Does D351 Unwiring-Integer patch now look like you envisioned?

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Does D351 Unwiring-Integer patch now look like you envisioned?

Simon Peyton Jones
I replied, but failed to press "Submit" (which is several screenfuls away). Sorry

The Maybe DataCon idea looks right to me.


FWIW I *hate* the way that TidyPgm is forced to predict what CorePrep will do.  I've created https://ghc.haskell.org/trac/ghc/ticket/9718 to explain.

Simon


|  -----Original Message-----
|  From: Herbert Valerio Riedel [mailto:hvriedel at gmail.com]
|  Sent: 24 October 2014 08:48
|  To: Simon Peyton Jones
|  Subject: Does D351 Unwiring-Integer patch now look like you
|  envisioned?
|  
|  Hello Simon,
|  
|  I was wondering if
|  
|    https://phabricator.haskell.org/D351
|  
|  looks the way you expected, and more specifically I'd like some
|  feedback on the `DataCon` vs. `Maybe Id` comment at
|  
|    https://phabricator.haskell.org/D351#8682
|  
|  Thanks!
|    hvr

Reply | Threaded
Open this post in threaded view
|

Does D351 Unwiring-Integer patch now look like you envisioned?

Herbert Valerio Riedel-3
Hello!

On 2014-10-24 at 10:08:00 +0200, Simon Peyton Jones wrote:
> I replied, but failed to press "Submit" (which is several screenfuls
> away). Sorry

Thanks!

(if you press the 'z'-key outside of any input-form on code-revision
pages, you get the screen horizontally tiled with the submit-form in the
bottom half - this is sometimes generally useful to have IMHO)

> The Maybe DataCon idea looks right to me.
>
>
> FWIW I *hate* the way that TidyPgm is forced to predict what CorePrep
> will do.  I've created https://ghc.haskell.org/trac/ghc/ticket/9718 to
> explain.

Btw, this reminds me I have a low-priority plan to improve the
`mkInteger` call interface, as the current one is terrible: it requires
splitting a large integer literal into 31-bit words (even when machine
wordsize is 64bit!), and then wrapping those into an ordinary
[Int]-list.

I'd rather like mkInteger to take an packed array of machine-size words,
similar to how [Char] literals are handled via `unpackCString#`, which
would allow for a more compact representation in object files as well as
possibly direct conversion of integer literals in GHC's Integer
backends... I've created a ticket to keep track of that idea:

  https://ghc.haskell.org/trac/ghc/ticket/9719

Cheers,
  hvr