This is the mail archive of the
mailing list for the GCC project.
Re: gzip performance test
- To: mark at codesourcery dot com (Mark Mitchell)
- Subject: Re: gzip performance test
- From: Joe Buck <jbuck at synopsys dot COM>
- Date: Fri, 25 May 2001 15:14:35 -0700 (PDT)
- Cc: dje at watson dot ibm dot com, jfm2 at club-internet dot fr, jbuck at synopsys dot COM, gcc at gcc dot gnu dot org
"David" == David Edelsohn <email@example.com> writes:
> David> All of this is un-scientific and not official results
> David> of any kind. By this comparison, gcc-3.0 looks pretty good
> David> at the basic C optimization level.
Mark Mitchell writes:
> Good news so far -- both gzip (for whatever it's worth) and your
> benchmarks indicate that we're not getting much worse.
There is a place where we know we're worse: the Stepanov tests, which
were near-perfect on 2.95.2 and are worse now. While this is not
enough to hold the release, I'd at least like to see an analysis of
why we're worse.
By the way, as David guessed, the gzip tests I ran used -O. Compiling
gzip with -O2 gives a very small improvement for Sparc:
Again the test is
gzip -c -9 SN452.tar > /dev/null
Platform 1: sparc-sun-solaris2.7
The server is an 8-processor server with 8GB of RAM; each CPU is a 400 MHz
gzip compiled with: -O -O2
2.95.2 58.59 57.45
3.0-prerelease 57.71 56.16
I haven't redone the tests on the Red Hat system yet.
By the way, someone asked about data for duplicating this test (Robert?
I deleted the msg). To get the input data, grab
and gunzip it. You should have 59146240 bytes. I picked it because it
was the biggest .gz file I had lying around.
Then get gzip-1.2.4a from your favorite GNU archive site.