Skip to content

Compute accumulate destination type based on the op #24

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
anicusan merged 2 commits into from
Feb 19, 2025
Merged

Compute accumulate destination type based on the op #24

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
fa3e748
Compute accumulate destination type based on the
pxl-th Feb 18, 2025
1b02e6e
Bump to 0.3.1
pxl-th Feb 18, 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
2 changes: 1 addition & 1 deletion Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
name = "AcceleratedKernels"
uuid = "6a4ca0a5-0e36-4168-a932-d9be78d558f1"
authors = ["Andrei-Leonard Nicusan <[email protected]> and contributors"]
version = "0.3.0"
version = "0.3.1"

[deps]
ArgCheck = "dce04be8-c92d-5529-be00-80e4d2c0e197"
Expand Down
8 changes: 4 additions & 4 deletions src/accumulate/accumulate.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,11 +71,11 @@ For compatibility with the `Base.accumulate!` function, we provide the two-array
we do not need the constraint of `dst` and `src` being different; to minimise memory use, we
recommend using the single-array interface (the first one above).

## CPU
## CPU
The CPU implementation is currently single-threaded; we are waiting on a multithreaded
implementation in OhMyThreads.jl ([issue](https://github.com/JuliaFolds2/OhMyThreads.jl/issues/129)).

## GPU
## GPU
For the 1D case (`dims=nothing`), the `alg` can be one of the following:
- `DecoupledLookback()`: the default algorithm, using opportunistic lookback to reuse earlier
blocks' results; requires device-level memory consistency guarantees, which Apple Metal does not
Expand Down Expand Up @@ -241,7 +241,7 @@ function accumulate(
temp::Union{Nothing, AbstractArray}=nothing,
temp_flags::Union{Nothing, AbstractArray}=nothing,
)
dst_type = promote_type(eltype(v), typeof(init))
dst_type = Base.promote_op(op, eltype(v), typeof(init))
vcopy = similar(v, dst_type)
copyto!(vcopy, v)
accumulate!(
Expand All @@ -252,7 +252,7 @@ function accumulate(
inclusive=inclusive,

alg=alg,

block_size=block_size,
temp=temp,
temp_flags=temp_flags,
Expand Down
Loading