Beta Release: GPT (cockpit texture extraction, remote cockpit control, shm mirror)
-
…
Curse my stupid head… i thought that start.bat was a test routine. Nevermind, nevermind… it’s working now. :rolleyes:
-
hello all
Thanks GiGurra for your great work.
I have an question, is it possible in an next version to replace the white empty box for MFD on DisplayReceiver by dark empty box? When i start my home cockpit before the PC Serveur with F4 (and F4 run in mission the first time), MFD Screen are white. i prefer black. If it is possible, it will be great.
Thanks.
-
+1 to black screens
And since we are in this request mood… and asking is for free…Could we have an option to display the rings an azimuth lines on RWR?
Thanks again for your work.
-
Well it’s interesting to see that this thread is still alive :).
I made gpt in its current state mostly just as an experiment for my home pit and it seems others are still using it.I’ve since pretty much stopped playing bms, now working on a ton of diff sw projects when I’m not stuck @ work.
I’ll see if I can remember to implement black screens. The fix probably is just a 1 minute thing but I’m afraid getting my development environment to support my gpt experiments might take a little while. -
I will be very gratefull too if you could spend that minute on this
-
Hello,
Still not managed to get it working by batch due to Java fault.
Now running Jave 7 update 55, but still same.Any one an idea?
Did have problem that I got white MFD’s on my slave display while before it was working fine.
Running BMS on main machine and Helios , MFDE, GPT display extract on slave PC.
Both connected via home network.Config setting:
falconhook_shm]
active = 0
name = “GiGurraTexturesSharedMemoryArea”[falconhook_socket]
active = 1
addr = “192.168.0.126”
port = 8051
max_kbps = 100000
max_hz = 50
jpegQual = 70After hours off looking into the problem finally discovered the IP of my slave was changed.
New IP solved the problem :woohoo:Remains a JAVA problem I cannot fix while starting KCR and SHM on my BMS machine via the batch.
Error occurred during initialization of VM
java/lang/NoClassDefFounderError: java/lang/invoke/AdapterMethodHandle.Running XP with JAVA 7 update 45.
Running the JAR files directly is working fine, only via batch is giving this error.
Anyone?
-
I would like to get the source code for this but the login / password given in the first post do not work, anyone?
-
I’m not sure if this will improve your situation, but you can try using the GPT DisplaysTransmitter, and set the ini file to use SHM instead of TCP messages (That way it will just download the displays texture and write it to SHM as raw RGBA instead of sending over TCP as JPEG). Then launch the DisplaysReceiver on the same computer. Though this will probably reduce your fps compared to BMS’s built mfd extractor, but it might allow you to play full screen, and might reduce stutters.
My specs are:
Primary PC (runs bms with the GPT DisplaysTransmitter, the GPT ShmTransmitter and the GPT KeyComReceiver)
generation 1 i7@3,2 GHz (i7 960)
pcie 2.0 16x
gtx 680Slave PC (runs all the other GPT tools and Helios)
generation 1 [email protected] GHz (i7 920)
2xNvidia GT210s (low end gpus, but enough for helios & mfds)––
At first I got the Gtx680 in order to try to alleviate the stutters and fps loss that I had before with my previous graphics card on my primary PC (using the built-in extractor) - back then I had the gt210s installed in my primary pc to run the mfd monitors (I also tried a gtx260 with the same results), but the performance wasn’t very good. I had hoped that the GTX680 would solve the problem (since then I didnt need any extra GPUs), but unfortunately it didnt - the stutters when using the built in extractor got almost worse. At that point I decided I would try to make GPT instead, and let a slave computer render the MFDs, which has really made my FPS a LOT smoother. I recently flew a campaign mission with all the cool stuff like HDR on and superb fps from the tarmac to the target, which I was never able to do before :).
But I expect it will be different from system to system how GPT performs. I expect it to work good on nvidia cards though, as that’s what I’ve been using. The important thing is that the texture download is issued directly after the buffer swap (d3d Present function), and is performed asynchronously up to the point where the sim starts painting the cockpit displays texture again.
this is very interesting. i’m in the same boat as Ice. i have a laptop which runs BMS single monitor (external) just fine (60fps at 1080p). i’d love to use the laptop display for MFDs etc but the frame rate is unbearable, 8fps.
are you saying using GPT on my exiting computer wouldn’t be any better than BMS’s native MFD extraction?
i think i have an old laptop laying around somewhere, i could plug that in as a 2nd i guess. i suspect the performance need on that isn’t great as its all 2d stuff. but what about latency? i have been using the android MFD thing but gave up because the latency was too slow and it freezes constantly (on the android side). would you run the tcp link over wifi or should it be a network cable?
thx
-
i read one comment in here, can’t find it anymore but basically GiGurra said that GPT wasn’t designed to speed up MFD extraction on a single PC. has anyone tried to do that? it seems at all the component are in GPT already, just need to strip out the jpg compression/decomp (not needed since no networking), strip out the networking, and make the whole thing a single app to read shared mem and do the display. how much cpu could that use? not much i suspect.
if no one has done that, maybe i’ll give it a go
by the way, on my system, the framerate dropped from 60 to 40 with the GPT stuff on the laptop display, far better than than the built in MFD which drops it down to sub 10.
i bet with a single ap, no jpeg version all in c++ this would only loose 2-5fps
-
does anyone have source for this? i tried this
You can download the source code of GPT from SVN: https://www.gigurra.se/svn/GEAR/GPT/
The public login is public/public.but it says site restricted (and my firewall doesn’t like me going there).
if anyone has it and can email it or give me working download like, please p.m. me and i’ll send you my email
thx steve
-
I’ve tried to contact him via email but, no response.
-
I’ve tried to contact him via email but, no response.
thanks, i just got a PM from the OP so he’s looking for it for me now.
-
Hi guys, my configuration is 3 monitor, via nvidia sorround and Falcon bms in the Primary PC, Slave Pc runnig Helios but the only problem is how to configure Helios in LAN with the GTP , in particulary I do not understand how to connect Helios with the GTP shared memory receiver, because I want put the analogue gauges, adi, hsi, ias, alt, in the second pc.
Any Idea?
Thanks in advance
Geebee -
Hi everyone,
I have been using the GPT tools to run a laptop with Helios over a LAN to my PC running BMS, which works great, however I have found recently that a Windows 7 update causes BMS not to see the keystrokes from Helios anymore.
For now I have wound back the batch of updates and it works again, but was wondering if anyone else had this problem and if there was a solution? (I should add that the updates were on the laptop, not the PC)
Thanks! -
Short announcement.
I’m re-opening GPT development. There will be a new release within 1-2 wks.
The main differences are quite small for the end user:
- GPT now has a proper build script and source code is available at https://github.com/GiGurra/gpt
- GPT now uses MGen and MNet (the new data modeling toolkit I’ve been working on)
- GPT now uses JSON for config files, instead of previous unflexible ini files (example: https://github.com/GiGurra/gpt/blob/master/gpt-displaystransmitter/gpt_displaystransmitter_cfg.json)
- GPT DisplaysReceiver: now starts with black screens
- GPT DisplaysTransnitter: now only streams over tcp (BMS handles sharedmemory case natively)
- GPT DisplaysTransmitter: can now stream to multiple network targets
- Passing command line parameters to the game itself
Feel free to use the source if you see anything useful :).
-
Short announcement.
I’m re-opening GPT development. There will be a new release within 1-2 wks.
The main differences are quite small for the end user:
- GPT now has a proper build script and source code is available at https://github.com/GiGurra/gpt
- GPT now uses MGen and MNet (the new data modeling toolkit I’ve been working on)
- The DisplaysReceiver: now starts with black screens
- GPT DisplaysTransnitter: now only streams over network (BMS now handles sharedmemory case natively)
- GPT DisplaysTransmitter: can now stream to multiple network targets
Feel free to use the source if you see anything useful :).
Thank you. U back is good news.
I would like to ask you to reconsider the limitation of display transmitter working only over the net.
This is the reason why.
Although bms has the built in extractor, if you want to use it bms is forced into windowed mode.This in turn penalizes a lot some nVidia users.
I first started using GPT in the standalone mode to be able to run bms on full screen mode.
At least on my rig, the performance difference between running bms Windowed or fill screen was huge.
Later I started using gpt on server client config, and so far I am a very happy user.
Thanks again for your tools.
-
Thank you. U back is good news.
I would like to ask you to reconsider the limitation of display transmitter working only over the net.
This is the reason why.
Although bms has the built in extractor, if you want to use it bms is forced into windowed mode.This in turn penalizes a lot some nVidia users.
I first started using GPT in the standalone mode to be able to run bms on full screen mode.
At least on my rig, the performance difference between running bms Windowed or fill screen was huge.
Later I started using gpt on server client config, and so far I am a very happy user.
Thanks again for your tools.
+1
GPT is the only standalone tool now to run full screen with the ability to alt tab to windows and back without losing sync in the mfd’s and ded.
Thanks GiGurra.
-
Thank you. U back is good news.
I would like to ask you to reconsider the limitation of display transmitter working only over the net.
This is the reason why.
Although bms has the built in extractor, if you want to use it bms is forced into windowed mode.This in turn penalizes a lot some nVidia users.
I first started using GPT in the standalone mode to be able to run bms on full screen mode.
At least on my rig, the performance difference between running bms Windowed or fill screen was huge.
Later I started using gpt on server client config, and so far I am a very happy user.
Thanks again for your tools.
Does bms not export textures to a shared memory already since a few patches back?
Or does that ALSO require running in windowed mode?The new displaystransmitter can stream to localhost. Of course direct shared memory is faster, but if your
gaming pc has 4 cores or more, it shouldn’t matter. BMS itself consumes maybe 1-2 cores and GPT’s texture and
streaming management maybe consumes 0,5 core if you export at 50+ Hz at double resolution cockpit displays (2x0,5 if you stream to localhost).
Empirical values from my own setup though, your results may vary (gaming on an i7 3770k). I have an old core2duo as a slave pc for
running my displays remotely - works no problem.The alt-tab story itself is a bit interesting actually. I first noticed that alt-tab (originally, in the first gpt prototypes) didn’t work.
I had to read up on the Direct3D and then, it turns out (at least for directx9, which is what bms uses) that direct3d destroys all
textures and lots of other resources when you alt tab (it also gives you a function callback), but as soon as you re-enter the game,
you (=the programmer) manually have to load up all your resources again. Guess this is one “optimization” that backfired royally
since so many games out there fail to handle the case of alt-tab. -
Does bms not export textures to a shared memory already since a few patches back?
Or does that ALSO require running in windowed mode?Sorry, beyond my knowledge.
All I know is that native bms does not support full screen mode. -
Are 1-2 BMS weeks? Lol.
I’m kidding, thaks very much for your work. I bought a computer just to use your software as a second rig on my pit because I needed a lot of displays and I don’t wanted to overload the main one.
Thanks again!