Version bump to 0.8.3 [skip ci]

Simplified logic for HNSW vacuum progress [skip ci]
Improved HNSW vacuum progress for round block sizes [skip ci]
2026-07-22 03:57:34 +08:00 · 2026-06-17 21:26:13 -07:00 · 2026-06-17 12:18:17 -07:00 · 2026-06-16 14:21:23 -07:00 · 2026-06-16 13:53:40 -07:00 · 2026-06-16 12:33:25 -07:00
44 changed files with 956 additions and 306 deletions
--- a/.dockerignore
+++ b/.dockerignore
@@ -1,8 +0,0 @@
-/.git/
-/dist/
-/results/
-/tmp_check/
-/sql/vector--?.?.?.sql
-regression.*
-*.o
-*.so
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -8,27 +8,29 @@ jobs:
      fail-fast: false
      matrix:
        include:
+          - postgres: 19
+            os: ubuntu-24.04
          - postgres: 18
            os: ubuntu-24.04
          - postgres: 17
            os: ubuntu-24.04
          - postgres: 16
-            os: ubuntu-22.04
+            os: ubuntu-24.04-arm
          - postgres: 15
            os: ubuntu-22.04
          - postgres: 14
-            os: ubuntu-20.04
+            os: ubuntu-22.04-arm
          - postgres: 13
-            os: ubuntu-20.04
+            os: ubuntu-22.04
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
      - uses: ankane/setup-postgres@v1
        with:
          postgres-version: ${{ matrix.postgres }}
          dev-files: true
      - run: make
        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare
+          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-sign-compare ${{ matrix.postgres >= 18 && '-Wno-missing-field-initializers' || '' }}
      - run: |
          export PG_CONFIG=`which pg_config`
          sudo --preserve-env=PG_CONFIG make install
@@ -46,18 +48,18 @@ jobs:
      fail-fast: false
      matrix:
        include:
-          - postgres: 17
-            os: macos-15
+          - postgres: 18
+            os: macos-26
          - postgres: 14
-            os: macos-13
+            os: macos-15-intel
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
      - uses: ankane/setup-postgres@v1
        with:
          postgres-version: ${{ matrix.postgres }}
      - run: make
        env:
-          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter
+          PG_CFLAGS: -DUSE_ASSERT_CHECKING -Wall -Wextra -Werror -Wno-unused-parameter -Wno-unknown-warning-option ${{ matrix.postgres >= 18 && '-Wno-missing-field-initializers' || '' }}
      - run: make install
      - run: make installcheck
      - if: ${{ failure() }}
@@ -70,27 +72,35 @@ jobs:
          tar xf $TAG.tar.gz
          mv postgres-$TAG postgres
        env:
-          TAG: ${{ matrix.postgres == 17 && 'REL_17_2' || 'REL_14_15' }}
+          TAG: ${{ matrix.postgres == 18 && 'REL_18_2' || 'REL_14_21' }}
      - run: make prove_installcheck PROVE_FLAGS="-I ./postgres/src/test/perl -I ./test/perl"
        env:
          PERL5LIB: /Users/runner/perl5/lib/perl5
      - run: make clean && $(brew --prefix llvm@$LLVM_VERSION)/bin/scan-build --status-bugs make
        env:
-          LLVM_VERSION: ${{ matrix.os == 'macos-15' && 18 || 15 }}
+          LLVM_VERSION: ${{ matrix.os == 'macos-26' && 20 || 18 }}
          PG_CFLAGS: -DUSE_ASSERT_CHECKING
  windows:
-    runs-on: windows-latest
+    runs-on: ${{ matrix.os }}
    if: ${{ !startsWith(github.ref_name, 'mac') }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - postgres: 17
+            os: windows-2025
+          - postgres: 14
+            os: windows-2022
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
      - uses: ankane/setup-postgres@v1
        with:
-          postgres-version: 14
+          postgres-version: ${{ matrix.postgres }}
      - run: |
-          call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
+          call "C:\Program Files\Microsoft Visual Studio\${{ matrix.os == 'windows-2025' && 18 || 2022 }}\Enterprise\VC\Auxiliary\Build\vcvars64.bat" && ^
          nmake /NOLOGO /F Makefile.win && ^
          nmake /NOLOGO /F Makefile.win install && ^
-          nmake /NOLOGO /F Makefile.win installcheck && ^
+          nmake /NOLOGO /F Makefile.win installcheck ${{ matrix.postgres != 17 && 'PG_REGRESS=$(PGROOT)\bin\pg_regress' || '' }} && ^
          nmake /NOLOGO /F Makefile.win clean && ^
          nmake /NOLOGO /F Makefile.win uninstall
        shell: cmd
@@ -123,10 +133,10 @@ jobs:
    if: ${{ !startsWith(github.ref_name, 'mac') && !startsWith(github.ref_name, 'windows') }}
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
      - uses: ankane/setup-postgres-valgrind@v1
        with:
-          postgres-version: 17
+          postgres-version: 18
          check-ub: yes
      - run: make OPTFLAGS=""
      - run: sudo --preserve-env=PG_CONFIG make install
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,19 @@
+## 0.8.3 (2026-06-17)
+
+- Fixed possible index corruption with HNSW vacuuming
+- Fixed performance regression with Hamming distance and Jaccard distance with Postgres 18
+
+## 0.8.2 (2026-02-25)
+
+- Fixed buffer overflow with parallel HNSW index build - [more info](https://github.com/pgvector/pgvector/issues/959)
+- Improved `install` target on Windows
+- Fixed `Index Searches` in `EXPLAIN` output for Postgres 18
+
+## 0.8.1 (2025-09-04)
+
+- Added support for Postgres 18 rc1
+- Improved performance of `binary_quantize` function
+
 ## 0.8.0 (2024-10-30)

 - Added support for iterative index scans
--- a/7
+++ b/7
@@ -1,8 +1,11 @@
+# syntax=docker/dockerfile:1
+
 ARG PG_MAJOR=17
-FROM postgres:$PG_MAJOR
+ARG DEBIAN_CODENAME=bookworm
+FROM postgres:$PG_MAJOR-$DEBIAN_CODENAME
 ARG PG_MAJOR

-COPY . /tmp/pgvector
+ADD https://github.com/pgvector/pgvector.git#v0.8.3 /tmp/pgvector

 RUN apt-get update && \
 		apt-mark hold locales && \
--- a/2
+++ b/2
@@ -1,4 +1,4 @@
-Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
+Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group

 Portions Copyright (c) 1994, The Regents of the University of California

--- a/META.json
+++ b/META.json
@@ -2,7 +2,7 @@
 	"name": "vector",
 	"abstract": "Open-source vector similarity search for Postgres",
 	"description": "Supports L2 distance, inner product, and cosine distance",
-	"version": "0.8.0",
+	"version": "0.8.3",
 	"maintainer": [
 		"Andrew Kane <andrew@ankane.org>"
 	],
@@ -20,7 +20,7 @@
 		"vector": {
 			"file": "sql/vector.sql",
 			"docfile": "README.md",
-			"version": "0.8.0",
+			"version": "0.8.3",
 			"abstract": "Open-source vector similarity search for Postgres"
 		}
 	},
--- a/14
+++ b/14
@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.8.0
+EXTVERSION = 0.8.3

 MODULE_big = vector
 DATA = $(wildcard sql/*--*--*.sql)
@@ -27,6 +27,11 @@ ifneq ($(filter ppc64%, $(shell uname -m)), )
 	OPTFLAGS =
 endif

+# RISC-V64 doesn't support -march=native
+ifeq ($(shell uname -m), riscv64)
+	OPTFLAGS =
+endif
+
 # For auto-vectorization:
 # - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html
 # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html
@@ -76,4 +81,9 @@ docker:
 .PHONY: docker-release

 docker-release:
-	docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) .
+	docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) --build-arg DEBIAN_CODENAME=bookworm -t pgvector/pgvector:pg$(PG_MAJOR) -t pgvector/pgvector:pg$(PG_MAJOR)-bookworm -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR) -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR)-bookworm .
+
+.PHONY: docker-release-trixie
+
+docker-release-trixie:
+	docker buildx build --push --pull --no-cache --platform linux/amd64,linux/arm64 --build-arg PG_MAJOR=$(PG_MAJOR) --build-arg DEBIAN_CODENAME=trixie -t pgvector/pgvector:pg$(PG_MAJOR)-trixie -t pgvector/pgvector:$(EXTVERSION)-pg$(PG_MAJOR)-trixie .
--- a/Makefile.win
+++ b/Makefile.win
@@ -1,5 +1,5 @@
 EXTENSION = vector
-EXTVERSION = 0.8.0
+EXTVERSION = 0.8.3

 DATA_built = sql\$(EXTENSION)--$(EXTVERSION).sql
 OBJS = src\bitutils.obj src\bitvec.obj src\halfutils.obj src\halfvec.obj src\hnsw.obj src\hnswbuild.obj src\hnswinsert.obj src\hnswscan.obj src\hnswutils.obj src\hnswvacuum.obj src\ivfbuild.obj src\ivfflat.obj src\ivfinsert.obj src\ivfkmeans.obj src\ivfscan.obj src\ivfutils.obj src\ivfvacuum.obj src\sparsevec.obj src\vector.obj
@@ -31,6 +31,9 @@ LIBDIR = $(PGROOT)\lib
 PKGLIBDIR = $(PGROOT)\lib
 SHAREDIR = $(PGROOT)\share

+# Use $(PGROOT)\bin\pg_regress for Postgres < 17
+PG_REGRESS = $(LIBDIR)\pgxs\src\test\regress\pg_regress
+
 CFLAGS = /nologo /I"$(INCLUDEDIR_SERVER)\port\win32_msvc" /I"$(INCLUDEDIR_SERVER)\port\win32" /I"$(INCLUDEDIR_SERVER)" /I"$(INCLUDEDIR)"

 CFLAGS = $(CFLAGS) $(PG_CFLAGS)
@@ -54,11 +57,11 @@ install: all
 	copy $(SHLIB) "$(PKGLIBDIR)"
 	copy $(EXTENSION).control "$(SHAREDIR)\extension"
 	copy sql\$(EXTENSION)--*.sql "$(SHAREDIR)\extension"
-	mkdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
+	if not exist "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)" mkdir "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"
 	for %f in ($(HEADERS)) do copy %f "$(INCLUDEDIR_SERVER)\extension\$(EXTENSION)"

 installcheck:
-	"$(BINDIR)\pg_regress" --bindir="$(BINDIR)" $(REGRESS_OPTS) $(REGRESS)
+	"$(PG_REGRESS)" --bindir="$(BINDIR)" $(REGRESS_OPTS) $(REGRESS)

 uninstall:
 	del /f "$(PKGLIBDIR)\$(SHLIB)"
--- a/README.md
+++ b/README.md
@@ -11,6 +11,8 @@ Store your vectors with the rest of your data. Supports:

 Plus [ACID](https://en.wikipedia.org/wiki/ACID) compliance, point-in-time recovery, JOINs, and all of the other [great features](https://www.postgresql.org/about/) of Postgres

+Have a lot of vectors? Use [quantization](#scaling) to scale
+
 [![Build Status](https://github.com/pgvector/pgvector/actions/workflows/build.yml/badge.svg)](https://github.com/pgvector/pgvector/actions)

 ## Installation
@@ -21,7 +23,7 @@ Compile and install the extension (supports Postgres 13+)

 ```sh
 cd /tmp
-git clone --branch v0.8.0 https://github.com/pgvector/pgvector.git
+git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
 cd pgvector
 make
 make install # may need sudo
@@ -29,31 +31,21 @@ make install # may need sudo

 See the [installation notes](#installation-notes---linux-and-mac) if you run into issues

-You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), [pkg](#pkg), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).
+You can also install it with [Docker](#docker), [Homebrew](#homebrew), [PGXN](#pgxn), [APT](#apt), [Yum](#yum), [pkg](#pkg), [APK](#apk), or [conda-forge](#conda-forge), and it comes preinstalled with [Postgres.app](#postgresapp) and many [hosted providers](#hosted-postgres). There are also instructions for [GitHub Actions](https://github.com/pgvector/setup-pgvector).

 ### Windows

-Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170#download-and-install-the-tools) is installed, and run:
+Ensure [C++ support in Visual Studio](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170#download-and-install-the-tools) is installed and run `x64 Native Tools Command Prompt for VS [version]` as administrator. Then use `nmake` to build:

 ```cmd
-call "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"
-```
-
-Note: The exact path will vary depending on your Visual Studio version and edition
-
-Then use `nmake` to build:
-
-```cmd
-set "PGROOT=C:\Program Files\PostgreSQL\16"
+set "PGROOT=C:\Program Files\PostgreSQL\18"
 cd %TEMP%
-git clone --branch v0.8.0 https://github.com/pgvector/pgvector.git
+git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
 nmake /F Makefile.win install
 ```

-Note: Postgres 17 is not supported with MSVC yet due to an [upstream issue](https://www.postgresql.org/message-id/flat/CAOdR5yF0krWrxycA04rgUKCgKugRvGWzzGLAhDZ9bzNv8g0Lag%40mail.gmail.com)
-
 See the [installation notes](#installation-notes---windows) if you run into issues

 You can also install it with [Docker](#docker) or [conda-forge](#conda-forge).
@@ -84,7 +76,7 @@ Get the nearest neighbors by L2 distance
 SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
 ```

-Also supports inner product (`<#>`), cosine distance (`<=>`), and L1 distance (`<+>`, added in 0.7.0)
+Also supports inner product (`<#>`), cosine distance (`<=>`), and L1 distance (`<+>`)

 Note: `<#>` returns the negative inner product since Postgres only supports `ASC` order index scans on operators

@@ -148,9 +140,9 @@ Supported distance functions are:
 - `<->` - L2 distance
 - `<#>` - (negative) inner product
 - `<=>` - cosine distance
- `<+>` - L1 distance (added in 0.7.0)
- `<~>` - Hamming distance (binary vectors, added in 0.7.0)
- `<%>` - Jaccard distance (binary vectors, added in 0.7.0)
+- `<+>` - L1 distance
+- `<~>` - Hamming distance (binary vectors)
+- `<%>` - Jaccard distance (binary vectors)

 Get the nearest neighbors to a row

@@ -237,19 +229,19 @@ Cosine distance
 CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
 ```

-L1 distance - added in 0.7.0
+L1 distance

 ```sql
 CREATE INDEX ON items USING hnsw (embedding vector_l1_ops);
 ```

-Hamming distance - added in 0.7.0
+Hamming distance

 ```sql
 CREATE INDEX ON items USING hnsw (embedding bit_hamming_ops);
 ```

-Jaccard distance - added in 0.7.0
+Jaccard distance

 ```sql
 CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
@@ -258,9 +250,9 @@ CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
 Supported types are:

 - `vector` - up to 2,000 dimensions
- `halfvec` - up to 4,000 dimensions (added in 0.7.0)
- `bit` - up to 64,000 dimensions (added in 0.7.0)
- `sparsevec` - up to 1,000 non-zero elements (added in 0.7.0)
+- `halfvec` - up to 4,000 dimensions
+- `bit` - up to 64,000 dimensions
+- `sparsevec` - up to 1,000 non-zero elements

 ### Index Options

@@ -314,13 +306,17 @@ Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on t

 Like other index types, it’s faster to create an index after loading your initial data

-Starting with 0.6.0, you can also speed up index creation by increasing the number of parallel workers (2 by default)
+You can also speed up index creation by increasing the number of parallel workers (2 by default)

 ```sql
 SET max_parallel_maintenance_workers = 7; -- plus leader
 ```

-For a large number of workers, you may also need to increase `max_parallel_workers` (8 by default)
+For a large number of workers, you may need to increase `max_parallel_workers` (8 by default)
+
+The [index options](#index-options) also have a significant impact on build time (use the defaults unless seeing low recall)
+
+Use [binary quantization](#binary-quantization) for faster build times at scale

 ### Indexing Progress

@@ -367,7 +363,7 @@ Cosine distance
 CREATE INDEX ON items USING ivfflat (embedding vector_cosine_ops) WITH (lists = 100);
 ```

-Hamming distance - added in 0.7.0
+Hamming distance

 ```sql
 CREATE INDEX ON items USING ivfflat (embedding bit_hamming_ops) WITH (lists = 100);
@@ -376,8 +372,8 @@ CREATE INDEX ON items USING ivfflat (embedding bit_hamming_ops) WITH (lists = 10
 Supported types are:

 - `vector` - up to 2,000 dimensions
- `halfvec` - up to 4,000 dimensions (added in 0.7.0)
- `bit` - up to 64,000 dimensions (added in 0.7.0)
+- `halfvec` - up to 4,000 dimensions
+- `bit` - up to 64,000 dimensions

 ### Query Options

@@ -451,13 +447,7 @@ Exact indexes work well for conditions that match a low percentage of rows. Othe
 CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
 ```

-With approximate indexes, filtering is applied *after* the index is scanned. If a condition matches 10% of rows, with HNSW and the default `hnsw.ef_search` of 40, only 4 rows will match on average. For more rows, increase `hnsw.ef_search`.
-
-```sql
-SET hnsw.ef_search = 200;
-```
-
-Starting with 0.8.0, you can enable [iterative index scans](#iterative-index-scans), which will automatically scan more of the index when needed.
+With approximate indexes, filtering is applied *after* the index is scanned. If a condition matches 10% of rows, with HNSW and the default `hnsw.ef_search` of 40, only 4 rows will match on average. For more rows, enable [iterative index scans](#iterative-index-scans), which will automatically scan more of the index when needed.

 ```sql
 SET hnsw.iterative_scan = strict_order;
@@ -477,8 +467,6 @@ CREATE TABLE items (embedding vector(3), category_id int) PARTITION BY LIST(cate

 ## Iterative Index Scans

-*Added in 0.8.0*
-
 With approximate indexes, queries with filtering can return less results since filtering is applied *after* the index is scanned. Starting with 0.8.0, you can enable iterative index scans, which will automatically scan more of the index until enough results are found (or it reaches `hnsw.max_scan_tuples` or `ivfflat.max_probes`).

 Iterative scans can use strict or relaxed ordering.
@@ -502,9 +490,11 @@ With relaxed ordering, you can use a [materialized CTE](https://www.postgresql.o
 ```sql
 WITH relaxed_results AS MATERIALIZED (
    SELECT id, embedding <-> '[1,2,3]' AS distance FROM items WHERE category_id = 123 ORDER BY distance LIMIT 5
-) SELECT * FROM relaxed_results ORDER BY distance;
+) SELECT * FROM relaxed_results ORDER BY distance + 0;
 ```

+Note: `+ 0` is needed for Postgres 17+
+
 For queries that filter by distance, use a materialized CTE and place the distance filter outside of it for best performance (due to the [current behavior](https://www.postgresql.org/message-id/flat/CAOdR5yGUoMQ6j7M5hNUXrySzaqZVGf_Ne%2B8fwZMRKTFxU1nbJg%40mail.gmail.com) of the Postgres executor)

 ```sql
@@ -549,8 +539,6 @@ Note: If this is lower than `ivfflat.probes`, `ivfflat.probes` will be used

 ## Half-Precision Vectors

-*Added in 0.7.0*
-
 Use the `halfvec` type to store half-precision vectors

 ```sql
@@ -559,8 +547,6 @@ CREATE TABLE items (id bigserial PRIMARY KEY, embedding halfvec(3));

 ## Half-Precision Indexing

-*Added in 0.7.0*
-
 Index vectors at half precision for smaller indexes

 ```sql
@@ -582,24 +568,16 @@ CREATE TABLE items (id bigserial PRIMARY KEY, embedding bit(3));
 INSERT INTO items (embedding) VALUES ('000'), ('111');
 ```

-Get the nearest neighbors by Hamming distance (added in 0.7.0)
+Get the nearest neighbors by Hamming distance

 ```sql
 SELECT * FROM items ORDER BY embedding <~> '101' LIMIT 5;
 ```

-Or (before 0.7.0)
-
-```sql
-SELECT * FROM items ORDER BY bit_count(embedding # '101') LIMIT 5;
-```
-
 Also supports Jaccard distance (`<%>`)

 ## Binary Quantization

-*Added in 0.7.0*
-
 Use expression indexing for binary quantization

 ```sql
@@ -622,8 +600,6 @@ SELECT * FROM (

 ## Sparse Vectors

-*Added in 0.7.0*
-
 Use the `sparsevec` type to store sparse vectors

 ```sql
@@ -657,8 +633,6 @@ You can use [Reciprocal Rank Fusion](https://github.com/pgvector/pgvector-python

 ## Indexing Subvectors

-*Added in 0.7.0*
-
 Use expression indexing to index subvectors

 ```sql
@@ -697,6 +671,10 @@ SHOW shared_buffers;

 Be sure to restart Postgres for changes to take effect.

+### Storing
+
+Use the `halfvec` type instead of `vector` for a smaller working set.
+
 ### Loading

 Use `COPY` for bulk loading data ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/loading/example.py)).
@@ -711,6 +689,8 @@ Add any indexes *after* loading the initial data for best performance.

 See index build time for [HNSW](#index-build-time) and [IVFFlat](#index-build-time-1).

+Use [binary quantization](#binary-quantization) for smaller indexes and faster build times at scale.
+
 In production environments, create indexes concurrently to avoid blocking writes.

 ```sql
@@ -719,10 +699,10 @@ CREATE INDEX CONCURRENTLY ...

 ### Querying

-Use `EXPLAIN ANALYZE` to debug performance.
+Use `EXPLAIN (ANALYZE, BUFFERS)` to debug performance.

 ```sql
-EXPLAIN ANALYZE SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
+EXPLAIN (ANALYZE, BUFFERS) SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
 ```

 #### Exact Search
@@ -741,6 +721,8 @@ SELECT * FROM items ORDER BY embedding <#> '[3,1,2]' LIMIT 5;

 #### Approximate Search

+Use [binary quantization](#binary-quantization) with re-ranking to keep indexes in-memory at scale.
+
 To speed up queries with an IVFFlat index, increase the number of inverted lists (at the expense of recall).

 ```sql
@@ -756,23 +738,20 @@ REINDEX INDEX CONCURRENTLY index_name;
 VACUUM table_name;
 ```

+## Scaling
+
+For a smaller working set:
+
+1. Use the `halfvec` type instead of `vector` for tables
+2. Use [binary quantization](#binary-quantization) for indexes (with re-ranking for search)
+
+Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
+
+Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus), [PgDog](https://github.com/pgdogdev/pgdog), or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus/example.py)).
+
 ## Monitoring

-Monitor performance with [pg_stat_statements](https://www.postgresql.org/docs/current/pgstatstatements.html) (be sure to add it to `shared_preload_libraries`).
-
-```sql
-CREATE EXTENSION pg_stat_statements;
-```
-
-Get the most time-consuming queries with:
-
-```sql
-SELECT query, calls, ROUND((total_plan_time + total_exec_time) / calls) AS avg_time_ms,
-    ROUND((total_plan_time + total_exec_time) / 60000) AS total_time_min
-    FROM pg_stat_statements ORDER BY total_plan_time + total_exec_time DESC LIMIT 20;
-```
-
-Note: Replace `total_plan_time + total_exec_time` with `total_time` for Postgres < 13
+Use existing tools like [pg_stat_statements](https://www.postgresql.org/docs/current/pgstatstatements.html) or [PgHero](https://github.com/ankane/pghero) to monitor performance.

 Monitor recall by comparing results from approximate search with exact search.

@@ -783,23 +762,18 @@ SELECT ...
 COMMIT;
 ```

-## Scaling
-
-Scale pgvector the same way you scale Postgres.
-
-Scale vertically by increasing memory, CPU, and storage on a single instance. Use existing tools to [tune parameters](#tuning) and [monitor performance](#monitoring).
-
-Scale horizontally with [replicas](https://www.postgresql.org/docs/current/hot-standby.html), or use [Citus](https://github.com/citusdata/citus) or another approach for sharding ([example](https://github.com/pgvector/pgvector-python/blob/master/examples/citus/example.py)).
-
 ## Languages

 Use pgvector from any language with a Postgres client. You can even generate and store vectors in one language and query them in another.

 Language | Libraries / Examples
 --- | ---
+Ada | [pgvector-ada](https://github.com/pgvector/pgvector-ada)
+Algol | [pgvector-algol](https://github.com/pgvector/pgvector-algol)
 C | [pgvector-c](https://github.com/pgvector/pgvector-c)
 C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp)
 C#, F#, Visual Basic | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
+COBOL | [pgvector-cobol](https://github.com/pgvector/pgvector-cobol)
 Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal)
 D | [pgvector-d](https://github.com/pgvector/pgvector-d)
 Dart | [pgvector-dart](https://github.com/pgvector/pgvector-dart)
@@ -811,19 +785,23 @@ Go | [pgvector-go](https://github.com/pgvector/pgvector-go)
 Haskell | [pgvector-haskell](https://github.com/pgvector/pgvector-haskell)
 Java, Kotlin, Groovy, Scala | [pgvector-java](https://github.com/pgvector/pgvector-java)
 JavaScript, TypeScript | [pgvector-node](https://github.com/pgvector/pgvector-node)
-Julia | [pgvector-julia](https://github.com/pgvector/pgvector-julia)
+Julia | [Pgvector.jl](https://github.com/pgvector/Pgvector.jl)
 Lisp | [pgvector-lisp](https://github.com/pgvector/pgvector-lisp)
 Lua | [pgvector-lua](https://github.com/pgvector/pgvector-lua)
 Nim | [pgvector-nim](https://github.com/pgvector/pgvector-nim)
 OCaml | [pgvector-ocaml](https://github.com/pgvector/pgvector-ocaml)
+Pascal | [pgvector-pascal](https://github.com/pgvector/pgvector-pascal)
 Perl | [pgvector-perl](https://github.com/pgvector/pgvector-perl)
 PHP | [pgvector-php](https://github.com/pgvector/pgvector-php)
+Prolog | [pgvector-prolog](https://github.com/pgvector/pgvector-prolog)
 Python | [pgvector-python](https://github.com/pgvector/pgvector-python)
 R | [pgvector-r](https://github.com/pgvector/pgvector-r)
+Racket | [pgvector-racket](https://github.com/pgvector/pgvector-racket)
 Raku | [pgvector-raku](https://github.com/pgvector/pgvector-raku)
 Ruby | [pgvector-ruby](https://github.com/pgvector/pgvector-ruby), [Neighbor](https://github.com/ankane/neighbor)
 Rust | [pgvector-rust](https://github.com/pgvector/pgvector-rust)
 Swift | [pgvector-swift](https://github.com/pgvector/pgvector-swift)
+Tcl | [pgvector-tcl](https://github.com/pgvector/pgvector-tcl)
 Zig | [pgvector-zig](https://github.com/pgvector/pgvector-zig)

 ## Frequently Asked Questions
@@ -838,11 +816,11 @@ Yes, pgvector uses the write-ahead log (WAL), which allows for replication and p

 #### What if I want to index vectors with more than 2,000 dimensions?

-You can use [half-precision indexing](#half-precision-indexing) to index up to 4,000 dimensions or [binary quantization](#binary-quantization) to index up to 64,000 dimensions. Another option is [dimensionality reduction](https://en.wikipedia.org/wiki/Dimensionality_reduction).
+You can use [half-precision vectors](#half-precision-vectors) or [half-precision indexing](#half-precision-indexing) to index up to 4,000 dimensions or [binary quantization](#binary-quantization) to index up to 64,000 dimensions. Other options are [indexing subvectors](#indexing-subvectors) (for models that support it) or [dimensionality reduction](https://en.wikipedia.org/wiki/Dimensionality_reduction).

 #### Can I store vectors with different dimensions in the same column?

-You can use `vector` as the type (instead of `vector(3)`).
+You can use `vector` as the type (instead of `vector(n)`).

 ```sql
 CREATE TABLE embeddings (model_id bigint, item_id bigint, embedding vector, PRIMARY KEY (model_id, item_id));
@@ -897,6 +875,8 @@ No, but like other index types, you’ll likely see better performance if they d
 SELECT pg_size_pretty(pg_relation_size('index_name'));
 ```

+Use [half-precision indexing](#half-precision-indexing) or [binary quantization](#binary-quantization) for smaller indexes.
+
 ## Troubleshooting

 #### Why isn’t a query using an index?
@@ -1090,7 +1070,7 @@ l2_normalize(sparsevec) → sparsevec | Normalize with Euclidean norm | 0.7.0
 If your machine has multiple Postgres installations, specify the path to [pg_config](https://www.postgresql.org/docs/current/app-pgconfig.html) with:

 ```sh
-export PG_CONFIG=/Library/PostgreSQL/17/bin/pg_config
+export PG_CONFIG=/Library/PostgreSQL/18/bin/pg_config
 ```

 Then re-run the installation instructions (run `make clean` before `make` if needed). If `sudo` is needed for `make install`, use:
@@ -1101,11 +1081,11 @@ sudo --preserve-env=PG_CONFIG make install

 A few common paths on Mac are:

- EDB installer - `/Library/PostgreSQL/17/bin/pg_config`
- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@17/bin/pg_config`
- Homebrew (x86-64) - `/usr/local/opt/postgresql@17/bin/pg_config`
+- EDB installer - `/Library/PostgreSQL/18/bin/pg_config`
+- Homebrew (arm64) - `/opt/homebrew/opt/postgresql@18/bin/pg_config`
+- Homebrew (x86-64) - `/usr/local/opt/postgresql@18/bin/pg_config`

-Note: Replace `17` with your Postgres server version
+Note: Replace `18` with your Postgres server version

 ### Missing Header

@@ -1114,14 +1094,20 @@ If compilation fails with `fatal error: postgres.h: No such file or directory`,
 For Ubuntu and Debian, use:

 ```sh
-sudo apt install postgresql-server-dev-17
+sudo apt install postgresql-server-dev-18
 ```

-Note: Replace `17` with your Postgres server version
+Note: Replace `18` with your Postgres server version

 ### Missing SDK

-If compilation fails and the output includes `warning: no such sysroot directory` on Mac, reinstall Xcode Command Line Tools.
+If compilation fails and the output includes `warning: no such sysroot directory` on Mac, your Postgres installation points to a path that no longer exists.
+
+```sh
+pg_config --cppflags
+```
+
+Reinstall Postgres to fix this.

 ### Portability

@@ -1139,6 +1125,14 @@ make OPTFLAGS=""

 If compilation fails with `Cannot open include file: 'postgres.h': No such file or directory`, make sure `PGROOT` is correct.

+### Mismatched Architecture
+
+If compilation fails with `error C2196: case value '4' already used`, make sure you’re using the `x64 Native Tools Command Prompt`. Then run `nmake /F Makefile.win clean` and re-run the installation instructions.
+
+### Missing Symbol
+
+If linking fails with `unresolved external symbol float_to_shortest_decimal_bufn` with Postgres 17.0-17.2, upgrade to Postgres 17.3+.
+
 ### Permissions

 If installation fails with `Access is denied`, re-run the installation instructions as an administrator.
@@ -1150,17 +1144,38 @@ If installation fails with `Access is denied`, re-run the installation instructi
 Get the [Docker image](https://hub.docker.com/r/pgvector/pgvector) with:

 ```sh
-docker pull pgvector/pgvector:pg17
+docker pull pgvector/pgvector:pg18-trixie
 ```

-This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `17` with your Postgres server version, and run it the same way).
+This adds pgvector to the [Postgres image](https://hub.docker.com/_/postgres) (replace `18` with your Postgres server version, and run it the same way).
+
+Supported tags are:
+
+- `pg18-trixie`, `0.8.3-pg18-trixie`
+- `pg18-bookworm`, `0.8.3-pg18-bookworm`, `pg18`, `0.8.3-pg18`
+- `pg17-trixie`, `0.8.3-pg17-trixie`
+- `pg17-bookworm`, `0.8.3-pg17-bookworm`, `pg17`, `0.8.3-pg17`
+- `pg16-trixie`, `0.8.3-pg16-trixie`
+- `pg16-bookworm`, `0.8.3-pg16-bookworm`, `pg16`, `0.8.3-pg16`
+- `pg15-trixie`, `0.8.3-pg15-trixie`
+- `pg15-bookworm`, `0.8.3-pg15-bookworm`, `pg15`, `0.8.3-pg15`
+- `pg14-trixie`, `0.8.3-pg14-trixie`
+- `pg14-bookworm`, `0.8.3-pg14-bookworm`, `pg14`, `0.8.3-pg14`
+- `pg13-trixie`, `0.8.3-pg13-trixie`
+- `pg13-bookworm`, `0.8.3-pg13-bookworm`, `pg13`, `0.8.3-pg13`

 You can also build the image manually:

 ```sh
-git clone --branch v0.8.0 https://github.com/pgvector/pgvector.git
+git clone --branch v0.8.3 https://github.com/pgvector/pgvector.git
 cd pgvector
-docker build --pull --build-arg PG_MAJOR=17 -t myuser/pgvector .
+docker build --pull --build-arg PG_MAJOR=18 -t myuser/pgvector .
+```
+
+If you increase `maintenance_work_mem`, make sure `--shm-size` is at least that size to avoid an error with parallel HNSW index builds.
+
+```sh
+docker run --shm-size=1g ...
 ```

 ### Homebrew
@@ -1171,7 +1186,7 @@ With Homebrew Postgres, you can use:
 brew install pgvector
 ```

-Note: This only adds it to the `postgresql@17` and `postgresql@14` formulas
+Note: This only adds it to the `postgresql@18` and `postgresql@17` formulas

 ### PGXN

@@ -1186,29 +1201,29 @@ pgxn install vector
 Debian and Ubuntu packages are available from the [PostgreSQL APT Repository](https://wiki.postgresql.org/wiki/Apt). Follow the [setup instructions](https://wiki.postgresql.org/wiki/Apt#Quickstart) and run:

 ```sh
-sudo apt install postgresql-17-pgvector
+sudo apt install postgresql-18-pgvector
 ```

-Note: Replace `17` with your Postgres server version
+Note: Replace `18` with your Postgres server version

 ### Yum

 RPM packages are available from the [PostgreSQL Yum Repository](https://yum.postgresql.org/). Follow the [setup instructions](https://www.postgresql.org/download/linux/redhat/) for your distribution and run:

 ```sh
-sudo yum install pgvector_17
+sudo yum install pgvector_18
 # or
-sudo dnf install pgvector_17
+sudo dnf install pgvector_18
 ```

-Note: Replace `17` with your Postgres server version
+Note: Replace `18` with your Postgres server version

 ### pkg

 Install the FreeBSD package with:

 ```sh
-pkg install postgresql16-pgvector
+pkg install postgresql17-pgvector
 ```

 or the port with:
@@ -1218,6 +1233,14 @@ cd /usr/ports/databases/pgvector
 make install
 ```

+### APK
+
+Install the Alpine package with:
+
+```sh
+apk add postgresql-pgvector
+```
+
 ### conda-forge

 With Conda Postgres, install from [conda-forge](https://anaconda.org/conda-forge/pgvector) with:
@@ -1250,36 +1273,6 @@ You can check the version in the current database with:
 SELECT extversion FROM pg_extension WHERE extname = 'vector';
 ```

-## Upgrade Notes
-
-### 0.6.0
-
-#### Postgres 12
-
-If upgrading with Postgres 12, remove this line from `sql/vector--0.5.1--0.6.0.sql`:
-
-```sql
-ALTER TYPE vector SET (STORAGE = external);
-```
-
-Then run `make install` and `ALTER EXTENSION vector UPDATE;`.
-
-#### Docker
-
-The Docker image is now published in the `pgvector` org, and there are tags for each supported version of Postgres (rather than a `latest` tag).
-
-```sh
-docker pull pgvector/pgvector:pg16
-# or
-docker pull pgvector/pgvector:0.6.0-pg16
-```
-
-Also, if you’ve increased `maintenance_work_mem`, make sure `--shm-size` is at least that size to avoid an error with parallel HNSW index builds.
-
-```sh
-docker run --shm-size=1g ...
-```
-
 ## Thanks

 Thanks to:
@@ -1336,7 +1329,7 @@ make clean && PG_CFLAGS="-DUSE_ASSERT_CHECKING" make && make install
 To enable benchmarking:

 ```sh
-make clean && PG_CFLAGS="-DIVFFLAT_BENCH" make && make install
+make clean && PG_CFLAGS="-DHNSW_BENCH -DIVFFLAT_BENCH" make && make install
 ```

 To show memory usage:
--- a/sql/vector--0.8.0--0.8.1.sql
+++ b/sql/vector--0.8.0--0.8.1.sql
@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.1'" to load this file. \quit
--- a/sql/vector--0.8.1--0.8.2.sql
+++ b/sql/vector--0.8.1--0.8.2.sql
@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.2'" to load this file. \quit
--- a/sql/vector--0.8.2--0.8.3.sql
+++ b/sql/vector--0.8.2--0.8.3.sql
@@ -0,0 +1,2 @@
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "ALTER EXTENSION vector UPDATE TO '0.8.3'" to load this file. \quit
--- a/src/bitutils.c
+++ b/src/bitutils.c
@@ -31,10 +31,12 @@
 #define BIT_TARGET_CLONES
 #endif

-/* Use built-ins when possible for inlining */
-#if defined(HAVE__BUILTIN_POPCOUNT) && defined(HAVE_LONG_INT_64)
+/* Use built-ins when possible for Postgres < 19 for inlining */
+#if PG_VERSION_NUM >= 190000
+#define popcount64(x) pg_popcount64(x)
+#elif defined(HAVE__BUILTIN_POPCOUNT) && (defined(HAVE_LONG_INT_64) || SIZEOF_LONG == 8)
 #define popcount64(x) __builtin_popcountl(x)
-#elif defined(HAVE__BUILTIN_POPCOUNT) && defined(HAVE_LONG_LONG_INT_64)
+#elif defined(HAVE__BUILTIN_POPCOUNT) && (defined(HAVE_LONG_LONG_INT_64) || SIZEOF_LONG_LONG == 8)
 #define popcount64(x) __builtin_popcountll(x)
 #elif !defined(_MSC_VER)
 /* Fails to resolve with MSVC */
@@ -169,7 +171,7 @@ BitJaccardDistanceAvx512Popcount(uint32 bytes, unsigned char *ax, unsigned char
 #endif

 TARGET_XSAVE static bool
-SupportsAvx512Popcount()
+SupportsAvx512Popcount(void)
 {
 	unsigned int exx[4] = {0, 0, 0, 0};

--- a/src/bitvec.c
+++ b/src/bitvec.c
@@ -2,6 +2,7 @@

 #include "bitutils.h"
 #include "bitvec.h"
+#include "fmgr.h"
 #include "utils/varbit.h"
 #include "vector.h"

--- a/src/halfutils.c
+++ b/src/halfutils.c
@@ -1,5 +1,7 @@
 #include "postgres.h"

+#include <math.h>
+
 #include "halfutils.h"
 #include "halfvec.h"

--- a/src/halfvec.c
+++ b/src/halfvec.c
@@ -13,12 +13,20 @@
 #include "port.h"				/* for strtof() */
 #include "sparsevec.h"
 #include "utils/array.h"
-#include "utils/builtins.h"
 #include "utils/float.h"
+#include "utils/fmgrprotos.h"
 #include "utils/lsyscache.h"
-#include "utils/numeric.h"
+#include "utils/varbit.h"
 #include "vector.h"

+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif
+
+#if PG_VERSION_NUM >= 170000
+#include "parser/scansup.h"
+#endif
+
 #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
 #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))

@@ -129,9 +137,9 @@ InitHalfVector(int dim)
 	return result;
 }

-/*
- * Check for whitespace, since array_isspace() is static
- */
+#if PG_VERSION_NUM >= 170000
+#define halfvec_isspace(ch) scanner_isspace(ch)
+#else
 static inline bool
 halfvec_isspace(char ch)
 {
@@ -144,6 +152,7 @@ halfvec_isspace(char ch)
 		return true;
 	return false;
 }
+#endif

 /*
 * Check state array
@@ -898,8 +907,21 @@ halfvec_binary_quantize(PG_FUNCTION_ARGS)
 	half	   *ax = a->x;
 	VarBit	   *result = InitBitVector(a->dim);
 	unsigned char *rx = VARBITS(result);
+	int			i = 0;
+	int			count = (a->dim / 8) * 8;

-	for (int i = 0; i < a->dim; i++)
+	/* Auto-vectorized on aarch64 */
+	for (; i < count; i += 8)
+	{
+		unsigned char result_byte = 0;
+
+		for (int j = 0; j < 8; j++)
+			result_byte |= (HalfToFloat4(ax[i + j]) > 0) << (7 - j);
+
+		rx[i / 8] = result_byte;
+	}
+
+	for (; i < a->dim; i++)
 		rx[i / 8] |= (HalfToFloat4(ax[i]) > 0) << (7 - (i % 8));

 	PG_RETURN_VARBIT_P(result);
--- a/src/hnsw.c
+++ b/src/hnsw.c
@@ -1,18 +1,25 @@
 #include "postgres.h"

 #include <float.h>
+#include <limits.h>
 #include <math.h>

 #include "access/amapi.h"
+#include "access/genam.h"
 #include "access/reloptions.h"
 #include "commands/progress.h"
 #include "commands/vacuum.h"
+#include "fmgr.h"
 #include "hnsw.h"
 #include "miscadmin.h"
+#include "nodes/pg_list.h"
+#include "storage/lwlock.h"
 #include "utils/float.h"
 #include "utils/guc.h"
+#include "utils/relcache.h"
 #include "utils/selfuncs.h"
 #include "utils/spccache.h"
+#include "vector.h"

 #if PG_VERSION_NUM < 150000
 #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
@@ -52,12 +59,20 @@ HnswInitLockTranche(void)
 								  sizeof(int) * 1,
 								  &found);
 	if (!found)
+	{
+#if PG_VERSION_NUM >= 190000
+		tranche_ids[0] = LWLockNewTrancheId("HnswBuild");
+#else
 		tranche_ids[0] = LWLockNewTrancheId();
+#endif
+	}
 	hnsw_lock_tranche_id = tranche_ids[0];
 	LWLockRelease(AddinShmemInitLock);

+#if PG_VERSION_NUM < 190000
 	/* Per-backend registration of the tranche ID */
 	LWLockRegisterTranche(hnsw_lock_tranche_id, "HnswBuild");
+#endif
 }

 /*
@@ -130,7 +145,7 @@ hnswcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	Relation	index;

 	/* Never use index without order */
-	if (path->indexorderbys == NULL)
+	if (path->indexorderbys == NIL)
 	{
 		*indexStartupCost = get_float8_infinity();
 		*indexTotalCost = get_float8_infinity();
@@ -252,6 +267,64 @@ FUNCTION_PREFIX PG_FUNCTION_INFO_V1(hnswhandler);
 Datum
 hnswhandler(PG_FUNCTION_ARGS)
 {
+#if PG_VERSION_NUM >= 190000
+	static const IndexAmRoutine amroutine = {
+		.type = T_IndexAmRoutine,
+		.amstrategies = 0,
+		.amsupport = 3,
+		.amoptsprocnum = 0,
+		.amcanorder = false,
+		.amcanorderbyop = true,
+		.amcanhash = false,
+		.amconsistentequality = false,
+		.amconsistentordering = false,
+		.amcanbackward = false,
+		.amcanunique = false,
+		.amcanmulticol = false,
+		.amoptionalkey = true,
+		.amsearcharray = false,
+		.amsearchnulls = false,
+		.amstorage = false,
+		.amclusterable = false,
+		.ampredlocks = false,
+		.amcanparallel = false,
+		.amcanbuildparallel = true,
+		.amcaninclude = false,
+		.amusemaintenanceworkmem = false,
+		.amsummarizing = false,
+		.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
+		.amkeytype = InvalidOid,
+
+		.ambuild = hnswbuild,
+		.ambuildempty = hnswbuildempty,
+		.aminsert = hnswinsert,
+		.aminsertcleanup = NULL,
+		.ambulkdelete = hnswbulkdelete,
+		.amvacuumcleanup = hnswvacuumcleanup,
+		.amcanreturn = NULL,
+		.amcostestimate = hnswcostestimate,
+		.amgettreeheight = NULL,
+		.amoptions = hnswoptions,
+		.amproperty = NULL,
+		.ambuildphasename = hnswbuildphasename,
+		.amvalidate = hnswvalidate,
+		.amadjustmembers = NULL,
+		.ambeginscan = hnswbeginscan,
+		.amrescan = hnswrescan,
+		.amgettuple = hnswgettuple,
+		.amgetbitmap = NULL,
+		.amendscan = hnswendscan,
+		.ammarkpos = NULL,
+		.amrestrpos = NULL,
+		.amestimateparallelscan = NULL,
+		.aminitparallelscan = NULL,
+		.amparallelrescan = NULL,
+		.amtranslatestrategy = NULL,
+		.amtranslatecmptype = NULL,
+	};
+
+	PG_RETURN_POINTER(&amroutine);
+#else
 	IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);

 	amroutine->amstrategies = 0;
@@ -259,6 +332,11 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->amoptsprocnum = 0;
 	amroutine->amcanorder = false;
 	amroutine->amcanorderbyop = true;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amcanhash = false;
+	amroutine->amconsistentequality = false;
+	amroutine->amconsistentordering = false;
+#endif
 	amroutine->amcanbackward = false;	/* can change direction mid-scan */
 	amroutine->amcanunique = false;
 	amroutine->amcanmulticol = false;
@@ -291,6 +369,9 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->amvacuumcleanup = hnswvacuumcleanup;
 	amroutine->amcanreturn = NULL;
 	amroutine->amcostestimate = hnswcostestimate;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amgettreeheight = NULL;
+#endif
 	amroutine->amoptions = hnswoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
 	amroutine->ambuildphasename = hnswbuildphasename;
@@ -311,5 +392,11 @@ hnswhandler(PG_FUNCTION_ARGS)
 	amroutine->aminitparallelscan = NULL;
 	amroutine->amparallelrescan = NULL;

+#if PG_VERSION_NUM >= 180000
+	amroutine->amtranslatestrategy = NULL;
+	amroutine->amtranslatecmptype = NULL;
+#endif
+
 	PG_RETURN_POINTER(amroutine);
+#endif
 }
--- a/src/hnsw.h
+++ b/src/hnsw.h
@@ -3,15 +3,29 @@

 #include "postgres.h"

+#include <math.h>
+
 #include "access/genam.h"
 #include "access/parallel.h"
 #include "lib/pairingheap.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
+#include "storage/bufpage.h"
+#include "storage/condition_variable.h"
+#include "storage/lwlock.h"
+#include "storage/s_lock.h"
 #include "utils/relptr.h"
 #include "utils/sampling.h"
 #include "vector.h"

+#ifdef HNSW_BENCH
+#include "portability/instr_time.h"
+#endif
+
+#if PG_VERSION_NUM >= 190000
+typedef Pointer Item;
+#endif
+
 #define HNSW_MAX_DIM 2000
 #define HNSW_MAX_NNZ 1000

@@ -68,6 +82,21 @@
 #define HnswPageGetOpaque(page)	((HnswPageOpaque) PageGetSpecialPointer(page))
 #define HnswPageGetMeta(page)	((HnswMetaPageData *) PageGetContents(page))

+#ifdef HNSW_BENCH
+#define HnswBench(name, code) \
+	do { \
+		instr_time	start; \
+		instr_time	duration; \
+		INSTR_TIME_SET_CURRENT(start); \
+		(code); \
+		INSTR_TIME_SET_CURRENT(duration); \
+		INSTR_TIME_SUBTRACT(duration, start); \
+		elog(INFO, "%s: %.3f ms", name, INSTR_TIME_GET_MILLISEC(duration)); \
+	} while (0)
+#else
+#define HnswBench(name, code) (code)
+#endif
+
 #if PG_VERSION_NUM >= 150000
 #define RandomDouble() pg_prng_double(&pg_global_prng_state)
 #define SeedRandom(seed) pg_prng_seed(&pg_global_prng_state, seed)
@@ -402,6 +431,7 @@ typedef struct HnswVacuumState
 	BufferAccessStrategy bas;
 	HnswNeighborTuple ntup;
 	HnswElementData highestPoint;
+	HnswElementData fallbackPoint;

 	/* Memory */
 	MemoryContext tmpCtx;
@@ -424,7 +454,7 @@ void	   *HnswAlloc(HnswAllocator * allocator, Size size);
 HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc);
 HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
 void		HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, HnswSupport * support, int m, int efConstruction, bool existing);
-HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement em, HnswQuery * q, Relation rel, HnswSupport * support, bool loadVec);
+HnswSearchCandidate *HnswEntryCandidate(char *base, HnswElement entryPoint, HnswQuery * q, Relation index, HnswSupport * support, bool loadVec);
 void		HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building);
 void		HnswSetNeighborTuple(char *base, HnswNeighborTuple ntup, HnswElement e, int m);
 void		HnswAddHeapTid(HnswElement element, ItemPointer heaptid);
--- a/src/hnswbuild.c
+++ b/src/hnswbuild.c
@@ -36,11 +36,14 @@
 */
 #include "postgres.h"

-#include <math.h>
+#include <limits.h>

+#include "access/genam.h"
 #include "access/parallel.h"
+#include "access/relscan.h"
 #include "access/table.h"
 #include "access/tableam.h"
+#include "access/tupdesc.h"
 #include "access/xact.h"
 #include "access/xloginsert.h"
 #include "catalog/index.h"
@@ -48,11 +51,19 @@
 #include "commands/progress.h"
 #include "hnsw.h"
 #include "miscadmin.h"
+#include "nodes/execnodes.h"
 #include "optimizer/optimizer.h"
 #include "storage/bufmgr.h"
+#include "storage/condition_variable.h"
 #include "tcop/tcopprot.h"
 #include "utils/datum.h"
 #include "utils/memutils.h"
+#include "utils/rel.h"
+#include "utils/snapmgr.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_progress.h"
@@ -69,6 +80,8 @@
 #define PARALLEL_KEY_HNSW_AREA			UINT64CONST(0xA000000000000002)
 #define PARALLEL_KEY_QUERY_TEXT			UINT64CONST(0xA000000000000003)

+#define HNSW_MAX_GRAPH_MEMORY (SIZE_MAX / 2)
+
 /*
 * Create the metapage
 */
@@ -398,7 +411,7 @@ UpdateNeighborsInMemory(char *base, HnswSupport * support, HnswElement e, int m)
 * Update graph in memory
 */
 static void
-UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, HnswBuildState * buildstate)
+UpdateGraphInMemory(HnswSupport * support, HnswElement element, int m, HnswElement entryPoint, HnswBuildState * buildstate)
 {
 	HnswGraph  *graph = buildstate->graph;
 	char	   *base = buildstate->hnswarea;
@@ -460,7 +473,7 @@ InsertTupleInMemory(HnswBuildState * buildstate, HnswElement element)
 	HnswFindElementNeighbors(base, element, entryPoint, NULL, support, m, efConstruction, false);

 	/* Update graph in memory */
-	UpdateGraphInMemory(support, element, m, efConstruction, entryPoint, buildstate);
+	UpdateGraphInMemory(support, element, m, entryPoint, buildstate);

 	/* Release entry lock */
 	LWLockRelease(entryLock);
@@ -481,6 +494,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	LWLock	   *flushLock = &graph->flushLock;
 	char	   *base = buildstate->hnswarea;
 	Datum		value;
+	Size		memoryMargin;

 	/* Form index value */
 	if (!HnswFormIndexValue(&value, values, isnull, buildstate->typeInfo, support))
@@ -489,6 +503,9 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	/* Get datum size */
 	valueSize = VARSIZE_ANY(DatumGetPointer(value));

+	/* In a parallel build, add a margin so allocations never fail */
+	memoryMargin = base == NULL ? 0 : 1024 * 1024;
+
 	/* Ensure graph not flushed when inserting */
 	LWLockAcquire(flushLock, LW_SHARED);

@@ -510,7 +527,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
 	 * Check that we have enough memory available for the new element now that
 	 * we have the allocator lock, and flush pages if needed.
 	 */
-	if (graph->memoryUsed >= graph->memoryTotal)
+	if (graph->memoryUsed + memoryMargin >= graph->memoryTotal)
 	{
 		LWLockRelease(&graph->allocatorLock);

@@ -545,7 +562,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn

 	/* Copy the datum */
 	memcpy(valuePtr, DatumGetPointer(value), valueSize);
-	HnswPtrStore(base, element->value, valuePtr);
+	HnswPtrStore(base, element->value, (char *) valuePtr);

 	/* Create a lock for the element */
 	LWLockInitialize(&element->lock, hnsw_lock_tranche_id);
@@ -603,7 +620,7 @@ InitGraph(HnswGraph * graph, char *base, Size memoryTotal)
 	HnswPtrStore(base, graph->head, (HnswElement) NULL);
 	HnswPtrStore(base, graph->entryPoint, (HnswElement) NULL);
 	graph->memoryUsed = 0;
-	graph->memoryTotal = memoryTotal;
+	graph->memoryTotal = Min(memoryTotal, HNSW_MAX_GRAPH_MEMORY);
 	graph->flushed = false;
 	graph->indtuples = 0;
 	SpinLockInit(&graph->lock);
@@ -644,9 +661,17 @@ static void *
 HnswSharedMemoryAlloc(Size size, void *state)
 {
 	HnswBuildState *buildstate = (HnswBuildState *) state;
-	void	   *chunk = buildstate->hnswarea + buildstate->graph->memoryUsed;
+	Size		alignedSize = MAXALIGN(size);
+	void	   *chunk;

-	buildstate->graph->memoryUsed += MAXALIGN(size);
+	if (alignedSize > 1024 * 1024)
+		elog(ERROR, "hnsw allocation too large");
+
+	if (buildstate->graph->memoryUsed + alignedSize > buildstate->graph->memoryTotal)
+		elog(ERROR, "hnsw allocator out of memory");
+
+	chunk = buildstate->hnswarea + buildstate->graph->memoryUsed;
+	buildstate->graph->memoryUsed += alignedSize;
 	return chunk;
 }

@@ -778,7 +803,11 @@ HnswParallelScanAndInsert(Relation heapRel, Relation indexRel, HnswShared * hnsw
 	buildstate.hnswarea = hnswarea;
 	InitAllocator(&buildstate.allocator, &HnswSharedMemoryAlloc, &buildstate);
 	scan = table_beginscan_parallel(heapRel,
-									ParallelTableScanFromHnswShared(hnswshared));
+									ParallelTableScanFromHnswShared(hnswshared)
+#if PG_VERSION_NUM >= 190000
+									,SO_NONE
+#endif
+		);
 	reltuples = table_index_build_scan(heapRel, indexRel, indexInfo,
 									   true, progress, BuildCallback,
 									   (void *) &buildstate, scan);
@@ -932,6 +961,8 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
 	if (esthnswarea > estother)
 		esthnswarea -= estother;

+	esthnswarea = Min(esthnswarea, HNSW_MAX_GRAPH_MEMORY);
+
 	shm_toc_estimate_chunk(&pcxt->estimator, esthnswarea);
 	shm_toc_estimate_keys(&pcxt->estimator, 2);

@@ -974,8 +1005,7 @@ HnswBeginParallel(HnswBuildState * buildstate, bool isconcurrent, int request)
 								  snapshot);

 	hnswarea = (char *) shm_toc_allocate(pcxt->toc, esthnswarea);
-	/* Report less than allocated so never fails */
-	InitGraph(&hnswshared->graphData, hnswarea, esthnswarea - 1024 * 1024);
+	InitGraph(&hnswshared->graphData, hnswarea, esthnswarea);

 	/*
 	 * Avoid base address for relptr for Postgres < 14.5
@@ -1054,7 +1084,7 @@ ComputeParallelWorkers(Relation heap, Relation index)
 * Build graph
 */
 static void
-BuildGraph(HnswBuildState * buildstate, ForkNumber forkNum)
+BuildGraph(HnswBuildState * buildstate)
 {
 	int			parallel_workers = 0;

@@ -1102,7 +1132,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,

 	InitBuildState(buildstate, heap, index, indexInfo, forkNum);

-	BuildGraph(buildstate, forkNum);
+	BuildGraph(buildstate);

 	if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM)
 		log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);
--- a/src/hnswinsert.c
+++ b/src/hnswinsert.c
@@ -1,13 +1,19 @@
 #include "postgres.h"

-#include <math.h>
-
+#include "access/genam.h"
 #include "access/generic_xlog.h"
 #include "hnsw.h"
+#include "nodes/execnodes.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
+#include "storage/lwlock.h"
 #include "utils/datum.h"
 #include "utils/memutils.h"
+#include "utils/rel.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 /*
 * Get the insert page
@@ -660,7 +666,7 @@ FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
 * Update graph on disk
 */
 static void
-UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
+UpdateGraphOnDisk(Relation index, HnswSupport * support, HnswElement element, int m, HnswElement entryPoint, bool building)
 {
 	BlockNumber newInsertPage = InvalidBlockNumber;

@@ -708,7 +714,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo

 	/* Create an element */
 	element = HnswInitElement(base, heaptid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	HnswPtrStore(base, element->value, DatumGetPointer(value));
+	HnswPtrStore(base, element->value, (char *) DatumGetPointer(value));

 	/* Prevent concurrent inserts when likely updating entry point */
 	if (entryPoint == NULL || element->level > entryPoint->level)
@@ -728,7 +734,7 @@ HnswInsertTupleOnDisk(Relation index, HnswSupport * support, Datum value, ItemPo
 	HnswFindElementNeighbors(base, element, entryPoint, index, support, m, efConstruction, false);

 	/* Update graph on disk */
-	UpdateGraphOnDisk(index, support, element, m, efConstruction, entryPoint, building);
+	UpdateGraphOnDisk(index, support, element, m, entryPoint, building);

 	/* Release lock */
 	UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
--- a/src/hnswscan.c
+++ b/src/hnswscan.c
@@ -1,12 +1,23 @@
 #include "postgres.h"

+#include <limits.h>
+
+#include "access/genam.h"
 #include "access/relscan.h"
 #include "hnsw.h"
+#include "lib/pairingheap.h"
+#include "miscadmin.h"
+#include "nodes/pg_list.h"
 #include "pgstat.h"
-#include "storage/bufmgr.h"
 #include "storage/lmgr.h"
 #include "utils/float.h"
 #include "utils/memutils.h"
+#include "utils/relcache.h"
+#include "utils/snapmgr.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 /*
 * Algorithm 5 from paper
@@ -142,7 +153,7 @@ hnswbeginscan(Relation index, int nkeys, int norderbys)
 	/* Calculate max memory */
 	/* Add 256 extra bytes to fill last block when close */
 	maxMemory = (double) work_mem * hnsw_scan_mem_multiplier * 1024.0 + 256;
-	so->maxMemory = Min(maxMemory, (double) SIZE_MAX);
+	so->maxMemory = Min(maxMemory, (double) (SIZE_MAX / 2));

 	scan->opaque = so;

@@ -193,6 +204,10 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)

 		/* Count index scan for stats */
 		pgstat_count_index_scan(scan->indexRelation);
+#if PG_VERSION_NUM >= 180000
+		if (scan->instrument)
+			scan->instrument->nsearches++;
+#endif

 		/* Safety check */
 		if (scan->orderByData == NULL)
--- a/src/hnswutils.c
+++ b/src/hnswutils.c
@@ -2,18 +2,24 @@

 #include <math.h>

+#include "access/genam.h"
 #include "access/generic_xlog.h"
-#include "catalog/pg_type.h"
-#include "catalog/pg_type_d.h"
 #include "common/hashfn.h"
 #include "fmgr.h"
 #include "hnsw.h"
 #include "lib/pairingheap.h"
+#include "nodes/pg_list.h"
+#include "port/atomics.h"
 #include "sparsevec.h"
 #include "storage/bufmgr.h"
 #include "utils/datum.h"
 #include "utils/memdebug.h"
 #include "utils/rel.h"
+#include "vector.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 #if PG_VERSION_NUM < 170000
 static inline uint64
@@ -256,7 +262,7 @@ HnswInitElement(char *base, ItemPointer heaptid, int m, double ml, int maxLevel,

 	HnswInitNeighbors(base, element, m, allocator);

-	HnswPtrStore(base, element->value, (Pointer) NULL);
+	HnswPtrStore(base, element->value, (char *) NULL);

 	return element;
 }
@@ -282,7 +288,7 @@ HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno)
 	element->blkno = blkno;
 	element->offno = offno;
 	HnswPtrStore(base, element->neighbors, (HnswNeighborArrayPtr *) NULL);
-	HnswPtrStore(base, element->value, (Pointer) NULL);
+	HnswPtrStore(base, element->value, (char *) NULL);
 	return element;
 }

@@ -508,7 +514,7 @@ HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHe
 		char	   *base = NULL;
 		Datum		value = datumCopy(PointerGetDatum(&etup->data), false, -1);

-		HnswPtrStore(base, element->value, DatumGetPointer(value));
+		HnswPtrStore(base, element->value, (char *) DatumGetPointer(value));
 	}
 }

@@ -540,6 +546,9 @@ HnswLoadElementImpl(BlockNumber blkno, OffsetNumber offno, double *distance, Hns

 	Assert(HnswIsElementTuple(etup));

+	if (unlikely(etup->deleted))
+		elog(ERROR, "cannot load deleted element");
+
 	/* Calculate distance */
 	if (distance != NULL)
 	{
@@ -922,7 +931,7 @@ HnswSearchLayer(char *base, HnswQuery * q, List *ep, int ef, int lc, Relation in
 					continue;
 			}

-			if (eElement == NULL || !(eDistance < f->distance || alwaysAdd))
+			if (!(eDistance < f->distance || alwaysAdd))
 			{
 				if (discarded != NULL)
 				{
--- a/src/hnswvacuum.c
+++ b/src/hnswvacuum.c
@@ -1,13 +1,22 @@
 #include "postgres.h"

-#include <math.h>
-
+#include "access/genam.h"
 #include "access/generic_xlog.h"
 #include "commands/vacuum.h"
 #include "hnsw.h"
+#include "nodes/pg_list.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
 #include "utils/memutils.h"
+#include "utils/rel.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif
+
+#if PG_VERSION_NUM >= 180000
+#define vacuum_delay_point() vacuum_delay_point(false)
+#endif

 /*
 * Check if deleted list contains an index TID
@@ -28,17 +37,20 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 {
 	BlockNumber blkno = HNSW_HEAD_BLKNO;
 	HnswElement highestPoint = &vacuumstate->highestPoint;
+	HnswElement fallbackPoint = &vacuumstate->fallbackPoint;
 	Relation	index = vacuumstate->index;
 	BufferAccessStrategy bas = vacuumstate->bas;
-	HnswElement entryPoint = HnswGetEntryPoint(vacuumstate->index);
 	IndexBulkDeleteResult *stats = vacuumstate->stats;

-	/* Store separately since highestPoint.level is uint8 */
+	/* Store separately since HnswElement level is uint8 */
 	int			highestLevel = -1;
+	int			fallbackLevel = -1;

-	/* Initialize highest point */
+	/* Initialize highest point and fallback point */
 	highestPoint->blkno = InvalidBlockNumber;
 	highestPoint->offno = InvalidOffsetNumber;
+	fallbackPoint->blkno = InvalidBlockNumber;
+	fallbackPoint->offno = InvalidOffsetNumber;

 	while (BlockNumberIsValid(blkno))
 	{
@@ -110,14 +122,31 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 				tidhash_insert(vacuumstate->deleted, ip, &found);
 				Assert(!found);
 			}
-			else if (etup->level > highestLevel && !(entryPoint != NULL && blkno == entryPoint->blkno && offno == entryPoint->offno))
+			else if (etup->level > highestLevel)
 			{
-				/* Keep track of highest non-entry point */
+				if (BlockNumberIsValid(highestPoint->blkno))
+				{
+					/* Current highest point becomes fallback */
+					fallbackPoint->blkno = highestPoint->blkno;
+					fallbackPoint->offno = highestPoint->offno;
+					fallbackPoint->level = highestPoint->level;
+					fallbackLevel = highestLevel;
+				}
+
+				/* Keep track of highest point */
 				highestPoint->blkno = blkno;
 				highestPoint->offno = offno;
 				highestPoint->level = etup->level;
 				highestLevel = etup->level;
 			}
+			else if (etup->level > fallbackLevel)
+			{
+				/* Keep track of second highest point */
+				fallbackPoint->blkno = blkno;
+				fallbackPoint->offno = offno;
+				fallbackPoint->level = etup->level;
+				fallbackLevel = etup->level;
+			}
 		}

 		blkno = HnswPageGetOpaque(page)->nextblkno;
@@ -129,6 +158,10 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)

 		UnlockReleaseBuffer(buf);
 	}
+
+#ifdef HNSW_MEMORY
+	elog(INFO, "memory: %zu KB", MemoryContextMemAllocated(CurrentMemoryContext, true) / 1024);
+#endif
 }

 /*
@@ -170,7 +203,12 @@ NeedsUpdated(HnswVacuumState * vacuumstate, HnswElement element)
 	/* Also update if layer 0 is not full */
 	/* This could indicate too many candidates being deleted during insert */
 	if (!needsUpdated)
+	{
+		/* Keep clang-tidy happy */
+		Assert(ntup->count > 0);
+
 		needsUpdated = !ItemPointerIsValid(&ntup->indextids[ntup->count - 1]);
+	}

 	UnlockReleaseBuffer(buf);

@@ -255,12 +293,27 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 		/* Get a shared lock */
 		LockPage(index, HNSW_UPDATE_LOCK, ShareLock);

-		/* Load element */
-		HnswLoadElement(highestPoint, NULL, NULL, index, support, true, NULL);
+		/* Get latest entry point */
+		entryPoint = HnswGetEntryPoint(index);

-		/* Repair if needed */
-		if (NeedsUpdated(vacuumstate, highestPoint))
-			RepairGraphElement(vacuumstate, highestPoint, HnswGetEntryPoint(index));
+		/* Use fallback point if highest point is entry point */
+		if (entryPoint != NULL && entryPoint->blkno == highestPoint->blkno && entryPoint->offno == highestPoint->offno)
+		{
+			highestPoint = &vacuumstate->fallbackPoint;
+
+			if (!BlockNumberIsValid(highestPoint->blkno))
+				highestPoint = NULL;
+		}
+
+		if (highestPoint != NULL)
+		{
+			/* Load element */
+			HnswLoadElement(highestPoint, NULL, NULL, index, support, true, NULL);
+
+			/* Repair if needed */
+			if (NeedsUpdated(vacuumstate, highestPoint))
+				RepairGraphElement(vacuumstate, highestPoint, entryPoint);
+		}

 		/* Release lock */
 		UnlockPage(index, HNSW_UPDATE_LOCK, ShareLock);
@@ -427,6 +480,99 @@ RepairGraph(HnswVacuumState * vacuumstate)
 		/* Reset memory context */
 		MemoryContextSwitchTo(oldCtx);
 		MemoryContextReset(vacuumstate->tmpCtx);
+
+#ifdef HNSW_VACUUM_PROGRESS
+		if (!BlockNumberIsValid(blkno) || (blkno - HNSW_HEAD_BLKNO) % 1000 == 0)
+		{
+			BlockNumber totalBlocks = RelationGetNumberOfBlocks(index);
+			BlockNumber currentBlocks = BlockNumberIsValid(blkno) ? blkno : totalBlocks;
+
+			elog(INFO, "hnsw vacuum progress: %.1f%%", 100.0 * currentBlocks / totalBlocks);
+		}
+#endif
+	}
+}
+
+/*
+ * Confirm graph was repaired
+ */
+static void
+ConfirmRepaired(HnswVacuumState * vacuumstate)
+{
+	BlockNumber blkno = HNSW_HEAD_BLKNO;
+	Relation	index = vacuumstate->index;
+	BufferAccessStrategy bas = vacuumstate->bas;
+
+	while (BlockNumberIsValid(blkno))
+	{
+		Buffer		buf;
+		Page		page;
+		OffsetNumber offno;
+		OffsetNumber maxoffno;
+
+		vacuum_delay_point();
+
+		buf = ReadBufferExtended(index, MAIN_FORKNUM, blkno, RBM_NORMAL, bas);
+		LockBuffer(buf, BUFFER_LOCK_SHARE);
+		page = BufferGetPage(buf);
+		maxoffno = PageGetMaxOffsetNumber(page);
+
+		/* Iterate over nodes */
+		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
+		{
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
+			HnswNeighborTuple ntup;
+			Buffer		nbuf;
+			Page		npage;
+			BlockNumber neighborPage;
+			OffsetNumber neighborOffno;
+
+			/* Skip neighbor tuples */
+			if (!HnswIsElementTuple(etup))
+				continue;
+
+			/* Skip if being deleted */
+			if (!ItemPointerIsValid(&etup->heaptids[0]))
+				continue;
+
+			/* Get neighbor page */
+			neighborPage = ItemPointerGetBlockNumber(&etup->neighbortid);
+			neighborOffno = ItemPointerGetOffsetNumber(&etup->neighbortid);
+
+			if (neighborPage == blkno)
+			{
+				nbuf = buf;
+				npage = page;
+			}
+			else
+			{
+				nbuf = ReadBufferExtended(index, MAIN_FORKNUM, neighborPage, RBM_NORMAL, bas);
+				LockBuffer(nbuf, BUFFER_LOCK_SHARE);
+				npage = BufferGetPage(nbuf);
+			}
+
+			ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno));
+
+			/* Check neighbors */
+			for (int i = 0; i < ntup->count; i++)
+			{
+				ItemPointer indextid = &ntup->indextids[i];
+
+				if (!ItemPointerIsValid(indextid))
+					continue;
+
+				/* Check if in deleted list */
+				if (DeletedContains(vacuumstate->deleted, indextid))
+					elog(ERROR, "hnsw graph not repaired");
+			}
+
+			if (nbuf != buf)
+				UnlockReleaseBuffer(nbuf);
+		}
+
+		blkno = HnswPageGetOpaque(page)->nextblkno;
+
+		UnlockReleaseBuffer(buf);
 	}
 }

@@ -520,8 +666,9 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 			ntup = (HnswNeighborTuple) PageGetItem(npage, PageGetItemId(npage, neighborOffno));

 			/* Overwrite element */
+			/* Use memset instead of MemSet to keep clang-tidy happy */
 			etup->deleted = 1;
-			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
+			memset(&etup->data, 0, VARSIZE_ANY(&etup->data));

 			/* Overwrite neighbors */
 			for (int i = 0; i < ntup->count; i++)
@@ -619,13 +766,16 @@ hnswbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
 	InitVacuumState(&vacuumstate, info, stats, callback, callback_state);

 	/* Pass 1: Remove heap TIDs */
-	RemoveHeapTids(&vacuumstate);
+	HnswBench("RemoveHeapTids", RemoveHeapTids(&vacuumstate));

 	/* Pass 2: Repair graph */
-	RepairGraph(&vacuumstate);
+	HnswBench("RepairGraph", RepairGraph(&vacuumstate));

-	/* Pass 3: Mark as deleted */
-	MarkDeleted(&vacuumstate);
+	/* Pass 3: Confirm repaired */
+	HnswBench("ConfirmRepaired", ConfirmRepaired(&vacuumstate));
+
+	/* Pass 4: Mark as deleted */
+	HnswBench("MarkDeleted", MarkDeleted(&vacuumstate));

 	FreeVacuumState(&vacuumstate);

--- a/src/ivfbuild.c
+++ b/src/ivfbuild.c
@@ -2,23 +2,37 @@

 #include <float.h>

+#include "access/genam.h"
+#include "access/generic_xlog.h"
+#include "access/itup.h"
+#include "access/relscan.h"
 #include "access/table.h"
 #include "access/tableam.h"
+#include "access/tupdesc.h"
 #include "access/parallel.h"
 #include "access/xact.h"
-#include "bitvec.h"
+#include "access/xloginsert.h"
 #include "catalog/index.h"
 #include "catalog/pg_operator_d.h"
 #include "catalog/pg_type_d.h"
 #include "commands/progress.h"
-#include "halfvec.h"
+#include "fmgr.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
+#include "nodes/execnodes.h"
 #include "optimizer/optimizer.h"
 #include "storage/bufmgr.h"
+#include "storage/condition_variable.h"
 #include "tcop/tcopprot.h"
 #include "utils/memutils.h"
-#include "vector.h"
+#include "utils/rel.h"
+#include "utils/sampling.h"
+#include "utils/snapmgr.h"
+#include "utils/tuplesort.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 #if PG_VERSION_NUM >= 140000
 #include "utils/backend_progress.h"
@@ -49,15 +63,13 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
 	Datum		value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));

 	/*
-	 * Normalize with KMEANS_NORM_PROC since spherical distance function
-	 * expects unit vectors
+	 * Check with KMEANS_NORM_PROC that the value can be normalized since
+	 * spherical distance function expects unit vectors
 	 */
 	if (buildstate->kmeansnormprocinfo != NULL)
 	{
 		if (!IvfflatCheckNorm(buildstate->kmeansnormprocinfo, buildstate->collation, value))
 			return;
-
-		value = IvfflatNormValue(buildstate->typeInfo, buildstate->collation, value);
 	}

 	if (samples->length < targsamples)
@@ -68,7 +80,7 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
 	else
 	{
 		if (buildstate->rowstoskip < 0)
-			buildstate->rowstoskip = reservoir_get_next_S(&buildstate->rstate, samples->length, targsamples);
+			buildstate->rowstoskip = reservoir_get_next_S(&buildstate->rstate, buildstate->samplerows, targsamples);

 		if (buildstate->rowstoskip <= 0)
 		{
@@ -84,6 +96,9 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)

 		buildstate->rowstoskip -= 1;
 	}
+
+	/* Increment after reservoir_get_next_S */
+	buildstate->samplerows += 1;
 }

 /*
@@ -120,6 +135,7 @@ SampleRows(IvfflatBuildState * buildstate)
 	int			targsamples = buildstate->samples->maxlen;
 	BlockNumber totalblocks = RelationGetNumberOfBlocks(buildstate->heap);

+	buildstate->samplerows = 0;
 	buildstate->rowstoskip = -1;

 	BlockSampler_Init(&buildstate->bs, totalblocks, targsamples, RandomInt());
@@ -129,8 +145,24 @@ SampleRows(IvfflatBuildState * buildstate)
 	{
 		BlockNumber targblock = BlockSampler_Next(&buildstate->bs);

+		/* Set anyvisible to false like table_index_build_scan */
 		table_index_build_range_scan(buildstate->heap, buildstate->index, buildstate->indexInfo,
-									 false, true, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+									 false, false, false, targblock, 1, SampleCallback, (void *) buildstate, NULL);
+	}
+
+	/* Normalize if needed */
+	if (buildstate->kmeansnormprocinfo != NULL)
+	{
+		VectorArray samples = buildstate->samples;
+
+		for (int i = 0; i < samples->length; i++)
+		{
+			Datum		value = PointerGetDatum(VectorArrayGet(samples, i));
+			Datum		normValue = IvfflatNormValue(buildstate->typeInfo, buildstate->collation, value);
+
+			VectorArraySet(samples, i, DatumGetPointer(normValue));
+			pfree(DatumGetPointer(normValue));
+		}
 	}
 }

@@ -138,7 +170,7 @@ SampleRows(IvfflatBuildState * buildstate)
 * Add tuple to sort
 */
 static void
-AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState * buildstate)
+AddTupleToSort(ItemPointer tid, Datum *values, IvfflatBuildState * buildstate)
 {
 	double		distance;
 	double		minDistance = DBL_MAX;
@@ -215,7 +247,7 @@ BuildCallback(Relation index, ItemPointer tid, Datum *values,
 	oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx);

 	/* Add tuple to sort */
-	AddTupleToSort(index, tid, values, buildstate);
+	AddTupleToSort(tid, values, buildstate);

 	/* Reset memory context */
 	MemoryContextSwitchTo(oldCtx);
@@ -361,9 +393,13 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
 	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 1, "list", INT4OID, -1, 0);
 	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 2, "tid", TIDOID, -1, 0);
 	TupleDescInitEntry(buildstate->sortdesc, (AttrNumber) 3, "vector", TupleDescAttr(buildstate->tupdesc, 0)->atttypid, -1, 0);
+#if PG_VERSION_NUM >= 190000
+	TupleDescFinalize(buildstate->sortdesc);
+#endif

 	buildstate->slot = MakeSingleTupleTableSlot(buildstate->sortdesc, &TTSOpsVirtual);

+	/* TODO Ensure within maintenance_work_mem */
 	buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions, buildstate->typeInfo->itemSize(buildstate->dimensions));
 	buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);

@@ -422,7 +458,7 @@ ComputeCenters(IvfflatBuildState * buildstate)
 	buildstate->samples = VectorArrayInit(numSamples, buildstate->dimensions, buildstate->centers->itemsize);
 	if (buildstate->heap != NULL)
 	{
-		SampleRows(buildstate);
+		IvfflatBench("sample rows", SampleRows(buildstate));

 		if (buildstate->samples->length < buildstate->lists)
 		{
@@ -470,8 +506,8 @@ CreateMetaPage(Relation index, int dimensions, int lists, ForkNumber forkNum)
 * Create list pages
 */
 static void
-CreateListPages(Relation index, VectorArray centers, int dimensions,
-				int lists, ForkNumber forkNum, ListInfo * *listInfo)
+CreateListPages(Relation index, VectorArray centers, int lists,
+				ForkNumber forkNum, ListInfo * *listInfo)
 {
 	Buffer		buf;
 	Page		page;
@@ -637,7 +673,11 @@ IvfflatParallelScanAndSort(IvfflatSpool * ivfspool, IvfflatShared * ivfshared, S
 	ivfspool->sortstate = InitBuildSortState(buildstate.sortdesc, sortmem, coordinate);
 	buildstate.sortstate = ivfspool->sortstate;
 	scan = table_beginscan_parallel(ivfspool->heap,
-									ParallelTableScanFromIvfflatShared(ivfshared));
+									ParallelTableScanFromIvfflatShared(ivfshared)
+#if PG_VERSION_NUM >= 190000
+									,SO_NONE
+#endif
+		);
 	reltuples = table_index_build_scan(ivfspool->heap, ivfspool->index, indexInfo,
 									   true, progress, BuildCallback,
 									   (void *) &buildstate, scan);
@@ -1004,7 +1044,7 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,

 	/* Create pages */
 	CreateMetaPage(index, buildstate->dimensions, buildstate->lists, forkNum);
-	CreateListPages(index, buildstate->centers, buildstate->dimensions, buildstate->lists, forkNum, &buildstate->listInfo);
+	CreateListPages(index, buildstate->centers, buildstate->lists, forkNum, &buildstate->listInfo);
 	CreateEntryPages(buildstate, forkNum);

 	/* Write WAL for initialization fork since GenericXLog functions do not */
@@ -1023,6 +1063,10 @@ ivfflatbuild(Relation heap, Relation index, IndexInfo *indexInfo)
 	IndexBuildResult *result;
 	IvfflatBuildState buildstate;

+#ifdef IVFFLAT_BENCH
+	SeedRandom(42);
+#endif
+
 	BuildIndex(heap, index, indexInfo, &buildstate, MAIN_FORKNUM);

 	result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));
--- a/src/ivfflat.c
+++ b/src/ivfflat.c
@@ -3,14 +3,19 @@
 #include <float.h>

 #include "access/amapi.h"
+#include "access/genam.h"
 #include "access/reloptions.h"
 #include "commands/progress.h"
 #include "commands/vacuum.h"
+#include "fmgr.h"
 #include "ivfflat.h"
+#include "nodes/pg_list.h"
 #include "utils/float.h"
 #include "utils/guc.h"
+#include "utils/relcache.h"
 #include "utils/selfuncs.h"
 #include "utils/spccache.h"
+#include "vector.h"

 #if PG_VERSION_NUM < 150000
 #define MarkGUCPrefixReserved(x) EmitWarningsOnPlaceholders(x)
@@ -92,7 +97,7 @@ ivfflatcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	Relation	index;

 	/* Never use index without order */
-	if (path->indexorderbys == NULL)
+	if (path->indexorderbys == NIL)
 	{
 		*indexStartupCost = get_float8_infinity();
 		*indexTotalCost = get_float8_infinity();
@@ -179,6 +184,64 @@ FUNCTION_PREFIX PG_FUNCTION_INFO_V1(ivfflathandler);
 Datum
 ivfflathandler(PG_FUNCTION_ARGS)
 {
+#if PG_VERSION_NUM >= 190000
+	static const IndexAmRoutine amroutine = {
+		.type = T_IndexAmRoutine,
+		.amstrategies = 0,
+		.amsupport = 5,
+		.amoptsprocnum = 0,
+		.amcanorder = false,
+		.amcanorderbyop = true,
+		.amcanhash = false,
+		.amconsistentequality = false,
+		.amconsistentordering = false,
+		.amcanbackward = false,
+		.amcanunique = false,
+		.amcanmulticol = false,
+		.amoptionalkey = true,
+		.amsearcharray = false,
+		.amsearchnulls = false,
+		.amstorage = false,
+		.amclusterable = false,
+		.ampredlocks = false,
+		.amcanparallel = false,
+		.amcanbuildparallel = true,
+		.amcaninclude = false,
+		.amusemaintenanceworkmem = false,
+		.amsummarizing = false,
+		.amparallelvacuumoptions = VACUUM_OPTION_PARALLEL_BULKDEL,
+		.amkeytype = InvalidOid,
+
+		.ambuild = ivfflatbuild,
+		.ambuildempty = ivfflatbuildempty,
+		.aminsert = ivfflatinsert,
+		.aminsertcleanup = NULL,
+		.ambulkdelete = ivfflatbulkdelete,
+		.amvacuumcleanup = ivfflatvacuumcleanup,
+		.amcanreturn = NULL,
+		.amcostestimate = ivfflatcostestimate,
+		.amgettreeheight = NULL,
+		.amoptions = ivfflatoptions,
+		.amproperty = NULL,
+		.ambuildphasename = ivfflatbuildphasename,
+		.amvalidate = ivfflatvalidate,
+		.amadjustmembers = NULL,
+		.ambeginscan = ivfflatbeginscan,
+		.amrescan = ivfflatrescan,
+		.amgettuple = ivfflatgettuple,
+		.amgetbitmap = NULL,
+		.amendscan = ivfflatendscan,
+		.ammarkpos = NULL,
+		.amrestrpos = NULL,
+		.amestimateparallelscan = NULL,
+		.aminitparallelscan = NULL,
+		.amparallelrescan = NULL,
+		.amtranslatestrategy = NULL,
+		.amtranslatecmptype = NULL,
+	};
+
+	PG_RETURN_POINTER(&amroutine);
+#else
 	IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);

 	amroutine->amstrategies = 0;
@@ -186,6 +249,11 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amoptsprocnum = 0;
 	amroutine->amcanorder = false;
 	amroutine->amcanorderbyop = true;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amcanhash = false;
+	amroutine->amconsistentequality = false;
+	amroutine->amconsistentordering = false;
+#endif
 	amroutine->amcanbackward = false;	/* can change direction mid-scan */
 	amroutine->amcanunique = false;
 	amroutine->amcanmulticol = false;
@@ -218,6 +286,9 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->amvacuumcleanup = ivfflatvacuumcleanup;
 	amroutine->amcanreturn = NULL;	/* tuple not included in heapsort */
 	amroutine->amcostestimate = ivfflatcostestimate;
+#if PG_VERSION_NUM >= 180000
+	amroutine->amgettreeheight = NULL;
+#endif
 	amroutine->amoptions = ivfflatoptions;
 	amroutine->amproperty = NULL;	/* TODO AMPROP_DISTANCE_ORDERABLE */
 	amroutine->ambuildphasename = ivfflatbuildphasename;
@@ -238,5 +309,11 @@ ivfflathandler(PG_FUNCTION_ARGS)
 	amroutine->aminitparallelscan = NULL;
 	amroutine->amparallelrescan = NULL;

+#if PG_VERSION_NUM >= 180000
+	amroutine->amtranslatestrategy = NULL;
+	amroutine->amtranslatecmptype = NULL;
+#endif
+
 	PG_RETURN_POINTER(amroutine);
+#endif
 }
--- a/src/ivfflat.h
+++ b/src/ivfflat.h
@@ -9,10 +9,15 @@
 #include "lib/pairingheap.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
+#include "storage/condition_variable.h"
 #include "utils/sampling.h"
 #include "utils/tuplesort.h"
 #include "vector.h"

+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif
+
 #if PG_VERSION_NUM >= 150000
 #include "common/pg_prng.h"
 #endif
@@ -21,6 +26,10 @@
 #include "portability/instr_time.h"
 #endif

+#if PG_VERSION_NUM >= 190000
+typedef Pointer Item;
+#endif
+
 #define IVFFLAT_MAX_DIM 2000

 /* Support functions */
@@ -73,9 +82,11 @@
 #if PG_VERSION_NUM >= 150000
 #define RandomDouble() pg_prng_double(&pg_global_prng_state)
 #define RandomInt() pg_prng_uint32(&pg_global_prng_state)
+#define SeedRandom(seed) pg_prng_seed(&pg_global_prng_state, seed)
 #else
 #define RandomDouble() (((double) random()) / MAX_RANDOM_VALUE)
 #define RandomInt() random()
+#define SeedRandom(seed) srandom(seed)
 #endif

 /* Variables */
@@ -203,7 +214,8 @@ typedef struct IvfflatBuildState
 	/* Sampling */
 	BlockSamplerData bs;
 	ReservoirStateData rstate;
-	int			rowstoskip;
+	double		samplerows;
+	double		rowstoskip;

 	/* Sorting */
 	Tuplesortstate *sortstate;
--- a/src/ivfinsert.c
+++ b/src/ivfinsert.c
@@ -2,11 +2,16 @@

 #include <float.h>

+#include "access/genam.h"
 #include "access/generic_xlog.h"
+#include "access/itup.h"
+#include "fmgr.h"
 #include "ivfflat.h"
+#include "nodes/execnodes.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
 #include "utils/memutils.h"
+#include "utils/rel.h"

 /*
 * Find the list that minimizes the distance function
@@ -65,7 +70,7 @@ FindInsertPage(Relation index, Datum *values, BlockNumber *insertPage, ListInfo
 * Insert a tuple into the index
 */
 static void
-InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
+InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid)
 {
 	const		IvfflatTypeInfo *typeInfo = IvfflatGetTypeInfo(index);
 	IndexTuple	itup;
@@ -204,7 +209,7 @@ ivfflatinsert(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid,
 	oldCtx = MemoryContextSwitchTo(insertCtx);

 	/* Insert tuple */
-	InsertTuple(index, values, isnull, heap_tid, heap);
+	InsertTuple(index, values, isnull, heap_tid);

 	/* Delete memory context */
 	MemoryContextSwitchTo(oldCtx);
--- a/src/ivfkmeans.c
+++ b/src/ivfkmeans.c
@@ -1,17 +1,19 @@
 #include "postgres.h"

 #include <float.h>
+#include <limits.h>
 #include <math.h>

-#include "bitvec.h"
-#include "halfutils.h"
-#include "halfvec.h"
+#include "access/genam.h"
+#include "fmgr.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
-#include "utils/builtins.h"
-#include "utils/datum.h"
 #include "utils/memutils.h"
-#include "vector.h"
+#include "utils/relcache.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 /*
 * Initialize with kmeans++
--- a/src/ivfscan.c
+++ b/src/ivfscan.c
@@ -2,15 +2,26 @@

 #include <float.h>

+#include "access/genam.h"
+#include "access/itup.h"
 #include "access/relscan.h"
+#include "access/tupdesc.h"
 #include "catalog/pg_operator_d.h"
 #include "catalog/pg_type_d.h"
+#include "fmgr.h"
 #include "lib/pairingheap.h"
 #include "ivfflat.h"
 #include "miscadmin.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
 #include "utils/memutils.h"
+#include "utils/rel.h"
+#include "utils/snapmgr.h"
+#include "utils/tuplesort.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 #define GetScanList(ptr) pairingheap_container(IvfflatScanList, ph_node, ptr)
 #define GetScanListConst(ptr) pairingheap_const_container(IvfflatScanList, ph_node, ptr)
@@ -287,6 +298,9 @@ ivfflatbeginscan(Relation index, int nkeys, int norderbys)
 	so->tupdesc = CreateTemplateTupleDesc(2);
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 1, "distance", FLOAT8OID, -1, 0);
 	TupleDescInitEntry(so->tupdesc, (AttrNumber) 2, "heaptid", TIDOID, -1, 0);
+#if PG_VERSION_NUM >= 190000
+	TupleDescFinalize(so->tupdesc);
+#endif

 	/* Prep sort */
 	so->sortstate = InitScanSortState(so->tupdesc);
@@ -355,6 +369,10 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)

 		/* Count index scan for stats */
 		pgstat_count_index_scan(scan->indexRelation);
+#if PG_VERSION_NUM >= 180000
+		if (scan->instrument)
+			scan->instrument->nsearches++;
+#endif

 		/* Safety check */
 		if (scan->orderByData == NULL)
--- a/src/ivfutils.c
+++ b/src/ivfutils.c
@@ -1,13 +1,19 @@
 #include "postgres.h"

+#include "access/genam.h"
 #include "access/generic_xlog.h"
-#include "bitvec.h"
-#include "catalog/pg_type.h"
 #include "fmgr.h"
 #include "halfutils.h"
 #include "halfvec.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
+#include "utils/relcache.h"
+#include "utils/varbit.h"
+#include "vector.h"
+
+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif

 /*
 * Allocate a vector array
@@ -259,8 +265,8 @@ VectorUpdateCenter(Pointer v, int dimensions, float *x)
 	SET_VARSIZE(vec, VECTOR_SIZE(dimensions));
 	vec->dim = dimensions;

-	for (int k = 0; k < dimensions; k++)
-		vec->x[k] = x[k];
+	for (int i = 0; i < dimensions; i++)
+		vec->x[i] = x[i];
 }

 static void
@@ -271,8 +277,8 @@ HalfvecUpdateCenter(Pointer v, int dimensions, float *x)
 	SET_VARSIZE(vec, HALFVEC_SIZE(dimensions));
 	vec->dim = dimensions;

-	for (int k = 0; k < dimensions; k++)
-		vec->x[k] = Float4ToHalfUnchecked(x[k]);
+	for (int i = 0; i < dimensions; i++)
+		vec->x[i] = Float4ToHalfUnchecked(x[i]);
 }

 static void
@@ -284,29 +290,33 @@ BitUpdateCenter(Pointer v, int dimensions, float *x)
 	SET_VARSIZE(vec, VARBITTOTALLEN(dimensions));
 	VARBITLEN(vec) = dimensions;

-	for (uint32 k = 0; k < VARBITBYTES(vec); k++)
-		nx[k] = 0;
+	for (uint32 i = 0; i < VARBITBYTES(vec); i++)
+		nx[i] = 0;

-	for (int k = 0; k < dimensions; k++)
-		nx[k / 8] |= (x[k] > 0.5 ? 1 : 0) << (7 - (k % 8));
+	for (int i = 0; i < dimensions; i++)
+		nx[i / 8] |= (x[i] > 0.5 ? 1 : 0) << (7 - (i % 8));
 }

 static void
 VectorSumCenter(Pointer v, float *x)
 {
 	Vector	   *vec = (Vector *) v;
+	int			dim = vec->dim;

-	for (int k = 0; k < vec->dim; k++)
-		x[k] += vec->x[k];
+	/* Auto-vectorized */
+	for (int i = 0; i < dim; i++)
+		x[i] += vec->x[i];
 }

 static void
 HalfvecSumCenter(Pointer v, float *x)
 {
 	HalfVector *vec = (HalfVector *) v;
+	int			dim = vec->dim;

-	for (int k = 0; k < vec->dim; k++)
-		x[k] += HalfToFloat4(vec->x[k]);
+	/* Auto-vectorized on aarch64 */
+	for (int i = 0; i < dim; i++)
+		x[i] += HalfToFloat4(vec->x[i]);
 }

 static void
@@ -314,8 +324,8 @@ BitSumCenter(Pointer v, float *x)
 {
 	VarBit	   *vec = (VarBit *) v;

-	for (int k = 0; k < VARBITLEN(vec); k++)
-		x[k] += (float) (((VARBITS(vec)[k / 8]) >> (7 - (k % 8))) & 0x01);
+	for (int i = 0; i < VARBITLEN(vec); i++)
+		x[i] += (float) (((VARBITS(vec)[i / 8]) >> (7 - (i % 8))) & 0x01);
 }

 /*
--- a/src/ivfvacuum.c
+++ b/src/ivfvacuum.c
@@ -1,9 +1,16 @@
 #include "postgres.h"

+#include "access/genam.h"
 #include "access/generic_xlog.h"
+#include "access/itup.h"
 #include "commands/vacuum.h"
 #include "ivfflat.h"
 #include "storage/bufmgr.h"
+#include "utils/relcache.h"
+
+#if PG_VERSION_NUM >= 180000
+#define vacuum_delay_point() vacuum_delay_point(false)
+#endif

 /*
 * Bulk delete tuples from the index
--- a/src/sparsevec.c
+++ b/src/sparsevec.c
@@ -5,18 +5,27 @@

 #include "catalog/pg_type.h"
 #include "common/shortest_dec.h"
-#include "common/string.h"
 #include "fmgr.h"
 #include "halfutils.h"
 #include "halfvec.h"
+#include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
 #include "sparsevec.h"
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/float.h"
+#include "utils/fmgrprotos.h"
 #include "utils/lsyscache.h"
 #include "vector.h"

+#if PG_VERSION_NUM >= 160000
+#include "varatt.h"
+#endif
+
+#if PG_VERSION_NUM >= 170000
+#include "parser/scansup.h"
+#endif
+
 typedef struct SparseInputElement
 {
 	int32		index;
@@ -150,9 +159,9 @@ InitSparseVector(int dim, int nnz)
 	return result;
 }

-/*
- * Check for whitespace, since array_isspace() is static
- */
+#if PG_VERSION_NUM >= 170000
+#define sparsevec_isspace(ch) scanner_isspace(ch)
+#else
 static inline bool
 sparsevec_isspace(char ch)
 {
@@ -165,6 +174,7 @@ sparsevec_isspace(char ch)
 		return true;
 	return false;
 }
+#endif

 /*
 * Compare indices
@@ -172,10 +182,10 @@ sparsevec_isspace(char ch)
 static int
 CompareIndices(const void *a, const void *b)
 {
-	if (((SparseInputElement *) a)->index < ((SparseInputElement *) b)->index)
+	if (((const SparseInputElement *) a)->index < ((const SparseInputElement *) b)->index)
 		return -1;

-	if (((SparseInputElement *) a)->index > ((SparseInputElement *) b)->index)
+	if (((const SparseInputElement *) a)->index > ((const SparseInputElement *) b)->index)
 		return 1;

 	return 0;
--- a/src/vector.c
+++ b/src/vector.c
@@ -16,16 +16,20 @@
 #include "port.h"				/* for strtof() */
 #include "sparsevec.h"
 #include "utils/array.h"
-#include "utils/builtins.h"
 #include "utils/float.h"
+#include "utils/fmgrprotos.h"
 #include "utils/lsyscache.h"
-#include "utils/numeric.h"
+#include "utils/varbit.h"
 #include "vector.h"

 #if PG_VERSION_NUM >= 160000
 #include "varatt.h"
 #endif

+#if PG_VERSION_NUM >= 170000
+#include "parser/scansup.h"
+#endif
+
 #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
 #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))

@@ -35,7 +39,11 @@
 #define VECTOR_TARGET_CLONES
 #endif

+#if PG_VERSION_NUM >= 180000
+PG_MODULE_MAGIC_EXT(.name = "vector", .version = "0.8.3");
+#else
 PG_MODULE_MAGIC;
+#endif

 /*
 * Initialize index options and variables
@@ -125,9 +133,9 @@ InitVector(int dim)
 	return result;
 }

-/*
- * Check for whitespace, since array_isspace() is static
- */
+#if PG_VERSION_NUM >= 170000
+#define vector_isspace(ch) scanner_isspace(ch)
+#else
 static inline bool
 vector_isspace(char ch)
 {
@@ -140,6 +148,7 @@ vector_isspace(char ch)
 		return true;
 	return false;
 }
+#endif

 /*
 * Check state array
@@ -920,11 +929,13 @@ vector_concat(PG_FUNCTION_ARGS)
 	CheckDim(dim);
 	result = InitVector(dim);

-	for (int i = 0; i < a->dim; i++)
+	/* Auto-vectorized */
+	for (int i = 0, imax = a->dim; i < imax; i++)
 		result->x[i] = a->x[i];

-	for (int i = 0; i < b->dim; i++)
-		result->x[i + a->dim] = b->x[i];
+	/* Auto-vectorized */
+	for (int i = 0, imax = b->dim, start = a->dim; i < imax; i++)
+		result->x[i + start] = b->x[i];

 	PG_RETURN_POINTER(result);
 }
@@ -940,8 +951,21 @@ binary_quantize(PG_FUNCTION_ARGS)
 	float	   *ax = a->x;
 	VarBit	   *result = InitBitVector(a->dim);
 	unsigned char *rx = VARBITS(result);
+	int			i = 0;
+	int			count = (a->dim / 8) * 8;

-	for (int i = 0; i < a->dim; i++)
+	/* Auto-vectorized */
+	for (; i < count; i += 8)
+	{
+		unsigned char result_byte = 0;
+
+		for (int j = 0; j < 8; j++)
+			result_byte |= (ax[i + j] > 0) << (7 - j);
+
+		rx[i / 8] = result_byte;
+	}
+
+	for (; i < a->dim; i++)
 		rx[i / 8] |= (ax[i] > 0) << (7 - (i % 8));

 	PG_RETURN_VARBIT_P(result);
--- a/test/expected/halfvec.out
+++ b/test/expected/halfvec.out
@@ -540,6 +540,12 @@ SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::halfvec)
 01001110101
 (1 row)

+SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::halfvec);
+   binary_quantize   
+---------------------
+ 1110110110011011011
+(1 row)
+
 SELECT subvector('[1,2,3,4,5]'::halfvec, 1, 3);
 subvector 
 -----------
--- a/test/expected/hnsw_vector.out
+++ b/test/expected/hnsw_vector.out
@@ -123,6 +123,12 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 [0,0,0]
 (3 rows)

+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+ val 
+-----
+(0 rows)
+
 RESET hnsw.iterative_scan;
 RESET hnsw.ef_search;
 DROP TABLE t;
--- a/test/expected/ivfflat_vector.out
+++ b/test/expected/ivfflat_vector.out
@@ -110,6 +110,15 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 [1,1,1]
 (2 rows)

+TRUNCATE t;
+NOTICE:  ivfflat index created with little data
+DETAIL:  This will cause low recall.
+HINT:  Drop the index until the table has more data.
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+ val 
+-----
+(0 rows)
+
 RESET ivfflat.iterative_scan;
 RESET ivfflat.max_probes;
 DROP TABLE t;
--- a/test/expected/vector_type.out
+++ b/test/expected/vector_type.out
@@ -576,6 +576,12 @@ SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::vector);
 01001110101
 (1 row)

+SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::vector);
+   binary_quantize   
+---------------------
+ 1110110110011011011
+(1 row)
+
 SELECT subvector('[1,2,3,4,5]'::vector, 1, 3);
 subvector 
 -----------
--- a/test/sql/halfvec.sql
+++ b/test/sql/halfvec.sql
@@ -121,6 +121,7 @@ SELECT l2_normalize('[65504]'::halfvec);

 SELECT binary_quantize('[1,0,-1]'::halfvec);
 SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::halfvec);
+SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::halfvec);

 SELECT subvector('[1,2,3,4,5]'::halfvec, 1, 3);
 SELECT subvector('[1,2,3,4,5]'::halfvec, 3, 2);
--- a/test/sql/hnsw_vector.sql
+++ b/test/sql/hnsw_vector.sql
@@ -70,6 +70,9 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 SET hnsw.iterative_scan = relaxed_order;
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';

+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+
 RESET hnsw.iterative_scan;
 RESET hnsw.ef_search;
 DROP TABLE t;
--- a/test/sql/ivfflat_vector.sql
+++ b/test/sql/ivfflat_vector.sql
@@ -59,6 +59,9 @@ SELECT * FROM t ORDER BY val <-> '[3,3,3]';
 SET ivfflat.max_probes = 2;
 SELECT * FROM t ORDER BY val <-> '[3,3,3]';

+TRUNCATE t;
+SELECT * FROM t ORDER BY val <-> '[3,3,3]';
+
 RESET ivfflat.iterative_scan;
 RESET ivfflat.max_probes;
 DROP TABLE t;
--- a/test/sql/vector_type.sql
+++ b/test/sql/vector_type.sql
@@ -128,6 +128,7 @@ SELECT l2_normalize('[3e38]'::vector);

 SELECT binary_quantize('[1,0,-1]'::vector);
 SELECT binary_quantize('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]'::vector);
+SELECT binary_quantize('[1,2,3,-4,5,6,-7,8,1,-2,-3,4,5,-6,7,8,-1,2,3]'::vector);

 SELECT subvector('[1,2,3,4,5]'::vector, 1, 3);
 SELECT subvector('[1,2,3,4,5]'::vector, 3, 2);
--- a/test/t/005_ivfflat_query_recall.pl
+++ b/test/t/005_ivfflat_query_recall.pl
@@ -16,29 +16,19 @@ $node->safe_psql("postgres",
 	"INSERT INTO tst SELECT i, ARRAY[random(), random(), random()] FROM generate_series(1, 100000) i;"
 );

-# Check each index type
-my @operators = ("<->", "<#>", "<=>");
-my @opclasses = ("vector_l2_ops", "vector_ip_ops", "vector_cosine_ops");
+# Add index
+$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v vector_l2_ops);");

-for my $i (0 .. $#operators)
+# Test 100% recall
+for (1 .. 20)
 {
-	my $operator = $operators[$i];
-	my $opclass = $opclasses[$i];
-
-	# Add index
-	$node->safe_psql("postgres", "CREATE INDEX ON tst USING ivfflat (v $opclass);");
-
-	# Test 100% recall
-	for (1 .. 20)
-	{
-		my $id = int(rand() * 100000);
-		my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $id;");
-		my $res = $node->safe_psql("postgres", qq(
-			SET enable_seqscan = off;
-			SELECT v FROM tst ORDER BY v <-> '$query' LIMIT 1;
-		));
-		is($res, $query);
-	}
+	my $id = int(rand() * 100000);
+	my $query = $node->safe_psql("postgres", "SELECT v FROM tst WHERE i = $id;");
+	my $res = $node->safe_psql("postgres", qq(
+		SET enable_seqscan = off;
+		SELECT v FROM tst ORDER BY v <-> '$query' LIMIT 1;
+	));
+	is($res, $query);
 }

 done_testing();
--- a/test/t/045_hnsw_low_memory_build.pl
+++ b/test/t/045_hnsw_low_memory_build.pl
@@ -0,0 +1,29 @@
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+# Initialize node
+my $node = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->start;
+
+# Create table
+$node->safe_psql("postgres", "CREATE EXTENSION vector;");
+$node->safe_psql("postgres", "CREATE TABLE tst (v vector(3));");
+$node->safe_psql("postgres",
+	"INSERT INTO tst SELECT ARRAY[random(), random(), random()] FROM generate_series(1, 1000) i;"
+);
+
+my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
+	SET client_min_messages = DEBUG;
+	SET maintenance_work_mem = '3073kB';
+	ALTER TABLE tst SET (parallel_workers = 1);
+	CREATE INDEX ON tst USING hnsw (v vector_l2_ops);
+));
+is($ret, 0, $stderr);
+like($stderr, qr/using \d+ parallel workers/);
+like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem after 0 tuples/);
+
+done_testing();
--- a/vector.control
+++ b/vector.control
@@ -1,4 +1,4 @@
 comment = 'vector data type and ivfflat and hnsw access methods'
-default_version = '0.8.0'
+default_version = '0.8.3'
 module_pathname = '$libdir/vector'
 relocatable = true
Author	SHA1	Message	Date
Andrew Kane	586e7515ba	Version bump to 0.8.3 [skip ci]	2026-06-17 21:26:13 -07:00
Andrew Kane	a619d0b34d	Simplified logic for HNSW vacuum progress [skip ci]	2026-06-17 12:18:17 -07:00
Andrew Kane	ad222abf48	Improved HNSW vacuum progress for round block sizes [skip ci]	2026-06-16 14:21:23 -07:00
Andrew Kane	8a81a3fe3a	Added HNSW_VACUUM_PROGRESS flag [skip ci]	2026-06-16 13:53:40 -07:00
Andrew Kane	fc804925a4	Added todo [skip ci]	2026-06-16 12:33:25 -07:00
Andrew Kane	98dc4aa6d0	Added pass to confirm graph was repaired for HNSW vacuuming	2026-06-16 10:51:54 -07:00
Andrew Kane	7b72aeeff3	Added check for deleted element [skip ci]	2026-06-16 10:13:40 -07:00
Andrew Kane	4682d51e5d	Added memory usage for HNSW vacuuming [skip ci]	2026-06-16 10:10:13 -07:00
Andrew Kane	91fe2e62e7	Added benchmarking for HNSW vacuuming [skip ci]	2026-06-16 09:53:47 -07:00
Andrew Kane	eedba7ee14	Fixed possible index corruption with HNSW vacuuming - resolves #988 Co-authored-by: Bhagyesh Chaturvedi <bhagyeshc@google.com>	2026-06-16 09:44:34 -07:00
Andrew Kane	32284ba28a	Updated CI for latest windows-2025 image [skip ci]	2026-06-15 19:24:02 -07:00
Andrew Kane	1f68c73c96	Updated changelog [skip ci]	2026-06-10 20:20:55 -07:00
Andrew Kane	f15bc0904f	Updated changelog [skip ci]	2026-06-10 20:13:55 -07:00
Andrew Kane	2b7ad083dc	Always use inline pg_popcount64 for Postgres 19+ - #985	2026-06-10 20:08:03 -07:00
Andrew Kane	421a35fcdc	Fixed performance regression with Hamming distance and Jaccard distance with Postgres 18 - fixes #985	2026-06-10 19:55:21 -07:00
Andrew Kane	12368bd79c	Updated readme [skip ci]	2026-05-30 11:58:10 -07:00
Andrew Kane	88a0085459	Improved casting [skip ci]	2026-05-26 14:09:19 -07:00
Andrew Kane	ea23884efd	Ran latest pgindent [skip ci]	2026-05-26 13:39:55 -07:00
Andrew Kane	3351f3d43e	Simplified comment [skip ci]	2026-05-26 13:39:20 -07:00
Andrew Kane	d238409bec	Merge pull request #979 from xingtanzjr/fix_ivf_create_concurrently Fix CREATE INDEX CONCURRENTLY crash on ivfflat with assertion-enabled PostgreSQL	2026-04-27 03:28:52 -07:00
Zhang Jinrui	529f37175b	Fix CREATE INDEX CONCURRENTLY crash on ivfflat with assertion-enabled PostgreSQL by passing anyvisible=false in ivfflat SampleRows	2026-04-27 10:15:32 +00:00
Andrew Kane	bce3946392	Added link to PgHero [skip ci]	2026-04-27 02:11:17 -07:00
Andrew Kane	13cb253d30	Added link to PgDog [skip ci]	2026-04-27 00:09:20 -07:00
Andrew Kane	41b3cdc011	Updated readme [skip ci]	2026-04-26 14:47:17 -07:00
Andrew Kane	ec02a96239	Updated readme [skip ci]	2026-04-26 14:45:12 -07:00
Andrew Kane	610d95b8d2	Moved scaling section [skip ci]	2026-04-26 13:12:28 -07:00
Andrew Kane	609d01f4c6	Added more scaling advice to readme [skip ci]	2026-04-26 12:56:07 -07:00
Andrew Kane	a7551a61ca	Updated style [skip ci]	2026-04-14 16:21:32 -07:00
lychen0613	5c9a97af3b	optimize: defer sample normalization when building ivfflat	2026-04-14 16:14:17 -07:00
Andrew Kane	8af675cd25	Added comment [skip ci]	2026-04-14 16:10:45 -07:00
Andrew Kane	97e0ed4464	Fixed compilation errors with Postgres 19	2026-04-14 15:46:10 -07:00
Andrew Kane	533ad160e0	Fixed CI	2026-04-14 13:51:04 -07:00
Andrew Kane	dfbd724a1f	Changed rowstoskip to double to match reservoir_get_next_S	2026-04-14 13:38:20 -07:00
lychen0613	5f270c9663	Add benchmark for ivfflat sample rows	2026-04-14 21:56:10 +08:00
lychen0613	352ec5be29	fix: correctly use reservoir_get_next_S when building ivfflat	2026-04-14 17:57:40 +08:00
Andrew Kane	17916cad00	Improved includes [skip ci]	2026-02-25 17:35:23 -08:00
Andrew Kane	284f2f0fec	Improved casting for HNSW scan memory limit [skip ci]	2026-02-25 12:59:17 -08:00
Andrew Kane	edd49863ba	Improved limit for HNSW scan memory	2026-02-25 12:52:29 -08:00
Andrew Kane	411291189d	Updated changelog [skip ci]	2026-02-25 11:28:43 -08:00
Andrew Kane	cab9da72c0	Version bump to 0.8.2 [skip ci]	2026-02-25 10:46:57 -08:00
Andrew Kane	d702ee31f2	Added test case for low memory HNSW index builds	2026-02-25 10:41:17 -08:00
Andrew Kane	9ed39fb94b	Hardened shared memory allocator Co-authored-by: chungkn from OneMount Group <kimngocchung.k2a@gmail.com>	2026-02-25 10:39:01 -08:00
Andrew Kane	b7e680d41a	Fixed integer wraparound leading to buffer overflow with parallel HNSW index build Reported-by: chungkn from OneMount Group <kimngocchung.k2a@gmail.com>	2026-02-25 10:35:14 -08:00
Andrew Kane	f5841f07fd	Updated CI	2026-02-23 14:06:35 -08:00
Andrew Kane	35ab919bf5	Switched to statically-allocated IndexAmRoutine for Postgres 19 [skip ci]	2026-01-21 16:40:35 -08:00
Andrew Kane	5c08be0291	Updated checkout action [skip ci]	2026-01-15 20:45:56 -08:00
Jack Huang	544686feb1	Add RISC-V64 architecture support in Makefile (#948 )	2026-01-14 06:11:58 -08:00
Andrew Kane	1a8ebb1129	Updated license year [skip ci]	2026-01-05 13:01:58 -08:00
Andrew Kane	aca89aec73	Switched to scanner_isspace for Postgres 17+ - resolves #942	2026-01-05 12:54:04 -08:00
Andrew Kane	c9875f5150	Simplified test case - resolves #943	2026-01-05 12:42:14 -08:00
Andrew Kane	df2c3d301f	Added link to pgvector-cobol [skip ci]	2025-12-13 11:33:35 -08:00
Andrew Kane	d557f1c9b1	Added link to pgvector-tcl [skip ci]	2025-12-11 23:22:56 -08:00
Andrew Kane	3d6dd20469	Removed unneeded check	2025-12-11 16:28:45 -08:00
Andrew Kane	c711da411c	Improved includes for indexes	2025-12-11 15:35:37 -08:00
Andrew Kane	3975e212cc	Improved includes for ivfkmeans [skip ci]	2025-12-11 13:26:36 -08:00
Andrew Kane	2b741ef068	Improved includes for types	2025-12-11 00:35:18 -08:00
Andrew Kane	f85a7d3fd0	Keep clang-tidy happy [skip ci]	2025-12-10 20:03:26 -08:00
Andrew Kane	5c25f81789	Fixed inconsistent parameter names [skip ci]	2025-12-10 18:49:00 -08:00
Andrew Kane	d72ee71f23	Fixed compilation errors with Postgres 19	2025-12-10 14:59:29 -08:00
Andrew Kane	b7c0085afe	Added instructions for APK [skip ci]	2025-12-01 22:52:23 -08:00
Andrew Kane	2b742576b4	Added links [skip ci]	2025-11-24 12:41:30 -08:00
Andrew Kane	a126c02184	Added link to pgvector-ada [skip ci]	2025-10-30 22:29:36 -07:00
Andrew Kane	744362305b	Updated changelog [skip ci]	2025-10-30 11:48:44 -07:00
Andrew Kane	69e78e36a3	Added link to pgvetor-pascal [skip ci]	2025-10-30 11:40:29 -07:00
Andrew Kane	8dd37d4fb2	Improved install target for Windows - #917 [skip ci]	2025-10-30 11:03:59 -07:00
Andrew Kane	a24125839a	Updated readme [skip ci]	2025-10-28 10:25:34 -07:00
Andrew Kane	2ee3113417	Updated readme [skip ci]	2025-10-22 13:10:58 -07:00
Andrew Kane	3ebb9a506e	Added varatt headers for Postgres 16+	2025-10-22 11:56:05 -07:00
Andrew Kane	727d4836d2	Fixed Index Searches in EXPLAIN output for Postgres 18	2025-10-22 11:47:11 -07:00
Andrew Kane	d823c44591	Removed unused parameter [skip ci]	2025-09-27 15:58:18 -07:00
Andrew Kane	8f4aa0751c	Removed unused parameters [skip ci]	2025-09-27 15:51:32 -07:00
Andrew Kane	1ff30720c7	Updated readme [skip ci]	2025-09-26 13:02:37 -07:00
Andrew Kane	4cef9213ec	Added Docker images for Postgres 18 [skip ci]	2025-09-25 20:00:10 -07:00
Andrew Kane	90add68d6f	Updated style to be consistent with Mac [skip ci]	2025-09-25 15:55:30 -07:00
Andrew Kane	bc05dbf312	Updated CI for Postgres 18 [skip ci]	2025-09-25 15:54:37 -07:00
Andrew Kane	02fefc0e3f	Updated readme for Postgres 18 [skip ci]	2025-09-25 15:46:28 -07:00
Andrew Kane	e7899b1250	Updated checkout action [skip ci]	2025-09-22 11:55:01 -07:00
Andrew Kane	9e29dff78c	Updated CI [skip ci]	2025-09-19 19:16:08 -07:00
Andrew Kane	13c0dbf530	Test with macos-15-intel on CI [skip ci]	2025-09-19 18:34:51 -07:00
Andrew Kane	c205a12107	Updated CI [skip ci]	2025-09-05 17:03:37 -07:00
Andrew Kane	db3755a58f	Added windows-2025 to CI [skip ci]	2025-09-05 16:41:22 -07:00
Andrew Kane	778dacf20c	Version bump to 0.8.1 [skip ci]	2025-09-04 17:51:09 -07:00
Andrew Kane	3f687687ee	Fixed compilation error with Postgres 19	2025-09-04 15:58:23 -07:00
Andrew Kane	815f48e489	Updated changelog [skip ci]	2025-09-04 15:36:18 -07:00
Andrew Kane	67e648b13e	Added another test for binary_quantize function [skip ci]	2025-08-29 01:09:18 -07:00
Andrew Kane	c3ff955231	Reordered supported Docker tags [skip ci]	2025-08-23 19:57:14 -07:00
Andrew Kane	6f46a1035d	Added new Docker images to readme [skip ci]	2025-08-23 19:46:07 -07:00
Andrew Kane	bbe66e821b	Added Docker images for Debian 13 / Trixie [skip ci]	2025-08-23 19:34:11 -07:00
Julien Rouhaud	dd3a1e9137	Use NIL for empty lists (#890 ) Postgres standard way to check for list emptiness is to compare a pointer to NIL rather than NULL.	2025-08-23 03:31:22 -07:00
Andrew Kane	ea4746f6c0	Updated comment [skip ci]	2025-08-19 14:26:36 -07:00
Andrew Kane	6aec80ccdb	Improved performance of binary_quantize function for halfvec	2025-08-19 13:47:11 -07:00
Luca Giacchino	0c9070ba82	Add autovectorized implementation of binary quantize	2025-08-19 13:18:50 -07:00
Andrew Kane	30382418da	Ran pgindent [skip ci]	2025-08-19 12:44:54 -07:00
Andrew Kane	26b50e536f	Updated FreeBSD package name in readme [skip ci]	2025-08-09 12:09:29 -07:00
Andrew Kane	e29fc3aa1a	Switched to PG_MODULE_MAGIC_EXT for Postgres 18+ [skip ci]	2025-07-31 19:59:27 -07:00
Andrew Kane	6ef7fccb5c	Added Postgres 19 to CI [skip ci]	2025-07-27 19:50:46 -07:00
Andrew Kane	5b8b68ba1d	Use consistent style [skip ci]	2025-07-27 18:19:28 -07:00
Andrew Kane	3be8693c13	Added supported Docker tags to readme [skip ci]	2025-07-27 13:44:59 -07:00
Andrew Kane	247c8dc8a5	Updated changelog [skip ci]	2025-07-27 13:25:51 -07:00
Andrew Kane	3600ab93e5	Added Docker images with -bookworm suffix [skip ci]	2025-07-27 13:14:29 -07:00
Andrew Kane	665db75a3c	Updated Dockerfile to use release [skip ci]	2025-07-27 12:34:23 -07:00
Andrew Kane	44163d0a97	Revert "Added OS to Dockerfile [skip ci]" This reverts commit `33ca8a61e2`.	2025-07-09 22:01:34 -07:00
Andrew Kane	33ca8a61e2	Added OS to Dockerfile [skip ci]	2025-07-09 16:25:05 -07:00
Andrew Kane	742e2d1d28	Synced .dockerignore with .gitignore [skip ci]	2025-07-09 16:22:43 -07:00
Andrew Kane	a7c49d8283	Updated readme [skip ci]	2025-07-01 11:13:29 -07:00
Andrew Kane	ae9ee81e4d	Fixed relaxed results example for Postgres 17 - #862 [skip ci]	2025-07-01 10:11:50 -07:00
Andrew Kane	fa1dee4e3b	Updated link [skip ci]	2025-07-01 03:35:26 -07:00
Andrew Kane	e6bad96a03	Ran pgindent [skip ci]	2025-06-18 20:07:46 -07:00
Jon Daniel	3a49d141b3	Vectorizing vector_concat for improved performance (#861 ) * Vectorizing vector_concat for improved performance On an ARM chip this should generate SIMD instructions to copy the two incoming vectors to the new vector as opposed to doing it all in software. * Moving declarations to above CheckDim * Removing const from dims * Formatting	2025-06-18 20:06:32 -07:00
Andrew Kane	ce09c9a27a	Improved variable names [skip ci]	2025-06-18 17:32:37 -07:00
Andrew Kane	870ca6724d	Fixed CI [skip ci]	2025-06-18 16:26:30 -07:00
Andrew Kane	8ad680f009	Ran pgindent [skip ci]	2025-06-18 16:11:15 -07:00
Jon Daniel	fe697e8788	vectorize: optimize VectorSumCenter and HalfvecSumCenter (#860 ) * vectorize: optimize VectorSumCenter and HalfvecSumCenter The functions VectorSumCenter and HalfvecSumCenter were not being vectorized by the compiler. A few slight changes will allow these optimizations to take place and get a performance boost by utilizing SIMD instructions. This optimization helps improve performance of vector operations in IVF index building and updating. * Removing const, commenting that it is only vectoirzed on ARM	2025-06-18 16:09:43 -07:00
Andrew Kane	bf28ed8176	Set random seed for IVFFLAT_BENCH [skip ci]	2025-06-18 10:57:18 -07:00
Andrew Kane	799cfebf70	Updated readme [skip ci]	2025-06-04 14:29:53 -07:00
Andrew Kane	3cd1f09f66	Updated Windows installation instructions - #593 [skip ci]	2025-06-04 14:18:53 -07:00
Andrew Kane	e2efe62fe5	Updated readme [skip ci]	2025-05-12 13:11:43 -07:00
Andrew Kane	7b58352336	Updated readme [skip ci]	2025-05-06 21:39:51 -07:00
Andrew Kane	83d410eae9	Updated readme [skip ci]	2025-04-13 13:07:31 -07:00
Andrew Kane	ebbfe8dba0	Fixed CI for Postgres 18 [skip ci]	2025-04-05 13:05:27 -07:00
Andrew Kane	e575866297	Revert "Fixed warnings with Postgres 18 [skip ci]" This reverts commit `32e95a8598`.	2025-04-05 12:56:00 -07:00
Andrew Kane	35f4f7fc80	Improved warning check [skip ci]	2025-04-05 12:38:30 -07:00
Andrew Kane	32e95a8598	Fixed warnings with Postgres 18 [skip ci]	2025-04-05 12:13:38 -07:00
Andrew Kane	a03dc5b7d0	Added fields to IndexAmRoutine for Postgres 18 [skip ci]	2025-04-05 11:31:57 -07:00
Andrew Kane	d365aac370	Added note about index options to index build time docs - #807 [skip ci]	2025-03-26 11:56:38 -07:00
Andrew Kane	05182479a2	Added ARM to CI [skip ci]	2025-03-24 19:26:52 -07:00
Andrew Kane	cc0958dec5	Removed deprecated runner image [skip ci]	2025-03-24 19:18:01 -07:00
Andrew Kane	4af2b06dc5	Improved installation note [skip ci]	2025-03-23 14:59:28 -07:00
Andrew Kane	880dc4d6b9	Added Windows installation note about mismatched architecture - #593 #797 #804 [skip ci]	2025-03-23 11:51:10 -07:00
Andrew Kane	fef635c9e5	Updated readme [skip ci]	2025-02-20 00:05:35 -08:00
Andrew Kane	78ed8f1157	Fixed compilation error with Postgres 18 - fixes #779	2025-02-17 16:38:14 -08:00
Andrew Kane	f11e4d7b20	Updated readme [skip ci]	2025-02-17 13:22:50 -08:00
Andrew Kane	aafdf4167d	Improved missing SDK docs [skip ci]	2025-02-17 13:16:31 -08:00
Andrew Kane	656b059258	Added tests for iterative index scan with empty index - resolves #679 [skip ci]	2025-02-13 16:28:14 -08:00
Andrew Kane	7cf9980696	Updated Windows installation notes for Postgres 17.3 - closes #669 [skip ci]	2025-02-13 10:04:22 -08:00