Optimizationsdisabled?

[Mike Jakubik]

David O'Brien said:

> On Fri, Jun 11, 2004 at 01:31:42PM -0400, Mike Jakubik wrote:
>> Hello,
>>
>> I have recntly aquired an AMD64 machine, and during an update to
>> -CURRENT
>> i noticed that the compiler options seem to eliminate a lot of
>> optimizations such as mfpmath=387 -mno-sse -mno-sse2 -mno-mmx -mno-3dnow
>> -msoft-float etc.. Does this mean all of these are disbled? Also,
>> setting
>
> One can't use FP in the kernel as in userland.  This is quite normal for
> UNIX kernels.  You don't see all the "-mno-" on FreeBSD/i386 because GCC
> doesn't default to the "ON" state for those by default [because older
> i386 family CPU's don't support those features].
>
> PowerPC and Sparc64 also use -msoft-float.
>
>> CPUTYPE to athlon-xp does not seem to affect the optimizations at all
>> (even when compiling ports). I was hoping to run a full 64bit system,
>> but
>> this is sort of discouraging. Can some shed some light as to why this
>> is?
>
> Why is this "so discouraging"?  Have you benchmarked anything on your
> 64-bit AMD64 machine?  You'll find it beats the sh*t out of other
> machines.

I ran some synthetic becnmarks (ubench) compared to an old athlon 1400mhz.
The CPU results were not as for apart as i hoped, but the memory
throughput was great. I also ran bytebench, but for some reason the
context switch test does not run correctly, and returns a 0. All other
results however were good.

> An Opteron is not an athlon-xp, please don't set CPUTYPE telling it so in
> 64-bit mode.  There is a single AMD64 CPU core, so GCC knows how to fully
> schedule for it by default (what other default could gcc have??).

Will do. Thanks for the explanation guys, and keep up the good work. Im
quite happy with the stability so far, we shall see once it goes into
production though :)