madVR - high quality video renderer (GPU assisted)

[ericgur]

the 3DLUT is for performing color gamut correction, so we could get accurate colors within our display's color gamut, which usually is different (in some cases much more wider) than the source's color gamuts.

I understood you perfectly, but the actual use case is not clear to me. As far as I know, when a TV/projector recieves RGB, it assumes a certain gamut (PAL/NTSC/sRGB) and display a given 8 bit RGB value as specified by that gamut even its native gamut is larger. This is improtant for color accuracy. The TV via post processing (e.g. Saturation control) can create colors outside the PAL/NTSC gamut. New TVs which support xvyCC (via HDMI 1.3) can also display an extended gamut, but xvYCC is passed as YCbCr not RGB.
PC monitors are another issue - but the use case for MadVR is for HTPCs not desktops.
Using current OSes (XP, Vista) how does manipulating the gamut helps? You can't expand it, only shrink it even further.

In order to output a wide gamut you'll need OS support for outputing YCbCr without ever going to RGB.

Am I missing something here? Are you reffering to using the EDID information from the display? Doesn't the GPU driver already do this if EDID information is valid?

[FoLLgoTT]

Using current OSes (XP, Vista) how does manipulating the gamut helps? You can't expand it, only shrink it even further.

Yes, this is what we want. There are projectors and other displays out there which have a wide native gamut which can not be shrinked to the ITU-R Rec. BT.709 (or BT.601), because the devices lack the necessary option. So the saturation of the primaries and certain color combinations are way too high when watching Blu-ray's or DVD's. The picture looks not as it is intended to. Our goal is to achieve the BT.709 as close as possible.

[yesgrey]

I understood you perfectly, but the actual use case is not clear to me.
...
You can't expand it, only shrink it even further.
...
Am I missing something here?
Yes, like FoLLgoTT said, we don't want to expand the color gamuts (that's not possible) we want to shrink them.
If you want to see it in pictures go here (http://www.avsforum.com/avs-vb/showpost.php?p=11705893&postcount=2).
You can see two pictures of my projector. The outer white triangle is my projector's color gamut, the interior triangles, are the standard's color gamut.

[leeperry]

Yes, like FoLLgoTT said, we don't want to expand the color gamuts (that's not possible) we want to shrink them.
from the same thread, this screenshot on the JVC RS1/HD1 has the native gamut checked as "reference" in Color.HCFR, and you can read that the man is quite happy : http://www.avsforum.com/avs-vb/showpost.php?p=11937064&postcount=90

ppl on HCFR are ecstatic when they find out about this, wide gamut is evil on SMPTE-C movies....your PS script is saving lives I tell you :p

[jmone]

Can you force MadVR to be loaded as the default Video Renderer in the same sort of way Reclock can be loaded as the default Audio Renderer for all apps?
Thanks
Nathan

[ericgur]

Can you force MadVR to be loaded as the default Video Renderer in the same sort of way Reclock can be loaded as the default Audio Renderer for all apps?
Thanks
Nathan

Short answer - no.
Some players like Media Player 9-11 and Media Center load specific renderers. PowerDVD and WinDVD have their own built-in renderers. In some players you can specify the renderer (e.g. ZoomPlayer). DirectShow filters in general can declare a 'merit' - this value is used by the DirectShow architecture when building graphs automatically. The higher the merit - the most likely the filter will be used in a graph.

[tetsuo55]

Hi madshi,

i am about to buy some new hardware, and your renderer has got me thinking twice about which videocard to buy.

Here is what i use my pc for:

*SD(xvid/realmedia) playback with upconversion to fullhd (my current cpu can hardly keep up with ffdshow's processing)
*DVD with deinterlacing and upconversion
*non-full hd(hdready) to fullhd scaling
*fullhd native (should be 1:1)

I already know that a difficult high bitrate h264 + TrueHD/DTS-MA decoding requires a +/- 3ghz C2D cpu.


Now, i'm guessing that madVR will do the whole upconvert-to-bear-hd thing in an upcoming version(it's almost there now anyway), What kind of CPU+Videocard combination should i at least be looking at to get all the features you have planned working?
(PS i doubt it matters at the moment, but my display supports up to 12bit over hdmi)

[leeperry]

well, there'll never be any DXVA in mVR as I understand it so CoreAVC CUDA would be a great asset IMHO.

BTW, my HR/mVR comparison on the previous page was...just that. the slight gamma difference is mostly due to ConvertToYUY2() I guess, or the difference in the YV12>RGB32 pipeline between the 2.

to truly compare 8bit/16bit dithered, I'd have to either try a 8bit LUT in mVR(prolly not supported) or use the same exact HR setup w/ both types of LUT, and then it'd make sense.

prolly w/ these settings in the Avisynth filter of ffdshow, w/ a 16-235 input and a LUT that does PC conversion+BT709 RGB :
yv12toyuy2(itype=2,threads=4)
t3dlut("C:\LUT.3dlut",threads=4)

there's many arguments for t3dlut() but they don't sound too familiar to me :o

BTW, the 16bit LUT doesn't use any CPU cycles in mVR, right?

[Mark_A_W]

Short answer - no.
Some players like Media Player 9-11 and Media Center load specific renderers. PowerDVD and WinDVD have their own built-in renderers. In some players you can specify the renderer (e.g. ZoomPlayer). DirectShow filters in general can declare a 'merit' - this value is used by the DirectShow architecture when building graphs automatically. The higher the merit - the most likely the filter will be used in a graph.

Umm....long answer yes?

Reclock manages to spoof the id's of the default directshow Audio Renderer and can be setup to get loaded in all situations (it used to be a bit of a PITA to control the damn thing).

But this sort of thing is a long way down the track. Once Madshi has implemented smoothness controls, MadVR will be ready for primetime, but at the moment he hasn't and it's too jerky.

[leeperry]

Once Madshi has implemented smoothness controls, MadVR will be ready for primetime, but at the moment he hasn't and it's too jerky.
all went fine for me in the past days, but yesterday it started great, and after 1H it dropped frames like crazy....like VMR9/EVR always do in MPC on XP in non-exclusive mode :(

this type of issue happens on many boxes(not just mine), mostly w/o exclusive mode XP can't keep up w/ the tight Reclock timings I think...only HR can, and only if you catch the VSYNC fliptime properly(which is pretty hard to begin with) :o

but yeah, this renderer is only a week old and madshi's got more tricks than a clown's pocket :p

Beliyaal has found a way to NEVER miss the VSYNC fliptime in VMR9/EVR from what I've seen, so it's all possible I guess...just requires some maad coding skillz :devil:

[Rectal Prolapse]

I don't think madVR could ever be set as a "default" video renderer until it supports the subpicture/subtitle pin.

[madshi]

The biggest advantage of the 3DLUT is that you know that it will never use more GPU resources that it currently uses.:)
Its biggest disadvantage, is the size.
The decision is yours. At any time, cr3dlut's code could be transformed in shader code, it would not be a very hard task...;)
I still don't like the idea. Personally, I've no clue how gamut correction works. So whenever there's a bug I'd have to work on a shader code for formulas I don't understand. No, I prefer using an external tool. If anything goes wrong (like the problems leeperry reported), it's not my responsibility to fix it, but I can point to someone else (you!) to fix it... :) Having cr3dlut external means that I can save any work related on 3dlut stuff. Plus, I still think that doing gamut correction via shaders would eat more resources than doing a simple 3D texture lookup. I'm less concerned with startup time and more concerned with playback experience.

Can you force MadVR to be loaded as the default Video Renderer in the same sort of way Reclock can be loaded as the default Audio Renderer for all apps?
Currently no.

i am about to buy some new hardware, and your renderer has got me thinking twice about which videocard to buy.
Personally, I'm planning to get an ATI HD4770 (released planned for early May) because the combination of power consumption (-> fan noise), performance and price looks extremely attractive to me. I'm not aware of any NVidia graphics cards which can hold up with the 4770 in terms of performance per Watt.

Also, I'm planning to get a low power Intel Quad core, just to be on the safe side of things. Maybe one of those new Xenons which have a really low TDP.

Probably the combination of Quad core Intel + 4770 is much more than is really needed. But I don't know yet which hardware is exactly required for madVR. We will only be able to say once basic playback works perfectly including smooth playback features. Only after that we can collect data on which combination of CPU and GPU works and which doesn't.

BTW, the 16bit LUT doesn't use any CPU cycles in mVR, right?
Right. Except loading the 3dlut file at startup time, of course.

[madshi]

madVR 0.5 released

http://madshi.net/madVR.zip

* DVD playback should work now, too
* improved switching times between fullscreen and windowed mode
* improved startup time
* chroma samples are now placed left instead of center luma
* added settings control to switch between TV and PC output levels
* added settings control to enable/disable 3dlut processing
* added settings control for selecting chroma resampling filter
* D3DX9_35.dll is not needed, anymore
* improved display refresh rate detection calculation
* 3dlut and settings file routines are using unicode instead of ansi now
* fixed: aspect ratio calculation truncated instead of rounding
* fixed: aspect ratio changes in the middle of the movie were ignored
Please try whether the display refresh rate detection works good now. It's important because it will later be needed for smooth playback. There's a number written right next to the refresh rate in brackets, e.g. "60Hz (20s)". This is the number of seconds the refresh rate calculation was able to work without any problems. The higher the number the better. Ideally it should simply continue to increase throughout the whole movie duration. The higher this number gets, the more reliable the calculated refresh rate is.

[nijiko]

Excute me. Why delete the "disable resample chroma" options?

[madshi]

Excute me. Why delete the "disable resample chroma" options?
Because this setting has been superseded by the "Chroma resampling" setting. Set "Chroma resampling" to "Bilinear" to achieve the same result as "disable resample chroma".

[nijiko]

There is a problem.
The subtitles can't be loaded when videoclips start playing.
The subtitles must be closed once and opend once, it just can be displayed.
(With VSFilter.dll)

[madshi]

Subtitles are not really supported by madVR yet. I know there are some ways to work around that, but I don't have any experience with them, so I can't really help with this problem yet.

[nijiko]

So...
Thank you all the same for your great works.

I think pre buffer like 0.4 is better, how about you?

[neoufo51]

I don't think madVR could ever be set as a "default" video renderer until it supports the subpicture/subtitle pin.
The latest SVN release adds support for madVR as a possible renderer.

Revision 1048 - Directory Listing
Modified Sat Apr 18 17:39:19 2009 UTC (13 hours, 30 minutes ago) by casimir666
Changed : merge with Beliyaal branch
Added : support for madVR (new Video Renderer)

I guess I'll try this new renderer now.

[madshi]

I think pre buffer like 0.4 is better
What do you mean with "pre buffer"?

[FoLLgoTT]

Great work! :)

The bugs I found so far:

- I still get a macrovision failure when I want to play a DVD (in MPC and Zoom Player).
- If I play a file and open another file the screen stays black. When I press stop there is the message "loading pixel shader failed". Only a restart of MPC solves this problem.

I'm running MPC under Vista with a Radeon 3850.

[KoD]

Subtitles with Vsfilter work fine here.

Bugs:

- I have a 1440x900 screen. I'm in ZoomPlayer, in fullscreen mode. I open a file that is 1920x1080. The video file is shown at a 1920x1080 resolution, and I see only the portion cropped to 1440x900 (no resizing took place). Playing a file with a smaller resolution than my video screen does get scaled to the screen size when playback starts, however.

- after closing a file while in ZoomPlayer, and then every second file opened in ZoomPlayer ends with and error from madVr: "loading pixel shader failed". For the error to be visible, I have to switch from windowed to fullscreen or from fullscreen to windowed. The only way to have playback, is to close the player and open it again. (this is the same error as FoLLgoTT's)

- madshi, do you happen to have any mpeg (with mpeg 1 video) files around ? Try to play them and you'll see the "time length" is 0:00 (at least in Zoom Player), and the screen is black. The splitter used is the Microsoft MPEG-I splitter. Every other renderer work fine, with the reported time length of the file being right in them.

Subjective note: isn't the rendering time longer in this version ? I think I'm getting more files unplayable now. I think this renderer will not become usable by laptop owners, it requires not only a very good graphic chip (which I already have), but also a very powerful CPU, as this renderer is more CPU demanding than Haali's renderer, for instance.

Good things: the timer besides the detected refresh rate gets increased for much longer than in previous versions (when it didn't last longer than 5 seconds). Now I've even seen it over 1 minute (haven't tested it for long, though)

[SpaceAgeHero]

According to MadVR settings menu I have MadVR v0.6 now. :p

[nijiko]

What do you mean with "pre buffer"?

Sorry for my English.
In 0.4, when open a videoclip, it will use some time to init the videoclip, isn't it pre buffering?
I think it's propitious to play some videoclips.

[madshi]

- I still get a macrovision failure when I want to play a DVD (in MPC and Zoom Player).
Weird. I've only one DVD (THX demo DVD) to test with, though, and I don't even know if it's protected. Anyway, this one DVD showed "Macrovision fail" with madVR 0.4 in MPC HC for me and it plays fine now with madVR 0.5. I guess I have to find a "real" movie DVD to test with. Which exact error message do you get?

- If I play a file and open another file the screen stays black. When I press stop there is the message "loading pixel shader failed". Only a restart of MPC solves this problem.
Argh.

- I have a 1440x900 screen. I'm in ZoomPlayer, in fullscreen mode. I open a file that is 1920x1080. The video file is shown at a 1920x1080 resolution, and I see only the portion cropped to 1440x900 (no resizing took place). Playing a file with a smaller resolution than my video screen does get scaled to the screen size when playback starts, however.
That sounds strange. It seems to work fine with MPC HC here. Will have to test this with ZoomPlayer.

- do you happen to have any mpeg (with mpeg 1 video) files around ?
No, do you have a small sample?

Subjective note: isn't the rendering time longer in this version ? I think I'm getting more files unplayable now.
Try checking/unchecking "use 3dlut". Does that make a difference? madVR 0.4 used the 3dlut files by default. madVR 0.5 uses shader math by default. But actually I thought that should speed up things, not slow them down...

Good things: the timer besides the detected refresh rate gets increased for much longer than in previous versions (when it didn't last longer than 5 seconds). Now I've even seen it over 1 minute (haven't tested it for long, though)
Good to hear, thanks.

According to MadVR settings menu I have MadVR v0.6 now. :p
You're too fast... ;)

In 0.4, when open a videoclip, it will use some time to init the videoclip, isn't it pre buffering?
No, that was just time wasted on initializing dozens of pixel shaders which were not needed. And 3dlut file loading.

[madshi]

madVR 0.6 released

http://madshi.net/madVR.zip

* fixed: only the first movie played fine, a 2nd movie stayed black

[KoD]

Reports:

- the 1920x1080 that is not resized to a smaller screen size is still there
- now I've also found a 704x400 file that it's not upsized when starting playback
- the incorrect time length, and black screen for mpg files is there
- it's hard to compare the rendering times when using 3dlut and when not; looks like what madVr displays is an average, and the values vary each time the file is played. A subjective impression is that there's more time spent on resampling textures and rendering in this version than in previous versions. Now even the 704x400 file mentioned above is unplayable (rendering times over 50 ms for short periods of time). In Haali, there's no problem.

Conditions to reproduce the "no resizing" bug in ZoomPlayer: starting ZoomPlayer, going fullscreen, and then opening a file.

I'm uploading a mpeg and that 704x400 mkv file above and I'll give you the links in a PM.

More: - and I'll even send you an avi file which can't get played no matter what.

[nijiko]

>>madshi

But I consider 0.4 is more smoothly than after.

And, the CPU usage is too high, isn't it?

[KoD]

There is a problem.
The subtitles can't be loaded when videoclips start playing.
The subtitles must be closed once and opend once, it just can be displayed.
(With VSFilter.dll)
I've just noticed this report of yours in the MPC HC thread as well. I use ZoomPlayer and I have no issues with subtitles when using VSfilter. What you are experiencing is probably MPC not loading VSfilter because it wants to use its internal subtitle renderer. This is not a madVr issue, and it's not a VSfilter issue, it's simply a consequence of MPC behavior.

[Mike5]

Please try whether the display refresh rate detection works good now. It's important because it will later be needed for smooth playback. There's a number written right next to the refresh rate in brackets, e.g. "60Hz (20s)". This is the number of seconds the refresh rate calculation was able to work without any problems. The higher the number the better. Ideally it should simply continue to increase throughout the whole movie duration. The higher this number gets, the more reliable the calculated refresh rate is.

ATI HD4650, Catalyst 9.3, Win7, MPC-HC, madVR 0.6.

Playing a mkv 72op 25fps with a refresh rate of 75Hz, madVR says strange refresh rates, as:
106.04204Hz (1s)

Hz change continously and 1s stays the same.

The other parameters in statistics are correct.

[Mike5]

I keep getting Macrovision Fail both with a DVD (Gladiator) in the drive and trying to open the .IFO file in a DVD copied to disk.

No problem with single .VOB files.

The problem with refresh rate detection is the same as with mkv.

[KoD]

Message sent, I've added more info there as well. The mpg issue was present in v0.4 too.

[FoLLgoTT]

Weird. I've only one DVD (THX demo DVD) to test with, though, and I don't even know if it's protected. Anyway, this one DVD showed "Macrovision fail" with madVR 0.4 in MPC HC for me and it plays fine now with madVR 0.5. I guess I have to find a "real" movie DVD to test with. Which exact error message do you get?

It is the same message as before: "DVD: Macrovision Fail". Even the not protected HCFR test disc produces this error.

[Mike5]

The other parameters in statistics are correct.

Actually, reading them with more attention, the vsync interval is wrong too, but coherent with the changing display parametr.

[nijiko]

I've just noticed this report of yours in the MPC HC thread as well. I use ZoomPlayer and I have no issues with subtitles when using VSfilter. What you are experiencing is probably MPC not loading VSfilter because it wants to use its internal subtitle renderer. This is not a madVr issue, and it's not a VSfilter issue, it's simply a consequence of MPC behavior.

Sorry. After more testings, I found this problem is in VSFilter.dll, not madVR.

[Mike5]

Even the not protected HCFR test disc produces this error.

It appears evident to me that is not a protection problem, because I also get it with .ifo files of copied (and thus unprotected) DVD.

It seems a sort of a Menu navigation problem, but I'm not an expert.

[nlnl]

VISTA 32, MPC-НС 1056 (supports madVR), Nvidia 9400, ffdshow yv12 output

Can not use madVR! MPC goes back to Video Renderer.
But the installation was sucsessful. Filter madVR is added to external filters (preffered).

[TinTime]

I'm getting an error now with v0.5/0.6 in Zoom Player. Whenever I start playback, or seek, the image is displayed incorrectly.

If the player is windowed it looks like this (http://i42.tinypic.com/4gmb08.png).

The image moves further out of the window the further from the top left of the screen the window is (if that makes sense). If I resize the window, or go to full screen, the image corrects itself. However if I seek it then breaks again.

If I start playback or seek in full screen it displays the image in the top left of the full screen like this (http://i40.tinypic.com/24v0h8j.png). I've set the background to grey in Zoom Player which is why it's, well, grey :)
Full screen resolution is 1440x900.

[ericgur]

version 0.6 still crashes when d3d device is lost.

[nijiko]

>>madshi

On my computer, here, 0.6 is slower than 0.4.
0.4 is more smoothly.
Can you improve it?

[Snowknight26]

The labels seem to have a a gray background even though they should be white like the panel background:
http://www.stfcc.org/misc/madvr.png

[noee]

multi-mon, XPSP3, MPC-HC 1056, ffdshow decoder, Reclock (no Vsync), madVR .6 - all perf options enabled, K8 X2 3.0Ghz, ATI HD2600XT CCC9.4

Secondary monitor refresh calc is dead on perfect: 24.00079-.00081 (reclock says 24.001). :) Plus, it goes for many, many seconds, >200s.

I can't get any 3Dluts to work. When I enable 3dluts while playing, the player has a hard crash. If I modify the settings, then restart MPC-HC, I get audio, but blank video.

I would agree that .6 is slower somewhat. For AVC SD encodes, I'm seeing ~18-20% CPU and 61% GPU. On .4, the same file shows CPU: ~15% and GPU: ~43%

[cyberbeing]

Having the 3dlut files named HD.3dlut & SD.3dlut gives me an error "madVR reports: - opening 3dlut file failed"

Do they have to be named something different now, or is there some other problem?

[yesgrey]

I also cannot make v0.6 work with 3dlut files...:(

[TinTime]

They work ok for me. They've got a name change, and there are four of them.
Now:
SD - PC.3dlut
SD - Video.3dlut
HD - PC.3dlut
HD - Video.3dlut

[hubblec4]

hello madshi

i like your work but i cant use your madVR.
i use MPC-HC newest version. i open a videofile and the movie start, i hear the sound but the screen stay black. when i push stop, i can see an error-message:

madVR reports:
- graphics card only supports power of 2 textures

My System:
Grafik: ATI Radeon X1950 GT
WinXP_SP3
DirectX9.0c
CoreAVC codec
Haali Splitter


Can anybody help me?


hubble

[madshi]

- the 1920x1080 that is not resized to a smaller screen size is still there
- now I've also found a 704x400 file that it's not upsized when starting playback
- the incorrect time length, and black screen for mpg files is there

Conditions to reproduce the "no resizing" bug in ZoomPlayer: starting ZoomPlayer, going fullscreen, and then opening a file.

I'm uploading a mpeg and that 704x400 mkv file above and I'll give you the links in a PM.

More: - and I'll even send you an avi file which can't get played no matter what.
Unfortunately I can not reproduce any of this, sadly. Even your samples all work perfectly fine here with both ZoomPlayer 5 and 6. Does everyone have these problems? If so, can you please help me reproducing them?

I'm getting an error now with v0.5/0.6 in Zoom Player. Whenever I start playback, or seek, the image is displayed incorrectly.
Seems like the same problem KoD is reporting. But I can't reproduce it!!!

Is this maybe a problem with secondary monitor again? Or does it also occur on primary?

- it's hard to compare the rendering times when using 3dlut and when not; looks like what madVr displays is an average, and the values vary each time the file is played. A subjective impression is that there's more time spent on resampling textures and rendering in this version than in previous versions. Now even the 704x400 file mentioned above is unplayable (rendering times over 50 ms for short periods of time).
On my computer, here, 0.6 is slower than 0.4.
0.4 is more smoothly.
I would agree that .6 is slower somewhat. For AVC SD encodes, I'm seeing ~18-20% CPU and 61% GPU. On .4, the same file shows CPU: ~15% and GPU: ~43%
I've done 3 things which might effect GPU load:

(1) not using 3dlut by default, anymore, instead shader math is used, but you can manually enable 3dluts in the settings panel
(2) added some shader math to allow for future brightness, contrast and saturation control, when using 3dlut, but it's really only one addition and one multiplication, I can't imagine it eats that much performance
(3) installed latest DirectX version with newer HLSL shader compiler

It seems that on my PC the GPU load is not any bigger compared to madVR 0.4. Maybe the new HLSL shader compiler version optimizes for current generation hardware now and produces less good results for older hardware? I don't really know. I guess I'll remove (2) in the next build to see if that was the problem. If not, I might have to go back to an older DirectX version in the hope that it fixes the problem. But please try using the 3dlut instead of the shader math (by activating the "use 3dlut" option).

ATI HD4650, Catalyst 9.3, Win7, MPC-HC, madVR 0.6.

Playing a mkv 72op 25fps with a refresh rate of 75Hz, madVR says strange refresh rates, as:
106.04204Hz (1s)

Hz change continously and 1s stays the same.
:( That's bad. Anybody else with this problem?

What happens if you change your refresh rate to e.g. 60Hz, does detection still fail? Thanks!

I keep getting Macrovision Fail both with a DVD (Gladiator) in the drive and trying to open the .IFO file in a DVD copied to disk.
It is the same message as before: "DVD: Macrovision Fail". Even the not protected HCFR test disc produces this error.
I guess I'll need to do some more investigating on this.

VISTA 32, MPC-НС 1056 (supports madVR), Nvidia 9400, ffdshow yv12 output

Can not use madVR! MPC goes back to Video Renderer.
But the installation was sucsessful. Filter madVR is added to external filters (preffered).
There's a setting in ffdshow which allows or blocks connection to "unknown" filters. Make sure that this option is set correctly.

version 0.6 still crashes when d3d device is lost.
True. Haven't had a chance to look into this yet. But it's high on my priority list.

Secondary monitor refresh calc is dead on perfect: 24.00079-.00081 (reclock says 24.001). :) Plus, it goes for many, many seconds, >200s.
Good, thanks.

I can't get any 3Dluts to work. When I enable 3dluts while playing, the player has a hard crash. If I modify the settings, then restart MPC-HC, I get audio, but blank video.
Weird, works perfectly fine here for me with MPC-HC 1056!

Having the 3dlut files named HD.3dlut & SD.3dlut gives me an error "madVR reports: - opening 3dlut file failed"

Do they have to be named something different now, or is there some other problem?
They have different names now, but madVR should automatically create them if they're missing! Have you removed the cr3dlut directory or removed or modified any files in it?

P.S: It might be a NTFS security problem. Can madVR write to its own folder on your PC?

[madshi]

madVR reports:
- graphics card only supports power of 2 textures

Grafik: ATI Radeon X1950 GT
It seems that your graphics card has a quite limited feature set. It does not support what madVR asks it to do (namely create textures in the resolution of the video file). Unfortunately for now there's no way for you to make madVR work, except by getting a newer graphics card. Maybe eventually perhaps I'll find a way to make madVR work on your graphics card, too, sooner or later, but I can't say for sure. Also I don't know how well it would perform, even if I found a way to make it work...

[leeperry]

@madshi : I see you were talking about PS scripts for the gamut conversion, do you think making 4x100mb of LUT data is the best option? ...especially when you can simply use a PS script to do the 3D LUT.

I believe tritical went for the LUT route when yesgrey gave him the idea that it'd be much faster than running ddcc in realtime(which it is, except for 16bit LUT's that are CPU hogs)...and mostly because tritical didn't want to use CUDA or anything GPU accelerated.

but because you're making a video renderer, I think using the PS script from yesgrey/JohnAd(posted on AVS : http://www.avsforum.com/avs-vb/showthread.php?t=912720 ) would yield identical results in real world footage...and maybe you could even do it in 16bit? I understand HSLS can run in 64FP.

I'm sure yesgrey could also add an option in cr3dlut to output a PS script(so we'd have 4 text files of 2k instead of 4x96MB), and there you go! who cares if the GPU runs one PS script for gamut conversion, I believe Crysis runs thousands of them.

[madshi]

@madshi : I see you were talking about PS scripts for the gamut conversion, do you think making 4x100mb of LUT data is the best option? ...especially when you can simply use a PS script to do the 3D LUT.
A simple PS script may do the trick for simple color corrections. But yesgrey3 has bigger plans, e.g. like allowing different corrections for different brightness levels and such funny stuff. Doing all the necessary calculations in a script probably costs too much performance. That's why I think the 3dlut file concept is the better solution.