• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Instructions?

Really? So most programs don't even use the current instruction sets SSE3, SSE4 etc? I wonder how much more performance you could squeeze out by using an optimized compiler as opposed to the lowest common denominator. I saw ivy bridge will have a hardware random number generation which seems like it could be a boon for scientific computing.

Isn't there like an ISO for CPU instructions? If you claim your computer is ISO 12222 it should support these instructions, kind of a deal? They should do that if they haven't.
 
Open64 is Linux only and GCC compiled software for Windows is in the minority. MS's compiler only goes to SSE2 and Intel's compiler won't be supporting AMD only extensions (XOP/FMA4).
 
How much difference in performance will new instructions make in the coming CPUs?

AVX2, FMA3&4, SSE5 etc.
AVX2 will start a revolution. It finally adds parallel load operations, allowing code to be parallelized much more easily and run up to eight times faster than serial code. And together with FMA you basically get the same instructions which make GPUs so powerful. The CPU is much more flexible though, so you get the best of both worlds.

And since it will be supported by OpenCL and C++ AMP implementations, developers will get easy access to this technology (while retaining compatibility with older CPUs).
 
Back
Top