[Ffmpeg-devel-irc] ffmpeg-devel.log.20171009

burek burek021 at gmail.com
Tue Oct 10 03:05:03 EEST 2017


[01:50:41 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:9c7a71145b88: configure: Add config option for libva2 (VAAPI 1)
[01:50:42 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:bd211bb866f8: vaapi: Remove H.264 baseline profile
[01:50:43 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:e339411691e3: vaapi: Always free parameter buffers after vaEndPicture() with libva2
[01:50:44 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:5f3978866879: hwcontext_vaapi: Factorise out common connection code
[01:50:45 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:f3602875b325: hwcontext_vaapi: Set message callbacks on internally-created devices
[01:50:46 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:309d660775e2: hwcontext: Perform usual initialisation on derived device contexts
[01:50:47 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:b2f256a9f5db: hwcontext_vaapi: Add support for mapping to DRM objects
[01:59:36 CEST] <cone-284> ffmpeg 03Mark Thompson 07master:c154c34c1b37: Revert "configure: Disable -Wbool-operation."
[02:24:12 CEST] <cone-284> ffmpeg 03Carl Eugen Hoyos 07master:c585b4e9f0b3: configure: Disable -Wbool-operation using check_disable_warning().
[03:01:56 CEST] <cone-284> ffmpeg 03Carl Eugen Hoyos 07master:14b15539398f: lavf/adp: Fix the probe function on systems with signed char.
[09:58:53 CEST] <durandal_170> how new ffmpeg release will be named?
[13:35:45 CEST] <jkqxz> wm4:  That's the encoder, not the decoder.  And yes, it does in the decoder (unless you have the allow-profile-mismatch flag to override it).
[13:37:53 CEST] <wm4> ooh
[13:38:12 CEST] <wm4> that should be fine then
[15:56:16 CEST] <cone-068> ffmpeg 03foo86 07master:23990950e35e: avcodec/dca_parser: revert to conservative sync distance estimation
[15:56:16 CEST] <cone-068> ffmpeg 03foo86 07master:73789b85a759: avcodec/dca_core: always limit frame size to data size
[17:06:48 CEST] <cone-068> ffmpeg 03bnnm 07master:1954e625b1d6: avcodec/wmaprodec: support multichannel XMA stream configurations
[19:57:53 CEST] <BBB> wbs: http://www.bbc.com/news/technology-41551546 < does that mean the presentation @ VDD was futile? :(
[19:59:10 CEST] <wm4> wasn't this long known
[20:00:02 CEST] <jkqxz> Aren't we still expecting desktop-Windows-on-aarch64 real-soon-now?
[20:00:06 CEST] <JEEB> ^
[20:00:24 CEST] <JEEB> I thought the presentation was mostly about the non-mobile aarch64
[20:00:35 CEST] <BBB> desktop-win-on-aarch64?
[20:00:37 CEST] <BBB> is that a thing?
[20:01:00 CEST] <JEEB> yea, binaries have already poked their head up
[20:01:33 CEST] <JEEB> not a full image yet though, just additional binaries that wbs found :) and then Certain People seem to have SDKs
[20:02:11 CEST] <JEEB> BBB: after all, they did already have an armv7 thing for quite a while
[20:03:20 CEST] <jamrial> does amd still have plans for their arm stuff, or are they focused on ryzen?
[20:04:45 CEST] <wm4> (was the ryzen bug with gcc compilation found yet?)
[20:05:23 CEST] <wm4> would be nice to try amd CPU+GPU instead of the old intel+nvidia crap
[20:05:34 CEST] <wm4> (nvidia still don't support wayland properly)
[20:05:51 CEST] <BtbN> There was no Ryzen bug with gcc i think
[20:05:52 CEST] <jkqxz> They haven't talked about it at all in the last year.  I assume it's dead.
[20:05:53 CEST] <BtbN> just broken chips
[20:06:15 CEST] <jkqxz> (The AMD custom ARM core.)
[20:06:16 CEST] <BtbN> jkqxz, they recently published a draft of the new API and had a whole conference about it.
[20:06:46 CEST] <BtbN> wm4, now if only AMD would sell useable GPUs...
[20:06:59 CEST] <wm4> I thought vega or something was good
[20:07:03 CEST] <BtbN> But everything they offer is inferior to what nvidia has for the same price
[20:07:06 CEST] <BtbN> lol, Vega is a joke
[20:07:14 CEST] <wm4> what new API?
[20:07:22 CEST] <jamrial> BtbN: how so?
[20:07:24 CEST] <BtbN> Vega is more expensive than a faster nvidia card, and uses twice as much power.
[20:08:10 CEST] <wm4> well, can't comment on that
[20:08:12 CEST] <jamrial> i think i read that undervolting helped vega 56 in a lot of scenarios
[20:08:17 CEST] <wm4> what does faster mean? game benchmarks?
[20:08:31 CEST] <BtbN> Overall benchmarks
[20:08:47 CEST] <wm4> nvidia almost surely have better GL drivers
[20:08:48 CEST] <jkqxz> BtbN:  New API for what, sorry?
[20:08:51 CEST] <wm4> but fortunately vulkan exists
[20:08:52 CEST] <BtbN> Wayland
[20:09:07 CEST] <wm4> if you mean nvidia's eglstream idea for wayland, please no
[20:09:08 CEST] <BtbN> The Vega64 is roughly somewhere between a GTX1070 and a GTX1080, but uses over 2 times the power a 1080 uses.
[20:09:25 CEST] <atomnuker> that's why its a monster and I want one
[20:09:40 CEST] <BtbN> just so you don't need a space heater?
[20:09:47 CEST] <atomnuker> though I guess I could settle for an rx460 too, that thing is a monster too
[20:09:50 CEST] <jamrial> it's not like you can buy a vega anyway. crypto currency miners made sure of that
[20:10:01 CEST] <wm4> lol
[20:10:30 CEST] <jkqxz> BtbN:  For what, though?  It's significantly better than that for compute (including mining, of course).
[20:10:38 CEST] <wm4> btw. nvidia's power management is persistently broken
[20:10:57 CEST] <wm4> I can't play video or play a game from 1996 without stuttering, unless I set the card to high perf mode
[20:12:19 CEST] <jamrial> atomnuker: i wouldn't get a rx460, be it gaming or video decoding. amd starts being a good purchase with the rx470/570
[20:12:30 CEST] <jamrial> if you can get them at a decent price because, again, miners
[20:15:05 CEST] <atomnuker> I like how amazon is ridiculously biased here, you search for rx570 and you get a few results with the rest being nvidia gtx 1050 or some other nvidia card
[20:15:19 CEST] <atomnuker> but if you look for nvidia its pages and pages of nvidia without a single amd card
[20:15:38 CEST] <BtbN> The algorithm favors what people actually end up buying
[20:19:00 CEST] <atomnuker> true, most are windows users who only play minecraft and don't mind installing +1gb drivers with fancy guis
[20:19:38 CEST] <BtbN> The Windows-Drivers for AMD and NVidia are equally horrible
[20:19:50 CEST] <BtbN> I prefer the nvidia one, but they are both horribly overloaded
[20:20:59 CEST] <atomnuker> I bet they look horrible too, just conservative spreadsheets of options, back in the day this was norma - http://img.hexus.net/v2/sff/biostar/200t/Egg.jpg
[20:21:05 CEST] <atomnuker> *normal
[20:22:04 CEST] <wm4> wat
[20:22:21 CEST] <atomnuker> actually nevermind, http://www.guru3d.com/articles-pages/geforce-gtx-1080-overclocking-guide-with-afterburner-4-3,2.html still shows overclock programs look crazy
[20:22:34 CEST] <BtbN> Dealt with the AMD one at work recently. They are horrible. It took me over 10 minutes to find where to rotate the screen
[20:22:56 CEST] <BtbN> Afterburner is a 3rd party tool.
[20:22:57 CEST] <wm4> lol gamer shit
[20:23:03 CEST] <jamrial> BtbN: best thing was when amd removed the old .net control panel in favor of the new one
[20:23:18 CEST] <jamrial> 200mb off the installer shaved in one release
[20:23:26 CEST] <nevcairiel> afterburner is the only tool that works properly however, but their new fancy skins are always dumb
[20:24:25 CEST] <atomnuker> its a blank canvas for programmers to express their graphical design creativity, like this one here -http://i.imgur.com/9UGgO.jpg
[20:25:46 CEST] <atomnuker> (though in 90% of the cases they end up copying car manufacturers because its too hard)
[20:54:51 CEST] <wbs> BBB: as JEEB said, that should be unrelated. they're allegedly working on windows-on-arm64 for both server and laptop use
[21:10:44 CEST] <wm4> arm servers?
[21:10:50 CEST] <wbs> yes
[21:11:00 CEST] <wm4> I thought everyone recognized that as nonsense
[21:11:17 CEST] <wbs> that doesn't stop people from pouring money and effort into it
[21:11:21 CEST] <JEEB> ^
[21:33:35 CEST] <kierank> wm4: what's wrong with arm servers
[21:36:34 CEST] <jkqxz> Nothing intrinsically, but if they are going to put shitty cores running slow in the things then they better have fuckloads of them.  This is a thing that ARM server vendors have historically failed at.
[21:37:08 CEST] <jkqxz> (And is why Cavium is basically the only one worth considering from those currently around.)
[21:37:59 CEST] <wm4> kierank: so far arm seems to have a good reputation for low energy use, but not exactly for being very powerful?
[21:38:14 CEST] <kierank> sure, but ok for most webserving and crap, no
[21:38:22 CEST] <wm4> and I don't think anyone ever tried to build servers with Atoms (when they were alive)
[21:52:31 CEST] <nevcairiel> actually such things exist
[21:53:07 CEST] <nevcairiel> they use Atom SoCs with high core count (16 or so) for NAS servers
[21:53:28 CEST] <JEEB> yea, 8+
[21:53:43 CEST] <nevcairiel> not every server necessarily needs high cpu performance
[21:54:09 CEST] <wm4> yeah, but there are ARM NASes too
[21:54:31 CEST] <kierank> my nas is atom based
[21:54:43 CEST] <JEEB> atom > ARM in various cases
[21:54:53 CEST] <jkqxz> That's really just because of the rest of the things on the SoC.  (E.g. <https://ark.intel.com/products/97927/> -> 16 SATA ports, 4x10GbE.)
[21:55:24 CEST] <llogan> speaking of nas, is synology still distributing a modified, nonfree ffmpeg?
[21:55:43 CEST] <TD-Linux> socionext has a new 24-core cortex-a53 board on 96boards
[00:00:00 CEST] --- Tue Oct 10 2017


More information about the Ffmpeg-devel-irc mailing list