ATI owners, please raise your hand back
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!I've uploaded, right there [LINK http://ompf.org/alpha/gpp-20060501-2.rar] , a little win32 binary doing some fashionable yet silly post processing.
Support for float buffers being what it is, hello bluescreen, i'm pretty sure it's not going to work everywhere. In fact i'm pretty sure it shouldn't work on mobile ATI for example.
That's why i made that silly thing (it loads an hdr pix and molest it a bit); i'd be delighted to hear reports about whether or not it's functionnal.
Works here on a 7800GT.
EDIT: link updated.
(L) [2006/05/01] [Ho Ho] [ATI owners, please raise your hand] Wayback!I'm currently sitting behind my brother's PC. It is athlonxp (==no SSE2) and has r9800se. I guess that no-sse2 thingie means I can't be of help.
In about 8-10h I'll get back home and there I can test it on another PC that has R9600xt and SSE2 support. I can test it there if noone else does it before me.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Ah, i should have made that clear, this thing is only for testing OpenGL, NPOT/float textures etc... So it doesn't require SSE at all. Just a somewhat recent 3D card.
(L) [2006/05/01] [Ho Ho] [ATI owners, please raise your hand] Wayback!Ah, the sweet old memories of Windows error messages [SMILEY Smile]
"This application has failed to start because the application configuration is incorrect. Reinstalling the application may fix the problem"
Is that what is supposed to happen when some of the features you use are not supported?
If I understood correctly the latest drivers are installed.
[edit]
From event viewer:
Type: error, Event ID: 59
Generate Activation Context failed for E:\New\gpp.exe. Reference error message: The operation completed successfully.
.
For more information, see Help and Support Center at [LINK http://go.microsoft.com/fwlink/events.asp.]
Product: Windows Operating System
Event ID: 59
Source: SideBySide
Version: 5.2
Symbolic Name: MSG_SXS_FUNCTION_CALL_FAIL
Message: %1 failed for %2. Reference error message: %3.
A component or manifest could not be activated.
Possible causes include:
The component or manifest depends on another program or component that is not installed.
The manifest contains XML content that is not valid.
The user does not have the correct permissions.
User Action
To resolve this problem, do one or all of the following:
Review the event logs to see whether any related events are logged.
Verify that all dependent components or programs are installed.
Review the XML content, and correct any that is not valid.
Verify that the user has the correct permissions on the computer.
Let me guess, you used MSVC later versions that need to have some weird config files together with .dll's?
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Not really [SMILEY Smile]
It's at least supposed to start and *then* complain. It sounds like there's a dependency missing, even if i thought i've included them all.
Please hold the line.
PS: got to love those informative messages.
EDIT: Ah i see, that's this <censored> msvc8 playing tricks in my back. Hmmpff. <scratches head> Now that was a debug build, maybe they're special.
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Could you try again with this binary, this time really statically built (but without debugging, gah).
[LINK http://ompf.org/alpha/gpp-ndebug.exe]
If that works i'll update the bundle.
(L) [2006/05/01] [Ho Ho] [ATI owners, please raise your hand] Wayback!That made this little error go away but shaders don't compile:
gpp::load_raw: 1024x768, about to read 9437184 bytes.
gpp::pass_t::compile: infolog...
Link failed. All shader objects have not been successfully compiled.
gpp::pass_t::compile: could not link tonemap.glsl
Isn't there a way to log the messages GLSL compiler gives?
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Yes, in the debug version [SMILEY Wink]
Ok, so the first problem was because my first binary wasn't statically linked and you don't have the msvc dll around.
If you got this far, it means that it should work (and that's either i've used an nvidia-ism or ATI is not compliant).
... building another binary... thanks for your patience.
EDIT: Static debug build here i am! [LINK http://ompf.org/alpha/gpp-debug.exe]
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!I've checked my code and i'm full of it; i try to get as much information for compilation as possible debug build or not.
I wonder how ATI think we're supposed to get that.
(L) [2006/05/01] [Ho Ho] [ATI owners, please raise your hand] Wayback!With that new binary I get the exact same output as before.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!I've googled a bit but found nothing relevant insofar.
So i don't know how to i'm supposed to get back compilation spam and don't know why such a brain-dead shader is a problem to begin with.
(L) [2006/05/01] [rontana] [ATI owners, please raise your hand] Wayback!seems to work here
got some glowing lights [SMILEY Smile] running at 59.9 fps (assuming im reading that right)
hardware:
Athlon 64 3500+, 1GB ram, 7800 GT 256MB
so do i win a prize? [SMILEY Very Happy]
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Close, but no cigar [SMILEY Razz]
Glad it ran on another box, but you have the same card as me so that's a tad too easy; that allows me to say that you have the vsync on and your desktop has a 60hz freq, eh.
Thanks for running it.
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Ho Ho, i think this one should catch compiler spam/message/hint better.
[LINK http://ompf.org/alpha/gpp-20060501-1.rar]
(L) [2006/05/01] [Ho Ho] [ATI owners, please raise your hand] Wayback!gpp::load_raw: 1024x768, about to read 9437184 bytes.
gpp::pass_t::compile: tonemap.glsl
*** failed to compile fragment shader
ERROR: 0:10: '=' : assigning non-constant to 'const 2-component vector of float'
ERROR: 0:12: '=' : assigning non-constant to 'const 3-component vector of float'
ERROR: 0:13: '=' : assigning non-constant to 'const 3-component vector of float'
ERROR: 0:14: '=' : assigning non-constant to 'const 3-component vector of float'
ERROR: 0:15: '=' : assigning non-constant to 'const 3-component vector of float'
ERROR: 5 compilation errors. No code generated.
*** failed to link.
Link failed. All shader objects have not been successfully compiled.
It seems like we're doomed, type something & hit return to continue or kill me.
Btw, have you tried GLSL validator on your programs? [LINK http://developer.3dlabs.com/downloads/]
I tried to install it but it was missing some DLL's.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Nope, but you put me on track.
GLI-SciTE, bundled with GLIntercept [LINK http://glintercept.nutty.org/] also includes that <insert bird name> validator/frontend. It also happens that ATI is using it.
So i can reproduce those errors and i'll put the blame on my const correctness habit.
Thanks!
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Updated bundle with smashing shaders uploaded, [LINK http://ompf.org/alpha/gpp-20060501-2.rar]
Let's see how far we get into before crashing this time.
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Semi-woot then [SMILEY Razz]
First, thanks a lot for going throught all this.
From what i see while your card/driver supports float textures, it doesn't want to filter them. I'm already mutilating data into 16 bit floats, so nvidia cards can filter but i can't do much more. Or better put that would require too much work (another path yada yada) just for some idiotic fashionable effect - even if tonemapping isn't that gimmicky -.
I don't know if it's just a transient issue (lack of driver support etc) or a fundamental hardware restriction, i'd need to google around a bit.
What's your card exactly?
(L) [2006/05/01] [Ho Ho] [ATI owners, please raise your hand] Wayback!Radeon 9800se
One more little thingie that might be useful to you: [LINK http://www.realtech-vr.com/glview/download.html OpenGL Extensions Viewer]
Info given about that GPU is here:
[LINK http://hoho.bafsoft.net/glview-RADEON-9800-SE-x86_MMX_3DNow_SSE.XML]
Just open that with the viewer and you should get all the information about the extensions you need [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.
Jan L.A. van de Snepscheut
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!I'll investigate.
But the functionality i'm looking for isn't exposed as an extension. I mean all the required extension were provided, FBO, NPOT, float textures etc... it's just that your box refuses to filter fp16, just like mine drops to 1fps when trying to filter fp32.
(L) [2006/05/01] [tbp] [ATI owners, please raise your hand] Wayback!Hmmk. Sounds like your card can't filter fp16 (or fp32). I could do it by hand in my shaders but that would run like <bleep>, i mean i already had to go down to fp16 and scale down my pingponging because it was barely acceptable otherwise.
Or i could get back to using good old integer textures, but that would be unfun [SMILEY Smile]
Time to invest...
[LINK http://www.digit-life.com/articles2/video/r520-part1.html]
(L) [2006/05/01] [rontana] [ATI owners, please raise your hand] Wayback!sorry for being ignorant, but what exactly are you trying to do anyway ?
the lights and stuff are all rather cool, i just get get why [SMILEY Smile] (even thou ive had a few of the may day beers)
(L) [2006/05/02] [tbp] [ATI owners, please raise your hand] Wayback!In one corner this thingy taking a float buffer (here an image with a wild dynamic) and doing various post-processing passes on it; there's no lighting, just post processing. It's a technological test, a much needed one as you can see.
In the other corner a real time raytracer already consuming oodles of memory bandwidth and cycles to produce a... float buffer.
It's pretty clear that the next step is to conjugate them [SMILEY Smile]
(L) [2006/05/02] [Lynx] [ATI owners, please raise your hand] Wayback!Interesting to see that it doesn't need any noticeable CPU time...
And is it just me, or is it somewhat ironic to do the 3d part on the CPU and the post-pro on the GPU? [SMILEY Very Happy]
I already see the first RTRT PS3 game where the GPU is degraded to post-effect and physics chip and Cell does all the rendering work, as i just learned it can do it better anyway...
-edit- still on a GeForce 6600 here btw...
(L) [2006/05/02] [tbp] [ATI owners, please raise your hand] Wayback!It's quite ironic indeed.
Now i wish those damn gpu could handle not-so-predictable texture fetches a bit more gracefuly and efficiently.
Sure it's nice to advertise <insert insanely large figure> giga/s, but it's a bit of a pain to hit that sweet spot when all you want is to apply rather large filters.
(L) [2006/05/02] [fpsunflower] [ATI owners, please raise your hand] Wayback!works on 6800 ultra too
(L) [2006/05/02] [tbp] [ATI owners, please raise your hand] Wayback!Grate.
I haven't hooked up everything, so by any chance could any of you GI people provide me with a wide dynamic picture and matching depth information?
(L) [2006/05/02] [tbp] [ATI owners, please raise your hand] Wayback!Seems puuuurfect (and puuurty). Thanks a lot.
Have i already said that picture was pretty?
If you have in stock pix with wider dynamic (they are hard to come by on the net and not "artificial" enough), sign me in [SMILEY Smile]
(L) [2006/05/02] [fpsunflower] [ATI owners, please raise your hand] Wayback!The credit is hardly deserved, that one was put together by one of my users. Had I picked the colors .... well ... programmer art syndrome [SMILEY Razz]
Can't wait to see it with some neat postproc filters applied. You're working on glare/tone mapping and stuff I presume ?
(L) [2006/05/02] [tbp] [ATI owners, please raise your hand] Wayback!Yep, because my rtrt wasn't gimmick compliant until now... so i'm in for the sinus scroll/lens flare of the XXIth century.
I can't run ATI demo that exhibits pseudo DOF, so i don't know how well it can be faked; i suppose it's horrible.
Well, at least i'll have a good tonemap to justify for that code & time [SMILEY Wink]
back
