R devel - Dec 2011 - Initializing a large data structure to be accessed strictly within a shared C library

Dear R-devel members,

The question:  Is it possible to initialize and later free a large data
structure strictly within a shared C library, to be used by a function in
the C library that I'll call from R--without ever having to pass data to
and from R? This is analogous to C++ object initialization/use/destruction,
but if possible I'd like to stay in C.

The context: I'm implementing a particle swarm optimization of a
60-dimension nonlinear transform, where the transform is defined in a
half-gigabyte dataset. By carefully initializing a C struct I can trim a
large amount of work from the PSO iteration stage. This is, of course,
straight forward if I implement the whole thing in a self-contained C
program--however, I'd like R to handle the optimization routines, and my
shared library to implement the value function.

So: what do folks think?

Cheers,

James

	[[alternative HTML version deleted]]

Dear R-help members,

*(My apologies for cross-posting to both R-help and R-devel -- this
question straddles both domains...)*

The question:  Is it possible to initialize and later free a large data
structure strictly within a shared C library, to be used by a function in
the C library that I'll call from R--without ever having to pass data to
and from R? This is analogous to C++ object initialization/use/destruction,
but if possible I'd like to stay in C.

The context: I'm implementing a particle swarm optimization of a
60-dimension nonlinear transform, where the transform is defined in a
half-gigabyte dataset. By carefully initializing a C struct I can trim a
large amount of work from the PSO iteration stage. This is, of course,
straight forward if I implement the whole thing in a self-contained C
program--however, I'd like R to handle the optimization routines, and my
shared library to implement the value function.

So: what do folks think?

Cheers,

James

	[[alternative HTML version deleted]]

Dear R-help members,

*(My apologies for cross-posting to both R-help and R-devel -- this
question straddles both domains...)*

The question:  Is it possible to initialize and later free a large data
structure strictly within a shared C library, to be used by a function in
the C library that I'll call from R--without ever having to pass data to
and from R? This is analogous to C++ object initialization/use/destruction,
but if possible I'd like to stay in C.

The context: I'm implementing a particle swarm optimization of a
60-dimension nonlinear transform, where the transform is defined in a
half-gigabyte dataset. By carefully initializing a C struct I can trim a
large amount of work from the PSO iteration stage. This is, of course,
straight forward if I implement the whole thing in a self-contained C
program--however, I'd like R to handle the optimization routines, and my
shared library to implement the value function.

So: what do folks think?

Cheers,

James

	[[alternative HTML version deleted]]

Dear R-help members,

*(My apologies for cross-posting to both R-help and R-devel -- this
question straddles both domains...)*

The question:  Is it possible to initialize and later free a large data
structure strictly within a shared C library, to be used by a function in
the C library that I'll call from R--without ever having to pass data to
and from R? This is analogous to C++ object initialization/use/destruction,
but if possible I'd like to stay in C.

The context: I'm implementing a particle swarm optimization of a
60-dimension nonlinear transform, where the transform is defined in a
half-gigabyte dataset. By carefully initializing a C struct I can trim a
large amount of work from the PSO iteration stage. This is, of course,
straight forward if I implement the whole thing in a self-contained C
program--however, I'd like R to handle the optimization routines, and my
shared library to implement the value function.

So: what do folks think?

Cheers,

James

On 11-12-27 5:58 PM, James Muller wrote:
> Dear R-devel members,
>
> The question:  Is it possible to initialize and later free a large data
> structure strictly within a shared C library, to be used by a function in
> the C library that I'll call from R--without ever having to pass data
to
> and from R? This is analogous to C++ object initialization/use/destruction,
> but if possible I'd like to stay in C.
Yes, if you use malloc or Calloc you'll get memory that you can manage 
yourself.  See Writing R Extensions, section 6.1.2.

Duncan Murdoch
>
> The context: I'm implementing a particle swarm optimization of a
> 60-dimension nonlinear transform, where the transform is defined in a
> half-gigabyte dataset. By carefully initializing a C struct I can trim a
> large amount of work from the PSO iteration stage. This is, of course,
> straight forward if I implement the whole thing in a self-contained C
> program--however, I'd like R to handle the optimization routines, and
my
> shared library to implement the value function.
>
> So: what do folks think?
>
> Cheers,
>
> James
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel