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

Accelerating MultiLayerQG on GPUs #373

Open
wants to merge 13 commits into
base: main
Choose a base branch
from

Conversation

mpudig
Copy link
Collaborator

@mpudig mpudig commented Oct 28, 2024

This pull request addresses accelerating the PV-stream function inversion in MultiLayerQG for arbitrary layers on a GPU, as discussed here.

I've used KernelAbstractions to optimize what used to be a loop over all (x, y). There is still a loop over number of layers squared. These changes greatly accelerate the code for certain set-ups with more than two layers based on some simple tests (seen here).

@navidcy
Copy link
Member

navidcy commented Oct 28, 2024

We should bump a patch release.

Comment on lines 116 to 118
# if dev == GPU() && nlayers > 2
# @warn """MultiLayerQG module is not optimized on the GPU yet for configurations with
# 3 fluid layers or more!
Copy link
Member

Choose a reason for hiding this comment

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

Why comment these out? Delete?

S, nlayers = params.S, params.nlayers
kernel!(qh, ψh, S, nlayers)

# This will ensure that no other operations occur until the kernel has finished
Copy link
Member

@navidcy navidcy Oct 28, 2024

Choose a reason for hiding this comment

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

Suggested change
# This will ensure that no other operations occur until the kernel has finished
# Ensure that no other operations occur until the kernel has finished

Comment on lines 626 to 645
"""
@kernel function streamfunctionfrompv_kernel!(ψh, qh, S⁻¹, nlayers)

Kernel for GPU acceleration of streamfunction from PV calculation, i.e., invert the PV to obtain
the Fourier transform of the streamfunction `ψh` in each layer from `qh` using `ψh = params.S⁻¹ qh`.
"""
@kernel function streamfunctionfrompv_kernel!(ψh, qh, S⁻¹, nlayers)
i, j = @index(Global, NTuple)

@unroll for k = 1:nlayers

@inbounds ψh[i, j, k] = 0

@unroll for m = 1:nlayers
@inbounds ψh[i, j, k] += S⁻¹[i, j][k, m] * qh[i, j, m]
end

end
end

Copy link
Member

Choose a reason for hiding this comment

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

are the two kernel functions identical except the order they want their arguments?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Looks like it yes. It'd probably make sense to write one kernel in the more general form.

Copy link
Member

Choose a reason for hiding this comment

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

I think so! Would make the code more robust.

Comment on lines 547 to 551
@unroll for k = 1:nlayers

@inbounds qh[i, j, k] = 0

@unroll for m = 1:nlayers
Copy link
Member

Choose a reason for hiding this comment

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

The @unroll don't do anything unless nlayers is known at compile time (this requires using Val, but I don't know if it will speed anything up... it might).

Kernel for GPU acceleration of PV from streamfunction calculation, i.e., obtaining the Fourier
transform of the PV from the streamfunction `ψh` in each layer using `qh = params.S * ψh`.
"""
@kernel function pvfromstreamfunction_kernel!(qh, ψh, S, nlayers)
Copy link
Member

Choose a reason for hiding this comment

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

Suggested change
@kernel function pvfromstreamfunction_kernel!(qh, ψh, S, nlayers)
@kernel function pvfromstreamfunction_kernel!(qh, ψh, S, ::Val{nlayers}) where nlayers

for @unroll you have to do this, and also pass Val(nlayers) rather than nlayers into the kernel when launching it. I don't know if it will speed things up though. It might.

Comment on lines +533 to +545
@kernel function PVinversion_kernel!(a, b, M, ::Val{nlayers}) where nlayers
i, j = @index(Global, NTuple)

@unroll for k = 1:nlayers

@inbounds a[i, j, k] = 0

@unroll for m = 1:nlayers
@inbounds a[i, j, k] += M[i, j][k, m] * b[i, j, m]
end

end
end
Copy link
Collaborator Author

Choose a reason for hiding this comment

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

I rewrote the kernel in more general form and added Val. The code has sped up slightly, but the 16-thread CPU still outperforms the GPU. Compare these benchmarks to what I showed here

  • GPU
    nlayers = 12; nx = 512; prob = MultiLayerQG.Problem(nlayers, GPU(); nx); @btime stepforward!(prob) 668.165 ms (2533 allocations: 191.19 KiB)

  • CPU with 16 threads
    nlayers = 12; nx = 512; prob = MultiLayerQG.Problem(nlayers, CPU(); nx); @btime stepforward!(prob) 444.419 ms (113 allocations: 5.61 KiB)

Copy link
Member

Choose a reason for hiding this comment

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

Are you sure you are timing the GPU properly?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

julia> nlayers = 12; nx = 512; prob = MultiLayerQG.Problem(nlayers, GPU(); nx); @benchmark CUDA.@sync CUDA.@time stepforward!(prob)

0.681338 seconds (57.95 k CPU allocations: 2.419 MiB) (18 GPU allocations: 345.250 MiB, 0.04% memmgmt time)
0.678481 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.676472 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.694825 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.678072 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.677693 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.678237 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.04% memmgmt time)
0.677198 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.676980 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.676189 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.678326 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.677010 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.677142 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.04% memmgmt time)
0.676321 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.04% memmgmt time)
0.678115 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.677461 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.678255 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.677168 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
0.677529 seconds (2.58 k CPU allocations: 194.141 KiB) (16 GPU allocations: 297.156 MiB, 0.03% memmgmt time)
BenchmarkTools.Trial: 8 samples with 1 evaluation.
Range (min … max):  676.718 ms … 678.645 ms  ┊ GC (min … max): 0.00% … 0.00%
Time  (median):     677.681 ms               ┊ GC (median):    0.00%
Time  (mean ± σ):   677.765 ms ± 618.941 μs  ┊ GC (mean ± σ):  0.00% ± 0.00%

█                     █  ██       █   █                 █   █  
█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁██▁▁▁▁▁▁▁█▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁█ ▁
677 ms           Histogram: frequency by time          679 ms <

Memory estimate: 199.41 KiB, allocs estimate: 2660.

Seems to be roughly the same as above? Unless I'm misunderstanding what this benchmark is doing...

Copy link
Member

Choose a reason for hiding this comment

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

That's disappointing...

The first thing I would try to figure out is whether this function is indeed the bottleneck. It might be better, infact, to simply benchmark this function in isolation.

I don't know if it matters but I saw the workgroup is 8, 8. Usually we use 16, 16.

I would also check 3 layers first perhaps. The double inner loop gets slower with more layers, and perhaps the computational costs scale differently on CPU vs GPU. That might give a clue.

The loop is over k, m --- which are the slowest (last) indices in a and b. That could be an issue. If you can benchmark this operation in isolation, then you can experiment with new arrays where k, m are the fastest / first indices in a and b. This experiment would tell you what kind of slow down that's incurring.

Maybe the @unroll is not working for some reason. When I've seen stuff like this before, people have used matrix / linear algebra via StaticArrays, rather than explicit loops as used here. If you are just testing the kernel in isolation, transforming a, b to arrays of StaticVector could also be something to experiment with.

In general with performance engineering, one has to really be persistent and creative and test test test. To make this easier you want to extract out the function you're trying to optimize and work with a very idealized test case that also allows you to change the data structure (rather than working with a FourierFlows script). Think of this as research. If you find that you need to rearrange memory differently, then we can come back to GeophysicalFlows and see whether that is feasible or not.

@navidcy
Copy link
Member

navidcy commented Nov 17, 2024

@mpudig any idea why CI breaks?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants