This is an automated email from the ASF dual-hosted git repository.

tisonkun pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/datasketches-rust.git


The following commit(s) were added to refs/heads/main by this push:
     new bb6ba57  perf(xor): reduce peak construction memory (#316)
bb6ba57 is described below

commit bb6ba571e6e24a0b64365a2ef5825b4db18b1eff
Author: tison <[email protected]>
AuthorDate: Fri Oct 30 18:09:34 2026 +0800

    perf(xor): reduce peak construction memory (#316)
---
 CHANGELOG.md                   | 2 +-
 datasketches/src/xor/filter.rs | 6 ++++++
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 376478f..cbc6000 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,7 +6,7 @@ All significant changes to this project will be documented in 
this file.
 
 ### New features
 
-* Add immutable `XorFilter` and `XorFilterBuilder` behind the `xor` feature, 
with 8-, 16-, and 32-bit fingerprints, consuming builder configuration with 
complete-input construction and bounded seeded retries, optional cardinality 
metadata for imported filters, standard serialization for all three widths 
(Java-compatible for 8/16-bit filters), and raw parts for application-defined 
storage.
+* Add immutable `XorFilter` and `XorFilterBuilder` behind the `xor` feature, 
with 8-, 16-, and 32-bit fingerprints, consuming builder configuration with 
complete-input construction, bounded seeded retries and compact temporary 
storage, optional cardinality metadata for imported filters, standard 
serialization for all three widths (Java-compatible for 8/16-bit filters), and 
raw parts for application-defined storage.
 
 ### Improvements
 
diff --git a/datasketches/src/xor/filter.rs b/datasketches/src/xor/filter.rs
index 0762b40..0bf297b 100644
--- a/datasketches/src/xor/filter.rs
+++ b/datasketches/src/xor/filter.rs
@@ -405,6 +405,12 @@ impl XorFilter {
             .with_context("num_items", num_items));
         }
 
+        // Assignment only needs the peel stack. Release the graph before 
allocating the
+        // fingerprint payload so these allocations do not overlap at peak 
memory usage.
+        drop(xor_mask);
+        drop(count);
+        drop(queue);
+
         let fingerprints = match filter_type {
             XorFilterType::Xor8 => {
                 let mut values = vec![0_u8; capacity].into_boxed_slice();


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to