Boost logo

Boost :

From: Paul Giaccone (paulg_at_[hidden])
Date: 2007-03-21 13:03:47


David Walthall wrote:
> After fixing those issues, I found that the radix_qsort and radix_sort
> are slower than standard sort. Maybe this is a result of when I
> commented out the specializations of insertion and r_insertion?
>
> [...]
> Specs: Windows XP Professional x64 SP2, Visual C++ 7.1, Intel Xeon
> dual core @ 2.66GHz, 2.0 GB RAM.
>
>
> Do you have anything that shows that your code beats std::sort?

Just a thought - is this code optimised? The compiler will go some way
to optimising the code, and presumably David used the same optimisation
flags in each implementation, but I wonder (I have not looked at the
code), Sam, what you have done to optimise the code and whether there is
anything else that can be done manually that the compiler is not smart
enough to do itself. That's assuming that it's possible to optimise the
code at all.

No doubt the standard library people have optimised their code to make
it run as fast as possible, so optimisation is necessary if you are to
beat the big guys.

Paul


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk