NVIDIA driver crash

I’m getting a crash on NVIDIA cards (TI 4400, FX 5800 Ultra) when using rectangle textures. What I do is:

create a rectangle texture with max size (4kx4k on the above cards). works

create another rectangle texture (it has the size Ax4k, where A is smaller than 4k. Here it crashes.

The texture format is GL_RGB.
My program works just fine on a radeon 9800 PRO (though there the 1st texture is 2k x 2x, since ATI has no support for 4k x 4k textures).

I double checked my memory allocation stuff and such, and I’m pretty sure that I am not overwriting any memory I’ve not allocated.

Any had similar problem?

Regards
-Lev

P.S. the driver version is 44.03

Just tested it with limiting texture size to 2k x 2k. Works flawlessly.

So I suppose it’s a bug. Anyone already submitted it? If no anyone knows where should I send that?

Regards
-Lev

Originally posted by Lev:
create another rectangle texture (it has the size Ax4k, where A is smaller than 4k. Here it crashes.

Lev, what’s that A exactly equal to?

You have to use the degree of 2 for the texture image dimensions, i.e. width = 2^n, height = 2^m (not including texture borders), where n,m are positive integers.

Originally posted by matt_weird:
You have to use the degree of 2 for the texture image dimensions, i.e. width = 2^n, height = 2^m (not including texture borders), where n,m are positive integers.

Not with texture rectangle. With that extension, you are not limited to a power of 2 dimension for the width and height. But the downside is that they may not use mipmap filtering and they are accessed by non-normalized tex coordinates. So instead of [0…1]x[0…1] it’s [0…w]x[0…h].

-SirKnight

argh, sorry, have to read more carefully next time(heh, someone’s “you don’t even read my posts!” makes sense though )

maybe this would help:

“When the target is TEXTURE_RECTANGLE_NV,the INVALID_VALUE error is generated if border is any value other than zero or the level is any value other than zero. Also when the target is TEXTURE_RECTANGLE_NV,the texture dimension restrictions specified by equations 3.11, 3.12, and 3.13 are ignored; however, if the width is less than zero or the height is less than zero, the error INVALID_VALUE is generated.
In the case of a rectangular texture, ws and hs equal the specified width and height respectively of the rectangular texture image while ds is 1.”

“Certain texture parameter values may not be specified for textures with a target of TEXTURE_RECTANGLE_NV. The error INVALID_ENUM is generated if the target is TEXTURE_RECTANGLE_NV and the TEXTURE_WRAP_S, TEXTURE_WRAP_T, or TEXTURE_WRAP_R parameter is set to REPEAT or MIRRORED_REPEAT_IBM. The error INVALID_ENUM is generated if the target is TEXTURE_RECTANGLE_NV and the TEXTURE_MIN_FILTER is set to a value other than NEAREST or LINEAR (no mipmap filtering is permitted). The error INVALID_ENUM is generated if the target is TEXTURE_RECTANGLE_NV and TEXTURE_BASE_LEVEL is set to any value other than zero.”

I mean, check glGetError return after!

[This message has been edited by matt_weird (edited 07-10-2003).]

matt_weird, how dare you to think Lev does such a mistake??? If he would do such a thing, i wouldn´t have trust in the extension loading library anymore!

Jan.

Originally posted by matt_weird:
I mean, check glGetError return after!

Well, I can’t, it gives me access violation in NTOGL.DLL on the glTexImage2D call. As I said, the problem is not that it doesn’t work correctly, it does not work at all.

Regards
-Lev

Originally posted by Lev:
[b]Just tested it with limiting texture size to 2k x 2k. Works flawlessly.

So I suppose it’s a bug. Anyone already submitted it? If no anyone knows where should I send that?

Regards
-Lev[/b]

Did you try

glGetIntegerv(GL_MAX_RECTANGLE_TEXTURE_SIZE_NV, &value);

Also, try proxy textures.

Dont send a bug report before beeing sure. And make a short demo for people here to try.

V-man:
[b]
Did you try

glGetIntegerv(GL_MAX_RECTANGLE_TEXTURE_SIZE_NV, &value);[/b]

Well, if Lev says 4k size does work, and the less size doesn’t – then this doesn’t help.

Jan2000:
matt_weird, how dare you to think Lev does such a mistake??? If he would do such a thing, i wouldn´t have trust in the extension loading library anymore!

That’s for sure, lol!

Lev:

If no anyone knows where should I send that?

How about to contact… Mark J. Kilgard!!! (mjk ‘at’ nvidia.com )

Please send me an app (with source if possible) that reproduces this problem and I’ll file a bug to get it fixed in a future driver release.

[This message has been edited by jra101 (edited 07-11-2003).]

No, really, i think you should contact Mark Kilgard first for that matter, as it was specified in specs for that extension(even despite jra101 works in nVidia too ):

[b] Contact

Mark J. Kilgard, NVIDIA Corporation (mjk ‘at’ nvidia.com)
[/b]

Ah, I didnt read the post well.

It sounds like a bug. But since you mentioned GL_RGB, make sure you dont have a data alignment issue.