]> mj.ucw.cz Git - libucw.git/commit
Rewritten handling of small chunks in the parallel radix-sorter.
authorMartin Mares <mj@ucw.cz>
Mon, 10 Sep 2007 18:05:10 +0000 (20:05 +0200)
committerMartin Mares <mj@ucw.cz>
Mon, 10 Sep 2007 18:05:10 +0000 (20:05 +0200)
commit73c12b4b33fc8017585439894aff79c20ef99ea8
treeef44256d4f606f1aa44e1aec1c258dfba92c7f14
parent9adbef3afa83fb9d9edf7054c98b0821f89cd831
Rewritten handling of small chunks in the parallel radix-sorter.

They are now immediately submitted to the worker thread instead of keeping
them in a list of delayed chunks for the whole course of the algorithm.
This avoids memory usage explosions caused by queueing lots of smallish
chunks. We prefer to wait for the small chunks to complete before starting
the next large chunk.
lib/sorter/array.c
lib/sorter/common.h