Skip to content

Remove unnecessary bit counting code from spark bit_count #18841

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
Nov 21, 2025
+6 −19
Merged

Remove unnecessary bit counting code from spark bit_count #18841

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
25 changes: 6 additions & 19 deletions datafusion/spark/src/function/bitwise/bit_count.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,24 +102,25 @@ fn spark_bit_count(value_array: &[ArrayRef]) -> Result<ArrayRef> {
DataType::Int8 => {
let result: Int32Array = value_array
.as_primitive::<Int8Type>()
.unary(|v| bit_count(v.into()));
.unary(|v| (v as i64).count_ones() as i32);
Ok(Arc::new(result))
}
DataType::Int16 => {
let result: Int32Array = value_array
.as_primitive::<Int16Type>()
.unary(|v| bit_count(v.into()));
.unary(|v| (v as i64).count_ones() as i32);
Ok(Arc::new(result))
}
DataType::Int32 => {
let result: Int32Array = value_array
.as_primitive::<Int32Type>()
.unary(|v| bit_count(v.into()));
.unary(|v| (v as i64).count_ones() as i32);
Ok(Arc::new(result))
}
DataType::Int64 => {
let result: Int32Array =
value_array.as_primitive::<Int64Type>().unary(bit_count);
let result: Int32Array = value_array
.as_primitive::<Int64Type>()
.unary(|v| v.count_ones() as i32);
Ok(Arc::new(result))
}
DataType::UInt8 => {
Expand Down Expand Up @@ -155,20 +156,6 @@ fn spark_bit_count(value_array: &[ArrayRef]) -> Result<ArrayRef> {
}
}

Copy link
Contributor Author

@pepijnve pepijnve Nov 21, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

FYI, this algorithm is a SWAR hamming weight implementation. Per the code comments in java.lang.Long, this comes from Hacker's Delight.

What's interesting is that the Rust compiler generates something very similar when calling count_ones().

With a sufficiently recent target architecture though you get popcnt instead.

Copy link
Contributor

@alamb alamb Nov 21, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The popcnt instruction is crazy fast -- we tested it in one example where we had a special codepath for no nulls, and I was worried that calculating the test if nulls.count_ones() == 0 would overwhelm the improvement

Nowhere close. 🚀

Hackers Delight is a classic -- I am not at all surprised that the Rust compiler includes all those tricks (and then some!)

// Here’s the equivalent Rust implementation of the bitCount function (similar to Apache Spark's bitCount for LongType)
// Spark: https://github.com/apache/spark/blob/ac717dd7aec665de578d7c6b0070e8fcdde3cea9/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/bitwiseExpressions.scala#L243
// Java impl: https://github.com/openjdk/jdk/blob/d226023643f90027a8980d161ec6d423887ae3ce/src/java.base/share/classes/java/lang/Long.java#L1584
fn bit_count(i: i64) -> i32 {
let mut u = i as u64;
u = u - ((u >> 1) & 0x5555555555555555);
u = (u & 0x3333333333333333) + ((u >> 2) & 0x3333333333333333);
u = (u + (u >> 4)) & 0x0f0f0f0f0f0f0f0f;
u = u + (u >> 8);
u = u + (u >> 16);
u = u + (u >> 32);
(u as i32) & 0x7f
}

#[cfg(test)]
mod tests {
use super::*;
Expand Down