floating point optimizations around SSE4, AVX etc
more and better floating point optimizations around SSE4, AVX etc
why this is a crucial feature for high performance programming --
either do it on x86 platform or support CUDA ..,. just to be competitive.
Is it many/multi-core CPU-GPU or GPU-CPU?
This is a CLR feature request. You should move your votes over the .NET section.
+1 for options to automatically generate AVX2 as with the /arch:AVX flag, FMA, and vectorization.
It would be nice if VC++ would also support the upcoming AVX2 and provide us with options to automatically generate the int 256 version of AVX.
First thing would just put in 80 bit doubles - I mean, it's like the stone age having to write C++ DLLs so I can use basic maths functions with a decent level of precision. As I am having to use Wolfram Mathematica libraries for not particularly exotic math functions, any chance of Microsoft providing a decent library of math functions ?