[haskell-llvm] LLVM.ST (Was: Upcoming release, high level binding state)

[Benjamin Saunders]

The exposed interface, being constrained to the manipulation of LLVM IR, is not known to provide facilities for executing pointer arithmetic. If you can demonstrate to the contrary, please do--or better still, provide a patch. If not, then this is not a productive line of discussion.

----- Original Message -----
From: "Henning Thielemann" <lemming at henning-thielemann.de>
To: "Benjamin Saunders" <besaunde at sfu.ca>
Cc: haskell-llvm at projects.haskell.org
Sent: Wednesday, May 15, 2013 1:46:18 PM
Subject: Re: [haskell-llvm] LLVM.ST (Was: Upcoming release,	high level binding state)


On Wed, 15 May 2013, Benjamin Saunders wrote:

> That is the purpose of LLVM.ST, yes. Note that this module is 
> experimental. If you find a case where purity is violated,

I don't have a concrete example at hand, but with some pointer arithmetics 
it should be simple to crash the machine, but runST would pretend that 
this is not possible, right?

> or believe you have a design improvement, please open an issue at 
> https://github.com/bos/llvm/issues

My design suggestion is to move this (experimental) code into a separate 
package. There is no need to bind version numbers of your experiments to 
the version number of LLVM.