Pass Discrete Sequence to GPU

[mcgrathcm]

In preparation for higher order sequences, this adds:

• Discrete sequences adapted and passed directly to GPU kernels
• Spatial gradient Bz term moved to its own helper function
• AbstractDiscreteSequence type added to allow dispatch when higher order sequences are added

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.02%. Comparing base (3a2aad0) to head (eeaeafc).
⚠️ Report is 5 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #652      +/-   ##
==========================================
+ Coverage   89.66%   90.02%   +0.35%     
==========================================
  Files          74       75       +1     
  Lines        5179     5181       +2     
==========================================
+ Hits         4644     4664      +20     
+ Misses        535      517      -18     

Flag	Coverage Δ
base	86.48% <100.00%> (ø)
core	90.67% <100.00%> (+2.65%)	⬆️
files	95.26% <ø> (ø)
komamri	88.23% <ø> (ø)
plots	90.88% <ø> (ø)

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	85.71% <ø> (ø)
...IBase/src/datatypes/simulation/DiscreteSequence.jl	93.10% <100.00%> (ø)
KomaMRICore/src/KomaMRICore.jl	100.00% <ø> (ø)
KomaMRICore/src/simulation/Functors.jl	77.27% <ø> (ø)
...re/src/simulation/SimMethods/Bloch/cpu/BlochCPU.jl	100.00% <100.00%> (ø)
...re/src/simulation/SimMethods/Bloch/gpu/BlochGPU.jl	100.00% <ø> (ø)
...e/src/simulation/SimMethods/BlochDict/BlochDict.jl	96.96% <ø> (ø)
...ation/SimMethods/BlochMagnus/cpu/BlochMagnusCPU.jl	100.00% <ø> (ø)
...c/simulation/SimMethods/BlochSimple/BlochSimple.jl	100.00% <ø> (ø)
KomaMRICore/src/simulation/SimulatorCore.jl	90.69% <ø> (ø)
... and 1 more

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

KomaMRI Benchmarks

Details

Benchmark suite	Current: 17e586e	Previous: 11d3180	Ratio
MRI Lab/Bloch/CPU/2 thread(s)	387199837.5 ns	337633469 ns	1.15
MRI Lab/Bloch/CPU/4 thread(s)	289067546 ns	276282606 ns	1.05
MRI Lab/Bloch/CPU/8 thread(s)	242467570.5 ns	209852165 ns	1.16
MRI Lab/Bloch/CPU/1 thread(s)	655549193 ns	555065232 ns	1.18
MRI Lab/Bloch/GPU/CUDA	20962776 ns	21231134 ns	0.99
MRI Lab/Bloch/GPU/oneAPI	85944808 ns	77053690 ns	1.12
MRI Lab/Bloch/GPU/Metal	104643083 ns	95540333 ns	1.10
MRI Lab/Bloch/GPU/AMDGPU	25102288 ns	24763685 ns	1.01
Slice Selection 3D/Bloch/CPU/2 thread(s)	1651013192 ns	1592087059 ns	1.04
Slice Selection 3D/Bloch/CPU/4 thread(s)	911694348 ns	889539516 ns	1.02
Slice Selection 3D/Bloch/CPU/8 thread(s)	599249406.5 ns	565401528.5 ns	1.06
Slice Selection 3D/Bloch/CPU/1 thread(s)	3109371230 ns	3029269071 ns	1.03
Slice Selection 3D/Bloch/GPU/CUDA	31822745 ns	32639703 ns	0.97
Slice Selection 3D/Bloch/GPU/oneAPI	133804680 ns	121489006.5 ns	1.10
Slice Selection 3D/Bloch/GPU/Metal	112492041 ns	111152750 ns	1.01
Slice Selection 3D/Bloch/GPU/AMDGPU	32559879.5 ns	32636098 ns	1.00

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

[cncastillo]

The macos fails are due to actions/runner-images#13341.

[cncastillo]

Compare benchmarks with #653.

Update: macos cache problem fixed.

[cncastillo]

Looks very good, no significant performance regressions, but some of them could be explained by some of the suggestions below.

[cncastillo]

This might generate some problems with Enzyme/Reactant, @KookiesNKareem what do you think?