NVIDIA/nvbench
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvbench.git synced 2026-05-13 09:45:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

Correct cuda cccl examples per change in api (#353)

Browse Source
This commit is contained in:
Oleksandr Pavlyk
2026-05-06 13:30:44 -05:00
committed by GitHub
parent f392725015
commit d13a0fde32
2 changed files with 27 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
python/examples/cuda_compute_segmented_reduce.py
View File

@@ -67,21 +67,26 @@ def segmented_reduce(state: bench.State):
add_op = OpKind.PLUS
alg = algorithms.make_segmented_reduce(
d_input, d_output, start_offsets, end_offsets, add_op, h_init
d_in=d_input,
d_out=d_output,
start_offsets_in=start_offsets,
end_offsets_in=end_offsets,
op=add_op,
h_init=h_init,
)
cccl_stream = state.get_stream()
# query size of temporary storage and allocate
temp_nbytes = alg(
None,
d_input,
d_output,
add_op,
n_rows,
start_offsets,
end_offsets,
h_init,
cccl_stream,
temp_storage=None,
d_in=d_input,
d_out=d_output,
op=add_op,
num_segments=n_rows,
start_offsets_in=start_offsets,
end_offsets_in=end_offsets,
h_init=h_init,
stream=cccl_stream,
)
h_init = np.zeros(tuple(), dtype=np.int32)
@@ -91,15 +96,15 @@ def segmented_reduce(state: bench.State):
def launcher(launch: bench.Launch):
s = launch.get_stream()
alg(
temp_storage,
d_input,
d_output,
add_op,
n_rows,
start_offsets,
end_offsets,
h_init,
s,
temp_storage=temp_storage,
d_in=d_input,
d_out=d_output,
op=add_op,
num_segments=n_rows,
start_offsets_in=start_offsets,
end_offsets_in=end_offsets,
h_init=h_init,
stream=s,
)
state.exec(launcher)

4
python/examples/cuda_coop_block_reduce.py
View File

@@ -58,7 +58,9 @@ def multi_block_bench(state: bench.State):
return
ring = BitsetRing()
block_reduce = coop.block.reduce(numba.uint64, threads_per_block, BitsetRing.add)
block_reduce = coop.block.make_reduce(
numba.uint64, threads_per_block, BitsetRing.add
)
@cuda.jit(link=block_reduce.files)
def kernel(inp_arr, out_arr):