This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

RE: i386 backend


>
> > What is being done to address these issues, if anything?
>
> The i386 backend has been completely redesigned for 3.0, so it

I'm glad to hear that. The page about gcc 3.0 didn't say anything about
this.

> should generate
> considerably better code.  In my measurements it is comparable
> with MSVC I have, but
> I usually don't get so drastical differences in benchmarks.
>

I don't necesarily get such a drastic difference always, but in calculation
intensitive , performance-critical integer based code, such as my chess
program, I get a difference of the orders mentioned.

> If you do have some code you really worry about and you can point
> out those simple
> optimizations missed, I may take a look at those.
>

I don't have any code, right here. The code I know about, are examples I
have seen in computer chess communities, where this issue is also being
discussed. Others know far more about this that I do.
That being said, I guess I could provide the assembler output of my chess
program from both compilers, if anyone is interested.

David


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]