Replies to post #12460 on Advanced Micro Devices Inc (AMD)

[wbmw]

Aixman, Re: SMP (and iHT) on the other hand, require algorithmic changes, sometimes quite serious. Programs have to be restructured, parallel branches balanced for good performance, variables need to be designated local/shared, etc. Of course, whatever programs are multi-threaded already would use SMP as before. Others will stay with single threads simply because SMP is too damn time consuming to code for. Certainly very very few will suddenly decide to rewrite for SMP just because Intel said so.

Unlike AMD, Intel writes their own software development tools, which reduce the complexity of optimizing software for their platform. When a company runs into problems because they have pointers among thousands of lines of code that don't translate very well into 64-bit, they won't have AMD on the other side of the tech support line. But if they were developing for multithreading, they'd have the tools and support from Intel.

http://www.intel.com/software/products/threading/

[kpf]

aixman

Thanks lot for your take from a programmer's perspective.

Converting to AMD64 is basically the same as making the code 32/64 bit clean" seems. Thus anybody dealing with 64-bit platforms can get the application to run under AMD64 in almost no time

I completely agree reading your these statements together.
We can see getting software to run seems to be possible in almost no time; but there are strong indications that there is (a whole lot?) more required to port software to AMD-64 to take full advantage of its capabilities - iaw providing real benefits.

I frankly admit I am very far from overseeing what the above "a whole lot" exactly is. And I am sure its not all on the programming side, but on the compiler- and library-side as well. As i see applications gaining performance from being ported to AMD-64 and applications dropping in performance, both using the same compiler, it sure seems there is something to be done on the programming-level - just that I would not be able to tell what it is.

In essence, maybe the misconception you see is not that huge at all, carefully said. K