Re: [boost] Please Help: Experienced very bad performance withMulti-index

Subject: Re: [boost] Please Help: Experienced very bad performance withMulti-index
From: Jarolin, Robert (Robert.Jarolin_at_[hidden])
Date: 2009-07-31 14:22:29

Ok.  I thought there was some options to specify for boost that removed any and all non-release related settings.  If not, then we are probably fine with our current test and release based settings.

I was hoping that there was some ability to specify your own hashing function to be called for each hash.  Add to that, I was hoping that the supplied hash function was given the ability to return some BOOST defined value like DO_NOT_HASH that would tell boost not to hash that value in the entry when building its table info.  That would be ideal, but if that (or something similar) is not possible, then I guess my compromise is the way for me to go.

Again, thanks.

-----Original Message-----
From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]] On Behalf Of Joaquin M Lopez Munoz
Sent: Friday, July 31, 2009 12:49 PM
To: boost_at_[hidden]
Subject: Re: [boost] Please Help: Experienced very bad performance withMulti-index

Jarolin, Robert <Robert.Jarolin <at> trueposition.com> writes:

> Joaquin,
> Thank you for the help you are providing.
> To answer your first question, I am still uncertain how to put the 
> boost library into a release mode for the usage, though I no longer 
> have those #define statements that caused the excessive debug 
> overhead.

It is not about putting the *Boost library* into release mode, but about compiling your code with release settings, usually implying that the optimizer is used among several other settings oriented toards making the resulting binary code faster and lighter.

Which environment are you using?

> I happened to go a route similar to what you stated in 2 below[...] I 
> only wish there was a way to tell the boost multiindex to not create a 
> hash on the item if it is a certain defined default value (ex. Empty 
> string, -1, ...)

Some hash value has to be associated, as the element is to be inserted anywhere into the index. Your mechanism is a good compromise.

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost