 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Op 24-9-2021 om 00:03 schreef Samuel B.:
> Thomas de Groot <tho### [at] degroot org> wrote:
>> Op 22/09/2021 om 02:47 schreef Samuel B.:
>>> Hey, a fellow i5 owner :) I've got a 6500 here. It's a really good chip.
>>>
>> <grin> I have an i5 8250 and an i7 8750 here. For some arcane and
>> absolutely trivial reasons, I am using the i5 more than the i7, but
>> there it is. :-)
>
> Based on user benchmarks, the i5-6500 seems to perform a bit better than the
> i5-8250(U), but the i7-8750(H) outperforms both. (The number of cores, however,
> do factor into all this.)
>
Indeed. the i5 has 4 cores / 8 logical processors (hence my +wt7) and
the i7 has 6 cores / 12 logical processors, where I generally use +wt10
or +wt11.
> From what I've seen, CPU performance has been rather incremental over the years
> when it comes to the hard stuff like raytracing... Certain things will always be
> slow to parse and/or render, which is why I try to find faster ways of doing
> things when possible...
>
I sometimes take up again old scenes which I found almost impossible to
render within a "acceptable"(?) time back then, and which now render
pretty fast.
>>> Ten minutes of parsing seems a little steep.(...) Thirty minutes of
>>> render time doesn't seem as terrible as it could be
>>>
>> Not sure as I have not been monitoring closely what is going on, I guess
>> it is the building of the individual mesh2 hairs which sums up. In a
>> next run , I shall save/read the hair meshes and that goes faster in the
>> end imo.
>
> What's the shape of the hair follicles? If they are round, how many sides do
> they have?
>
Using Robert's settings here, without change:
#declare Upoints = 50; // Number of points along
#declare Vpoints = 5; // Number of points around
> And if you're using includes, then perhaps they are the reason for your high
> parse times. So if it's at all possible, maybe try copying all the necessary
> macros &/ functions into your scene file, since using includes in any heavy way
> is much slower than referencing local code. (But I think you knew this already.)
>
Yes, I generally try to put all relevant, repeating, macros and
functions in the scene file. I probably could gain more parse time by
picking out Sweepspline.inc and more especially meshmaker.inc, but I
didn't do that (laziness).
>> But, not trivial, I had the laptop battery in "best battery
>> life" mode; "best performance" mode is certainly faster indeed.
>
> I've got a 2012 laptop, and I've been using it without its battery for a number
> of years now. (The battery's fine, but I like to extend its life by not using it
> ;D) The laptop just stays plugged in with high performance mode enabled at all
> times. Enabling best/high performance really does make a difference.
>
Yes, I increasingly do that too, especially when doing heavy duty renders.
> An unfortunate thing happened recently... I've got the laptop perched on two
> small planks (for air flow) on top of my desktop computer. It's up high so that
> the laptop can get a better signal. Well, one night I was a little careless and
> didn't notice the planks were slipping... Anyway, the laptop fell off and I put
> it back thinking nothing of it. But the next day I realized the earbuds jack was
> askew... Turns out the port got bent when everything fell over, and now it only
> transmits a signal through one channel D: And I can't observe the contacts
> without taking the whole thing apart (I've opened the laptop plenty of times,
> but never removed /everything/). My only other option is cutting a hole through
> the plastic over the problematic area. but at least the thing is still
> functional in all other respects :P
>
Yeah... one of those "stupid" accidents. I got my touchpad blocked
(something heavy fell on it - not my hammer - and since then it is
unoperable) but as I much prefer a good mouse I just leave it at that.
>>> Did I ever post my experiments with hair? I was using an .obj-to-.pov converter
>>> and trying to grow hair from a mesh. The way I had it set up was if a triangle
>>> was too small it only had a small chance to grow a hair, otherwise it would try
>>> to grow a certain number of hairs for a given triangle's area. (...)
>>>
>> I don't remember, but that looks interesting, especially the approach
>> using triangle sizes. I had not thought of that aspect. I simply used a
>> simple trace() function on the skullcap from randomly generated points
>> on an external sphere. Pretty fast by itself.
>
> The code for the hair still exists somewhere, but it is not readily available. I
> do, however, have at hand the program I wrote to convert .obj files to
> POV-readable data. (I may have posted it once; I'll have to check.) The program
> exports arrays, with options for additional data such as neighboring faces and
> vertices. (If you remember some preliminary stuff I posted a while back for an
> IRTC competition, that's what I was using to make flagstones/stone walls out of
> blobs.)
>
I didn't check yet, but I may have that somewhere in my database, filed
under "Sam Benge" :-)
>>> P.S. Thanks for reminding me of Wallace and Gromit. It's a great series. Just
>>> watched A Close Shave. What a gem of a flick :D
>>>
>> Oh yes, I love them. I was reminded of the scene showing Wallace tasting
>> a piece of Moon (on toast) and musing: "Wensleydale?"
>
> Lol. Based on how often he mentions that particular cheese, I'd say it's his
> favorite :)
>
> After you originally mentioned W&G I got to thinking... There might have a way
> to make a convincing clay material, complete with fingerprints. It involves the
> use of slope patterns and projected normals...
>
Now, that might be interesting indeed. Based on blobs I presume?
--
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Il 21/09/2021 14:24, Thomas de Groot ha scritto:
> Following an old discussion in 2014:
>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.52dff97746aac96b7a3e03fe0@news.povray.org%3E/?mtop=391452&moff=21
>
>
> I took up Robert McGregor's code to see how it could be tweaked to
> obtain a nice haircut. 10000 hairs were planted here. To give a rough
> approximation: parsing took about 10 minutes and complete (stochastic)
> render about half an hour with an i5 machine (Win10; Pov version 3.8).
> The small highlights on the hairs resulted from a combination
> normal/finish choices.
>
He is definitely Der Struwwelpeter
(https://de.wikipedia.org/wiki/Struwwelpeter)!
A nice job,
Paolo
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thomas de Groot <tho### [at] degrootorg> wrote:
> Following an old discussion in 2014:
>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.52dff97746aac96b7a3e03fe0@news.povray.org%3E/?mtop=391452
&moff=21
>
> I took up Robert McGregor's code to see how it could be tweaked to
> obtain a nice haircut.
I love this-- especially with all the hair shadows falling over the face. No
fakery there!
I'm curious: In the news thread you mentioned (or maybe an offshoot of it), one
of the ideas was to color each hair along its individual curl, like uv-mapping,
using some clever code. I can't really discern that effect from your posted
render, but did you do that as well?
> Based on user benchmarks, the i5-6500 seems to perform a bit better than the
> i5-8250(U), but the i7-8750(H) outperforms both. (The number of cores, however,
> do factor into all this.)
>
Indeed. the i5 has 4 cores / 8 logical processors (hence my +wt7) and
the i7 has 6 cores / 12 logical processors, where I generally use +wt10
or +wt11.
Naw-- an i7 with 8 cores/16 threads is the way to go :-P (My previous machine
was a Win7 box with 'only' 2 cores! How did I ever manage with that?!)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Op 25/09/2021 om 02:07 schreef Kenneth:
> Thomas de Groot <tho### [at] degrootorg> wrote:
>> Following an old discussion in 2014:
>>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.52dff97746aac96b7a3e03fe0@news.povray.org%3E/?mtop=391452
> &moff=21
>>
>> I took up Robert McGregor's code to see how it could be tweaked to
>> obtain a nice haircut.
>
> I love this-- especially with all the hair shadows falling over the face. No
> fakery there!
>
> I'm curious: In the news thread you mentioned (or maybe an offshoot of it), one
> of the ideas was to color each hair along its individual curl, like uv-mapping,
> using some clever code. I can't really discern that effect from your posted
> render, but did you do that as well?
>
Hmm, good question. Each mesh2 hair is uv-mapped by meshmaker when
created, and the pigment is uv-mapped on the mesh2... so, that seems to
be happening indeed. I shall come back about this later as I do not have
the code at hand at the moment. What I can say though: meshmaker is your
friend! Last year, I used it for converting parametric grass blades (Uwe
Gleiss: Lawnmaker) to mesh2 blades.
>> Based on user benchmarks, the i5-6500 seems to perform a bit better than the
>> i5-8250(U), but the i7-8750(H) outperforms both. (The number of cores, however,
>> do factor into all this.)
>>
> Indeed. the i5 has 4 cores / 8 logical processors (hence my +wt7) and
> the i7 has 6 cores / 12 logical processors, where I generally use +wt10
> or +wt11.
>
> Naw-- an i7 with 8 cores/16 threads is the way to go :-P (My previous machine
> was a Win7 box with 'only' 2 cores! How did I ever manage with that?!)
>
Waaah! [throws machine in the dustbin]
I know. Not so long ago, I also had to manage with two cores, and there
was a time with only one... We are spoiled. ;-)
--
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Op 25-9-2021 om 09:04 schreef Thomas de Groot:
> Op 25/09/2021 om 02:07 schreef Kenneth:
>> Thomas de Groot <tho### [at] degrootorg> wrote:
>>> Following an old discussion in 2014:
>>>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.52dff97746aac96b7a3e03fe0@news.povray.org%3E/?mtop=391452
>>>
>> &moff=21
>>>
>>> I took up Robert McGregor's code to see how it could be tweaked to
>>> obtain a nice haircut.
>>
>> I love this-- especially with all the hair shadows falling over the
>> face. No
>> fakery there!
>>
>> I'm curious: In the news thread you mentioned (or maybe an offshoot of
>> it), one
>> of the ideas was to color each hair along its individual curl, like
>> uv-mapping,
>> using some clever code. I can't really discern that effect from your
>> posted
>> render, but did you do that as well?
>>
> Hmm, good question. Each mesh2 hair is uv-mapped by meshmaker when
> created, and the pigment is uv-mapped on the mesh2... so, that seems to
> be happening indeed. I shall come back about this later as I do not have
> the code at hand at the moment. What I can say though: meshmaker is your
> friend! Last year, I used it for converting parametric grass blades (Uwe
> Gleiss: Lawnmaker) to mesh2 blades.
>
Checked. Yes the individual mesh2 hairs are uv-mapped by SweepSpline.inc
(Mike Williams) which uses makemesh.inc (Ingo Janssen) to build and
write the files. All this is present in the original scene file by
Robert McGregor. I did not change anything in there, except for the hair
pigment.
Later in the discussion, I suggested to increase render speed by
texturing the whole union of meshes instead of the individual hairs, but
as I added, that would result in a far less interesting hairdo.
--
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Op 25-9-2021 om 00:18 schreef Paolo Gibellini:
> Il 21/09/2021 14:24, Thomas de Groot ha scritto:
> > Following an old discussion in 2014:
> >
>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.52dff97746aac96b7a3e03fe0@news.povray.org%3E/?mtop=391452&moff=21
>
> >
> >
> > I took up Robert McGregor's code to see how it could be tweaked to
> > obtain a nice haircut. 10000 hairs were planted here. To give a rough
> > approximation: parsing took about 10 minutes and complete (stochastic)
> > render about half an hour with an i5 machine (Win10; Pov version 3.8).
> > The small highlights on the hairs resulted from a combination
> > normal/finish choices.
> >
>
> He is definitely Der Struwwelpeter
> (https://de.wikipedia.org/wiki/Struwwelpeter)!
>
> A nice job,
> Paolo
Yes Paolo, you are absolutely right: Der Struwwelpeter it is! :-)
--
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thomas de Groot <tho### [at] degrootorg> wrote:
> Op 24-9-2021 om 00:03 schreef Samuel B.:
> > I do, however, have at hand the program I wrote to convert .obj files to
> > POV-readable data.
> >
> I didn't check yet, but I may have that somewhere in my database, filed
> under "Sam Benge" :-)
Well, I didn't manage to find the program on the POV-Ray server when looking
through the usual groups, so I don't think I ever released it. But when looking
at the binary yesterday on my hard drive I decided to do a couple virus scans...
All was good on the Windows Defender front, but when I scanned it on VirusTotal,
one of its engines flagged it as malware :S Some kind of worm that uses shell
commands to work its malfeasance. I really don't believe it's infected, I just
think that the MaxSecure engine sometimes likes to throw false positives. But
it's enough to make me not want to publish it. After all, I did use a 3rd party
..obj-reading library. But what I think really happened is MaxSecure saw the
program's ability to take command line arguments and decided it was up to no
good. So, unless everyone's willing to take a risk, I'll be withholding it until
I can do a recompile using updated libraries. And if /that/ doesn't appease
MaxSecure, I'll try making the program read from a .cfg file instead of taking
command line arguments :|
> > After you originally mentioned W&G I got to thinking... There might have a way
> > to make a convincing clay material, complete with fingerprints. It involves the
> > use of slope patterns and projected normals...
> >
> Now, that might be interesting indeed. Based on blobs I presume?
Blobs, meshes, whatever. The material should work with anything.
Sam
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Op 26/09/2021 om 00:00 schreef Samuel B.:
> Thomas de Groot <tho### [at] degrootorg> wrote:
>> Op 24-9-2021 om 00:03 schreef Samuel B.:
>>> I do, however, have at hand the program I wrote to convert .obj files to
>>> POV-readable data.
>>>
>> I didn't check yet, but I may have that somewhere in my database, filed
>> under "Sam Benge" :-)
>
> Well, I didn't manage to find the program on the POV-Ray server when looking
> through the usual groups, so I don't think I ever released it. But when looking
> at the binary yesterday on my hard drive I decided to do a couple virus scans...
> All was good on the Windows Defender front, but when I scanned it on VirusTotal,
> one of its engines flagged it as malware :S Some kind of worm that uses shell
> commands to work its malfeasance. I really don't believe it's infected, I just
> think that the MaxSecure engine sometimes likes to throw false positives. But
> it's enough to make me not want to publish it. After all, I did use a 3rd party
> ..obj-reading library. But what I think really happened is MaxSecure saw the
> program's ability to take command line arguments and decided it was up to no
> good. So, unless everyone's willing to take a risk, I'll be withholding it until
> I can do a recompile using updated libraries. And if /that/ doesn't appease
> MaxSecure, I'll try making the program read from a .cfg file instead of taking
> command line arguments :|
>
You may want to have a second opinion on this. Eset has a one time,
free, online scanner: https://www.eset.com/uk/home/online-scanner/
I have been using Eset (one of the paid versions) for the last ten years
or so with complete satisfaction, so no tricks there.
>>> After you originally mentioned W&G I got to thinking... There might have a way
>>> to make a convincing clay material, complete with fingerprints. It involves the
>>> use of slope patterns and projected normals...
>>>
>> Now, that might be interesting indeed. Based on blobs I presume?
>
> Blobs, meshes, whatever. The material should work with anything.
>
Ah, ok, a /material/. Somehow I thought the physical stuff. :-) However,
anyway, that would be highly interesting indeed.
--
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thomas de Groot <tho### [at] degrootorg> wrote:
> Op 26/09/2021 om 00:00 schreef Samuel B.:
> > But when looking at the binary yesterday on my hard drive I decided to
> > do a couple virus scans... All was good on the Windows Defender front,
> > but when I scanned it on VirusTotal, one of its engines flagged it as
> > malware :S
>
> You may want to have a second opinion on this.
I might try downloading the AV you recommended, but it's just another point of
reference, and it may miss something.
The thing I like about VirusTotal is that it's not just a /second/ opinion, it's
a 3rd, 4th, etc. It gathers results from many different virus scanners. But I
really wonder about some of its results. Windows Defender never seems to fail
me, though. (Key word: 'seems.')
I was able to find the original source for the program, so I can now recompile
it. But the program I used to compile it originally was Code::Blocks. I just
downloaded the new version, but when I scanned its installer, VirusTotal
reported it as potentially containing malware! Gah. There was only one engine
that engine detected trouble (not MaxSecure this time), but how can I ever be
sure of anything at all? Argh :( Maybe VirusTotal isn't too careful when it
comes to choosing its virus-detection engines?
> > Blobs, meshes, whatever. The material should work with anything.
>
> Ah, ok, a /material/. Somehow I thought the physical stuff. :-) However,
> anyway, that would be highly interesting indeed.
Ya, I think it will look good but I need to make a suitable showcase model
before implementing the material.
Sam
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Naw-- an i7 with 8 cores/16 threads is the way to go :-P (My previous machine
> was a Win7 box with 'only' 2 cores! How did I ever manage with that?!)
>
>
A whole two cores ? WOW !!!
My previous one was a P4 pre-hyperthread, agp4 video and a whole 3Gb of
RAM running Win7.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
