Tim Peters wrote:
>
> [Jeremy Hylton]
> > I wanted to be sure that some other change to the dictionary code
> > didn't have the unintended consequence of slowing down insertdict.
>
> Have you looked at insertdict? Again, nothing has changed in it since 2.0,
> and it's a simple little function anyway. Here it is in its entirety:
>
> static void
> insertdict(register dictobject *mp, PyObject *key, long hash, PyObject
> *value)
> {
> PyObject *old_value;
> register dictentry *ep;
> ep = (mp->ma_lookup)(mp, key, hash);
> if (ep->me_value != NULL) {
> old_value = ep->me_value;
> ep->me_value = value;
> Py_DECREF(old_value); /* which **CAN** re-enter */
> Py_DECREF(key);
> }
> else {
> if (ep->me_key == NULL)
> mp->ma_fill++;
> else
> Py_DECREF(ep->me_key);
> ep->me_key = key;
> ep->me_hash = hash;
> ep->me_value = value;
> mp->ma_used++;
> }
> }
>
> There's not even a loop. Unless Py_DECREF got a lot slower, there's nothing
> at all time-consuming in inserdict proper.
>
> > There is a real and measurable slowdown in MAL's DictCreation
> > microbenchmark, which needs to be explained somehow. insertdict
> > sounds more plausible than many other explanations.
>
> Given the code above, and that it hasn't changed at all, do you still think
> it's plausible? At this point I can only repeat my last msg: perhaps your
> profiler is mistakenly charging the time for this line:
>
> ep = (mp->ma_lookup)(mp, key, hash);
>
> to insertdict; perhaps the profiler is plain buggy; perhaps you didn't
> measure what you think you did; perhaps it's an accidental I-cache conflict;
> all *I* can be sure of is that it's not due to any change in insertdict
> <wink>.
It doesn't have anything to do with icache conflicts or
other esoteric magic at dye-level. The reason for the slowdown
is that the benchmark uses integers as keys and these have to
go through the whole rich compare machinery to find their way into
the dictionary.
Please see my other post on the subject -- I think we need
an optimized API especially for checking for equality.
--
Marc-Andre Lemburg
______________________________________________________________________
Company: http://www.egenix.com/
Consulting: http://www.lemburg.com/
Python Pages: http://www.lemburg.com/python/
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4