Skip to content

perf(wasi-nn): avoid extra allocations in tensor byte conversions #12490

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
ANtutov wants to merge 3 commits into
base: main
Choose a base branch
from
Open

perf(wasi-nn): avoid extra allocations in tensor byte conversions #12490

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2c772f9
Update onnx.rs
ANtutov Feb 2, 2026
56b4815
Update main.rs
ANtutov Feb 2, 2026
b3d437a
Update main.rs
ANtutov Feb 2, 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
14 changes: 7 additions & 7 deletions crates/wasi-nn/examples/classification-component-onnx/src/main.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,13 +116,13 @@ fn main() {
}

pub fn bytes_to_f32_vec(data: Vec<u8>) -> Vec<f32> {
let chunks: Vec<&[u8]> = data.chunks(4).collect();
let v: Vec<f32> = chunks
.into_iter()
.map(|c| f32::from_le_bytes(c.try_into().unwrap()))
.collect();

v.into_iter().collect()
assert_eq!(data.len() % 4, 0);
data.chunks(4)
.map(|c| {
let arr: [u8; 4] = c.try_into().unwrap();
f32::from_le_bytes(arr)
})
.collect()
}

// Take the image located at 'path', open it, resize it to height x width, and then converts
Expand Down
14 changes: 7 additions & 7 deletions crates/wasi-nn/examples/classification-example-winml/src/main.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,13 +121,13 @@ fn postprocess(output_tensor: Vec<f32>) -> Vec<f32> {
}

pub fn bytes_to_f32_vec(data: Vec<u8>) -> Vec<f32> {
let chunks: Vec<&[u8]> = data.chunks(4).collect();
let v: Vec<f32> = chunks
.into_iter()
.map(|c| f32::from_ne_bytes(c.try_into().unwrap()))
.collect();

v.into_iter().collect()
assert_eq!(data.len() % 4, 0);
data.chunks(4)
.map(|c| {
let arr: [u8; 4] = c.try_into().unwrap();
f32::from_ne_bytes(arr)
})
.collect()
}

// A wrapper for class ID and match probabilities.
Expand Down
22 changes: 12 additions & 10 deletions crates/wasi-nn/src/backend/onnx.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -437,19 +437,21 @@ fn to_input_value(slot: &TensorSlot) -> Result<[SessionInputValue<'_>; 1], Backe
}

pub fn f32_vec_to_bytes(data: Vec<f32>) -> Vec<u8> {
let chunks: Vec<[u8; 4]> = data.into_iter().map(|f| f.to_le_bytes()).collect();
let result: Vec<u8> = chunks.iter().flatten().copied().collect();
result
let mut bytes = Vec::with_capacity(data.len() * 4);
for f in data {
bytes.extend_from_slice(&f.to_le_bytes());
}
bytes
}

pub fn bytes_to_f32_vec(data: Vec<u8>) -> Vec<f32> {
let chunks: Vec<&[u8]> = data.chunks(4).collect();
let v: Vec<f32> = chunks
.into_iter()
.map(|c| f32::from_le_bytes(c.try_into().unwrap()))
.collect();

v.into_iter().collect()
assert_eq!(data.len() % 4, 0);
data.chunks(4)
.map(|c| {
let arr: [u8; 4] = c.try_into().unwrap();
f32::from_le_bytes(arr)
})
.collect()
}

/// Returns whether the dimension is dynamic.
Expand Down