Given that this compiler is obviously targeted at and optimised for Intel CPUs, I am curious to know if it is "beneficial" when the executables are run on non-Intel CPUs like AMD?
Does anyone know?
Are there any actual tests/benchmarks that highlight the improvement (or otherwise) of using Intel's compiler instead of GCC, clang etc. on other vendor's CPUs?
- Development Tools
- Intel® C++ Compiler
- Intel® Parallel Studio XE
- Intel® System Studio
- Parallel Computing
I cannot really say. But what I know for sure is that Intel MKL is not optimized for the latest AMD CPUs and you can get better performance using another BLAS library say BLIS.
One guy working for AMD claimed to me that gcc gives better code than Intel C fro AMD CPUs. But for our numerical intensive code switching to another BLAS library while staying with Intel C for AMD CPUs did the trick.
This is intended to occur, that AVX will be used if available. For AMD, it may not be possible, in my opinion, to catch cases where AVX128 may be better than AVX256, although for basic AVX without AVX2, (sandy bridge) , possibly unaligned loads and stores will be split down to AVX128.