Android tests

[xoft]

http://mc-server.xoft.cz/Minecraft%20Old%20Versions/
Soon enough. DL the appropriate file, rename to minecraft.jar and replace the one in %APPDATA%\.minecraft\bin

[NiLSPACE]

no problem

[FakeTruth]

Oh yes! It works!


Half flat, half fancy terrain! awesome



Why hello there cow!! (and flaming squid)

[NiLSPACE]

you still use the marble caves that uses much world generation speed:P

[xoft]

Actually on ARM processors MarbleCaves are not so bad, they take about 8% of the total runtime, unlike on x86, where they take about 50 %. ARM can calculate cNoise really fast (because it has many registers), but it's too slow accesing the memory, that's why lighting is such a slow hog on ARM.

[NiLSPACE]

but worm-nest is still faster?

[xoft]

I'm not sure, haven't tested. I'd say they are comparable in speed.

[FakeTruth]

Well with flat world I get 150ch/s lighting is not so fast, I'd say about 50 chunks per second

[NiLSPACE]

can't wait to try MCS on my Wildfire S

---

i just found out that classic HeightGen is about as fast as Flat Heightgen.
i had 2 world.ini files one for flat and one classic.

world.ini Flat chunks per sec=369.00

[Generator]

BiomeGen=Constant
ConstantBiome=Plains
HeightGen=Flat
FlatHeight=32
CompositionGen=Classic
ClassicSeaLevel=0
ClassicBeachHeight=-1
Structures=Ravines,Wormnestcaves
Finishers=
BiomeGenCacheSize=0
HeightGenCacheSize=0

world.ini classic chunks per sec=344.23

[Generator]
BiomeGen=Constant
ConstantBiome=Plains
HeightGen=Classic
FlatHeight=32
CompositionGen=Classic
ClassicSeaLevel=0
ClassicBeachHeight=-1
Structures=Ravines,Wormnestcaves
Finishers=
BiomeGenCacheSize=0
HeightGenCacheSize=0
ClassicBeachDepth=4
ClassicBlockTop=grass
ClassicBlockMiddle=dirt
ClassicBlockBottom=stone
ClassicBlockBeach=sand
ClassicBlockBeachBottom=sandstone
ClassicBlockSea=9
ClassicHeightFreq1=0.100000
ClassicHeightFreq2=1.000000
ClassicHeightFreq3=2.000000
ClassicHeightAmp1=1.000000
ClassicHeightAmp2=0.500000
ClassicHeightAmp3=0.500000

i think classic heightgen is usefull to use on the android devices.

[xoft]

You can't compare the performance on an ARM to performance on an Intel. The two have a very different architecture, what works well for one doesn't work that well for the other. So until you test on the actual hardware, there's no way to tell what's performing well and what not.
My guess is that DistortedVoronoi Biomes with Biomal heightgen and compositiongen will be the least of your problems, since the ARM has enough registers to calculate the underlying cNoise very fast.

Here's my profiling results from running on a RaspberryPi, which is an ARM-based device:

Code:

%   cumulative   self              self     total          
time   seconds   seconds    calls   s/call   s/call  name    
17.84     10.22    10.22      398     0.03     0.03  cLightingThread::PrepareSkyLight()
12.50     17.38     7.16      882     0.01     0.02  deflate_slow
10.77     23.55     6.17  1217115     0.00     0.00  cCaveTunnel::ProcessChunk()
10.10     29.34     5.79    15968     0.00     0.00  fill_window
  8.38     34.14     4.80  1345982     0.00     0.00  longest_match
  8.25     38.87     4.73      422     0.01     0.04  cLightingThread::LightChunk()
  5.85     42.22     3.35   370359     0.00     0.00  cHeiGenBiomal::GetHeightAt()
  5.69     45.48     3.26     8901     0.00     0.00  cLightingThread::CalcLightStep()
  4.08     47.82     2.34      484     0.00     0.00  cChunk::CreateBlockEntities()
  2.34     49.16     1.34      478     0.00     0.02  cStructGenTrees::GenStructures()
  2.09     50.36     1.20  1775634     0.00     0.00  cNoise::CubicNoise2D() const
  1.75     51.36     1.00     6123     0.00     0.00  adler32
  1.66     52.31     0.95   726731     0.00     0.00  cBioGenVoronoi::VoronoiBiome()
  1.43     53.13     0.82   426000     0.00     0.00  cNoise::CubicNoise3D() const
  1.20     53.82     0.69     1814     0.00     0.00  cIniFile::FindValue() const

The results on any ARM-based Android shouldn't be too far from these. And as you can see, the narrowest bottlenecks are lighting, compression and wormnest caves; all the biome-related code ran in under 6 %.