[ffmpeg] Backport of armv6 and vfp optmisations for DTS #3016
Conversation
|
fine be me. @aballier any comments? |
|
needs testing under iOS since it uses ASLR and this will cause issues if symbols are not handled correctly. |
|
fine by me; I don't know much about arm asm and since those come from upstream they should be fine |
|
don't these patches need to be added in the patch file folder too? Edit: |
|
yes since they are likely not in the next target ffmpeg. On Mon, Jul 29, 2013 at 8:40 PM, Martijn Kaijser
|
|
@elupus Can anyone test this on iOS or android? jenkins build this please |
|
reminder, this does NOT go in until tested under iOS. |
|
Understood. Jenkins did give a failure on ATV2 (although it disappeared with the rebase to bump the patch numbers). |
|
Hmmm. The error reported by jenkins: http://jenkins.xbmc.org/job/XBMC-IOS-ATV2/273/console
doesn't seem to match the code. Do we have a way to view files on jenkins? Just in case the rebase (renumbering of the patches, which shoudn't be used in build) caused a build issue, I'll try again. jenkins build this please |
|
You can watch the workspace of a Job/Slave here http://jenkins.xbmc.org/job/XBMC-IOS-ATV2/ws/ Though this is the workspace of the current Build now - not of this Pr (you need to Look at it when it is Building...) |
|
Same errors again. Can anyone offer an explanation as to what's wrong on ATV2 build? |
|
lib/ffmpeg/patches/0045-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-i.patch this patch adds a reference to TCOS_D0_HEAD |
|
@popcornmix - please try there newest gas-preprocessor.pl from here: http://git.libav.org/?p=gas-preprocessor.git;a=log just stick it in tools/depends/native/gas-preprocessor-native and redo a jenkins build with your branch afterwards. |
|
Trying with updated gas-preprocessor.pl |
|
Now its going crazy on droid and ios/atv2. Someone hasn't altered the xcode project files for all darwin platforms as it seems. For android i don't know what the issue is: "cp: cannot stat `xbmc/xbmc.d': No such file or directory Nevertheless - the gas-preprocessor fixed the ffmpeg compile for ios ... |
|
As is see davilla fixed the xcode project. A rebase to master might fix the build for ios/atv2. |
|
@Memphiz this i get sometimes under linux too for some time now, removing the sourcedir and rebuild helps in this case (here) |
…r_float
Before After
Mean StdDev Mean StdDev Change
This function 9295.0 114.9 4853.2 83.5 +91.5%
Overall 23699.8 397.6 19285.5 292.0 +22.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
…oat_fmul_scalar
Before After
Mean StdDev Mean StdDev Change
This function 1175.0 4.4 366.2 18.3 +220.8%
Overall 19285.5 292.0 18420.5 489.1 +4.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
…ul_array8 This is similar to int32_to_float_fmul_scalar, but loads a new scalar multiplier every 8 input samples. This enables the use of much larger input arrays, which is important for pipelining on some CPUs (such as ARMv6). Signed-off-by: Martin Storsjö <martin@martin.st>
Signed-off-by: Martin Storsjö <martin@martin.st>
…oat_fmul_array8
Before After
Mean StdDev Mean StdDev Change
This function 366.2 18.3 277.8 13.7 +31.9%
Overall 18420.5 489.1 17049.5 408.2 +8.0%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 2653.0 28.5 1108.8 51.4 +139.3%
Overall 17049.5 408.2 15973.0 223.2 +6.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 868.2 33.5 436.0 27.0 +99.1%
Overall 15973.0 223.2 15577.5 83.2 +2.5%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 1389.3 4.2 967.8 35.1 +43.6%
Overall 15577.5 83.2 15400.0 336.4 +1.2%
Signed-off-by: Martin Storsjö <martin@martin.st>
This does most of the work formerly carried out by the static function qmf_32_subbands() in dcadec.c. Signed-off-by: Martin Storsjö <martin@martin.st>
…ands
Before After
Mean StdDev Mean StdDev Change
This function 1323.0 98.0 746.2 60.6 +77.3%
Overall 15400.0 336.4 14147.5 288.4 +8.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
…p assembly files Reviewed-by: Kostya Shishkov Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
|
Rebased. |
|
@sraue - shouldn't be an issue on jenkins though as we git clean -xfd the tree before building (in some cases without depends though ...) |
|
wow, gas-preprocessor.pl got bumped :) |
|
Woo-hoo! Do you want the gas-preprocessor.pl bump in a separate PR, or is it okay here? |
|
It is ok in here imo. I pulled the debs for ios and atv2 from the buildslave and will upload it on my dropbox in a min for willing testers ... (not sure if i myself have time to try it). @davilla do you think this one has potential to panic the atv2 kernel again? ^^ |
|
@Memphiz , until it gets tested, I worry above any asm changes in ffmpeg. it's just so picky. |
|
testing ... |
|
This is good to go for atv2 and ios. I've tested DTS 2 stereo decode and it worked flawless. Tough DTS passthrough shouldn't be affected by this - i confirmed it is still working too. |
[ffmpeg] Backport of armv6 and vfp optmisations for DTS
|
Thanks for testing Memphiz. |
There have been a sequence of optimisations accepted to upstream libavcodec/ffmpeg designed to make software decode of DTS run better on the Raspberry Pi.
Since DTS is a floating point codec, the optimisations mostly rely upon hand-scheduled VFP code and the use
of short vectors.
Overall this reduces the time spent in DTS by about 36%. This makes software DTS decode on Pi much more feasible without overclocking.
These patches can be dropped when ffmpeg is updated to a sufficiently new version. These patches aren't present in v.2.0, which may be the next logical upgrade step for ffmpeg, so I'd like to get these cherry-picked, rather than waiting for an ffmpeg bump.
The patches have been present in rbej builds for a few weeks and no regressions have been reported.