x86_64 packages vs non-64 packages
 

I'm new to Linux, trying the full-conversion tactic...
I have erased my hard drive of Windows and dove head first into Suse 10.

I'm trying to get the hang of installing new applications...
RPMs are generally simple, compiling from source is often not so easy.

I have a 64bit processor, so I've been trying to get packages labelled 64...

If one isn't 64, and I compile it myself, will it be optimized for my 64bit processor?
If I were to compile my own 64 version, as opposed to running a normal x86 version, would I see any real difference?


Thanks.

>>If one isn't 64, and I compile it myself, will it be optimized for my 64bit processor?
If I were to compile my own 64 version, as opposed to running a normal x86 version, would I see any real difference?

Compileing yourself is always better because it's customized for your system then but as for whether you notice a real difference I think it depends on the software program and if it's been designed to actually make the calls necessary to take advantage of 64bit or HyperThreading or whatever you have.

Here's a Windows example of a test I ran tonight using my Intel P4 HT EM64T (running 32bit mode though in MC2005). The product I tested was a program by Ulead for video editing. They tout that the program is designed to take advantage of HT and Dual Core processors for "high speed"!

I tested that claim by timeing the conversion of a 1.5 minute 35MB .mpg clip to .divx. It took it the same amount of time as the Divx Converter program I have which is not (at least they don't say it is) optimized: 2mins35 seconds.

How does this relate? It shows that sometimes the differences are barely noticeable. It depends on the application. I imagine the same applies to 64bit vs. 32 bit apps.

That being said, in your case I'd always get the 64bit apps or compile myself just for system stability. Yea it can be a pain because of dependencies but you may want whatever its dependent on for another package someday anyway so you may as well track them down.

That pretty much sums up some of my assumptions.
Thanks for sharing the results of your test...

I guess overall it matters how an application was written and not necessarily how it was compiled.