Re: [Boost-users] metafunction invocation performance

From: Hicham Mouline (hicham_at_[hidden])
Date: 2008-06-08 16:44:08

Hello,
By lambda, I assume you refer to the placeholder expressions _  and _1 and
_2.
Do you mean the one in pair_with, in triple_with, in sequence_product, in
seq_seqpair_product,
or in all of them?

A custom metafunction class that includes a nested apply template?

I apologize if my questions are trivial...

Rds,

-----Original Message-----
From: boost-users-bounces_at_[hidden]
[mailto:boost-users-bounces_at_[hidden]] On Behalf Of Steven Watanabe
Sent: 04 June 2008 23:31
To: boost-users_at_[hidden]
Subject: Re: [Boost-users] metafunction invocation performance

AMDG

Hicham Mouline wrote:
> Hello,
> I submitted a post a few days ago 
> http://lists.boost.org/boost-users/2008/06/36852.php
> and I went to investigate on my own.
>
> Having not reached a solution to speed up the compilation,
> I looked again at the post and realized the code didn't compile
> due to a wrong paste. Apologies (if anyone tried),
>
> Anyways, here is the version that is too slow and runs out of mem during
> compilation
> on a P4 512Mb minimal redhat box with g++4.1
> while,
> on core2duo 2Gb vista64 with intel 10.1, it compiles in 40s and runtime is
> correct.
>   

I'm trying to process it using the tool I wrote at BoostCon:
http://svn.boost.org/svn/boost/sandbox/tools/profile_templates

Unfortunately, I had to kill the compiler (gcc 4.3.0 on cygwin)
after 2 hours.  I'll post the results when I'm done processing
the output.  (I have less than 10 GB to go)

In the meantime, it might help to write a custom metafunction class
rather than using MPL lambda.

In Christ,
Steven Watanabe