Author: Armin Rigo <[email protected]> Branch: extradoc Changeset: r4028:874eed76d72e Date: 2012-01-13 15:40 +0100 http://bitbucket.org/pypy/extradoc/changeset/874eed76d72e/
Log: Rewrote this using the official term. diff --git a/blog/draft/tm.rst b/blog/draft/tm.rst --- a/blog/draft/tm.rst +++ b/blog/draft/tm.rst @@ -101,9 +101,10 @@ Events might also be *mostly independent*, i.e. they rarely access the same object concurrently. Of course, in a multi-threaded world we would still need -a lock to ensure correctness, but the point is that without the lock the -program would run correctly "most of the time" (and likely segfault the rest -of the time). +locks to ensure correctness, but the point is that the locks are rarely causing +pauses: `lock contention`_ is low. + +.. _`lock contention`: http://en.wikipedia.org/wiki/Lock_%28computer_science%29 Consider again the Twisted example I gave above. There are often several events pending in the dispatch queue (assuming the program is using 100% _______________________________________________ pypy-commit mailing list [email protected] http://mail.python.org/mailman/listinfo/pypy-commit
