| 00:14.37 | *** join/#bzflag WhatsUpDanger (0ca73d02@gateway/web/cgi-irc/kiwiirc.com/ip.12.167.61.2) |
| 05:07.46 | *** part/#bzflag kierra (~quassel@104.243.40.186) |
| 05:11.06 | *** join/#bzflag moriah (~quassel@unaffiliated/kierra) |
| 07:26.32 | BZNotify | bzflag: 03atupone commented on pull request #232 "Vertex buffers object handling" (06https://git.io/Je9oR): > ... |
| 07:29.12 | Flash | tupone: if you can wait for a couple of weeks, I'll have some time during the Christmas break |
| 07:36.17 | macsforme | Flash should give us all a crash course on modern C++ :-) |
| 07:36.33 | Flash | heh. I just did :P |
| 07:38.39 | Flash | https://youtu.be/JfmTagWcqoE |
| 07:39.02 | macsforme | I take it writing C++ must be part of your day job? |
| 07:39.08 | Flash | https://youtu.be/xnqTKD8uD64 |
| 07:39.16 | Flash | writing, reviewing, mentoring |
| 07:39.33 | Flash | https://channel9.msdn.com/posts/C-and-Beyond-2011-Herb-Sutter-Why-C |
| 07:39.44 | Flash | pretty much anything by Herb Sutter :) |
| 07:57.25 | BZNotify | bzflag: 03atupone 11synchronized pull request #232 "Vertex buffers object handling" (06https://git.io/Je9kl) |
| 08:05.00 | BZNotify | bzflag: 03atupone 11synchronized pull request #232 "Vertex buffers object handling" (06https://git.io/Je9kl) |
| 08:08.03 | macsforme | tupone: I didn't dive in as deep as Flash did (yet), but I had a hard time understanding your model and what requirements you are trying to fulfill |
| 08:09.06 | tupone | Flash: I can wait :) This class is needed, and is not a rewriting of something is already on bzflag. Coding is welcomed and you have access to my repository so we can fix the code before merging into the main repo |
| 08:10.05 | macsforme | I understand there is some need to centralize VBO creation and destruction so we can do it when we create or destroy a GL context, and I understand tracking indices and tracking memory... but otherwise I don't understand why it is so centralized |
| 08:10.28 | tupone | macsforme: managing internal buffers and avoid switching between them |
| 08:11.11 | macsforme | and actually, you did rewrite a lot of what I did in the embedded port, where I wrote a class to manage vertex arrays... that makes me sad |
| 08:11.30 | macsforme | https://github.com/macsforme/bzflag-embedded/blob/embedded/include/DrawArrays.h |
| 08:11.45 | macsforme | https://github.com/macsforme/bzflag-embedded/blob/embedded/src/ogl/DrawArrays.cxx |
| 08:13.07 | *** join/#bzflag Sgeo (~Sgeo@ool-18b982ad.dyn.optonline.net) |
| 08:15.23 | tupone | maybe VAO can reduce the need of avoid switching but I have not used it yet |
| 08:16.02 | macsforme | we don't have VAOs until OpenGL 3 |
| 08:16.18 | tupone | yes, that is one of the reason :) |
| 08:17.39 | macsforme | nor in OpenGL ES until version 3 |
| 08:19.07 | tupone | So I created big buffers and used it for all the glDrawArrays, changing only the indexes |
| 08:20.49 | tupone | at the beginning I used Vertex/Normal/Texture/Color buffers for all, but then I realize that is a waste of memory. Some Draw uses only vertex and color, some only vertex and normal ... |
| 08:21.25 | tupone | so I created a set for VTN, one for VTC, one for VT ... as needed |
| 08:24.06 | macsforme | what kinds of different components will share a VBO? I could understand for drawing a basic quad, but I don't understand further... |
| 08:37.04 | macsforme | I also don't understand why you are tracking VBO names (allocated and free) and reusing them rather than freeing them when done and then asking for a new one when needed |
| 09:07.16 | macsforme | and you are storing vertices, normals, and texcoords in different buffers, rather than interleaving them into one? |
| 09:09.42 | macsforme | I don't think that's how we want to do it... |
| 13:40.26 | *** join/#bzflag sehh (~sehh@unaffiliated/sehh) |
| 13:41.22 | sehh | hey there |
| 13:41.30 | blast007 | hello |
| 13:42.08 | sehh | anyone knows if its possible to compile just the server part? |
| 13:42.25 | blast007 | yes. what operating system are using? |
| 13:42.31 | sehh | centos 8 |
| 13:43.04 | blast007 | you can disable the graphical client and/or bzadmin by passing some flags to the configure script: ./configure --disable-client --disable-bzadmin |
| 13:43.39 | blast007 | so, if building from our git, it'd be: ./autogen.sh && ./configure --disable-client --disable-bzadmin && make |
| 13:43.49 | sehh | ah nice, I'll give that a try and see how far I get. |
| 13:44.46 | sehh | thanks for the info |
| 13:52.01 | blast007 | you're welcome :) |
| 14:07.15 | sehh | erm |
| 14:07.24 | sehh | configure: error: A c++ compiler with C++0x support is required to build BZFlag |
| 14:07.28 | sehh | I'm missing something |
| 14:11.55 | blast007 | needs g++ installed |
| 14:12.35 | *** join/#bzflag I_Died_Once (~I_Died_On@unaffiliated/idiedonce/x-1828535) |
| 14:12.48 | blast007 | see README.Linux for the needed dependencies, though some of those are only needed if you're building everything |
| 14:13.53 | blast007 | probably this is enough: yum install gcc-c++ libtool c-ares-devel libcurl-devel zlib-devel |
| 14:14.06 | sehh | I believe I'm missing gcc-c++ |
| 14:15.00 | sehh | I had done 'dnf install gcc cpp make.. etc' only and missed gcc-c++ |
| 14:15.16 | sehh | its compiling right now |
| 14:17.58 | sehh | while waiting to compile, I'll now go read README.Linux :] |
| 14:38.37 | *** part/#bzflag sehh (~sehh@unaffiliated/sehh) |
| 16:25.16 | *** join/#bzflag WhatsUpDanger (0ca73d02@gateway/web/cgi-irc/kiwiirc.com/ip.12.167.61.2) |
| 17:10.27 | tupone | macsforme: my first trial was with interleaving, but on my PC was slower then separate buffers |
| 17:11.56 | tupone | and I use a single vbo for a lot of user buffers. This should avoid as most as possible to save call to gl. |
| 17:12.19 | tupone | not to save, to do. I want to save call :) |
| 17:26.12 | BZNotify | bzflag: 03atupone 11synchronized pull request #232 "Vertex buffers object handling" (06https://git.io/Je9kl) |
| 17:35.44 | BZNotify | bzflag: 03atupone 11synchronized pull request #232 "Vertex buffers object handling" (06https://git.io/Je9kl) |
| 17:38.18 | macsforme | tupone: optimizations should not be based on only one GPU/driver/OS combination... especially not an integrated GPU in Linux, which is not representative of the average user's system |
| 17:38.55 | macsforme | my impression was that interleaving vertex data is most optimal, generally speaking |
| 17:40.13 | macsforme | this could be why others who tested got worse performance than with the original code, while you saw an improvement |
| 17:45.42 | macsforme | as far as using a single VBO rather than many, I recognize that you are trying to save GL calls by doing optimizations on the CPU side... but I am skeptical whether the overall benefit will be seen, and will be worth the added code complexity (such as memory management issues, as Flash pointed out) |
| 17:53.25 | BZNotify | bzflag: 03atupone 11synchronized pull request #232 "Vertex buffers object handling" (06https://git.io/Je9kl) |
| 18:01.38 | macsforme | moriah: did you think of any large event replays that would be good for a stress test/benchmark? |
| 18:04.05 | allejo | oh yea, someone nag me to put up a replay server for some events on Troy's server |
| 18:07.07 | blast007 | allejo: hey, put up a replay server. hey, listen. |
| 18:07.10 | blast007 | ducks |
| 18:08.22 | allejo | i mean, you're not wrong. but like, later today when I have access to my server lol |
| 18:11.06 | the_map | well, it's later. you have access to your server? |
| 18:11.12 | the_map | :p |
| 18:20.35 | tupone | macsforme: I did to save GL call so GPU can benefit from that. |
| 18:22.05 | tupone | Maybe I'm wrong, But at the beginnning of my work, I was switching and did that change (not switching buffers only if possible) and that improved the FPS |
| 18:25.54 | tupone | I have a doubt about using GL_POLYGON on the GPU buffers. It seems that simple mesh like the one in xs.bzexcess.com have that and changing from client buffers to GPU buffers loose half of FPS. I cannot explain it |
| 18:27.07 | tupone | have 3 buffers clients changed with 3 GPU buffers :/ |
| 18:36.45 | blast007 | fun bit of info about game performance. on linux with a Radeon RX 570 8GB, at 1920x1200 on Rats Nest, with radar turned off I was getting 550FPS. With the fast radar, it drops to about 300, fast sorted to around 270, enhanced to 200, and normal to 120. |
| 18:42.43 | The_Noah | :O |
| 18:56.01 | macsforme | I like how things like the console and radar have more of an impact on framerate than actually rendering the 3D view :-P |
| 18:57.57 | macsforme | allejo: maybe I could just ask you... what is the biggest event (as far as number of players) that we have a replay of? |
| 18:58.52 | allejo | kierra was asking me about that, she thinks some of the replays I have for Troy's server might have some large player counts |
| 18:59.10 | allejo | we'll know once I get the replay server up for it (hopefully tonight) |
| 19:19.13 | blast007 | the stagedSceneGenerator could also be useful as a benchmark. could have a bunch of fixed player/shot locations (though the shots may be limited as they're server-side) |
| 19:19.32 | blast007 | world weapons* |
| 19:31.42 | macsforme | I guess it also matters what kind of objects are in the map, if we're benchmarking the renderers for different object types |
| 20:04.01 | allejo | oooo do the box tower |
| 20:37.13 | blast007 | allejo: this one? https://static.bzexcess.com/boxie_warz.bzw |
| 20:37.42 | blast007 | https://static.bzexcess.com/misc_screenshots/BoxieBenchmark/BoxieBenchmark001.png |
| 20:39.09 | allejo | yup! |
| 20:39.35 | blast007 | and here's the lite version: https://static.bzexcess.com/misc_screenshots/boxes01.png |
| 20:39.46 | allejo | lol |
| 20:46.39 | blast007 | there's also a bunch of old maps here: https://static.bzexcess.com/downloads/bzWorlds/ |
| 20:47.20 | blast007 | (some of them are deal links since I had to build that from what was grabbed by archive.org) |
| 20:48.24 | blast007 | kinda forgot about the old script/program that converted images to BZFlag maps (see the Calvin&Hobbes maps there) |
| 20:49.01 | blast007 | JungleFighter1.0.bzw has a large number of objects |
| 21:57.05 | tupone | Is there a way to reprocess mesh to remove polygon? They are deprecated (and slow) |
| 22:06.23 | WhatsUpDanger | just draw it as a triangle fan |
| 22:15.58 | WhatsUpDanger | void split_poly(const std::vector<Vector3> &verts) |
| 22:38.24 | The_Noah | I just tested the radar and console on Urban Jungle and turning them off makes a bug-huge difference. |
| 23:32.13 | *** join/#bzflag Zehra (~Zehra@unaffiliated/zehra) |
| 23:52.40 | *** join/#bzflag spldart (~spldart@bzflag/contributor/spldart) |
| 23:52.40 | *** mode/#bzflag [+v spldart] by ChanServ |
| 23:52.58 | moriah | macsforme: this coming sunday we're having a community event on Troy server to commemorate those gone....T-roy's bday was yest. Sent out pms to all reg'd players & posted to forums. Hopefully it will be large event. Allejo says he'll have replay server for it |
