Boost logo

Boost :

From: Greg Colvin (gcolvin_at_[hidden])
Date: 2000-09-12 13:55:27

From: William Kempf <sirwillard_at_[hidden]>
> I've updated the Win32 implementation again. This version was
> compiled with gcc (using Dev-C++ IDE, which I *think* uses mingwin),
> so the code is closer to ANSI compliant. Other than that, the
> implementation is the same.
> Included is a test harness for doing speed comparisons of Win32
> critical sections, Win32 mutexes and the libraries implementation of
> mutex. The results of the timings (one run, so give or take some):
> Compiled with VC++
> boost::mutex was 5.68 x faster than Win32 mutex
> boost::mutex was 9.93 x slower than Win32 critical section
> Compiled with gcc
> boost::mutex was 2.63 x faster than Win32 mutex
> boost::mutex was 18.03 x slower than Win32 critical section
> I assume the terrible results using gcc can be attributed to my
> ignorance of the compiler. Having no idea what the command line
> options are I assume there were no optimizations used by the
> compiler. If anyone knows of a windows .hlp file for gcc available
> on the net, I'd appreciate an e-mail. If I'm going to work on Boost
> libraries, I think it's time I became familiar with a more compliant
> compiler such as gcc.

In my experience last spring, the GCC 2.95 compiler did a bad
job of generating Pentium code.

Boost list run by bdawes at, gregod at, cpdaniel at, john at