On Nov 8, 2004, at 1:34 AM, Leopold Toetsch wrote:
Jeff Clites <[EMAIL PROTECTED]> wrote:
OTOH it doesn't really matter, if the context structure is in the frame too. We'd just need to skip that gap. REG_INT(64) or I64 is as valid as I0 or I4, as long as it's assured, that it's exactly addressing the incoming argument area of the called function.
A problem with this is that you can't be sure that you can actually have the "next" frame of registers adjacent to the current frame--they might already be taken. Imagine A calls B, then B creates a continuation and stores it in a global, then returns.
Please read the proposal summary by Miroslav Silovic, keyword "watermark".
If frames aren't adjacent, normal argument copying can be done anyway.
This would seem to require the same types of runtime checks that you are objecting to below, unless user code is expected to explicitly check whether it's supposed to be assigning to I3 v I(3 + 32x?). And that still seems to require copying, in my case of a function a() which calls a function b() with the exact same arguments.
Keep the old-scheme registers inside the interpreter structure, *as well as* the new indirect registers. Call the registers in the interpreter I0 to I31, and the indirect registers I32 to whatever.
That would need two different addressing modes depending on the register
number. That'll lead to considerable code bloat: we'd have all possible
permutations for direct/indirect registers. Doing it at runtime only
would be a serious slowdown.
I discussed that in item (1) under "Further notes".
It's not needed. I've a better scheme in mind, which addressess efficieny as well as argument passing.
And spilling?
JEff
