Skip to content

Commit d90d074

Browse files
kumarUjjawalkumarUjjawal
authored
perf: Optimize round scalar performance (#19831)
## Which issue does this PR close? <!-- We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123. --> - Part of apache/datafusion-comet#2986 ## Rationale for this change The round function currently converts scalar inputs to arrays before processing, even when both value and decimal_places are scalar values. This adds unnecessary overhead for constant folding scenarios like <!-- Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed. Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes. --> ## What changes are included in this PR? - Add scalar fast path in RoundFunc::invoke_with_args for Float64 and Float32 inputs - Directly compute the result when both inputs are scalars, avoiding array conversion overhead - Add benchmark <!-- There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR. --> ## Are these changes tested? Yes Type | Before | After | Speedup -- | -- | -- | -- round_f64_scalar | 570 ns | 195 ns | 2.9x round_f32_scalar | 564 ns | 192 ns | 2.9x <!-- We typically require tests for all PRs in order to: 1. Prevent the code from being accidentally broken by subsequent changes 2. Serve as another way to document the expected behavior of the code If tests are not included in your PR, please explain why (for example, are they covered by existing tests)? --> ## Are there any user-facing changes? No <!-- If there are user-facing changes then we may require documentation to be updated before approving the PR. --> <!-- If there are any breaking changes to public APIs, please add the `api change` label. -->
1 parent 05451da commit d90d074

File tree

3 files changed

+221
-2
lines changed

3 files changed

+221
-2
lines changed

‎datafusion/functions/Cargo.toml‎

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -320,3 +320,8 @@ required-features = ["math_expressions"]
320320
harness = false
321321
name = "floor_ceil"
322322
required-features = ["math_expressions"]
323+
324+
[[bench]]
325+
harness = false
326+
name = "round"
327+
required-features = ["math_expressions"]

‎datafusion/functions/benches/round.rs‎

Lines changed: 154 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,154 @@
1+
// Licensed to the Apache Software Foundation (ASF) under one
2+
// or more contributor license agreements. See the NOTICE file
3+
// distributed with this work for additional information
4+
// regarding copyright ownership. The ASF licenses this file
5+
// to you under the Apache License, Version 2.0 (the
6+
// "License"); you may not use this file except in compliance
7+
// with the License. You may obtain a copy of the License at
8+
//
9+
// http://www.apache.org/licenses/LICENSE-2.0
10+
//
11+
// Unless required by applicable law or agreed to in writing,
12+
// software distributed under the License is distributed on an
13+
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14+
// KIND, either express or implied. See the License for the
15+
// specific language governing permissions and limitations
16+
// under the License.
17+
18+
extern crate criterion;
19+
20+
use arrow::datatypes::{DataType, Field, Float32Type, Float64Type};
21+
use arrow::util::bench_util::create_primitive_array;
22+
use criterion::{Criterion, SamplingMode, criterion_group, criterion_main};
23+
use datafusion_common::ScalarValue;
24+
use datafusion_common::config::ConfigOptions;
25+
use datafusion_expr::{ColumnarValue, ScalarFunctionArgs};
26+
use datafusion_functions::math::round;
27+
use std::hint::black_box;
28+
use std::sync::Arc;
29+
use std::time::Duration;
30+
31+
fn criterion_benchmark(c: &mut Criterion) {
32+
let round_fn = round();
33+
let config_options = Arc::new(ConfigOptions::default());
34+
35+
for size in [1024, 4096, 8192] {
36+
let mut group = c.benchmark_group(format!("round size={size}"));
37+
group.sampling_mode(SamplingMode::Flat);
38+
group.sample_size(10);
39+
group.measurement_time(Duration::from_secs(10));
40+
41+
// Float64 array benchmark
42+
let f64_array = Arc::new(create_primitive_array::<Float64Type>(size, 0.1));
43+
let batch_len = f64_array.len();
44+
let f64_args = vec![
45+
ColumnarValue::Array(f64_array),
46+
ColumnarValue::Scalar(ScalarValue::Int32(Some(2))),
47+
];
48+
49+
group.bench_function("round_f64_array", |b| {
50+
b.iter(|| {
51+
let args_cloned = f64_args.clone();
52+
black_box(
53+
round_fn
54+
.invoke_with_args(ScalarFunctionArgs {
55+
args: args_cloned,
56+
arg_fields: vec![
57+
Field::new("a", DataType::Float64, true).into(),
58+
Field::new("b", DataType::Int32, false).into(),
59+
],
60+
number_rows: batch_len,
61+
return_field: Field::new("f", DataType::Float64, true).into(),
62+
config_options: Arc::clone(&config_options),
63+
})
64+
.unwrap(),
65+
)
66+
})
67+
});
68+
69+
// Float32 array benchmark
70+
let f32_array = Arc::new(create_primitive_array::<Float32Type>(size, 0.1));
71+
let f32_args = vec![
72+
ColumnarValue::Array(f32_array),
73+
ColumnarValue::Scalar(ScalarValue::Int32(Some(2))),
74+
];
75+
76+
group.bench_function("round_f32_array", |b| {
77+
b.iter(|| {
78+
let args_cloned = f32_args.clone();
79+
black_box(
80+
round_fn
81+
.invoke_with_args(ScalarFunctionArgs {
82+
args: args_cloned,
83+
arg_fields: vec![
84+
Field::new("a", DataType::Float32, true).into(),
85+
Field::new("b", DataType::Int32, false).into(),
86+
],
87+
number_rows: batch_len,
88+
return_field: Field::new("f", DataType::Float32, true).into(),
89+
config_options: Arc::clone(&config_options),
90+
})
91+
.unwrap(),
92+
)
93+
})
94+
});
95+
96+
// Scalar benchmark (the optimization we added)
97+
let scalar_f64_args = vec![
98+
ColumnarValue::Scalar(ScalarValue::Float64(Some(std::f64::consts::PI))),
99+
ColumnarValue::Scalar(ScalarValue::Int32(Some(2))),
100+
];
101+
102+
group.bench_function("round_f64_scalar", |b| {
103+
b.iter(|| {
104+
let args_cloned = scalar_f64_args.clone();
105+
black_box(
106+
round_fn
107+
.invoke_with_args(ScalarFunctionArgs {
108+
args: args_cloned,
109+
arg_fields: vec![
110+
Field::new("a", DataType::Float64, false).into(),
111+
Field::new("b", DataType::Int32, false).into(),
112+
],
113+
number_rows: 1,
114+
return_field: Field::new("f", DataType::Float64, false)
115+
.into(),
116+
config_options: Arc::clone(&config_options),
117+
})
118+
.unwrap(),
119+
)
120+
})
121+
});
122+
123+
let scalar_f32_args = vec![
124+
ColumnarValue::Scalar(ScalarValue::Float32(Some(std::f32::consts::PI))),
125+
ColumnarValue::Scalar(ScalarValue::Int32(Some(2))),
126+
];
127+
128+
group.bench_function("round_f32_scalar", |b| {
129+
b.iter(|| {
130+
let args_cloned = scalar_f32_args.clone();
131+
black_box(
132+
round_fn
133+
.invoke_with_args(ScalarFunctionArgs {
134+
args: args_cloned,
135+
arg_fields: vec![
136+
Field::new("a", DataType::Float32, false).into(),
137+
Field::new("b", DataType::Int32, false).into(),
138+
],
139+
number_rows: 1,
140+
return_field: Field::new("f", DataType::Float32, false)
141+
.into(),
142+
config_options: Arc::clone(&config_options),
143+
})
144+
.unwrap(),
145+
)
146+
})
147+
});
148+
149+
group.finish();
150+
}
151+
}
152+
153+
criterion_group!(benches, criterion_benchmark);
154+
criterion_main!(benches);

‎datafusion/functions/src/math/round.rs‎

Lines changed: 62 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -31,7 +31,7 @@ use arrow::error::ArrowError;
3131
use datafusion_common::types::{
3232
NativeType, logical_float32, logical_float64, logical_int32,
3333
};
34-
use datafusion_common::{Result, ScalarValue, exec_err};
34+
use datafusion_common::{Result, ScalarValue, exec_err, internal_err};
3535
use datafusion_expr::sort_properties::{ExprProperties, SortProperties};
3636
use datafusion_expr::{
3737
Coercion, ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature,
@@ -141,7 +141,67 @@ impl ScalarUDFImpl for RoundFunc {
141141
&default_decimal_places
142142
};
143143

144-
round_columnar(&args.args[0], decimal_places, args.number_rows)
144+
// Scalar fast path for float and decimal types - avoid array conversion overhead
145+
if let (ColumnarValue::Scalar(value_scalar), ColumnarValue::Scalar(dp_scalar)) =
146+
(&args.args[0], decimal_places)
147+
{
148+
if value_scalar.is_null() || dp_scalar.is_null() {
149+
return ColumnarValue::Scalar(ScalarValue::Null)
150+
.cast_to(args.return_type(), None);
151+
}
152+
153+
let dp = if let ScalarValue::Int32(Some(dp)) = dp_scalar {
154+
*dp
155+
} else {
156+
return internal_err!(
157+
"Unexpected datatype for decimal_places: {}",
158+
dp_scalar.data_type()
159+
);
160+
};
161+
162+
match value_scalar {
163+
ScalarValue::Float32(Some(v)) => {
164+
let rounded = round_float(*v, dp)?;
165+
Ok(ColumnarValue::Scalar(ScalarValue::from(rounded)))
166+
}
167+
ScalarValue::Float64(Some(v)) => {
168+
let rounded = round_float(*v, dp)?;
169+
Ok(ColumnarValue::Scalar(ScalarValue::from(rounded)))
170+
}
171+
ScalarValue::Decimal128(Some(v), precision, scale) => {
172+
let rounded = round_decimal(*v, *scale, dp)?;
173+
let scalar =
174+
ScalarValue::Decimal128(Some(rounded), *precision, *scale);
175+
Ok(ColumnarValue::Scalar(scalar))
176+
}
177+
ScalarValue::Decimal256(Some(v), precision, scale) => {
178+
let rounded = round_decimal(*v, *scale, dp)?;
179+
let scalar =
180+
ScalarValue::Decimal256(Some(rounded), *precision, *scale);
181+
Ok(ColumnarValue::Scalar(scalar))
182+
}
183+
ScalarValue::Decimal64(Some(v), precision, scale) => {
184+
let rounded = round_decimal(*v, *scale, dp)?;
185+
let scalar =
186+
ScalarValue::Decimal64(Some(rounded), *precision, *scale);
187+
Ok(ColumnarValue::Scalar(scalar))
188+
}
189+
ScalarValue::Decimal32(Some(v), precision, scale) => {
190+
let rounded = round_decimal(*v, *scale, dp)?;
191+
let scalar =
192+
ScalarValue::Decimal32(Some(rounded), *precision, *scale);
193+
Ok(ColumnarValue::Scalar(scalar))
194+
}
195+
_ => {
196+
internal_err!(
197+
"Unexpected datatype for value: {}",
198+
value_scalar.data_type()
199+
)
200+
}
201+
}
202+
} else {
203+
round_columnar(&args.args[0], decimal_places, args.number_rows)
204+
}
145205
}
146206

147207
fn output_ordering(&self, input: &[ExprProperties]) -> Result<SortProperties> {

0 commit comments

Comments
 (0)