l1 cache and register memory use in my program ...

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

Hello …

i want’s to know that how can i use L1 cache and register memory in my cuda fortran code ?

i can use shared memory but i dont know how to use register and local and L1 cache memory …

i’ll be appreciate if you help me in this issue …


Best regard …

2

Hi Miri,

L1 cache and register usage is not under user control. It’s used, but the back-end device compiler performs the register allocation and the runtime manages the L1 cache. You can however limit the number of registers used via the “-Mcuda=maxregcount:” flag.

  • Mat