search

c3sr / comm_scope

NUMA-aware multi-CPU multi-GPU data transfer benchmarks
https://github.com/c3sr/scope
Apache License 2.0
21 stars 3 forks source link

dcbf should clobber memory #11

Closed cwpearson closed 5 years ago

cwpearson commented 5 years ago

https://github.com/c3sr/comm_scope/blob/eb289cf4797777fb36036e553eb9c1939427e6a1/src/utils/cache_control.hpp#L17

Something like this probably:

asm volatile ("dcbf 0, %0" : : "r"(p) : "memory");

cwpearson commented 5 years ago

In 8bb88f21a870516ba653c38086dc8037a0efafac we make this the same as the linux kernel implementation:

https://github.com/torvalds/linux/blob/d7197a5ad8528642cb70f1d27d4d5c7332a2b395/arch/powerpc/include/asm/cache.h#L96