Skip to content

[CUBLAS] Update wrapppers to use the ILP64 API #2845

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

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

[CUBLAS] Update wrapppers to use the ILP64 API #2845

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c5da1f5
Update wrapppers to use the ILP64 API
amontoison Aug 12, 2025
ace28c3
Update wrappers.jl
amontoison Aug 12, 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
54 changes: 37 additions & 17 deletions lib/cublas/wrappers.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -400,8 +400,8 @@ for (fname, fname_64, elty) in ((:cublasSrotm_v2, :cublasSrotm_v2_64, :Float32),
end

## rotmg
for (fname, elty) in ((:cublasSrotmg_v2, :Float32),
(:cublasDrotmg_v2, :Float64))
for (fname, fname_64, elty) in ((:cublasSrotmg_v2, :cublasSrotmg_v2_64, :Float32),
(:cublasDrotmg_v2, :cublasSrotmg_v2_64, :Float64))
@eval begin
function rotmg!(d1::$elty,
d2::$elty,
Expand All @@ -412,7 +412,11 @@ for (fname, elty) in ((:cublasSrotmg_v2, :Float32),
ref_d2 = CuRef(d2)
ref_x = CuRef(x)
ref_y = CuRef(y)
$fname(handle(), ref_d1, ref_d2, ref_x, ref_y, param)
if CUBLAS.version() >= v"12.0"
$fname_64(handle(), ref_d1, ref_d2, ref_x, ref_y, param)
else
$fname(handle(), ref_d1, ref_d2, ref_x, ref_y, param)
end
ref_d1[], ref_d2[], ref_x[], ref_y[], param
end
end
Expand Down Expand Up @@ -545,8 +549,8 @@ end
for (fname, fname_64, eltyin, eltyout) in (
(:cublasDgemvBatched, :cublasDgemvBatched_64, :Float64, :Float64),
(:cublasSgemvBatched, :cublasSgemvBatched_64, :Float32, :Float32),
(:cublasHSHgemvBatched, :cublasHSHgemvBatched, :Float16, :Float16),
(:cublasHSSgemvBatched, :cublasHSSgemvBatched, :Float16, :Float32),
(:cublasHSHgemvBatched, :cublasHSHgemvBatched_64, :Float16, :Float16),
(:cublasHSSgemvBatched, :cublasHSSgemvBatched_64, :Float16, :Float32),
(:cublasZgemvBatched, :cublasZgemvBatched_64, :ComplexF64, :ComplexF64),
(:cublasCgemvBatched, :cublasCgemvBatched_64, :ComplexF32, :ComplexF32),
)
Expand Down Expand Up @@ -594,8 +598,8 @@ end
for (fname, fname_64, eltyin, eltyout) in (
(:cublasDgemvStridedBatched, :cublasDgemvStridedBatched_64, :Float64, :Float64),
(:cublasSgemvStridedBatched, :cublasSgemvStridedBatched_64, :Float32, :Float32),
(:cublasHSHgemvStridedBatched, :cublasHSHgemvStridedBatched, :Float16, :Float16),
(:cublasHSSgemvStridedBatched, :cublasHSSgemvStridedBatched, :Float16, :Float32),
(:cublasHSHgemvStridedBatched, :cublasHSHgemvStridedBatched_64, :Float16, :Float16),
(:cublasHSSgemvStridedBatched, :cublasHSSgemvStridedBatched_64, :Float16, :Float32),
(:cublasZgemvStridedBatched, :cublasZgemvStridedBatched_64, :ComplexF64, :ComplexF64),
(:cublasCgemvStridedBatched, :cublasCgemvStridedBatched_64, :ComplexF32, :ComplexF32),
)
Expand Down Expand Up @@ -1116,7 +1120,7 @@ end
## (GE) general matrix-matrix multiplication
for (fname, fname_64, elty) in ((:cublasDgemm_v2, :cublasDgemm_v2_64, :Float64),
(:cublasSgemm_v2, :cublasSgemm_v2_64, :Float32),
(:cublasHgemm, :cublasHgemm, :Float16),
(:cublasHgemm, :cublasHgemm_64, :Float16),
(:cublasZgemm_v2, :cublasZgemm_v2_64, :ComplexF64),
(:cublasCgemm_v2, :cublasCgemm_v2_64, :ComplexF32))
@eval begin
Expand Down Expand Up @@ -1527,7 +1531,7 @@ end
## (GE) general matrix-matrix multiplication batched
for (fname, fname_64, elty) in ((:cublasDgemmBatched, :cublasDgemmBatched_64, :Float64),
(:cublasSgemmBatched, :cublasSgemmBatched_64, :Float32),
(:cublasHgemmBatched, :cublasHgemmBatched, :Float16),
(:cublasHgemmBatched, :cublasHgemmBatched_64, :Float16),
(:cublasZgemmBatched, :cublasZgemmBatched_64, :ComplexF64),
(:cublasCgemmBatched, :cublasCgemmBatched_64, :ComplexF32))
@eval begin
Expand Down Expand Up @@ -1594,7 +1598,7 @@ end
## (GE) general matrix-matrix multiplication strided batched
for (fname, fname_64, elty) in ((:cublasDgemmStridedBatched, :cublasDgemmStridedBatched_64, :Float64),
(:cublasSgemmStridedBatched, :cublasSgemmStridedBatched_64, :Float32),
(:cublasHgemmStridedBatched, :cublasHgemmStridedBatched, :Float16),
(:cublasHgemmStridedBatched, :cublasHgemmStridedBatched_64, :Float16),
(:cublasZgemmStridedBatched, :cublasZgemmStridedBatched_64, :ComplexF64),
(:cublasCgemmStridedBatched, :cublasCgemmStridedBatched_64, :ComplexF32))
@eval begin
Expand Down Expand Up @@ -1945,11 +1949,11 @@ function her2k(uplo::Char, trans::Char,
end

## (TR) Triangular matrix and vector multiplication and solution
for (mmname, smname, elty) in
((:cublasDtrmm_v2,:cublasDtrsm_v2,:Float64),
(:cublasStrmm_v2,:cublasStrsm_v2,:Float32),
(:cublasZtrmm_v2,:cublasZtrsm_v2,:ComplexF64),
(:cublasCtrmm_v2,:cublasCtrsm_v2,:ComplexF32))
for (mmname, mmname_64, elty) in
((:cublasDtrmm_v2, :cublasDtrmm_v2_64, :Float64),
(:cublasStrmm_v2, :cublasStrmm_v2_64, :Float32),
(:cublasZtrmm_v2, :cublasZtrmm_v2_64, :ComplexF64),
(:cublasCtrmm_v2, :cublasCtrmm_v2_64, :ComplexF32))
@eval begin
# Note: CUBLAS differs from BLAS API for trmm
# BLAS: inplace modification of B
Expand All @@ -1972,10 +1976,22 @@ for (mmname, smname, elty) in
lda = max(1,stride(A,2))
ldb = max(1,stride(B,2))
ldc = max(1,stride(C,2))
$mmname(handle(), side, uplo, transa, diag, m, n, alpha, A, lda, B, ldb, C, ldc)
if CUBLAS.version() >= v"12.0"
$mmname_64(handle(), side, uplo, transa, diag, m, n, alpha, A, lda, B, ldb, C, ldc)
else
$mmname(handle(), side, uplo, transa, diag, m, n, alpha, A, lda, B, ldb, C, ldc)
end
C
end
end
end

for (smname, smname_64, elty) in
((:cublasDtrsm_v2, :cublasDtrsm_v2_64, :Float64),
(:cublasStrsm_v2, :cublasStrsm_v2_64, :Float32),
(:cublasZtrsm_v2, :cublasZtrsm_v2_64, :ComplexF64),
(:cublasCtrsm_v2, :cublasCtrsm_v2_64, :ComplexF32))
@eval begin
function trsm!(side::Char,
uplo::Char,
transa::Char,
Expand All @@ -1990,7 +2006,11 @@ for (mmname, smname, elty) in
if nA != (side == 'L' ? m : n) throw(DimensionMismatch("trsm!")) end
lda = max(1,stride(A,2))
ldb = max(1,stride(B,2))
$smname(handle(), side, uplo, transa, diag, m, n, alpha, A, lda, B, ldb)
if CUBLAS.version() >= v"12.0"
$smname_64(handle(), side, uplo, transa, diag, m, n, alpha, A, lda, B, ldb)
else
$smname(handle(), side, uplo, transa, diag, m, n, alpha, A, lda, B, ldb)
end
B
end
end
Expand Down