How to use Incise Flow for rivers in Wilbur correctly?

[waldronate]

One "feature" of the erosion model in Wilbur is that it's a global process. In this sense, it means that each part relies on the other parts of the system and a trivial partitioning (for example, splitting into quarters) followed by processing won't yield the same results as processing of the whole surface.

One of the benefits of using the described multi-scale processing operations is that the global features are established when surface processing is quick. One of the problems with the multi-scale process is that results have some peculiar features that result from upscaling the lower-resolution system. It can also be tough to get a solid handle on how the final surface will look just by looking at the first few processed surfaces in the sequence.

[waldronate]

One other detail that can cause poor performance is having large nearly-flat areas. The amount of time required to find the flow is proportional to the surface size, but also to the average length of rivers. The flow-finder follows the terrain gradient upstream to count cells: the longer the river, the more steps are required. The worst case would be a single river that encompasses the entire surface, as happens with an "ocean" covering the surface when no selection or ocean is used to constrain the computed areas. Precipiton erosion will only happen in the areas inside of a selection, and drainage areas (as used by incise flow) will only be calculated where the land level is above the water level (water level is defined as 0 for this calculation). The selection requirement is why you should select all (or select areas above water level) before using precipiton erosion. Similarly, if your world has an "ocean", you should make sure that parts that you don't want incised via incise flow or filled via fill basins should be below zero.

I mention this because I just noticed that your original screen shots show what is likely "ocean" as green, meaning that it's above sea level. The presence of incised areas in parts that you are likely to discard indicates a whole lot of work being done that's going to be discarded. The nearly-flat character of those areas also means that the work to be discarded will be some of the most expensive work on the whole surface. Lowering your surface until the subsea areas are below zero may speed things up dramatically for you.

Yes, things aren't symmetric from operation to operation. It's what happens during evolution over a long time (Wilbur is more than 20 years old now).

[xeyla]

You were correct in that my screenshots did have the ocean as above sea-level, as per that tutorial I am following. Then I found the trick for selecting just my land areas, which did help some, but the erosion processes still takes time. I did try that "cutting my map into quarters" thought, and the rivers didn't line up on the edges, so I scrapped that copy. My process right now is, 'Precipitation Erosion, to erode the coastline. Then Generate Noise 3%, Fill Basins, repeat Noise 3%, Fill Basins, incise erosion".

Here is the map I am trying to work on:


It's my world map for my Pathfinder game I am currently running. I did want to add more mountain chains, but I was afraid that would look not too realistic? I wanted an ocean that would touch every piece of continent so my players could use ships as the main form of traveling, if they wanted too. I didn't know how rivers would look on the map, hence using Wilbur to make them. And I didn't know Wilbur was 20 years old! Its a fantastic program, in my opinion, and I will be using it a lot for my maps to make them look awesome. I'm still a newbie to it though, but I'm practicing. I will also make note of making my oceans lower than 0 for now on. Thank you for that tip.

[waldronate]

Practice is what it takes to get good at Wilbur.

(I missed any intended inland seas or lakes and the biomes are probably all wrong as well).

[johnvanvliet]

as to the "not responding" issue

how big is the image ? 2048x2048 ? 4096x4096? 8192x8192 ?

and the machine you are using
a desktop ?
a laptop?
a tablet ?

and the system specs -- as in how much ram

i had no real issues on a OLD pentium 4 with 1 gig ram ( a old version of wilbur from 6 to 10 years ago

and i have not had issues with the current 64 bit running in the WINE emulator on a I5 desktop with 8 gig ram

[xeyla]

as to the "not responding" issue

how big is the image ? 2048x2048 ? 4096x4096? 8192x8192 ?

and the machine you are using
a desktop ?
a laptop?
a tablet ?

and the system specs -- as in how much ram

i had no real issues on a OLD pentium 4 with 1 gig ram ( a old version of wilbur from 6 to 10 years ago

and i have not had issues with the current 64 bit running in the WINE emulator on a I5 desktop with 8 gig ram

I have a desktop gaming computer that is 5 years old:
Win 7 Pro 64bit
16gb of RAM
AMD FX-9590 Eight-Core Processor 4.7ghz
NVidia GEForce GTX 770 graphics card with 4gb ram

And my image is 4000x4000 pixels, as pictured in my last post.

[johnvanvliet]

you have twice the ram and twice the cores as i do

You might really want to turn off unneeded back ground services and uninstall any preinstalled crapware and disable those services

a 2048x2048 image takes almost 10 sec to run incise
and 5 sec for the perception

4096x4096

incise = 45 sec
precpt = 22 sec

8192 x 8192

incise = just under 3 min.
and rain 1.5 min.

[xeyla]

I use System Mechanic Pro to turn off my un-needed background services, and I actually don't have any preinstalled crapware. When I built the computer on Cyberpowerpc.com, I selected all the hardware, and the software installed. The only crapware they added was the stupid Norton Antivirus 360, which I promptly uninstalled the day I hooked up my new baby. I never installed the Norton, which I learned was a virus itself from previous experience with my last computer.

So, in light of this post, I really don't know what is causing the hangups. I even upped Wilburs priority to normal, which it was below normal at default. I'm just dealing with it for now, until a solution is found in the future.

[woodb3kmaster]

I've run into similar issues as xeyla on my new-ish refurbished server (Win 10 Pro x64, 2x Intel Xeon E5430 Quad-Cores, 32 GB RAM, GeForce GT 730 with 2 GB VRAM) when trying to erode my 12000x9200 map. My old computer, with far less impressive specs, could run incise flow on it in around 20 minutes; the new computer takes hours to do the same calculation. I, likewise, had no pre-installed crapware, as the hard drives had been wiped. I can only speculate as to why more powerful computers run Wilbur more slowly.

[waldronate]

I would guess that the most likely cause of "more powerful computers running [flow finding] more slowly" is memory latency. The flow finding algorithm probably has a memory access pattern that the memory prefetch system can't predict, causing performance to be largely dictated by random-access memory speed. It would probably get worse on a NUMA system where memory has an extra hop.