Invalid texture reference

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,All

My cuda code can run very well, but I want to use the float-point version automicAdd()function in kernel function. As we know the floating-point version of atomicAdd() is only supported by devices of compute capability 2.x. So I change my project setting to :GPU Architecture(1) sm_20, GPU Architecture(2) sm_20. When I run my program there was an error said that Runtime API error: invalid texure reference.

I just don’t know why, Anyone could help me.

Thank you very much in advance.

2

What is the compute capability of your device? You can’t set that arbitrarily based on what functionality you want to use, it has to match your GPU.

3

Thank you, my device is GeForce 9300M GS, its computer capability is 1.1. It means I couldn’t use the float point version atomicAdd()function, is there any other solution can solve my problem?

4

Yes.
Just rewrite the double atomicAdd() routine from appendix B.11 of the programming guide for float. Or use a reduction scheme instead.