how to use 48k-shared memory in gtx480?

Accelerated Computing CUDA CUDA Programming and Performance
21

Try this NVCCFLAGS:

-O3 --ptxas-options=-v -arch sm_20

Do not include the common.mk, you are probably overriding the flags you just set.

22

Ya, I change the NVCCFLAGS to the one you mentioned. But when I remove the include of common.mk to avoid overriding flags, it reports errors like

error: cutil_inline.h: No such file or directory

It seems I lack something in makefile to include this…

23

Ya, I change the NVCCFLAGS to the one you mentioned. But when I remove the include of common.mk to avoid overriding flags, it reports errors like

error: cutil_inline.h: No such file or directory

It seems I lack something in makefile to include this…

24

You need to add the include folder to nvcc. Cutils folder lies in the SDK folder instead of the folder the toolkits headers are put in.

25

You need to add the include folder to nvcc. Cutils folder lies in the SDK folder instead of the folder the toolkits headers are put in.

26

When I add -arch=sm_20 the code compiles, but it takes significant longer time to compile… Without the flag, it compiles in 1second, but with the “-arch=sm_20” flag, it takes 45secounds to compile. Is this expected?

Thanks!