Skip to content

Use Arrays.compareUnsigned in IDVersionSegmentTermsEnum and OrdsSegmentTermsEnum. #13782

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
vsop-479 merged 9 commits into from
Jan 22, 2026
Merged

Use Arrays.compareUnsigned in IDVersionSegmentTermsEnum and OrdsSegmentTermsEnum. #13782

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
6149b5d
Use Arrays.compareUnsigned instead of iterating compare in IDVersionS…
vsop-479 Sep 13, 2024
f46f5de
Add changes entry.
vsop-479 Sep 13, 2024
9917b0d
Move change entry to Optimizations.
vsop-479 Sep 13, 2024
cf2e7f6
Merge branch 'main' into use_arrays_compare_unsigned
vsop-479 Sep 13, 2024
4d3d219
Merge changes entries into one.
vsop-479 Sep 14, 2024
2cbf5ff
Use Arrays.compareUnsigned instead of loop compare suffix.
vsop-479 Sep 18, 2024
be35a14
Update change entry.
vsop-479 Sep 19, 2024
ef0e46b
Merge branch 'main' into use_arrays_compare_unsigned
vsop-479 Jan 5, 2026
a8545c1
Split changes entry.
vsop-479 Jan 21, 2026
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
2 changes: 2 additions & 0 deletions lucene/CHANGES.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -85,6 +85,8 @@ Improvements

Optimizations
---------------------
* GITHUB#13782: Replace handwritten loops compare with Arrays.compareUnsigned in TermsEnum and TermsEnumFrame classes. (Zhou Hui)

* GITHUB#14011: Reduce allocation rate in HNSW concurrent merge. (Viliam Durina)
* GITHUB#14022: Optimize DFS marking of connected components in HNSW by reducing stack depth, improving performance and reducing allocations. (Viswanath Kuchibhotla)

Expand Down
88 changes: 33 additions & 55 deletions lucene/codecs/src/java/org/apache/lucene/codecs/blocktreeords/OrdsSegmentTermsEnum.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@

import java.io.IOException;
import java.io.PrintStream;
import java.util.Arrays;
import org.apache.lucene.codecs.BlockTermState;
import org.apache.lucene.codecs.blocktreeords.FSTOrdsOutputs.Output;
import org.apache.lucene.index.BaseTermsEnum;
Expand Down Expand Up @@ -183,19 +184,19 @@ OrdsSegmentTermsEnumFrame pushFrame(FST.Arc<Output> arc, long fp, int length, lo
// " isFloor?=" + f.isFloor + " hasTerms=" + f.hasTerms + " pref=" + term + " nextEnt=" +
// f.nextEnt + " targetBeforeCurrentLength=" + targetBeforeCurrentLength + " term.length=" +
// term.length + " vs prefix=" + f.prefix);
if (f.prefix > targetBeforeCurrentLength) {
if (f.prefixLength > targetBeforeCurrentLength) {
// System.out.println(" do rewind!");
f.rewind();
} else {
// if (DEBUG) {
// System.out.println(" skip rewind!");
// }
}
assert length == f.prefix;
assert length == f.prefixLength;
assert termOrd == f.termOrdOrig;
} else {
f.nextEnt = -1;
f.prefix = length;
f.prefixLength = length;
f.state.termBlockOrd = 0;
f.termOrdOrig = termOrd;
// System.out.println("set termOrdOrig=" + termOrd);
Expand Down Expand Up @@ -308,31 +309,18 @@ public boolean seekExact(final BytesRef target) throws IOException {
}

if (cmp == 0) {
final int targetUptoMid = targetUpto;

// Second compare the rest of the term, but
// don't save arc/output/frame; we only do this
// to find out if the target term is before,
// equal or after the current term
final int targetLimit2 = Math.min(target.length, term.length());
while (targetUpto < targetLimit2) {
cmp =
(term.byteAt(targetUpto) & 0xFF) - (target.bytes[target.offset + targetUpto] & 0xFF);
// if (DEBUG) {
// System.out.println(" cycle2 targetUpto=" + targetUpto + " (vs limit=" +
// targetLimit + ") cmp=" + cmp + " (targetLabel=" + (char) (target.bytes[target.offset +
// targetUpto]) + " vs termLabel=" + (char) (term.bytes[targetUpto]) + ")");
// }
if (cmp != 0) {
break;
}
targetUpto++;
}

if (cmp == 0) {
cmp = term.length() - target.length;
}
targetUpto = targetUptoMid;
cmp =
Arrays.compareUnsigned(
term.bytes(),
targetUpto,
term.length(),
target.bytes,
target.offset + targetUpto,
target.offset + target.length);
}

if (cmp < 0) {
Expand Down Expand Up @@ -424,7 +412,7 @@ public boolean seekExact(final BytesRef target) throws IOException {
// toHex(targetLabel));
// }

validIndexPrefix = currentFrame.prefix;
validIndexPrefix = currentFrame.prefixLength;
// validIndexPrefix = targetUpto;

currentFrame.scanToFloorFrame(target);
Expand Down Expand Up @@ -484,7 +472,7 @@ public boolean seekExact(final BytesRef target) throws IOException {
}

// validIndexPrefix = targetUpto;
validIndexPrefix = currentFrame.prefix;
validIndexPrefix = currentFrame.prefixLength;

currentFrame.scanToFloorFrame(target);

Expand Down Expand Up @@ -600,28 +588,16 @@ public SeekStatus seekCeil(final BytesRef target) throws IOException {
}

if (cmp == 0) {
final int targetUptoMid = targetUpto;
// Second compare the rest of the term, but
// don't save arc/output/frame:
final int targetLimit2 = Math.min(target.length, term.length());
while (targetUpto < targetLimit2) {
cmp =
(term.byteAt(targetUpto) & 0xFF) - (target.bytes[target.offset + targetUpto] & 0xFF);
// if (DEBUG) {
// System.out.println(" cycle2 targetUpto=" + targetUpto + " (vs limit=" + targetLimit
// + ") cmp=" + cmp + " (targetLabel=" + (char) (target.bytes[target.offset + targetUpto])
// + " vs termLabel=" + (char) (term.bytes[targetUpto]) + ")");
// }
if (cmp != 0) {
break;
}
targetUpto++;
}

if (cmp == 0) {
cmp = term.length() - target.length;
}
targetUpto = targetUptoMid;
cmp =
Arrays.compareUnsigned(
term.bytes(),
targetUpto,
term.length(),
target.bytes,
target.offset + targetUpto,
target.offset + target.length);
}

if (cmp < 0) {
Expand Down Expand Up @@ -710,7 +686,7 @@ public SeekStatus seekCeil(final BytesRef target) throws IOException {
// toHex(targetLabel));
// }

validIndexPrefix = currentFrame.prefix;
validIndexPrefix = currentFrame.prefixLength;
// validIndexPrefix = targetUpto;

currentFrame.scanToFloorFrame(target);
Expand Down Expand Up @@ -771,7 +747,7 @@ public SeekStatus seekCeil(final BytesRef target) throws IOException {
}

// validIndexPrefix = targetUpto;
validIndexPrefix = currentFrame.prefix;
validIndexPrefix = currentFrame.prefixLength;

currentFrame.scanToFloorFrame(target);

Expand Down Expand Up @@ -809,7 +785,7 @@ private void printSeekState(PrintStream out) throws IOException {
while (true) {
OrdsSegmentTermsEnumFrame f = getFrame(ord);
assert f != null;
final BytesRef prefix = new BytesRef(term.bytes(), 0, f.prefix);
final BytesRef prefix = new BytesRef(term.bytes(), 0, f.prefixLength);
if (f.nextEnt == -1) {
out.println(
" frame "
Expand All @@ -820,7 +796,7 @@ private void printSeekState(PrintStream out) throws IOException {
+ f.fp
+ (f.isFloor ? (" (fpOrig=" + f.fpOrig + ")") : "")
+ " prefixLen="
+ f.prefix
+ f.prefixLength
+ " prefix="
+ ToStringUtils.bytesRefToString(prefix)
+ (f.nextEnt == -1 ? "" : (" (of " + f.entCount + ")"))
Expand Down Expand Up @@ -850,7 +826,7 @@ private void printSeekState(PrintStream out) throws IOException {
+ f.fp
+ (f.isFloor ? (" (fpOrig=" + f.fpOrig + ")") : "")
+ " prefixLen="
+ f.prefix
+ f.prefixLength
+ " prefix="
+ ToStringUtils.bytesRefToString(prefix)
+ " nextEnt="
Expand All @@ -877,12 +853,14 @@ private void printSeekState(PrintStream out) throws IOException {
}
if (fr.index != null) {
assert !isSeekFrame || f.arc != null : "isSeekFrame=" + isSeekFrame + " f.arc=" + f.arc;
if (f.prefix > 0 && isSeekFrame && f.arc.label() != (term.byteAt(f.prefix - 1) & 0xFF)) {
if (f.prefixLength > 0
&& isSeekFrame
&& f.arc.label() != (term.byteAt(f.prefixLength - 1) & 0xFF)) {
out.println(
" broken seek state: arc.label="
+ (char) f.arc.label()
+ " vs term byte="
+ (char) (term.byteAt(f.prefix - 1) & 0xFF));
+ (char) (term.byteAt(f.prefixLength - 1) & 0xFF));
throw new RuntimeException("seek state is broken");
}
Output output = Util.get(fr.index, prefix);
Expand Down Expand Up @@ -911,7 +889,7 @@ private void printSeekState(PrintStream out) throws IOException {
if (f == currentFrame) {
break;
}
if (f.prefix == validIndexPrefix) {
if (f.prefixLength == validIndexPrefix) {
isSeekFrame = false;
}
ord++;
Expand Down Expand Up @@ -993,7 +971,7 @@ public BytesRef next() throws IOException {

// Note that the seek state (last seek) has been
// invalidated beyond this depth
validIndexPrefix = Math.min(validIndexPrefix, currentFrame.prefix);
validIndexPrefix = Math.min(validIndexPrefix, currentFrame.prefixLength);
// if (DEBUG) {
// System.out.println(" reset validIndexPrefix=" + validIndexPrefix);
// }
Expand Down
Loading