[concurrency-interest] On A Formal Definition of 'Data-Race'

Boehm, Hans hans.boehm at hp.com
Wed Apr 17 14:52:36 EDT 2013


From: Nathan Reynolds

>  Many data races can exist without violating A, C or I and hence be harmless.
> These data races are great ways to improve scalability.
Absolute performance, probably.  Scalability, I doubt it.  Fences on x86 seem to involve just local work which doesn't involve the memory system.  Programs whose performance is limited by fences tend to scale well.  See my RACES 12 workshop paper http://www.hpl.hp.com/techreports/2012/HPL-2012-218.pdf.

> Setting the String hash value is racy yet harmless
Probably.  Aside from the fact that we know that the specified semantics for racy accesses are not correct, and hence are not quite sure what the real rules are.

Hans




More information about the Concurrency-interest mailing list