Hello,
I have attached my draft application.
I would like to know your opinion!
https://docs.google.com/document/d/1f_7v1tnZO7aBM5TbOzV5rWpqqxbtPfSqXX4lrz3GcKk/edit?usp=sharing
Thank you,
Rajath Shashidhara
On Fri, Mar 18, 2016 at 3:34 AM, Vassily Litvinov <[email protected]> wrote:
> Hi Rajath,
>
> On 03/17/16 13:30, Rajath Shashidhara wrote:
>
>> ......
>> [1] Why is the module "Futures" referenced in the mail above not used
>> currently ?
>> Are there any particular flaws in this module ? This will help me
>> identify the pitfalls and understand how my implementation must be
>> different from it.
>>
>
> It is not used because its implementation has not reached the state where
> we are happy to merge it onto our master branch.
>
> Also its implementation is based on lambda constructs. If possible, we
> would like to avoid relying on lambdas because it is not a well-implemented
> feature at the moment.
>
> [2] Is there any reason behind why "single" qualifier is not supported
>> for record types ?
>> i.e., Is it not implemented or is it not supported deliberately for
>> some reason ?
>>
>
> Good question. It is deliberate because we do not have a good answer to
> the following scenario:
>
> Consider a record r with fields x,y,z. Let is have a "single" type and not
> be "full" yet. Say you execute r.x = 5. Should that make r be "full" ?
>
> [3] What work needs to be done on the run-time library for begin
>> expressions support ?
>> As Michael has stated in the previous mail, the library is
>> thread-safe. If thread-safety is not the problem, I can't think of any
>> other problem.
>>
>
> I do not see anything immediately. I cannot guarantee that it will not
> surface during the course of the project.
>
> [4] Can begin expressions contain blocks ? i.e., a = begin {z=fn(b);
>> fn1(z)} + begin fn(c); [...]
>>
>
> That will surely be valuable.
>
> Aside:
>> Is there a way to print the AST produced by the compiler ?
>>
>
> When you are executing the compiler from the debugger, you can use the
> functions print_view, list_view et al. passing them either a pointer to
> BaseAST or an integer node ID as argument. We have a couple of handy gdb
> aliases in compiler/etc/gdb.commands
>
> Otherwise there is the option -d or --log that you can pass to the
> compiler (see chpl --help --devel for more details and related options). It
> will create a bunch of per-pass per-module files in the log directory.
>
> Vassily
>
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Chapel-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-developers
