Skip to content

Add narrow-precision floats to cudaDataType #3180

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
AntonOresten wants to merge 2 commits into
base: main
Choose a base branch
from
+14 −0
Open

Add narrow-precision floats to cudaDataType #3180

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
62cea0f
Add narrow-precision floats to cudaDataType [only julia] [skip benchm…
AntonOresten Jun 21, 2026
603e01b
Add R_8F_UE4M3 alias of R_8F_E4M3 [only julia] [skip benchmarks]
AntonOresten Jun 21, 2026
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
8 changes: 8 additions & 0 deletions CUDACore/src/library_types.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,8 +33,16 @@ end
C_64I = 25 # complex as a pair of signed 64-bit int numbers
R_64U = 26 # real as a unsigned 64-bit int
C_64U = 27 # complex as a pair of unsigned 64-bit int numbers
R_8F_E4M3 = 28 # real as a nv_fp8_e4m3
R_8F_E5M2 = 29 # real as a nv_fp8_e5m2
R_8F_UE8M0 = 30 # real as an exponent-only unsigned nv_fp8_e8m0
R_6F_E2M3 = 31 # real as a nv_fp6_e2m3
R_6F_E3M2 = 32 # real as a nv_fp6_e3m2
R_4F_E2M1 = 33 # real as a nv_fp4_e2m1
end

const R_8F_UE4M3 = R_8F_E4M3 # real as an unsigned nv_fp8_e4m3

function Base.convert(::Type{cudaDataType}, T::DataType)
if T === Float16
return R_16F
Expand Down
6 changes: 6 additions & 0 deletions test/core/utils.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,4 +30,10 @@ end
end
@test_throws ArgumentError convert(CUDACore.cudaDataType, BigFloat)
@test_throws ArgumentError Type(CUDACore.R_4I) # adjust once we support 4-bit Ints
# narrow-precision floats: enum members exist,
# but no Julia type maps to them in CUDACore.
for c_type in (CUDACore.R_8F_E4M3, CUDACore.R_8F_E5M2, CUDACore.R_8F_UE8M0,
CUDACore.R_6F_E2M3, CUDACore.R_6F_E3M2, CUDACore.R_4F_E2M1)
@test_throws ArgumentError Type(c_type)
end
end