Skip to content

Avoid relying on a RNG in the back-end. #687

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
maleadt merged 1 commit into from
Mar 22, 2026
+12 −12
Merged

Avoid relying on a RNG in the back-end. #687

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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 examples/matmul.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,8 +26,8 @@ function matmul!(output, a, b)
return
end

a = rand!(allocate(backend, Float32, 256, 123))
b = rand!(allocate(backend, Float32, 123, 45))
a = copyto!(allocate(backend, Float32, 256, 123), rand(Float32, 256, 123))
b = copyto!(allocate(backend, Float32, 123, 45), rand(Float32, 123, 45))
output = KernelAbstractions.zeros(backend, Float32, 256, 45)

matmul!(output, a, b)
Expand Down
2 changes: 1 addition & 1 deletion examples/naive_transpose.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ end
res = 1024

# creating initial arrays
b = rand!(allocate(backend, Float32, res, res))
b = copyto!(allocate(backend, Float32, res, res), rand(Float32, res, res))
a = KernelAbstractions.zeros(backend, Float32, res, res)

naive_transpose!(a, b)
Expand Down
8 changes: 4 additions & 4 deletions examples/numa_aware.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,11 +27,11 @@ function measure_membw(

a = dtype(3.1415)
if init == :serial
X = rand!(zeros(dtype, N))
Y = rand!(zeros(dtype, N))
X = rand(dtype, N)
Y = rand(dtype, N)
else
X = rand!(KernelAbstractions.zeros(backend, dtype, N))
Y = rand!(KernelAbstractions.zeros(backend, dtype, N))
X = copyto!(KernelAbstractions.zeros(backend, dtype, N), rand(dtype, N))
Y = copyto!(KernelAbstractions.zeros(backend, dtype, N), rand(dtype, N))
end
workgroup_size = 1024

Expand Down
6 changes: 3 additions & 3 deletions examples/performance.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,7 +145,7 @@ for block_dims in ((TILE_DIM, TILE_DIM), (TILE_DIM * TILE_DIM, 1), (1, TILE_DIM
("transpose", simple_transpose_kernel!(backend, block_dims)),
)
NVTX.@range "Simple $name $block_dims" let
input = rand!(allocate(backend, T, N, N))
input = copyto!(allocate(backend, T, N, N), rand(T, N, N))
output = similar(input)

# compile kernel
Expand All @@ -165,7 +165,7 @@ for (name, kernel) in (
)
for bank in (true, false)
NVTX.@range "Localmem $name ($TILE_DIM, $TILE_DIM) bank=$bank" let
input = rand!(allocate(backend, T, N, N))
input = copyto!(allocate(backend, T, N, N), rand(T, N, N))
output = similar(input)

# compile kernel
Expand All @@ -185,7 +185,7 @@ for (name, kernel) in (
)
for bank in (true, false)
NVTX.@range "Localmem + multiple elements $name ($TILE_DIM, $BLOCK_ROWS) bank=$bank" let
input = rand!(allocate(backend, T, N, N))
input = copyto!(allocate(backend, T, N, N), rand(T, N, N))
output = similar(input)

# We want a number of blocks equivalent to (TILE_DIM, TILE_DIM)
Expand Down
4 changes: 2 additions & 2 deletions examples/performant_matmul.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,8 +79,8 @@ end
N = 1024
R = 512
M = 2048
A = rand!(allocate(backend, Float32, N, R))
B = rand!(allocate(backend, Float32, R, M))
A = copyto!(allocate(backend, Float32, N, R), rand(Float32, N, R))
B = copyto!(allocate(backend, Float32, R, M), rand(Float32, R, M))
C = KernelAbstractions.zeros(backend, Float32, N, M)

workgroupsize = (TILE_DIM, TILE_DIM)
Expand Down
Loading