Android port efforts

For people working on the C++ code.
MoNTE48
Member
Posts: 321
Joined: Sat Apr 06, 2013 11:58
GitHub: MoNTE48
In-game: MoNTE48
Location: Internet

by MoNTE48 » Post

stu, I have tested it on many phones. I am well versed in Android, make ROMs, and I am sure that is nothing more than a game to crash happens can't.
MultiCraft Open Sourсe – https://github.com/MultiCraft

User avatar
hunterdelyx1
Member
Posts: 27
Joined: Fri Aug 10, 2012 02:14

by hunterdelyx1 » Post

+ Spoiler
That's what i got on my tab 3 7.0".

I used both apk from this topic and builded by myself.

If you need logs or something i can provide it. But it looks like one famous irrlicht bug...
Last edited by hunterdelyx1 on Tue Mar 11, 2014 20:11, edited 1 time in total.
Still waiting for utf 8 minetest formspec support, looking at freeminer.

Jordach
Member
Posts: 4525
Joined: Mon Oct 03, 2011 17:58
GitHub: Jordach
IRC: Jordach
In-game: Jordach
Location: Blender Scene

by Jordach » Post

hunterdelyx1 wrote: That's what i got on my tab 3 7.0".

I used both apk from this topic and builded by myself.

If you need logs or something i can provide it. But it looks like one famous irrlicht bug...
Had this before on a system which seems to be missing a certain OpenGL feature.
viewtopic.php?f=10&t=19056 Solar Plains Dev Server

User avatar
stu
Member
Posts: 923
Joined: Sat Feb 02, 2013 02:51
GitHub: stujones11
Location: United Kingdom

by stu » Post

hunterdelyx1 wrote: That's what i got on my tab 3 7.0".

I used both apk from this topic and builded by myself.

If you need logs or something i can provide it. But it looks like one famous irrlicht bug...
Interesting, have you tried any of the other ports? eg sapier's or that other fork
Last edited by stu on Tue Mar 11, 2014 22:06, edited 1 time in total.

User avatar
hunterdelyx1
Member
Posts: 27
Joined: Fri Aug 10, 2012 02:14

by hunterdelyx1 » Post

stu wrote:
hunterdelyx1 wrote: That's what i got on my tab 3 7.0".

I used both apk from this topic and builded by myself.

If you need logs or something i can provide it. But it looks like one famous irrlicht bug...
Interesting, have you tried any of the other ports? eg sapier's or that other fork
I tried "Lime minetest" but it crashes on startup loading( black screen before menu ).

I builded sapier's port. I have the same problem with it as with yours.
Still waiting for utf 8 minetest formspec support, looking at freeminer.

User avatar
thetoon
Member
Posts: 106
Joined: Tue Dec 11, 2012 12:55

by thetoon » Post

Kudos stu, the latest apk works like a charm on my Nexus 4. Most of the previous glitches I've seen are gone, and I'm fond of the controls like they are now : swipe to look around, long-click to dig, short-click to build. Small lags between chunk generation are gone too. I even got used to running-and-jumping (tip of the thumb on the up arrow, and when in need of jump just lower the whole thumb).

User avatar
spillz
Member
Posts: 138
Joined: Thu Feb 13, 2014 05:11

by spillz » Post

Hate to be the one bursting bubbles, but having tested the various APKs on a few devices now (nexus 7 I, galaxy note 2, memo pad HD 7) there is still a lot of work to do. Controls need a lot of tuning and while I can get a decent frame rate on all devices (20 to 40 fps) the draw distance s***s b***s (20 to 40 nodes). I get a solid 150+ node draw distances on both minecraft PE and Survival craft on all 3 devices, which even my Linux PC struggles with on minetest. So much for the superiority of a C++ engine. ;)

User avatar
Inocudom
Member
Posts: 3096
Joined: Sat Sep 29, 2012 01:14
IRC: Inocudom
In-game: Inocudom

by Inocudom » Post

spillz wrote:Hate to be the one bursting bubbles, but having tested the various APKs on a few devices now (nexus 7 I, galaxy note 2, memo pad HD 7) there is still a lot of work to do. Controls need a lot of tuning and while I can get a decent frame rate on all devices (20 to 40 fps) the draw distance s***s b***s (20 to 40 nodes). I get a solid 150+ node draw distances on both minecraft PE and Survival craft on all 3 devices, which even my Linux PC struggles with on minetest. So much for the superiority of a C++ engine. ;)
I take it that the number of visible polygons is not the sole fps dropper with Minetest?
Last edited by Inocudom on Sat Mar 15, 2014 03:54, edited 1 time in total.
You can now find my videos at BitChute: https://www.bitchute.com/channel/some_cheeky_jinuskian/

User avatar
sfan5
Moderator
Posts: 3933
Joined: Wed Aug 24, 2011 09:44
GitHub: sfan5
IRC: sfan5
Location: Germany

by sfan5 » Post

spillz wrote:I get a solid 150+ node draw distances on both minecraft PE and Survival craft on all 3 devices
MCPE and Survivalcraft where written with mobile device performance in mind, you can't expect something that was originally made for PC to just work on mobile devices like software that was designed for them.
Mods: Mesecons | WorldEdit | Nuke & Minetest builds for Windows (32-bit & 64-bit)

User avatar
spillz
Member
Posts: 138
Joined: Thu Feb 13, 2014 05:11

by spillz » Post

sfan5 wrote: MCPE and Survivalcraft where written with mobile device performance in mind, you can't expect something that was originally made for PC to just work on mobile devices like software that was designed for them.
Well aware and had no such expectation it would work out of the box. As I said, "there is still a lot of work to do" - it is called feedback and the last line was (mostly) a joke, hence the ;). Probably bugs that need fixing and performance tuning needed in irrlicht-android too.

User avatar
stu
Member
Posts: 923
Joined: Sat Feb 02, 2013 02:51
GitHub: stujones11
Location: United Kingdom

by stu » Post

To add to sfan5's comments, I would also point out that MCPE is not the same as Minecraft, AFAIK you cannot play on MC servers.
This (when finished) will be as fully featured as the pc game. I realize that performance will not be great on lower end devices atm, however, with mobile technology advancing at the rate it is, It is only a matter of time before performance levels for the average phone will reach that of your pc.
Already my 60 Euro tablet outer performs my old p4 desktop with onboard vga.

I do appreciate all feedback and agree that there is still a lot that could be done to improve performance in the short term, for both mobile and pc versions of minetest.
Last edited by stu on Sat Mar 15, 2014 17:37, edited 1 time in total.

User avatar
celeron55
Administrator
Posts: 465
Joined: Tue Apr 19, 2011 10:10
GitHub: celeron55
IRC: celeron55

by celeron55 » Post

If somebody is interested in improving Minetest's performance on whatever hardware it happens to perfporm poorly on:

Here are three important bottlenecks in Minetests rendering at the moment:
  • 1) Texture switches: The texture atlas was removed about a year ago because it was buggy, nobody was interested in fixing it and nobody realized it was actually pretty damn useful. However, in situations when there are a ton of different nodes to render at the same time (any city or forest with variation or almost any interesting place whatsoever), it can make a difference. Doing a comparison is simple: just define the textures of every node in the game to be the same (which auto-generated texture atlases effectively do from the hardware standpoint) and see how FPS differs from when they're not the same. The solution to this is to bring back auto-generated texture atlases. This affects GPUs quite variedly and unpredictably.
  • 2) Data transfer to the GPU: Currently zero data is stored on the GPU (doing such is often called using a VBO). It can be enabled, but due to how Irrlicht handles them and how Minetest manages meshes, there are memory consumption issues. This certainly affects high-end GPUs; however I don't know about mobile devices. Somebody should test it on those. It affects GPUs that draw faster than the interface they are sitting on transfers data.
  • 3) Drawing of nodeboxes that are buried underground: There is no code to detect such ones and thus they are fully drawn. On maps that have underground pipes and such, this raises the general vertex count, and due to all of them probably using different textures, the effect of this is multiplied by (1). Thus fixing (1) makes this a lot less important.

User avatar
Inocudom
Member
Posts: 3096
Joined: Sat Sep 29, 2012 01:14
IRC: Inocudom
In-game: Inocudom

by Inocudom » Post

celeron55 wrote:If somebody is interested in improving Minetest's performance on whatever hardware it happens to perfporm poorly on:

Here are three important bottlenecks in Minetests rendering at the moment:
  • 1) Texture switches: The texture atlas was removed about a year ago because it was buggy, nobody was interested in fixing it and nobody realized it was actually pretty damn useful. However, in situations when there are a ton of different nodes to render at the same time (any city or forest with variation or almost any interesting place whatsoever), it can make a difference. Doing a comparison is simple: just define the textures of every node in the game to be the same (which auto-generated texture atlases effectively do from the hardware standpoint) and see how FPS differs from when they're not the same. The solution to this is to bring back auto-generated texture atlases. This affects GPUs quite variedly and unpredictably.
  • 2) Data transfer to the GPU: Currently zero data is stored on the GPU (doing such is often called using a VBO). It can be enabled, but due to how Irrlicht handles them and how Minetest manages meshes, there are memory consumption issues. This certainly affects high-end GPUs; however I don't know about mobile devices. Somebody should test it on those. It affects GPUs that draw faster than the interface they are sitting on transfers data.
  • 3) Drawing of nodeboxes that are buried underground: There is no code to detect such ones and thus they are fully drawn. On maps that have underground pipes and such, this raises the general vertex count, and due to all of them probably using different textures, the effect of this is multiplied by (1). Thus fixing (1) makes this a lot less important.
Have you ever tested the VBO feature that Freeminer has, celeron55? I wish I could test it myself, but builds of that fork of Minetest are not very common.
You can now find my videos at BitChute: https://www.bitchute.com/channel/some_cheeky_jinuskian/

User avatar
PilzAdam
Member
Posts: 4026
Joined: Fri Jul 20, 2012 16:19
GitHub: PilzAdam
IRC: PilzAdam
Location: Germany

by PilzAdam » Post

Inocudom wrote:
celeron55 wrote:If somebody is interested in improving Minetest's performance on whatever hardware it happens to perfporm poorly on:

Here are three important bottlenecks in Minetests rendering at the moment:
  • 1) Texture switches: The texture atlas was removed about a year ago because it was buggy, nobody was interested in fixing it and nobody realized it was actually pretty damn useful. However, in situations when there are a ton of different nodes to render at the same time (any city or forest with variation or almost any interesting place whatsoever), it can make a difference. Doing a comparison is simple: just define the textures of every node in the game to be the same (which auto-generated texture atlases effectively do from the hardware standpoint) and see how FPS differs from when they're not the same. The solution to this is to bring back auto-generated texture atlases. This affects GPUs quite variedly and unpredictably.
  • 2) Data transfer to the GPU: Currently zero data is stored on the GPU (doing such is often called using a VBO). It can be enabled, but due to how Irrlicht handles them and how Minetest manages meshes, there are memory consumption issues. This certainly affects high-end GPUs; however I don't know about mobile devices. Somebody should test it on those. It affects GPUs that draw faster than the interface they are sitting on transfers data.
  • 3) Drawing of nodeboxes that are buried underground: There is no code to detect such ones and thus they are fully drawn. On maps that have underground pipes and such, this raises the general vertex count, and due to all of them probably using different textures, the effect of this is multiplied by (1). Thus fixing (1) makes this a lot less important.
Have you ever tested the VBO feature that Freeminer has, celeron55? I wish I could test it myself, but builds of that fork of Minetest are not very common.
Actually me and celeron55 created the VBO patch together, but it has memory leaks so it isnt in master.

User avatar
spillz
Member
Posts: 138
Joined: Thu Feb 13, 2014 05:11

by spillz » Post

A few issues with stu's latest android build, tested on a few servers with a bunch of mods enabled. probably should be surprised that it works at all...

1. Starting a client, app hangs on main screen trying play if I click connect instead of pressing enter on the password dialog.

2. Crash if trying to connect to a server with wifi disabled (oops)

3. Unified inventory mod doesn't appear to work correctly. Most buttons unresponsive.

4. Lots of actions don't work (e.g. punching mobs doesn't kill them, door won't open, many blocks can't be placed, mining items doesn't add them to inventory). It should just be tap screen to use right?

5. What happened to my rails! :)

Image


The Lime Minetest guy is a total tool for posting his fork to google play.
Last edited by spillz on Mon Mar 17, 2014 23:21, edited 1 time in total.

User avatar
spillz
Member
Posts: 138
Joined: Thu Feb 13, 2014 05:11

by spillz » Post

stu wrote:To add to sfan5's comments, I would also point out that MCPE is not the same as Minecraft, AFAIK you cannot play on MC servers.
This (when finished) will be as fully featured as the pc game.
That's true, but MC pocket still supports local multiplayer. If you are playing on a server that's taking a lot of load off the mobile device. But performance needs to be better than it is, but I'm confident you guys will get there!
I do appreciate all feedback and agree that there is still a lot that could be done to improve performance in the short term, for both mobile and pc versions of minetest.
Hopefully celeron55's tips will prove fruitful.

Is the buggy code for the texture atlas still around? That sounds like a promising place to start. EDIT: here's the commit that removed it: https://github.com/minetest/minetest/co ... 1596140526
Last edited by spillz on Tue Mar 18, 2014 00:48, edited 1 time in total.

MoNTE48
Member
Posts: 321
Joined: Sat Apr 06, 2013 11:58
GitHub: MoNTE48
In-game: MoNTE48
Location: Internet

by MoNTE48 » Post

See, the game will go far. I communicate with different people, looking for a way normal unpacking assets. I will share the results here.
UPD: Don't work on more models. Galaxy S2 and S4 too.
MultiCraft Open Sourсe – https://github.com/MultiCraft

MoNTE48
Member
Posts: 321
Joined: Sat Apr 06, 2013 11:58
GitHub: MoNTE48
In-game: MoNTE48
Location: Internet

by MoNTE48 » Post

Please, look this:
java.lang.RuntimeException: Unable to start activity ComponentInfo{com.Lime.Minetest/android.app.NativeActivity}: java.lang.IllegalArgumentException: Unable to find native library: Minetest
at android.app.ActivityThread.performLaunchActivity(ActivityThread.java:2059)
at android.app.ActivityThread.handleLaunchActivity(ActivityThread.java:2084)
at android.app.ActivityThread.access$600(ActivityThread.java:130)
at android.app.ActivityThread$H.handleMessage(ActivityThread.java:1195)
at android.os.Handler.dispatchMessage(Handler.java:99)
at android.os.Looper.loop(Looper.java:137)
at android.app.ActivityThread.main(ActivityThread.java:4745)
at java.lang.reflect.Method.invokeNative(Native Method)
at java.lang.reflect.Method.invoke(Method.java:511)
at com.android.internal.os.ZygoteInit$MethodAndArgsCaller.run(ZygoteInit.java:786)
at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:553)
at dalvik.system.NativeStart.main(Native Method)
Caused by: java.lang.IllegalArgumentException: Unable to find native library: Minetest
at android.app.NativeActivity.onCreate(NativeActivity.java:198)
at android.app.Activity.performCreate(Activity.java:5008)
at android.app.Instrumentation.callActivityOnCreate(Instrumentation.java:1079)
at android.app.ActivityThread.performLaunchActivity(ActivityThread.java:2023)
... 11 more
MultiCraft Open Sourсe – https://github.com/MultiCraft

User avatar
stu
Member
Posts: 923
Joined: Sat Feb 02, 2013 02:51
GitHub: stujones11
Location: United Kingdom

by stu » Post

Martin_Devil wrote:Please, look this:
java.lang.RuntimeException: Unable to start activity ComponentInfo{com.Lime.Minetest/android.app.NativeActivity}: java.lang.IllegalArgumentException: Unable to find native library: Minetest
Looks like the same problem Vanessa was having, you should maybe try sapier's version, It looks like that will become the 'official' version sometime soon (TM)
I will probably continue this at my leisure since it's a good learning exercise, also I don't like the UI of sapier's version and would have preferred to stay clear of java (that's not such a big deal though)

spillz: Thanks for the updates, I will look into those issues you mention. Some of them may have already fixed in my android_rebase branch, you could check that out but beware that the screen pointer no longer works in multi-player (at least not for me) this is due to #062de11
Also note that I added cURL support, although it does not crash, I am still not seeing the online server list YMMV please let me know if you try it.

btw, these changes are not in the current apk. You would need to build it yourself.

User avatar
sfan5
Moderator
Posts: 3933
Joined: Wed Aug 24, 2011 09:44
GitHub: sfan5
IRC: sfan5
Location: Germany

by sfan5 » Post

stu wrote:It looks like that will become the 'official' version sometime soon (TM)
Everyone ignores my version :-(
Mods: Mesecons | WorldEdit | Nuke & Minetest builds for Windows (32-bit & 64-bit)

User avatar
stu
Member
Posts: 923
Joined: Sat Feb 02, 2013 02:51
GitHub: stujones11
Location: United Kingdom

by stu » Post

sfan5 wrote:
stu wrote:It looks like that will become the 'official' version sometime soon (TM)
Everyone ignores my version :-(
Sorry sfan5, I guess yours IS the official version at the moment. That was the first one I tried but it seemed rather awkward and featureless.
xyz gave us a real head start with something almost playable. Since c55 merged that into his sailfish branch I felt that was the most appropriate choice as a base to work from.
Last edited by stu on Tue Mar 18, 2014 22:20, edited 1 time in total.

MoNTE48
Member
Posts: 321
Joined: Sat Apr 06, 2013 11:58
GitHub: MoNTE48
In-game: MoNTE48
Location: Internet

by MoNTE48 » Post

Stu. I build new version you code... More warning and error on line 70 (build.sh). help me...
MultiCraft Open Sourсe – https://github.com/MultiCraft

Amaz
Member
Posts: 354
Joined: Wed May 08, 2013 08:26
GitHub: Amaz1
IRC: Amaz
In-game: Amaz

by Amaz » Post

How do you build the app from the source code? Just wondering, as I would like to try out some of the builds that don't have an apk file...

User avatar
hoodedice
Member
Posts: 1372
Joined: Sat Jul 06, 2013 06:33
GitHub: hoodedice
IRC: hoodedice
In-game: hoodedice
Location: world
Contact:

by hoodedice » Post

Amaz wrote:How do you build the app from the source code? Just wondering, as I would like to try out some of the builds that don't have an apk file...
http://developer.android.com/tools/building/index.html
7:42 PM - Bauglio: I think if you go to staples you could steal firmware from a fax machine that would run better than win10 does on any platform
7:42 PM - Bauglio: so fudge the stable build
7:43 PM - Bauglio: get the staple build

Amaz
Member
Posts: 354
Joined: Wed May 08, 2013 08:26
GitHub: Amaz1
IRC: Amaz
In-game: Amaz

by Amaz » Post

hoodedice wrote:
Amaz wrote:How do you build the app from the source code? Just wondering, as I would like to try out some of the builds that don't have an apk file...
http://developer.android.com/tools/building/index.html
Thanks!

Post Reply

Who is online

Users browsing this forum: Baidu [Spider] and 1 guest