f7e5b0cc8f
Remove duplication of TRANSPOSE8 macro, also removing some redundant load/stores
...
Originally committed as revision 15355 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-18 03:08:33 +00:00
7b16a97969
Make a separate macro for writing an entire block from registers to memory
...
Originally committed as revision 15354 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-18 03:08:13 +00:00
3f429d11da
Factorize SSE2_(Row|Column)_IDCT into one macro
...
Originally committed as revision 15353 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-18 03:07:41 +00:00
3c1ebfd56d
Cosmetics: remove excessive blank lines
...
Originally committed as revision 15351 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 20:52:54 +00:00
8cfd78ce8f
Ensure MMX/SSE2 VP3 IDCT selection isn't disabled when only Theora is enabled
...
Originally committed as revision 15350 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 19:49:31 +00:00
c3c5bba1e2
Cosmetics: reindent
...
Originally committed as revision 15349 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 19:38:44 +00:00
f7fc9e499c
Rewrite SSE2 VP3 IDCT in inline asm
...
Originally committed as revision 15347 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 19:30:51 +00:00
437e3f4dbc
Rewrite MMX VP3 IDCT in inline asm
...
Originally committed as revision 15346 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 19:30:33 +00:00
ccd3ec82b8
MMX/SSE2 VP3 IDCT are bitexact now that the dequantization matrices are permutated correctly
...
Originally committed as revision 15345 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 19:30:03 +00:00
987903826b
Globally rename the header inclusion guard names.
...
Consistently apply this rule: the guard name is obtained from the
filename by stripping the leading "lib", converting '/' and '.' to
'_' and uppercasing the resulting name. Guard names in the root
directory have to be prefixed by "FFMPEG_".
Originally committed as revision 15120 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-31 07:39:47 +00:00
b4c3d83584
Use ff_vp3_idct_data in vp3dsp_mmx.c rather than duplicating it
...
Originally committed as revision 15118 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-31 07:05:55 +00:00
43410eaf7c
Declare ff_vp3_idct_data to be uint16_t
...
Originally committed as revision 15117 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-31 07:05:29 +00:00
587ae47aa7
Don't declare SSE vp3 idct data static, so it can be used in the mmx version
...
Originally committed as revision 15116 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-31 07:05:08 +00:00
c979fa030f
Remove unused dequantization code from SSE VP3 IDCT
...
Originally committed as revision 15054 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-30 19:47:47 +00:00
167029a73a
Use ff_pw_8 in MMX/SSE VP3 IDCT
...
Originally committed as revision 15053 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-30 19:41:42 +00:00
21383da8c4
Let ff_pw_8 be used as an SSE constant
...
Originally committed as revision 15052 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-30 19:40:21 +00:00
2ccddc0211
Add explicit (int) cast to i386 optimized MUL* macros.
...
Wrong result is returned when 16-bit value is passed as value.
Also fixes "Warning: using `%edx' instead of `%dx' due to `l' suffix".
Originally committed as revision 14981 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-26 19:38:17 +00:00
dad6afb4cb
stricter constraints of asm() blocks
...
All these variables are used as left operands of a movd instruction,
which does accept only memory or register operands while the "g"
constraint also allows immediates. Use "rm" instead.
Patch by Alexis Ballier %alexis P ballier A gmail P com%
Originally committed as revision 14941 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-24 08:41:20 +00:00
7ca7d5fae0
file which should have been added in r14749
...
Originally committed as revision 14751 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-14 05:00:25 +00:00
75ac287517
missing prototype
...
Originally committed as revision 14750 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-14 04:41:02 +00:00
ebceaa1cd5
gcc chokes on the 7 registers needed for float_to_int16_interleave6 (even inside HAVE_7REGS), so write it in yasm
...
Originally committed as revision 14749 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-14 04:40:46 +00:00
ee46753739
gcc chokes on xmm constraints, so pessimize int32_to_float_fmul_scalar_sse a little
...
Originally committed as revision 14748 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-14 04:39:59 +00:00
675872382f
special case 6 channel version of float_to_int16_interleave
...
5% faster ac3
Originally committed as revision 14744 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-13 23:36:37 +00:00
911e21a306
simd int->float
...
20% faster ac3 if downmixing, 15% if not
Originally committed as revision 14743 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-13 23:35:40 +00:00
ac2e556456
simd downmix
...
13% faster ac3 if downmixing
Originally committed as revision 14742 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-13 23:33:48 +00:00
862b98d42c
cosmetics in dsp init
...
Originally committed as revision 14704 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-12 00:51:45 +00:00
0a570e826d
remove mdct tmp buffer
...
Originally committed as revision 14702 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-12 00:36:36 +00:00
46803f4f67
optimize imdct_half:
...
remove tmp buffer.
skip fft reinterleave pass, leaving data in a format more convenient for simd.
merge post-rotate with post-reorder.
Originally committed as revision 14700 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-12 00:33:34 +00:00
5d0ddd1a9f
split-radix FFT
...
c is 1.9x faster than previous c (on various x86 cpus), sse is 1.6x faster than previous sse.
Originally committed as revision 14698 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-12 00:26:58 +00:00
bafad220a7
import yasm macros from x264
...
Originally committed as revision 14697 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-11 23:54:09 +00:00
f769b746aa
Mark add_png_paeth_prediction_* functions which are only used within this file
...
as static. patch by Uoti Urpala, uoti.urpala pp1.inet fi
Originally committed as revision 14509 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-02 17:32:55 +00:00
4f20b45fbe
Fix h264_loop_filter_strength_mmx2() so it works with PAFF.
...
fixed at least:
CVFI1_Sony_D.jsv
CVFI1_SVA_C.264
MR6_BT_B.h264
Originally committed as revision 14310 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-19 21:53:54 +00:00
5eb0f2a425
float_to_int16_interleave: change src to an array of pointers instead of assuming it's contiguous.
...
this has no immediate effect, but will allow it to be used in more codecs.
Originally committed as revision 14252 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-16 00:50:12 +00:00
4342a7f30b
10l, float_to_int16_interleave_sse/3dnow wrote the wrong samples
...
Originally committed as revision 14236 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-15 04:11:30 +00:00
b9fa32082c
exploit mdct symmetry
...
2% faster vorbis on conroe, k8. 7% on celeron.
Originally committed as revision 14207 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-13 15:03:58 +00:00
f27e1d645e
simplify vorbis windowing
...
Originally committed as revision 14205 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-13 14:56:01 +00:00
d7e1fc4254
SSE2 optimizations for Monkey's Audio decoder vector functions
...
Originally committed as revision 14161 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-11 04:48:38 +00:00
bc31447225
Make the function prototype visible to comply with C99 inline.
...
Fixes building with gcc -std=gnu99.
Originally committed as revision 14140 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-09 17:51:57 +00:00
e98750c373
float_to_int16_sse2()
...
20% faster than sse
Originally committed as revision 14138 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-09 07:21:12 +00:00
1835cda65a
Make LOAD4/STORE4 macros more generic.
...
Patch by Victor Pollex victor pollex web de
Original thread: [PATCH] mmx implementation of vc-1 inverse transformations
Date: 06/21/2008 03:37 PM
Originally committed as revision 14108 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-08 09:24:11 +00:00
35ee72b1d7
1 c-asm loop less and 1x unroll of float_to_int16_sse()
...
25% faster
Originally committed as revision 14104 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-07 21:25:18 +00:00
560fa9bf51
Fix x86-64
...
Originally committed as revision 14103 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-07 21:04:29 +00:00
63b737d4f9
dont use C-asm loops and unroll once float_to_int16_3dnow()
...
30% faster
Originally committed as revision 14102 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-07 20:46:03 +00:00
74fd9022b5
Realign newlines.
...
Originally committed as revision 14023 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-06-28 18:30:50 +00:00
00969e1c59
Use MANGLE() instead of memory operands to read globals.
...
(fixes out of registers with apple gcc 4.2)
Originally committed as revision 14022 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-06-28 18:27:31 +00:00
00eebe3d6a
Fix add_bytes_mmx and add_bytes_l2_mmx for w < 16
...
Originally committed as revision 13877 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-06-22 07:05:40 +00:00
0bd134abd3
Simplify vsad16_mmx2().
...
Originally committed as revision 13193 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:36:44 +00:00
6bf6a9301b
Simplify vsad16_mmx().
...
Originally committed as revision 13191 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:35:14 +00:00
e13810223a
Simplify vsad_intra16_mmx2()
...
Originally committed as revision 13189 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:33:01 +00:00
06bb35f94c
Simplify vsad_intra16_mmx()
...
Originally committed as revision 13188 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:31:10 +00:00
David Conrad
Stefano Sabatini
Vladimir Voroshilov
Alexis Ballier
Loren Merritt
Uoti Urpala
Michael Niedermayer
Kostya Shishkov
Alexander Strange
Victor Pollex
Reimar Döffinger