shared memory size how much shared memory is available

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

is there a general way or a rule of thmub how much shared memory is available for every block? I have a GTS 240 with 1GB.

Kind Regards

2

On compute capability 1.x devices like your GTS 240 it’s 16 kbytes per multiprocessor. Thus a block can use up to 16kb.
If you want more blocks per multiprocessor, you need to restrict the shared memory use further.

Check the Programming Guide for details.

3

I verified this using deviceQuery. Thank you very much.