Fix build issues with MSVC for AVX512
There are couple build issues with MSVC currently. I've seen it in this pipeline. This addresses the ones I've seen so far. Test are still building. MSVC is being real slow to build and consuming quite a bit of memory.
Edit: I've seen a single process of cl.exe taking more than 60GB of memory. I think it will be more prudent to disable the recent optimizations added for AVX512 (GemmKernel and TrsmKernel) to avoid build issues on Windows with MSVC. I think this might be related to the template recursion used in both kernels. Compilation aborts with fatal error below.
fatal error C1002: compiler is out of heap space in pass 2
Edited by aaraujom