Adding MPI test #518
Adding MPI test #518
Conversation
michel2323
force-pushed
the
ms/mpi
branch
3 times, most recently
from
8d782c4 to
b46613f
Compare
michel2323
force-pushed
the
ms/mpi
branch
3 times, most recently
from
409d930 to
b9e0668
Compare
|
Can you rebase once more? |
|
@vchuravy @wsmoses The MPI test crashes now. With 1.9 we have the finalizer issue, and with 1.8 the logs and Manifest.toml are attached. |
michel2323
force-pushed
the
ms/mpi
branch
2 times, most recently
from
0f2cbdb to
6754c67
Compare
|
Can you open an issue with the error message? |
Codecov ReportAll modified and coverable lines are covered by tests ✅
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## main #518 +/- ##
===========================================
+ Coverage 75.39% 93.43% +18.03%
===========================================
Files 35 7 -28
Lines 10653 259 -10394
===========================================
- Hits 8032 242 -7790
+ Misses 2621 17 -2604 ☔ View full report in Codecov by Sentry. |
wsmoses
force-pushed
the
ms/mpi
branch
2 times, most recently
from
0e829b9 to
a1745a5
Compare
|
@michel2323 i rebased this PR with the jll with your fix, if it passes let's merge this! |
|
@vchuravy this still fails presumably for the need for: #669 from While I do agree custom rules are good (and we can redo that API to be a custom global invariant rule, the issue of that MPI.Double being hidden behind a separate julia-specific global int is hindering the analysis here (and also likely would for other libXYZ calls potentially, for example CUDA). |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| buf = @view x[1:1] | ||
| push!(requests, MPI.Isend(x[2:2], MPI.COMM_WORLD; dest=rank-1, tag=0)) | ||
| push!(requests, MPI.Irecv!(buf, MPI.COMM_WORLD; source=rank-1, tag=0)) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| buf = @view x[1:1] | |
| push!(requests, MPI.Isend(x[2:2], MPI.COMM_WORLD; dest=rank-1, tag=0)) | |
| push!(requests, MPI.Irecv!(buf, MPI.COMM_WORLD; source=rank-1, tag=0)) | |
| buf = @view x[1:1] | |
| push!(requests, MPI.Isend(x[2:2], MPI.COMM_WORLD; dest=rank - 1, tag=0)) | |
| push!(requests, MPI.Irecv!(buf, MPI.COMM_WORLD; source=rank - 1, tag=0)) |
| if rank != np-1 | ||
| buf = @view x[end:end] | ||
| push!(requests, MPI.Isend(x[end-1:end-1], MPI.COMM_WORLD; dest=rank+1, tag=0)) | ||
| push!(requests, MPI.Irecv!(buf, MPI.COMM_WORLD; source=rank+1, tag=0)) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| if rank != np-1 | |
| buf = @view x[end:end] | |
| push!(requests, MPI.Isend(x[end-1:end-1], MPI.COMM_WORLD; dest=rank+1, tag=0)) | |
| push!(requests, MPI.Irecv!(buf, MPI.COMM_WORLD; source=rank+1, tag=0)) | |
| if rank != np - 1 | |
| buf = @view x[end:end] | |
| push!(requests, | |
| MPI.Isend(x[(end - 1):(end - 1)], MPI.COMM_WORLD; dest=rank + 1, tag=0)) | |
| push!(requests, MPI.Irecv!(buf, MPI.COMM_WORLD; source=rank + 1, tag=0)) |
| n = np*10 | ||
| n1 = Int(round(rank / np * (n+np))) - rank | ||
| n2 = Int(round((rank + 1) / np * (n+np))) - rank | ||
| nl = rank == 0 ? n1+1 : n1 | ||
| nr = rank == np-1 ? n2-1 : n2 | ||
| nlocal = nr-nl+1 |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| n = np*10 | |
| n1 = Int(round(rank / np * (n+np))) - rank | |
| n2 = Int(round((rank + 1) / np * (n+np))) - rank | |
| nl = rank == 0 ? n1+1 : n1 | |
| nr = rank == np-1 ? n2-1 : n2 | |
| nlocal = nr-nl+1 | |
| n = np * 10 | |
| n1 = Int(round(rank / np * (n + np))) - rank | |
| n2 = Int(round((rank + 1) / np * (n + np))) - rank | |
| nl = rank == 0 ? n1 + 1 : n1 | |
| nr = rank == np - 1 ? n2 - 1 : n2 | |
| nlocal = nr - nl + 1 |
| fill!(context.x, Float64(rank)) | ||
| halo(context) | ||
| if rank != 0 | ||
| @test context.x[1] == Float64(rank-1) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| @test context.x[1] == Float64(rank-1) | |
| @test context.x[1] == Float64(rank - 1) |
| if rank != np-1 | ||
| @test context.x[end] == Float64(rank+1) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| if rank != np-1 | |
| @test context.x[end] == Float64(rank+1) | |
| if rank != np - 1 | |
| @test context.x[end] == Float64(rank + 1) |
| try | ||
| include("mpi.jl") | ||
| mpiexec() do cmd | ||
| run(`$cmd -n 2 $(Base.julia_cmd()) --project=$testdir $testdir/mpi.jl`) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| run(`$cmd -n 2 $(Base.julia_cmd()) --project=$testdir $testdir/mpi.jl`) | |
| return run(`$cmd -n 2 $(Base.julia_cmd()) --project=$testdir $testdir/mpi.jl`) |
| end | ||
| @test mpi_test | ||
| end | ||
|
|
||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
test/runtests.jl
Outdated
| b = Float64[11, 13] | ||
| db = zero(b) | ||
|
|
||
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, Duplicated, Duplicated{typeof(A)}, Duplicated{typeof(b)}) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, Duplicated, Duplicated{typeof(A)}, Duplicated{typeof(b)}) | |
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, | |
| Duplicated, Duplicated{typeof(A)}, | |
| Duplicated{typeof(b)}) |
test/runtests.jl
Outdated
|
|
||
| db = zero(b) | ||
|
|
||
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, Duplicated, Const{typeof(A)}, Duplicated{typeof(b)}) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, Duplicated, Const{typeof(A)}, Duplicated{typeof(b)}) | |
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, | |
| Duplicated, Const{typeof(A)}, | |
| Duplicated{typeof(b)}) |
test/runtests.jl
Outdated
|
|
||
| dA = zero(A) | ||
|
|
||
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, Duplicated, Duplicated{typeof(A)}, Const{typeof(b)}) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, Duplicated, Duplicated{typeof(A)}, Const{typeof(b)}) | |
| forward, pullback = Enzyme.autodiff_thunk(ReverseSplitNoPrimal, Const{typeof(\)}, | |
| Duplicated, Duplicated{typeof(A)}, | |
| Const{typeof(b)}) |
|
I get this now on the most recent Enzyme build: This is the request issue @vchuravy mentioned. How do we proceed with MPI in Julia? Should we start an Enzyme extension for MPI/MPI extension for Enzyme? |
Benchmark Results
Benchmark PlotsA plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR. |
Let me know if you want to keep the trailing whitespaces.