As someone who does not know, can someone please explain to me how EU differs from RF, and how the IC2 power system differs from other tech mods?
I do have familiarity with Reika's RotaryCraft, if that helps.
As someone who does not know, can someone please explain to me how EU differs from RF, and how the IC2 power system differs from other tech mods?
I do have familiarity with Reika's RotaryCraft, if that helps.
Ok, lets address these one at a time.
First, you might want to have different flags for client and server. They will be doing different things, with different needs, especially with the memory.
Lets start here. Try starting the server with this:
-XX:+UseG1GC -Xmx3400m
and see if it works.
Now, are you able to use JVisualVM to examine the memory heap? (Upper right corner of the process monitor tab)
Basically, if it works, I'd want to see either a screenshot of the memory usage, or a copy of the garbage collection log, as the first thing to try to figure out why you are running out of memory during startup with some GC's and not other GC's.
He is very mischievous and often likes fail, if Vanilla Forge server to run next to the game and it will be better.
Sorry, the auto-translation came out gibberish. I did not understand you.
during the generation of "Old gen" will be filled with the process and either hang or fall.
Not really sure I understand this either.
If old gen fills, that's an out of memory error.
The point of garbage collection should be to remove enough junk to prevent that.
CMS garbage collection really, *really* needs to avoid premature promotion.
The best way to have that is with a giant new space.
honestly at this size (186), the number of garbage collections is about ~5250(all types throughout heap), with a total duration of 3 minutes(1 collect an average of ~34 ms), per 30 minutes games. The increase in the younger generation reduces the number of garbage collections, but unfortunately the problem with 1 FPS that does not solve already checked.
What I can say is that, having done a lot of testing, minecraft garbage collects best when there is enough space in New such that objects live in Survivor for at least 30 seconds (to die off as junk) before being promoted to Tenured. That needs a lot of space in New.
use GLSL/OpenCL as an easier option than CUDA?
Sure. Code written to use OpenCL or the equivalent will run fine on any system with OpenCL drivers. Not all systems have it; older systems did not. When minecraft was developed, I don't think it was widespread at all.
Now, if you want to use OpenCL as a way to run code that was not designed for OpenCL? I don't think you can (I am not familiar enough with it).
Quoteonly run with the parameters: -XX:PermSize=300m -XX:MaxPermSize=300m -Xmx3400m
First, on J8, PermSize arguments are ignored.
Second, "-XX:PermSize" should never be mentioned. "-XX:MaxPermSize" is sufficient; the garbage collection will reallocate memory as needed, and it avoids wasting memory that is not needed for the permgen.
Minecraft has leaks. Even server-side (dedicated server). So yea, no GC will get rid of all the garbage, because it does not fit the definition of garbage (there are still pointers in hashmaps that point to it). Yet either you are saying that the world won't generate because of leaks without these options, or I'm misunderstanding you.
I can say that CMS wants a large new generation. 186 MB is not enough. I use 800, and my modpacks are not that huge. The biggest key for me is to get the large collections down to 30 seconds or longer.
First, you are horribly over-specifying stuff to the JVM. Do you even know what those options are doing, or did you just copy/paste? Did you test to see if those are even right? Most of those have "auto-tune" behavior by the VM.
Second: Moving work from the CPU to the GPU is not trivial. Minecraft was written to primarily use the CPU and not the GPU; getting a vendor-independent set of routines to make use of a GPU that is not known at the time you write code is really, really hard.
FPS is determined by the client thread, and how long it takes to do things, primarily redrawing. TPS is determined by the server thread, and how long it takes to do things, primarily ticking blocks and entities. It is very possible for the server work to be short, and generate massive redraw work for the client. The client is very good at only doing a few things when it has a big backlog, and prioritizes stuff close to you in your field of vision; yet it can still find that one task will take a large fraction of a second.
I would strongly recommend removing most of those args to start. You mention memory leaks, but there's nothing I can see there that deals with any attempt to diagnose that.
Is fastcraft supposed to work in dev/Eclipse?
Here's what I got with 1.21:
https://gist.github.com/keybounce/5f673d7ad37315a9aff3
There's a very similar error using the new ctest-17 version (the addresses on the net.minecraft.world.World line differ).
Attempting to use this with Mystcraft NEI plugin results in crashing with out of memory during texture registration (3 GB heap).
Mystcraft NEI: http://binarymage.com/forum/viewtopic.php?p=23473#p23473
Crashlog: https://gist.github.com/keybounce/b9dda1a11f342adf4ad0
I don't know.
If I did not have a wiki telling me exactly what to cross-breed with what to get what, I probably would not bother with bees and trees. The idea of "I'll get something every time" means that I never have to worry about "failure, failure, failure, failure, gee, I wish a had a clue, faulure, fauilure, failure, ...".
Knowing that every combination gives me something, and I can look at the resultant numbers, mean that I'll get something "good enough" eventually.
Think of vanilla horse breeding. There's randomness. But even without a WAILA plug-in to tell me what the numbers are, I can test the speed with a redstone track, I can test the height of jump with some snow layers, and I know I can breed the higher numbers for potentially higher children. And repeat with donkeys. Cross for mules, now I have good quality movement and storage.
QuoteFine, a player could spend hours and lots of resources in their survival mode world to get what they want
You mean, like they do now with bees, trees, flowers, etc; with horses, donkeys, mules; etc? Those are what I know of; this is just more of the same.
If there's some underlying order -- some sense of "the output numbers depend on the input numbers" -- something for players to work with as "we're learning what works/does not work", then great.
Otherwise, sure, you're having to play real-world "inventor" (really, tester), trying object after object after object ... something has to polarize the light (hey, a mirror), or twist the polarized light (... failed optical glass).
When you say that the tick timing mod can run into problematic corner cases, what are these problem corner spots? In other words, what things should I look out for?
Generally, my tps stays up at 20 unless I'm either doing new chunk gen, or running info a massive gas leak (a floating forge liquid that will eventually float up to Y=255, or fill up a sealed cave system.) I've got one test world where that gas is close to out of control (the bug is now fixed, but there's about 4-9 chunks with a ton of the stuff out in the open air), moving into range of the gas moves memory consumption from about 200K to 1.6 GB (that's tenured only), and ticks drop to an average of 1-2 tps.
Is there a way to disable Ic2 ore generation in the overworld / nether / end? Yes, that means twilight forest and mystcraft ages for ores.