3D texture problem on linux

Graphics / Linux Linux
1

I have some code runs on Windows fine, and I have dual boot on the same machine that runs Linux but the code is not running, it generates unknown error. I tracked down the line that makes the kernel to be aborted, it is this line:
surf3Dwrite(e, surfRef, width * sizeof(float2), height, depth, cudaBoundaryModeTrap);

I tried on running the code on different platforms and cards: CUDA 4.2, and CUDA 5.0, GTX 480, GTX 460, NVS 4200M, and Tesla 2050 with no luck on the unix side.

Anybody faced this issue with 3d texture on Unix? Giving that I have modified the code from 1D texture and it was working fine on all platforms.

2

Hi eHailat,

Do you have a standalone test case we could run? Please also run nvidia-bug-report.sh and attach the resulting nvidia-bug-report.log.gz file.

3

Hi, For some reason I can’t login to my account with the same username and password. I created a new one and hopefully this will be working next time I try to log in.

aplattner,
Log file is attached, please change the extension to .log. Your help is appreciated.

Thanks,

4

Hi eyhailat,

As Aaron said, we would need to be able to run your test case in order to attempt to reproduce the issue. Can you provide us with the code you’re running?

Thanks,

  • Pierre-Loup
5

Sorry for not responding earlier with my test case.
Here is the output from this code:

Values at location 0.000000 are x = 10000000.000000 and y = 10000000.000000.
Values at location 0.200000 are x = 2902735.000000 and y = 34853276.000000.
Values at location 0.400000 are x = 656.234741 and y = 8194.456055.
Values at location 0.600000 are x = -0.523052 and y = 14.010744.
Values at location 0.800000 are x = -0.545429 and y = -2.635774.
Values at location 1.000000 are x = -0.142661 and y = -0.823045.
Values at location 1.200000 are x = -0.052134 and y = -0.308617.
Values at location 1.400000 are x = -0.021802 and y = -0.130090.
Values at location 1.600000 are x = 0.000000 and y = 0.000000.
Values at location 1.800000 are x = 0.000000 and y = 0.000000.
Values at location (86.0, 0.0, 43.0) are x = -1998397155538108416.000000 and y = -1998397155538108416.000000.

The last line should have 0.000000 for both x and y. It works correctly on Windows.

Please change the file extension to .cu