Closed Mixaill closed 4 years ago
According to AMD specs (https://www.amd.com/system/files/TechDocs/25481.pdf (page 20)), for FMA4 detection we should use 16th bit of ECX register in 8000_0001 function.
It fixes crash on Zen2 CPUs, which do not support FMA4.
Fixed in this commit.
Thanks for finding this one!
I had it correct in this blog post, but it was pretty much always wrong in the DirectXMathFMA4.h header :(
According to AMD specs (https://www.amd.com/system/files/TechDocs/25481.pdf (page 20)), for FMA4 detection we should use 16th bit of ECX register in 8000_0001 function.
It fixes crash on Zen2 CPUs, which do not support FMA4.