"warning : Double is not supported. Demoting to float" on GTX295 under ubuntu 9.10

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
i’ve just made a fresh installation of Ubuntu 9.10 64bit with the latest version of cuda driver(256.40), cuda toolkit and sdk (all in 64bit version).

Every example from cuda SDK or piece of code i compile i get this error:

“warning : Double is not supported. Demoting to float”

but i’m working on 2 GTX295 and, looking at the deviceQuery, the device is correct (gtx295 with computing capabilty 1.3).

Previously i had a working 8.04 Ubuntu 64bit with cuda driver of realease 2.0.

Any idea to identify/solve this problem?

More info are available in the attached nvidia bug report file.

Thanks in advance,

Danielta

2

I’ve seen this on my GTX 280 (Cuda 2.3) but it doesn’t seem to occur on my Tesla. (Cuda 3.1)

-Nachiket

3

I’ve seen this on my GTX 280 (Cuda 2.3) but it doesn’t seem to occur on my Tesla. (Cuda 3.1)

-Nachiket

4

Are you compiling with [font=“Courier New”]-arch sm_13[/font]?

5

Are you compiling with [font=“Courier New”]-arch sm_13[/font]?

6

Yes, I am.

7

Yes, I am.

8

Hi im in 64 bits, and i cant use memcpy, it says memory or pointers are too large to fit in 32bits, what flags should i use? or do you know something about my problem? it is not memory cause im only transfering 1024*sizeof(int) bytes.

9

Hi im in 64 bits, and i cant use memcpy, it says memory or pointers are too large to fit in 32bits, what flags should i use? or do you know something about my problem? it is not memory cause im only transfering 1024*sizeof(int) bytes.

10

Hi im in 64 bits, and i cant use memcpy(devicetohost), it says memory or pointers are too large to fit in 32bits, what flags should i use? or do you know something about my problem? it is not memory cause im only transfering 1024*sizeof(int) bytes.

11

Hi im in 64 bits, and i cant use memcpy(devicetohost), it says memory or pointers are too large to fit in 32bits, what flags should i use? or do you know something about my problem? it is not memory cause im only transfering 1024*sizeof(int) bytes.