Lock granularity, bubbles, shaders, texture filtering
-
Hey sthalik,
What I wanted here was a technical discussion, hence the brought-up points. Alas, they’ve been ignored, and the discussion has since turned into a political one
they’ve not been ignored!!!
Cheers
Biker -
What I wanted here was a technical discussion, hence the brought-up points. Alas, they’ve been ignored, and the discussion has since turned into a political one
There’s no political discussion bud. Just reminding Arty what this community use to be like.
If you can improve on the items you’ve mentioned feel free to do so and send us the work. We are not prejudice or bias towards our own efforts, so if you’re effort is an improvement it will get in.
cheers
-
Dunc said he’d release the HLSL shader source. Don’t wanna pull his tongue, but until this happens, there’s nothing I can do, besides giving advice how to structure the code for perf… But this can only be done basing on the leaked F4/SP code, no doubt touched and restructured by you somewhat, so whatever I can muster is of little value…
If the code was released you would no longer have a community. Look at the divergence with AF, FF and OF
That’s an argument for monoculture. Additionally, good code can survive on its own merits…
-
What I wanted here was a technical discussion, hence the brought-up points. Alas, they’ve been ignored, and the discussion has since turned into a political one
I can’t argue with you on technical stuff because you are looking at a black box and yet you tell what’s inside, without seeing it.
You speak of bubble as if it were some problem… bubble is there for a reason and a compromise between real simulation and performance (performance, it terms of HW limitations).
We can start a story about it and run it for a couple of pages easily. But I’ll short it out.
I’m not a CS guy, but I think it can all be sum to this:
The only serious optimization that Falcon code can take will be real multi-core and/or multi-GPU implementation. That will make some real difference.Question is - Who is able to do that? considering the man hours probably required and the fact that there is no money behind it. It’s not some 10 minutes work, that I promise you… and, If it can be so easily implemented, then why it’s not in some commercial products like DCS and FSX which are starving for more cycles??
EDIT: Let’s get practical, code will not be released in any shape.
-
Dunc said he’d release the HLSL shader source. Don’t wanna pull his tongue, but until this happens, there’s nothing I can do, besides giving advice how to structure the code for perf… But this can only be done basing on the leaked F4/SP code, no doubt touched and restructured by you somewhat, so whatever I can muster is of little value…
If the code was released you would no longer have a community. Look at the divergence with AF, FF and OF
That’s an argument for monoculture. Additionally, good code can survive on its own merits…
want an advice ?
If you are so good in programming , so fast and efficient, and if you want to contribute to Bms , you should code a replacement for LODEDITOR, it is really needed, all 3D developpers will worship you.
and you should do it in full partnership with WD to demonstrate your capability to work as a team partner…
Just saying 'BMS is shit" and “i am better than BMS coders and i can give them advises” will not be suffisiant to go somewhere
-
That’s an argument for monoculture…
Call that as you like … but it is BMS policy. Nothing you can do against that point.
And AFAIK … there is still an FFocs team working on another Falcon4 with a different “flavor”.
So … No mono-culture of Falcon4 … just a mono-culture of BMS.
BTW … with that kind of statements, I would say (personal opinion) that you are the “worst” (do not understand that as an personal attack … I just have no other better word in my poor English) candidate for a future potential team member… whatever your talent.
-
and you should do it in full partnership with WD to demonstrate your capability to work as a team partner…
100%
If you are so good in programming , so fast and efficient, and if you want to contribute to Bms , you should code a replacement for LODEDITOR, it is really needed, all 3D developpers will worship you.
+100!
-
I can’t argue with you on technical stuff because you are looking at a black box and yet you tell what’s inside, without seeing it.
I dimly recall it was you who said some year (?) ago that lock granularity is bad and too often threads are spawned… To whoever BMS member said that, I recommended refactor. Never got a response…
bubble is there for a reason and a compromise between real simulation and performance
Don’t you get bottlenecks in the executed code? If you do get any, either CPU or GPU, bubble can be further extended without perf loss.
The only serious optimization that Falcon code can take will be real multi-core and/or multi-GPU implementation. That will make some real difference.
It’s doable Takes some man-hours (And I’m anything but fast and efficient, more methodical, triple-checking), but done both 3D (bicubic texture filtering in software, wined3d hacking on 2 occasions) and parallelization work (commercial software, FTNOIR locking bugs correction from 1.6 to now) so it’s possible.
For instance, do you know a program called apitrace? It’s a GPU profiler, a good one at that. Together with Wine programmer, Stefan Dossinger, we found out why Falcon BMS worked so slow on ATI Linux driver.
Question is - Who is able to do that? considering the man hours probably required and the fact that there is money behind it.
Let’s see about the LODEditor first… As for me, never could take money from Falcon dev work, not being a snake, or a snake-man…
I’ll be reading LODEditor to find out what it does, what are the data structures, etc…
and you should do it in full partnership with WD to demonstrate your capability to work as a team partner…
Who is WD?
Just saying 'BMS is shit" and “i am better than BMS coders and i can give them advises” will not be suffisiant to go somewhere
Never did say it was shit… In fact, it isn’t shit so why would I say that? There are areas for improvement, but if there are none, software is obsolete already
As for giving advice, I can of course give advice only in areas where I possess expert knowledge in… There aren’t that many of these areas, and team members surpass me by much in most of the code’s areas (flight model for instance)…
OK guys, time to work on lodeditor… Read the manual of BaldEagle and so on… I’m putting the terrain tools on hold because of this.
One question - How different is LOD format from Cobra-RV, FF and F4.0 from BMS 4.32?
And AFAIK … there is still an FFocs team working on another Falcon4 with a different “flavor”.
There hasn’t been a single commit in 2-3 months IIRC… And these folk use .sln files in an “open-source” project instead of something more appropriate.
For the latter, it’s bearable, but lack of activity is worrying.
-
Who is WD?
I can only guess, maybe WaveyDave.
-
Let’s see about the LODEditor first… I’ll be reading LODEditor to find out what it does, what are the data structures, etc…
OK guys, time to work on lodeditor… Read the manual of BaldEagle and so on… I’m putting the terrain tools on hold because of this.
Pleased to read this
but lack of activity is worrying.
mmm … BMS too was also qualified a “vapor-ware” you know …
-
OK, so for all recent Max yearly releases, an importer is needed, that is, compatible with the Max API for given year? Or do modders prefer one given version?
Is there source code for the existing LOD editor importer? This would ease the burden of writing stuff for it significantly.
Moreover, for the actual LOD Editor replacement, what features do you need? Is it enough to export everything to Max in one format or another, along with viewing functionality, as well as exporting to .LOD format, -or- do you want it done some other way?
If someone could make a quick, informal description, I’d be grateful.
Additionally, if someone could draw up the UI screens on a piece of paper, how they’d IDEALLY look like (within reason, of course), that’d limit my role to a coder, as I know rather little about what you need from the tool… having a spec is helpful
Edit: Why write the actual LOD editor replacement and not just ‘better’ importer/exporter?
-
IMHO … those questions should be addressed to WaveyDave …
-
OK, so for all recent Max yearly releases, an importer is needed, that is, compatible with the Max API for given year? Or do modders prefer one given version?
Is there source code for the existing LOD editor importer? This would ease the burden of writing stuff for it significantly.
Moreover, for the actual LOD Editor replacement, what features do you need? Is it enough to export everything to Max in one format or another, along with viewing functionality, as well as exporting to .LOD format, -or- do you want it done some other way?
If someone could make a quick, informal description, I’d be grateful.
Additionally, if someone could draw up the UI screens on a piece of paper, how they’d IDEALLY look like (within reason, of course), that’d limit my role to a coder, as I know rather little about what you need from the tool… having a spec is helpful
Edit: Why write the actual LOD editor replacement and not just ‘better’ importer/exporter?
Importer exporter is already in development, contact waveydave for team work
-
Dave is away until next week.
-
As for me, never could take money from Falcon dev work, not being a snake, or a snake-man…
I’ll be reading LODEditor to find out what it does, what are the data structures, etc…
OK guys, time to work on lodeditor… Read the manual of BaldEagle and so on… I’m putting the terrain tools on hold because of this.Nooooooo, … I was so much in hope that you finish what you’ve started first.
There is already a good working 3DSMAX to LOD export tool by WD,
and sure, LodEditor could become more user friendly, but it works.
… and that’s why there are so many new 3D models around.On the other hand look at the terrain tools.
There are just a few old specific tools out there, which we could just feed with old data at the start.
Dem2Terrain for example, feeded with old SRTM and “unfiltered” feature data (.e00).… and that’s why there are no finished terrains/theaters out there,
because developing a theater is a pain in the the as job with the current tools available.So IMHO those new terrain tools are way more needed than a new LE.
Now if you put the terrain tools on hold to code a new LE,
and maybe become a BMS coder, then (I’m almost sure) the community will never see new terrain tools from you.… just my 2 euro cent
Cheers,
LS -
I think we dont need another Lodeditor. WD 3dmax exporter works like a charm. LODED as is is sufficient to edit some parents or CTs for testing and setup for new models. All other is done via 3ddb compiler in data repository. I will tell you sthalik, what we desparately need. We need tool like TacEdit with source so we can expand save game format and db and not to be stuck with original TacEdit. To me that tool is far more needed than another Loded. I can imagine if done properly it can be considered as entry project to the team. Just my 2 cents.
As i saw you like to work with data structures and get how all that works. For new TacEdit you dont need BMS code. SP3 exe code that is/wasavailable for download on internet is suffcient for such a tool as this formats are untouched in BMS code for obvious reason to do not break compatibility with original TacEdit from what we do not have source of course.
-
I agree with ab, as waveydave exporter importer is so well advanced tacedit replacement would be more usefull
-
A particle system tool that would allow people to run and visually see the effect they made would be great to have as well.
-
Hi sthalik
I dimly recall it was you who said some year (?) ago that lock granularity is bad and too often threads are spawned… To whoever BMS member said that, I recommended refactor. Never got a response…
Nope… not me. To start with, I’m not a professional coder, and TBH not an expert to threads and processes.
Don’t you get bottlenecks in the executed code? If you do get any, either CPU or GPU, bubble can be further extended without perf loss.
TBH, I don’t know (I know bubble code mainly from reading some)… but assuming the main bubble function(s) took some changes (probably some rewrites) through the years, I don’t think there are serious bottlenecks, as if there were, they would have been detected by profiling and a fix would have been at least tried or searched for, and that isn’t the case, AFAIK.
It’s doable Takes some man-hours (And I’m anything but fast and efficient, more methodical, triple-checking), but done both 3D (bicubic texture filtering in software, wined3d hacking on 2 occasions) and parallelization work (commercial software, FTNOIR locking bugs correction from 1.6 to now) so it’s possible.
For instance, do you know a program called apitrace? It’s a GPU profiler, a good one at that. Together with Wine programmer, Stefan Dossinger, we found out why Falcon BMS worked so slow on ATI Linux driver.
I’ve done some MPI and Open MP coding in some parallel processing university course (Although not CS or SW degree) and that’s my main knowledge of parallel processing. Never tried to check if something like that can fit Falcon code. I guess that if someone with expertise in multi threading and multi-core coding methods will try to get something implemented, it could work, even locally in some high processing load areas of the code (if that makes any sense…)
a breakthrough in that direction will be revolutionary for a sim like Falcon which can starve sometimes for CPU cycles.
I think we dont need another Lodeditor. WD 3dmax exporter works like a charm. LODED as is is sufficient to edit some parents or CTs for testing and setup for new models. All other is done via 3ddb compiler in data repository. I will tell you sthalik, what we desparately need. We need tool like TacEdit with source so we can expand save game format and db and not to be stuck with original TacEdit. To me that tool is far more needed than another Loded. I can imagine if done properly it can be considered as entry project to the team. Just my 2 cents.
As i saw you like to work with data structures and get how all that works. For new TacEdit you dont need BMS code. SP3 exe code that is/wasavailable for download on internet is suffcient for such a tool as this formats are untouched in BMS code for obvious reason to do not break compatibility with original TacEdit from what we do not have source of course.
100% agree!
A particle system tool that would allow people to run and visually see the effect they made would be great to have as well.
You can do that already D
.sfx command, no?
-
Hehe…I hate using that. :mrgreen:
I’d rather have a tool so I can work on my laptop without having BMS installed.