Subject: Re: [boost] [Endian] Performance
From: Beman Dawes (bdawes_at_[hidden])
Date: 2011-09-07 07:18:57
On Wed, Sep 7, 2011 at 7:12 AM, Stewart, Robert <Robert.Stewart_at_[hidden]> wrote:
> Mathias Gaunard wrote:
>> On 09/07/2011 01:54 AM, Phil Endecott wrote:
>> > You mean the performance of these endian operations in
>> > general is not important?
>> Not that much no. It's important that it's reasonably fast,
>> but it needn't be performance critical.
> Examples were given recently of swapping bytes in large data sets, where each extra cycle quickly accumulates into noticeable delays. Another example arises in high frequency trading where latency is of the utmost importance. Every little bit that can be shaved off the critical path, of which byte swapping often is a part, can mean the difference between making money or not. Thus, performance of low level functions like these can be critical.
> It is not necessary that Boost provide the fastest implementation on every platform, but merely provide interfaces that can be implemented in the fastest way on at least some significant platforms to prove the point. Later standardization, should it occur, can leave platform-specific optimization to implementers.
I've worked on commercial GIS software where byte swapping was one of
the critical performance bottlenecks.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk