| Time |
Nick |
Message |
| 00:08 |
|
jwmhjwmh joined #minetest-dev |
| 00:31 |
|
jwmhjwmh joined #minetest-dev |
| 00:40 |
|
kilbith joined #minetest-dev |
| 00:52 |
|
kilbith joined #minetest-dev |
| 00:56 |
MTDiscord |
<savilli> Haha, I started to read opengl code to prove my point, and decreased minetest CPU usage with Nvidia GPU in half instead |
| 01:03 |
|
kilbith joined #minetest-dev |
| 01:06 |
schwarzwald[m] |
Do you mean you halved Minetest's CPU resources when Minetest is running on an Nvidia GPU? |
| 01:06 |
schwarzwald[m] |
Just now? |
| 01:06 |
schwarzwald[m] |
It's only been 2 hours. |
| 01:18 |
MTDiscord |
<savilli> Yeah, I've patched Nvidia opengl driver, and CPU usage decreased in half |
| 01:18 |
MTDiscord |
<savilli> Probably Nvidia didn't care about CPU when wrote their driver |
| 01:19 |
schwarzwald[m] |
bruh |
| 01:20 |
schwarzwald[m] |
How can I get this patch? |
| 01:20 |
MTDiscord |
<savilli> Do you use linux? |
| 01:20 |
schwarzwald[m] |
Too bad it won't affect my new laptop - it has a Radeon GPU. |
| 01:21 |
schwarzwald[m] |
Yeah, I use Linux. |
| 01:23 |
MTDiscord |
<savilli> Well, I can send you patched libnvidia-glcore.so file. But I personally wouldn't recommend running so files from strangers. |
| 01:23 |
MTDiscord |
<savilli> I'll check if that part has been open-sourced. |
| 01:40 |
MTDiscord |
<savilli> Well, it's not. 100% proprietary. |
| 01:42 |
MTDiscord |
<savilli> Who would even open their kernel driver, but don't open their user-space libraries. Damn Nvidia. |
| 02:47 |
schwarzwald[m] |
How'd you even patch it in the first place? lol |
| 04:00 |
|
MTDiscord joined #minetest-dev |
| 04:37 |
|
calcul0n_ joined #minetest-dev |
| 04:52 |
|
YuGiOhJCJ joined #minetest-dev |
| 05:29 |
|
olliy joined #minetest-dev |
| 06:51 |
pgimeno |
@savilli I'm interested in that patched file |
| 08:44 |
|
appguru joined #minetest-dev |
| 09:14 |
|
appguru1 joined #minetest-dev |
| 09:54 |
|
appguru joined #minetest-dev |
| 11:03 |
MisterE[m] |
celeron55: your pr has been closed with adoption needed due to inactivity. Could you pick it back up again? We need that feature. |
| 11:03 |
|
kilbith joined #minetest-dev |
| 11:30 |
celeron55 |
MisterE[m]: of course... once i get a new vacation |
| 11:32 |
|
Taoki joined #minetest-dev |
| 12:25 |
|
cranezhou joined #minetest-dev |
| 12:51 |
|
troller joined #minetest-dev |
| 13:40 |
|
Fixer joined #minetest-dev |
| 14:54 |
kilbith |
btw pyrollo might be coming back as he'll work with me very soon |
| 16:43 |
sfan5 |
rubenwardy: reminder #12789 |
| 16:43 |
ShadowBot |
https://github.com/minetest/minetest/issues/12789 -- Revise bump_version.sh script to address shortcomings by sfan5 |
| 16:48 |
|
ely-the-kitsune joined #minetest-dev |
| 17:40 |
|
ely-the-kitsune joined #minetest-dev |
| 17:48 |
|
ely-the-kitsune joined #minetest-dev |
| 18:31 |
|
kilbith joined #minetest-dev |
| 18:48 |
|
vampirefrog joined #minetest-dev |
| 19:15 |
|
ely-the-kitsune joined #minetest-dev |
| 20:17 |
|
Taoki joined #minetest-dev |
| 20:20 |
diceLibrarian |
anything in the 5.6 code that could cause a memory leak? |
| 20:22 |
rubenwardy |
it's C++, anything could |
| 20:22 |
sfan5 |
any programming language can leak memory even |
| 20:22 |
|
ely-the-kitsune joined #minetest-dev |
| 20:26 |
diceLibrarian |
let's say it only STARTED in 5.6? |
| 20:26 |
diceLibrarian |
I'm sure you saw the stacktrace in the engine channel |
| 20:27 |
rubenwardy |
C++ is especially bad at it, given manual memory management |
| 20:28 |
sfan5 |
the point is that the answer to this question is never "no", if you have concrete pointers to a regression open a bug report |
| 20:28 |
sfan5 |
or post the details here to have them looked at |
| 20:29 |
diceLibrarian |
I have no idea what caused it since I'm programming illiterate |
| 20:33 |
|
fluxionary joined #minetest-dev |
| 20:52 |
sfan5 |
me neither because you didn't provide any details at all |
| 20:55 |
MTDiscord |
<exe_virus> This is a modding question that hits on engine internals. I am using a long tool chain to make a markdown reader writer for Formspecs. It's using lulPeg, which gets pretty intense in terms of using every function in the Lua lib. I am getting different output from the same codebase using plain Jane luaJIT, and Minetest (with luaJIT). I suspect we might be overriding some base library functionality for our Lua environment, such as |
| 20:55 |
MTDiscord |
table.concat or similar. Can someone point me to possible places we change behavior of base functions? |
| 20:57 |
sfan5 |
https://github.com/minetest/minetest/blob/master/builtin/common/misc_helpers.lua |
| 20:57 |
sfan5 |
https://github.com/minetest/minetest/blob/master/src/script/cpp_api/s_security.cpp#L75 |
| 21:09 |
MTDiscord |
<exe_virus> Great, thank you! We overwrite a frick ton, dang. Okay well I'll start hunting. I suspect it's somewhere around the loadfile, setlocale, io lines, dofile, but I'll have to test. Maybe insecure is a good test here |
| 21:09 |
sfan5 |
it is |
| 21:10 |
MTDiscord |
<exe_virus> Wait, you have an idea of which one? Maybe I can find a workaround? |
| 21:10 |
sfan5 |
that was to "insecure is a good test" |
| 21:10 |
MTDiscord |
<exe_virus> Gotcha, yep |
| 21:10 |
sfan5 |
all other stuff should be harmless additions that don't exist in later lua versions either |
| 21:10 |
MTDiscord |
<exe_virus> Yeah, the misc seems harmless |
| 21:18 |
diceLibrarian |
ran some tests |
| 21:18 |
diceLibrarian |
the OOM stops occuring once I disable dynamic shadows |
| 21:23 |
|
lebruhgamer[m] joined #minetest-dev |
| 21:34 |
MTDiscord |
<exe_virus> Narrowed down a lot of my errors, luckily, none so far require insecure. One error remains. Seems like JIT and MT have different table sorting or something deep in the reader, as **test** is being read as a bulleted list before being seen as a bold word in MT, but works as expected in JIT. |
| 21:37 |
MTDiscord |
<exe_virus> Lol, and devouring the first two letters of any emphasis used. These are fun |
| 22:35 |
|
panwolfram joined #minetest-dev |
| 23:54 |
|
diceLibrarian joined #minetest-dev |
