Atomic add on quadro fx 4800

Accelerated Computing CUDA CUDA Programming and Performance
1

I apologize in advance for this simple question.
I’m using quadro fx 4800, a 1.3 capacity card.
I’m conducting a project and hit a point where a need
atomic add otherwise I will launch too many kernels.
My question is : Is it possible to do float atomic add?
I found It was possible with integers.
Thank you!

2

I apologize in advance for this simple question.
I’m using quadro fx 4800, a 1.3 capacity card.
I’m conducting a project and hit a point where a need
atomic add otherwise I will launch too many kernels.
My question is : Is it possible to do float atomic add?
I found It was possible with integers.
Thank you!

3

I think float atomic add is only capability>=2.0 (see sm_20_atomic_functions.h).

This workaround seems to work for me on a 4800 though…

4

I think float atomic add is only capability>=2.0 (see sm_20_atomic_functions.h).

This workaround seems to work for me on a 4800 though…

5

Hi,

Thank you for your answer.

The page you showed was very helpful.

6

Hi,

Thank you for your answer.

The page you showed was very helpful.