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.

Sign up for GitHub

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

Jump to bottom

Basic SparseMatrixCSC calculations do not work. #1381

Open
ToPo-ToPo-ToPo opened this issue Apr 4, 2024 · 9 comments
Open

Basic SparseMatrixCSC calculations do not work. #1381

ToPo-ToPo-ToPo opened this issue Apr 4, 2024 · 9 comments
Labels
more information needed

Comments

@ToPo-ToPo-ToPo
Copy link

ToPo-ToPo-ToPo commented Apr 4, 2024 •
edited by wsmoses

I would like to use the basic calculations in SparseMatrixCSC, but they do not work.
The following is a simple SparseMatrixCSC-to-SparseMatrixCSC addition program.

using Enzyme
using LinearAlgebra
using SparseArrays
#--------------------------------------------------------------------------------
# Function to be differentiated
#--------------------------------------------------------------------------------
function eval(x::Vector{Float64})
    A = sparse([1, 1, 2, 3], [1, 3, 2, 3], [2.0*x[2]^3.0, 1.0-x[1], 2.0+x[3], -1.0])
    B = sparse([1, 1, 2, 3], [1, 3, 2, 3], [2.0*x[2], 1.0-x[1], 2.0+x[3], -1.0])
    C = A + B
    return C[1, 1] + C[2, 2]
end
#--------------------------------------------------------------------------------
# main
#--------------------------------------------------------------------------------
x = rand(Float64, 3)
dx = [0.0, 0.0, 0.0]

val = eval(x)

# compute gradient
autodiff(Reverse, eval, Duplicated(x, dx))
@show dx 

df_numeric = similar(x)
for i in eachindex(x)
    x_tmp = x[i]
    x_p = copy(x)
    x_p[i] += 1.0e-05
    x_m = copy(x)
    x_m[i] -= 1.0e-05
    df_numeric[i] = (eval(x_p) - eval(x_m)) / (2.0 * 1.0e-05)
end

display(df_numeric)

error = norm(dx-df_numeric)
println("error = $(error)")

The errors that occur when executing the program are as follows:

ERROR: BoundsError: attempt to access 4-element Vector{Int64} at index [0]
Stacktrace:
  [1] _deleteend!
    @ ./array.jl:1081 [inlined]
  [2] resize!
    @ ./array.jl:1320 [inlined]
  [3] trimstorage!
    @ ~/.julia/juliaup/julia-1.10.2+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:135
  [4] _map_zeropres!
    @ ~/.julia/juliaup/julia-1.10.2+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:361
  [5] _noshapecheck_map
    @ ~/.julia/juliaup/julia-1.10.2+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:184
  [6] map
    @ ~/.julia/juliaup/julia-1.10.2+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:1187 [inlined]
  [7] +
    @ ~/.julia/juliaup/julia-1.10.2+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/sparsematrix.jl:2242 [inlined]
  [8] eval
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:11 [inlined]
  [9] diffejulia_eval_2011wrap
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:0
 [10] macro expansion
    @ ~/.julia/packages/Enzyme/l4FS0/src/compiler.jl:5378 [inlined]
 [11] enzyme_call
    @ ~/.julia/packages/Enzyme/l4FS0/src/compiler.jl:5056 [inlined]
 [12] CombinedAdjointThunk
    @ ~/.julia/packages/Enzyme/l4FS0/src/compiler.jl:4998 [inlined]
 [13] autodiff
    @ ~/.julia/packages/Enzyme/l4FS0/src/Enzyme.jl:215 [inlined]
 [14] autodiff
    @ ~/.julia/packages/Enzyme/l4FS0/src/Enzyme.jl:238 [inlined]
 [15] autodiff(mode::ReverseMode{false, FFIABI}, f::typeof(eval), args::Duplicated{Vector{Float64}})
    @ Enzyme ~/.julia/packages/Enzyme/l4FS0/src/Enzyme.jl:224
 [16] top-level scope
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:23
@wsmoses
Copy link
Member

wsmoses commented Apr 20, 2024

Should now be fixed on main

@wsmoses wsmoses closed this as completed Apr 20, 2024
@ToPo-ToPo-ToPo
Copy link
Author

I have been told that this has been supported, but in my environment, the following error occurred and the calculation could not be performed.
Enzyme.jl: Enzyme v0.12.7

┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/kqxyC/src/utils.jl:59
ERROR: BoundsError: attempt to access 4-element Vector{Int64} at index [0]
Stacktrace:
  [1] _deleteend!
    @ ./array.jl:1081 [inlined]
  [2] resize!
    @ ./array.jl:1320 [inlined]
  [3] trimstorage!
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:135
  [4] _map_zeropres!
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:361
  [5] _noshapecheck_map
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:184
  [6] map
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:1187 [inlined]
  [7] +
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/sparsematrix.jl:2242 [inlined]
  [8] eval
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:11 [inlined]
  [9] diffejulia_eval_6074wrap
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:0
 [10] macro expansion
    @ ~/.julia/packages/Enzyme/UKzrf/src/compiler.jl:5787 [inlined]
 [11] enzyme_call
    @ ~/.julia/packages/Enzyme/UKzrf/src/compiler.jl:5453 [inlined]
 [12] CombinedAdjointThunk
    @ ~/.julia/packages/Enzyme/UKzrf/src/compiler.jl:5332 [inlined]
 [13] autodiff
    @ ~/.julia/packages/Enzyme/UKzrf/src/Enzyme.jl:291 [inlined]
 [14] autodiff
    @ ~/.julia/packages/Enzyme/UKzrf/src/Enzyme.jl:315 [inlined]
 [15] autodiff(mode::ReverseMode{false, FFIABI, false}, f::typeof(eval), args::Duplicated{Vector{Float64}})
    @ Enzyme ~/.julia/packages/Enzyme/UKzrf/src/Enzyme.jl:300
 [16] top-level scope
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:23

@wsmoses wsmoses reopened this May 22, 2024
@wsmoses
Copy link
Member

wsmoses commented May 22, 2024

@ToPo-ToPo-ToPo what is your version of Julia / OS.

The code above works for me on Enzyme#main

@wsmoses
Copy link
Member

wsmoses commented May 22, 2024

Also please show the package environments as well

@ToPo-ToPo-ToPo
Copy link
Author

I use 1.10.3+0.aarch64.apple.darwin14 (release).
The packages used are as follows:

Status `~/.julia/environments/v1.10/Project.toml`
  [6e4b80f9] BenchmarkTools v1.5.0
  [082447d4] ChainRules v1.66.0
  [d360d2e6] ChainRulesCore v1.23.0
  [7da242da] Enzyme v0.12.7
  [95c220a8] ExtendableSparse v1.4.0
  [f6369f11] ForwardDiff v0.10.36
  [42fd0dbc] IterativeSolvers v0.9.4
  [ba0b0d4f] Krylov v0.9.6
  [7ed4a6bd] LinearSolve v2.30.0
  [76087f3c] NLopt v1.0.2
  [b8a86587] NearestNeighbors v0.4.16
  [91a5bcdd] Plots v1.40.4
  [37e2e3b7] ReverseDiff v1.15.3
  [e56a9233] Sparspak v0.3.9
  [64499a7a] WriteVTK v1.19.1
  [e88e6eb3] Zygote v0.6.70

@wsmoses
Copy link
Member

wsmoses commented May 24, 2024

Just to double check, since I'm having difficulty reproducing: what happens if you're in an enviroment with only Enzyme.

@ToPo-ToPo-ToPo
Copy link
Author

All packages were removed, leaving only Enzyme.jl available.
When the same program was run, the error was not resolved.

Enzyme.jl: Enzyme v0.12.8

┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
┌ Warning: TODO reverse jl_array_del_end zero-set used memset rather than runtime type
└ @ Enzyme.Compiler ~/.julia/packages/GPUCompiler/nWT2N/src/utils.jl:59
ERROR: BoundsError: attempt to access 4-element Vector{Int64} at index [0]
Stacktrace:
  [1] _deleteend!
    @ ./array.jl:1081 [inlined]
  [2] resize!
    @ ./array.jl:1320 [inlined]
  [3] trimstorage!
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:135
  [4] _map_zeropres!
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:361
  [5] _noshapecheck_map
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:184
  [6] map
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/higherorderfns.jl:1187 [inlined]
  [7] +
    @ ~/.julia/juliaup/julia-1.10.3+0.aarch64.apple.darwin14/share/julia/stdlib/v1.10/SparseArrays/src/sparsematrix.jl:2242 [inlined]
  [8] eval
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:11 [inlined]
  [9] diffejulia_eval_2098wrap
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:0
 [10] macro expansion
    @ ~/.julia/packages/Enzyme/UZsMX/src/compiler.jl:5855 [inlined]
 [11] enzyme_call
    @ ~/.julia/packages/Enzyme/UZsMX/src/compiler.jl:5521 [inlined]
 [12] CombinedAdjointThunk
    @ ~/.julia/packages/Enzyme/UZsMX/src/compiler.jl:5400 [inlined]
 [13] autodiff
    @ ~/.julia/packages/Enzyme/UZsMX/src/Enzyme.jl:291 [inlined]
 [14] autodiff
    @ ~/.julia/packages/Enzyme/UZsMX/src/Enzyme.jl:315 [inlined]
 [15] autodiff(mode::ReverseMode{false, FFIABI, false}, f::typeof(eval), args::Duplicated{Vector{Float64}})
    @ Enzyme ~/.julia/packages/Enzyme/UZsMX/src/Enzyme.jl:300
 [16] top-level scope
    @ ~/my_program/GitHub/JuliaOptOS/tests/Enzyme/sparse_array_tests/basic_add.jl:23

@wsmoses
Copy link
Member

wsmoses commented May 25, 2024

I tried it out on an arm macbook and it still worked for me:

julia> versioninfo()
Julia Version 1.10.2
Commit bd47eca2c8a (2024-03-01 10:14 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: macOS (arm64-apple-darwin22.4.0)
  CPU: 10 × Apple M1 Max
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, apple-m1)
Threads: 1 default, 0 interactive, 1 GC (on 8 virtual cores)

I'm not really sure how to help more here since I can't reproduce, if you have any ideas.

Perhaps hop on a call or something

@ToPo-ToPo-ToPo
Copy link
Author

My environment is as follows. I tried the same version of julia as you, but the error was not resolved.
I don't know what is wrong.

julia> versioninfo()
Julia Version 1.10.2
Commit bd47eca2c8a (2024-03-01 10:14 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: macOS (arm64-apple-darwin22.4.0)
  CPU: 8 × Apple M2
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, apple-m1)
Threads: 1 default, 0 interactive, 1 GC (on 4 virtual cores)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
more information needed
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@wsmoses @ToPo-ToPo-ToPo