macro  movw    rd, val
93 
-        mov     \rd, \val &  255
94 
-        orr     \rd, \val & ~255
95 
-.endm
96 
-#endif
97 
-
98 
-.macro  mov32   rd, val
99 
-#if HAVE_ARMV6T2
100 
-        movw            \rd, #(\val) & 0xffff
101 
-    .if (\val) >> 16
102 
-        movt            \rd, #(\val) >> 16
103 
-    .endif
104 
-#else
105 
-        ldr             \rd, =\val
106 
-#endif
107 
-.endm
108 
-
109 
-.macro  movrel rd, val
110 
-#if HAVE_ARMV6T2 && !CONFIG_PIC && !defined(__APPLE__)
111 
-        movw            \rd, #:lower16:\val
112 
-        movt            \rd, #:upper16:\val
113 
-#else
114 
-        ldr             \rd, =\val
115 
-#endif
116 
-.endm
117 
-
118 
-.macro  ldr_pre         rt,  rn,  rm:vararg
119 
-A       ldr             \rt, [\rn, \rm]!
120 
-T       add             \rn, \rn, \rm
121 
-T       ldr             \rt, [\rn]
122 
-.endm
123 
-
124 
-.macro  ldr_dpre        rt,  rn,  rm:vararg
125 
-A       ldr             \rt, [\rn, -\rm]!
126 
-T       sub             \rn, \rn, \rm
127 
-T       ldr             \rt, [\rn]
128 
-.endm
129 
-
130 
-.macro  ldr_nreg        rt,  rn,  rm:vararg
131 
-A       ldr             \rt, [\rn, -\rm]
132 
-T       sub             \rt, \rn, \rm
133 
-T       ldr             \rt, [\rt]
134 
-.endm
135 
-
136 
-.macro  ldr_post        rt,  rn,  rm:vararg
137 
-A       ldr             \rt, [\rn], \rm
138 
-T       ldr             \rt, [\rn]
139 
-T       add             \rn, \rn, \rm
140 
-.endm
141 
-
142 
-.macro  ldrd_reg        rt,  rt2, rn,  rm
143 
-A       ldrd            \rt, \rt2, [\rn, \rm]
144 
-T       add             \rt, \rn, \rm
145 
-T       ldrd            \rt, \rt2, [\rt]
146 
-.endm
147 
-
148 
-.macro  ldrd_post       rt,  rt2, rn,  rm
149 
-A       ldrd            \rt, \rt2, [\rn], \rm
150 
-T       ldrd            \rt, \rt2, [\rn]
151 
-T       add             \rn, \rn, \rm
152 
-.endm
153 
-
154 
-.macro  ldrh_pre        rt,  rn,  rm
155 
-A       ldrh            \rt, [\rn, \rm]!
156 
-T       add             \rn, \rn, \rm
157 
-T       ldrh            \rt, [\rn]
158 
-.endm
159 
-
160 
-.macro  ldrh_dpre       rt,  rn,  rm
161 
-A       ldrh            \rt, [\rn, -\rm]!
162 
-T       sub             \rn, \rn, \rm
163 
-T       ldrh            \rt, [\rn]
164 
-.endm
165 
-
166 
-.macro  ldrh_post       rt,  rn,  rm
167 
-A       ldrh            \rt, [\rn], \rm
168 
-T       ldrh            \rt, [\rn]
169 
-T       add             \rn, \rn, \rm
170 
-.endm
171 
-
172 
-.macro  ldrb_post       rt,  rn,  rm
173 
-A       ldrb            \rt, [\rn], \rm
174 
-T       ldrb            \rt, [\rn]
175 
-T       add             \rn, \rn, \rm
176 
-.endm
177 
-
178 
-.macro  str_post       rt,  rn,  rm:vararg
179 
-A       str             \rt, [\rn], \rm
180 
-T       str             \rt, [\rn]
181 
-T       add             \rn, \rn, \rm
182 
-.endm
183 
-
184 
-.macro  strb_post       rt,  rn,  rm:vararg
185 
-A       strb            \rt, [\rn], \rm
186 
-T       strb            \rt, [\rn]
187 
-T       add             \rn, \rn, \rm
188 
-.endm
189 
-
190 
-.macro  strd_post       rt,  rt2, rn,  rm
191 
-A       strd            \rt, \rt2, [\rn], \rm
192 
-T       strd            \rt, \rt2, [\rn]
193 
-T       add             \rn, \rn, \rm
194 
-.endm
195 
-
196 
-.macro  strh_pre        rt,  rn,  rm
197 
-A       strh            \rt, [\rn, \rm]!
198 
-T       add             \rn, \rn, \rm
199 
-T       strh            \rt, [\rn]
200 
-.endm
201 
-
202 
-.macro  strh_dpre       rt,  rn,  rm
203 
-A       strh            \rt, [\rn, -\rm]!
204 
-T       sub             \rn, \rn, \rm
205 
-T       strh            \rt, [\rn]
206 
-.endm
207 
-
208 
-.macro  strh_post       rt,  rn,  rm
209 
-A       strh            \rt, [\rn], \rm
210 
-T       strh            \rt, [\rn]
211 
-T       add             \rn, \rn, \rm
212 
-.endm
213 
-
214 
-.macro  strh_dpost       rt,  rn,  rm
215 
-A       strh            \rt, [\rn], -\rm
216 
-T       strh            \rt, [\rn]
217 
-T       sub             \rn, \rn, \rm
218 
-.endm
219 
-
220 
-#if HAVE_VFP_ARGS
221 
-        .eabi_attribute 28, 1
222 
-#   define VFP
223 
-#   define NOVFP @
224 
-#else
225 
-#   define VFP   @
226 
-#   define NOVFP
227 
-#endif
228 
-
229 
-#define GLUE(a, b) a ## b
230 
-#define JOIN(a, b) GLUE(a, b)
231 
-#define X(s) JOIN(EXTERN_ASM, s)
libavcodec/arm/dcadsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 function ff_dca_lfe_fir_neon, export=1
24 24 
         push            {r4-r6,lr}
libavcodec/arm/dsputil_arm.S
History View file @ 7e22514
... ... 
@@ -20,7 +20,7 @@
20 20 
 @
21 21
22 22 
 #include "config.h"
23 
-#include "asm.S"
23 
+#include "libavutil/arm/asm.S"
24 24
25 25 
         preserve8
26 26
libavcodec/arm/dsputil_armv6.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
         preserve8
24 24
libavcodec/arm/dsputil_init_neon.c
History View file @ 7e22514
... ... 
@@ -150,7 +150,6 @@ void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
150 150 
 void ff_vp3_v_loop_filter_neon(uint8_t *, int, int *);
151 151 
 void ff_vp3_h_loop_filter_neon(uint8_t *, int, int *);
152 152
153 
-void ff_vector_fmul_neon(float *dst, const float *src0, const float *src1, int len);
154 153 
 void ff_vector_fmul_window_neon(float *dst, const float *src0,
155 154 
                                 const float *src1, const float *win, int len);
156 155 
 void ff_vector_fmul_scalar_neon(float *dst, const float *src, float mul,
... ... 
@@ -328,7 +327,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
328 328 
         c->vp3_idct_dc_add   = ff_vp3_idct_dc_add_neon;
329 329 
     }
330 330
331 
-    c->vector_fmul                = ff_vector_fmul_neon;
332 331 
     c->vector_fmul_window         = ff_vector_fmul_window_neon;
333 332 
     c->vector_fmul_scalar         = ff_vector_fmul_scalar_neon;
334 333 
     c->vector_fmac_scalar         = ff_vector_fmac_scalar_neon;
libavcodec/arm/dsputil_init_vfp.c
History View file @ 7e22514
... ... 
@@ -18,20 +18,13 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "libavutil/arm/cpu.h"
22 21 
 #include "libavcodec/dsputil.h"
23 22 
 #include "dsputil_arm.h"
24 23
25 
-void ff_vector_fmul_vfp(float *dst, const float *src0,
26 
-                        const float *src1, int len);
27 24 
 void ff_vector_fmul_reverse_vfp(float *dst, const float *src0,
28 25 
                                 const float *src1, int len);
29 26
30 27 
 void ff_dsputil_init_vfp(DSPContext* c, AVCodecContext *avctx)
31 28 
 {
32 
-    int cpu_flags = av_get_cpu_flags();
33 
-
34 
-    if (!have_vfpv3(cpu_flags))
35 
-        c->vector_fmul = ff_vector_fmul_vfp;
36 29 
     c->vector_fmul_reverse = ff_vector_fmul_reverse_vfp;
37 30 
 }
libavcodec/arm/dsputil_neon.S
History View file @ 7e22514
... ... 
@@ -20,7 +20,7 @@
20 20 
  */
21 21
22 22 
 #include "config.h"
23 
-#include "asm.S"
23 
+#include "libavutil/arm/asm.S"
24 24
25 25 
         preserve8
26 26
... ... 
@@ -534,45 +534,6 @@ function ff_add_pixels_clamped_neon, export=1
534 534 
         bx              lr
535 535 
 endfunc
536 536
537 
-function ff_vector_fmul_neon, export=1
538 
-        subs            r3,  r3,  #8
539 
-        vld1.32         {d0-d3},  [r1,:128]!
540 
-        vld1.32         {d4-d7},  [r2,:128]!
541 
-        vmul.f32        q8,  q0,  q2
542 
-        vmul.f32        q9,  q1,  q3
543 
-        beq             3f
544 
-        bics            ip,  r3,  #15
545 
-        beq             2f
546 
-1:      subs            ip,  ip,  #16
547 
-        vld1.32         {d0-d1},  [r1,:128]!
548 
-        vld1.32         {d4-d5},  [r2,:128]!
549 
-        vmul.f32        q10, q0,  q2
550 
-        vld1.32         {d2-d3},  [r1,:128]!
551 
-        vld1.32         {d6-d7},  [r2,:128]!
552 
-        vmul.f32        q11, q1,  q3
553 
-        vst1.32         {d16-d19},[r0,:128]!
554 
-        vld1.32         {d0-d1},  [r1,:128]!
555 
-        vld1.32         {d4-d5},  [r2,:128]!
556 
-        vmul.f32        q8,  q0,  q2
557 
-        vld1.32         {d2-d3},  [r1,:128]!
558 
-        vld1.32         {d6-d7},  [r2,:128]!
559 
-        vmul.f32        q9,  q1,  q3
560 
-        vst1.32         {d20-d23},[r0,:128]!
561 
-        bne             1b
562 
-        ands            r3,  r3,  #15
563 
-        beq             3f
564 
-2:      vld1.32         {d0-d1},  [r1,:128]!
565 
-        vld1.32         {d4-d5},  [r2,:128]!
566 
-        vst1.32         {d16-d17},[r0,:128]!
567 
-        vmul.f32        q8,  q0,  q2
568 
-        vld1.32         {d2-d3},  [r1,:128]!
569 
-        vld1.32         {d6-d7},  [r2,:128]!
570 
-        vst1.32         {d18-d19},[r0,:128]!
571 
-        vmul.f32        q9,  q1,  q3
572 
-3:      vst1.32         {d16-d19},[r0,:128]!
573 
-        bx              lr
574 
-endfunc
575 
-
576 537 
 function ff_vector_fmul_window_neon, export=1
577 538 
         push            {r4,r5,lr}
578 539 
         ldr             lr,  [sp, #12]
libavcodec/arm/dsputil_vfp.S
History View file @ 7e22514
... ... 
@@ -19,7 +19,7 @@
19 19 
  */
20 20
21 21 
 #include "config.h"
22 
-#include "asm.S"
22 
+#include "libavutil/arm/asm.S"
23 23
24 24 
 /*
25 25 
  * VFP is a floating point coprocessor used in some ARM cores. VFP11 has 1 cycle
... ... 
@@ -37,53 +37,6 @@
37 37 
  */
38 38
39 39 
 /**
40 
- * ARM VFP optimized implementation of 'vector_fmul_c' function.
41 
- * Assume that len is a positive number and is multiple of 8
42 
- */
43 
-@ void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1, int len)
44 
-function ff_vector_fmul_vfp, export=1
45 
-        vpush           {d8-d15}
46 
-        fmrx            r12, fpscr
47 
-        orr             r12, r12, #(3 << 16) /* set vector size to 4 */
48 
-        fmxr            fpscr, r12
49 
-
50 
-        vldmia          r1!, {s0-s3}
51 
-        vldmia          r2!, {s8-s11}
52 
-        vldmia          r1!, {s4-s7}
53 
-        vldmia          r2!, {s12-s15}
54 
-        vmul.f32        s8,  s0,  s8
55 
-1:
56 
-        subs            r3,  r3,  #16
57 
-        vmul.f32        s12, s4,  s12
58 
-        itttt           ge
59 
-        vldmiage        r1!, {s16-s19}
60 
-        vldmiage        r2!, {s24-s27}
61 
-        vldmiage        r1!, {s20-s23}
62 
-        vldmiage        r2!, {s28-s31}
63 
-        it              ge
64 
-        vmulge.f32      s24, s16, s24
65 
-        vstmia          r0!, {s8-s11}
66 
-        vstmia          r0!, {s12-s15}
67 
-        it              ge
68 
-        vmulge.f32      s28, s20, s28
69 
-        itttt           gt
70 
-        vldmiagt        r1!, {s0-s3}
71 
-        vldmiagt        r2!, {s8-s11}
72 
-        vldmiagt        r1!, {s4-s7}
73 
-        vldmiagt        r2!, {s12-s15}
74 
-        ittt            ge
75 
-        vmulge.f32      s8,  s0,  s8
76 
-        vstmiage        r0!, {s24-s27}
77 
-        vstmiage        r0!, {s28-s31}
78 
-        bgt             1b
79 
-
80 
-        bic             r12, r12, #(7 << 16) /* set vector size back to 1 */
81 
-        fmxr            fpscr, r12
82 
-        vpop            {d8-d15}
83 
-        bx              lr
84 
-endfunc
85 
-
86 
-/**
87 40 
  * ARM VFP optimized implementation of 'vector_fmul_reverse_c' function.
88 41 
  * Assume that len is a positive number and is multiple of 8
89 42 
  */
libavcodec/arm/fft_fixed_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 .macro  bflies          d0,  d1,  r0,  r1
24 24 
         vrev64.32       \r0, \d1                @ t5, t6, t1, t2
libavcodec/arm/fft_neon.S
History View file @ 7e22514
... ... 
@@ -24,7 +24,7 @@
24 24 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 25 
  */
26 26
27 
-#include "asm.S"
27 
+#include "libavutil/arm/asm.S"
28 28
29 29 
 #define M_SQRT1_2 0.70710678118654752440
30 30
libavcodec/arm/fmtconvert_neon.S
History View file @ 7e22514
... ... 
@@ -20,7 +20,7 @@
20 20 
  */
21 21
22 22 
 #include "config.h"
23 
-#include "asm.S"
23 
+#include "libavutil/arm/asm.S"
24 24
25 25 
         preserve8
26 26
libavcodec/arm/fmtconvert_vfp.S
History View file @ 7e22514
... ... 
@@ -19,7 +19,7 @@
19 19 
  */
20 20
21 21 
 #include "config.h"
22 
-#include "asm.S"
22 
+#include "libavutil/arm/asm.S"
23 23
24 24 
 /**
25 25 
  * ARM VFP optimized float to int16 conversion.
libavcodec/arm/h264cmc_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 /* chroma_mc8(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y) */
24 24 
 .macro  h264_chroma_mc8 type, codec=h264
libavcodec/arm/h264dsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22 
 #include "neon.S"
23 23
24 24 
         /* H.264 loop filter */
libavcodec/arm/h264idct_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
         preserve8
24 24
libavcodec/arm/h264pred_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
         .macro ldcol.8  rd,  rs,  rt,  n=8,  hi=0
24 24 
 .if \n == 8 || \hi == 0
libavcodec/arm/int_neon.S
History View file @ 7e22514
... ... 
@@ -19,7 +19,7 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
-#include "asm.S"
22 
+#include "libavutil/arm/asm.S"
23 23
24 24 
         preserve8
25 25 
         .fpu neon
libavcodec/arm/jrevdct_arm.S
History View file @ 7e22514
... ... 
@@ -25,7 +25,7 @@
25 25
26 26 
 */
27 27
28 
-#include "asm.S"
28 
+#include "libavutil/arm/asm.S"
29 29
30 30 
 #define FIX_0_298631336 2446
31 31 
 #define FIX_0_541196100 4433
libavcodec/arm/mdct_fixed_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
         preserve8
24 24
libavcodec/arm/mdct_neon.S
History View file @ 7e22514
... ... 
@@ -19,7 +19,7 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
-#include "asm.S"
22 
+#include "libavutil/arm/asm.S"
23 23
24 24 
         preserve8
25 25
libavcodec/arm/mpegaudiodsp_fixed_armv6.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 .macro  skip            args:vararg
24 24 
 .endm
libavcodec/arm/mpegvideo_armv5te_s.S
History View file @ 7e22514
... ... 
@@ -20,7 +20,7 @@
20 20 
  */
21 21
22 22 
 #include "config.h"
23 
-#include "asm.S"
23 
+#include "libavutil/arm/asm.S"
24 24
25 25 
 /*
26 26 
  * Special optimized version of dct_unquantize_h263_helper_c, it
libavcodec/arm/mpegvideo_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22 
 #include "asm-offsets.h"
23 23
24 24 
 function ff_dct_unquantize_h263_inter_neon, export=1
libavcodec/arm/rdft_neon.S
History View file @ 7e22514
... ... 
@@ -19,7 +19,7 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
-#include "asm.S"
22 
+#include "libavutil/arm/asm.S"
23 23
24 24 
         preserve8
25 25
libavcodec/arm/rv34dsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22 
 #include "neon.S"
23 23
24 24 
 .macro rv34_inv_transform    r0
libavcodec/arm/rv40dsp_neon.S
History View file @ 7e22514
... ... 
@@ -19,7 +19,7 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
-#include "asm.S"
22 
+#include "libavutil/arm/asm.S"
23 23 
 #include "neon.S"
24 24
25 25 
 .macro  qpel_lowpass    r0,  r1,  rc1, rc2, shift
libavcodec/arm/sbrdsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 function ff_sbr_sum64x5_neon, export=1
24 24 
         push            {lr}
libavcodec/arm/simple_idct_arm.S
History View file @ 7e22514
... ... 
@@ -23,7 +23,7 @@
23 23 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 24 
  */
25 25
26 
-#include "asm.S"
26 
+#include "libavutil/arm/asm.S"
27 27
28 28 
 /* useful constants for the algorithm, they are save in __constant_ptr__ at */
29 29 
 /* the end of the source code.*/
libavcodec/arm/simple_idct_armv5te.S
History View file @ 7e22514
... ... 
@@ -21,7 +21,7 @@
21 21 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 22 
  */
23 23
24 
-#include "asm.S"
24 
+#include "libavutil/arm/asm.S"
25 25
26 26 
 #define W1  22725   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
27 27 
 #define W2  21407   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
libavcodec/arm/simple_idct_armv6.S
History View file @ 7e22514
... ... 
@@ -21,7 +21,7 @@
21 21 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 22 
  */
23 23
24 
-#include "asm.S"
24 
+#include "libavutil/arm/asm.S"
25 25
26 26 
 #define W1  22725   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
27 27 
 #define W2  21407   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
libavcodec/arm/simple_idct_neon.S
History View file @ 7e22514
... ... 
@@ -23,7 +23,7 @@
23 23 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 24 
  */
25 25
26 
-#include "asm.S"
26 
+#include "libavutil/arm/asm.S"
27 27
28 28 
 #define W1  22725  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
29 29 
 #define W2  21407  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
libavcodec/arm/synth_filter_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
         preserve8
24 24
libavcodec/arm/vp3dsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 const   vp3_idct_constants, align=4
24 24 
 .short 64277, 60547, 54491, 46341, 36410, 25080, 12785
libavcodec/arm/vp56dsp_neon.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 .macro  vp6_edge_filter
24 24 
         vdup.16         q3,  r2                 @ t
libavcodec/arm/vp8_armv6.S
History View file @ 7e22514
... ... 
@@ -18,7 +18,7 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
-#include "asm.S"
21 
+#include "libavutil/arm/asm.S"
22 22
23 23 
 .macro rac_get_prob     h, bs, buf, cw, pr, t0, t1
24 24 
         adds            \bs, \bs, \t0
libavcodec/arm/vp8dsp_armv6.S
History View file @ 7e22514
... ... 
@@ -52,7 +52,7 @@
52 52 
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
53 53 
  */
54 54
55 
-#include "asm.S"
55 
+#include "libavutil/arm/asm.S"
56 56
57 57 
 @ idct
58 58
libavcodec/arm/vp8dsp_neon.S
History View file @ 7e22514
... ... 
@@ -21,7 +21,7 @@
21 21 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 22 
  */
23 23
24 
-#include "asm.S"
24 
+#include "libavutil/arm/asm.S"
25 25 
 #include "neon.S"
26 26
27 27 
 function ff_vp8_luma_dc_wht_neon, export=1
libavcodec/atrac3.c
History View file @ 7e22514
... ... 
@@ -36,9 +36,9 @@
36 36 
 #include <stddef.h>
37 37 
 #include <stdio.h>
38 38
39 
+#include "libavutil/float_dsp.h"
39 40 
 #include "avcodec.h"
40 41 
 #include "get_bits.h"
41 
-#include "dsputil.h"
42 42 
 #include "bytestream.h"
43 43 
 #include "fft.h"
44 44 
 #include "fmtconvert.h"
... ... 
@@ -125,13 +125,13 @@ typedef struct {
125 125
126 126 
     FFTContext          mdct_ctx;
127 127 
     FmtConvertContext   fmt_conv;
128 
+    AVFloatDSPContext   fdsp;
128 129 
 } ATRAC3Context;
129 130
130 131 
 static DECLARE_ALIGNED(32, float, mdct_window)[MDCT_SIZE];
131 132 
 static VLC              spectral_coeff_tab[7];
132 133 
 static float            gain_tab1[16];
133 134 
 static float            gain_tab2[31];
134 
-static DSPContext       dsp;
135 135
136 136
137 137 
 /**
... ... 
@@ -164,7 +164,7 @@ static void IMLT(ATRAC3Context *q, float *pInput, float *pOutput, int odd_band)
164 164 
     q->mdct_ctx.imdct_calc(&q->mdct_ctx,pOutput,pInput);
165 165
166 166 
     /* Perform windowing on the output. */
167 
-    dsp.vector_fmul(pOutput, pOutput, mdct_window, MDCT_SIZE);
167 
+    q->fdsp.vector_fmul(pOutput, pOutput, mdct_window, MDCT_SIZE);
168 168
169 169 
 }
170 170
... ... 
@@ -1039,7 +1039,7 @@ static av_cold int atrac3_decode_init(AVCodecContext *avctx)
1039 1039 
         q->matrix_coeff_index_next[i] = 3;
1040 1040 
     }
1041 1041
1042 
-    ff_dsputil_init(&dsp, avctx);
1042 
+    avpriv_float_dsp_init(&q->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
1043 1043 
     ff_fmt_convert_init(&q->fmt_conv, avctx);
1044 1044
1045 1045 
     q->pUnits = av_mallocz(sizeof(channel_unit)*q->channels);
libavcodec/dsputil.c
History View file @ 7e22514
... ... 
@@ -2471,12 +2471,6 @@ WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)
2471 2471 
 WRAPPER8_16_SQ(rd8x8_c, rd16_c)
2472 2472 
 WRAPPER8_16_SQ(bit8x8_c, bit16_c)
2473 2473
2474 
-static void vector_fmul_c(float *dst, const float *src0, const float *src1, int len){
2475 
-    int i;
2476 
-    for(i=0; i<len; i++)
2477 
-        dst[i] = src0[i] * src1[i];
2478 
-}
2479 
-
2480 2474 
 static void vector_fmul_reverse_c(float *dst, const float *src0, const float *src1, int len){
2481 2475 
     int i;
2482 2476 
     src1 += len-1;
... ... 
@@ -3054,7 +3048,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
3054 3054 
 #if CONFIG_AC3_DECODER
3055 3055 
     c->ac3_downmix = ff_ac3_downmix_c;
3056 3056 
 #endif
3057 
-    c->vector_fmul = vector_fmul_c;
3058 3057 
     c->vector_fmul_reverse = vector_fmul_reverse_c;
3059 3058 
     c->vector_fmul_add = vector_fmul_add_c;
3060 3059 
     c->vector_fmul_window = vector_fmul_window_c;
libavcodec/dsputil.h
History View file @ 7e22514
... ... 
@@ -403,7 +403,6 @@ typedef struct DSPContext {
403 403 
     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
404 404 
     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
405 405 
     /* assume len is a multiple of 16, and arrays are 32-byte aligned */
406 
-    void (*vector_fmul)(float *dst, const float *src0, const float *src1, int len);
407 406 
     void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
408 407 
     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
409 408 
     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
libavcodec/nellymoserenc.c
History View file @ 7e22514
... ... 
@@ -35,6 +35,7 @@
35 35 
  * http://wiki.multimedia.cx/index.php?title=Nellymoser
36 36 
  */
37 37
38 
+#include "libavutil/float_dsp.h"
38 39 
 #include "libavutil/mathematics.h"
39 40 
 #include "nellymoser.h"
40 41 
 #include "avcodec.h"
... ... 
@@ -55,6 +56,7 @@ typedef struct NellyMoserEncodeContext {
55 55 
     AVCodecContext  *avctx;
56 56 
     int             last_frame;
57 57 
     DSPContext      dsp;
58 
+    AVFloatDSPContext fdsp;
58 59 
     FFTContext      mdct_ctx;
59 60 
     AudioFrameQueue afq;
60 61 
     DECLARE_ALIGNED(32, float, mdct_out)[NELLY_SAMPLES];
... ... 
@@ -120,11 +122,11 @@ static void apply_mdct(NellyMoserEncodeContext *s)
120 120 
     float *in1 = s->buf + NELLY_BUF_LEN;
121 121 
     float *in2 = s->buf + 2 * NELLY_BUF_LEN;
122 122
123 
-    s->dsp.vector_fmul        (s->in_buff,                 in0, ff_sine_128, NELLY_BUF_LEN);
123 
+    s->fdsp.vector_fmul       (s->in_buff,                 in0, ff_sine_128, NELLY_BUF_LEN);
124 124 
     s->dsp.vector_fmul_reverse(s->in_buff + NELLY_BUF_LEN, in1, ff_sine_128, NELLY_BUF_LEN);
125 125 
     s->mdct_ctx.mdct_calc(&s->mdct_ctx, s->mdct_out, s->in_buff);
126 126
127 
-    s->dsp.vector_fmul        (s->in_buff,                 in1, ff_sine_128, NELLY_BUF_LEN);
127 
+    s->fdsp.vector_fmul       (s->in_buff,                 in1, ff_sine_128, NELLY_BUF_LEN);
128 128 
     s->dsp.vector_fmul_reverse(s->in_buff + NELLY_BUF_LEN, in2, ff_sine_128, NELLY_BUF_LEN);
129 129 
     s->mdct_ctx.mdct_calc(&s->mdct_ctx, s->mdct_out + NELLY_BUF_LEN, s->in_buff);
130 130 
 }
... ... 
@@ -172,6 +174,7 @@ static av_cold int encode_init(AVCodecContext *avctx)
172 172 
     if ((ret = ff_mdct_init(&s->mdct_ctx, 8, 0, 32768.0)) < 0)
173 173 
         goto error;
174 174 
     ff_dsputil_init(&s->dsp, avctx);
175 
+    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
175 176
176 177 
     /* Generate overlap window */
177 178 
     ff_init_ff_sine_windows(7);
libavcodec/ppc/dsputil_altivec.c
History View file @ 7e22514
... ... 
@@ -24,9 +24,9 @@
24 24 
 #if HAVE_ALTIVEC_H
25 25 
 #include <altivec.h>
26 26 
 #endif
27 
+#include "libavutil/ppc/types_altivec.h"
28 
+#include "libavutil/ppc/util_altivec.h"
27 29 
 #include "libavcodec/dsputil.h"
28 
-#include "util_altivec.h"
29 
-#include "types_altivec.h"
30 30 
 #include "dsputil_altivec.h"
31 31
32 32 
 static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
libavcodec/ppc/fft_altivec.c
History View file @ 7e22514
... ... 
@@ -19,9 +19,10 @@
19 19 
  * License along with FFmpeg; if not, write to the Free Software
20 20 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 21 
  */
22 
+
23 
+#include "libavutil/ppc/types_altivec.h"
24 
+#include "libavutil/ppc/util_altivec.h"
22 25 
 #include "libavcodec/fft.h"
23 
-#include "util_altivec.h"
24 
-#include "types_altivec.h"
25 26
26 27 
 /**
27 28 
  * Do a complex FFT with the parameters defined in ff_fft_init(). The
libavcodec/ppc/float_altivec.c
History View file @ 7e22514
... ... 
@@ -18,25 +18,10 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
+#include "libavutil/ppc/util_altivec.h"
21 22 
 #include "libavcodec/dsputil.h"
22 23
23 24 
 #include "dsputil_altivec.h"
24 
-#include "util_altivec.h"
25 
-
26 
-static void vector_fmul_altivec(float *dst, const float *src0, const float *src1, int len)
27 
-{
28 
-    int i;
29 
-    vector float d0, d1, s, zero = (vector float)vec_splat_u32(0);
30 
-    for(i=0; i<len-7; i+=8) {
31 
-        d0 = vec_ld(0, src0+i);
32 
-        s = vec_ld(0, src1+i);
33 
-        d1 = vec_ld(16, src0+i);
34 
-        d0 = vec_madd(d0, s, zero);
35 
-        d1 = vec_madd(d1, vec_ld(16,src1+i), zero);
36 
-        vec_st(d0, 0, dst+i);
37 
-        vec_st(d1, 16, dst+i);
38 
-    }
39 
-}
40 25
41 26 
 static void vector_fmul_reverse_altivec(float *dst, const float *src0,
42 27 
                                         const float *src1, int len)
... ... 
@@ -124,7 +109,6 @@ static void vector_fmul_window_altivec(float *dst, const float *src0, const floa
124 124
125 125 
 void ff_float_init_altivec(DSPContext* c, AVCodecContext *avctx)
126 126 
 {
127 
-    c->vector_fmul = vector_fmul_altivec;
128 127 
     c->vector_fmul_reverse = vector_fmul_reverse_altivec;
129 128 
     c->vector_fmul_add = vector_fmul_add_altivec;
130 129 
     if(!(avctx->flags & CODEC_FLAG_BITEXACT)) {
libavcodec/ppc/fmtconvert_altivec.c
History View file @ 7e22514
... ... 
@@ -20,8 +20,8 @@
20 20
21 21 
 #include "libavcodec/fmtconvert.h"
22 22
23 
+#include "libavutil/ppc/util_altivec.h"
23 24 
 #include "dsputil_altivec.h"
24 
-#include "util_altivec.h"
25 25
26 26 
 static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src, float mul, int len)
27 27 
 {
libavcodec/ppc/gmc_altivec.c
History View file @ 7e22514
... ... 
@@ -20,9 +20,9 @@
20 20 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 21 
  */
22 22
23 
+#include "libavutil/ppc/types_altivec.h"
24 
+#include "libavutil/ppc/util_altivec.h"
23 25 
 #include "libavcodec/dsputil.h"
24 
-#include "util_altivec.h"
25 
-#include "types_altivec.h"
26 26 
 #include "dsputil_altivec.h"
27 27
28 28 
 /*
libavcodec/ppc/h264_altivec.c
History View file @ 7e22514
... ... 
@@ -19,13 +19,13 @@
19 19 
  */
20 20
21 21 
 #include "libavutil/cpu.h"
22 
+#include "libavutil/ppc/types_altivec.h"
23 
+#include "libavutil/ppc/util_altivec.h"
22 24 
 #include "libavcodec/dsputil.h"
23 25 
 #include "libavcodec/h264data.h"
24 26 
 #include "libavcodec/h264dsp.h"
25 27
26 28 
 #include "dsputil_altivec.h"
27 
-#include "util_altivec.h"
28 
-#include "types_altivec.h"
29 29
30 30 
 #define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
31 31 
 #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
libavcodec/ppc/idct_altivec.c
History View file @ 7e22514
... ... 
@@ -41,8 +41,8 @@
41 41 
 #if HAVE_ALTIVEC_H
42 42 
 #include <altivec.h>
43 43 
 #endif
44 
+#include "libavutil/ppc/types_altivec.h"
44 45 
 #include "libavcodec/dsputil.h"
45 
-#include "types_altivec.h"
46 46 
 #include "dsputil_altivec.h"
47 47
48 48 
 #define IDCT_HALF                                       \
libavcodec/ppc/int_altivec.c
History View file @ 7e22514
... ... 
@@ -28,12 +28,11 @@
28 28 
 #include <altivec.h>
29 29 
 #endif
30 30
31 
+#include "libavutil/ppc/types_altivec.h"
31 32 
 #include "libavcodec/dsputil.h"
32 33
33 34 
 #include "dsputil_altivec.h"
34 35
35 
-#include "types_altivec.h"
36 
-
37 36 
 static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2,
38 37 
                                      int size) {
39 38 
     int i, size16;
libavcodec/ppc/mpegaudiodec_altivec.c
History View file @ 7e22514
... ... 
@@ -20,7 +20,7 @@
20 20 
  */
21 21
22 22 
 #include "dsputil_altivec.h"
23 
-#include "util_altivec.h"
23 
+#include "libavutil/ppc/util_altivec.h"
24 24 
 #include "libavcodec/dsputil.h"
25 25 
 #include "libavcodec/mpegaudiodsp.h"
26 26
libavcodec/ppc/mpegvideo_altivec.c
History View file @ 7e22514
... ... 
@@ -23,12 +23,13 @@
23 23
24 24 
 #include <stdlib.h>
25 25 
 #include <stdio.h>
26 
+
26 27 
 #include "libavutil/cpu.h"
28 
+#include "libavutil/ppc/types_altivec.h"
29 
+#include "libavutil/ppc/util_altivec.h"
27 30 
 #include "libavcodec/dsputil.h"
28 31 
 #include "libavcodec/mpegvideo.h"
29 32
30 
-#include "util_altivec.h"
31 
-#include "types_altivec.h"
32 33 
 #include "dsputil_altivec.h"
33 34
34 35 
 /* AltiVec version of dct_unquantize_h263
libavcodec/ppc/types_altivec.h
History View file @ 7e22514
35 36 
deleted file mode 100644
... ... 
@@ -1,47 +0,0 @@
1 
-/*
2 
- * Copyright (c) 2006 Guillaume Poirier <gpoirier@mplayerhq.hu>
3 
- *
4 
- * This file is part of FFmpeg.
5 
- *
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.
10 
- *
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.
15 
- *
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
19 
- */
20 
-
21 
-#ifndef AVCODEC_PPC_TYPES_ALTIVEC_H
22 
-#define AVCODEC_PPC_TYPES_ALTIVEC_H
23 
-
24 
-/***********************************************************************
25 
- * Vector types
26 
- **********************************************************************/
27 
-#define vec_u8  vector unsigned char
28 
-#define vec_s8  vector signed char
29 
-#define vec_u16 vector unsigned short
30 
-#define vec_s16 vector signed short
31 
-#define vec_u32 vector unsigned int
32 
-#define vec_s32 vector signed int
33 
-#define vec_f   vector float
34 
-
35 
-/***********************************************************************
36 
- * Null vector
37 
- **********************************************************************/
38 
-#define LOAD_ZERO const vec_u8 zerov = vec_splat_u8( 0 )
39 
-
40 
-#define zero_u8v  (vec_u8)  zerov
41 
-#define zero_s8v  (vec_s8)  zerov
42 
-#define zero_u16v (vec_u16) zerov
43 
-#define zero_s16v (vec_s16) zerov
44 
-#define zero_u32v (vec_u32) zerov
45 
-#define zero_s32v (vec_s32) zerov
46 
-
47 
-#endif /* AVCODEC_PPC_TYPES_ALTIVEC_H */
libavcodec/ppc/util_altivec.h
History View file @ 7e22514
48 1 
deleted file mode 100644
... ... 
@@ -1,118 +0,0 @@
1 
-/*
2 
- * This file is part of FFmpeg.
3 
- *
4 
- * FFmpeg is free software; you can redistribute it and/or
5 
- * modify it under the terms of the GNU Lesser General Public
6 
- * License as published by the Free Software Foundation; either
7 
- * version 2.1 of the License, or (at your option) any later version.
8 
- *
9 
- * FFmpeg is distributed in the hope that it will be useful,
10 
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12 
- * Lesser General Public License for more details.
13 
- *
14 
- * You should have received a copy of the GNU Lesser General Public
15 
- * License along with FFmpeg; if not, write to the Free Software
16 
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 
- */
18 
-
19 
-/**
20 
- * @file
21 
- * Contains misc utility macros and inline functions
22 
- */
23 
-
24 
-#ifndef AVCODEC_PPC_UTIL_ALTIVEC_H
25 
-#define AVCODEC_PPC_UTIL_ALTIVEC_H
26 
-
27 
-#include <stdint.h>
28 
-
29 
-#include "config.h"
30 
-
31 
-#if HAVE_ALTIVEC_H
32 
-#include <altivec.h>
33 
-#endif
34 
-
35 
-#include "types_altivec.h"
36 
-
37 
-// used to build registers permutation vectors (vcprm)
38 
-// the 's' are for words in the _s_econd vector
39 
-#define WORD_0 0x00,0x01,0x02,0x03
40 
-#define WORD_1 0x04,0x05,0x06,0x07
41 
-#define WORD_2 0x08,0x09,0x0a,0x0b
42 
-#define WORD_3 0x0c,0x0d,0x0e,0x0f
43 
-#define WORD_s0 0x10,0x11,0x12,0x13
44 
-#define WORD_s1 0x14,0x15,0x16,0x17
45 
-#define WORD_s2 0x18,0x19,0x1a,0x1b
46 
-#define WORD_s3 0x1c,0x1d,0x1e,0x1f
47 
-
48 
-#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
49 
-#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
50 
-
51 
-// vcprmle is used to keep the same index as in the SSE version.
52 
-// it's the same as vcprm, with the index inversed
53 
-// ('le' is Little Endian)
54 
-#define vcprmle(a,b,c,d) vcprm(d,c,b,a)
55 
-
56 
-// used to build inverse/identity vectors (vcii)
57 
-// n is _n_egative, p is _p_ositive
58 
-#define FLOAT_n -1.
59 
-#define FLOAT_p 1.
60 
-
61 
-
62 
-// Transpose 8x8 matrix of 16-bit elements (in-place)
63 
-#define TRANSPOSE8(a,b,c,d,e,f,g,h) \
64 
-do { \
65 
-    vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \
66 
-    vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \
67 
- \
68 
-    A1 = vec_mergeh (a, e); \
69 
-    B1 = vec_mergel (a, e); \
70 
-    C1 = vec_mergeh (b, f); \
71 
-    D1 = vec_mergel (b, f); \
72 
-    E1 = vec_mergeh (c, g); \
73 
-    F1 = vec_mergel (c, g); \
74 
-    G1 = vec_mergeh (d, h); \
75 
-    H1 = vec_mergel (d, h); \
76 
- \
77 
-    A2 = vec_mergeh (A1, E1); \
78 
-    B2 = vec_mergel (A1, E1); \
79 
-    C2 = vec_mergeh (B1, F1); \
80 
-    D2 = vec_mergel (B1, F1); \
81 
-    E2 = vec_mergeh (C1, G1); \
82 
-    F2 = vec_mergel (C1, G1); \
83 
-    G2 = vec_mergeh (D1, H1); \
84 
-    H2 = vec_mergel (D1, H1); \
85 
- \
86 
-    a = vec_mergeh (A2, E2); \
87 
-    b = vec_mergel (A2, E2); \
88 
-    c = vec_mergeh (B2, F2); \
89 
-    d = vec_mergel (B2, F2); \
90 
-    e = vec_mergeh (C2, G2); \
91 
-    f = vec_mergel (C2, G2); \
92 
-    g = vec_mergeh (D2, H2); \
93 
-    h = vec_mergel (D2, H2); \
94 
-} while (0)
95 
-
96 
-
97 
-/** @brief loads unaligned vector @a *src with offset @a offset
98 
-    and returns it */
99 
-static inline vector unsigned char unaligned_load(int offset, uint8_t *src)
100 
-{
101 
-    register vector unsigned char first = vec_ld(offset, src);
102 
-    register vector unsigned char second = vec_ld(offset+15, src);
103 
-    register vector unsigned char mask = vec_lvsl(offset, src);
104 
-    return vec_perm(first, second, mask);
105 
-}
106 
-
107 
-/**
108 
- * loads vector known misalignment
109 
- * @param perm_vec the align permute vector to combine the two loads from lvsl
110 
- */
111 
-static inline vec_u8 load_with_perm_vec(int offset, uint8_t *src, vec_u8 perm_vec)
112 
-{
113 
-    vec_u8 a = vec_ld(offset, src);
114 
-    vec_u8 b = vec_ld(offset+15, src);
115 
-    return vec_perm(a, b, perm_vec);
116 
-}
117 
-
118 
-#endif /* AVCODEC_PPC_UTIL_ALTIVEC_H */
libavcodec/ppc/vc1dsp_altivec.c
History View file @ 7e22514
... ... 
@@ -19,12 +19,11 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
+#include "libavutil/ppc/types_altivec.h"
23 
+#include "libavutil/ppc/util_altivec.h"
22 24 
 #include "libavcodec/dsputil.h"
23 25 
 #include "libavcodec/vc1dsp.h"
24 26
25 
-#include "util_altivec.h"
26 
-#include "dsputil_altivec.h"
27 
-
28 27 
 // main steps of 8x8 transform
29 28 
 #define STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_rnd) \
30 29 
 do { \
libavcodec/ppc/vp3dsp_altivec.c
History View file @ 7e22514
... ... 
@@ -18,9 +18,9 @@
18 18 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 19 
  */
20 20
21 
+#include "libavutil/ppc/types_altivec.h"
22 
+#include "libavutil/ppc/util_altivec.h"
21 23 
 #include "libavcodec/dsputil.h"
22 
-#include "util_altivec.h"
23 
-#include "types_altivec.h"
24 24 
 #include "dsputil_altivec.h"
25 25
26 26 
 static const vec_s16 constants =
libavcodec/ppc/vp8dsp_altivec.c
History View file @ 7e22514
... ... 
@@ -21,10 +21,10 @@
21 21 
  */
22 22
23 23 
 #include "libavutil/cpu.h"
24 
+#include "libavutil/ppc/types_altivec.h"
25 
+#include "libavutil/ppc/util_altivec.h"
24 26 
 #include "libavcodec/vp8dsp.h"
25 27 
 #include "dsputil_altivec.h"
26 
-#include "types_altivec.h"
27 
-#include "util_altivec.h"
28 28
29 29 
 #define REPT4(...) { __VA_ARGS__, __VA_ARGS__, __VA_ARGS__, __VA_ARGS__ }
30 30
libavcodec/ra288.c
History View file @ 7e22514
... ... 
@@ -19,6 +19,7 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
+#include "libavutil/float_dsp.h"
22 23 
 #include "avcodec.h"
23 24 
 #define BITSTREAM_READER_LE
24 25 
 #include "get_bits.h"
... ... 
@@ -26,7 +27,6 @@
26 26 
 #include "lpc.h"
27 27 
 #include "celp_math.h"
28 28 
 #include "celp_filters.h"
29 
-#include "dsputil.h"
30 29
31 30 
 #define MAX_BACKWARD_FILTER_ORDER  36
32 31 
 #define MAX_BACKWARD_FILTER_LEN    40
... ... 
@@ -38,6 +38,7 @@
38 38 
 typedef struct {
39 39 
     AVFrame frame;
40 40 
     DSPContext dsp;
41 
+    AVFloatDSPContext fdsp;
41 42 
     DECLARE_ALIGNED(32, float,   sp_lpc)[FFALIGN(36, 16)];   ///< LPC coefficients for speech data (spec: A)
42 43 
     DECLARE_ALIGNED(32, float, gain_lpc)[FFALIGN(10, 16)];   ///< LPC coefficients for gain        (spec: GB)
43 44
... ... 
@@ -62,7 +63,7 @@ static av_cold int ra288_decode_init(AVCodecContext *avctx)
62 62 
 {
63 63 
     RA288Context *ractx = avctx->priv_data;
64 64 
     avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
65 
-    ff_dsputil_init(&ractx->dsp, avctx);
65 
+    avpriv_float_dsp_init(&ractx->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
66 66
67 67 
     avcodec_get_frame_defaults(&ractx->frame);
68 68 
     avctx->coded_frame = &ractx->frame;
... ... 
@@ -137,7 +138,7 @@ static void do_hybrid_window(RA288Context *ractx,
137 137 
                                             MAX_BACKWARD_FILTER_LEN   +
138 138 
                                             MAX_BACKWARD_FILTER_NONREC, 16)]);
139 139
140 
-    ractx->dsp.vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16));
140 
+    ractx->fdsp.vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16));
141 141
142 142 
     convolve(buffer1, work + order    , n      , order);
143 143 
     convolve(buffer2, work + order + n, non_rec, order);
... ... 
@@ -164,7 +165,7 @@ static void backward_filter(RA288Context *ractx,
164 164 
     do_hybrid_window(ractx, order, n, non_rec, temp, hist, rec, window);
165 165
166 166 
     if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1))
167 
-        ractx->dsp.vector_fmul(lpc, lpc, tab, FFALIGN(order, 16));
167 
+        ractx->fdsp.vector_fmul(lpc, lpc, tab, FFALIGN(order, 16));
168 168
169 169 
     memmove(hist, hist + n, move_size*sizeof(*hist));
170 170 
 }
libavcodec/twinvq.c
History View file @ 7e22514
... ... 
@@ -19,6 +19,7 @@
19 19 
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 20 
  */
21 21
22 
+#include "libavutil/float_dsp.h"
22 23 
 #include "avcodec.h"
23 24 
 #include "get_bits.h"
24 25 
 #include "dsputil.h"
... ... 
@@ -176,6 +177,7 @@ typedef struct TwinContext {
176 176 
     AVCodecContext *avctx;
177 177 
     AVFrame frame;
178 178 
     DSPContext      dsp;
179 
+    AVFloatDSPContext fdsp;
179 180 
     FFTContext mdct_ctx[3];
180 181
181 182 
     const ModeTab *mtab;
... ... 
@@ -787,8 +789,8 @@ static void read_and_decode_spectrum(TwinContext *tctx, GetBitContext *gb,
787 787 
             dec_bark_env(tctx, bark1[i][j], bark_use_hist[i][j], i,
788 788 
                          tctx->tmp_buf, gain[sub*i+j], ftype);
789 789
790 
-            tctx->dsp.vector_fmul(chunk + block_size*j, chunk + block_size*j, tctx->tmp_buf,
791 
-                                  block_size);
790 
+            tctx->fdsp.vector_fmul(chunk + block_size*j, chunk + block_size*j,
791 
+                                   tctx->tmp_buf, block_size);
792 792
793 793 
         }
794 794
... ... 
@@ -809,7 +811,7 @@ static void read_and_decode_spectrum(TwinContext *tctx, GetBitContext *gb,
809 809 
         dec_lpc_spectrum_inv(tctx, lsp, ftype, tctx->tmp_buf);
810 810
811 811 
         for (j = 0; j < mtab->fmode[ftype].sub; j++) {
812 
-            tctx->dsp.vector_fmul(chunk, chunk, tctx->tmp_buf, block_size);
812 
+            tctx->fdsp.vector_fmul(chunk, chunk, tctx->tmp_buf, block_size);
813 813 
             chunk += block_size;
814 814 
         }
815 815 
     }
... ... 
@@ -1156,6 +1158,7 @@ static av_cold int twin_decode_init(AVCodecContext *avctx)
1156 1156 
     }
1157 1157
1158 1158 
     ff_dsputil_init(&tctx->dsp, avctx);
1159 
+    avpriv_float_dsp_init(&tctx->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
1159 1160 
     if ((ret = init_mdct_win(tctx))) {
1160 1161 
         av_log(avctx, AV_LOG_ERROR, "Error initializing MDCT\n");
1161 1162 
         twin_decode_close(avctx);
libavcodec/vc1dsp.c
History View file @ 7e22514
... ... 
@@ -561,7 +561,7 @@ static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int
561 561 
 /** Function used to do motion compensation with bicubic interpolation
562 562 
  */
563 563 
 #define VC1_MSPEL_MC(OP, OPNAME)\
564 
-static void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
564 
+static av_always_inline void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
565 565 
 {\
566 566 
     int     i, j;\
567 567 
 \
libavcodec/vorbisdec.c
History View file @ 7e22514
... ... 
@@ -30,6 +30,7 @@
30 30 
 #include <math.h>
31 31
32 32 
 #define BITSTREAM_READER_LE
33 
+#include "libavutil/float_dsp.h"
33 34 
 #include "avcodec.h"
34 35 
 #include "get_bits.h"
35 36 
 #include "dsputil.h"
... ... 
@@ -128,6 +129,7 @@ typedef struct vorbis_context_s {
128 128 
     AVFrame frame;
129 129 
     GetBitContext gb;
130 130 
     DSPContext dsp;
131 
+    AVFloatDSPContext fdsp;
131 132 
     FmtConvertContext fmt_conv;
132 133
133 134 
     FFTContext mdct[2];
... ... 
@@ -987,6 +989,7 @@ static av_cold int vorbis_decode_init(AVCodecContext *avccontext)
987 987
988 988 
     vc->avccontext = avccontext;
989 989 
     ff_dsputil_init(&vc->dsp, avccontext);
990 
+    avpriv_float_dsp_init(&vc->fdsp, avccontext->flags & CODEC_FLAG_BITEXACT);
990 991 
     ff_fmt_convert_init(&vc->fmt_conv, avccontext);
991 992
992 993 
     if (avccontext->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
... ... 
@@ -1609,7 +1612,7 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
1609 1609 
     for (j = vc->audio_channels-1;j >= 0; j--) {
1610 1610 
         ch_floor_ptr = vc->channel_floors   + j           * blocksize / 2;
1611 1611 
         ch_res_ptr   = vc->channel_residues + res_chan[j] * blocksize / 2;
1612 
-        vc->dsp.vector_fmul(ch_floor_ptr, ch_floor_ptr, ch_res_ptr, blocksize / 2);
1612 
+        vc->fdsp.vector_fmul(ch_floor_ptr, ch_floor_ptr, ch_res_ptr, blocksize / 2);
1613 1613 
         mdct->imdct_half(mdct, ch_res_ptr, ch_floor_ptr);
1614 1614 
     }
1615 1615
libavcodec/x86/dsputil_mmx.c
History View file @ 7e22514
... ... 
@@ -2581,11 +2581,6 @@ int  ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src,
2581 2581
2582 2582 
 float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order);
2583 2583
2584 
-void ff_vector_fmul_sse(float *dst, const float *src0, const float *src1,
2585 
-                        int len);
2586 
-void ff_vector_fmul_avx(float *dst, const float *src0, const float *src1,
2587 
-                        int len);
2588 
-
2589 2584 
 void ff_vector_fmul_reverse_sse(float *dst, const float *src0,
2590 2585 
                                 const float *src1, int len);
2591 2586 
 void ff_vector_fmul_reverse_avx(float *dst, const float *src0,
... ... 
@@ -2915,7 +2910,6 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags)
2915 2915 
     c->vorbis_inverse_coupling = vorbis_inverse_coupling_sse;
2916 2916 
     c->ac3_downmix             = ac3_downmix_sse;
2917 2917 
 #if HAVE_YASM
2918 
-    c->vector_fmul         = ff_vector_fmul_sse;
2919 2918 
     c->vector_fmul_reverse = ff_vector_fmul_reverse_sse;
2920 2919 
     c->vector_fmul_add     = ff_vector_fmul_add_sse;
2921 2920 
 #endif
... ... 
@@ -3077,7 +3071,6 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
3077 3077 
         }
3078 3078 
     }
3079 3079 
     c->butterflies_float_interleave = ff_butterflies_float_interleave_avx;
3080 
-    c->vector_fmul = ff_vector_fmul_avx;
3081 3080 
     c->vector_fmul_reverse = ff_vector_fmul_reverse_avx;
3082 3081 
     c->vector_fmul_add = ff_vector_fmul_add_avx;
3083 3082 
 #endif
libavcodec/x86/dsputil_yasm.asm
History View file @ 7e22514
... ... 
@@ -1130,38 +1130,6 @@ VECTOR_CLIP_INT32 6, 1, 0, 0
1130 1130 
 %endif
1131 1131
1132 1132 
 ;-----------------------------------------------------------------------------
1133 
-; void vector_fmul(float *dst, const float *src0, const float *src1, int len)
1134 
-;-----------------------------------------------------------------------------
1135 
-%macro VECTOR_FMUL 0
1136 
-cglobal vector_fmul, 4,4,2, dst, src0, src1, len
1137 
-    lea       lenq, [lend*4 - 2*mmsize]
1138 
-ALIGN 16
1139 
-.loop
1140 
-    mova      m0,   [src0q + lenq]
1141 
-    mova      m1,   [src0q + lenq + mmsize]
1142 
-    mulps     m0, m0, [src1q + lenq]
1143 
-    mulps     m1, m1, [src1q + lenq + mmsize]
1144 
-    mova      [dstq + lenq], m0
1145 
-    mova      [dstq + lenq + mmsize], m1
1146 
-
1147 
-    sub       lenq, 2*mmsize
1148 
-    jge       .loop
1149 
-%if mmsize == 32
1150 
-    vzeroupper
1151 
-    RET
1152 
-%else
1153 
-    REP_RET
1154 
-%endif
1155 
-%endmacro
1156 
-
1157 
-INIT_XMM sse
1158 
-VECTOR_FMUL
1159 
-%if HAVE_AVX
1160 
-INIT_YMM avx
1161 
-VECTOR_FMUL
1162 
-%endif
1163 
-
1164 
-;-----------------------------------------------------------------------------
1165 1133 
 ; void vector_fmul_reverse(float *dst, const float *src0, const float *src1,
1166 1134 
 ;                          int len)
1167 1135 
 ;-----------------------------------------------------------------------------
libavutil/Makefile
History View file @ 7e22514
... ... 
@@ -63,6 +63,7 @@ OBJS = adler32.o                                                        \
63 63 
        eval.o                                                           \
64 64 
        fifo.o                                                           \
65 65 
        file.o                                                           \
66 
+       float_dsp.o                                                      \
66 67 
        imgutils.o                                                       \
67 68 
        intfloat_readwrite.o                                             \
68 69 
        inverse.o                                                        \
libavutil/arm/Makefile
History View file @ 7e22514
... ... 
@@ -1 +1,8 @@
1 1 
 OBJS += arm/cpu.o                                                       \
2 
+        arm/float_dsp_init_arm.o                                        \
3 
+
4 
+ARMVFP-OBJS += arm/float_dsp_init_vfp.o                                 \
5 
+               arm/float_dsp_vfp.o                                      \
6 
+
7 
+NEON-OBJS += arm/float_dsp_init_neon.o                                  \
8 
+             arm/float_dsp_neon.o                                       \
libavutil/arm/asm.S
History View file @ 7e22514
2 9 
new file mode 100644
... ... 
@@ -0,0 +1,231 @@
0 
+/*
1 
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "config.h"
21 
+
22 
+#ifdef __ELF__
23 
+#   define ELF
24 
+#else
25 
+#   define ELF @
26 
+#endif
27 
+
28 
+#if CONFIG_THUMB
29 
+#   define A @
30 
+#   define T
31 
+#else
32 
+#   define A
33 
+#   define T @
34 
+#endif
35 
+
36 
+#if   HAVE_NEON
37 
+        .arch           armv7-a
38 
+#elif HAVE_ARMV6T2
39 
+        .arch           armv6t2
40 
+#elif HAVE_ARMV6
41 
+        .arch           armv6
42 
+#elif HAVE_ARMV5TE
43 
+        .arch           armv5te
44 
+#endif
45 
+
46 
+#if   HAVE_NEON
47 
+        .fpu            neon
48 
+#elif HAVE_ARMVFP
49 
+        .fpu            vfp
50 
+#endif
51 
+
52 
+        .syntax unified
53 
+T       .thumb
54 
+
55 
+.macro  require8 val=1
56 
+ELF     .eabi_attribute 24, \val
57 
+.endm
58 
+
59 
+.macro  preserve8 val=1
60 
+ELF     .eabi_attribute 25, \val
61 
+.endm
62 
+
63 
+.macro  function name, export=0
64 
+    .macro endfunc
65 
+ELF     .size   \name, . - \name
66 
+        .endfunc
67 
+        .purgem endfunc
68 
+    .endm
69 
+        .text
70 
+        .align          2
71 
+    .if \export
72 
+        .global EXTERN_ASM\name
73 
+EXTERN_ASM\name:
74 
+    .endif
75 
+ELF     .type   \name, %function
76 
+        .func   \name
77 
+\name:
78 
+.endm
79 
+
80 
+.macro  const   name, align=2
81 
+    .macro endconst
82 
+ELF     .size   \name, . - \name
83 
+        .purgem endconst
84 
+    .endm
85 
+        .section        .rodata
86 
+        .align          \align
87 
+\name:
88 
+.endm
89 
+
90 
+#if !HAVE_ARMV6T2
91 
+.macro  movw    rd, val
92 
+        mov     \rd, \val &  255
93 
+        orr     \rd, \val & ~255
94 
+.endm
95 
+#endif
96 
+
97 
+.macro  mov32   rd, val
98 
+#if HAVE_ARMV6T2
99 
+        movw            \rd, #(\val) & 0xffff
100 
+    .if (\val) >> 16
101 
+        movt            \rd, #(\val) >> 16
102 
+    .endif
103 
+#else
104 
+        ldr             \rd, =\val
105 
+#endif
106 
+.endm
107 
+
108 
+.macro  movrel rd, val
109 
+#if HAVE_ARMV6T2 && !CONFIG_PIC && !defined(__APPLE__)
110 
+        movw            \rd, #:lower16:\val
111 
+        movt            \rd, #:upper16:\val
112 
+#else
113 
+        ldr             \rd, =\val
114 
+#endif
115 
+.endm
116 
+
117 
+.macro  ldr_pre         rt,  rn,  rm:vararg
118 
+A       ldr             \rt, [\rn, \rm]!
119 
+T       add             \rn, \rn, \rm
120 
+T       ldr             \rt, [\rn]
121 
+.endm
122 
+
123 
+.macro  ldr_dpre        rt,  rn,  rm:vararg
124 
+A       ldr             \rt, [\rn, -\rm]!
125 
+T       sub             \rn, \rn, \rm
126 
+T       ldr             \rt, [\rn]
127 
+.endm
128 
+
129 
+.macro  ldr_nreg        rt,  rn,  rm:vararg
130 
+A       ldr             \rt, [\rn, -\rm]
131 
+T       sub             \rt, \rn, \rm
132 
+T       ldr             \rt, [\rt]
133 
+.endm
134 
+
135 
+.macro  ldr_post        rt,  rn,  rm:vararg
136 
+A       ldr             \rt, [\rn], \rm
137 
+T       ldr             \rt, [\rn]
138 
+T       add             \rn, \rn, \rm
139 
+.endm
140 
+
141 
+.macro  ldrd_reg        rt,  rt2, rn,  rm
142 
+A       ldrd            \rt, \rt2, [\rn, \rm]
143 
+T       add             \rt, \rn, \rm
144 
+T       ldrd            \rt, \rt2, [\rt]
145 
+.endm
146 
+
147 
+.macro  ldrd_post       rt,  rt2, rn,  rm
148 
+A       ldrd            \rt, \rt2, [\rn], \rm
149 
+T       ldrd            \rt, \rt2, [\rn]
150 
+T       add             \rn, \rn, \rm
151 
+.endm
152 
+
153 
+.macro  ldrh_pre        rt,  rn,  rm
154 
+A       ldrh            \rt, [\rn, \rm]!
155 
+T       add             \rn, \rn, \rm
156 
+T       ldrh            \rt, [\rn]
157 
+.endm
158 
+
159 
+.macro  ldrh_dpre       rt,  rn,  rm
160 
+A       ldrh            \rt, [\rn, -\rm]!
161 
+T       sub             \rn, \rn, \rm
162 
+T       ldrh            \rt, [\rn]
163 
+.endm
164 
+
165 
+.macro  ldrh_post       rt,  rn,  rm
166 
+A       ldrh            \rt, [\rn], \rm
167 
+T       ldrh            \rt, [\rn]
168 
+T       add             \rn, \rn, \rm
169 
+.endm
170 
+
171 
+.macro  ldrb_post       rt,  rn,  rm
172 
+A       ldrb            \rt, [\rn], \rm
173 
+T       ldrb            \rt, [\rn]
174 
+T       add             \rn, \rn, \rm
175 
+.endm
176 
+
177 
+.macro  str_post       rt,  rn,  rm:vararg
178 
+A       str             \rt, [\rn], \rm
179 
+T       str             \rt, [\rn]
180 
+T       add             \rn, \rn, \rm
181 
+.endm
182 
+
183 
+.macro  strb_post       rt,  rn,  rm:vararg
184 
+A       strb            \rt, [\rn], \rm
185 
+T       strb            \rt, [\rn]
186 
+T       add             \rn, \rn, \rm
187 
+.endm
188 
+
189 
+.macro  strd_post       rt,  rt2, rn,  rm
190 
+A       strd            \rt, \rt2, [\rn], \rm
191 
+T       strd            \rt, \rt2, [\rn]
192 
+T       add             \rn, \rn, \rm
193 
+.endm
194 
+
195 
+.macro  strh_pre        rt,  rn,  rm
196 
+A       strh            \rt, [\rn, \rm]!
197 
+T       add             \rn, \rn, \rm
198 
+T       strh            \rt, [\rn]
199 
+.endm
200 
+
201 
+.macro  strh_dpre       rt,  rn,  rm
202 
+A       strh            \rt, [\rn, -\rm]!
203 
+T       sub             \rn, \rn, \rm
204 
+T       strh            \rt, [\rn]
205 
+.endm
206 
+
207 
+.macro  strh_post       rt,  rn,  rm
208 
+A       strh            \rt, [\rn], \rm
209 
+T       strh            \rt, [\rn]
210 
+T       add             \rn, \rn, \rm
211 
+.endm
212 
+
213 
+.macro  strh_dpost       rt,  rn,  rm
214 
+A       strh            \rt, [\rn], -\rm
215 
+T       strh            \rt, [\rn]
216 
+T       sub             \rn, \rn, \rm
217 
+.endm
218 
+
219 
+#if HAVE_VFP_ARGS
220 
+        .eabi_attribute 28, 1
221 
+#   define VFP
222 
+#   define NOVFP @
223 
+#else
224 
+#   define VFP   @
225 
+#   define NOVFP
226 
+#endif
227 
+
228 
+#define GLUE(a, b) a ## b
229 
+#define JOIN(a, b) GLUE(a, b)
230 
+#define X(s) JOIN(EXTERN_ASM, s)
libavutil/arm/float_dsp_arm.h
History View file @ 7e22514
0 231 
new file mode 100644
... ... 
@@ -0,0 +1,29 @@
0 
+/*
1 
+ * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#ifndef AVUTIL_ARM_FLOAT_DSP_ARM_H
21 
+#define AVUTIL_ARM_FLOAT_DSP_ARM_H
22 
+
23 
+#include "libavutil/float_dsp.h"
24 
+
25 
+void ff_float_dsp_init_vfp (AVFloatDSPContext *fdsp);
26 
+void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp);
27 
+
28 
+#endif /* AVUTIL_ARM_FLOAT_DSP_ARM_H */
libavutil/arm/float_dsp_init_arm.c
History View file @ 7e22514
0 29 
new file mode 100644
... ... 
@@ -0,0 +1,33 @@
0 
+/*
1 
+ * ARM optimized DSP utils
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "libavutil/arm/cpu.h"
21 
+#include "libavutil/float_dsp.h"
22 
+#include "float_dsp_arm.h"
23 
+
24 
+void ff_float_dsp_init_arm(AVFloatDSPContext *fdsp)
25 
+{
26 
+    int cpu_flags = av_get_cpu_flags();
27 
+
28 
+    if (have_vfp(cpu_flags))
29 
+        ff_float_dsp_init_vfp(fdsp);
30 
+    if (have_neon(cpu_flags))
31 
+        ff_float_dsp_init_neon(fdsp);
32 
+}
libavutil/arm/float_dsp_init_neon.c
History View file @ 7e22514
0 33 
new file mode 100644
... ... 
@@ -0,0 +1,32 @@
0 
+/*
1 
+ * ARM NEON optimised Float DSP functions
2 
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
3 
+ *
4 
+ * This file is part of Libav.
5 
+ *
6 
+ * Libav 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.
10 
+ *
11 
+ * Libav 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.
15 
+ *
16 
+ * You should have received a copy of the GNU Lesser General Public
17 
+ * License along with Libav; if not, write to the Free Software
18 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 
+ */
20 
+
21 
+#include <stdint.h>
22 
+
23 
+#include "libavutil/float_dsp.h"
24 
+#include "float_dsp_arm.h"
25 
+
26 
+void ff_vector_fmul_neon(float *dst, const float *src0, const float *src1, int len);
27 
+
28 
+void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp)
29 
+{
30 
+    fdsp->vector_fmul = ff_vector_fmul_neon;
31 
+}
libavutil/arm/float_dsp_init_vfp.c
History View file @ 7e22514
0 32 
new file mode 100644
... ... 
@@ -0,0 +1,34 @@
0 
+/*
1 
+ * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "libavutil/arm/cpu.h"
21 
+#include "libavutil/float_dsp.h"
22 
+#include "float_dsp_arm.h"
23 
+
24 
+void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1,
25 
+                        int len);
26 
+
27 
+void ff_float_dsp_init_vfp(AVFloatDSPContext *fdsp)
28 
+{
29 
+    int cpu_flags = av_get_cpu_flags();
30 
+
31 
+    if (!have_vfpv3(cpu_flags))
32 
+        fdsp->vector_fmul = ff_vector_fmul_vfp;
33 
+}
libavutil/arm/float_dsp_neon.S
History View file @ 7e22514
0 34 
new file mode 100644
... ... 
@@ -0,0 +1,64 @@
0 
+/*
1 
+ * ARM NEON optimised Float DSP functions
2 
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
3 
+ *
4 
+ * This file is part of Libav.
5 
+ *
6 
+ * Libav 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.
10 
+ *
11 
+ * Libav 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.
15 
+ *
16 
+ * You should have received a copy of the GNU Lesser General Public
17 
+ * License along with Libav; if not, write to the Free Software
18 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 
+ */
20 
+
21 
+#include "config.h"
22 
+#include "asm.S"
23 
+
24 
+        preserve8
25 
+
26 
+function ff_vector_fmul_neon, export=1
27 
+        subs            r3,  r3,  #8
28 
+        vld1.32         {d0-d3},  [r1,:128]!
29 
+        vld1.32         {d4-d7},  [r2,:128]!
30 
+        vmul.f32        q8,  q0,  q2
31 
+        vmul.f32        q9,  q1,  q3
32 
+        beq             3f
33 
+        bics            ip,  r3,  #15
34 
+        beq             2f
35 
+1:      subs            ip,  ip,  #16
36 
+        vld1.32         {d0-d1},  [r1,:128]!
37 
+        vld1.32         {d4-d5},  [r2,:128]!
38 
+        vmul.f32        q10, q0,  q2
39 
+        vld1.32         {d2-d3},  [r1,:128]!
40 
+        vld1.32         {d6-d7},  [r2,:128]!
41 
+        vmul.f32        q11, q1,  q3
42 
+        vst1.32         {d16-d19},[r0,:128]!
43 
+        vld1.32         {d0-d1},  [r1,:128]!
44 
+        vld1.32         {d4-d5},  [r2,:128]!
45 
+        vmul.f32        q8,  q0,  q2
46 
+        vld1.32         {d2-d3},  [r1,:128]!
47 
+        vld1.32         {d6-d7},  [r2,:128]!
48 
+        vmul.f32        q9,  q1,  q3
49 
+        vst1.32         {d20-d23},[r0,:128]!
50 
+        bne             1b
51 
+        ands            r3,  r3,  #15
52 
+        beq             3f
53 
+2:      vld1.32         {d0-d1},  [r1,:128]!
54 
+        vld1.32         {d4-d5},  [r2,:128]!
55 
+        vst1.32         {d16-d17},[r0,:128]!
56 
+        vmul.f32        q8,  q0,  q2
57 
+        vld1.32         {d2-d3},  [r1,:128]!
58 
+        vld1.32         {d6-d7},  [r2,:128]!
59 
+        vst1.32         {d18-d19},[r0,:128]!
60 
+        vmul.f32        q9,  q1,  q3
61 
+3:      vst1.32         {d16-d19},[r0,:128]!
62 
+        bx              lr
63 
+endfunc
libavutil/arm/float_dsp_vfp.S
History View file @ 7e22514
0 64 
new file mode 100644
... ... 
@@ -0,0 +1,68 @@
0 
+/*
1 
+ * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net>
2 
+ *
3 
+ * This file is part of FFmpeg
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "config.h"
21 
+#include "asm.S"
22 
+
23 
+/**
24 
+ * Assume that len is a positive number and is multiple of 8
25 
+ */
26 
+@ void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1, int len)
27 
+function ff_vector_fmul_vfp, export=1
28 
+        vpush           {d8-d15}
29 
+        fmrx            r12, fpscr
30 
+        orr             r12, r12, #(3 << 16) /* set vector size to 4 */
31 
+        fmxr            fpscr, r12
32 
+
33 
+        vldmia          r1!, {s0-s3}
34 
+        vldmia          r2!, {s8-s11}
35 
+        vldmia          r1!, {s4-s7}
36 
+        vldmia          r2!, {s12-s15}
37 
+        vmul.f32        s8,  s0,  s8
38 
+1:
39 
+        subs            r3,  r3,  #16
40 
+        vmul.f32        s12, s4,  s12
41 
+        itttt           ge
42 
+        vldmiage        r1!, {s16-s19}
43 
+        vldmiage        r2!, {s24-s27}
44 
+        vldmiage        r1!, {s20-s23}
45 
+        vldmiage        r2!, {s28-s31}
46 
+        it              ge
47 
+        vmulge.f32      s24, s16, s24
48 
+        vstmia          r0!, {s8-s11}
49 
+        vstmia          r0!, {s12-s15}
50 
+        it              ge
51 
+        vmulge.f32      s28, s20, s28
52 
+        itttt           gt
53 
+        vldmiagt        r1!, {s0-s3}
54 
+        vldmiagt        r2!, {s8-s11}
55 
+        vldmiagt        r1!, {s4-s7}
56 
+        vldmiagt        r2!, {s12-s15}
57 
+        ittt            ge
58 
+        vmulge.f32      s8,  s0,  s8
59 
+        vstmiage        r0!, {s24-s27}
60 
+        vstmiage        r0!, {s28-s31}
61 
+        bgt             1b
62 
+
63 
+        bic             r12, r12, #(7 << 16) /* set vector size back to 1 */
64 
+        fmxr            fpscr, r12
65 
+        vpop            {d8-d15}
66 
+        bx              lr
67 
+endfunc
libavutil/float_dsp.c
History View file @ 7e22514
0 68 
new file mode 100644
... ... 
@@ -0,0 +1,42 @@
0 
+/*
1 
+ * This file is part of Libav.
2 
+ *
3 
+ * Libav is free software; you can redistribute it and/or
4 
+ * modify it under the terms of the GNU Lesser General Public
5 
+ * License as published by the Free Software Foundation; either
6 
+ * version 2.1 of the License, or (at your option) any later version.
7 
+ *
8 
+ * Libav is distributed in the hope that it will be useful,
9 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11 
+ * Lesser General Public License for more details.
12 
+ *
13 
+ * You should have received a copy of the GNU Lesser General Public
14 
+ * License along with Libav; if not, write to the Free Software
15 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16 
+ */
17 
+
18 
+#include "config.h"
19 
+
20 
+#include "float_dsp.h"
21 
+
22 
+static void vector_fmul_c(float *dst, const float *src0, const float *src1,
23 
+                          int len)
24 
+{
25 
+    int i;
26 
+    for (i = 0; i < len; i++)
27 
+        dst[i] = src0[i] * src1[i];
28 
+}
29 
+
30 
+void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int bit_exact)
31 
+{
32 
+    fdsp->vector_fmul = vector_fmul_c;
33 
+
34 
+#if ARCH_ARM
35 
+    ff_float_dsp_init_arm(fdsp);
36 
+#elif ARCH_PPC
37 
+    ff_float_dsp_init_ppc(fdsp, bit_exact);
38 
+#elif ARCH_X86
39 
+    ff_float_dsp_init_x86(fdsp);
40 
+#endif
41 
+}
libavutil/float_dsp.h
History View file @ 7e22514
0 42 
new file mode 100644
... ... 
@@ -0,0 +1,53 @@
0 
+/*
1 
+ * This file is part of FFmpeg.
2 
+ *
3 
+ * FFmpeg is free software; you can redistribute it and/or
4 
+ * modify it under the terms of the GNU Lesser General Public
5 
+ * License as published by the Free Software Foundation; either
6 
+ * version 2.1 of the License, or (at your option) any later version.
7 
+ *
8 
+ * FFmpeg is distributed in the hope that it will be useful,
9 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11 
+ * Lesser General Public License for more details.
12 
+ *
13 
+ * You should have received a copy of the GNU Lesser General Public
14 
+ * License along with FFmpeg; if not, write to the Free Software
15 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16 
+ */
17 
+
18 
+#ifndef AVUTIL_FLOAT_DSP_H
19 
+#define AVUTIL_FLOAT_DSP_H
20 
+
21 
+typedef struct AVFloatDSPContext {
22 
+    /**
23 
+     * Calculate the product of two vectors of floats and store the result in
24 
+     * a vector of floats.
25 
+     *
26 
+     * @param dst  output vector
27 
+     *             constraints: 32-byte aligned
28 
+     * @param src0 first input vector
29 
+     *             constraints: 32-byte aligned
30 
+     * @param src1 second input vector
31 
+     *             constraints: 32-byte aligned
32 
+     * @param len  number of elements in the input
33 
+     *             constraints: multiple of 16
34 
+     */
35 
+    void (*vector_fmul)(float *dst, const float *src0, const float *src1,
36 
+                        int len);
37 
+} AVFloatDSPContext;
38 
+
39 
+/**
40 
+ * Initialize a float DSP context.
41 
+ *
42 
+ * @param fdsp    float DSP context
43 
+ * @param strict  setting to non-zero avoids using functions which may not be IEEE-754 compliant
44 
+ */
45 
+void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int strict);
46 
+
47 
+
48 
+void ff_float_dsp_init_arm(AVFloatDSPContext *fdsp);
49 
+void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int strict);
50 
+void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp);
51 
+
52 
+#endif /* AVUTIL_FLOAT_DSP_H */
libavutil/ppc/Makefile
History View file @ 7e22514
... ... 
@@ -1 +1,4 @@
1 1 
 OBJS += ppc/cpu.o                                                       \
2 
+        ppc/float_dsp_init.o                                            \
3 
+
4 
+ALTIVEC-OBJS += ppc/float_dsp_altivec.o                                 \
libavutil/ppc/float_dsp_altivec.c
History View file @ 7e22514
2 5 
new file mode 100644
... ... 
@@ -0,0 +1,38 @@
0 
+/*
1 
+ * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "util_altivec.h"
21 
+#include "float_dsp_altivec.h"
22 
+
23 
+void ff_vector_fmul_altivec(float *dst, const float *src0, const float *src1,
24 
+                            int len)
25 
+{
26 
+    int i;
27 
+    vector float d0, d1, s, zero = (vector float)vec_splat_u32(0);
28 
+    for (i = 0; i < len - 7; i += 8) {
29 
+        d0 = vec_ld( 0, src0 + i);
30 
+        s  = vec_ld( 0, src1 + i);
31 
+        d1 = vec_ld(16, src0 + i);
32 
+        d0 = vec_madd(d0, s, zero);
33 
+        d1 = vec_madd(d1, vec_ld(16, src1 + i), zero);
34 
+        vec_st(d0,  0, dst + i);
35 
+        vec_st(d1, 16, dst + i);
36 
+    }
37 
+}
libavutil/ppc/float_dsp_altivec.h
History View file @ 7e22514
0 38 
new file mode 100644
... ... 
@@ -0,0 +1,27 @@
0 
+/*
1 
+ * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#ifndef AVUTIL_PPC_FLOAT_DSP_ALTIVEC_H
21 
+#define AVUTIL_PPC_FLOAT_DSP_ALTIVEC_H
22 
+
23 
+extern void ff_vector_fmul_altivec(float *dst, const float *src0,
24 
+                                   const float *src1, int len);
25 
+
26 
+#endif /* AVUTIL_PPC_FLOAT_DSP_ALTIVEC_H */
libavutil/ppc/float_dsp_init.c
History View file @ 7e22514
0 27 
new file mode 100644
... ... 
@@ -0,0 +1,36 @@
0 
+/*
1 
+ * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "config.h"
21 
+#include "libavutil/cpu.h"
22 
+#include "libavutil/float_dsp.h"
23 
+#include "float_dsp_altivec.h"
24 
+
25 
+void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int bit_exact)
26 
+{
27 
+#if HAVE_ALTIVEC
28 
+    int mm_flags = av_get_cpu_flags();
29 
+
30 
+    if (!(mm_flags & AV_CPU_FLAG_ALTIVEC))
31 
+        return;
32 
+
33 
+    fdsp->vector_fmul = ff_vector_fmul_altivec;
34 
+#endif
35 
+}
libavutil/ppc/types_altivec.h
History View file @ 7e22514
0 36 
new file mode 100644
... ... 
@@ -0,0 +1,47 @@
0 
+/*
1 
+ * Copyright (c) 2006 Guillaume Poirier <gpoirier@mplayerhq.hu>
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#ifndef AVUTIL_PPC_TYPES_ALTIVEC_H
21 
+#define AVUTIL_PPC_TYPES_ALTIVEC_H
22 
+
23 
+/***********************************************************************
24 
+ * Vector types
25 
+ **********************************************************************/
26 
+#define vec_u8  vector unsigned char
27 
+#define vec_s8  vector signed char
28 
+#define vec_u16 vector unsigned short
29 
+#define vec_s16 vector signed short
30 
+#define vec_u32 vector unsigned int
31 
+#define vec_s32 vector signed int
32 
+#define vec_f   vector float
33 
+
34 
+/***********************************************************************
35 
+ * Null vector
36 
+ **********************************************************************/
37 
+#define LOAD_ZERO const vec_u8 zerov = vec_splat_u8( 0 )
38 
+
39 
+#define zero_u8v  (vec_u8)  zerov
40 
+#define zero_s8v  (vec_s8)  zerov
41 
+#define zero_u16v (vec_u16) zerov
42 
+#define zero_s16v (vec_s16) zerov
43 
+#define zero_u32v (vec_u32) zerov
44 
+#define zero_s32v (vec_s32) zerov
45 
+
46 
+#endif /* AVUTIL_PPC_TYPES_ALTIVEC_H */
libavutil/ppc/util_altivec.h
History View file @ 7e22514
0 47 
new file mode 100644
... ... 
@@ -0,0 +1,118 @@
0 
+/*
1 
+ * This file is part of FFmpeg.
2 
+ *
3 
+ * FFmpeg is free software; you can redistribute it and/or
4 
+ * modify it under the terms of the GNU Lesser General Public
5 
+ * License as published by the Free Software Foundation; either
6 
+ * version 2.1 of the License, or (at your option) any later version.
7 
+ *
8 
+ * FFmpeg is distributed in the hope that it will be useful,
9 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11 
+ * Lesser General Public License for more details.
12 
+ *
13 
+ * You should have received a copy of the GNU Lesser General Public
14 
+ * License along with FFmpeg; if not, write to the Free Software
15 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16 
+ */
17 
+
18 
+/**
19 
+ * @file
20 
+ * Contains misc utility macros and inline functions
21 
+ */
22 
+
23 
+#ifndef AVUTIL_PPC_UTIL_ALTIVEC_H
24 
+#define AVUTIL_PPC_UTIL_ALTIVEC_H
25 
+
26 
+#include <stdint.h>
27 
+
28 
+#include "config.h"
29 
+
30 
+#if HAVE_ALTIVEC_H
31 
+#include <altivec.h>
32 
+#endif
33 
+
34 
+#include "types_altivec.h"
35 
+
36 
+// used to build registers permutation vectors (vcprm)
37 
+// the 's' are for words in the _s_econd vector
38 
+#define WORD_0 0x00,0x01,0x02,0x03
39 
+#define WORD_1 0x04,0x05,0x06,0x07
40 
+#define WORD_2 0x08,0x09,0x0a,0x0b
41 
+#define WORD_3 0x0c,0x0d,0x0e,0x0f
42 
+#define WORD_s0 0x10,0x11,0x12,0x13
43 
+#define WORD_s1 0x14,0x15,0x16,0x17
44 
+#define WORD_s2 0x18,0x19,0x1a,0x1b
45 
+#define WORD_s3 0x1c,0x1d,0x1e,0x1f
46 
+
47 
+#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
48 
+#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
49 
+
50 
+// vcprmle is used to keep the same index as in the SSE version.
51 
+// it's the same as vcprm, with the index inversed
52 
+// ('le' is Little Endian)
53 
+#define vcprmle(a,b,c,d) vcprm(d,c,b,a)
54 
+
55 
+// used to build inverse/identity vectors (vcii)
56 
+// n is _n_egative, p is _p_ositive
57 
+#define FLOAT_n -1.
58 
+#define FLOAT_p 1.
59 
+
60 
+
61 
+// Transpose 8x8 matrix of 16-bit elements (in-place)
62 
+#define TRANSPOSE8(a,b,c,d,e,f,g,h) \
63 
+do { \
64 
+    vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \
65 
+    vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \
66 
+ \
67 
+    A1 = vec_mergeh (a, e); \
68 
+    B1 = vec_mergel (a, e); \
69 
+    C1 = vec_mergeh (b, f); \
70 
+    D1 = vec_mergel (b, f); \
71 
+    E1 = vec_mergeh (c, g); \
72 
+    F1 = vec_mergel (c, g); \
73 
+    G1 = vec_mergeh (d, h); \
74 
+    H1 = vec_mergel (d, h); \
75 
+ \
76 
+    A2 = vec_mergeh (A1, E1); \
77 
+    B2 = vec_mergel (A1, E1); \
78 
+    C2 = vec_mergeh (B1, F1); \
79 
+    D2 = vec_mergel (B1, F1); \
80 
+    E2 = vec_mergeh (C1, G1); \
81 
+    F2 = vec_mergel (C1, G1); \
82 
+    G2 = vec_mergeh (D1, H1); \
83 
+    H2 = vec_mergel (D1, H1); \
84 
+ \
85 
+    a = vec_mergeh (A2, E2); \
86 
+    b = vec_mergel (A2, E2); \
87 
+    c = vec_mergeh (B2, F2); \
88 
+    d = vec_mergel (B2, F2); \
89 
+    e = vec_mergeh (C2, G2); \
90 
+    f = vec_mergel (C2, G2); \
91 
+    g = vec_mergeh (D2, H2); \
92 
+    h = vec_mergel (D2, H2); \
93 
+} while (0)
94 
+
95 
+
96 
+/** @brief loads unaligned vector @a *src with offset @a offset
97 
+    and returns it */
98 
+static inline vector unsigned char unaligned_load(int offset, uint8_t *src)
99 
+{
100 
+    register vector unsigned char first = vec_ld(offset, src);
101 
+    register vector unsigned char second = vec_ld(offset+15, src);
102 
+    register vector unsigned char mask = vec_lvsl(offset, src);
103 
+    return vec_perm(first, second, mask);
104 
+}
105 
+
106 
+/**
107 
+ * loads vector known misalignment
108 
+ * @param perm_vec the align permute vector to combine the two loads from lvsl
109 
+ */
110 
+static inline vec_u8 load_with_perm_vec(int offset, uint8_t *src, vec_u8 perm_vec)
111 
+{
112 
+    vec_u8 a = vec_ld(offset, src);
113 
+    vec_u8 b = vec_ld(offset+15, src);
114 
+    return vec_perm(a, b, perm_vec);
115 
+}
116 
+
117 
+#endif /* AVUTIL_PPC_UTIL_ALTIVEC_H */
libavutil/x86/Makefile
History View file @ 7e22514
... ... 
@@ -1 +1,4 @@
1 1 
 OBJS += x86/cpu.o                                                       \
2 
+        x86/float_dsp_init.o                                            \
3 
+
4 
+YASM-OBJS += x86/float_dsp.o                                            \
libavutil/x86/float_dsp.asm
History View file @ 7e22514
2 5 
new file mode 100644
... ... 
@@ -0,0 +1,55 @@
0 
+;*****************************************************************************
1 
+;* x86-optimized Float DSP functions
2 
+;*
3 
+;* This file is part of Libav.
4 
+;*
5 
+;* Libav is free software; you can redistribute it and/or
6 
+;* modify it under the terms of the GNU Lesser General Public
7 
+;* License as published by the Free Software Foundation; either
8 
+;* version 2.1 of the License, or (at your option) any later version.
9 
+;*
10 
+;* Libav is distributed in the hope that it will be useful,
11 
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+;* Lesser General Public License for more details.
14 
+;*
15 
+;* You should have received a copy of the GNU Lesser General Public
16 
+;* License along with Libav; if not, write to the Free Software
17 
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+;******************************************************************************
19 
+
20 
+%include "x86inc.asm"
21 
+
22 
+SECTION .text
23 
+
24 
+;-----------------------------------------------------------------------------
25 
+; void vector_fmul(float *dst, const float *src0, const float *src1, int len)
26 
+;-----------------------------------------------------------------------------
27 
+%macro VECTOR_FMUL 0
28 
+cglobal vector_fmul, 4,4,2, dst, src0, src1, len
29 
+    lea       lenq, [lend*4 - 2*mmsize]
30 
+ALIGN 16
31 
+.loop
32 
+    mova      m0,   [src0q + lenq]
33 
+    mova      m1,   [src0q + lenq + mmsize]
34 
+    mulps     m0, m0, [src1q + lenq]
35 
+    mulps     m1, m1, [src1q + lenq + mmsize]
36 
+    mova      [dstq + lenq], m0
37 
+    mova      [dstq + lenq + mmsize], m1
38 
+
39 
+    sub       lenq, 2*mmsize
40 
+    jge       .loop
41 
+%if mmsize == 32
42 
+    vzeroupper
43 
+    RET
44 
+%else
45 
+    REP_RET
46 
+%endif
47 
+%endmacro
48 
+
49 
+INIT_XMM sse
50 
+VECTOR_FMUL
51 
+%if HAVE_AVX
52 
+INIT_YMM avx
53 
+VECTOR_FMUL
54 
+%endif
libavutil/x86/float_dsp_init.c
History View file @ 7e22514
0 55 
new file mode 100644
... ... 
@@ -0,0 +1,41 @@
0 
+/*
1 
+ * This file is part of FFmpeg.
2 
+ *
3 
+ * FFmpeg is free software; you can redistribute it and/or
4 
+ * modify it under the terms of the GNU Lesser General Public
5 
+ * License as published by the Free Software Foundation; either
6 
+ * version 2.1 of the License, or (at your option) any later version.
7 
+ *
8 
+ * FFmpeg is distributed in the hope that it will be useful,
9 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11 
+ * Lesser General Public License for more details.
12 
+ *
13 
+ * You should have received a copy of the GNU Lesser General Public
14 
+ * License along with FFmpeg; if not, write to the Free Software
15 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16 
+ */
17 
+
18 
+#include "config.h"
19 
+
20 
+#include "libavutil/cpu.h"
21 
+#include "libavutil/float_dsp.h"
22 
+
23 
+extern void ff_vector_fmul_sse(float *dst, const float *src0, const float *src1,
24 
+                               int len);
25 
+extern void ff_vector_fmul_avx(float *dst, const float *src0, const float *src1,
26 
+                               int len);
27 
+
28 
+void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
29 
+{
30 
+#if HAVE_YASM
31 
+    int mm_flags = av_get_cpu_flags();
32 
+
33 
+    if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) {
34 
+        fdsp->vector_fmul = ff_vector_fmul_sse;
35 
+    }
36 
+    if (mm_flags & AV_CPU_FLAG_AVX && HAVE_AVX) {
37 
+        fdsp->vector_fmul = ff_vector_fmul_avx;
38 
+    }
39 
+#endif
40 
+}