How can I switch between cudaFuncCachePreferShared/cudaFuncCachePreferL1 for Thrust algorithms, an example turn off for sort?
PS And an additional if you know, how to specify compilation flag in MS VS 2010 to turn off L1 cache in cuda, where must i write (-Xptxas -dlcm=cg)?
You can use the cudaFuncSetCacheConfig
runtime function to set your preference for L1 cache or shared memory.