[Top][All Lists]

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

Re: gnustep-make experiment

From: Nicola Pero
Subject: Re: gnustep-make experiment
Date: Mon, 12 Feb 2007 22:38:21 +0100 (CET)

> - this search would be executed for /every/ make invocation instead of
> once during configure.

It would be really slow if we were trying to compile something against the 
for every make invocation.  Subprocesses are very slow, and the compiler is 
particularly slow. :-/

But if we're just checking if 4 or 10 files exist, that's a completely different
operation -- that is a fast operation.  Reading very little data from disk
becomes unmeasurably fast very quickly ... reading the information from disk 
will be
really fast the first time it's done; subsequent checks will hit the OS cache
and the disk won't even spin, so it takes almost zero time. :-)

The general rule of thumb for makefile performance is:

 * forking a subprocess is slow

 * reading a small file is reasonably fast

 * checking if a file exist is really fast

The reason that reading is usually fast is that it can benefit from caching. ;-)

> But I see that other projects that don't yet already need ./configure
> for other purposes like GDL2 would profit from this approach.
> [...]
> So, do you have suggestion on how to handle the
> LIBRARY_COMBO/GNUSTEP_TARGET_DIR or will it be function approach?

The actual function would need to be defined inside common.make and the list 
of paths that are searched will mirror exactly the list of -L flags that we 
build in there.  It will be easy to keep them in sync since they are defined
in the same place. :-)


reply via email to

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