Author: Armin Rigo <ar...@tunes.org> Branch: extradoc Changeset: r4231:db258d235ec4 Date: 2012-06-30 20:19 +0200 http://bitbucket.org/pypy/extradoc/changeset/db258d235ec4/
Log: The opposite catch. (thanks fijal) diff --git a/talk/ep2012/stm/stm.txt b/talk/ep2012/stm/stm.txt --- a/talk/ep2012/stm/stm.txt +++ b/talk/ep2012/stm/stm.txt @@ -106,6 +106,7 @@ Implemented in pypy-stm --- slowly, but who cares? :-) when you have an unlimited supply of cores... (ok, I agree we care anyway.) + How? ---- @@ -129,6 +130,21 @@ as the transactions don't conflict with each other. +The opposite catch +------------------ + +"Fool-proof" in the sense of always giving correct results, but you +may need tweaks to avoid systematic conflicts + +E.g. don't update a global counter from every single transaction! +That's a systematic conflict. + +Not the final best answer, but at least approaching the problem from the +"safe" side: "it always gives correct results, but usually you need to +work to reduce conflicts". As opposed to using threads: "it is always +fast, but usually you need to work to remove synchronization bugs". + + What about CPython? ------------------- _______________________________________________ pypy-commit mailing list pypy-commit@python.org http://mail.python.org/mailman/listinfo/pypy-commit
