I wanted to know what does the value, Global Memory Bandwidth shown by cudaProf mean?
My application has large number of Global Memory accesses and I wanted to compare the memory on the CPU and the GPU. Thus, wanted to know the exact meaning of the value returned by cudaProf.
Is there any way to know the bandwidth of shared memory? Its mentioned in the Programming Guide that it is 100 times faster than global memory but can I know the exact value?