Re: gcc-3.4.0: mt_allocator crashes due to alignment problems onSPARC targets when sizeof(type) <= 2

[Paolo Carlini <pcarlini at suse dot de>]

Brad Spencer wrote:

> The numbers in the table are the number of times faster the
> mt_allocator-based operator new() is relative to the default operator
> new() on a sparc-sun-solaris2.8 Sun Netra-T1 (a relatively slow box).
>  
Thanks! Very useful data, in my opinion!

Any chance you can also do the test for our baseline kind of 
pool_allocation,
enabled with --enable-libstdcxx-allocator=pool ? Unfortunately, in order to
do that you need either a snapshot of 3.4.1 or mainline...

> If I set GLIBCXX_FORCE_NEW, then I should see values very close to
> 1.0.  Strangely, this isn't the case for 1024-byte objects which seem
> to not only be about the same speed as mt_allocator, but are actually
> _faster_ using mt_allocator's "forced new" than when they just call
> new directly.  I can't explain this yet.
I can't either :( Perhaps knowing a bit more about your benchmark would
help...

Thanks again,
Paolo.