Summary of containers patches

[Ian Lynagh]

On Thu, Sep 23, 2010 at 09:12:21PM -0400, Edward Kmett wrote:
> If I am reading this right, the numbers for alter/delete/update are rather unsettling in your proposal.
> 
> Is there a compromise point that does not cause those to suffer so much?

It's really late to be doing research for the GHC 7.0 release. We're
just trying to get the builds to go through so we can get the RC out.

Simon Marlow wrote:
> So what to do for GHC 7.0.  The options are:
> 
>  - make the INLINABLE changes, re-do the measurements to make sure
>    performance hasn't got worse, and push the patches.
> 
>  - roll back the first set of patches.

Milan Straka wrote:
> So, the INLINABLE does not help at all.

The safest option is probably to roll back. There weren't any INLINE
pragmas in the containers that came with 6.12, so I believe this is
unlikely to be a regression, even with the changes to the inliner.

I don't think we need to roll back the other parts of the patches,
though; just remove the pragmas.

Then we can do the research and make a less rushed decision for the HEAD
(and perhaps 7.0.2).


Thanks
Ian