Re: gcc compile-time performance take 2

[Diego Novillo <dnovillo at redhat dot com>]

On Sun, 2002-05-19 at 02:19, Andreas Jaeger wrote:

> Jan Hubicka asked me last week to do this and yesterday evening I
> started enhancing my scripts to visualize:
> - the bootstrap time of GCC
> - the build time for each SPEC program
> 
I will probably be doing similar modifications to the SPEC95 tests.  In
the meantime I've done some quick grepping on the logs we have.  I
plotted three things:

- SPECint95 build times from Mar/01 to May/02.
- GCC bootstrap times (C front end) from Apr/01 to Sep/01.
- GCC bootstrap times (C and F77 front ends) from Mar/02 to May/02.

I have only included times for successful SPEC runs.  The more
consistent timings are for SPEC builds.  I never consistently
bootstrapped GCC (and for a period I bootstrapped with C and C++).

Attached are the raw files I got from the logs in case somebody wants to
do more analysis.  The first column is the date, the second is the time
in seconds.  I plotted them and put the results in
http://people.redhat.com/dnovillo/SPEC/build-times/

Bootstrap times don't seem to change significantly for the C front end.
But, we seem to have gotten slower when bootstrapping both C and F77
starting in Apr/02.

If I weren't such an optimist, I'd say that the SPECint build times are
getting progressively worse.


Diego.

Attachment: c-bootstrap-times.dat
Description: MPEG movie

Attachment: c-f77-bootstrap-times.dat
Description: MPEG movie

Attachment: spec-build-time.dat
Description: MPEG movie