Skip to content

Pack float32 comparison results into bitmap words #10741

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

Draft
connortsui20 wants to merge 1 commit into
base: develop
Choose a base branch
from
+7 −7
Draft

Pack float32 comparison results into bitmap words #10741

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
14 changes: 7 additions & 7 deletions vortex-array/src/scalar_fn/fns/binary/compare/primitive.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: Apache-2.0
// SPDX-FileCopyrightText: Copyright the Vortex contributors

//! Native comparison of primitive arrays with specialized bitmap packing for 8-bit inputs.
//! Native comparison of primitive arrays with packed bitmap output.

use vortex_buffer::BitBuffer;
use vortex_buffer::BufferAllocatorRef;
Expand Down Expand Up @@ -30,8 +30,8 @@ use crate::scalar_fn::fns::operators::CompareOperator;

/// Compare two primitive arrays of the same [`PType`].
///
/// Floats compare with Vortex's total ordering: `NaN` is the largest value, `-0.0 < +0.0`, and
/// equality is bitwise.
/// Floats use a total ordering over their bit patterns, including NaN signs and payloads.
/// Equality is bitwise, so signed zeros and distinct NaN payloads remain distinct.
pub(super) fn compare_primitive(
lhs: &ArrayRef,
rhs: &ArrayRef,
Expand Down Expand Up @@ -153,9 +153,9 @@ fn collect_bits_dispatch<T: NativePType>(
f: impl Fn(T) -> bool,
allocator: &BufferAllocatorRef,
) -> BitBuffer {
// This type check folds away during monomorphization. Wider masks keep the lane kernel:
// byte packing regresses 64-bit comparisons on AVX2.
if matches!(T::PTYPE, PType::I8 | PType::U8) {
// This type check folds away during monomorphization. Word packing improves float32
// comparisons; wider types retain the lane kernel to avoid AVX2 constant-case regressions.
if matches!(T::PTYPE, PType::I8 | PType::U8 | PType::F32) {
collect_bits_narrow(values, f, allocator)
} else {
collect_bits(values, f, allocator)
Expand All @@ -168,7 +168,7 @@ fn collect_zip_bits_dispatch<T: NativePType>(
f: impl Fn(T, T) -> bool,
allocator: &BufferAllocatorRef,
) -> BitBuffer {
if matches!(T::PTYPE, PType::I8 | PType::U8) {
if matches!(T::PTYPE, PType::I8 | PType::U8 | PType::F32) {
collect_zip_bits_narrow(lhs, rhs, f, allocator)
} else {
collect_zip_bits(lhs, rhs, f, allocator)
Expand Down
Loading