[Yhc] CAF vs 0-arity function question

[Tom Shackell]

Hi Rob,

The constant table item constants are somewhat of a legacy. The original 
constants were chosen to correspond to nhc's constants, however as far 
as the Yhc runtime is concerned:

- A and Z are simply references to heap nodes and are treated in exactly 
the same way.
- F, 0, C, P, X are all references to Info structures and are also 
treated in exactly the same way.

However, you are quite right, looking at the C code 0 is mistakenly 
included with the A&Z code. This has likely not proved a problem because 
'0' is infact entirely redundant. The only thing you could do with a 
0-arity FInfo is make an application to it, but why would you want to 
when you can just push the CAF instead?

Ultimately we should tidy up the constants to a more simple

- Some constant value (i, l, f, d, s)
- References to heap nodes (N)
- References to FInfo or CInfo (I)

For the moment I shall change the C code to make using '0' an error :-)

Thanks

Tom



Robert Dockins wrote:

>The HBC bytecode format has different constant tags for the folloing: 
>
>1) CAF, tag 'A'
>2) 0-arity function, tag '0'
>
>http://haskell.org/haskellwiki/Yhc/RTS/hbc
>
>Why the distinction?  Maybe I don't fully understand, but I thought that a 
>0-arity function _was_ a CAF?
>
>The runtime seems to treat them very much the same (although I can't be quite 
>sure -- reading C gives me a headache ;)
>
>
>Rob Dockins
>_______________________________________________
>Yhc mailing list
>Yhc at haskell.org
>http://www.haskell.org//mailman/listinfo/yhc
>  
>