The current xbyak_aarch64 encodes fcpy(z0.s, p0, 0) as fcpy(z0.s, p0, 1).
It's very misleading and debug-hard behavior.
So I fixed xbyak_aarch64 to throw an exception unless the immediate value can be correctly encoded. And, I simplified the original algorithm.
The current xbyak_aarch64 encodes
fcpy(z0.s, p0, 0)asfcpy(z0.s, p0, 1). It's very misleading and debug-hard behavior. So I fixed xbyak_aarch64 to throw an exception unless the immediate value can be correctly encoded. And, I simplified the original algorithm.