[libav-commits] x86: dcadsp: implement int8x8_fmul_int32

Module: libav Branch: master Commit: 5b59a9fc6152169599561f04b4f66370edda5c9c Author:Christophe Gisquet Committer: Janne Grunau Date: Fri May 11 11:25:30 2012 +0200 x86: dcadsp: implement int8x8_fmul_int32 For the callable function (as opposed to the inline one): C SSE

[libav-commits] dcadsp: add int8x8_fmul_int32 to dsp context

Module: libav Branch: master Commit: 2bd44cb705340c4f7bd7e459a1efed5074bf45fc Author:Christophe Gisquet Committer: Janne Grunau Date: Fri May 11 11:17:36 2012 +0200 dcadsp: add int8x8_fmul_int32 to dsp context It is currently declared as a macro who is set to inlinable functions

[libav-commits] dcadsp: split lfe_dir cases

Module: libav Branch: master Commit: 5fdbfcb5b793f5849c496214668094a8ec99fa07 Author:Christophe Gisquet Committer: Janne Grunau Date: Wed Feb 5 23:40:52 2014 + dcadsp: split lfe_dir cases The x86 runs short on registers because numerous elements are not static. In addition

[libav-commits] build: fix examples compilation when EXESUF is not empty

Module: libav Branch: master Commit: 682b2a809b446c4ca45f88a84c1d92ccfcbcfbd1 Author:Christophe Gisquet Committer: Janne Grunau Date: Mon Feb 10 00:23:28 2014 +0100 build: fix examples compilation when EXESUF is not empty Due to a wrong substitution doc/examples is not added as a

[libav-commits] dca: replace some memcpy by AV_COPY128

Module: libav Branch: master Commit: ef010f08ae53479c54e2f16be5a7e1a809a9e268 Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:10 2014 + dca: replace some memcpy by AV_COPY128 Signed-off-by: Janne Grunau --- libavcodec/dcadec.c |4 +--- 1 file

[libav-commits] x86: float dsp: unroll SSE versions

Module: libav Branch: master Commit: 996697e266c8adc0ad9b7fc7568406c7529c97cf Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:12 2014 + x86: float dsp: unroll SSE versions vector_fmul and vector_fmac_scalar are guaranteed that they can process in batch of

[libav-commits] dcadec: simplify decoding of VQ high frequencies

Module: libav Branch: master Commit: 4cb6964244fd6c099383d8b7e99731e72cc844b9 Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:13 2014 + dcadec: simplify decoding of VQ high frequencies The vector dequantization has a test in a loop preventing effective

[libav-commits] x86: dcadsp: implement SSE lfe_dir

Module: libav Branch: master Commit: ad507d7907457e678900bac132122ba7be4644cb Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:07 2014 + x86: dcadsp: implement SSE lfe_dir Results for Arrandale/Windows: 32: 1670 -> 316 64: 728 -> 298 Signed-off-by:

[libav-commits] x86: synth filter float: implement SSE2 version

Module: libav Branch: master Commit: 08e3ea60ff4059341b74be04a428a38f7c3630b0 Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:09 2014 + x86: synth filter float: implement SSE2 version Timings for Arrandale: CSSE win32: 2108 334 win64

[libav-commits] dcadec: remove scaling in lfe_interpolation_fir

Module: libav Branch: master Commit: 87ec849fe9acba075c843e67bcd01f256f481a18 Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:06 2014 + dcadec: remove scaling in lfe_interpolation_fir The scaling factor is constant so it is faster to scale the FIR

[libav-commits] dcadsp: scan coefficients linearly in dca_lfe_fir

Module: libav Branch: master Commit: 57b1eb9f75b04571063ddec316e290c216c114ac Author:Christophe Gisquet Committer: Janne Grunau Date: Fri Feb 14 15:03:08 2014 + dcadsp: scan coefficients linearly in dca_lfe_fir This change is inspired by x86 asm where it frees a register. Signed

[libav-commits] dca: replace some memcpy by AV_COPY128

Module: libav Branch: release/10 Commit: 2897481f64c8ab856b2121ae900d822930d0a5fc Author:Christophe Gisquet Committer: Reinhard Tartler Date: Fri Feb 14 15:03:10 2014 + dca: replace some memcpy by AV_COPY128 Signed-off-by: Janne Grunau (cherry picked from commit

[libav-commits] x86: deduplicate some constants

Module: libav Branch: master Commit: 3c504bc3599f00bfc5923adc114beef34bce11d0 Author:Christophe Gisquet Committer: Anton Khirnov Date: Wed Aug 6 07:43:39 2014 + x86: deduplicate some constants Signed-off-by: Anton Khirnov --- libavcodec/x86/constants.c |2

[libav-commits] aac: avoid a memcpy in sbr_qmf_analysis

Module: libav Branch: master Commit: e32bea8eb479d39359042cb43fa0cb9f92863e6b Author:Christophe Gisquet Committer: Luca Barbato Date: Wed Nov 28 22:47:25 2012 +0100 aac: avoid a memcpy in sbr_qmf_analysis Swapping buffer indices allows saving one memcpy that accounts for 1% of the

[libav-commits] SBR DSP x86: implement SSE sbr_hf_gen

Module: libav Branch: master Commit: 2aef3d66c9cdf9d9ad3ecec2fb0c6b3020e9d3b0 Author:Christophe Gisquet Committer: Diego Biurrun Date: Fri Feb 24 22:11:19 2012 +0100 SBR DSP x86: implement SSE sbr_hf_gen Start and end index are multiple of 2, therefore guaranteeing aligned access

[libav-commits] AAC SBR: use AVFloatDSPContext's vector_fmul

Module: libav Branch: master Commit: 9a16359c3af6beab7e688fc55eaa60c835ec Author:Christophe Gisquet Committer: Diego Biurrun Date: Fri Nov 30 21:49:33 2012 +0100 AAC SBR: use AVFloatDSPContext's vector_fmul Around 5% speedup on the code block using 'vector_fmul_add'

[libav-commits] x86: sbrdsp: Implement SSE qmf_post_shuffle

Module: libav Branch: master Commit: 4f50646697606df39317b93c2a427603b77636ee Author:Christophe Gisquet Committer: Diego Biurrun Date: Sun Nov 25 07:55:28 2012 +0100 x86: sbrdsp: Implement SSE qmf_post_shuffle 255 to 174 cycles on Arrandale / Win64. Unrolling yields no gain. Signed

[libav-commits] x86: sbrdsp: Implement SSE sum64x5

Module: libav Branch: master Commit: 44a0036d10579ed91e48df24859e54b08a582742 Author:Christophe Gisquet Committer: Diego Biurrun Date: Sat Nov 24 23:15:07 2012 +0100 x86: sbrdsp: Implement SSE sum64x5 698 to 174 cycles on Arrandale. Unrolling is a 6 cycles gain. Signed-off-by: Diego

[libav-commits] x86: sbrdsp: Implement SSE neg_odd_64

Module: libav Branch: master Commit: f4b0d12f5b3fb2fe825fb2ac9b2d9d9374baa5c5 Author:Christophe Gisquet Committer: Diego Biurrun Date: Fri Dec 21 18:11:06 2012 +0100 x86: sbrdsp: Implement SSE neg_odd_64 Timing on Arrandale: C SSE Win32: 57 44 Win64: 47 38 Unrolling

[libav-commits] x86inc: Fix number of operands for cmp* instructions

Module: libav Branch: master Commit: 2e81acc687e64d15dd93c74793060bb5a233f44d Author:Christophe Gisquet Committer: Diego Biurrun Date: Sun Apr 7 19:29:52 2013 + x86inc: Fix number of operands for cmp* instructions cmp{p,s}{s,d} instructions do take an imm8 operand. Signed-off

[libav-commits] x86: float dsp: butterflies_float SSE

Module: libav Branch: master Commit: 566b7a20fd0cab44d344329538d314454a0bcc2f Author:Christophe Gisquet Committer: Anton Khirnov Date: Fri Apr 12 21:07:01 2013 +0200 x86: float dsp: butterflies_float SSE 97c -> 49c Some codecs could benefit from more unrolling, but AAC does

[libav-commits] x86: sbrdsp: Implement SSE2 qmf_deint_bfly

Module: libav Branch: master Commit: 5a97469a4fd12ef0327292ad6062f89a5e055a62 Author:Christophe Gisquet Committer: Diego Biurrun Date: Tue Apr 9 21:57:07 2013 + x86: sbrdsp: Implement SSE2 qmf_deint_bfly Sandybridge: 47 cycles Having a loop counter is a 7 cycle gain. Unrolling

[libav-commits] sbrdsp: Unroll and use integer operations

Module: libav Branch: master Commit: 4a7af92cc80ced8498626401ed21f25ffe6740c8 Author:Christophe Gisquet Committer: Diego Biurrun Date: Tue Apr 9 21:57:10 2013 + sbrdsp: Unroll and use integer operations This patch can be controversial, by assuming floats are IEEE-754 and

[libav-commits] sbrdsp: Unroll sbr_autocorrelate_c

Module: libav Branch: master Commit: 8394d9a676db1ffa1c8e4f6dd239e154d051ce03 Author:Christophe Gisquet Committer: Diego Biurrun Date: Tue Apr 9 21:57:09 2013 + sbrdsp: Unroll sbr_autocorrelate_c 1410 cycles to 1148 on Arrandale/Win64 Signed-off-by: Diego Biurrun

[libav-commits] x86: sbrdsp: implement SSE2 qmf_pre_shuffle

Module: libav Branch: master Commit: 2c299d4165cd9653153e12270971c2368551b79e Author:Christophe Gisquet Committer: Anton Khirnov Date: Sun Nov 25 09:10:36 2012 +0100 x86: sbrdsp: implement SSE2 qmf_pre_shuffle >From 253 to 51 cycles on Arrandale and Win64. 44 cycles on SandyBri

[libav-commits] fmtconvert: Explicitly use int32_t instead of int

Module: libav Branch: master Commit: b6293e2798afab60596a87010b6163fcb4ca3086 Author:Christophe Gisquet Committer: Martin Storsjö Date: Thu Dec 27 22:33:51 2012 +0100 fmtconvert: Explicitly use int32_t instead of int Signed-off-by: Martin Storsjö --- libavcodec/ac3dec.c

[libav-commits] vp8dsp x86: perform rounding shift with a single instruction

Module: libav Branch: master Commit: f9888520cc853fe27cbe6f3827b2723c63299a4f Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Mon Mar 26 02:29:09 2012 +0200 vp8dsp x86: perform rounding shift with a single instruction Signed-off-by: Ronald S. Bultje --- libavcodec/x86

[libav-commits] dsputil x86: remove deprecated parameter from scalarproduct_int16 prototype

Module: libav Branch: master Commit: cd88105f6fddf04fb04fe6c57cbaac5f6037c6a6 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Mon Mar 19 16:34:39 2012 +0100 dsputil x86: remove deprecated parameter from scalarproduct_int16 prototype Signed-off-by: Ronald S. Bultje

[libav-commits] dsputil x86: use SSE float instruction instead of SSE2 integer equivalent

Module: libav Branch: master Commit: 6b81da2fd030dbc03d36b38558b58e8f0f667501 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Wed Mar 21 20:46:29 2012 +0100 dsputil x86: use SSE float instruction instead of SSE2 integer equivalent All the more required since the users are

[libav-commits] rv40dsp x86: use only one register, for both increment and loop counter

Module: libav Branch: master Commit: 2130bd8f5b6504ea14cd41e33f5d4f431eb724f3 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Tue Mar 20 16:13:55 2012 +0100 rv40dsp x86: use only one register, for both increment and loop counter Around 10 cycles faster for luma. Signed

[libav-commits] rv40dsp: implement prescaled versions for biweight.

Module: libav Branch: master Commit: 272b252c0110225188c7d7f31167941210aac197 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Mon Mar 19 22:46:28 2012 +0100 rv40dsp: implement prescaled versions for biweight. Quite often, the original weights are multiple of 512. By

[libav-commits] rv34dsp x86: implement MMX2 inverse transform

Module: libav Branch: master Commit: fe5ed69dc7f6acf6928c27d614f2be3fc89ed176 Author:Christophe Gisquet Committer: Ronald S. Bultje Date: Thu Mar 29 22:37:41 2012 +0200 rv34dsp x86: implement MMX2 inverse transform 141 cycles down to 51. Signed-off-by: Ronald S. Bultje

[libav-commits] dsputil x86: revert a test back to its previous value

Module: libav Branch: master Commit: e75d1d4f7321b097f4b09799bf28a22ad259fea6 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Wed Mar 28 23:22:22 2012 +0200 dsputil x86: revert a test back to its previous value Commit 356ee8d caused the initial inversion. Signed-off-by

[libav-commits] rv40: don't always do the full prev_type search

Module: libav Branch: master Commit: 34e1b0754638ea6e0356239aad01e74de3a582a5 Author:Christophe Gisquet Committer: Ronald S. Bultje Date: Tue Apr 17 18:44:48 2012 + rv40: don't always do the full prev_type search 120->100 cycles. Signed-off-by: Ronald S

[libav-commits] rv34: remove constant parameter

Module: libav Branch: master Commit: 8802719332b2e6b27f357b3eb12d7e0abdad Author:Christophe Gisquet Committer: Ronald S. Bultje Date: Wed Apr 4 10:22:28 2012 +0200 rv34: remove constant parameter is_block2 was always 0, so just remove it, and change accordingly the code. Signed

[libav-commits] rv40: change a logical test into a bitwise one.

Module: libav Branch: master Commit: 0195ab5f5ff216a8797623e62e2703f3df2624c2 Author:Christophe Gisquet Committer: Ronald S. Bultje Date: Tue Apr 17 18:44:50 2012 + rv40: change a logical test into a bitwise one. Signed-off-by: Ronald S. Bultje --- libavcodec/rv34.c |2

[libav-commits] rv40: perform bitwise checks in loop filter

Module: libav Branch: master Commit: d834ab0f2a5d5689cdd323e935534fda918319ef Author:Christophe Gisquet Committer: Ronald S. Bultje Date: Tue Apr 17 18:44:51 2012 + rv40: perform bitwise checks in loop filter Down from 95 kcycles to 93 (including all called functions). Signed

[libav-commits] rv34dsp: factorize a multiplication in the noround inverse transform

Module: libav Branch: master Commit: 7fb8b491e5c10b3fbad53ffc89c60d3541114416 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Tue Apr 17 18:44:43 2012 + rv34dsp: factorize a multiplication in the noround inverse transform Signed-off-by: Ronald S. Bultje

[libav-commits] ape: Use unsigned integer maths

Module: libav Branch: master Commit: 706b998cdcea97c50fad2228f67488de0e06b2a2 Author:Christophe Gisquet Committer: Diego Biurrun Date: Tue Feb 28 17:42:12 2012 +0100 ape: Use unsigned integer maths This involves a division that should be a shift. Signed-off-by: Diego Biurrun

[libav-commits] rv40dsp x86: MMX/MMX2/3DNow/SSE2/SSSE3 implementations of MC

Module: libav Branch: master Commit: 110d0cdc9d1ec414a658f841a3fbefbf6f796d61 Author:Christophe Gisquet Committer: Diego Biurrun Date: Thu Apr 19 22:36:17 2012 +0200 rv40dsp x86: MMX/MMX2/3DNow/SSE2/SSSE3 implementations of MC Code mostly inspired by vp8's MC, however: - its

[libav-commits] x86: fft: replace call to memcpy by a loop

Module: libav Branch: master Commit: a5bfa66df516b7be55fd08fc62c2b012fc18e340 Author:Christophe Gisquet Committer: Mans Rullgard Date: Tue Jun 26 16:10:33 2012 +0200 x86: fft: replace call to memcpy by a loop The function call was a mess to handle, and memcpy cannot make the

[libav-commits] hevc: derive partially amvp list

Module: libav Branch: master Commit: 3505b19652ea34089a4cd08d4d3358fcc0a3db8b Author:Christophe Gisquet Committer: Anton Khirnov Date: Fri Aug 8 18:21:00 2014 + hevc: derive partially amvp list When the candidate has been found, no need to derive others. Signed-off-by: Anton

[libav-commits] hevc: reorder loops

Module: libav Branch: master Commit: d1b1c3bb5ef1bee409c06c3bcacfb8674e1cb574 Author:Christophe Gisquet Committer: Anton Khirnov Date: Fri Aug 8 18:20:58 2014 + hevc: reorder loops iterate over memory in a more continuous order Signed-off-by: Anton Khirnov --- libavcodec

[libav-commits] hevc: use intreadwrite

Module: libav Branch: master Commit: cf6090dc6252f2b276aa4133e3d73a89f4c6046c Author:Christophe Gisquet Committer: Anton Khirnov Date: Fri Aug 8 18:21:02 2014 + hevc: use intreadwrite When dealing with MVs, both components may be processed at a time. Signed-off-by: Anton

[libav-commits] hevc: derive partial merge list

Module: libav Branch: master Commit: ed53cc217f6a6e4ddb35e3f01a79496091dc82dc Author:Christophe Gisquet Committer: Anton Khirnov Date: Fri Aug 8 18:21:01 2014 + hevc: derive partial merge list The merge list only needs to be derived up to the merge index. Signed-off-by: Anton

[libav-commits] hevc_mvs: avoid deriving tmvp in amvp

Module: libav Branch: master Commit: 3187fa14a326908f9471a038e3b5b24c0eaf655e Author:Christophe Gisquet Committer: Anton Khirnov Date: Fri Aug 8 18:21:04 2014 + hevc_mvs: avoid deriving tmvp in amvp Reduces the number of calls to tmvp derivation from 933685 to 586271 on a

[libav-commits] hevc: wait proper position for tmvp

Module: libav Branch: master Commit: 14ca0aa69b5d954ce56d0826ab9f1f70e535c92a Author:Christophe Gisquet Committer: Anton Khirnov Date: Sun Jul 20 09:34:58 2014 +0200 hevc: wait proper position for tmvp The position is either rounded or not checked, so delay the wait to check the

[libav-commits] hevc: simplify rounding

Module: libav Branch: master Commit: 09182b3224370e1d636631e77f201a8388766bd6 Author:Christophe Gisquet Committer: Anton Khirnov Date: Thu Jul 10 06:47:55 2014 + hevc: simplify rounding Signed-off-by: Anton Khirnov --- libavcodec/hevc_mvs.c |8 1 file changed, 4

[libav-commits] proresenc: Report buffer overflow

Module: libav Branch: master Commit: 58b68e4fdea22e22178e237bda950b09cc6f363a Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:22 2014 + proresenc: Report buffer overflow If the allocated size, despite best efforts, is too small, exit with the appropriate

[libav-commits] proresenc: Realloc if buffer is too small

Module: libav Branch: master Commit: 45ce880a9b3e50cfa088f111dffaf8685bd7bc6b Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:23 2014 + proresenc: Realloc if buffer is too small The buffer allocation may be incorrect (e.g. with an alpha plane), and

[libav-commits] proresenc: Properly account for alpha plane

Module: libav Branch: master Commit: 41e1354c101004ccd46dc08d3dd6e956e83a6b51 Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:24 2014 + proresenc: Properly account for alpha plane The packet buffer allocation considers the alpha channel as DCT-coded

[libav-commits] proresenc: Remove unneeded parameters from encode_alpha_plane()

Module: libav Branch: master Commit: b16699f2da9c1d41eff852ec3a0c81f74fd44421 Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:21 2014 + proresenc: Remove unneeded parameters from encode_alpha_plane() Signed-off-by: Diego Biurrun Signed-off-by: Luca

[libav-commits] proresenc: Realloc if buffer is too small

Module: libav Branch: release/10 Commit: a437298de55c6a6a4f06b12335b3891bf4459082 Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:23 2014 + proresenc: Realloc if buffer is too small The buffer allocation may be incorrect (e.g. with an alpha plane), and

[libav-commits] proresenc: Report buffer overflow

Module: libav Branch: release/10 Commit: e912b0777b24133df27836b6c529faa89af588dc Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:22 2014 + proresenc: Report buffer overflow If the allocated size, despite best efforts, is too small, exit with the

[libav-commits] proresenc: Properly account for alpha plane

Module: libav Branch: release/10 Commit: f25f5f8c62ec7728ee7f5dcc8f1abd0dc6235735 Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:24 2014 + proresenc: Properly account for alpha plane The packet buffer allocation considers the alpha channel as DCT-coded

[libav-commits] proresenc: Remove unneeded parameters from encode_alpha_plane()

Module: libav Branch: release/10 Commit: b3f48a5044fd04539337e91d28022207c9d3b9e8 Author:Christophe Gisquet Committer: Luca Barbato Date: Mon Aug 18 14:15:21 2014 + proresenc: Remove unneeded parameters from encode_alpha_plane() Signed-off-by: Diego Biurrun Signed-off-by: Luca

[libav-commits] x86inc: Fix instantiation of YMM registers

Module: libav Branch: master Commit: f5e486f6f8c242bb2be01ad3ae952b5733ba1113 Author:Christophe Gisquet Committer: Anton Khirnov Date: Sat Aug 1 17:27:34 2015 +0200 x86inc: Fix instantiation of YMM registers Signed-off-by: Henrik Gramner Signed-off-by: Anton Khirnov

[libav-commits] dnxhddec: Decode and use interlace mb flag

Module: libav Branch: master Commit: c49cbecbae5a42f4ca004197b0118cc50aaaca2e Author:Christophe Gisquet Committer: Vittorio Giovara Date: Sat Sep 26 17:38:22 2015 +0200 dnxhddec: Decode and use interlace mb flag This bit is 1 in some samples, and seems to coincide with interlaced mbs

[libav-commits] dnxhddata: Deduplicate cid 1256 chroma table

Module: libav Branch: master Commit: e54d7e4e8ef7d778e2ddc5a60bf7307ed55d6333 Author:Christophe Gisquet Committer: Vittorio Giovara Date: Thu Oct 1 16:54:57 2015 +0200 dnxhddata: Deduplicate cid 1256 chroma table CID 1256 is specified as using the same table for luma and chroma

[libav-commits] rv34: joint coefficient decoding and dequantization

Module: libav Branch: master Commit: 98f24ecd6cfc9c57a555aae6bfcd3d9a4ce9503d Author:Christophe GISQUET Committer: Kostya Shishkov Date: Sun Jan 1 15:28:47 2012 +0100 rv34: joint coefficient decoding and dequantization Perform dequantization while decoding coefficients instead of

[libav-commits] rv34: DC-only inverse transform

Module: libav Branch: master Commit: 3faa303a47e0c3b59a53988e0f76018930c6cb1a Author:Christophe GISQUET Committer: Kostya Shishkov Date: Sun Jan 1 18:33:22 2012 +0100 rv34: DC-only inverse transform When decoding coefficients, detect whether the block is DC-only, and take advantage

[libav-commits] rv34: Inter/intra MB code split

Module: libav Branch: master Commit: 3eeb7557637e8e48fbc64e844a94775edb496906 Author:Christophe GISQUET Committer: Janne Grunau Date: Mon Jan 2 20:53:54 2012 +0100 rv34: Inter/intra MB code split Split inter/intra macroblock handling code. This will allow further optimizations such

[libav-commits] rv34: Intra 16x16 handling

Module: libav Branch: master Commit: d78062386e425deafe9a08d109cff70b7a2de22c Author:Christophe GISQUET Committer: Janne Grunau Date: Tue Jan 3 00:22:11 2012 +0100 rv34: Intra 16x16 handling Extract processing of intra 16x16 blocks from intra macroblock processing. Also implement a

[libav-commits] rv34: 1-pass inter MB reconstruction

Module: libav Branch: master Commit: 9ba9c3402499d90e54f8aa111b62c278206d11af Author:Christophe GISQUET Committer: Janne Grunau Date: Tue Jan 3 20:38:29 2012 +0100 rv34: 1-pass inter MB reconstruction Implement 1-pass inverse transform and reconstruction for inter blocks

[libav-commits] x86 dsputil: provide SSE2/SSSE3 versions of bswap_buf

Module: libav Branch: master Commit: 6b039003822a03add20c7ba91fc857dca52b0a03 Author:Christophe Gisquet Committer: Diego Biurrun Date: Thu Jan 19 21:48:39 2012 +0100 x86 dsputil: provide SSE2/SSSE3 versions of bswap_buf While pshufb allows emulating bswap on XMM registers for SSSE3

[libav-commits] rv40: x86 SIMD for biweight

Module: libav Branch: master Commit: e5c9de2ab78cf18636eac2fa8e059e58a71e512b Author:Christophe Gisquet Committer: Diego Biurrun Date: Thu Jan 12 00:11:15 2012 +0100 rv40: x86 SIMD for biweight Provide MMX, SSE2 and SSSE3 versions, with a fast-path when the weights are multiples of

[libav-commits] mpegaudio: replace memcpy by SIMD code

Module: libav Branch: master Commit: f3e084909bff422f0c853507a82f92ff2efc0d28 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Wed Jan 18 22:34:29 2012 +0100 mpegaudio: replace memcpy by SIMD code By replacing memcpy with an unrolled loop using the alignment knowledge it

[libav-commits] prores: use natural integer type for the codebook index

Module: libav Branch: master Commit: b5696ff2b872b3556a71fe3591c245b4caff08b5 Author:Christophe GISQUET Committer: Diego Biurrun Date: Tue Feb 21 22:36:15 2012 +0100 prores: use natural integer type for the codebook index The operations that use it require it to be promoted to a

[libav-commits] SBR DSP x86: implement SSE sbr_hf_g_filt

Module: libav Branch: master Commit: 2784d187919b48022a89633fb3b5a99ca97cf869 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Thu Feb 23 20:12:39 2012 +0100 SBR DSP x86: implement SSE sbr_hf_g_filt Unrolling the main loop to process, instead of 4 elements: - 8: minor gain

[libav-commits] SBR DSP x86: implement SSE sbr_sum_square_sse

Module: libav Branch: master Commit: 34454c761f01275d4adaf40df6d70a59011c4a6c Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Thu Feb 23 19:48:58 2012 +0100 SBR DSP x86: implement SSE sbr_sum_square_sse The 32bits targets have been compiled with -mfpmath=sse for proper

[libav-commits] SBR DSP: use intptr_t for the ixh parameter.

Module: libav Branch: master Commit: 2e74a5abc2fda6cfbc86589852d6194d502332cb Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Thu Feb 23 22:25:48 2012 +0100 SBR DSP: use intptr_t for the ixh parameter. Signed-off-by: Ronald S. Bultje --- libavcodec/arm

[libav-commits] aacsbr: use a swap index for the Y matrix rather than copy buffers.

Module: libav Branch: master Commit: cc412b71047ebf77c7e810c90b044f018a1c0c2d Author:Christophe Gisquet Committer: Alex Converse Date: Thu Feb 23 22:25:44 2012 + aacsbr: use a swap index for the Y matrix rather than copy buffers. Signed-off-by: Alex Converse --- libavcodec

[libav-commits] rv34: clean a bit availability checks.

Module: libav Branch: master Commit: 7104c23bd1a1dcb8a7d9e2c8838c7ce55c30a331 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Mon Jan 16 20:54:26 2012 +0100 rv34: clean a bit availability checks. Signed-off-by: Ronald S. Bultje --- libavcodec/rv34.c | 20

[libav-commits] AAC SBR: group some writes.

Module: libav Branch: master Commit: ba36f14e5d34bcca6ecf77ba1c7512d1b51a9740 Author:Christophe Gisquet Committer: Ronald S. Bultje Date: Tue Feb 28 11:20:02 2012 +0100 AAC SBR: group some writes. Signed-off-by: Ronald S. Bultje --- libavcodec/aacsbr.c | 18

[libav-commits] dsputil: remove shift parameter from scalarproduct_int16

Module: libav Branch: master Commit: 7e1ce6a6acd83cf2d5b21df94d2134b1553635ef Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Sat Mar 3 15:09:36 2012 +0100 dsputil: remove shift parameter from scalarproduct_int16 There is only one caller, which does not need the shifting

[libav-commits] SBR DSP: unroll sum_square

Module: libav Branch: master Commit: dabf8dd34afdbb6dc9dc7603d7a5228fc67de4c8 Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Wed Feb 22 17:48:59 2012 +0100 SBR DSP: unroll sum_square The length is even, so some unrolling can be performed. Timings are for x86: - 32bits

[libav-commits] rv34: remove dead code in intra availability check

Module: libav Branch: master Commit: 294c05ce8a7fbe3de74023065c264c1d720ec3cd Author:Christophe GISQUET Committer: Ronald S. Bultje Date: Thu Feb 23 19:10:32 2012 +0100 rv34: remove dead code in intra availability check This was an incorrect copy-and-paste to a code not needing the