Small code clean up

Hi,

that patch removes useless code.

Cheers,
Gwen

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

On Thu, Jun 06, 2013 at 11:40:09AM +0200, Gwenaël Casaccio wrote:

> From 4e074f382e7477c4e54ef50203f47af9a291f151 Mon Sep 17 00:00:00 2001
> From: Gwenael Casaccio <[hidden email]>
> Date: Thu, 6 Jun 2013 11:27:19 +0200
> Subject: [PATCH] Remove useless optimizations

Commit messages are read in the future and they should enclose as much
information as relevant. E.g. it should explain of why today a simple
'^instVar' is faster (e.g. avoiding to create an activation context) but
also attempt to figure out if "^ip yourself" has ever been the preferred
way.

I added a sentence to your commit message (so I am not going the entire
way) and it is currently beeing compiled and checked by travis-ci.


> - "This funny implementation thwarts the interpreter's optimizing effort"

thanks for removing that!



> Subject: [PATCH] Removes old IBM code

"IBM code" is very generic. The first time I read it I thought you will
patch the iconv support to remove IBM charset conversion code. But you
are removing VisualAge for Smalltalk code. In general VAST is still around.

The commit message should explain why it is not needed. E.g. did newer
versions of VAST remove these selectors, do we have no interest to import
code from VAST anymore.

I am not applying this patch right now as the commit message does not
provide any fact that help me to judge if it is useless code or not.


thanks
        holger

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

On 10/06/2013 08:10, Holger Hans Peter Freyther wrote:
I see no interest to have code that is not tested and nobody is going to
use.
If it was used everyday okay keep it but it's not the case and won't be.


_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

Il 06/06/2013 11:40, Gwenaël Casaccio ha scritto:
> Hi,
>
> that patch removes useless code.

Please revert this patch.  It is not enough to fix the problems that
Holger reported with the JIT, but it is wrong.

The point is not to optimize #ip and #sp.  It is to _prevent_ optimizing
it, and ensure that the cached values are committed by the interpreter
to the object before reading the variable.

Paolo

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk