Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: OT Multithreading, STL, and profiling FOLLOWUP

Author: Pat King

Date: 06:23:36 08/12/01

Go up one level in this thread


I had thought that a map ensured unique entries. Apparently not when
multithreading in windows with mingw. Things worked again normally when I
guarded the access to the STL code to prevent simul access. Slower of course,
but you expect that in profiling.

On August 11, 2001 at 21:02:47, Pat King wrote:

>Apologies if this is more or less a repeat, I had trouble posting.
>
>I'm using a windows port of gnu c++, no profiling tools. I have a helper class
>which I use to generate function counts. This class, and the reporting
>functions, use an STL map to hold function names and counts. When I implemented
>ppondering, and thus introduced threads to my program, suddenly my profiling
>report showed repeats of functions, with different counts.
>
>a) Different counts for different threads?
>
>b) Is the gnu STL really thread safe? Can these counts be trusted at all?
>
>c) What does "thread safe" mean anyway? It doesn't crash, but is it right?
>
>inquiring minds want to know.
>
>TIA
>
>Pat



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.