Sender: graham AT delorie DOT com
Message-ID: <37289E1B.2C066C92@home.com>
Date: Thu, 29 Apr 1999 10:59:55 -0700
From: Graham TerMarsch <gtermars AT home DOT com>
Organization: Internet specialist for hire.
X-Mailer: Mozilla 4.51 [en] (X11; I; Linux 2.2.2 i586)
X-Accept-Language: en
MIME-Version: 1.0
To: pgcc AT delorie DOT com
Subject: Performance question about PGCC and AMD-K6
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: pgcc AT delorie DOT com

Just sat down here and rebuilt egcs-1.1.2 and pgcc-1.1.3 so that I could try
out some early benchmarks of my own to see what performance gain I'd get out
of moving most of the stuff on my box over to a pgcc compile.  So, tried a
test with 'gzip' to see what it'd come back with.

However, stupid as it was, the egcs compile was actually about 3% faster than
the k6 compile.  Running stuff here on an AMD-K6 233Mhz, glibc2, 2.2.2
kernel.  Compile options were as follows:

egcs
----
-O6

pgcc
----
-O6 -k6 -march=k6

I would've thought that even out of some instruction reordering that I might
get a bit of a performance boost out of this, so was quite surprised to see
that the pgcc version came out a bit slower.  Tests were done by doing a

test ./gzip -c [a 50MB tarball] > /dev/null

repeatedly for each compiled version.

So, uh, I guess the real question is; should I have seen any performance boost
on this one?  And, if so, would someone be able to provide me with tips on how
to make pgcc output code that'll run faster on my K6?

-- 
Graham TerMarsch

// -----------------------------------------------------------------
// I'm having a BIG BANG THEORY!!   
// -----------------------------------------------------------------