Re: [FFmpeg-devel] avcodec/qtrle : improve decoding speed of 24bpp and 32 bpp

> > This seems to have broken all qtrle tests on sparc64. > > See http://fate.ffmpeg.org/history.cgi?slot=sparc64-linux-gcc-4.9 and > others. > Thanks for reporting this. Like i can't test on sparc64, doesn't know what line create the bus error. Patch in attach (untested on sparc64), can probably

[FFmpeg-devel] avcodec/proresenc (aw and ks) : use correct bitstream version in frame header

Hello Patch in attach change bitstream version in prores frame header (based on RDD36) 0 for 422 1 for 444 with ou without alpha 001 : Fix for prores_aw (update 444 fate test) 002 : Fix for prores_ks (no fate update because 444 is not tested for ks encoder) Pass fate for me (x86_64 macos) Marti

Re: [FFmpeg-devel] [PATCH] News: Removal of libndi

>+Support for the nonfree NDI protocol has been removed, it had > been a common source of GPL violations. > > This doesn't justify to break user tools (who respect the ffmpeg licence) and remove contributor's work. Martin ___ ffmpeg-devel mailing l

Re: [FFmpeg-devel] [PATCH] News: Removal of libndi

> > >+Support for the nonfree NDI protocol has been removed, it had > > > been a common source of GPL violations. > > > > > This doesn't justify to break user tools (who respect the ffmpeg licence) > > tools who depend on a license violation are violating the license too... > FFmpeg can be

Re: [FFmpeg-devel] [PATCH] avcodec/exr detect invalid line offset table and recreate table

Hello, I tested your patch on mac, seems to work, and doesn't break exr fate test. Some comments : > --- > libavcodec/exr.c | 19 +++ > 1 file changed, 19 insertions(+) > > diff --git a/libavcodec/exr.c b/libavcodec/exr.c > index 034920f..dec21af 100644 > --- a/libavcodec/exr.c

Re: [FFmpeg-devel] libavcodec/psd : add support for rgb/gray float

2017-03-08 2:15 GMT+01:00 Dzung Hoang : > I implemented a 32-bit float GBRAP/GBRP pixel format and used it in > vf_zscale.c and vf_overlay.c. I also implemented an EXR encoder using > miniexr that uses the new float pixel format. I can contribute this if > there is interest. > I used the above mod

Re: [FFmpeg-devel] libavcodec/exr : add support for uint32

2017-03-07 23:05 GMT+01:00 Martin Vignali : > Hello, > > In attach patch for decoding uint32 channels in exr files. > > Following previous comments, i doesn't use float/color transformation for > this kind of pixel data > in this new version. > > Comments welcome &

[FFmpeg-devel] fate/exr : add test for uint32

Hello, In attach patch to add fate test for uint 32 exr Sample can be found here : https://we.tl/iB6xWicb4X and need to be put inside fate-suite/exr/ Need to be apply after the patch : libavcodec/exr add support for uint32 Martin Jokyo Images 0002-fate-exr-add-test-for-uint32-data.patch De

Re: [FFmpeg-devel] libavcodec/exr : add support for uint32

> > Just a nitpick: > > > +} else { > > +//UINT 32 > > The comment indicates this should be "else if (pixel_type == UINT)" > > Not sure it's necessary to replace else by else if, because OpenExr "planar" can only have half, float or uint32 data But new patch attach. Martin 0

Re: [FFmpeg-devel] [PATCH] avcodec/exr detect invalid line offset table and recreate table

Hello, The latest patch appear to have a problem : fatal: corrupt patch at line 6 And it doesn't work for me (head variable seems to be important). I put in attach a patch, based on the previous one, who works for me. Martin 0001-libavcodec-exr-add-support-for-scanline-file-where-o.patch Des

[FFmpeg-devel] libavcodec/exr : cosmetics : rename variable

Hello, In attach a patch who rename tile variable to better follow ffmpeg coding style Martin -- Martin Vignali Jokyo Images 18 rue du Transvaal 69008 Lyon 06 99 89 33 30 m.vign...@jokyo-images.com http://www.jokyo-images.com/ 0003-libavcodec-exr-cosmetics-variable-name.patch Description

Re: [FFmpeg-devel] libavcodec/exr : add support for uint32

Ping for apply Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] fate/exr : add test for exr with offsets set to zero

/0jRSKpmIQu And need to be put inside ./fate-suite/exr The patch need to be apply after applying patch in the discussion : avcodec/exr detect invalid line offset table and recreate table Martin Vignali Jokyo Images 0002-fate-exr-add-test-for-scanline-file-with-offsets-set.patch Description: Binary

[FFmpeg-devel] fate/cineform : add test for yuv 10b

Hello, In attach patch to add fate test for cineform decoder Sample can be found here : https://we.tl/y06cKrKh3a there is two sample in the download link The test is about : cineform_yuv10b_hd.mov this sample need to be put inside ./fate-suite/cineform/ (the folder doesn't exist) and can be te

Re: [FFmpeg-devel] libavcodec/exr : add support for uint32

2017-03-27 10:10 GMT+02:00 Carl Eugen Hoyos : > 2017-03-25 13:09 GMT+01:00 Martin Vignali : > > Ping for apply > > Please send your public key to Michael and ask him to add you > as committer, you have many commits in FFmpeg. > > Hello, I'm not enough familiar with g

Re: [FFmpeg-devel] fate/exr : add test for uint32

Hello, This fate test can now be apply (the patch for uint32 exr have been apply) Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] fate/exr : add test for exr with offsets set to zero

Hello, This patch can now be apply (the patch avcodec/exr detect invalid line offset table and recreate table have been apply) Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] fate/exr : add test for exr with offsets set to zero

> > failed test on mips qemu > thus not pushed > > [exr @ 0x1e6e900] Warning: not compiled with thread support, using thread > emulation > [exr @ 0x1e6e900] Unsupported channel VRaySamplerInfo.A. > [exr @ 0x1e6e900] Unsupported channel VRaySamplerInfo.B. > [exr @ 0x1e6e900] Unsupported channel VRay

[FFmpeg-devel] fate/exr : fix pix_fmt for pxr24 half/uint32

Hello, In attach a patch to fix a fate test, the sample doesn't have alpha Martin 0003-fate-exr-fix-pix_fmt.patch Description: Binary data ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] fate/pixlet : add test for rgb

Hello, In attach patch to add a fate test for pixlet decoder Sample can be found here : https://we.tl/HYuFmmmfIe and need to be put inside ./fate-suite/pixlet (the folder doesn't exist) test can be run with make fate-pixlet SAMPLES=fate-suite/ Martin 0004-fate-pixlet-add-test-for-rgb.patch

[FFmpeg-devel] fate/dv : add tests for dvcpro hd decoding

Hello, in attach patch for testing dvcpro hd decoding Samples can be found here : https://we.tl/STetjfjCpy and need to be put inside ./fate-suite/dv/ (folder doesn't exist) test can be run with make fate-dv SAMPLES=./fate-suite/ Martin 0005-fate-dv-add-test-for-dvcprohd-decoding.patch Descr

Re: [FFmpeg-devel] fate/dv : add tests for dvcpro hd decoding

Hello, Thanks for testing. I will try to find a way to test on big endian. Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] fate/exr : fix pix_fmt for pxr24 half/uint32

2017-04-02 23:00 GMT+02:00 Martin Vignali : > Hello, > > In attach a patch to fix a fate test, the sample doesn't have alpha > > Martin > ping ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] fate/exr : fix pix_fmt for pxr24 half/uint32

> > How does it fix it ? > if it doesnt contain alpha but tests alpha thats a bit odd and may > test things that otherwise wouldnt be tested. > But "fix" sounds like theres something wrong, not just odd > Can you clarify/elaborate ? > > thx > > Hello, I made a copy and paste of a previous line whe

[FFmpeg-devel] libavcodec/exr : fix piz uncompress on big endian

Hello, In attach a patch who i think fix decoding of piz file on big endian. I doesn't find a way to test it on mips qemu but if i "enable" the HAVE_BIG_ENDIAN part on x86 i have the same wrong crc previously reported in discussion : https://ffmpeg.org/pipermail/ffmpeg-devel/2016-April/193471.htm

[FFmpeg-devel] Fate/exr : add test for piz uncompress

Hello, Patch in attach for fate test of piz Sample can be found here : https://we.tl/k1olNPK11t Need to be put inside ./fate-suite/exr half* files are probably already in fate-suite, but not the float file. (rgb_scanline_float_piz_48x32.exr) Can be test with make fate-exr SAMPLES=fate-suite/

Re: [FFmpeg-devel] libavcodec/exr : fix piz uncompress on big endian

> > There is bswap dsp function that runs on array of bytes. > > Thanks for your answer New patch in attach, using bswapdsp. I only enable bswapdsp on big endian, no need for it on little endian. Martin 0001-libavcodec-exr-fix-piz-uncompress-on-big-endian.patch Description: Binary data

[FFmpeg-devel] fate/exr : add test for float negative value conversion

Hello, in attach patch to add a fate test for negative float value Need to be apply after patch in discussion : Fate/exr : add test for piz uncompress libavcodec/exr : fix float to uint16 conversion for negative value Sample can be found here https://we.tl/HWgF0SFEll and need to be put inside f

[FFmpeg-devel] libavcodec/exr : fix float to uint16 conversion for negative value

Hello, In attach a patch who fix this sample : https://we.tl/HWgF0SFEll This sample contain negative value in the red channel. Before this patch, negative red float are converted to max red instead of zero. The previous hack doesn't seems to work for float to uint so i add an explicit sign chec

Re: [FFmpeg-devel] libavcodec/exr : fix float to uint16 conversion for negative value

> > did you try to make exp int32_t to avoid the 2nd check ? > > [...] > New patch in attach, with exp (and v) in int32_t Also fix the sample for me Martin 0001-libavcodec-exr-fix-float-to-uint16-conversion-for-ne.patch Description: Binary data ___ f

Re: [FFmpeg-devel] libavcodec/exr : fix piz uncompress on big endian

> > Tested-by: michael on qemu mips > > Ping for apply this patch (and the two fate tests) Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] fate/exr : add test for Y, B44A negative, datawindow != display window

Hello, In attach a patch to add fate tests for exr samples can be found here https://we.tl/ItuIX0BMfk and need to be put inside fate-suite/exr can be test with make fate-exr SAMPLES=fate-suite/ Theses tests increase coverage of exr.c : - rgb_b44a_half_negative_4x4 : test negative half value i

Re: [FFmpeg-devel] fate/exr : add test for Y, B44A negative, datawindow != display window

2017-05-01 14:31 GMT+02:00 Martin Vignali : > Hello, > > In attach a patch to add fate tests for exr > > samples can be found here > https://we.tl/ItuIX0BMfk > > and need to be put inside fate-suite/exr > > can be test with make fate-exr SAMPLES=fate-suite/ > &g

Re: [FFmpeg-devel] libavcodec/exr : cosmetics : rename variable

2017-03-25 13:08 GMT+01:00 Martin Vignali : > > Hello, > > In attach a patch who rename tile variable to better follow ffmpeg coding > style > Ping for apply ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.or

[FFmpeg-devel] libavcodec/exr : simplify reorder_pixels

->current_channel_offset and s->current_channel_offset is increment by 2 or 4 (following pixel_type) Comments welcome Martin Jokyo Images -- Martin Vignali Jokyo Images 18 rue du Transvaal 69008 Lyon 06 99 89 33 30 m.vign...@jokyo-images.com http://www.jokyo-images.com/ 0001-libavcodec-exr-si

Re: [FFmpeg-devel] fate/exr : add test for Y, B44A negative, datawindow != display window

2017-05-05 4:22 GMT+02:00 Michael Niedermayer : > On Mon, May 01, 2017 at 02:31:28PM +0200, Martin Vignali wrote: > > Hello, > > > > In attach a patch to add fate tests for exr > > > > samples can be found here > > https://we.tl/ItuIX0BMfk > &

Re: [FFmpeg-devel] libavcodec/exr : simplify reorder_pixels

2017-05-05 22:28 GMT+02:00 Martin Vignali : > Hello, > > in attach a patch who simplify the reorder_pixels func > > reorder_pixels is call by rle_uncompress and zip_uncompress > with size == uncompress_size > > uncompress_size is a multiple of 2 (because exr store data >

Re: [FFmpeg-devel] libavcodec/exr : cosmetics : rename variable

2017-05-05 22:32 GMT+02:00 Martin Vignali : > 2017-03-25 13:08 GMT+01:00 Martin Vignali : > >> >> Hello, >> >> In attach a patch who rename tile variable to better follow ffmpeg coding >> style >> > > Ping for apply > ping for apply _

Re: [FFmpeg-devel] [PATCH] Add FITS Decoder

Hello, You can create a FipsContext struct to store decoding data instead of using Fips Header struct (you can take a look to other image decoder) Also you should need to split the patch for libavcodec part and libavformat part Nit : The coding style for if/else is if { } else { } Not if {

[FFmpeg-devel] Maintainers : add myself for exr

Hello, Following Michael Niedermayer request, patch in attach to add myself for exr decoder Martin 0001-Maintainers-add-myself-for-exr.patch Description: Binary data ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/list

Re: [FFmpeg-devel] libavcodec/huffyuvdsp(enc) : add avx2 version for add(diff)_int16

2017-10-22 0:26 GMT+02:00 Martin Vignali : > Hello, > > In attach patch to add avx2 version for huffyuv dsp and huffyuvdsp enc > for add_int16 and diff_int16 func > > Check asm result for add_int16 (Kaby Lake, os 10.12) > add_int16_128_c: 1607.9 > add_int16_128_sse2: 44

Re: [FFmpeg-devel] libavcodec/utvideodsp : add avx2 version

2017-10-25 21:53 GMT+02:00 Martin Vignali : > > > 2017-10-22 14:05 GMT+02:00 Martin Vignali : > >> Hello, >> >> In attach patch to add AVX2 version for the utvideodsp >> >> Checkasm result (Kaby Lake, os 10.12) >> restore_rgb_pla

[FFmpeg-devel] remove last use of deprecated avcodec_get_chroma_sub_sample

Hello, patch in attach remove last use of this func replace it by av_pix_fmt_get_chroma_sub_sample, and return check remove several warning based on previous patch by Nicolas Frattaroli also remove the comment, where it's recommanded to use the deprecated func pass fate test for me Martin

[FFmpeg-devel] libavcodec/mpegvideo : remove warning introduce by previous patch

Hello, Sorry, i put the ret variable declaration in the previous patch, at the wrong place (at the start of the func, instead of inside the FF_API_DEBUG_MV part patch in attach fix warning, Martin 0001-libavcodec-mepgvideo-remove-warning-introduce-by-pre.patch Description: Binary data

Re: [FFmpeg-devel] libavcodec/mpegvideo : remove warning introduce by previous patch

2017-11-07 3:38 GMT+01:00 James Almer : > On 11/6/2017 8:05 PM, Martin Vignali wrote: > > Hello, > > > > Sorry, i put the ret variable declaration in the previous patch, at the > > wrong place (at the start of the func, instead of inside the > > FF_API_DEBUG_MV

[FFmpeg-devel] avcodec/losslessvideodsp : add avx2 for add_left_pred

Hello, patchs in attach add avx2 version for add_left_pred Pass fate test for me checkasm test (kaby lake os 10.12) with width = 1024 add_left_pred_zero_c: 1229.7 add_left_pred_zero_ssse3: 269.7 add_left_pred_zero_avx2: 217.9 0001-checkasm-llviddsp-test-return-of-add_left_pred-16 Modify the ch

[FFmpeg-devel] fate/magicyuv : add test for magicyuv encoding

Hello, Patch in attach add tests for magicyuv encoding (for the 3 prediction (left, median, gradient)) 0001-fate-magicyuv-move-decoding-test-to-magicyuv.mak-fil move magicyuvdecoding test inside a new file 0002-fate-magicyuvenc-add-test-for-magicyuv-encoding add test for encoding (based on utvid

Re: [FFmpeg-devel] [PATCH 10/11] avcodec/blockdsp: add AVX-512 version of clear_block(s)

2017-11-09 12:58 GMT+01:00 James Darnley : > From: James Darnley > > Also adjust alignment requirements where nessecary. > --- > Whether this patch is committed or not the change to 4xm.c should be > picked to > master because the alignment is wrong for the AVX version of this > function. I > as

Re: [FFmpeg-devel] [PATCH 11/11] avcodec/lossless_videodsp: add AVX-512 version of add_bytes

2017-11-09 12:58 GMT+01:00 James Darnley : > --- > libavcodec/x86/lossless_videodsp.asm| 5 + > libavcodec/x86/lossless_videodsp_init.c | 5 + > 2 files changed, 10 insertions(+) > > diff --git a/libavcodec/x86/lossless_videodsp.asm > b/libavcodec/x86/lossless_videodsp.asm > index ba4

Re: [FFmpeg-devel] [PATCH 11/11] avcodec/lossless_videodsp: add AVX-512 version of add_bytes

2017-11-09 20:37 GMT+01:00 Martin Vignali : > > > 2017-11-09 12:58 GMT+01:00 James Darnley : > >> --- >> libavcodec/x86/lossless_videodsp.asm| 5 + >> libavcodec/x86/lossless_videodsp_init.c | 5 + >> 2 files changed, 10 insertions(

Re: [FFmpeg-devel] [PATCH 09/11] avcodec/blockdsp: roll-up x86asm preprocessor loop

2017-11-09 12:58 GMT+01:00 James Darnley : > From: James Darnley > > --- > libavcodec/x86/blockdsp.asm | 11 --- > 1 file changed, 4 insertions(+), 7 deletions(-) > > diff --git a/libavcodec/x86/blockdsp.asm b/libavcodec/x86/blockdsp.asm > index 9d203df8f5..9d0e8a3242 100644 > --- a/liba

Re: [FFmpeg-devel] [PATCH 08/11] avcodec/v210enc: add AVX-512 10-bit line pack function

2017-11-09 12:58 GMT+01:00 James Darnley : > --- > libavcodec/x86/v210enc.asm| 5 + > libavcodec/x86/v210enc_init.c | 7 +++ > 2 files changed, 12 insertions(+) > > diff --git a/libavcodec/x86/v210enc.asm b/libavcodec/x86/v210enc.asm > index 965f2bea3c..5068af27f8 100644 > --- a/libav

Re: [FFmpeg-devel] [PATCH 08/11] avcodec/v210enc: add AVX-512 10-bit line pack function

2017-11-10 22:13 GMT+01:00 James Darnley : > On 2017-11-10 14:32, James Darnley wrote: > > I mentioned previously that using ZMM registers will cause the CPU to > > reduce its frequency. > > > > Gramner said on IRC that a user should spend 20-30% of time in > > AVX-512/ZMM code for it to be a net

Re: [FFmpeg-devel] fate/magicyuv : add test for magicyuv encoding

2017-11-07 21:52 GMT+01:00 Paul B Mahol : > On 11/7/17, Martin Vignali wrote: > > Hello, > > > > Patch in attach add tests for magicyuv encoding (for the 3 prediction > > (left, median, gradient)) > > > > 0001-fate-magicyuv-move-decoding-test-to-magicyuv

Re: [FFmpeg-devel] libavcodec/hap : add HapAlphaOnly decoding/encoding

2017-10-16 16:37 GMT+02:00 Tom Butterworth : > > >> Patches 0001, 0002, 0006 and 0007 LGTM and are uncontentious. As they > are > >> required for Hap Q Alpha support I will commit these shortly, assuming > >> nobody objects. > >> > >> > > Ok for that, so we can discuss in parallel about HAPQAlpha

Re: [FFmpeg-devel] libavcodec/hapdec : add support for hapqa decoding

2017-10-21 21:37 GMT+02:00 Carl Eugen Hoyos : > 2017-10-21 21:32 GMT+02:00 Martin Vignali : > > 2017-10-21 21:23 GMT+02:00 Carl Eugen Hoyos : > > > >> 2017-10-21 19:35 GMT+02:00 Martin Vignali : > >> > Hello, > >> > > >> > I split the

[FFmpeg-devel] avcodec/hapqa_extract_bsf : add bsf filter for haqqa (to hapq or hapalpha only) conversion (WIP)

Hello, In attach patch to add a new bitstream filter The goal is to convert HAPQA file to HAPQ (removing alpha) or HAPAlphaOnly (remove rgb) HAPQA data, is separate in two part, one for RGB data and one for alpha data so we can make the conversion without losses, by copying the right part. curr

Re: [FFmpeg-devel] avcodec/hapqa_extract_bsf : add bsf filter for haqqa (to hapq or hapalpha only) conversion (WIP)

2017-11-14 1:26 GMT+01:00 Carl Eugen Hoyos : > 2017-11-13 22:43 GMT+01:00 Martin Vignali : > > > In attach patch to add a new bitstream filter > > > > The goal is to convert HAPQA file to HAPQ (removing alpha) > > or HAPAlphaOnly (remove rgb) > > Could you ex

Re: [FFmpeg-devel] [PATCH] avfilter: add normalize filter

Hello, > Maybe there's some other way to send a patch (base64, attached zip file, > ???) > > > https://ffmpeg.org/git-howto.html#Preparing-a-patchset try git format-patch origin/master and put in attach the file(s) Martin __

[FFmpeg-devel] avcodec/hapdec : use gray8 for HapAlphaOnly decoding instead of RGB0

Hello, Following previous discussion patch in attach change pix_fmt for hap alpha only decoding to use gray8 instead of RGB0 0005-avcodec-texturedsp-add-rgtc1u-gray-decoding.patch add rgtc1u_gray func in order to decode in a gray8 picture 0006-avcodec-hapdec-use-gray8-for-HapAlphaOnly-decoding

Re: [FFmpeg-devel] fate/hapdec : add test for hap alpha only

2017-09-28 21:53 GMT+02:00 Martin Vignali : > > > 2017-09-24 11:53 GMT+02:00 Michael Niedermayer : > >> On Sat, Sep 23, 2017 at 09:53:45PM +0200, Martin Vignali wrote: >> > Hello, >> > >> > sample can be found here : >> > https://we.tl/cP0pW

Re: [FFmpeg-devel] fate/hapdec : add test for hap alpha only

> > New patch in attach (use gray8 pix_fmt) > Need to be apply after patch in discussion > avcodec/hapdec : use gray8 for HapAlphaOnly decoding instead of RGB0 > > Martin > With the attachment 0007-fate-hapAlphaOnly-add-test-for-hap-alpha-only-decodi.patch Description: Binary data __

Re: [FFmpeg-devel] [PATCH] dvenc: Prevent out-of-bounds read

2017-11-17 17:20 GMT+01:00 Derek Buitenhuis : > mb_area_start has 5 entries, and 'a' is iterated through from 0 to 3. > 'a2' is set to 'a + 1', and mb_area_start[a2 + 1] is accessed, so if > a is 3, then we try to access mb_area_start[5]. > > Signed-off-by: Derek Buitenhuis > --- > I'm not 100% s

Re: [FFmpeg-devel] libavcodec/huffyuvdsp(enc) : add avx2 version for add(diff)_int16

2017-11-04 19:31 GMT+01:00 Martin Vignali : > > > 2017-10-22 0:26 GMT+02:00 Martin Vignali : > >> Hello, >> >> In attach patch to add avx2 version for huffyuv dsp and huffyuvdsp enc >> for add_int16 and diff_int16 func >> >> Check asm result for ad

Re: [FFmpeg-devel] libavcodec/utvideodsp : add avx2 version

2017-11-04 19:33 GMT+01:00 Martin Vignali : > > > 2017-10-25 21:53 GMT+02:00 Martin Vignali : > >> >> >> 2017-10-22 14:05 GMT+02:00 Martin Vignali : >> >>> Hello, >>> >>> In attach patch to add AVX2 version for the utvideodsp >>

Re: [FFmpeg-devel] libavcodec/hapdec : add support for hapqa decoding

2017-11-13 20:06 GMT+01:00 Martin Vignali : > > > 2017-10-21 21:37 GMT+02:00 Carl Eugen Hoyos : > >> 2017-10-21 21:32 GMT+02:00 Martin Vignali : >> > 2017-10-21 21:23 GMT+02:00 Carl Eugen Hoyos : >> > >> >> 2017-10-21 19:35 GMT+02:00 Martin Vign

Re: [FFmpeg-devel] libavcodec/utvideodsp : add avx2 version

Hello, > > If noone reviews, and you tested&reviewed it then it should be ok to > apply Ok will apply AVX2 for utvideodsp, huffyuv(enc)dsp and hapqa decoding (and fate) > especially considering you waited a month (which is longer than > needed generally) > > > > What is the recommanded time, t

Re: [FFmpeg-devel] libavcodec/utvideodsp : add avx2 version

Hello, > > > Checkasm result (Kaby Lake, os 10.12) > > restore_rgb_planes_c: 8371.0 > > restore_rgb_planes_sse2: 6583.7 > > restore_rgb_planes_avx2: 3596.5 > > > > restore_rgb_planes10_c: 16735.7 > > restore_rgb_planes10_sse2: 11478.5 > > restore_rgb_planes10_avx2: 7193.7 > > Curious, on my Haswel

[FFmpeg-devel] checkasm/utvideo : be more explicit for WIDTH_PADDED define

Hello, Patch in attach Martin 0001-checkasm-utvideo-be-more-explicit-to-the-WIDTH_PADDE.patch Description: Binary data ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] avcodec/x86/exrdsp : use ymm constant for pb_80 instead of vbroadcasti128

Hello, After patch by James Almer (pb_80 now fit an ymm) The two mode (SSE, AVX2) for constant loading can be remove speed seems to be similar to me Martin 0002-avcodec-x86-exrdsp-use-ymm-constant-for-pb_80.patch Description: Binary data ___ ffmpeg-

Re: [FFmpeg-devel] avcodec/x86/exrdsp : use ymm constant for pb_80 instead of vbroadcasti128

2017-11-22 18:21 GMT+01:00 James Almer : > On 11/21/2017 6:09 PM, Martin Vignali wrote: > > Hello, > > > > After patch by James Almer > > (pb_80 now fit an ymm) > > > > The two mode (SSE, AVX2) for constant loading can be remove > > > > spee

Re: [FFmpeg-devel] avcodec/hapdec : use gray8 for HapAlphaOnly decoding instead of RGB0

2017-11-23 3:46 GMT+01:00 Carl Eugen Hoyos : > 2017-11-16 23:41 GMT+01:00 Martin Vignali : > > Hello, > > > > Following previous discussion > > patch in attach change pix_fmt for hap alpha only decoding to use gray8 > > instead of RGB0 > > Looks like what y

Re: [FFmpeg-devel] fate/hapdec : add test for hap alpha only

> > LGTM > > thx > > > > Pushed, thanks Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] avcodec/x86/bswapdsp : convert pb_bswap32 to ymm constant in order to simplify code

Hello, In attach patch to convert pb_bswap32 to ymm constant and remove the vbroadcasti128 part Speed seems to be similar to me Martin 0004-avcodec-x86-bswapdsp-convert-pb_bswap32-to-ymm.patch Description: Binary data ___ ffmpeg-devel mailing list f

[FFmpeg-devel] avcodec/x86/lossless_videodsp : add_left_pred AVX2 v2

Hello, New patchs in attach in order to add avx2 version for add_left_pred Change since the v1 patch - use ymm constant - use 3 operandes mode Check asm result add_left_pred_rnd_acc_c: 1279.8 add_left_pred_rnd_acc_ssse3: 261.3 add_left_pred_rnd_acc_avx2: 209.8 add_left_pred_zero_c: 1284.8 add_l

[FFmpeg-devel] fate/hap : add test for hap encoding

Hello, Patch in attach, add test for hap encoding (currently not cover) (patch 002) and move decoding tests to a separate file (patch 001) decoding can be test with make fate-hap SAMPLES=fate-suite/ and encoding can be test with make fate-hapenc SAMPLES=fate-suite/ Hap encoding need ffmpeg comp

[FFmpeg-devel] avcodec/huffyuvenc : try to call dsp with aligned data, and remove code duplication

Hello, in attach patchs 0001-avcodec-huffyuvenc-increase-scalar-loop-count and 0003-avcodec-huffyuvenc-sub_left_prediction_bgr32-call-ds like diff_bytes and diff_bytes16, have AVX2 version, increase the scalar loop to call the aligned version in most case 0002-avcodec-huffyuvenc-remove-code-d

[FFmpeg-devel] avcodec/utvideodec : use dsp add_median_pred for second line

Hello, Patch in attach dsp func need align16 data make only the start of the line in scalar, and call the dsp for the rest instead of process the entire line in scalar pass make fate-utvideo for me Martin 0001-avcodec-utvideodec-use-dsp-add_median_pred-for-secon.patch Description: Binary dat

[FFmpeg-devel] avcodec/utvideodec : add x86 SIMD (SSSE3) for gradient prediction

Hello, Patch in attach add SIMD (SSSE 3) for gradient prediction and a checkasm test Checkasm result (width = 1024) (kaby lake, macos 10.12) add_gradient_pred_c: 1708.8 add_gradient_pred_ssse3: 533.0 Benchmark on a 3 min HD File in gradient (422) without SIMD : bench: utime=102.695s bench: maxrs

Re: [FFmpeg-devel] avcodec/x86/bswapdsp : convert pb_bswap32 to ymm constant in order to simplify code

2017-11-27 17:59 GMT+01:00 Henrik Gramner : > On Sat, Nov 25, 2017 at 9:53 PM, Martin Vignali > wrote: > > Hello, > > > > In attach patch to convert pb_bswap32 to ymm constant > > and remove the vbroadcasti128 part > > > > Speed seems to be similar to m

[FFmpeg-devel] avutil/x86util : add macro for 128 bits constant load

Hello Following suggestion by Henrik Gramner in attach a patch to add a macro in x86_utils.asm in order to load a 128 bits constantes in an XMM register or in each part of a ZMM register Not sure about the name of this macro, and the position in the x86utils file Patch 002 : Use this new macro,

Re: [FFmpeg-devel] avutil/x86util : add macro for 128 bits constant load

2017-11-28 21:04 GMT+01:00 Henrik Gramner : > On Mon, Nov 27, 2017 at 11:37 PM, James Almer wrote: > > On 11/27/2017 7:33 PM, James Darnley wrote: > >> If the condition was made "mmsize > 16" would this work correctly for > >> zmm registers? (Assume I finally push my AVX-512 patches). > > > > No

Re: [FFmpeg-devel] avcodec/huffyuvenc : try to call dsp with aligned data, and remove code duplication

2017-11-26 19:07 GMT+01:00 Martin Vignali : > Hello, > > in attach patchs > > 0001-avcodec-huffyuvenc-increase-scalar-loop-count > and > 0003-avcodec-huffyuvenc-sub_left_prediction_bgr32-call-ds > > like diff_bytes and diff_bytes16, have AVX2 version, increase the

Re: [FFmpeg-devel] [PATCH] avfilter/vf_threshold: add x86 SIMD

Hello, + > +INIT_XMM sse4 > Maybe use a macro (AVX2 can probably, be easy to add) > +cglobal threshold8, 10, 13, 5, in, threshold, min, max, out, ilinesize, > tlinesize, flinesize, slinesize, olinesize, w, h, x > +mov wd, dword wm > +mov hd, dword hm > Maybe you can use p

Re: [FFmpeg-devel] [PATCH] avfilter/vf_threshold: add x86 SIMD

> >> > > Do you need pxor m0, m4 and pxor m1, m4 ? > > Yes, > I need it. > > Yes, you're right, sorry for the noise Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] avfilter: add hflip x86 SIMD

> + > +%include "libavutil/x86/x86util.asm" > + > +SECTION_RODATA > + > +pb_flip_byte: times 16 db 15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0 > +pb_flip_short: times 16 db 14,15,12,13,10,11,8,9,6,7,4,5,2,3,0,1 > + > times 16 ? Martin ___ ffmpeg-devel mailin

Re: [FFmpeg-devel] avutil/x86util : add macro for 128 bits constant load

2017-12-02 13:13 GMT+01:00 Henrik Gramner : > On Fri, Dec 1, 2017 at 9:03 PM, Martin Vignali > wrote: > > If no one have objections, i will push these patch tomorrow. > > > > Martin > > Follow James' suggestion to use >16 instead of ==32, otherwise OK. &g

Re: [FFmpeg-devel] avcodec/huffyuvenc : try to call dsp with aligned data, and remove code duplication

> > requiring FFMIN() to be evaluated per iteration could be slower > if the compiler fails to factor it out > > > > New patchs in attach : 001 : unchanged 002 : add "int min_width = FFMIN(w, 32)" at the start of the func 003 : add "int min_width = FFMIN(w, 8)" at the start of the func Pass fate

Re: [FFmpeg-devel] avcodec/x86/lossless_videodsp : add_left_pred AVX2 v2

New patch in attach 001, 002 : unchanged 003 : use VBROADCASTI128 macro for constant loading en XMM/YMM instead of 256 bits constants. Martin 0001-checkasm-llviddsp-test-return-of-add_left_pred-16.patch Description: Binary data 0002-avcodec-x86-lossless_videodsp.asm-make-macro-for.patch Desc

[FFmpeg-devel] avcodec/utvideodec : add SIMD (SSSE3 and AVX2) for gradient_pred (V2)

Hello, New patchs in attach for adding gradient pred SIMD (SSSE3 and AVX2) (use by utvideo dec now (more use will be add later)) Checkasm result (width = 1024) add_gradient_pred_c: 2070.2 add_gradient_pred_ssse3: 602.4 add_gradient_pred_avx2: 385.7 Need to be apply after add_left_pred AVX2 pat

[FFmpeg-devel] avfilter/vf_threshold : add checkasm and avx2 version for threshold8

Hello, In attach patch to add a checkasm test for the recently added threshold8 SIMD also add AVX2 version Checkasm result : threshold8_c: 584.8 threshold8_sse4: 65.0 threshold8_avx2: 43.5 001 : create ff_threshold_init_func in order to simplify checkasm write (more like previous checkasm test

Re: [FFmpeg-devel] [PATCH] avfilter: add hflip x86 SIMD

Hello, Maybe you can use a macro for byte and short version, only few lines are different in each version Martin ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] avfilter/vf_threshold : add checkasm and avx2 version for threshold8

> > > 002 : Add checkasm test for vf_threshold > > Why is this GPL? > > Because i copy paste the header of vf_blend checkasm Will change to LGPL. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] avfilter: add hflip x86 SIMD

2017-12-03 17:46 GMT+01:00 Paul B Mahol : > On 12/3/17, Martin Vignali wrote: > > Hello, > > > > Maybe you can use a macro for byte and short version, > > only few lines are different in each version > > Sure, feel free to send patches. > > I'm not

Re: [FFmpeg-devel] avfilter/vf_threshold : add checkasm and avx2 version for threshold8

> > >> > 002 : Add checkasm test for vf_threshold >> >> Why is this GPL? >> >> Because i copy paste the header of vf_blend checkasm > Will change to LGPL. > > in fact most(all ?) checkasm test are under GPL licence. and the checkasm exe, is under GPL too. So i will let the current GPL licence Mar

Re: [FFmpeg-devel] [PATCH] avfilter: add hflip x86 SIMD

> 2017-12-03 17:46 GMT+01:00 Paul B Mahol : > >> On 12/3/17, Martin Vignali wrote: >> > Hello, >> > >> > Maybe you can use a macro for byte and short version, >> > only few lines are different in each version >> >> Sure, feel free to sen

Re: [FFmpeg-devel] avfilter/vf_threshold : add checkasm and avx2 version for threshold8

2017-12-03 18:01 GMT+01:00 Paul B Mahol : > On 12/3/17, Martin Vignali wrote: > >> > > >>> > 002 : Add checkasm test for vf_threshold > >>> > >>> Why is this GPL? > >>> > >>> Because i copy paste the header of vf_b

[FFmpeg-devel] avfilter/x86/vf_threshold : add SSE4 and AVX2 for threshold 16

Hello, Patch in attach add SIMD for threshold16 Checkasm result : threshold16_c: 304.5 threshold16_sse4: 60.5 threshold16_avx2: 45.0 001 : modify threshold macro, and add threshold 16 002 : add checkasm test for threshold16 Martin 0001-avfilter-x86-vf_threshold-add-threshold16-SIMD-SSE4-.p

Re: [FFmpeg-devel] [PATCH] avfilter: add hflip x86 SIMD

> Can you post a disassembly of hflip_byte_c? > > > in O1 : clang -S -O1 test_asm_gen.c .section__TEXT,__text,regular,pure_instructions .macosx_version_min 10, 12 .globl_hflip_byte_c .p2align4, 0x90 _hflip_byte_c: ## @hflip_byte_c .cfi_start

Re: [FFmpeg-devel] [PATCH] avfilter: add hflip x86 SIMD

> > In any case, if clang or gcc can generate better code, then the hand > written version needs to be optimized to be as fast or faster. > > > Quick test : pass checkasm (but probably only because width = 256) hflip_byte_c: 26.4 hflip_byte_ssse3: 20.4 INIT_XMM ssse3 cglobal hflip_byte, 3, 5, 2,

<    1   2   3   4   5   6   >