[m-dev.] using set_tree234 as the default implementation of sets

[Paul Bone]

On Tue, Sep 29, 2009 at 03:35:04PM +1000, Zoltan Somogyi wrote:
> I tried this, and on tools/speedtest, it led to a 6% slowdown when the compiler
> was compiled in hlc.gc, and a 14% slowdown when the compiler was compiled in
> asm_fast.gc. These results tell me we should continue to use set_tree234
> in places where sets can become very big, but we should continue to use
> plain sets in places where they are very unlikely to become big, which in
> the compiler is most places.
> 

Thanks for this,  Thibaut and I had been wondering how these implementations
compared for typical uses of sets.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 489 bytes
Desc: Digital signature
URL: <http://lists.mercurylang.org/archives/developers/attachments/20090929/8edfee3f/attachment.sig>