@Antique

Quote:

Antique wrote: @olegil The newer gfx cards have built in 1080p decoding. But i guess the wallet needs to be big to get amd to write drivers/give us required docs.

Both AMD and nVidia have stated earlier that reason why video decoding docs aren't aviable is that they don't own tech used on it, they are just licensing it. Thought if my memory serves me right some AMD's oss drivers have reverse engineered support for it, but don't know how complete it is.

@KimmoK

Quote:

KimmoK wrote: @Antique Unless we know some Amigan (ex is enough) at AMD and could convice him to do the needed binary... (+arrange prebuilt dev system + bounty for the driver)

Would have to be a hell of bounty to someone risk losing their job for using corporate secrets for personal project. Only way to to do what you suggest would be negotiate deal with AMD and that propably wouldn't be any cheaper.


Quote:

+IIRC, in future it should be possible to use Gallium3D to access the GPU's ebgines to accelerate video decoding.

It has been WIP since 2008 and still not ready. Don't hold your breath while waiting...

@tommysammy

Quote:

tommysammy wrote: Loading and display the moon takes here under 5 sec

I would suggest to try another software for displaying pictures The moon gets displayed here in 1.4s (Mac mini, showgirls). I am sure, the X1000 can do that with similar speed (i.e less than 2s)

Last edited by Zylesea on 05-Feb-2012 at 12:01 PM.

_________________
My programs: via.bckrs.de
MorphOS user since V0.4 (2001)

@Rose

"Would have to be a hell of bounty to someone risk losing their job for using corporate secrets for personal project. Only way to to do what you suggest would be negotiate deal with AMD and that propably wouldn't be any cheaper."

Ofcourse it would need to have aproval from AMD.
I imagine it would not be impossible for some AMD developer to do the binary on his own time.

We do not know without asking.
(AmigaDE for mediaterminal would not have started without asking, same for spotify for Amiga, etc...)
It helps greatly if you know someone inside a company. Also there are hobby clubs inside every bigger IT firm, also they could do some helping, but we would need some Amigan contact first to go beyond the barrier.

Last edited by KimmoK on 05-Feb-2012 at 12:10 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Quote:

Ofcourse it would need to have aproval from AMD. I imagine it would not be impossible for some AMD developer to do the binary on his own time.

There's still problem that drivers haves code that AMD doesn't own.

Quote:

(AmigaDE for mediaterminal would not have started without asking, same for spotify for Amiga, etc...)

Mediaterminal never came out, Spotify project is still starting and devs doesn't even have hardware yet so bit early to advertise it as success (remember open office?).

about benchmarks:
Traditionally people have done linux kernel compilation tests on x86.
Could someone try similar (repeatable) compilation test on x1000?
(some linuxkernelcompilationtestsource.lha could be put to Aminet for that purpose, to ensure people use exact same files in testing ...)


@Rose

With your kind of negativity, nothing is started and nothing will happen.

(btw. mediaterminal did game out, but the Nokia dev guys did not manage to release AmigaDE with it before their R&D site was closed)

Last edited by KimmoK on 05-Feb-2012 at 12:23 PM.
Last edited by KimmoK on 05-Feb-2012 at 12:21 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Quote:

@Rose With your kind of negativity, nothing is started and nothing will happen. (btw. mediaterminal did game out, but the Nokia dev guys did not manage to release AmigaDE with it before their R&D site was closed)

With being overly positive about on going projects which lead to nothing there won't be competing projects since people will lull themselves that "No reason to start project since X will take care of it". Being overtly positive about everything hurts platform as much as negativity.

(btw. I same sense Walker got released... If it's not out in the wild, it really doesn't matter)

@Zylesea

not sure, but as far as I know WarpView is the fastest image displayer around...

_________________
Geomarketing at
www.geobiz.de
www.gebietsplanung.net
www.geomarketing-consultant.de

@amigadave

Quote:

Will improved video card drivers for OS4.2 in the future improve video playback or streaming video performance? Or is this totally dependent on the PA6T and the design of the X1000's motherboard and components that make it unable to process and display 1080p video content? (I know you can't predict how much video card drivers can be improved in the future, but is the current state of OS4 video card drivers one of the limiting factors preventing 1080p video playback?)

If we strictly speak of CPU decoding, the only way to enhance that is to use the second core since ffmpeg can be multithreaded (but in a real SMP environment, else it could be quite tricky in some hacky SMP/AMP implementation).

The thing is this test was only about the video decoding. In the real world, you still have to display it, and there, it depends on the video output driver in MPlayer (p96pip is at least 30-40% slower than it could be) and the video bus speed (in CPU mode it reaches 160MB/s which is not great, but expected, since DMA transfers are the way to go).

Of course, GPU decoding would be great, but i don't see it happening too soon. :)

Last edited by Fab on 05-Feb-2012 at 02:11 PM.

@Toaks

Quote:

Toaks wrote: @realize Quote: realize wrote: @amigang There will be a second run. Stated now by Trevor and AmigaKit. I hope denil gets one. got a source?

See my posting #83...

_________________
Ciao

Dandy
__________________________________________
If someone enjoys marching to military music, then I already despise him.
He got his brain accidently - the bone marrow in his back would have been sufficient for him!
(Albert Einstein)

@Sam & Co.

try Quake 1 in Dosbox in the highest resolution

_________________

@tommysammy

thank you very much for testing LoView...
I ask to try it only because it have a time counter and WarpView and other no...
no matter what is the faster :)

However The loadtime that you can see on LoView was the effective LoadTime not the time from load to display...

However nice result, I only hope that X1000 CPU will be colcked at 2GHz at last(and possibly also a bit more) so ew can gain about a 20% more speed :)

_________________
Simone"Tuxedo"Monsignori, Perugia, ITALY.

If a GPU hasHardawre support for HDTV decoding.. Why do you need a driver? why cant you pass a small data structure containing a pointer to the file system and the file to be decoded? Worse case, have multiple Gig of RAM, copy a VOB, which is 1 gig in size into RAM, pass address of start of VOB to teh HDTV decode hardware?

Why does nobody do things simply these days, Printers with USB should not need drivers, they should be able to use USB massstorage, especially given Rasberry Pi pricing of on board CPUs powerful enough to handle ASCII, JPEG, PDF etc natively.

Better still, why wont GPU makers add a SATA port to the card, so we dont NEED an OS to play HDTV files from optical disks?

_________________
The older and more respected a scientist is, the longer it takes to prove him wrong.

@RodTerl

GPU has hardware that can ASSIST with video decoding. There's still a lot that isn't pure matrix math in a decoder.

But I wouldn't think you absolutely needed a driver for it if the hardware was fast enough.

We have the bandwidth now to do offloading of specific functions to the GPU, by direct GPU hardware banging. Wouldn't think you would need an official driver to get at least SOME benefit.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@thread

In case someone did not notice...
2D benchmarks of x1000 have started to appear here:
http://hdrlab.org.nz/benchmark/gfxbench2d/OS/AmigaOS?start=0

Expect to see huge improvements on that department.


btw. SDLbench on SAM440ep-mini 666Mhz
Pitch = 320
Hardware surfaces avail = 1
Window manager avail = 1
Blitter hardware = 1
Colorkey blit hardware = 0
Alpha blit hardware = 0
Software->Hardware accel = 0
Video memory = 0
320x240 320x240 640x480 640x480
software hardware software hardware
Slow points (frames/sec): 2.8975 123.077 0.610874 33.195
Fast points (frames/sec): 93.0909 68.7618 26.2861 17.2333
Rect fill (rects/sec): 6681.89 178087 1760.96 97523.8
32x32 blits (blits/sec): 23272.7 141241 19598.1 132129
http://os4depot.net/share/utility/benchmark/sdlbench.lha
Is this relevant test app any more?

Or the CubeBM:
10128 frames drawn in 8.7 seconds
1338 fps (0.747 milliseconds per frame)
1178 fps (0.848 milliseconds per frame with swap)

And for 68k on 440/666: (sysinfo from Aminet)
Sysinfo gives 1679 dhrystones in RuninUAE (does not run on petunia)
More to try: http://aminet.net/search?query=benchmark

I thought AOS4 SWAP does not work ... but at least with Multiview I could open this wtc_photo http://aminet.net/package/pix/misc/wtc-photo slowly, buy steadily.
In 68k times SWAPping with third party tools did not slow down the rest of the system... but on AOS4 mouse jitters ... (windows emulator built in?) ... I hope we see improvements there. I would like to set how much physical RAM some naughty application gets etc... just like on 68k with VMEM/QMEM/whatever.

Last edited by KimmoK on 08-Feb-2012 at 03:26 PM.
Last edited by KimmoK on 08-Feb-2012 at 03:24 PM.
Last edited by KimmoK on 08-Feb-2012 at 03:12 PM.
Last edited by KimmoK on 08-Feb-2012 at 02:55 PM.
Last edited by KimmoK on 08-Feb-2012 at 02:48 PM.
Last edited by KimmoK on 08-Feb-2012 at 02:40 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

Quote:

Fab wrote: @paolone It makes sure the loopfilter won't be disabled. It's not by default, but some people might have put the skiploopfilter=all or so in their config file, which would have compromised the results. And this option should definitely work on AROS too. This option exists for years in MPlayer, and since the AROS port is based on the MorphOS port, i don't see any reason why it wouldn't work. What CPU does your virtual machine run on?

this option don't work also on Windows version..

edit:
i tried mplayer with -lavodopts option on old XP 32bit and works...

Last edited by Seiya on 11-Feb-2012 at 02:33 PM.

_________________

ta daaa

Last edited by Seiya on 11-Feb-2012 at 03:39 PM.

_________________

@Seiya

Nice pics, was that SAM really a flex 733 ? Mine does the blender bricks and water about 10s faster than that.

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@broadblues

AmigaOne G4 1,26 Ghz :

Saved: ram:pool0001.jpg Time: 00:13.32

Sam440Ep o/c at 733 Mhz (FSB 147) and Sam440 Setup 1.3 :

Saved: ram:pool0001.jpg Time: 00:55.23

Last edited by K-L on 11-Feb-2012 at 05:05 PM.

_________________
PowerMac G5 2,7Ghz - 2GB - Radeon 9650 - MorphOS 3.14
AmigaONE X1000, 2GB, Sapphire Radeon HD 7700
FPGA Replay + DB 68060 at 85Mhz

@Seiya

Comparison with x86 confirms my expectations (Blender, Lame). Mplayer benchmark show that our current software isn´t able to use its hardware efficiently: 1 GHz G4 is rougly as fast as 900 MHz Athlon (excluding AltiVec), yet it loses in this test even when using AltiVec.

I hoped for better performance of X1000 in Blender benchmark, test with greater workload (more complex scene) would be interesting.