[Caml-list] "noalloc" + enter/leave blocking section - safe?

[Caml-list] "noalloc" + enter/leave blocking section - safe?

[Markus Mottl]

Hi,

one important question: is it safe to use "noalloc" with
C-functions that contain calls to "caml_enter_blocking_section" and
"caml_leave_blocking_section" to allow other POSIX-treads?

I have the following external function:

  (** Synchronize all filesystem buffers with disk. *)
  external sync : unit -> unit = "unix_sync" "noalloc"

It is implemented as follows:

  CAMLprim value unix_sync()
  {
    caml_enter_blocking_section();
      sync();
    caml_leave_blocking_section();
    return Val_unit;
  }

When removing the "noalloc" attribute, I cannot reproduce the problem
(crash) reported in bug report #3019 anymore, but I'm not sure whether
this is just a coincidence.

Of course, functions with this attribute shouldn't allocate anything on
the OCaml-heap or throw exceptions.  But I thought that context-switches
would be safe?

Best regards,
Markus

-- 
Markus Mottl          http://www.oefai.at/~markus          markus@oefai.at

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] "noalloc" + enter/leave blocking section - safe?

[Markus Mottl]

On Mon, 02 Aug 2004, Markus Mottl wrote:
> one important question: is it safe to use "noalloc" with
> C-functions that contain calls to "caml_enter_blocking_section" and
> "caml_leave_blocking_section" to allow other POSIX-treads?

Ok, after having done some experiments on my own now, it seems that this
is _not_ safe.  Here is a small example that demonstrates this:

file: foo.ml
---------------------------------------------------------------------------
external foo : 'a -> unit = "foo_stub" "noalloc"

let thread () =
  let ar = Array.init 1 (fun _ -> "") in
  foo ();
  let ar = Array.init 1 (fun _ -> "") in
  ()

let () =
  let n = 10 in
  let ar = Array.create n () in
  let tids = Array.map (Thread.create thread) ar in
  Array.iter Thread.join tids
---------------------------------------------------------------------------


file: foo_stubs.c
---------------------------------------------------------------------------
#include <caml/mlvalues.h>
#include <caml/signals.h>

CAMLprim value foo_stub(value v) {
  caml_enter_blocking_section();
    int i;
    for (i = 0; i < 100000000; i++);
  caml_leave_blocking_section();
  return Val_unit;
}
---------------------------------------------------------------------------

Compiling and linking the above example to native code (i.e. with
POSIX-threads) and running it will almost immediately produce a segfault
on my machine.  Removing "noalloc" solves this problem.

Could the documentation on the C-interface please be updated with a
small section on the correct use of "noalloc"?  It's not documented at
all right now!  (This also closes bug report #3019, I guess).

Regards,
Markus

-- 
Markus Mottl          http://www.oefai.at/~markus          markus@oefai.at

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] "noalloc" + enter/leave blocking section - safe?

[Jere Sanisalo]

On Mon, Aug 02, 2004 at 05:15:02PM +0200, Markus Mottl wrote:
>---------------------------------------------------------------------------
>external foo : 'a -> unit = "foo_stub" "noalloc"

Hmm, I wasn't aware that there was a "noalloc" directive. What does it
specifically declare? Does it make ocaml call the natives in a more
efficient manner (and of course the function must not allocate anything)?
Are there any other directives I should be aware of?

And according to the documents the "noalloc" should really be the name of
the native function? Am I really missing something here?

-- 
Jere Sanisalo [xm@xmunkki.org] - http://www.xmunkki.org/

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] "noalloc" + enter/leave blocking section - safe?

[Falk Hueffner]

Jere Sanisalo <xm@xmunkki.org> writes:

> On Mon, Aug 02, 2004 at 05:15:02PM +0200, Markus Mottl wrote:
>>---------------------------------------------------------------------------
>>external foo : 'a -> unit = "foo_stub" "noalloc"
>
> Hmm, I wasn't aware that there was a "noalloc" directive. What does
> it specifically declare? Does it make ocaml call the natives in a
> more efficient manner (and of course the function must not allocate
> anything)?

It calls the C function directly, with zero overhead. Otherwise, the
call goes through "caml_c_call", which puts the GC state into global
variables for C to pick up, and restores it later from there.

> Are there any other directives I should be aware of?

There's "float", which means the function works on unboxed floats.

-- 
	Falk

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] "noalloc" + enter/leave blocking section - safe?

[Richard Jones]

On Mon, Aug 02, 2004 at 05:49:53PM +0200, Falk Hueffner wrote:
> Jere Sanisalo <xm@xmunkki.org> writes:
> 
> > On Mon, Aug 02, 2004 at 05:15:02PM +0200, Markus Mottl wrote:
> >>---------------------------------------------------------------------------
> >>external foo : 'a -> unit = "foo_stub" "noalloc"
> >
> > Hmm, I wasn't aware that there was a "noalloc" directive. What does
> > it specifically declare? Does it make ocaml call the natives in a
> > more efficient manner (and of course the function must not allocate
> > anything)?
> 
> It calls the C function directly, with zero overhead. Otherwise, the
> call goes through "caml_c_call", which puts the GC state into global
> variables for C to pick up, and restores it later from there.
>
> > Are there any other directives I should be aware of?
> 
> There's "float", which means the function works on unboxed floats.

Cool features!

Rich.

-- 
Richard Jones. http://www.annexia.org/ http://www.j-london.com/
Merjis Ltd. http://www.merjis.com/ - improving website return on investment
PTHRLIB is a library for writing small, efficient and fast servers in C.
HTTP, CGI, DBI, lightweight threads: http://www.annexia.org/freeware/pthrlib/

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] "noalloc" + enter/leave blocking section - safe?

[Markus Mottl]

Falk Hueffner schrieb am Montag, den 02. August 2004:
> > Are there any other directives I should be aware of?
> 
> There's "float", which means the function works on unboxed floats.

If I understand asmcomp/cmmgen.ml correctly, this means that the function
arguments and also the return value must all be (unboxed) floats?

Best regards,
Markus

-- 
Markus Mottl          http://www.oefai.at/~markus          markus@oefai.at

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] "noalloc" + enter/leave blocking section - safe?

[Falk Hueffner]

Markus Mottl <markus@oefai.at> writes:

> Falk Hueffner schrieb am Montag, den 02. August 2004:
>> > Are there any other directives I should be aware of?
>> 
>> There's "float", which means the function works on unboxed floats.
>
> If I understand asmcomp/cmmgen.ml correctly, this means that the function
> arguments and also the return value must all be (unboxed) floats?

That's what I presumed too when I stumbled over it. I haven't examined
it in detail, though... :-)

-- 
	Falk

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners