This is the mail archive of the
kawa@sourceware.org
mailing list for the Kawa project.
Re: Request for GSoC project Typed/optimized arithmetic with unit
Thanks, I'll take a look.
On 4 Mar 2018 12:46 p.m., "Per Bothner" <per@bothner.com> wrote:
> On 03/03/2018 09:35 AM, SANAN ANSARI wrote:
>
>> Hi , I'm Sanan Ansari. I am pursing my B.E in computer Engineering from
>> Mumbai university India.I just read the project idea of Typed/optimized
>> arithmetic with unit. I have knowledge about Java.I would love to work on
>> the project. Please .If you can give more documentation about project it
>> will help me.
>>
>> By the way this my first GSoC and I’m new to open source , real world
>> project.
>>
>> Hope you will help me as I’m a beginner.
>>
>
> Welcome! The work needed to implement "Types for units" is moderate, but
> it does require going fairly deep into low-level parts of the Kawa
> compiler.
>
> Most of the work for "Types for units" will be in the Java language, but
> it won't make sense without at least basic familiarity with the Kawa
> dialect of Scheme.
> Plus test-cases and examples would be written in Scheme. So start by
> downloading,
> building, and using Kawa. Write scripts and programs in it. Perhaps
> focus in
> on some Kawa feature that seems interesting/fun and go a bit deeper: For
> example
> Kawa as a web server - or "Composable pictures" - or using it in place of
> a shell script.
>
> Also look at the Kawa number hierarchy, especially the Quantities page:
> https://www.gnu.org/software/kawa/Quantities.html
> This doesn't go into depth, but should point in the right direction.
> Also look at how Quantities are implemented - look in gnu/math in the
> source code.
>
> As a model of what we want to do, look at how Kawa implements the unsigned
> types
> ulong/uin/ushort/ubyte. Look at the dis-assembled bytecode (using either
> the class gnu.byteode.dump or the standard javap). Understand how ''ulong'
> gets compiled to a plain long (when the types are known) vs how it is
> represented
> as a gnu.math.ULog (when we don't know the type at compile-time).
>
> In the same way that a 'ulong' is either a long or a gnu.math.ULong,
> we want the type quantity[cm^2] to be either a double or a
> gnu.math.DQuantity.
> Part of the goal is compile-time type-checking; part of the goal is to
> generate
> efficient code, using just double arithmetic when possible. In the latter
> case,
> all of the unit checking and combing would be done at compile-time.
>
> All of this stuff is in the Kawa manual at https://www.gnu.org/software/k
> awa/
> --
> --Per Bothner
> per@bothner.com http://per.bothner.com/
>