On Thu, May 16, 2002 at 04:25:39PM -0300, Daniel Grunblatt wrote: > > On Thu, 9 May 2002, Dan Sugalski wrote: > > > Anyone care to take this task on? It doesn't have to be at all fancy. > > In fact, all it needs to do is handle: > > > > *) Label offset calcs > > *) Explicit, fully-qualified, opcode names > > *) Opcode args as either labels or numbers > > > > So we'd only pass it: > > > > foo: sub_i_i_ic 1, 2, 5 > > ge_i_ic_l 1, 12, foo > > > > and not > > > > foo: sub I1, I2, 5 > > ge I, 12, foo > > > > Any opcodes you might think you need can be made available, just ask > > and we'll get 'em defined. > > -- > > Dan > > ok, it's committed at languages/parrot_compiler/ > > I added index and pack, I don't know if pack should check the sizeof($3), > what do you think? > > Besides labels or numbers opcode args could also be strings (The first > character after the "," will be used as the string delimiter). > > Arguments must be separated by "," no spaces are supported between them, > yet. > > I will rewrite some parts once I (or someone) gets out a bug in the hashes > (I created a hash with 722 key/value pairs and ~360 got lost)
I'll fix that bug as soon as someone fixes the bug where 6 or so of my 200 keys get lost. :-) Apparently a bug or two in string_concat. It's hard to avoid concat when you're writing stress tests for the hash code, though I suppose I could just generate huge gobs of constants.