From: Matthew Dobson <colpatch@us...> - 2001-10-05 23:28:49
I was just reading through the numa_api.html again and wanted to
add another $.02 to the communal change pile.
My comments regard the following functions:
int bindtonode(unsigned long nodes, int behavior, numasubset_t
int bindmemory(unsigned long start, size_t len, unsigned long nodes, int
behavior, numasubset_t *restrict);
I feel that these functions should be rolled into a single function:
int bindmemtonode(unsigned long nodes, int behavior, numasubset_t
*restrict, unsigned long start, size_t len);
My rationale is that bindtonode() sounds like we're binding the
*process* to a node, which is something (as the document mentions)
totally orthoganal, but at the same time very likely. When we want to
bind ALL of a proc's memory to a node (or set of nodes) we could use
constants for the start and len arguments (MEM_START and MEM_LENGTH or
Thoughts, feelings, questions?