Problem with Atomic Functions

Accelerated Computing CUDA CUDA Programming and Performance
1

I have a GeForce 470 GTX and have no problem compiling and running the simpleAtomicIntrinsics example using Visual Studio 2008 in 32-bit Windows XP. I tried then using the addAtomic function in another project I was working on and got the following error:

Error 2 error: identifier “atomicAdd” is undefined

I could not find any difference in the project settings and no obvious #include statements were missing as well. I then tried adding a call to addAtomic into the pre-canned cppIntegration project and got the exact same error so its apparently not just my project.

Has anyone else seen this or know how to fix it? Thanks

2

Are you compiling with [font=“Courier New”]-arch=sm_20[/font]?

3

Not explicitly no, but I don’t see where they are doing that in the canned simpleAtomicIntrinsics demo either. Where is the Visual Studio project properties would one add that? Thanks

4

Project / Properties / Configuration Properties / CUDA Build Rule / General / GPU Architecture

5

Project / Properties / Configuration Properties / CUDA Build Rule / General / GPU Architecture

6

Thanks for the help, the problem is solved. In the CUDA Build Rules there are 3 places for GPU Architecture and the initial settings that didn’t were:

GPU Architecture (1) sm_10

GPU Architecture (2) sm_20

GPU Architecture (3) 0

To Fix it I had to modify it to the following:

GPU Architecture (1) 0

GPU Architecture (2) sm_20

GPU Architecture (3) 0

7

Thanks for the help, the problem is solved. In the CUDA Build Rules there are 3 places for GPU Architecture and the initial settings that didn’t were:

GPU Architecture (1) sm_10

GPU Architecture (2) sm_20

GPU Architecture (3) 0

To Fix it I had to modify it to the following:

GPU Architecture (1) 0

GPU Architecture (2) sm_20

GPU Architecture (3) 0