On 2/12/21 6:03 AM, Ash Holsenback wrote:
> On 2/12/21 2:33 AM, Thomas de Groot wrote:
>> Op 11/02/2021 om 16:45 schreef Alain Martel:
>>> Le 2021-02-09 à 02:30, Thomas de Groot a écrit :
>>> When using the default method 3, when you have two values for
>>> samples, the second is silently ignored.
>> The wiki documentation then needs some updating here or a further
> already there: about 2/3 of the way down...
Maybe my old eyes, but I see on that page more or less a block of text
for method 3; then one for method 1 and last one for method 2.
For samples min, max in the block for method 1 I see:
"The samples Min, Max keyword specifies the minimum and maximum number
of samples taken per interval. The default values are samples 1,1. The
value for Max may be omitted, in which case the range Min = Max will be
Which for method 3 is:
- Wrong in suggesting the min and max are both used. Only min is used.
- Wrong in suggesting the default values for method 3 are 1,1. For
method 3 it is 10. (A little bell that the wrong values are listed
somewhere else too in the docs)
- Wrong in saying samples controls the min and max samples taken per
interval. In method three the number of samples ultimately taken (where
the adaptive method is actually working...) is controlled by
aa_threshold and aa_level. The aa_level really being the limit on the
recursive depth for the algorithm.
I think at a minimum Alain's sentence should be added to the block
related to method 3 at the top. Plus a note about the default samples
with method 3 for the single starting interval being the minimum 10.
Aside: As people talk about media, it is kicking loose some of the items
on my mental "media isn't quite right" list.
a) In multiple medias which I mentioned earlier I think the problem with
mixing media types had to do with intervals. The first/last setting gets
carried into the rest or similar because it gets set at parse time. It's
always wrong for some of the media - if you mix types.
b) Do NOT use jitter with method 3. Well, unless you just want
additional noise - and often slower media. It never fit the method 3
approach and there are bugs with the actual jitter implementation in
method 3. See, for example, https://github.com/POV-Ray/povray/issues/348.
Quite a bit of this post has been discussed previously elsewhere - a
good chunk of it at least was public (newsgroups / github).
Post a reply to this message