Re: [boost] Accelerating algorithms with SIMD - Segmented iterators and alternatives

Subject: Re: [boost] Accelerating algorithms with SIMD - Segmented iterators and alternatives
From: Thomas Klimpel (Thomas.Klimpel_at_[hidden])
Date: 2010-10-12 15:22:58

DE wrote:
> for example intel compiler for win32 allows to generate two versions
> binaries from one source: plain x86 and say sse2 which are switched at
> runtime depending on cpuid
> 
> so the question still stands

Question to Joel:
Will this compiler automatically do the correct thing with NT2 without any further user intervention?

If I understood correctly, NT2 will know at compile time which SIMD instruction set is available, and generate the appropriate code. Now if the intel compiler compiles the code multiple times for different target architectures, I imagine that NT2 would generate the appropriate code during the different compile runs for the different architectures, so that the resulting executable should work optimal right out of the box. Is this correct?

Regards,
Thomas