Re: One line performance tweak

[Peter Bienstman <Peter dot Bienstman at rug dot ac dot be>]

Zack Weinberg wrote:

>This patch produces a 10% performance improvement on Peter Bienstman's
>test case (see http://gcc.gnu.org/ml/gcc/2001-01/msg00218.html).

Thanks for tracking that down!

>This is old old code, and I don't properly understand its algorithm.
>But my guess would be that your code generates a small number of huge
>functions and fixup_var_refs_insns has to walk over the entire insn
>list looking for references to variables, which may only occur in a
>few places.

Strange, I'm not aware of any huge functions in my code. First I thought
they might be due to overactive inlining of the Blitz template library,
but dropping down to -O2 to disable inlining gives essentially the same
compile times.

This is definitely a regression with respect to gcc 2.95.2. Perhaps it has
something to do with the function-at-a-time mode?

>This isn't to say that 10% improvement on a one-line change is bad,
>just to encourage people not to forget about this test case.

When the dust settles down on this thread, I will file it into GNATS, to
make sure people don't forget ;-)

Thanks again,

Peter

------------------------------------- 
Peter Bienstman 
Department of Information Technology 
INTEC/IMEC - Ghent University
St.-Pietersnieuwstraat 41 
B-9000 Gent - Belgium 
E-mail: Peter.Bienstman@rug.ac.be 
Tel: +32 9 264 3445 
Fax: +32 9 264 3593
-------------------------------------