automake
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: fast install released already?


From: Christopher Sean Morrison
Subject: Re: fast install released already?
Date: Wed, 22 Oct 2008 19:54:06 -0400

 
>From: Ralf Wildenhues <address@hidden>
>
>Wow.  It would be interesting to see how much faster things get with the
>chances.  Are the libraries all created with libtool?  Because then I
>guess they will end up being the remaining bottleneck (that would be
>interesting to see, too).
>
>> If interested, I could also do some performance measurements with and
>> without this...
>
>Yes, that would be cool.

I'd also be pretty interested in helping test the changes with some performance 
measurements on BRL-CAD.  Our build time takes anywhere from 2 minutes to 2 
hours, depending on the compilation environment and hardware.  Example 8-core 
Mac build:  1'57" for configure, 16'21" for make -j8 (massively I/O limited),  
and 3'45" for install.  Those numbers swing all over the place for other 
platforms/environments.  

COMPILATION PRODUCTS:
     42 BRL-CAD Libraries
             26 Installed
             16 Not Installed
    392 BRL-CAD Applications
            367 Installed
             25 Not Installed
     15 3rd Party Installed Libraries
     67 3rd Party Installed Applications

FILESYSTEM ORGANIZATION:
   4988 BRL-CAD Files
   1492 BRL-CAD Directories
   3044 3rd Party Files
    755 3rd Party Directories

>> we currently use automake 1.9.6, and it would be best if
>> we don't have to touch any of our files (autoconf 2.59, libtool 1.5.24,
>> m4 1.4.4), is this possible?

Our build is a hybrid recursive make with some subdirectories non-recursive 
using autoconf+automake+libtool.  We generate a mix of static and 
dynamic/shared lib types (most of our core libs generate both).

We support an even wider breadth with build tool support at automake 1.6+, 
autoconf 2.52+, and libtool 1.5+.  This is done for a variety of configuration 
management reasons including simple default configuration flexibility for a 
very broad developer base, dozens of varied regression test platforms that are 
used for development but aren't desirable or easy to change configuration-wise, 
and simple out-of-the-box functionality on major platforms (even for our own 
developers).

We also already have built-in timing mechanisms to know how long configuration, 
compilation, and installation take so it should be pretty easy to compare 
things before and after.

I'll keep an eye out on the list for some word on it being a good time to test 
things out.  Nice to hear the focus on performance in general, thanks for the 
efforts.

Cheers!
Sean





reply via email to

[Prev in Thread] Current Thread [Next in Thread]