Skip to content
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

Jump to bottom

Add one benchmark to measure performance of MaskedScatterNDOfShape #182

Merged
merged 4 commits into from
May 16, 2024
Merged

Add one benchmark to measure performance of MaskedScatterNDOfShape #182

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9823bfe
documentation
xadupre May 8, 2024
e79c5bc
improve the code
xadupre May 8, 2024
17182ec
add bigger unit test
xadupre May 9, 2024
ad9429f
fix implementation
xadupre May 9, 2024
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 .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@ Testing/*
plot_*.csv
plot_*.xlsx
*.data
test_ort_version/*
test_ort_version*
x64/*
CMakeFiles/*
dist/*
Expand Down
12 changes: 10 additions & 2 deletions _doc/benchmarks.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,8 +104,16 @@ plot_op_scatternd_cuda

See :ref:`l-example-op-scatternd_cuda`.

The benchmark compares two operators Mul profiles
with their fusion into a single operator.
The benchmark compares two operators ScatterND, using
atomic, no atomic.

plot_op_scatternd_mask_cuda
+++++++++++++++++++++++++++

See :ref:`l-example-op-scatternd_mask_cuda`.

The benchmark compares three operators ScatterND to update
a matrix.

plot_op_transpose2dcast_cuda
++++++++++++++++++++++++++++
Expand Down
15 changes: 10 additions & 5 deletions _doc/examples/plot_op_scatternd_cuda.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@

Where the shapes are:

* zeros: 32000x4906
* zeros: 32000x4096
* indices: 2x1024x1
* updates: 2x1024x4096
"""
Expand Down Expand Up @@ -337,14 +337,18 @@ def benchmark(

print(f"sizes={sizes}")

data_nd1 = benchmark(sess1, sizes, script_args.config, "Atomic", itype=itype)
data_nd1 = benchmark(
sess1, sizes, script_args.config, "Atomic/Not Fused", itype=itype
)

#######################################
# Fused.

if sess2 is not None:

data_nd2 = benchmark(sess2, sizes, script_args.config, "No Atomic", itype=itype)
data_nd2 = benchmark(
sess2, sizes, script_args.config, "No Atomic/Fused", itype=itype
)


##########################################
Expand All @@ -364,10 +368,11 @@ def benchmark(
if sess2 is not None:

pivot = df.pivot(index="size", columns="label", values="time")
pivot["ratio"] = pivot["Atomic"] / pivot["No Atomic"]
pivot["ratio"] = pivot["Atomic/Not Fused"] / pivot["No Atomic/Fused"]
print("Speed up compare to the onnx standaed.")
print(pivot)

ax = pivot[["Atomic", "No Atomic"]].plot(
ax = pivot[["Atomic/Not Fused", "No Atomic/Fused"]].plot(
logx=True,
logy=True,
title=f"Atomic/No-Atomic implementation for ScatterND on CUDA\nitype={itype}",
Expand Down
Loading
Loading