Skip to content

Refactor some duplication in unary ops #664

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
merged 1 commit into from
Apr 13, 2025
Merged

Refactor some duplication in unary ops #664

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
64 changes: 12 additions & 52 deletions src/ops/unary_elementwise.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -179,10 +179,10 @@ fn par_unary_op_in_place<T: Copy + Send, VF: Fn(&mut [T]) + Send + Sync, SF: Fn(
}
}

/// Define an operator which supports float tensors and is optimize using SIMD
/// Define an operator which supports float tensors and is optimized using SIMD
/// and multithreading.
macro_rules! parallel_unary_float_op {
($op_name:ident, $func_name:ident, $in_place_func_name:ident, $impl_func:expr, $impl_in_place_func:expr, $impl_scalar:expr) => {
($op_name:ident, $func_name:ident, $in_place_func_name:ident, $simd_kernel:expr) => {
#[derive(Debug)]
pub struct $op_name {}

Expand Down Expand Up @@ -214,11 +214,13 @@ macro_rules! parallel_unary_float_op {
}

pub fn $func_name(pool: &TensorPool, input: TensorView) -> Tensor {
par_unary_op(pool, input, $impl_func)
let kernel = $simd_kernel;
par_unary_op(pool, input, |src, dst| kernel.map(src, dst))
}

pub fn $in_place_func_name(input: TensorViewMut) {
par_unary_op_in_place(input, $impl_in_place_func, $impl_scalar);
let kernel = $simd_kernel;
par_unary_op_in_place(input, |dst| kernel.map_mut(dst), |x| kernel.scalar_eval(x));
}
};
}
Expand Down Expand Up @@ -399,32 +401,11 @@ pub fn elu_in_place(input: TensorViewMut, alpha: f32) {
Elu { alpha }.apply(input)
}

parallel_unary_float_op!(
Erf,
erf,
erf_in_place,
|src, dest| vecmath::Erf {}.map(src, dest),
|src| vecmath::Erf {}.map_mut(src),
|x| vecmath::Erf {}.scalar_eval(x)
);
parallel_unary_float_op!(
Exp,
exp,
exp_in_place,
|src, dest| vecmath::Exp {}.map(src, dest),
|src| vecmath::Exp {}.map_mut(src),
|x| vecmath::Exp {}.scalar_eval(x)
);
parallel_unary_float_op!(Erf, erf, erf_in_place, vecmath::Erf {});
parallel_unary_float_op!(Exp, exp, exp_in_place, vecmath::Exp {});
unary_float_op!(Floor, floor, floor_in_place, |val: f32| val.floor());

parallel_unary_float_op!(
Gelu,
gelu,
gelu_in_place,
|src, dest| vecmath::Gelu {}.map(src, dest),
|src| vecmath::Gelu {}.map_mut(src),
|x| vecmath::Gelu {}.scalar_eval(x)
);
parallel_unary_float_op!(Gelu, gelu, gelu_in_place, vecmath::Gelu {});

#[derive(Debug)]
pub struct HardSigmoid {
Expand Down Expand Up @@ -575,14 +556,7 @@ pub fn round_in_place(x: TensorViewMut) {
Round {}.apply(x)
}

parallel_unary_float_op!(
Sigmoid,
sigmoid,
sigmoid_in_place,
|src, dest| vecmath::Sigmoid {}.map(src, dest),
|src| vecmath::Sigmoid {}.map_mut(src),
|x| vecmath::Sigmoid {}.scalar_eval(x)
);
parallel_unary_float_op!(Sigmoid, sigmoid, sigmoid_in_place, vecmath::Sigmoid {});

// Sigmoid Linear Unit (SiLU) function.
//
Expand All @@ -591,14 +565,7 @@ parallel_unary_float_op!(
//
// Not an official ONNX operator, but used in popular object detection models.
// See https://github.com/onnx/onnx/issues/4854.
parallel_unary_float_op!(
Silu,
silu,
silu_in_place,
|src, dest| vecmath::Silu {}.map(src, dest),
|src| vecmath::Silu {}.map_mut(src),
|x| vecmath::Silu {}.scalar_eval(x)
);
parallel_unary_float_op!(Silu, silu, silu_in_place, vecmath::Silu {});

/// Swish function (<https://en.wikipedia.org/wiki/Swish_function>).
///
Expand Down Expand Up @@ -682,14 +649,7 @@ unary_float_op!(Softplus, softplus, softplus_in_place, |val: f32| {
val.exp().ln_1p()
});
unary_float_op!(Tan, tan, tan_in_place, |val: f32| val.tan());
parallel_unary_float_op!(
Tanh,
tanh,
tanh_in_place,
|src, dest| vecmath::Tanh {}.map(src, dest),
|src| vecmath::Tanh {}.map_mut(src),
|x| vecmath::Tanh {}.scalar_eval(x)
);
parallel_unary_float_op!(Tanh, tanh, tanh_in_place, vecmath::Tanh {});

#[cfg(test)]
mod tests {
Expand Down