[Top][All Lists]

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

apparent slowdown from 3.75 to 3.79.1

From: Craig A. Handler
Subject: apparent slowdown from 3.75 to 3.79.1
Date: Mon, 02 Apr 2001 21:25:02 -0400

Hello.  I am suddenly running into difficulty with the modification-in-
the-future check in 3.75 of Gnu make.  I am running on Windows NT 4.0,
and writing to network disk.  I am not getting the same error on our
Unix boxes.

I don't actually expect support on the old version, so I downloaded
3.79.1, and built it.  The logic in remake.c is much more sophisticated,
and in fact it does not get the spurious errors 
that the 3.75 make does.

However, there is a problem.  The 3.79 make appears to be significantly
slower than 3.75.  I ran a quick experiment with debug (-d) turned on.
An operation that ran within the same minute with 3.75 crossed 3 
minute boundaries using 3.79.1.  I realize that this is not terribly
precise, but that's the joy of Windows!

The experiment was:

  R:...> time /t & make -n -d independent 2>&1 > .make.old.out & time /t


  R:...> time /t & make.new -n -d independent 2>&1 > .make.new.out & time /t

The output of the builds appears to be identical, with the exception 
of changes like the word "prerequisite" replace "dependency" in the 
debug strings, etc.

Interestingly, the same experiment on Unix takes .85 (3.75) or .89 
(3.79.1) seconds elapsed time.  Weird.  The %cpu was higher on the
second run, suggesting slightly less efficiency from 3.79.1.  Given
the impressive run times, however, it seems like splitting hairs.

I am attaching that output for your perusal, not that I think it will 
tell you terribly much.  I will provide what help I can, given my 
workload, blah, blah, blah.

Craig Handler

Attachment: .make.new.out
Description: .make.new.out

Attachment: .make.old.out
Description: .make.old.out

reply via email to

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