What's problem with simpleTextureDrv ?

Accelerated Computing CUDA CUDA Programming and Performance
1

Original “simpleTextureDrv” code in CUDA 2.0 Beta2:

int

main( int argc, char** argv) 

{

    runTest( argc, argv);

    CUT_EXIT(argc, argv);

}

I modified the “main” function as follow:

==>

int

main( int argc, char** argv) 

{

    int i = 1;

    while(true)

    {

        printf("Running No. =  %d\n", i++);

        runTest( argc, argv);

    }

}

The problem is : every time when the i=16378 , this program will crash(crash spot is “cuCtxCreate”).

Why I can not put the “runTest” in a loop?

And why it will crash when i=16378?

Related info:

Windows XP SP2 (32bit)

VS2005

Card1: Nvidia NVS 290 (Primary card for display)

Card2: XFX 8800GT 512MB

2

If I modify the example “simpleTexture” in the same way, that’s OK. I run it over ten hours. There is no crash.
Anyone tested in the same way?
That’s really weird!
It seems that application written with runtime API is different from application written with device API.

3

Is there anyone can explain this?
I suspect that some bug behinds this.

4

May b, You have really hit the tip of an iceberg! :-)

5

:biggrin:

6

How do I ask this question and modify the sample in that way? Because I wrote an application with CUDA 2.0 Beta2. And, when I run my application about thousands times later, it will crash. That happens always. I suspect that there is some bug in the CUDA. I modified the CUDA sample, finally , the CUDA sample also crashed. Wish some Nvidia people could pay attention to this problem.
Thank you!