x64Emitter: optimize immediate sizes

[FioraAeterna]

A nice alternative than trying to do it throughout the JIT.

[Tilka]

lgtm

[shuffle2]

lgtm, however considering #749, it looks like jit statements may be implicitly written as a shorter form than the writer of a jit statement expects...? Not sure if that's an issue or not.

[FioraAeterna]

Could that ever be a bad thing? I think most assemblers already do that internally -- is there some place we want longer forms?

[shuffle2]

It depends if anyone has written jit code with presumptions about
instruction length (for example manually calculating some pointers and
placing other instructions there). I doubt it, just pointing it out.

On Tue, Aug 12, 2014 at 2:07 PM, Fiora notifications@github.com wrote:

Could that ever be a bad thing? I think most assemblers already do that
internally -- is there some place we want longer forms?

—
Reply to this email directly or view it on GitHub
#754 (comment).

[FioraAeterna]

That sounds like it'd be really dangerous, since it'd depend on whether an instruction had a REX prefix or not

[delroth]

FYI I know that some of the backpatch stuff has/had assumptions of
instruction size to insert NOPs. Might be worth rechecking.

On Tue, Aug 12, 2014 at 2:19 PM, shuffle2 notifications@github.com wrote:

Merged #754 #754.

—
Reply to this email directly or view it on GitHub
#754 (comment).

Pierre "delroth" Bourdon delroth@gmail.com
Software Engineer @ Zürich, Switzerland
http://code.delroth.net/

[FioraAeterna]

I think it should be okay; the backpatch code doesn't seem to use immediates at all, and it already has a check to make sure the size of the code is sufficient (it inserts nops otherwise).