> Date: Thu, 8 Oct 2015 10:29:47 -0600 > From: Paul Gilmartin <[email protected]> > Subject: Re: BXLE considered harmful > > So, should every implied register reference be documented in situ, perhaps > by a line comment? Can any implied register reference be regarded as > "common knowledge"? What about Multiply which makes an implied reference > to the odd register adjacent to its explicit target? (I assume the admirable > cross-reference list reports that.) > > -- gil > > ------------------------------ Good point. I have two comments.
First, the answer is no. It is perfectly valid, IMO, to expect your readers to "know" precisely how an instruction works. That said, given two possible code constructs which are *reasonably close* in efficiency, I would prefer that which uses "less complicated" instructions, as it means the reader can press on without having to look things up (in an xref table or in POP). Now, multiply (M, MR, etc), unfortunately, don't have efficient, simpler alternatives. So I tend to use the GF variants of multiply, as they don't require multiple, implicitly-defined registers. And ED[MK] don't have good alternatives either (though their programmer UX is apalling). Brent Longborough Sent from my typewriter -- please forgive any grammatical errors
