Hi10P Support Proposed For VDPAU

[Vincenzov]

Hello everyone, we have news on the decoding hi10p by the gpu?
Thanks

[fritsch]

Hehe, the thread seemed to derail there somewhat. :P

Have there been any updates on this? Can anyone here explicitly confirm that they've tested it & it works? And even if it does; given that apparently nvidia hardware can't do this, is it likely that it would actually get accepted in upstream mesa?

Someone wrote me an email. Told he would be a video programmer and had lots of clue in this direction. I gave him all the Code and added what is left to do .... never heard anything back of him. I am super busy with xbmc currently and won't have any time the next months to make these patches ready for the different departments.

[aphirst]

Hehe, the thread seemed to derail there somewhat. :P

Have there been any updates on this? Can anyone here explicitly confirm that they've tested it & it works? And even if it does; given that apparently nvidia hardware can't do this, is it likely that it would actually get accepted in upstream mesa?

[Luke]

Many laptops need to be regularly blown free of dist

Like i said

But i have seen stuff like this happen. My Dell D630 had its mobo replaced, my collegue's Lenovo R61 had its mobo replaced (both were used in proper conditions, no gaming or anything and the bios didnt report very high temps) and i have seen many other laptops that had burnt chips. Thats the reality. If you have laptops that work 24/7, good for you, but thats not the norm for everyone else.

My sister's laptop was always running hot when she came over to visit, always cool after I blew out the dust from the cooler ports. I taught her how to keep it clean, now it doesn't run hot anymore. Desktops too: I've seen a Pentium 4 Prescott chip die after running an unknown amount of time with a totally blocked heatsink. Yes, I've seen a Core2 Quad survive booting with NO heatsink at all, but no doubt it was throttling way back to control heat. Long enough runtime on the threshold of issues-even below the throtting point- can speed up electromigration and finish off an old chip. If this was not true, chips overclocked right to the limit and runnning hot would last just as long as stock-clocked chips, and just as often. I do overclock, but not to the the voltage limits and always with enough cooler to stay under stock temps. Still had an Athlon 64 gradually lose speed capability over a few years, hope my FX-8120 does not do much of that.

[mmstick]

Like i said

But i have seen stuff like this happen. My Dell D630 had its mobo replaced, my collegue's Lenovo R61 had its mobo replaced (both were used in proper conditions, no gaming or anything and the bios didnt report very high temps) and i have seen many other laptops that had burnt chips. Thats the reality. If you have laptops that work 24/7, good for you, but thats not the norm for everyone else.

I'm afraid not because I've never seen this happen on any machine I've worked on in the past. It's even taught in A+ courses that such is not possible. If such is to happen, the company could be sued for having improper cooling. The only other way for that to happen is the user is obstructing the cooling to the system. Even then, machines today are designed to have the BIOS shutoff the system if temperatures are too high.

[gradinaruvasile]

Like i said

You might say it didnt happen to you

But i have seen stuff like this happen. My Dell D630 had its mobo replaced, my collegue's Lenovo R61 had its mobo replaced (both were used in proper conditions, no gaming or anything and the bios didnt report very high temps) and i have seen many other laptops that had burnt chips. Thats the reality. If you have laptops that work 24/7, good for you, but thats not the norm for everyone else.

[mmstick]

Some low end chips might be able to barely handle 1080p in software, but its not good, especially if you have a laptop, to have them running on full CPU load for hours. Some mobile cpus can get really hot and after many heating cycles the chip could fry or the soldering on the mobo fail. You might say it didnt happen to you but it does happen sometimes, especially on lower end models (of which we talk about here mainly). For example even my Dell D630 (which is a high end model) reports ~80-90 C after a few minutes of full load (core2duo@2GHz). It improves somewhat with a cooler, but it still reaches high temps. Other laptops such as Lenovo T61 do the same.

PS. Didnt know about mpv. Seems to be better than mplayer for casual viewing with its simple osd. And doesnt mess with the master pulse volume + it automatically uses software decoding if the hardware cant decode that particular codec (these 2 were my biggest beefs with mplayer). Had to adjust the config file to enable hw decoding and had to manually add mpeg2/4 codecs to the hw decode codec list to use them too. Performance wise doesnt seem to be that much different vs mplayer, in fact it seems to use a tiny bit more cpu, but mostly its about the same.

If your system gets that hot from running 100% load then it either has a faulty design or it is packed full of dust. It's actually illegal to sell systems that can get that hot. All systems should be able to sustain 100% CPU usage 24/7 without problems. For example, for many years now all of my systems use 100% 24/7 since I run distributed computing projects like BOINC and x264 encoding -- laptop included.

[gradinaruvasile]

Some low end chips might be able to barely handle 1080p in software, but its not good, especially if you have a laptop, to have them running on full CPU load for hours. Some mobile cpus can get really hot and after many heating cycles the chip could fry or the soldering on the mobo fail. You might say it didnt happen to you but it does happen sometimes, especially on lower end models (of which we talk about here mainly). For example even my Dell D630 (which is a high end model) reports ~80-90 C after a few minutes of full load (core2duo@2GHz). It improves somewhat with a cooler, but it still reaches high temps. Other laptops such as Lenovo T61 do the same.

PS. Didnt know about mpv. Seems to be better than mplayer for casual viewing with its simple osd. And doesnt mess with the master pulse volume + it automatically uses software decoding if the hardware cant decode that particular codec (these 2 were my biggest beefs with mplayer). Had to adjust the config file to enable hw decoding and had to manually add mpeg2/4 codecs to the hw decode codec list to use them too. Performance wise doesnt seem to be that much different vs mplayer, in fact it seems to use a tiny bit more cpu, but mostly its about the same.

[mmstick]

I am not sure E350 has more processing power than your Turion (uses less power, but thats another story). Turion was the top of the line mobile CPU back then for AMD.
Maybe the 1007 or the like Celerons have the oomph to decode 1080p in software, i dont know. But the E350 or the Atom certainly isnt up to the task. If the hi10p is anything like the regular 1080p, that is. Do you have a clip somewhere that we can test?

I have a SU3500 [email protected] which chokes on just about anything HD that isnt hardware decoded (it has the GMA45 video chipset which has some basic h264 decoding after i compiled libva manually). And that wasnt a cheap CPU when it was released. Also i have an atom 1.6GHz single core that barely handles anything. I installed Windows on another dual core 450 Atom that was crap also.

As long as you aren't rendering the whole video through software, as in using xv or opengl output, and your graphics chip/drivers can handle it, it should play fine. In the case of AMD, I've seen Catalyst cause severe lag with high resolution videos when the open source drivers play perfectly fine. It also helps to use something like mpv instead of VLC since it uses much less CPU for decoding, though I use mpv instead of VLC anyway since I've been encoding my stuff into MKV + 10-bit H.264 + Opus audio these days which VLC still can't properly decode. On my FX-8120 desktop, decoding a 1080p 10-bit video takes half of one the eight cores, which sounds fairly close to how my laptop only needs to use one of the two core to play smoothly. I don't know too much about the Atom, other than that I've heard a lot of bad things about the Atom being massively underpowered. Some online benchmarks of the E-350 places it somewhat slower than my Turion, but it should be passable with 1080p 10-bit as long as it's using multithreaded decoding.

[gradinaruvasile]

What video player and output mode are you using though? Even ARM CPUs in Android phones can play 1080p 10-bit fairly well today. You can easily go out and buy the cheapest AMD laptop on the market today, of which I got an old 1.6Ghz AMD Turion 64 X2 for $300 4 years ago, that plays 1080p 10-bit perfectly smooth with only ~half the CPU being used. AMD's E-Series should fair a lot better than that.

I am not sure E350 has more processing power than your Turion (uses less power, but thats another story). Turion was the top of the line mobile CPU back then for AMD.
Maybe the 1007 or the like Celerons have the oomph to decode 1080p in software, i dont know. But the E350 or the Atom certainly isnt up to the task. If the hi10p is anything like the regular 1080p, that is. Do you have a clip somewhere that we can test?

I have a SU3500 [email protected] which chokes on just about anything HD that isnt hardware decoded (it has the GMA45 video chipset which has some basic h264 decoding after i compiled libva manually). And that wasnt a cheap CPU when it was released. Also i have an atom 1.6GHz single core that barely handles anything. I installed Windows on another dual core 450 Atom that was crap also.