Multithreading support and examples

[filipinascimento]

These examples illustrate how to interact with fury using two different threads.

Offscreen example uses the main thread to render the scene, while a second thread keeps saving a snapshot at random intervals.

The other example opens a window in the main thread and updates the scene form a second thread. Note that the iren loop had to be done manually (showm.start() will not work in this case).

This is a work in progress PR.

[pep8speaks]

Hello @filipinascimento! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! :beers:

Comment last updated at 2022-06-12 04:53:08 UTC

[codecov[bot]]

Codecov Report

Merging #502 (e7263d7) into master (add8cc0) will decrease coverage by 0.02%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #502      +/-   ##
==========================================
- Coverage   87.95%   87.92%   -0.03%     
==========================================
  Files          62       63       +1     
  Lines       13030    13201     +171     
  Branches     1308     1330      +22     
==========================================
+ Hits        11460    11607     +147     
- Misses       1198     1212      +14     
- Partials      372      382      +10     

Impacted Files	Coverage Δ
fury/fury/window.py	82.32% <0.00%> (-0.89%)	:arrow_down:
fury/fury/actor.py	88.28% <0.00%> (-0.39%)	:arrow_down:
fury/fury/tests/test_actors.py	91.40% <0.00%> (-0.27%)	:arrow_down:
fury/fury/material.py	74.60% <0.00%> (ø)
fury/fury/actors/peak.py	94.25% <0.00%> (ø)
fury/fury/tests/test_thread.py	91.42% <0.00%> (ø)
fury/fury/shaders/tests/test_base.py	93.08% <0.00%> (+0.03%)	:arrow_up:
fury/fury/shaders/base.py	92.92% <0.00%> (+0.14%)	:arrow_up:
fury/fury/utils.py	91.89% <0.00%> (+0.22%)	:arrow_up:
fury/fury/tests/test_utils.py	94.74% <0.00%> (+0.47%)	:arrow_up:

[Garyfallidis]

@filipinascimento the CI fails are not relevant to your code. Can you remove all unrelated or old examples from this PR?

[Garyfallidis]

Ping @filipinascimento ! It would be great to have this merged asap. Please make the required changes.

[filipinascimento]

I'm having some trouble with the test on the mac. It seems like something from vtk is running in another thread. The problem seems to be caused because snapshot is being called outside of the main thread. I have an example of offscreen rendering that works so I'm figuring it out what is different.

On Feb 9, 2022, at 1:06 PM, Eleftherios Garyfallidis @.***> wrote:

Ping @filipinascimento https://github.com/filipinascimento ! It would be great to have this merged asap. Please make the required changes.

— Reply to this email directly, view it on GitHub https://github.com/fury-gl/fury/pull/502#issuecomment-1034048454, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAESXHJMYFTHKBN2REPVLMDU2KUKZANCNFSM5EVIRVKQ. You are receiving this because you were mentioned.

[Garyfallidis]

Ping @filipinascimento !!! Let's get this PR to move forward! TY!

[filipinascimento]

Hi @filipinascimento,

See below for some comments. The tests are failing on macOS but it works on my local machine. We need to figure out what is the issue. the error is illegal instruction 4 which is usually due to an architecture issue. The operating system in our CIs is macOS 11.6.5 with:

• Included Software: https://github.com/actions/virtual-environments/blob/macOS-11/20220508.1/images/macos/macos-11-Readme.md
• Image Release: https://github.com/actions/virtual-environments/releases/tag/macOS-11%2F20220508.1 maybe you can try via a virtual machine.

Thank you in advance for your feedback

Not sure how to address this problem. It happens in the threading.py when calling thread.run(), maybe related to https://github.com/tqdm/tqdm/issues/1037 . I'm not sure why threading.py would have any illegal instructions, unless that macOS VM runs in unsupported CPU microarchitecture. It is not clear which CPU model they are using in the macOS VMs.

Edit: My mistake, the problem is happening when snapshot is being called from another thread. Need more time to investigate the issue. Problem is that I have no access to an intel mac to test the macOS VM.

[m-agour]

Hello @filipinascimento, About the FPS issue not being accurate while using threads, it's as we suspected an OS related problem. Linux kernels have minimal sleep interval down to 1ms. But it's higher in windows (14ms). This got fixed in Python 3.11 and the time.sleep() has the same accuracy regarding the OS, but since we need to maintain older versions and I think VTK is yet to be ported to python 3.11, we really can't make use of that. A way to fix this for Windows so is setting the minimum timer resolution to 1ms:

import ctypes
import sys

if sys.platform == 'win32' and sys.version_info[1] < 11:
    winmm = ctypes.WinDLL('winmm')
    winmm.timeBeginPeriod(1)

This fixed the problem for me.

[filipinascimento]

Hello @filipinascimento, About the FPS issue not being accurate while using threads, it's as we suspected an OS related problem. Linux kernels have minimal sleep interval down to 1ms. But it's higher in windows (14ms). This got fixed in Python 3.11 and the time.sleep() has the same accuracy regarding the OS, but since we need to maintain older versions and I think VTK is yet to be ported to python 3.11, we really can't make use of that. A way to fix this for Windows so is setting the minimum timer resolution to 1ms:

import ctypes
import sys

if sys.platform == 'win32' and sys.version_info[1] < 11:
    winmm = ctypes.WinDLL('winmm')
    winmm.timeBeginPeriod(1)

This fixed the problem for me.

Thanks, this may work as a temporary solution.