Currently, the emitter has support for emitting 4-byte AVX instructions throughout most of the emitOutput* code paths in the emitter.
However, the support for emitting 4-byte SSE instructions only exists in a couple of the emitOutput* code paths and requires non-trivial work to enable elsewhere.
We should plumb the rest of this support through the emitter in order to enable several of the 4-byte SSE4.1 and SSE4.2 intrinsics on machines which don't support the VEX encoding (non-AVX machines).
This will also allow the Math.Round, Math.Ceiling, and Math.Floor intrinsics to be enabled for these architectures as well.
Currently, the emitter has support for emitting 4-byte AVX instructions throughout most of the
emitOutput*code paths in the emitter.However, the support for emitting 4-byte SSE instructions only exists in a couple of the
emitOutput*code paths and requires non-trivial work to enable elsewhere.We should plumb the rest of this support through the emitter in order to enable several of the 4-byte SSE4.1 and SSE4.2 intrinsics on machines which don't support the VEX encoding (non-AVX machines).
This will also allow the
Math.Round,Math.Ceiling, andMath.Floorintrinsics to be enabled for these architectures as well.