Hi Watson,
Sorry for late response. I think your benchmark code contains too much side
effects (such as the allocation of 'a' string and hash internals during
rehash). If you just benchmark h[42]=42 you should see a difference.
Anyways, I corrected the change to use the mask flag in r5187.
Lau
Hi Laurent,
There is only the extremely slight difference.
I think that I can ignore this.
How do you think?
I attached a patch and benchmark script.
** Result of r5185
$ macruby bm_hash.rb
Rehearsal
13.64 0.21 13.85 ( 9.114591)
---
Hi Watson,
This isn't good, rhash_modify() must be very fast, so calling OBJ_FROZEN and
OBJ_UNTRUSTED is not good there.
Can we look up the mask flag as before?
Laurent
On Jan 21, 2011, at 7:51 AM, source_chan...@macosforge.org wrote:
> Revision
> 5186
> Author
> watson1...@gmail.com
> Date
>
