Malcolm Wallace wrote:
Simon Marlow <[EMAIL PROTECTED]> wrote:
Calling a function on entry to every single BCO would be
very costly.
The interpreter could check before entering a BCO whether there was a
breakpoint set on that BCO. That's nearly equivalent to the
implementation I proposed before (overwriting the BCO with a
breakpoint object) and might be simpler, I'm not sure.
Just storing a flag byte in the header of every BCO would be very cheap
I imagine. And there must be some overhead in entering the interpreter
for a BCO anyway, so adding an extra check on the "breakpoint" flag at
the beginning of the interpreter loop should not cause a great slowdown?
I don't know why people are so interested in this minor aspect of the
design! But sure, adding a flag to the BCO object is another
possibility. Or just changing its header from a BCO to a BCO_BREAK object.
Some other questions are: how to map from source location to BCO (GHC
doesn't retain enough information except at top-level functions right
now), and how to add information about free variables to the BCO. These
questions are sidestepped by Lemmih's implementation, fortunately.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-ghc
