Inconsistent Mesh Uploads - Aditi and Agni

[Chic Aeon]

ANSWER:

Well you may have guessed that it was the cube physics and Firestorm although I have made a lot of stairs lately with no issues and now OMG land impacts.

I used planes physics(thanks to Aquila) and the stairs came in at 2 on both Aditi and Agni. So evidently the Havok code is not the same on Agni as Aditi . And while things LOOK like they will be fine, once changed to prims the cost is way overblown. I am going to check the cube physics using the Linden viewer and see what I get, but it will be VERY good when Firestorm has the new code included (maybe September). IF things hadn't of worked so well on Aditi, I would have figured it out sooner. Not all that great :D.  My build just got 18 land impact lighter ^^. This is good. 

 

CUBE PHYSICS uploaded with the Linden Viewer came in at 3 land impact and cost $14 to upload ON THE MAIN GRID --

2 land impact and $11 on ADITI. So STILL inconsistant

 

I have been noticing this for a long while now and mentioned it a time or two (or at least meant to) but it is easy enough to convince yourself that you simply did something wrong. We all make errors.  But today, this is a BIG difference and I  repeated the process and double checked my outcomes. 

The beta grid is still good for testing physics and textures and mapping and such but when it comes to costs (both upload AND land impact) the numbers aren't always matching. This doesn't seem to happen on anything small or perhaps it is the "use a cube for the physics idea" that is the difference. Here's what I got today.

Lani over on the beta grid testing stairs. The stairs uploaded with the default LOD settings (I checked and the default LOD settings are the same on both grids) and the exact same physics model (cube physics not plane and the physic is great on both grids). The Aditi version came in at 5 after being changed to PRIM status. It cost $14 to upload.

Over on the main grid, the same stairs file with the same default settings and the same physics model came in at 11 land impact and cost $22 to upload (twice - LOL). 

Am I missing something here?   

I am not so concerned about the extra 7 lindens but the land impact is over TWICE on the main grid version. 

 

ADDING INFO: 

I went into the MORE INFO tabs in the build menu and here are the results.

Unlinking the builds shows that it is the STAIRS PHYSICS on the main grid that is uping the land impact. Even the platform and railing are different though. 2.1 in Aditi and 2.9 in Agni. 

 

 

Edited August 4, 2017 by Chic Aeon

[Whirly Fizzle]

Your image shows the Aditi mesh is set to physics shape type None, not Prim.

[Beq Janus]

6 minutes ago, Whirly Fizzle said:

Your image shows the Aditi mesh is set to physics shape type None, not Prim.

As Whirly notes, based on the screen shots the Aditi build will be charged only for the root mesh physics, as the root cannot be None. While on Agni you get charged for both parts. Also, while it ultimately makes no difference to the result if physics is enabled for all parts, the switching of the root prim to be the deck on Agni, while it was the steps on Aditi would give you different results if you had both sets set to None (Agni would carry the Deck Physics cost while Aditi would carry the steps physics cost).

Regards

Beq

[Chic Aeon]

3 hours ago, Whirly Fizzle said:

Your image shows the Aditi mesh is set to physics shape type None, not Prim.

Thanks. That wasn't it though. I did this  MANY times. But a good catch. That seems kind odd as she walked up the steps. But it WAS an odd day. I wrote "the answer" up top. 

 

[Chic Aeon]

3 hours ago, Beq Janus said:

As Whirly notes, based on the screen shots the Aditi build will be charged only for the root mesh physics, as the root cannot be None. While on Agni you get charged for both parts. Also, while it ultimately makes no difference to the result if physics is enabled for all parts, the switching of the root prim to be the deck on Agni, while it was the steps on Aditi would give you different results if you had both sets set to None (Agni would carry the Deck Physics cost while Aditi would carry the steps physics cost).

Regards

Beq

Thanks. My error but that wasn't the problem. Many uploads with MANY set to PRIM (correctly).  

If I uploaded with the Linden viewer (ANALYZED cube physics) I "still" got inconsistencies which were slight but it certainly seems like things should be the SAME on both grids. But I can see that new codes may get entered and things just aren't the same. 

In the end I used planes physics as it came in at "2" and I think $14 linden :D. The cube physics uploaded with the Linden viewer came in at 3. Happily this mess will hopefully be over in a couple of months. 

What puzzled me the most was that there actually WERE walkable mesh (very nice walkable mesh of both floor and stairs -- tested, tested, tested -- that came in on Aditi as 5 for the linkset.  

I got that AGAIN on Agni using planes. Cubes would have added only one more USING THE LINDEN VIEWER.  IF things would have be truly high on Aditi then I would have been alerted about the need for Linden viewer or Planes. Oddly though the physics worked just fine even though the land impact was WAY TOO HIGH. I guess that is a bit of a plus. I haven't seen that before though. 

 

Sorry for pasting a confusing screenshot though; that was SO not helpful :D. 

 

Edited August 5, 2017 by Chic Aeon

[arton Rotaru]

I can assure you, the Havok code is exactly the same on Aditi, and on Agni. It hasn't changed for years, on neither grid. These things you are encountering are to 99.9 % user error, or expected behavior due to the nondeterministic design of the importer.

To me the best feature of the viewer is the use of the .slm file. Once I'm happy with the models on Aditi, I just load the high LOD, and everything else is set by the .slm file exactly like it was on Aditi. The only difference in land impact between the 2 imports is in the decimals, probably due to minor float rounding errors. But my .slm Agni imports were always slighty less than the original 1st import. So I wouldn't complain about that either.

[Chic Aeon]

Yes, I remember the slm file :D. Thanks. I did these experiments probably three or four times on each grid and got the same "inconsistent" results with the the very same files and uploading methods (full full default default) on the LOD scales. So I really can't say what was going on but this isn't the first time for me.

I DO believe in user error. Just difficult to believe it could be the same error six or eight times in a row LOL. So I am sticking to my story much like I stuck to the "we really ARE losing inventory" long long ago and have just now be vindicated :D.

Whatever, the Planes version is up and working well. The model that was 12 LI  in Firestorm uploaded with the same settings at 2 in the Linden Viewer. That's pretty much all I know. 

 

 

[ChinRey]

2 hours ago, Chic Aeon said:

If I uploaded with the Linden viewer (ANALYZED cube physics) I "still" got inconsistencies which were slight but it certainly seems like things should be the SAME on both grids. But I can see that new codes may get entered and things just aren't the same.

From context I get it that it's the physics weight that is causing the issue.

Unless there are some complex hulls with more than six faces, analyzed physics will always weigh in at 0.360 per hull so if there are any inconsistencies, it has to be the number of hulls that changes. Check how many of them there are and also use the preview spread to see if they have the shapes, sizes and positions you want them to have. Unfortunatley the uploader does sometimes split and/or merge parts of the physics model even when they seem to be clearly defined separate cubes and I suppose that's what happens here. One trick that often fixes it is to do a % simplification with a fairly high value (typically around 80%) That trick is a bit unpredictable but you can always use the preview spread to check if it worked before you upload.

Edited August 5, 2017 by ChinRey

[Chic Aeon]

14 minutes ago, ChinRey said:

From context I get it that it's the physics weight that is causing the issue.

Unless there are some complex hulls with more than six faces, analyzed physics will always weigh in at 0.360 per hull so if there are any inconsistencies, it has to be the number of hulls that changes. Check how many of them there are and also use the preview spread to see if they have the shapes, sizes and positions you want them to have. Unfortunatley the uploader does sometimes split and/or merge parts of the physics model even when they seem to be clearly defined separate cubes and I suppose that's what happens here. One trick that often fixes it is to do a % simplification with a fairly high value (typically around 80%) That trick is a bit unpredictable but you can always use the preview spread to check if it worked before you upload.

For the Analyzed CUBE physics I had seven quadralaterals cubes(so maybe those are "hulls"). BOTH the 12 li version (Firestorm in Agni) and the 3li version (same files and LOD settings but using the Linden viewer) had perfect (well good enough for me) physics. WORKING was never the issue.  There were NO pieces of the cube version of the mesh with more than six faces. All were cubes to start with but simply changed in shape. No edges were touching. I used the CUBE physics as the basis for the planes physics just deleting some faces and joining some vertices.  

I actually spent well over an hour and quite a few lindens retesting this over and over with the same results, so for "me" it is "truth" LOL and I am moving on anyway.  Happy that Aquila taught me planes physics as it worked (at 2li) with EITHER Linden or FS viewer.   

My biggest issue -- even if the uploader had a bad day -- was that repeatedly (REPEATEDLY) the same files and settings were good on Aditi and over the moon land impact on Agni (with Firestorm). 

Again. Sticking with my story. Maybe in another eight years it will be "understood" . 

 

Cheers and thanks for the reply. 

[ChinRey]

2 minutes ago, Chic Aeon said:

For the Analyzed CUBE physics I had seven quadralaterals cubes(so maybe those are "hulls").

You get the number of hulls at the bottom of the uplaod window after you've clicked the Calculate Weights & Fee button:

 

[Chic Aeon]

1 minute ago, ChinRey said:

You get the number of hulls at the bottom of the uplaod window after you've clicked the Calculate Weights & Fee button:

 

I get no hull info in Firestorm  and "I" at least know it is a Firestorm issue -- in part anyway. 

Here is what I have. 

Again, the real point is that everything WORKS with the Linden viewer although land impact and upload fees were STILL different for me even in the linden viewer. I have no clue WHY.

I did this a few more times this evening with non-matching upload data and on the third try ( new log in if that might matter) I finally got matching statement. I am NOT uploading again to test. Three times it seemed to be fine and once you changed to prim the land impact went up past 12. 

 

 

[arton Rotaru]

The download weight in these 2 last screenshots is different already. So the input data isn't the same. You might have entered the same numbers for the lower LODs, but the importer is nondeterministic. It can make something different each time you do this. To avoid that you will have to load all 4 LODs from file.

It's expected behavior and to disdain it is user error.

The physics model in the last image isn't used in the importers preview as well. It looks pretty much like the High LOD model is loaded as the physics there.

You can stick to which ever story you like indeed.

Just saying.

Edited August 5, 2017 by arton Rotaru

[Whirly Fizzle]

11 hours ago, Chic Aeon said:

I get no hull info in Firestorm  and "I" at least know it is a Firestorm issue -- in part anyway. 

Your images below from Firestorm show the hull info though....

[Chic Aeon]

3 hours ago, Whirly Fizzle said:

Your images below from Firestorm show the hull info though....

Ah OK. I see that AFTER I analyze!  Was looking at all the empty  Sorry. WAY too many times with this. 

The stairs look fantastic and came in at 2 with longer than necessary LODS for inside so happy with the end result. Looking forward to the new physics integration A LOT, but I am pretty good at planes physics now so something learned and that's a good thing :D.

 

Thanks. 

[Chic Aeon]

And OBVIOUSLY I LOVE FIRESTORM :D.    THAT is not the issue. The issue is closer to I would rather learn planes physics than deal with the Linden Viewer.   I DID write down what I need to do to get my skies working (on a sticky note) but then forgot that I had written it down. Rough day, but hey, those happen. Good end results and today I am working on arty and simpler stuff. (Actually the stairs WERE simple until they turned into "prim monsters" :D. )

 

[Chic Aeon]

One more thought on this thread.

Last night as I was taking those final pictures of the cube physics model where the land impact actually matched, I thought it was quite odd that the analyzed versions looked quite different that they did in the early morning. The night time models when analyzed looked  like the physics model had been painted on more or less -- no signs of the outlines of the cubes which for me is the norm. I am not sure how the difference happened but maybe someone else does.

I only made ONE cube physics model all day. I used the original one when I tested with the Linden viewer as well as my final upload. The stairs were the same at the end as the beginning except that I moved some UVs around and had to make a new UV map - hence new version of the model. 

I was watching the tutorial that I made while doing this project and noted the photo in the video and how different it was then the one at the end of the day. 

So pasting that in here in case it makes some difference. This is a shot of the original stair's upload on the beta grid -- which worked perfectly (walking up and down stairs and bumping up against the sides etc.). I think they came in as 3 land impact.  The red arrow is part of the screenshot from the video, not something I am trying to point out. 

On a close inspection (I hadn't noticed this before, it looks like the ENDS by the floor don't have a physics model surrounding them. They WORKS as if they did however and since there was only ONE cube physics model the pieces must have been there. So maybe just a trick of the coloring in the window.