[Gc] GC_get_bytes_since_gc locks

Juan Jose Garcia-Ripoll juanjose.garciaripoll at googlemail.com
Wed Aug 24 12:50:30 PDT 2011


On Wed, Aug 24, 2011 at 9:04 PM, Ivan Maidanski <ivmai at mail.ru> wrote:

> So, you concern is purely about performance, right? Or, this is a part of
> the story?


Here is my list of concerns

- Concern #1 efficiency: getters are called typically in groups, as shown by
Guile; locking takes some time; locking imposes some problems w.r.t.
interrupts (need to disable them, slowing down the code)

- Concern #2 design consistency: either all or no getter should lock.

- Concern #3 backwards compatibility: if the original functions were unsafe
leave it like that.

- Concern #4 arbitrary changing the semantics: one thing that annoyed me
about this change was the suprising nature of it. I am subscribed to the
mailing list but can not read all emails; however, I believe that semantic
changing should result from some dialogue with users and some "heads up"
warning.

- Concern #5 behavior of function is hidden (relates #2 and #3): no way to
know which one locks without reading the header and knowing the version of
the library. Makes someone who reads the code unaware of what goes
underneath and the problems of that function. In my experience it is better
to make such things explicit in user code.

Juanjo

-- 
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
http://juanjose.garciaripoll.googlepages.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://napali.hpl.hp.com/pipermail/gc/attachments/20110824/8a0a9900/attachment.htm


More information about the Gc mailing list