Miscompilation for sse4-i8x16

[nurmukhametov]

The following code

void SetByte(int8& dest, int8 src)
{
    uniform uint32 mask = lanemask();
    unmasked
    {
        dest = select(((1 << programIndex) & mask) != 0, src, dest);
    }
}

is compiled for sse4-i8x16 target to

SetByte___REFvytvyt:                    # @SetByte___REFvytvyt
        ret

It looks like miscompilation.

[nurmukhametov]

This was actually fixed in v1.24.0 by #2800