Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

New Motion #442

Merged
merged 100 commits into from
Sep 19, 2024
Merged

New Motion #442

merged 100 commits into from
Sep 19, 2024

Conversation

pvillacorta
Copy link
Collaborator

@pvillacorta pvillacorta commented Jul 11, 2024

This PR is intended to address #410

Still to be done:

  • Revise and change docstrings
  • Improve tests
  • Benchmarking (simulation times, allocations, accuracy...)
  • Test with flow phantoms
  • Remake phantoms from /examples/2.phantoms:
    • brain_motion
    • contracting_ring
    • flow phantom
  • Define the Nspins parameter for all motion types (I don't know if for this PR) -> EDIT: Now called SpinRange (solves #376)

@rkierulf
Copy link
Collaborator

I think I've found the cause of the Simple and Arbitrary Motion test failures for Metal: in UnitTime.jl, this line is processed incorrectly:

return min.(max.((t .- t_start) ./ (t_end - t_start), zero(T)), oneunit(T))

It seems as if Metal has issues compiling the min and max broadcast together. However, if they are separated then it works as expected:

tmp = max.((t .- t_start) ./ (t_end - t_start), zero(T))
return min.(tmp, oneunit(T))

Making this change fixed the failing tests on my computer. I've also written JuliaGPU/Metal.jl#389 since this appears to be a Metal issue with Julia 1.10.

Copy link

codecov bot commented Aug 1, 2024

Codecov Report

Attention: Patch coverage is 80.33827% with 93 lines in your changes missing coverage. Please review.

Project coverage is 89.86%. Comparing base (4862b40) to head (e416800).
Report is 101 commits behind head on master.

Files with missing lines Patch % Lines
KomaMRIPlots/src/ui/DisplayFunctions.jl 75.90% 20 Missing ⚠️
KomaMRIBase/src/motion/motionlist/SpinSpan.jl 26.08% 17 Missing ⚠️
KomaMRIBase/src/motion/nomotion/NoMotion.jl 23.80% 16 Missing ⚠️
KomaMRIFiles/src/Phantom/Phantom.jl 85.29% 10 Missing ⚠️
KomaMRIBase/src/motion/motionlist/MotionList.jl 88.23% 8 Missing ⚠️
KomaMRIBase/src/datatypes/Phantom.jl 76.47% 4 Missing ⚠️
KomaMRIBase/src/motion/motionlist/TimeSpan.jl 78.94% 4 Missing ⚠️
...e/src/motion/motionlist/actions/ArbitraryAction.jl 89.47% 4 Missing ⚠️
KomaMRIBase/src/motion/motionlist/Motion.jl 85.00% 3 Missing ⚠️
.../motion/motionlist/actions/simpleactions/Rotate.jl 86.36% 3 Missing ⚠️
... and 2 more
Additional details and impacted files

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #442      +/-   ##
==========================================
- Coverage   90.93%   89.86%   -1.07%     
==========================================
  Files          53       54       +1     
  Lines        2934     3050     +116     
==========================================
+ Hits         2668     2741      +73     
- Misses        266      309      +43     
Flag Coverage Δ
base 86.24% <76.95%> (-1.96%) ⬇️
core 93.07% <98.11%> (+0.45%) ⬆️
files 91.69% <85.29%> (-1.86%) ⬇️
komamri 93.98% <ø> (ø)
plots 88.43% <75.90%> (-0.88%) ⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines Coverage Δ
KomaMRIBase/src/KomaMRIBase.jl 100.00% <ø> (ø)
KomaMRIBase/src/motion/motionlist/Action.jl 100.00% <100.00%> (ø)
...Base/src/motion/motionlist/actions/SimpleAction.jl 100.00% <100.00%> (ø)
...tion/motionlist/actions/simpleactions/HeartBeat.jl 100.00% <100.00%> (ø)
KomaMRICore/ext/KomaoneAPIExt.jl 92.59% <ø> (ø)
KomaMRICore/src/KomaMRICore.jl 100.00% <ø> (ø)
KomaMRICore/src/simulation/Flow.jl 100.00% <100.00%> (ø)
KomaMRICore/src/simulation/Functors.jl 91.30% <100.00%> (+5.59%) ⬆️
...RICore/src/simulation/SimMethods/Bloch/BlochCPU.jl 100.00% <100.00%> (ø)
...RICore/src/simulation/SimMethods/Bloch/BlochGPU.jl 100.00% <100.00%> (ø)
... and 15 more

... and 4 files with indirect coverage changes

Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

KomaMRI Benchmarks

Benchmark suite Current: e416800 Previous: a26d598 Ratio
MRI Lab/Bloch/CPU/2 thread(s) 226540521 ns 224991845.5 ns 1.01
MRI Lab/Bloch/CPU/4 thread(s) 137009550 ns 174863887 ns 0.78
MRI Lab/Bloch/CPU/8 thread(s) 143972736.5 ns 90572834 ns 1.59
MRI Lab/Bloch/CPU/1 thread(s) 414395326 ns 347400444 ns 1.19
MRI Lab/Bloch/GPU/CUDA 56269762.5 ns 57092571.5 ns 0.99
MRI Lab/Bloch/GPU/oneAPI 500507813 ns 522866366 ns 0.96
MRI Lab/Bloch/GPU/Metal 551246770.5 ns 568303458 ns 0.97
MRI Lab/Bloch/GPU/AMDGPU 34821151 ns 36981171 ns 0.94
Slice Selection 3D/Bloch/CPU/2 thread(s) 1019622442 ns 1151416902 ns 0.89
Slice Selection 3D/Bloch/CPU/4 thread(s) 624166270 ns 580233533 ns 1.08
Slice Selection 3D/Bloch/CPU/8 thread(s) 390068757.5 ns 341164837 ns 1.14
Slice Selection 3D/Bloch/CPU/1 thread(s) 2262459750.5 ns 1930414196.5 ns 1.17
Slice Selection 3D/Bloch/GPU/CUDA 100820723.5 ns 101438582.5 ns 0.99
Slice Selection 3D/Bloch/GPU/oneAPI 640338956 ns 636188156.5 ns 1.01
Slice Selection 3D/Bloch/GPU/Metal 554214166.5 ns 565478250 ns 0.98
Slice Selection 3D/Bloch/GPU/AMDGPU 59085335 ns 60929383 ns 0.97

This comment was automatically generated by workflow using github-action-benchmark.

@cncastillo cncastillo added the documentation Improvements to docs., it also triggers doc preview label Aug 1, 2024
@cncastillo
Copy link
Member

You can preview the docs here (when they run):
https://juliahealth.org/KomaMRI.jl/previews/PR442

Copy link
Member

@cncastillo cncastillo left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I left some minor comments, most of them are renaming stuff. The documentation part can be a different PR, but we may want to create an issue so we don't forget. At least let's make the CI happy.

Also, I'm not sure if writing/reading phantoms work. I assume it does as it is passing the tests.

I am not sure yet about the best way to specific to which part of the object the motion applies, it would be better for another PR.

Did you try Bloch and BlochSimple?

We may want to add a benchmark that includes motion.

KomaMRIBase/src/datatypes/phantom/Motion.jl Outdated Show resolved Hide resolved
KomaMRIBase/src/datatypes/phantom/Motion.jl Outdated Show resolved Hide resolved
KomaMRIBase/test/runtests.jl Outdated Show resolved Hide resolved
@cncastillo
Copy link
Member

cncastillo commented Aug 6, 2024

Btw, the min(max) problem was already fixed in GPUCompiler.jl (JuliaGPU/GPUCompiler.jl#602), thanks to Ryan's report. So this is no longer necessary:

tmp = max.((t .- t_start) ./ (t_end - t_start), zero(T))
return min.(tmp, oneunit(T))

@pvillacorta
Copy link
Collaborator Author

Btw, the min(max) problem was already fixed in GPUCompiler.jl (JuliaGPU/GPUCompiler.jl#602), thanks to Ryan's report. So this is no longer necessary:

tmp = max.((t .- t_start) ./ (t_end - t_start), zero(T))
return min.(tmp, oneunit(T))

How can I apply those changes? Tests are still failing if I don't define the tmp variable

@cncastillo
Copy link
Member

cncastillo commented Aug 9, 2024

Nothing, the latest version of all the packages will be used. If the latest version of GPUCompiler is not being used (v0.27.0) that means there are some compatibility issues, and the Project.toml of KomaMRICore needs to be updated.

In the CI there are some errors due to renaming

LoadError: UndefVarError: `AbstractMotionList` not defined

@cncastillo
Copy link
Member

cncastillo commented Aug 9, 2024

Metal and oneAPI changed their compact to Julia 1.10 (next LTS), so we may want to do the same. If the problems are in Julia 1.9 that could be the reason.

Edit: metal has not registered a new version compatible with GPUcompiler 0.27 yet, so you can leave that part as it was for now.

@pvillacorta
Copy link
Collaborator Author

oneAPI is now the one that fails. I should revise it

@pvillacorta pvillacorta linked an issue Sep 16, 2024 that may be closed by this pull request
KomaMRICore/src/simulation/Flow.jl Show resolved Hide resolved
KomaMRICore/src/simulation/Flow.jl Outdated Show resolved Hide resolved
KomaMRICore/test/runtests.jl Show resolved Hide resolved
KomaMRICore/src/simulation/Flow.jl Outdated Show resolved Hide resolved
@cncastillo
Copy link
Member

cncastillo commented Sep 17, 2024

Also @pvillacorta, I just remembered, but there needs to be a pretty important change in how the Phantoms are plotted. Please dispatch based on Phantom{Motion} or something similar to change plot_phantom_map, as we need to use plot_koma instead of Plot, as the last one breaks compatibility with VSCode and Pluto. This is more important than all the other comments.

@pvillacorta
Copy link
Collaborator Author

pvillacorta commented Sep 17, 2024

Also @pvillacorta, I just remembered, but there needs to be a pretty important change in how the Phantoms are plotted. Please dispatch based on Phantom{Motion} or something similar to change plot_phantom_map, as we need to use plot_koma instead of Plot, as the last one breaks compatibility with VSCode and Pluto. This is more important than all the other comments.

@cncastillo Okay, I will take this into account. I remember when I tried to solve this problem with the Plot and plot_koma: the main problem was the time slider, which somehow did not work with the plot function (the one called by koma_plot), and only worked with Plot. So there is two options for now:

  1. Remove the slider. The plot will still be able to move (in fact, I am not 100% sure about this), but it would not be possible to control the time.
  2. Dispatch. Use plot_koma for phantoms with NoMotion, and Plot for dynamic phantoms.

In terms of code stability, option 1 is better, with the disadvantage that we would lose the slider...
Tell me what you think

@cncastillo
Copy link
Member

I prefer to have a separate method for plot_phantom_map if Phantom{Motion} has motion, so we don't loose the slider.

There will be code repetition, unless we refactor the functions, or make the slider work. For now this is the easiest way, we can clean it up later.

@cncastillo
Copy link
Member

CUDA seems to be segfaulting on BuildKite, but I don't think that is our fault, so I re-run that job to see if it passes, also for Github's CI (Julia 1 / Ubuntu).

@cncastillo
Copy link
Member

cncastillo commented Sep 18, 2024

@pvillacorta The CI is saying this

┌ KomaMRICore
│  WARNING: method definition for #outflow_spin_reset!#54 at /var/lib/buildkite-agent/builds/gpuci-15/julialang/komamri-dot-jl/KomaMRICore/src/simulation/Flow.jl:18 declares type variable T but does not use it.

Also, the CodeCov is failing because there's no test for the new plot_phantom_map with motion.

@pvillacorta
Copy link
Collaborator Author

pvillacorta commented Sep 18, 2024

Yep, I removed it in the commits where I tried to make spin_reset a Matrix{Bool}, but I am not pushing them and I forgot to change that.

Everything passes now, except one random test for oneAPI in buildkite. It is due to a @view call inside get_spin_coords, but it never happened before and there is nothing I have changed in the latest commits that could make it fail now. @cncastillo can you re-run that test?

commit 3662bfbcaf07669e658d0b5a9d0b000bc9b3d748
Author: Pablo Villacorta <[email protected]>
Date:   Thu Sep 19 10:46:20 2024 +0200

    Modify `naive_cumsum!` kernel (@rkierulf)

commit 88e63e42aea4d3fe1d283b1596a94177a81aa874
Author: Pablo Villacorta Aylagas <[email protected]>
Date:   Wed Sep 18 19:25:20 2024 +0200

    Change ArbitraryAction.jl to allow for using scalars as query times

commit 8f636623b09c05cef15f1a9cc0eaa5138ab28f88
Author: Pablo Villacorta Aylagas <[email protected]>
Date:   Wed Sep 18 19:07:16 2024 +0200

    Allow interpolation nodes and coefficients to have different eltype
@cncastillo
Copy link
Member

cncastillo commented Sep 19, 2024

It looks like a proper error, but I will re-run it.

EDIT: It may have been a fluke, as I realized that we are running Julia 1.10 twice now, as 1.11 is not out yet.

examples/3.tutorials/lit-05-SimpleMotion.jl Outdated Show resolved Hide resolved
examples/3.tutorials/lit-05-SimpleMotion.jl Outdated Show resolved Hide resolved
KomaMRICore/src/simulation/Functors.jl Outdated Show resolved Hide resolved
@pvillacorta pvillacorta linked an issue Sep 19, 2024 that may be closed by this pull request
@cncastillo
Copy link
Member

cncastillo commented Sep 19, 2024

I am a little worried about this @pvillacorta

Benchmark suite Current Previous Ratio
MRI Lab/Bloch/CPU/8 thread(s) 143972736.5 ns 90572834 ns 1.59

There seems to be a performance regression. I will merge it anyway, but if this is still the case (after the benchmarks are performed in master), I would need to ask you to investigate.

@cncastillo cncastillo merged commit 927de27 into master Sep 19, 2024
20 checks passed
@cncastillo cncastillo deleted the new-motion branch September 19, 2024 17:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
documentation Improvements to docs., it also triggers doc preview
Projects
None yet
3 participants