Win32 versus X64 speedup

[jaeger0]

I think to speedup my application from win32 to a x64 bit application.

I do a lot of computations (finte element analysis), which takes about 30 minutes.
So I use the Intel MKL library.
My question is, will the change to x64 result in a remarcable speedup or not ?

Another question. I use the f90gl library, which is build in win32, so I think I have to recompile the library to x64. This is a lot of work. Has anyone experience with that ? Any problems with x64,

[Andrew_Smith]

Quoting - jaeger0

I think to speedup my application from win32 to a x64 bit application.

I do a lot of computations (finte element analysis), which takes about 30 minutes.
So I use the Intel MKL library.
My question is, will the change to x64 result in a remarcable speedup or not ?

Another question. I use the f90gl library, which is build in win32, so I think I have to recompile the library to x64. This is a lot of work. Has anyone experience with that ? Any problems with x64,

Basicaly No. A 64 bit application still goes at the same clock speed and calculation rate as a 32 bit app. It could actually be slower in 64 bit because the memory addressing uses bigger pointers.

However, a 64 bit FE application could be faster for you if it allows you to do big solves in core.A 32 bit FE solvewill require an out of core solve i.e. one that required temporary files to substitute for memory once you go above about 2Gb.

Andy Smith

[Steve_Nuchia]

Beg to differ, sir. The datapaths for FP don't change but the number of available registers does. This often results in markedly better code generation. Not orders of magnitude, but good.

Integer code, like bignum libraries, can see a 4:1 speedup or a bit better moving to 64-bit processing. FP won't see anywhere near that but I'd be surprised if you didn't see measurable improvement.