Skip to content

Use optimize_gemm! from oneAPI.jl #77

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

Merged
amontoison merged 3 commits into from
Aug 27, 2025
Merged

Use optimize_gemm! from oneAPI.jl #77

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ddb59b1
Use optimize_gemm! from oneAPI.jl
amontoison Aug 22, 2025
75cf724
Update the tolerance on GPU
amontoison Aug 27, 2025
219270f
Update ext/oneAPI/operators.jl
amontoison Aug 27, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ KrylovPreconditionersCUDAExt = "CUDA"
KrylovPreconditionersOneAPIExt = "oneAPI"

[compat]
AMDGPU = "2"
AMDGPU = "2.0.1"
Adapt = "4"
CUDA = "5.7.3"
Graphs = "1.13.0"
Expand All @@ -34,7 +34,7 @@ Random = "1.10"
SparseArrays = "1.10"
Test = "1.10"
julia = "1.10"
oneAPI = "2.0.1"
oneAPI = "2.1.0"

[extras]
Krylov = "ba0b0d4f-ebba-5204-a429-3ac8c609bfb7"
Expand Down
3 changes: 2 additions & 1 deletion ext/oneAPI/operators.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,8 +16,9 @@ for (SparseMatrixType, BlasType) in ((:(oneSparseMatrixCSR{T}), :BlasFloat),)
m,n = size(A)
if nrhs == 1
oneMKL.sparse_optimize_gemv!(transa, A)
else
oneMKL.sparse_optimize_gemm!(transa, 'N', nrhs, A)
end
# sparse_optimize_gemm! is only available with oneAPI > v2024.1.0
return INTEL_KrylovOperator{T}(T, m, n, nrhs, transa, A)
end

Expand Down
18 changes: 11 additions & 7 deletions test/gpu/gpu.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ Random.seed!(666)
function test_ic0(FC, V, M)
n = 100
R = real(FC)
tol = eps(R) |> sqrt
A_cpu = rand(FC, n, n)
A_cpu = A_cpu * A_cpu'
A_cpu = sparse(A_cpu)
Expand All @@ -19,9 +20,9 @@ function test_ic0(FC, V, M)
r_gpu = b_gpu - A_gpu * x_gpu
@test stats.niter ≤ 5
if (FC <: ComplexF64) && V.body.name.name == :ROCArray
@test_broken norm(r_gpu) ≤ 1e-6
@test_broken norm(r_gpu) ≤ tol
else
@test norm(r_gpu) ≤ 1e-8
@test norm(r_gpu) ≤ tol
end

A_gpu = M(A_cpu + 200*I)
Expand All @@ -30,15 +31,16 @@ function test_ic0(FC, V, M)
r_gpu = b_gpu - A_gpu * x_gpu
@test stats.niter ≤ 5
if (FC <: ComplexF64) && V.body.name.name == :ROCArray
@test_broken norm(r_gpu) ≤ 1e-6
@test_broken norm(r_gpu) ≤ tol
else
@test norm(r_gpu) ≤ 1e-8
@test norm(r_gpu) ≤ tol
end
end

function test_ilu0(FC, V, M)
n = 100
R = real(FC)
tol = eps(R) |> sqrt
A_cpu = rand(FC, n, n)
A_cpu = sparse(A_cpu)
b_cpu = rand(FC, n)
Expand All @@ -50,17 +52,19 @@ function test_ilu0(FC, V, M)
x_gpu, stats = gmres(A_gpu, b_gpu, N=P, ldiv=true)
r_gpu = b_gpu - A_gpu * x_gpu
@test stats.niter ≤ 5
@test norm(r_gpu) ≤ 1e-8
@test norm(r_gpu) ≤ tol

A_gpu = M(A_cpu + 200*I)
update!(P, A_gpu)
x_gpu, stats = gmres(A_gpu, b_gpu, N=P, ldiv=true)
r_gpu = b_gpu - A_gpu * x_gpu
@test stats.niter ≤ 5
@test norm(r_gpu) ≤ 1e-8
@test norm(r_gpu) ≤ tol
end

function test_operator(FC, V, DM, SM)
R = real(FC)
tol = eps(R) |> sqrt
m = 200
n = 100
A_cpu = rand(FC, n, n)
Expand All @@ -74,7 +78,7 @@ function test_operator(FC, V, DM, SM)
x_gpu, stats = gmres(opA_gpu, b_gpu)
r_gpu = b_gpu - A_gpu * x_gpu
@test stats.solved
@test norm(r_gpu) ≤ 1e-8
@test norm(r_gpu) ≤ tol

A_cpu = rand(FC, m, n)
A_cpu = sparse(A_cpu)
Expand Down
Loading