Leo~ On 8/10/05, Leopold Toetsch <[EMAIL PROTECTED]> wrote: > [ /me warnock fighting - and update below for method calls ] > > -------- Original Message -------- > Subject: call opcodes cleanup > Date: Thu, 21 Jul 2005 09:52:07 +0200 > From: Leopold Toetsch <[EMAIL PROTECTED]> > To: Perl 6 Internals <perl6-internals@perl.org> > > > There are still some leftovers of the old calling scheme, which I'd like > to get rid of (1,2) or change slightly (3): > > 1) bare invoke > > This opcode assumes P0 = Sub, P1 = Continuation > > 2) bare invokecc > > This opcode assumes P0 = Sub > > 3) invoke Px > > This opcode assumes P1 = Continuation, but it has a double usage: call a > coroutine (not the first time) and yield from a coroutine with this code > sequence: > > interpinfo Px, .INTERPINFO_CURRENT_SUB > invoke Px > > Instead I'd like to have distinct opcodes, that carry the effects on > operands clearly in their signature. The replacement for above would be: > > 1) invoke PSub, PCont # call sub with given continuation > 2) invokecc PSub # call sub, create continuation - exists > 3a) invoke PCoro # call sub, leave continuation as is > 3b) yield # yield results from a coroutine > > 3a) exists too, but it currently picks up a continuation in P1. > > Rational: the register allocator has to track the usage of all symbols. > For all the implicit register usage, extra code is needed that tracks > register usage. See trunk imcc/cfg.c for details.
I like this, especially the motivation. > ------------- > > Method call opcodes > > We currently have these opcodes: > > 4a) callmethod > > Implicit registers used: S0 = method, P1 = continuation, P2 = > invocant/object > > 4b) callmethod Smeth > > 5a) callmethodcc > 5b) callmethodcc Smeth > > We probably need just one opcode: > > 5) callmethodcc Pinvocant, Smeth > (maybe callmethod Pinvocant, Smeth, Pcont) I think that we should have the second option (as that is an easy way to have tailmethods). We might also want a C< callmethod Pinvocant, Pmeth, Pcont > if we allow methods that are pmcs. (This might allow hoisting the string lookup outside a loop or something). Matt -- "Computer Science is merely the post-Turing Decline of Formal Systems Theory." -Stan Kelly-Bootle, The Devil's DP Dictionary
