64 bit F7

Jim Cornette fc-cornette at insight.rr.com
Wed Jul 18 10:46:02 UTC 2007


Chris Jones wrote:
> 
>>> As they say: "Where is the killer app for 64-bit computing?"
>>
>> Depends on who the "user" is, and what a "killer app" is to that user.
>>
>> Most things that consumers need don't approach requiring (or seriously
>> benefitting from) 64 bits. But some servers and workstations can get a
>> lot from the wider data path and bigger address space.
> 
> Basically, code built in 64 bit mode is normally faster than the same 
> code built in 32 bit mode. I say is normally, as it does depend on the 
> code. Also, if you search the web for this you will find lots of 
> interesting discussions about whether it is really the 64 bits versus 32 
> that make the difference, or the fact that 64 bit gcc uses SSE/SSE2 by 
> default whereas 32 bit gcc builds do not.
> 
> Bottom line for me is I have a scientific number crunching application - 
> Compiling and running the same code, on the same machine in 32 or 64 bit 
> mode, I find the 64 bit build is about 40% faster (about 20% if I enable 
> SSE/SSE2 in 32 bit mode). This is very significant and for me makes 64 
> bit builds well worth it.
> 
> Chris
> 

Since requirements for computers are more intense now, building 32-bit 
with SSE/SSE2 by default would make sense, especially if a 20% 
improvement is noticed in comparison.

If the code does not make a difference now, changing the default would 
allow implementing code in the future that takes advantage of the 
features. Also both 64-bit and 32-bit builds would be closer in line.

Interesting information, though the debate ws already encountered for 
i386 vs. i686.

Jim

-- 
The less time planning, the more time programming.




More information about the fedora-list mailing list