>>>>> "KUGAN" == #KUGAN VIVEKANANDARAJAH# <kugan@pmail.ntu.edu.sg> writes:
KUGAN> Thanks for the information, basically i am interested
KUGAN> in knowing how linux kernel manages data allocation so
KUGAN> that conflict misses in the data cache is minimized,
KUGAN> the one you pointed out might also do the same, i am
KUGAN> looking in there now.
What is a "conflict miss" ?
I. As for you question, the kernel tries to improve the cache
utilization by slab coloring, as Jan described, in order to avoid
the situation where some number of cache lines are contended, while
other cache lines stay "idle".
http://www.usenix.org/publications/library/proceedings/usenix01/bonwick.html
The Linux kernel slab allocator is based on a predecessor paper by
the same author, it may have evolved since, though.
II. Another thing is "false sharing", which is when two or more
unrelated pieces of data are allocated on the same cache line and
cause bounces even when each CPU accesses only one datum. This is
solved by allocating data on L1 cache line aligned address.
Hope either I or II answers you question.
~velco
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
