R help - Oct 2002 - performance issues: gcc 2.95.3 vs. gcc 3.2.0

Hello,

I recently installed gcc 3.2.0 next to my old gcc 2.95.3 . I understood that
this should give me some performance gain. I indeed found some, but in a few
cases I actually observed a considerable loss in performance. I would be
interested to learn if others have similar experiences, and what the general
opinion is concerning what version (and flags) to use for R!

1) gcc 3.2.0 -O3 can be much slower than gcc 3.2.0 -O2, causing a collection of
simple
tests to take over 50s instead of under 20s. This is due to -finline-functions!!

2) I have my own graphing/analysis programme that contains an expression
language with a
byte compiler (i.e. the parser constructs a call graph with the callback
function pointers,
memory for arguments and other relevant information). Again, I mostly find small
performance
gains, but also 1 particular construct (calling a procedure via the
'call' routine) that
is almost 10x slower with gcc 3.2.0 than with gcc 2.95.3 . (I don't
understand why it is so
fast with 2.95.3 but that's another issue :)). No influence of
-finline-functions here.
I imagine that this could be an issue for R too!

R.B.
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Benchmarking is of course a tricky business. One thought -- some
optimizations will increase code size and might bump a given loop out of
the on-chip or secondary cache. You'll see a big slow down in that case.

T.

On Wed, 2002-10-16 at 11:10, RenE J.V. Bertin wrote:
> Hello,
> 
> I recently installed gcc 3.2.0 next to my old gcc 2.95.3 . I understood
that this should give me some performance gain. I indeed found some, but in a
few cases I actually observed a considerable loss in performance. I would be
interested to learn if others have similar experiences, and what the general
opinion is concerning what version (and flags) to use for R!
> 
> 1) gcc 3.2.0 -O3 can be much slower than gcc 3.2.0 -O2, causing a
collection of simple
> tests to take over 50s instead of under 20s. This is due to
-finline-functions!!
> 
> 2) I have my own graphing/analysis programme that contains an expression
language with a
> byte compiler (i.e. the parser constructs a call graph with the callback
function pointers,
> memory for arguments and other relevant information). Again, I mostly find
small performance
> gains, but also 1 particular construct (calling a procedure via the
'call' routine) that
> is almost 10x slower with gcc 3.2.0 than with gcc 2.95.3 . (I don't
understand why it is so
> fast with 2.95.3 but that's another issue :)). No influence of
-finline-functions here.
> I imagine that this could be an issue for R too!
> 
> R.B.
>
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
> r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
> Send "info", "help", or "[un]subscribe"
> (in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
>
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
-- 
Timothy H. Keitt
Section of Integrative Biology
University of Texas at Austin

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._