Enable AVX2 on Clang (GCC Frontend) builds in Windows

[hmartinez82]

No description provided.

[azonenberg]

Can we get @bvernoux or any other Windows users/devs to weigh in on this? I see nothing wrong with the patch from a code style/quality perspective, it's a trivial one-line edit, but don't actively use ngscopeclient on Windows so can't easily test if it breaks things.

[bvernoux]

All seems fine but I have not tested Clang version
The modification is small anyway with "&& !defined(clang)" to avoid removing AVX on Clang

[azonenberg]

Yes, the modification is trivial. But we need someone to actually thoroughly stress test a clang-compiled ngscopeclient on windows to verify that it doesn't hit any edge cases and crash due to AVX issues.

[bvernoux]

So far I have no way to Build and Test it with Clang on Windows X86_64 (My setup is only to build with MingW64 with GCC and I do not have really the time to configure all for Clang, I'm not even sure there is any interest to keep Clang as that requires to test on different Toolchain which is quite time consuming especially with all feature present in ngscopeclient ...) maybe we can merge it and test on the automatic build and ask to other guys with Windows X86/X64 ?

[azonenberg]

Update, we've deprecated all usage of avx_mathfun which was I believe the only code that passed AVX registers across function call boundaries. I believe this is what broke, not just using AVX in a local function - so we should in theory be able to enable AVX for all x86 platforms at this time.

But again, not a Windows user so can't test.