The current code causes floating point miscalculations on AMD Ryzen. PR port-amd64/52966: amd64 FPU handling broken on AMD