A new v2 beta (version Jan-19-2009) is available for download here: http://moi3d.com/beta.htm

You will need to enter your v1 license key to download the v2 beta.

The v2 beta installs into its own folder, it is not a problem to have v1 and any v2 beta installed on the same machine, in fact I recommend that so that you can fall back to v1 if you run into some temporary bad bug in v2.


New stuff for this version:


Major overhaul of mesh exporting to greatly increase mesh generation speed and to make use of multiple CPU cores.

Speed improvement for dynamic feedback during transforms (for example dragging or mirroring) of heavier and denser objects.

A variety of other smaller tune-ups, I'll post a completed list tomorrow.


EDIT: list of other new things posted here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=2333.22


- Michael

It would be great to get some timing benchmarks of the new meshing speed!

If you have a heavier model that previously took quite a while to save to a mesh, if you could take 3 different benchmarks that would help to gather some information.

One measurement on version 1.0, one measurement with the new Jan-19 v2 beta, and if you have multi-core then one measurement with a setting in moi.ini edited to limit to it to only 1 CPU core would be helpful.

That final one will help me to see how well it is scaling across CPU cores on some different people's machines.

The moi.ini setting for limiting it to only 1 CPU core is:

[Mesh Export]
ThreadLimit=

Set that to ThreadLimit=1 to limit it to 1 CPU core

Close down MoI first before editing moi.ini and the moi.ini file can be found in WinXP here: C:\Documents and Settings\Michael\Application Data\Moi\moi.ini


Then after you're done set it back to a blank thread limit again to get back to full performance.


Some other notes - previously the meshing dialog would be frozen when it first appeared and you could only adjust settings after the initial mesh was finished being calculated. This is no longer the case, you can adjust settings and the current mesh job will automatically be canceled and a new job started with your new settings. Also I've removed the "Update mesh" button, instead now any changes you make will be applied automatically.

There is a new progress bar for the meshing job in progress, which is shown in the upper-right area of the main window underneath where the points and polygon count readouts are at.

- Michael

Attached here is a script copier utility for the new v2 beta.

If you run the attached ScriptCopier_Jan_19_2009.exe program, it will copy any custom commands that you had installed into MoI version 1.0 over to the new v2 beta's installation folder for you.

- Michael

HI Michael,

Did a little test and the result is the new MoI Jan-19-2009 is many times faster compare to previous beta version. I dont know exactly how many times. Any idea the best way to measure the time ?

I have test on meshing and mirror a big object. Attached is the file. You can use this file as benchmark test.

In SWX, we can use "feature statistic" so its very easy to compare the exact different time.

Good Job Michael :)

Hi Anis, you have to just use a stopwatch for gathering a measurement right now, there isn't any built in method.

Thanks for posting your robot, that is a really fun model by the way!

Over here on my test quad-core machine (Intel Q6600 running Vista), I get these times for your model at default settings:

v1:	25.03 seconds
v2 Jan-19:	1.25 seconds

So that's a factor of 20 times difference ( 25.03 / 1.25 ) not too bad huh? :)

Then with limiting to 1 thread, I get 3.62 seconds. This last measurement is primarily of interest just to me to understand if there is any problems with the scaling to multiple CPUs in particular situations like on particular models or on a particular kind of machine configuration.

It is possible for the multi-core stuff to be sensitive to a variety of factors so it may be helpful for me to collect some data on what speed increases various other people are seeing on their own machines and with their own complex models. If it is a lot different than what I see here there may be some things for me to tune up, but hopefully not.

- Michael

OMG! :o Thanks Michael! brilliant stuff!

Meshing Robby's head default mesh, 2 cores @ 3.2 G.

V1: I had to stop at 15 min because I couldn't wait any longer, 1 core of course.

V2.0 beta Nov-21-2008: 1 Min 46 sec, 1 core.

V2.0 beta Jan-19-2009: 10 sec, 2 cores :)

V2.0 beta Jan-19-2009: 19 sec, 1 core :)

Oh yeah, I like the addition of the progress bar and poly count as it's calculating.




Cheers
~Danny~

Hi All...

My Machine :

Processor: Intel(R) Core(TM)2 Duo CPU T7300 @ 2.00GHz (2 CPUs)
Memory: 3072MB RAM
Page File: 897MB used, 4059MB available

Card name: NVIDIA Quadro FX 1600M
Display Memory: 512.0 MB

Here is my test result,

Open anis`s robot :

V1 = 3.97
V2 Jan 1 Core = 3.2
V2 Jan = 3.01

Meshing anis`s robot :

V1 = 33.48
V2 Jan 1 Core = 4.78
V2 Jan = 4.53

Seems you have made a mistake first line of your post ;)
< A new v2 beta (version Nov-21-2008)

Holy crapsicles that is awesome..:O)

Great job on the meshing, I will run the hydraulic connector when I get home in all three tests on my dual quad. I ran a smaller component(46000polys) on my work clunker and v.1 time was 1:24 and v.2beta was 8 seconds. Great job Michael.

Fantastic speed improvement!!

I just wish that it placed triangles better in quad+tri mode now... Right now it makes a mesh that is very hard to UV unwrap.

Michael,

As a test case, I meshed a 60MB IGES file into an 88MB OBJ file. The last MoI beta took 26.7 minutes while the new beta took only 6.5 minutes on an old dual Xeon processor system.

I'm impressed so keep up the good work!

Al Greynolds
www.ruda.com

Xeon 4cores

Robot - Angle 4 - Divide larger than 2 - Avoid smaller 0.1 - Ngones 1117863 Ngones - 1855461 Points


13DEC : 2min27Sec
19JAN : 1min14sec (1 thread)
19JAN : 30Sec

 

 

Fighter - Angle 4 - Quads&Tri's 3713420 Polys - 3670177 Points


13DEC : 2min45Sec
19JAN : 49Sec (1 thread)
19JAN : 16Sec

 

Takes moslty less time to generate the mesh than save it on my HD :o)

The only limit I see now it's the memory, we need a MoI64 !!!

Hi LOTRJ,

> I just wish that it placed triangles better in quad+tri mode
> now... Right now it makes a mesh that is very hard to
> UV unwrap.

I don't know if it will help any or not, but there is one setting in the moi.ini file which tweaks how triangulation is done:

[Mesh Export]
CentroidTriangulation=y

With the default of =y then it will try to produce a kind of radial triangulation connecting points on an n-gon outline to a centroid point if possible. With it set to =n then the triangulation will only be connecting points on the outer n-gon outline to one another without adding in a new center point.


If you have an example of an existing shape that is not getting triangulated how you would like, could you please post the .3dm file of it, and maybe show how you wish it was triangulated instead? That would help me to think about how to tune it up further.


One thing that I have thought about before is trying to carve up a larger outline into some sub chunks to try and do centroid style triangulation on some sub pieces of a single big outline.


By the way the meshing bug you reported a few days ago here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=2323.1
should be fixed now with this new version, if you still have that original file around can you please verify that case is working now?

- Michael

Here's my times for meshing this:




Time are minutes:seconds:

V1 = 2:38
V2 Nov-21 = 1:54
V2 Jan-19 1 core = 0:19
V2 Jan-19 2 core = 0:10

Processor Core 2 Duo E6600 @ 2.4 GHz

That's a massive improvement. Great stuff Michael!

Plus the meshing problem I had is now gone too.

Hi Tony,

> That's a massive improvement. Great stuff Michael!
>

Thanks for testing it!

This level of speed increase should help to make it more convenient when working on higher detailed models I hope.


> Plus the meshing problem I had is now gone too.

That's great! Wow - not only so much faster but also better quality too - normally I am supposed to make you pick only one or the other of those, not both! :)

- Michael

OK here is my test with my home dual quad xeon using the default mesh setting for my connector http://moi3d.com/forum/messages.php?webtag=MOI&msg=2284.1

V.1 15:17
v.2 single core 1:34
v.2 eight cores :23

441,983 polys

Great job michael. I did get one out of memory indicator that shut down the meshing the first time but I closed MOI and re-opened the file and have not been able to replicate it again.

Hi Grendel, thanks for testing it!

Hmm - from V1 15 minutes to V2 23 seconds is a factor of nearly 40! That's certainly a record so far.

I tend to get something around 20 to 25x on my quad core over here, so your measurement is fairly consistent with around double that, but your measurement with the new engine limited to a single core seems to show that there may be some further tuning possible. I'm not very sure that I will be able to do that right now though, further tuning beyond this may be fairly hard work. One thing that is very good though is that it does not choke with 8 cores running, it is actually possible with multi-threaded code to get cases where increasing the number of cores actually makes things worse instead of speeding things up, if there is a problem of too much "contention" between the cores.

Re: running out of memory - yeah since now you can more easily spew out a huge number of polygons there is also an improvement to handle out of memory conditions better, instead of just crashing or something it will show that warning when it ran out of memory when trying to complete the current task.

I guess that probably happened when you got most or all of the 8 cores happening on crunching away on those long helix surfaces, those will kind of spike to a higher level of memory consumption during their subdivision process, so getting a bunch of those all going at once will put a pretty big spike in memory consumption. It will bump back down after each one is finished and the final mesh is extracted but if you bumped up beyond your full system memory during the spike, that will be a problem.

That's kind of an interesting problem with a higher number of cores that I hadn't really considered too much.

I think you mentioned that you are running XP 32 bit? Are you already running with the /3GB switch in boot.ini? (http://www.microsoft.com/whdc/system/platform/server/PAE/PAEmem.mspx)

If not then this may be something you'll need to use to increase the available memory for MoI to allocate when you have some longer surfaces in your model plus a high number of cores.

Anyway, it is pretty cool to go from something that took 15 minutes to less than 30 seconds on the same machine! :)

- Michael

Uncool in the discussion?

Brian REMEMBERED how to add the plugin/shortcuts to get them to work!

Well---you have to have some slow learners!

Brian

WOW!

I'll try to go back and run some "actual" benchmarks, but seeing the same results here (multiple times faster meshing -- opening of files seems a little snappier too (might just be the excitement))!!!

Excellent work MG!!!

Hi AW,

> opening of files seems a little snappier too (might just be
> the excitement))!!!

I think that opening of files should probably be the same... Probably just some feeling of momentum carrying over from the other speedups! :)

But now that I have a thread-safe meshing mechanism it will open up the door for me to do some things that will speed up file opening for real though.

I should be able to get the display mesher to run on multiple cores now - currently the display mesher (which is used during file open) does not use multiple cores, only the file export mesher is doing that now.

Then the other thing that I've been thinking of doing is to do an initial pretty rough display mesh (or even skip doing it and have only edge wires initially if it takes more than a couple of seconds), and then do a higher density display mesh calculation in the background. The higher density display meshes would kind of spool in while you are able to do other work on the model.

Those are some ideas anyway! Actually I don't think these will be all that difficult, the difficult part of these was to make the meshing engine thread safe which is all done now. But kind of a good first stage is to get that tested here first with the export meshing in this new release.

- Michael