Int32 and boxing

Int32 and boxing

[Jon Harrop]

Are int32s are boxed on a 64-bit machine?

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
OCaml for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists/?e

Re: [Caml-list] Int32 and boxing

[Richard Jones]

On Mon, Aug 13, 2007 at 05:08:09AM +0100, Jon Harrop wrote:
> 
> Are int32s are boxed on a 64-bit machine?

Unfortunately yes:

# Obj.is_block (Obj.repr 1_l);;
- : bool = true
# Sys.word_size;;
- : int = 64

I deal a lot with int32 and int64 types when interfacing with other
languages, so this causes an annoyance.  I'm not sure what the
implications would be of having a different representation on 32 and
64 bit machines.  I guess Marshal would have to go through hoops for a
start.

Rich.

-- 
Richard Jones
Red Hat

Re: [Caml-list] Int32 and boxing

[Jon Harrop]

On Monday 13 August 2007 18:55:01 Richard Jones wrote:
> I deal a lot with int32 and int64 types when interfacing with other
> languages, so this causes an annoyance.  I'm not sure what the
> implications would be of having a different representation on 32 and
> 64 bit machines.  I guess Marshal would have to go through hoops for a
> start.

Now that I think about it in more detail, this would probably require deeper 
compiler changes than I had thought. You wouldn't need to alter the run-time 
but you'd need to emit different load and store code for unboxed float32s, 
tagging and untagging them, and do arithmetic in 32-bit registers.

I'm not sure Marshal would need changing though. The representation would be 
embedded in nativeint so it would just be a type difference.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
OCaml for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists/?e