Less space [skip ci]

And rewrite some of the comments in InsertTuple(), to also give more
of a high-level overview of the flow.

.github/workflows/build.yml

@@ -62,7 +62,7 @@ jobs:
           wget -q https://github.com/postgres/postgres/archive/refs/tags/REL_14_5.tar.gz
           tar xf REL_14_5.tar.gz
       - run: make prove_installcheck PROVE_FLAGS="-I ./postgres-REL_14_5/src/test/perl" PERL5LIB="/Users/runner/perl5/lib/perl5"
-      - run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make
+      - run: make clean && /usr/local/opt/llvm@15/bin/scan-build --status-bugs make PG_CFLAGS="-DUSE_ASSERT_CHECKING"
   windows:
     runs-on: windows-latest
     if: ${{ !startsWith(github.ref_name, 'mac') }}

CHANGELOG.md

@@ -1,6 +1,10 @@
-## 0.6.0 (unreleased)
+## 0.5.2 (unreleased)
 
-- Added support for sparse vectors
+- Improved performance of HNSW
+- Added support for parallel index builds for HNSW
+- Reduced memory usage for HNSW index builds
+- Reduced WAL generation for HNSW index builds
+- Fixed `invalid memory alloc request size` error with HNSW index build
 
 ## 0.5.1 (2023-10-10)
 

Makefile

@@ -3,8 +3,8 @@ EXTVERSION = 0.5.1
 
 MODULE_big = vector
 DATA = $(wildcard sql/*--*.sql)
-OBJS = src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/svector.o src/vector.o
+OBJS = src/hnsw.o src/hnswbuild.o src/hnswinsert.o src/hnswscan.o src/hnswutils.o src/hnswvacuum.o src/ivfbuild.o src/ivfflat.o src/ivfinsert.o src/ivfkmeans.o src/ivfscan.o src/ivfutils.o src/ivfvacuum.o src/vector.o
-HEADERS = src/svector.h src/vector.h
+HEADERS = src/vector.h
 
 TESTS = $(wildcard test/sql/*.sql)
 REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))

Makefile.win

@@ -1,8 +1,8 @@
 EXTENSION = vector
 EXTVERSION = 0.5.1
 
-OBJS = 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\svector.obj src\vector.obj
+OBJS = 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\vector.obj
-HEADERS = src\svector.h src\vector.h
+HEADERS = src\vector.h
 
 REGRESS = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged
 REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)

README.md

@@ -10,7 +10,7 @@ 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
 
-[![Build Status](https://github.com/pgvector/pgvector/workflows/build/badge.svg?branch=master)](https://github.com/pgvector/pgvector/actions)
+[![Build Status](https://github.com/pgvector/pgvector/actions/workflows/build.yml/badge.svg)](https://github.com/pgvector/pgvector/actions)
 
 ## Installation
 
@@ -161,8 +161,97 @@ You can add an index to use approximate nearest neighbor search, which trades so
 
 Supported index types are:
 
-- [IVFFlat](#ivfflat)
 - [HNSW](#hnsw) - added in 0.5.0
+- [IVFFlat](#ivfflat)
+
+## HNSW
+
+An HNSW index creates a multilayer graph. It has better query performance than IVFFlat (in terms of speed-recall tradeoff), but has slower build times and uses more memory. Also, an index can be created without any data in the table since there isn’t a training step like IVFFlat.
+
+Add an index for each distance function you want to use.
+
+L2 distance
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
+```
+
+Inner product
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding vector_ip_ops);
+```
+
+Cosine distance
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
+```
+
+Vectors with up to 2,000 dimensions can be indexed.
+
+### Index Options
+
+Specify HNSW parameters
+
+- `m` - the max number of connections per layer (16 by default)
+- `ef_construction` - the size of the dynamic candidate list for constructing the graph (64 by default)
+
+```sql
+CREATE INDEX ON items USING hnsw (embedding vector_l2_ops) WITH (m = 16, ef_construction = 64);
+```
+
+A higher value of `ef_construction` provides better recall at the cost of index build time / insert speed.
+
+### Query Options
+
+Specify the size of the dynamic candidate list for search (40 by default)
+
+```sql
+SET hnsw.ef_search = 100;
+```
+
+A higher value provides better recall at the cost of speed.
+
+Use `SET LOCAL` inside a transaction to set it for a single query
+
+```sql
+BEGIN;
+SET LOCAL hnsw.ef_search = 100;
+SELECT ...
+COMMIT;
+```
+
+### Index Build Time
+
+Indexes build significantly faster when the graph fits into `maintenance_work_mem`
+
+```sql
+SET maintenance_work_mem = '8GB';
+```
+
+A notice is shown when the graph no longer fits
+
+```text
+NOTICE:  hnsw graph no longer fits into maintenance_work_mem after 100000 tuples
+DETAIL:  Building will take significantly more time.
+HINT:  Increase maintenance_work_mem to speed up builds.
+```
+
+Note: Do not set `maintenance_work_mem` so high that it exhausts the memory on the server
+
+### Indexing Progress
+
+Check [indexing progress](https://www.postgresql.org/docs/current/progress-reporting.html#CREATE-INDEX-PROGRESS-REPORTING) with Postgres 12+
+
+```sql
+SELECT phase, round(100.0 * blocks_done / nullif(blocks_total, 0), 1) AS "%" FROM pg_stat_progress_create_index;
+```
+
+The phases for HNSW are:
+
+1. `initializing`
+2. `loading tuples`
 
 ## IVFFlat
 
@@ -232,75 +321,6 @@ The phases for IVFFlat are:
 
 Note: `%` is only populated during the `loading tuples` phase
 
-## HNSW
-
-An HNSW index creates a multilayer graph. It has slower build times and uses more memory than IVFFlat, but has better query performance (in terms of speed-recall tradeoff). There’s no training step like IVFFlat, so the index can be created without any data in the table.
-
-Add an index for each distance function you want to use.
-
-L2 distance
-
-```sql
-CREATE INDEX ON items USING hnsw (embedding vector_l2_ops);
-```
-
-Inner product
-
-```sql
-CREATE INDEX ON items USING hnsw (embedding vector_ip_ops);
-```
-
-Cosine distance
-
-```sql
-CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
-```
-
-Vectors with up to 2,000 dimensions can be indexed.
-
-### Index Options
-
-Specify HNSW parameters
-
-- `m` - the max number of connections per layer (16 by default)
-- `ef_construction` - the size of the dynamic candidate list for constructing the graph (64 by default)
-
-```sql
-CREATE INDEX ON items USING hnsw (embedding vector_l2_ops) WITH (m = 16, ef_construction = 64);
-```
-
-### Query Options
-
-Specify the size of the dynamic candidate list for search (40 by default)
-
-```sql
-SET hnsw.ef_search = 100;
-```
-
-A higher value provides better recall at the cost of speed.
-
-Use `SET LOCAL` inside a transaction to set it for a single query
-
-```sql
-BEGIN;
-SET LOCAL hnsw.ef_search = 100;
-SELECT ...
-COMMIT;
-```
-
-### Indexing Progress
-
-Check [indexing progress](https://www.postgresql.org/docs/current/progress-reporting.html#CREATE-INDEX-PROGRESS-REPORTING) with Postgres 12+
-
-```sql
-SELECT phase, round(100.0 * blocks_done / nullif(blocks_total, 0), 1) AS "%" FROM pg_stat_progress_create_index;
-```
-
-The phases for HNSW are:
-
-1. `initializing`
-2. `loading tuples`
-
 ## Filtering
 
 There are a few ways to index nearest neighbor queries with a `WHERE` clause
@@ -318,8 +338,7 @@ CREATE INDEX ON items (category_id);
 Or a [partial index](https://www.postgresql.org/docs/current/indexes-partial.html) on the vector column for approximate search
 
 ```sql
-CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 100)
+CREATE INDEX ON items USING hnsw (embedding vector_l2_ops) WHERE (category_id = 123);
-    WHERE (category_id = 123);
 ```
 
 Use [partitioning](https://www.postgresql.org/docs/current/ddl-partitioning.html) for approximate search on many different values of the `WHERE` columns
@@ -369,26 +388,6 @@ To speed up queries with an IVFFlat index, increase the number of inverted lists
 CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 1000);
 ```
 
-## Sparse Vectors
-
-Create a sparse vector column with 10 dimensions
-
-```sql
-CREATE TABLE items (id bigserial PRIMARY KEY, embedding svector(10));
-```
-
-Insert vectors
-
-```sql
-INSERT INTO items (embedding) VALUES ('(0,1),(1,2),(2,3)|10|'), ('(0,4),(1,5),(4,6)|10|');
-```
-
-Get the nearest neighbors by L2 distance
-
-```sql
-SELECT * FROM items ORDER BY embedding <-> '(0,3),(1,1),(2,2)|10|' LIMIT 5;
-```
-
 ## 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.
@@ -397,7 +396,7 @@ Language | Libraries / Examples
 --- | ---
 C | [pgvector-c](https://github.com/pgvector/pgvector-c)
 C++ | [pgvector-cpp](https://github.com/pgvector/pgvector-cpp)
-C# | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
+C#, F#, Visual Basic | [pgvector-dotnet](https://github.com/pgvector/pgvector-dotnet)
 Crystal | [pgvector-crystal](https://github.com/pgvector/pgvector-crystal)
 Dart | [pgvector-dart](https://github.com/pgvector/pgvector-dart)
 Elixir | [pgvector-elixir](https://github.com/pgvector/pgvector-elixir)
@@ -406,8 +405,10 @@ 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)
+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)
 Perl | [pgvector-perl](https://github.com/pgvector/pgvector-perl)
 PHP | [pgvector-php](https://github.com/pgvector/pgvector-php)
 Python | [pgvector-python](https://github.com/pgvector/pgvector-python)
@@ -480,6 +481,14 @@ and query with:
 SELECT * FROM items ORDER BY embedding::vector(3) <-> '[3,1,2]' LIMIT 5;
 ```
 
+#### Do indexes need to fit into memory?
+
+No, but like other index types, you’ll likely see better performance if they do. You can get the size of an index with:
+
+```sql
+SELECT pg_size_pretty(pg_relation_size('index_name'));
+```
+
 ## Troubleshooting
 
 #### Why isn’t a query using an index?
@@ -579,10 +588,10 @@ 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-15
+sudo apt install postgresql-server-dev-16
 ```
 
-Note: Replace `15` with your Postgres server version
+Note: Replace `16` with your Postgres server version
 
 ### Windows
 
@@ -597,7 +606,7 @@ Note: The exact path will vary depending on your Visual Studio version and editi
 Then use `nmake` to build:
 
 ```cmd
-set "PGROOT=C:\Program Files\PostgreSQL\15"
+set "PGROOT=C:\Program Files\PostgreSQL\16"
 git clone --branch v0.5.1 https://github.com/pgvector/pgvector.git
 cd pgvector
 nmake /F Makefile.win
@@ -647,22 +656,22 @@ 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-15-pgvector
+sudo apt install postgresql-16-pgvector
 ```
 
-Note: Replace `15` with your Postgres server version
+Note: Replace `16` 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_15
+sudo yum install pgvector_16
 # or
-sudo dnf install pgvector_15
+sudo dnf install pgvector_16
 ```
 
-Note: Replace `15` with your Postgres server version
+Note: Replace `16` with your Postgres server version
 
 ### conda-forge
 

sql/vector--0.5.1--0.6.0.sql

@@ -1,79 +0,0 @@
--- complain if script is sourced in psql, rather than via CREATE EXTENSION
-\echo Use "ALTER EXTENSION vector UPDATE TO '0.6.0'" to load this file. \quit
-
-CREATE TYPE svector;
-
-CREATE FUNCTION svector_in(cstring, oid, integer) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_out(svector) RETURNS cstring
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_typmod_in(cstring[]) RETURNS integer
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_recv(internal, oid, integer) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_send(svector) RETURNS bytea
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE TYPE svector (
-	INPUT     = svector_in,
-	OUTPUT    = svector_out,
-	TYPMOD_IN = svector_typmod_in,
-	RECEIVE   = svector_recv,
-	SEND      = svector_send,
-	STORAGE   = external
-);
-
-CREATE FUNCTION l2_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION inner_product(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION cosine_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION jaccard_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_jaccard_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_l2_squared_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_negative_inner_product(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector(svector, integer, boolean) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION vector_to_svector(vector, integer, boolean) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_to_vector(svector, integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE CAST (svector AS svector)
-	WITH FUNCTION svector(svector, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (svector AS vector)
-	WITH FUNCTION svector_to_vector(svector, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (vector AS svector)
-	WITH FUNCTION vector_to_svector(vector, integer, boolean) AS IMPLICIT;
-
-CREATE OPERATOR <-> (
-	LEFTARG = svector, RIGHTARG = svector, PROCEDURE = l2_distance,
-	COMMUTATOR = '<->'
-);
-
-CREATE OPERATOR <#> (
-	LEFTARG = svector, RIGHTARG = svector, PROCEDURE = svector_negative_inner_product,
-	COMMUTATOR = '<#>'
-);
-
-CREATE OPERATOR <=> (
-	LEFTARG = svector, RIGHTARG = svector, PROCEDURE = cosine_distance,
-	COMMUTATOR = '<=>'
-);

sql/vector.sql

@@ -290,92 +290,3 @@ CREATE OPERATOR CLASS vector_cosine_ops
 	OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops,
 	FUNCTION 1 vector_negative_inner_product(vector, vector),
 	FUNCTION 2 vector_norm(vector);
-
---- svector type
-
-CREATE TYPE svector;
-
-CREATE FUNCTION svector_in(cstring, oid, integer) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_out(svector) RETURNS cstring
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_typmod_in(cstring[]) RETURNS integer
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_recv(internal, oid, integer) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_send(svector) RETURNS bytea
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE TYPE svector (
-	INPUT     = svector_in,
-	OUTPUT    = svector_out,
-	TYPMOD_IN = svector_typmod_in,
-	RECEIVE   = svector_recv,
-	SEND      = svector_send,
-	STORAGE   = external
-);
-
--- svector functions
-
-CREATE FUNCTION l2_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_l2_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION inner_product(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_inner_product' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION cosine_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_cosine_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION jaccard_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME', 'svector_jaccard_distance' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
--- svector private functions
-
-CREATE FUNCTION svector_l2_squared_distance(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_negative_inner_product(svector, svector) RETURNS float8
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
--- svector cast functions
-
-CREATE FUNCTION svector(svector, integer, boolean) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION vector_to_svector(vector, integer, boolean) RETURNS svector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
-CREATE FUNCTION svector_to_vector(svector, integer, boolean) RETURNS vector
-	AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-
--- svector casts
-
-CREATE CAST (svector AS svector)
-	WITH FUNCTION svector(svector, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (svector AS vector)
-	WITH FUNCTION svector_to_vector(svector, integer, boolean) AS IMPLICIT;
-
-CREATE CAST (vector AS svector)
-	WITH FUNCTION vector_to_svector(vector, integer, boolean) AS IMPLICIT;
-
--- svector operators
-
-CREATE OPERATOR <-> (
-	LEFTARG = svector, RIGHTARG = svector, PROCEDURE = l2_distance,
-	COMMUTATOR = '<->'
-);
-
-CREATE OPERATOR <#> (
-	LEFTARG = svector, RIGHTARG = svector, PROCEDURE = svector_negative_inner_product,
-	COMMUTATOR = '<#>'
-);
-
-CREATE OPERATOR <=> (
-	LEFTARG = svector, RIGHTARG = svector, PROCEDURE = cosine_distance,
-	COMMUTATOR = '<=>'
-);

src/hnsw.c

@@ -14,14 +14,45 @@
 #endif
 
 int			hnsw_ef_search;
+int			hnsw_lock_tranche_id;
 static relopt_kind hnsw_relopt_kind;
 
+/*
+ * Assign a tranche ID for our LWLocks. This only needs to be done by one
+ * backend, as the tranche ID is remembered in shared memory.
+ *
+ * This shared memory area is very small, so we just allocate it from the
+ * "slop" that PostgreSQL reserves for small allocations like this. If
+ * this grows bigger, we should use a shmem_request_hook and
+ * RequestAddinShmemSpace() to pre-reserve space for this.
+ */
+static void
+HnswInitLockTranche(void)
+{
+	int		   *tranche_ids;
+	bool		found;
+
+	LWLockAcquire(AddinShmemInitLock, LW_EXCLUSIVE);
+	tranche_ids = ShmemInitStruct("hnsw LWLock ids",
+								  sizeof(int) * 1,
+								  &found);
+	if (!found)
+		tranche_ids[0] = LWLockNewTrancheId();
+	hnsw_lock_tranche_id = tranche_ids[0];
+	LWLockRelease(AddinShmemInitLock);
+
+	/* Per-backend registration of the tranche ID */
+	LWLockRegisterTranche(hnsw_lock_tranche_id, "HnswBuild");
+}
+
 /*
  * Initialize index options and variables
  */
 void
 HnswInit(void)
 {
+	HnswInitLockTranche();
+
 	hnsw_relopt_kind = add_reloption_kind();
 	add_int_reloption(hnsw_relopt_kind, "m", "Max number of connections",
 					  HNSW_DEFAULT_M, HNSW_MIN_M, HNSW_MAX_M

src/hnsw.h

@@ -4,9 +4,11 @@
 #include "postgres.h"
 
 #include "access/generic_xlog.h"
+#include "access/parallel.h"
 #include "access/reloptions.h"
 #include "nodes/execnodes.h"
 #include "port.h"				/* for random() */
+#include "utils/relptr.h"
 #include "utils/sampling.h"
 #include "vector.h"
 
@@ -14,6 +16,10 @@
 #error "Requires PostgreSQL 11+"
 #endif
 
+#if PG_VERSION_NUM < 120000
+#include "access/relscan.h"
+#endif
+
 #define HNSW_MAX_DIM 2000
 
 /* Support functions */
@@ -59,7 +65,7 @@
 
 #define HNSW_MAX_SIZE (BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - sizeof(ItemIdData))
 
-#define HNSW_ELEMENT_TUPLE_SIZE(_dim)	MAXALIGN(offsetof(HnswElementTupleData, vec) + VECTOR_SIZE(_dim))
+#define HNSW_ELEMENT_TUPLE_SIZE(size)	MAXALIGN(offsetof(HnswElementTupleData, data) + (size))
 #define HNSW_NEIGHBOR_TUPLE_SIZE(level, m)	MAXALIGN(offsetof(HnswNeighborTupleData, indextids) + ((level) + 2) * (m) * sizeof(ItemPointerData))
 
 #define HnswPageGetOpaque(page)	((HnswPageOpaque) PageGetSpecialPointer(page))
@@ -67,8 +73,10 @@
 
 #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)
 #else
 #define RandomDouble() (((double) random()) / MAX_RANDOM_VALUE)
+#define SeedRandom(seed) srandom(seed)
 #endif
 
 #if PG_VERSION_NUM < 130000
@@ -86,31 +94,64 @@
 #define HnswGetMl(m) (1 / log(m))
 
 /* Ensure fits on page and in uint8 */
-#define HnswGetMaxLevel(m) Min(((BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - offsetof(HnswNeighborTupleData, indextids) - sizeof(ItemIdData)) / (sizeof(ItemPointerData)) / m) - 2, 255)
+#define HnswGetMaxLevel(m) Min(((BLCKSZ - MAXALIGN(SizeOfPageHeaderData) - MAXALIGN(sizeof(HnswPageOpaqueData)) - offsetof(HnswNeighborTupleData, indextids) - sizeof(ItemIdData)) / (sizeof(ItemPointerData)) / (m)) - 2, 255)
+
+#define HnswGetValue(base, element) PointerGetDatum(HnswPtrAccess(base, (element)->value))
+
+#if PG_VERSION_NUM < 140005
+#define relptr_offset(rp) ((rp).relptr_off - 1)
+#endif
+
+/* Pointer macros */
+#define HnswPtrAccess(base, hp) ((base) == NULL ? (hp).ptr : relptr_access(base, (hp).relptr))
+#define HnswPtrStore(base, hp, value) ((base) == NULL ? (void) ((hp).ptr = (value)) : (void) relptr_store(base, (hp).relptr, value))
+#define HnswPtrIsNull(base, hp) ((base) == NULL ? (hp).ptr == NULL : relptr_is_null((hp).relptr))
+#define HnswPtrEqual(base, hp1, hp2) ((base) == NULL ? (hp1).ptr == (hp2).ptr : relptr_offset((hp1).relptr) == relptr_offset((hp2).relptr))
+
+/* For code paths dedicated to each type */
+#define HnswPtrPointer(hp) (hp).ptr
+#define HnswPtrOffset(hp) relptr_offset((hp).relptr)
 
 /* Variables */
 extern int	hnsw_ef_search;
+extern int	hnsw_lock_tranche_id;
 
+typedef struct HnswElementData HnswElementData;
 typedef struct HnswNeighborArray HnswNeighborArray;
 
+#define HnswPtrDeclare(type, relptrtype, ptrtype) \
+	relptr_declare(type, relptrtype); \
+	typedef union { type *ptr; relptrtype relptr; } ptrtype;
+
+/* Pointers that can be absolute or relative */
+/* Use char for DatumPtr so works with Pointer */
+HnswPtrDeclare(HnswElementData, HnswElementRelptr, HnswElementPtr);
+HnswPtrDeclare(HnswNeighborArray, HnswNeighborArrayRelptr, HnswNeighborArrayPtr);
+HnswPtrDeclare(HnswNeighborArrayPtr, HnswNeighborsRelptr, HnswNeighborsPtr);
+HnswPtrDeclare(char, DatumRelptr, DatumPtr);
+
 typedef struct HnswElementData
 {
-	List	   *heaptids;
+	HnswElementPtr next;
+	ItemPointerData heaptids[HNSW_HEAPTIDS];
+	uint8		heaptidsLength;
 	uint8		level;
 	uint8		deleted;
-	HnswNeighborArray *neighbors;
+	uint32		hash;
+	HnswNeighborsPtr neighbors;
 	BlockNumber blkno;
 	OffsetNumber offno;
 	OffsetNumber neighborOffno;
 	BlockNumber neighborPage;
-	Vector	   *vec;
+	DatumPtr	value;
+	LWLock		lock;
 }			HnswElementData;
 
 typedef HnswElementData * HnswElement;
 
 typedef struct HnswCandidate
 {
-	HnswElement element;
+	HnswElementPtr element;
 	float		distance;
 	bool		closer;
 }			HnswCandidate;
@@ -119,7 +160,7 @@ typedef struct HnswNeighborArray
 {
 	int			length;
 	bool		closerSet;
-	HnswCandidate *items;
+	HnswCandidate items[FLEXIBLE_ARRAY_MEMBER];
 }			HnswNeighborArray;
 
 typedef struct HnswPairingHeapNode
@@ -136,6 +177,70 @@ typedef struct HnswOptions
 	int			efConstruction; /* size of dynamic candidate list */
 }			HnswOptions;
 
+typedef struct HnswGraph
+{
+	/* Graph state */
+	slock_t		lock;
+	HnswElementPtr head;
+	double		indtuples;
+
+	/* Entry state */
+	LWLock		entryLock;
+	HnswElementPtr entryPoint;
+
+	/* Allocations state */
+	LWLock		allocatorLock;
+	long		memoryUsed;
+	long		memoryTotal;
+
+	/* Flushed state */
+	LWLock		flushLock;
+	bool		flushed;
+}			HnswGraph;
+
+typedef struct HnswShared
+{
+	/* Immutable state */
+	Oid			heaprelid;
+	Oid			indexrelid;
+	bool		isconcurrent;
+
+	/* Worker progress */
+	ConditionVariable workersdonecv;
+
+	/* Mutex for mutable state */
+	slock_t		mutex;
+
+	/* Mutable state */
+	int			nparticipantsdone;
+	double		reltuples;
+	HnswGraph	graphData;
+
+#if PG_VERSION_NUM < 120000
+	ParallelHeapScanDescData heapdesc;	/* must come last */
+#endif
+}			HnswShared;
+
+#if PG_VERSION_NUM >= 120000
+#define ParallelTableScanFromHnswShared(shared) \
+	(ParallelTableScanDesc) ((char *) (shared) + BUFFERALIGN(sizeof(HnswShared)))
+#endif
+
+typedef struct HnswLeader
+{
+	ParallelContext *pcxt;
+	int			nparticipanttuplesorts;
+	HnswShared *hnswshared;
+	Snapshot	snapshot;
+	char	   *hnswarea;
+}			HnswLeader;
+
+typedef struct HnswAllocator
+{
+	void	   *(*alloc) (Size size, void *state);
+	void	   *state;
+}			HnswAllocator;
+
 typedef struct HnswBuildState
 {
 	/* Info */
@@ -159,16 +264,21 @@ typedef struct HnswBuildState
 	Oid			collation;
 
 	/* Variables */
-	List	   *elements;
+	HnswGraph	graphData;
-	HnswElement entryPoint;
+	HnswGraph  *graph;
 	double		ml;
 	int			maxLevel;
-	double		maxInMemoryElements;
-	bool		flushed;
 	Vector	   *normvec;
 
 	/* Memory */
+	MemoryContext graphCtx;
 	MemoryContext tmpCtx;
+	HnswAllocator allocator;
+
+	/* Parallel builds */
+	HnswLeader *hnswleader;
+	HnswShared *hnswshared;
+	char	   *hnswarea;
 }			HnswBuildState;
 
 typedef struct HnswMetaPageData
@@ -204,7 +314,7 @@ typedef struct HnswElementTupleData
 	ItemPointerData heaptids[HNSW_HEAPTIDS];
 	ItemPointerData neighbortid;
 	uint16		unused2;
-	Vector		vec;
+	Vector		data;
 }			HnswElementTupleData;
 
 typedef HnswElementTupleData * HnswElementTuple;
@@ -250,7 +360,7 @@ typedef struct HnswVacuumState
 	Oid			collation;
 
 	/* Variables */
-	HTAB	   *deleted;
+	struct tidhash_hash *deleted;
 	BufferAccessStrategy bas;
 	HnswNeighborTuple ntup;
 	HnswElementData highestPoint;
@@ -264,31 +374,29 @@ int			HnswGetM(Relation index);
 int			HnswGetEfConstruction(Relation index);
 FmgrInfo   *HnswOptionalProcInfo(Relation index, uint16 procnum);
 bool		HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result);
-void		HnswCommitBuffer(Buffer buf, GenericXLogState *state);
 Buffer		HnswNewBuffer(Relation index, ForkNumber forkNum);
 void		HnswInitPage(Buffer buf, Page page);
-void		HnswInitRegisterPage(Relation index, Buffer *buf, Page *page, GenericXLogState **state);
 void		HnswInit(void);
-List	   *HnswSearchLayer(Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
+List	   *HnswSearchLayer(char *base, Datum q, List *ep, int ef, int lc, Relation index, FmgrInfo *procinfo, Oid collation, int m, bool inserting, HnswElement skipElement);
 HnswElement HnswGetEntryPoint(Relation index);
 void		HnswGetMetaPageInfo(Relation index, int *m, HnswElement * entryPoint);
-HnswElement HnswInitElement(ItemPointer tid, int m, double ml, int maxLevel);
+void	   *HnswAlloc(HnswAllocator * allocator, Size size);
-void		HnswFreeElement(HnswElement element);
+HnswElement HnswInitElement(char *base, ItemPointer tid, int m, double ml, int maxLevel, HnswAllocator * alloc);
 HnswElement HnswInitElementFromBlock(BlockNumber blkno, OffsetNumber offno);
-void		HnswInsertElement(HnswElement element, HnswElement entryPoint, Relation index, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
+void		HnswFindElementNeighbors(char *base, HnswElement element, HnswElement entryPoint, Relation index, FmgrInfo *procinfo, Oid collation, int m, int efConstruction, bool existing);
-HnswElement HnswFindDuplicate(HnswElement e);
+HnswCandidate *HnswEntryCandidate(char *base, HnswElement em, Datum q, Relation rel, FmgrInfo *procinfo, Oid collation, bool loadVec);
-HnswCandidate *HnswEntryCandidate(HnswElement em, Datum q, Relation rel, FmgrInfo *procinfo, Oid collation, bool loadVec);
+void		HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum, bool building);
-void		HnswUpdateMetaPage(Relation index, int updateEntry, HnswElement entryPoint, BlockNumber insertPage, ForkNumber forkNum);
+void		HnswSetNeighborTuple(char *base, HnswNeighborTuple ntup, HnswElement e, int m);
-void		HnswSetNeighborTuple(HnswNeighborTuple ntup, HnswElement e, int m);
 void		HnswAddHeapTid(HnswElement element, ItemPointer heaptid);
-void		HnswInitNeighbors(HnswElement element, int m);
+void		HnswInitNeighbors(char *base, HnswElement element, int m, HnswAllocator * alloc);
-bool		HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel);
+bool		HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel, bool building);
-void		HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting);
+void		HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building);
 void		HnswLoadElementFromTuple(HnswElement element, HnswElementTuple etup, bool loadHeaptids, bool loadVec);
 void		HnswLoadElement(HnswElement element, float *distance, Datum *q, Relation index, FmgrInfo *procinfo, Oid collation, bool loadVec);
-void		HnswSetElementTuple(HnswElementTuple etup, HnswElement element);
+void		HnswSetElementTuple(char *base, HnswElementTuple etup, HnswElement element);
-void		HnswUpdateConnection(HnswElement element, HnswCandidate * hc, int m, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
+void		HnswUpdateConnection(char *base, HnswElement element, HnswCandidate * hc, int m, int lc, int *updateIdx, Relation index, FmgrInfo *procinfo, Oid collation);
 void		HnswLoadNeighbors(HnswElement element, Relation index, int m);
+PGDLLEXPORT void HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc);
 
 /* Index access methods */
 IndexBuildResult *hnswbuild(Relation heap, Relation index, IndexInfo *indexInfo);
@@ -306,4 +414,54 @@ void		hnswrescan(IndexScanDesc scan, ScanKey keys, int nkeys, ScanKey orderbys,
 bool		hnswgettuple(IndexScanDesc scan, ScanDirection dir);
 void		hnswendscan(IndexScanDesc scan);
 
+static inline HnswNeighborArray *
+HnswGetNeighbors(char *base, HnswElement element, int lc)
+{
+	HnswNeighborArrayPtr *neighborList = HnswPtrAccess(base, element->neighbors);
+
+	Assert(element->level >= lc);
+
+	return HnswPtrAccess(base, neighborList[lc]);
+}
+
+/* Hash tables */
+typedef struct TidHashEntry
+{
+	ItemPointerData tid;
+	char		status;
+}			TidHashEntry;
+
+#define SH_PREFIX tidhash
+#define SH_ELEMENT_TYPE TidHashEntry
+#define SH_KEY_TYPE ItemPointerData
+#define SH_SCOPE extern
+#define SH_DECLARE
+#include "lib/simplehash.h"
+
+typedef struct PointerHashEntry
+{
+	uintptr_t	ptr;
+	char		status;
+}			PointerHashEntry;
+
+#define SH_PREFIX pointerhash
+#define SH_ELEMENT_TYPE PointerHashEntry
+#define SH_KEY_TYPE uintptr_t
+#define SH_SCOPE extern
+#define SH_DECLARE
+#include "lib/simplehash.h"
+
+typedef struct OffsetHashEntry
+{
+	Size		offset;
+	char		status;
+}			OffsetHashEntry;
+
+#define SH_PREFIX offsethash
+#define SH_ELEMENT_TYPE OffsetHashEntry
+#define SH_KEY_TYPE Size
+#define SH_SCOPE extern
+#define SH_DECLARE
+#include "lib/simplehash.h"
+
 #endif

src/hnswinsert.c

@@ -5,6 +5,7 @@
 #include "hnsw.h"
 #include "storage/bufmgr.h"
 #include "storage/lmgr.h"
+#include "utils/datum.h"
 #include "utils/memutils.h"
 
 /*
@@ -92,7 +93,7 @@ HnswFreeOffset(Relation index, Buffer buf, Page page, HnswElement element, Size
  * Add a new page
  */
 static void
-HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState *state, Page page)
+HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState *state, Page page, bool building)
 {
 	/* Add a new page */
 	LockRelationForExtension(index, ExclusiveLock);
@@ -100,7 +101,11 @@ HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState
 	UnlockRelationForExtension(index, ExclusiveLock);
 
 	/* Init new page */
+	if (building)
+		*npage = BufferGetPage(*nbuf);
+	else
 		*npage = GenericXLogRegisterBuffer(state, *nbuf, GENERIC_XLOG_FULL_IMAGE);
+
 	HnswInitPage(*nbuf, *npage);
 
 	/* Update previous buffer */
@@ -111,7 +116,7 @@ HnswInsertAppendPage(Relation index, Buffer *nbuf, Page *npage, GenericXLogState
  * Add to element and neighbor pages
  */
 static void
-WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPage, BlockNumber *updatedInsertPage)
+AddElementOnDisk(Relation index, HnswElement e, int m, BlockNumber insertPage, BlockNumber *updatedInsertPage, bool building)
 {
 	Buffer		buf;
 	Page		page;
@@ -123,16 +128,16 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 	Size		minCombinedSize;
 	HnswElementTuple etup;
 	BlockNumber currentPage = insertPage;
-	int			dimensions = e->vec->dim;
 	HnswNeighborTuple ntup;
 	Buffer		nbuf;
 	Page		npage;
 	OffsetNumber freeOffno = InvalidOffsetNumber;
 	OffsetNumber freeNeighborOffno = InvalidOffsetNumber;
 	BlockNumber newInsertPage = InvalidBlockNumber;
+	char	   *base = NULL;
 
 	/* Calculate sizes */
-	etupSize = HNSW_ELEMENT_TUPLE_SIZE(dimensions);
+	etupSize = HNSW_ELEMENT_TUPLE_SIZE(VARSIZE_ANY(HnswPtrAccess(base, e->value)));
 	ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(e->level, m);
 	combinedSize = etupSize + ntupSize + sizeof(ItemIdData);
 	maxSize = HNSW_MAX_SIZE;
@@ -140,11 +145,11 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 
 	/* Prepare element tuple */
 	etup = palloc0(etupSize);
-	HnswSetElementTuple(etup, e);
+	HnswSetElementTuple(base, etup, e);
 
 	/* Prepare neighbor tuple */
 	ntup = palloc0(ntupSize);
-	HnswSetNeighborTuple(ntup, e, m);
+	HnswSetNeighborTuple(base, ntup, e, m);
 
 	/* Find a page (or two if needed) to insert the tuples */
 	for (;;)
@@ -152,8 +157,16 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 		buf = ReadBuffer(index, currentPage);
 		LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
 
+		if (building)
+		{
+			state = NULL;
+			page = BufferGetPage(buf);
+		}
+		else
+		{
 			state = GenericXLogStart(index);
 			page = GenericXLogRegisterBuffer(state, buf, 0);
+		}
 
 		/* Keep track of first page where element at level 0 can fit */
 		if (!BlockNumberIsValid(newInsertPage) && PageGetFreeSpace(page) >= minCombinedSize)
@@ -173,7 +186,12 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 		if (HnswFreeOffset(index, buf, page, e, ntupSize, &nbuf, &npage, &freeOffno, &freeNeighborOffno, &newInsertPage))
 		{
 			if (nbuf != buf)
+			{
+				if (building)
+					npage = BufferGetPage(nbuf);
+				else
 					npage = GenericXLogRegisterBuffer(state, nbuf, 0);
+			}
 
 			break;
 		}
@@ -182,7 +200,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 		/* Skip if both tuples can fit on the same page */
 		if (combinedSize > maxSize && PageGetFreeSpace(page) >= etupSize && !BlockNumberIsValid(HnswPageGetOpaque(page)->nextblkno))
 		{
-			HnswInsertAppendPage(index, &nbuf, &npage, state, page);
+			HnswInsertAppendPage(index, &nbuf, &npage, state, page, building);
 			break;
 		}
 
@@ -191,6 +209,7 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 		if (BlockNumberIsValid(currentPage))
 		{
 			/* Move to next page */
+			if (!building)
 				GenericXLogAbort(state);
 			UnlockReleaseBuffer(buf);
 		}
@@ -199,22 +218,33 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 			Buffer		newbuf;
 			Page		newpage;
 
-			HnswInsertAppendPage(index, &newbuf, &newpage, state, page);
+			HnswInsertAppendPage(index, &newbuf, &newpage, state, page, building);
 
 			/* Commit */
+			if (building)
+				MarkBufferDirty(buf);
+			else
 				GenericXLogFinish(state);
 
 			/* Unlock previous buffer */
 			UnlockReleaseBuffer(buf);
 
 			/* Prepare new buffer */
-			state = GenericXLogStart(index);
 			buf = newbuf;
+			if (building)
+			{
+				state = NULL;
+				page = BufferGetPage(buf);
+			}
+			else
+			{
+				state = GenericXLogStart(index);
 				page = GenericXLogRegisterBuffer(state, buf, 0);
+			}
 
 			/* Create new page for neighbors if needed */
 			if (PageGetFreeSpace(page) < combinedSize)
-				HnswInsertAppendPage(index, &nbuf, &npage, state, page);
+				HnswInsertAppendPage(index, &nbuf, &npage, state, page, building);
 			else
 			{
 				nbuf = buf;
@@ -268,6 +298,13 @@ WriteNewElementPages(Relation index, HnswElement e, int m, BlockNumber insertPag
 	}
 
 	/* Commit */
+	if (building)
+	{
+		MarkBufferDirty(buf);
+		if (nbuf != buf)
+			MarkBufferDirty(nbuf);
+	}
+	else
 		GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 	if (nbuf != buf)
@@ -302,12 +339,14 @@ ConnectionExists(HnswElement e, HnswNeighborTuple ntup, int startIdx, int lm)
  * Update neighbors
  */
 void
-HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting)
+HnswUpdateNeighborsOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement e, int m, bool checkExisting, bool building)
 {
+	char	   *base = NULL;
+
 	for (int lc = e->level; lc >= 0; lc--)
 	{
 		int			lm = HnswGetLayerM(m, lc);
-		HnswNeighborArray *neighbors = &e->neighbors[lc];
+		HnswNeighborArray *neighbors = HnswGetNeighbors(base, e, lc);
 
 		for (int i = 0; i < neighbors->length; i++)
 		{
@@ -320,11 +359,12 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
 			Size		ntupSize;
 			int			idx = -1;
 			int			startIdx;
-			OffsetNumber offno = hc->element->neighborOffno;
+			HnswElement neighborElement = HnswPtrAccess(base, hc->element);
+			OffsetNumber offno = neighborElement->neighborOffno;
 
 			/* Get latest neighbors since they may have changed */
 			/* Do not lock yet since selecting neighbors can take time */
-			HnswLoadNeighbors(hc->element, index, m);
+			HnswLoadNeighbors(neighborElement, index, m);
 
 			/*
 			 * Could improve performance for vacuuming by checking neighbors
@@ -334,17 +374,25 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
 			 */
 
 			/* Select neighbors */
-			HnswUpdateConnection(e, hc, lm, lc, &idx, index, procinfo, collation);
+			HnswUpdateConnection(NULL, e, hc, lm, lc, &idx, index, procinfo, collation);
 
 			/* New element was not selected as a neighbor */
 			if (idx == -1)
 				continue;
 
 			/* Register page */
-			buf = ReadBuffer(index, hc->element->neighborPage);
+			buf = ReadBuffer(index, neighborElement->neighborPage);
 			LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
+			if (building)
+			{
+				state = NULL;
+				page = BufferGetPage(buf);
+			}
+			else
+			{
 				state = GenericXLogStart(index);
 				page = GenericXLogRegisterBuffer(state, buf, 0);
+			}
 
 			/* Get tuple */
 			itemid = PageGetItemId(page, offno);
@@ -352,7 +400,7 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
 			ntupSize = ItemIdGetLength(itemid);
 
 			/* Calculate index for update */
-			startIdx = (hc->element->level - lc) * m;
+			startIdx = (neighborElement->level - lc) * m;
 
 			/* Check for existing connection */
 			if (checkExisting && ConnectionExists(e, ntup, startIdx, lm))
@@ -386,9 +434,12 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
 					elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
 				/* Commit */
+				if (building)
+					MarkBufferDirty(buf);
+				else
 					GenericXLogFinish(state);
 			}
-			else
+			else if (!building)
 				GenericXLogAbort(state);
 
 			UnlockReleaseBuffer(buf);
@@ -400,23 +451,34 @@ HnswUpdateNeighborPages(Relation index, FmgrInfo *procinfo, Oid collation, HnswE
  * Add a heap TID to an existing element
  */
 static bool
-HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup)
+AddDuplicateOnDisk(Relation index, HnswElement element, HnswElement dup, bool building)
 {
 	Buffer		buf;
 	Page		page;
 	GenericXLogState *state;
-	Size		etupSize = HNSW_ELEMENT_TUPLE_SIZE(dup->vec->dim);
+	ItemId		itemid;
 	HnswElementTuple etup;
+	Size		etupSize;
 	int			i;
 
 	/* Read page */
 	buf = ReadBuffer(index, dup->blkno);
 	LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
+	if (building)
+	{
+		state = NULL;
+		page = BufferGetPage(buf);
+	}
+	else
+	{
 		state = GenericXLogStart(index);
 		page = GenericXLogRegisterBuffer(state, buf, 0);
+	}
 
 	/* Find space */
-	etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, dup->offno));
+	itemid = PageGetItemId(page, dup->offno);
+	etup = (HnswElementTuple) PageGetItem(page, itemid);
+	etupSize = ItemIdGetLength(itemid);
 	for (i = 0; i < HNSW_HEAPTIDS; i++)
 	{
 		if (!ItemPointerIsValid(&etup->heaptids[i]))
@@ -426,19 +488,23 @@ HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup)
 	/* Either being deleted or we lost our chance to another backend */
 	if (i == 0 || i == HNSW_HEAPTIDS)
 	{
+		if (!building)
 			GenericXLogAbort(state);
 		UnlockReleaseBuffer(buf);
 		return false;
 	}
 
 	/* Add heap TID */
-	etup->heaptids[i] = *((ItemPointer) linitial(element->heaptids));
+	etup->heaptids[i] = element->heaptids[0];
 
 	/* Overwrite tuple */
 	if (!PageIndexTupleOverwrite(page, dup->offno, (Item) etup, etupSize))
 		elog(ERROR, "failed to add index item to \"%s\"", RelationGetRelationName(index));
 
 	/* Commit */
+	if (building)
+		MarkBufferDirty(buf);
+	else
 		GenericXLogFinish(state);
 	UnlockReleaseBuffer(buf);
 
@@ -446,62 +512,73 @@ HnswAddDuplicate(Relation index, HnswElement element, HnswElement dup)
 }
 
 /*
- * Write changes to disk
+ * Find duplicate element
+ */
+static bool
+FindDuplicateOnDisk(Relation index, HnswElement element, bool building)
+{
+	char	   *base = NULL;
+	HnswNeighborArray *neighbors = HnswGetNeighbors(base, element, 0);
+	Datum		value = HnswGetValue(base, element);
+
+	for (int i = 0; i < neighbors->length; i++)
+	{
+		HnswCandidate *neighbor = &neighbors->items[i];
+		HnswElement neighborElement = HnswPtrAccess(base, neighbor->element);
+		Datum		neighborValue = HnswGetValue(base, neighborElement);
+
+		/* Exit early since ordered by distance */
+		if (!datumIsEqual(value, neighborValue, false, -1))
+			return false;
+
+		if (AddDuplicateOnDisk(index, element, neighborElement, building))
+			return true;
+	}
+
+	return false;
+}
+
+/*
+ * Update graph on disk
  */
 static void
-WriteElement(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement dup, HnswElement entryPoint)
+UpdateGraphOnDisk(Relation index, FmgrInfo *procinfo, Oid collation, HnswElement element, int m, int efConstruction, HnswElement entryPoint, bool building)
 {
 	BlockNumber newInsertPage = InvalidBlockNumber;
 
-	/* Try to add to existing page */
+	/* Look for duplicate */
-	if (dup != NULL)
+	if (FindDuplicateOnDisk(index, element, building))
-	{
-		if (HnswAddDuplicate(index, element, dup))
 		return;
-	}
 
-	/* Write element and neighbor tuples */
+	/* Add element */
-	WriteNewElementPages(index, element, m, GetInsertPage(index), &newInsertPage);
+	AddElementOnDisk(index, element, m, GetInsertPage(index), &newInsertPage, building);
 
 	/* Update insert page if needed */
 	if (BlockNumberIsValid(newInsertPage))
-		HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM);
+		HnswUpdateMetaPage(index, 0, NULL, newInsertPage, MAIN_FORKNUM, building);
 
 	/* Update neighbors */
-	HnswUpdateNeighborPages(index, procinfo, collation, element, m, false);
+	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, false, building);
 
-	/* Update metapage if needed */
+	/* Update entry point if needed */
 	if (entryPoint == NULL || element->level > entryPoint->level)
-		HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM);
+		HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM, building);
 }
 
 /*
  * Insert a tuple into the index
  */
 bool
-HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
+HnswInsertTupleOnDisk(Relation index, Datum value, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel, bool building)
 {
-	Datum		value;
-	FmgrInfo   *normprocinfo;
 	HnswElement entryPoint;
 	HnswElement element;
 	int			m;
 	int			efConstruction = HnswGetEfConstruction(index);
 	FmgrInfo   *procinfo = index_getprocinfo(index, 1, HNSW_DISTANCE_PROC);
 	Oid			collation = index->rd_indcollation[0];
-	HnswElement dup;
 	LOCKMODE	lockmode = ShareLock;
-
+	char	   *base = NULL;
-	/* Detoast once for all calls */
-	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
-
-	/* Normalize if needed */
-	normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
-	if (normprocinfo != NULL)
-	{
-		if (!HnswNormValue(normprocinfo, collation, &value, NULL))
-			return false;
-	}
 
 	/*
 	 * Get a shared lock. This allows vacuum to ensure no in-flight inserts
@@ -514,8 +591,8 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
 	HnswGetMetaPageInfo(index, &m, &entryPoint);
 
 	/* Create an element */
-	element = HnswInitElement(heap_tid, m, HnswGetMl(m), HnswGetMaxLevel(m));
+	element = HnswInitElement(base, heap_tid, m, HnswGetMl(m), HnswGetMaxLevel(m), NULL);
-	element->vec = DatumGetVector(value);
+	HnswPtrStore(base, element->value, DatumGetPointer(value));
 
 	/* Prevent concurrent inserts when likely updating entry point */
 	if (entryPoint == NULL || element->level > entryPoint->level)
@@ -531,14 +608,11 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
 		entryPoint = HnswGetEntryPoint(index);
 	}
 
-	/* Insert element in graph */
+	/* Find neighbors for element */
-	HnswInsertElement(element, entryPoint, index, procinfo, collation, m, efConstruction, false);
+	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, false);
 
-	/* Look for duplicate */
+	/* Update graph on disk */
-	dup = HnswFindDuplicate(element);
+	UpdateGraphOnDisk(index, procinfo, collation, element, m, efConstruction, entryPoint, building);
-
-	/* Write to disk */
-	WriteElement(index, procinfo, collation, element, m, efConstruction, dup, entryPoint);
 
 	/* Release lock */
 	UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
@@ -546,6 +620,30 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
 	return true;
 }
 
+/*
+ * Insert a tuple into the index
+ */
+static void
+HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, Relation heapRel)
+{
+	Datum		value;
+	FmgrInfo   *normprocinfo;
+	Oid			collation = index->rd_indcollation[0];
+
+	/* Detoast once for all calls */
+	value = PointerGetDatum(PG_DETOAST_DATUM(values[0]));
+
+	/* Normalize if needed */
+	normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
+	if (normprocinfo != NULL)
+	{
+		if (!HnswNormValue(normprocinfo, collation, &value, NULL))
+			return;
+	}
+
+	HnswInsertTupleOnDisk(index, value, values, isnull, heap_tid, heapRel, false);
+}
+
 /*
  * Insert a tuple into the index
  */

src/hnswscan.c

@@ -21,6 +21,7 @@ GetScanItems(IndexScanDesc scan, Datum q)
 	List	   *w;
 	int			m;
 	HnswElement entryPoint;
+	char	   *base = NULL;
 
 	/* Get m and entry point */
 	HnswGetMetaPageInfo(index, &m, &entryPoint);
@@ -28,15 +29,15 @@ GetScanItems(IndexScanDesc scan, Datum q)
 	if (entryPoint == NULL)
 		return NIL;
 
-	ep = list_make1(HnswEntryCandidate(entryPoint, q, index, procinfo, collation, false));
+	ep = list_make1(HnswEntryCandidate(base, entryPoint, q, index, procinfo, collation, false));
 
 	for (int lc = entryPoint->level; lc >= 1; lc--)
 	{
-		w = HnswSearchLayer(q, ep, 1, lc, index, procinfo, collation, m, false, NULL);
+		w = HnswSearchLayer(base, q, ep, 1, lc, index, procinfo, collation, m, false, NULL);
 		ep = w;
 	}
 
-	return HnswSearchLayer(q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
+	return HnswSearchLayer(base, q, ep, hnsw_ef_search, 0, index, procinfo, collation, m, false, NULL);
 }
 
 /*
@@ -184,19 +185,19 @@ hnswgettuple(IndexScanDesc scan, ScanDirection dir)
 
 	while (list_length(so->w) > 0)
 	{
+		char	   *base = NULL;
 		HnswCandidate *hc = llast(so->w);
+		HnswElement element = HnswPtrAccess(base, hc->element);
 		ItemPointer heaptid;
 
 		/* Move to next element if no valid heap TIDs */
-		if (list_length(hc->element->heaptids) == 0)
+		if (element->heaptidsLength == 0)
 		{
 			so->w = list_delete_last(so->w);
 			continue;
 		}
 
-		heaptid = llast(hc->element->heaptids);
+		heaptid = &element->heaptids[--element->heaptidsLength];
-
-		hc->element->heaptids = list_delete_last(hc->element->heaptids);
 
 		MemoryContextSwitchTo(oldCtx);
 

src/hnswvacuum.c

@@ -12,12 +12,9 @@
  * Check if deleted list contains an index TID
  */
 static bool
-DeletedContains(HTAB *deleted, ItemPointer indextid)
+DeletedContains(tidhash_hash * deleted, ItemPointer indextid)
 {
-	bool		found;
+	return tidhash_lookup(deleted, *indextid) != NULL;
-
-	hash_search(deleted, indextid, HASH_FIND, &found);
-	return found;
 }
 
 /*
@@ -62,7 +59,8 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 		/* Iterate over nodes */
 		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
+			ItemId		itemid = PageGetItemId(page, offno);
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
 			int			idx = 0;
 			bool		itemUpdated = false;
 
@@ -93,7 +91,7 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 
 				if (itemUpdated)
 				{
-					Size		etupSize = HNSW_ELEMENT_TUPLE_SIZE(etup->vec.dim);
+					Size		etupSize = ItemIdGetLength(itemid);
 
 					/* Mark rest as invalid */
 					for (int i = idx; i < HNSW_HEAPTIDS; i++)
@@ -109,11 +107,13 @@ RemoveHeapTids(HnswVacuumState * vacuumstate)
 			if (!ItemPointerIsValid(&etup->heaptids[0]))
 			{
 				ItemPointerData ip;
+				bool		found;
 
 				/* Add to deleted list */
 				ItemPointerSet(&ip, blkno, offno);
 
-				(void) hash_search(vacuumstate->deleted, &ip, HASH_ENTER, NULL);
+				tidhash_insert(vacuumstate->deleted, ip, &found);
+				Assert(!found);
 			}
 			else if (etup->level > highestLevel && !(entryPoint != NULL && blkno == entryPoint->blkno && offno == entryPoint->offno))
 			{
@@ -199,21 +199,22 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	BufferAccessStrategy bas = vacuumstate->bas;
 	HnswNeighborTuple ntup = vacuumstate->ntup;
 	Size		ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(element->level, m);
+	char	   *base = NULL;
 
 	/* Skip if element is entry point */
 	if (entryPoint != NULL && element->blkno == entryPoint->blkno && element->offno == entryPoint->offno)
 		return;
 
 	/* Init fields */
-	HnswInitNeighbors(element, m);
+	HnswInitNeighbors(base, element, m, NULL);
-	element->heaptids = NIL;
+	element->heaptidsLength = 0;
 
-	/* Add element to graph, skipping itself */
+	/* Find neighbors for element, skipping itself */
-	HnswInsertElement(element, entryPoint, index, procinfo, collation, m, efConstruction, true);
+	HnswFindElementNeighbors(base, element, entryPoint, index, procinfo, collation, m, efConstruction, true);
 
 	/* Update neighbor tuple */
 	/* Do this before getting page to minimize locking */
-	HnswSetNeighborTuple(ntup, element, m);
+	HnswSetNeighborTuple(base, ntup, element, m);
 
 	/* Get neighbor page */
 	buf = ReadBufferExtended(index, MAIN_FORKNUM, element->neighborPage, RBM_NORMAL, bas);
@@ -230,7 +231,7 @@ RepairGraphElement(HnswVacuumState * vacuumstate, HnswElement element, HnswEleme
 	UnlockReleaseBuffer(buf);
 
 	/* Update neighbors */
-	HnswUpdateNeighborPages(index, procinfo, collation, element, m, true);
+	HnswUpdateNeighborsOnDisk(index, procinfo, collation, element, m, true, false);
 }
 
 /*
@@ -286,7 +287,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 			 * point is outdated and empty, the entry point will be empty
 			 * until an element is repaired.
 			 */
-			HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_ALWAYS, highestPoint, InvalidBlockNumber, MAIN_FORKNUM);
+			HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_ALWAYS, highestPoint, InvalidBlockNumber, MAIN_FORKNUM, false);
 		}
 		else
 		{
@@ -301,7 +302,7 @@ RepairGraphEntryPoint(HnswVacuumState * vacuumstate)
 			{
 				/* Reset neighbors from previous update */
 				if (highestPoint != NULL)
-					highestPoint->neighbors = NULL;
+					HnswPtrStore((char *) NULL, highestPoint->neighbors, (HnswNeighborArrayPtr *) NULL);
 
 				RepairGraphElement(vacuumstate, entryPoint, highestPoint);
 			}
@@ -419,7 +420,7 @@ RepairGraph(HnswVacuumState * vacuumstate)
 			 * was replaced and highest point was outdated.
 			 */
 			if (entryPoint == NULL || element->level > entryPoint->level)
-				HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM);
+				HnswUpdateMetaPage(index, HNSW_UPDATE_ENTRY_GREATER, element, InvalidBlockNumber, MAIN_FORKNUM, false);
 
 			/* Release lock */
 			UnlockPage(index, HNSW_UPDATE_LOCK, lockmode);
@@ -477,7 +478,8 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 		/* Update element and neighbors together */
 		for (offno = FirstOffsetNumber; offno <= maxoffno; offno = OffsetNumberNext(offno))
 		{
-			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, PageGetItemId(page, offno));
+			ItemId		itemid = PageGetItemId(page, offno);
+			HnswElementTuple etup = (HnswElementTuple) PageGetItem(page, itemid);
 			HnswNeighborTuple ntup;
 			Size		etupSize;
 			Size		ntupSize;
@@ -505,7 +507,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 				continue;
 
 			/* Calculate sizes */
-			etupSize = HNSW_ELEMENT_TUPLE_SIZE(etup->vec.dim);
+			etupSize = ItemIdGetLength(itemid);
 			ntupSize = HNSW_NEIGHBOR_TUPLE_SIZE(etup->level, vacuumstate->m);
 
 			/* Get neighbor page */
@@ -528,7 +530,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 
 			/* Overwrite element */
 			etup->deleted = 1;
-			MemSet(&etup->vec.x, 0, etup->vec.dim * sizeof(float));
+			MemSet(&etup->data, 0, VARSIZE_ANY(&etup->data));
 
 			/* Overwrite neighbors */
 			for (int i = 0; i < ntup->count; i++)
@@ -563,7 +565,7 @@ MarkDeleted(HnswVacuumState * vacuumstate)
 	}
 
 	/* Update insert page last, after everything has been marked as deleted */
-	HnswUpdateMetaPage(index, 0, NULL, insertPage, MAIN_FORKNUM);
+	HnswUpdateMetaPage(index, 0, NULL, insertPage, MAIN_FORKNUM, false);
 }
 
 /*
@@ -573,7 +575,6 @@ static void
 InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkDeleteResult *stats, IndexBulkDeleteCallback callback, void *callback_state)
 {
 	Relation	index = info->index;
-	HASHCTL		hash_ctl;
 
 	if (stats == NULL)
 		stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
@@ -595,10 +596,7 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
 	HnswGetMetaPageInfo(index, &vacuumstate->m, NULL);
 
 	/* Create hash table */
-	hash_ctl.keysize = sizeof(ItemPointerData);
+	vacuumstate->deleted = tidhash_create(CurrentMemoryContext, 256, NULL);
-	hash_ctl.entrysize = sizeof(ItemPointerData);
-	hash_ctl.hcxt = CurrentMemoryContext;
-	vacuumstate->deleted = hash_create("hnswbulkdelete indextids", 256, &hash_ctl, HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
 }
 
 /*
@@ -607,7 +605,7 @@ InitVacuumState(HnswVacuumState * vacuumstate, IndexVacuumInfo *info, IndexBulkD
 static void
 FreeVacuumState(HnswVacuumState * vacuumstate)
 {
-	hash_destroy(vacuumstate->deleted);
+	tidhash_destroy(vacuumstate->deleted);
 	FreeAccessStrategy(vacuumstate->bas);
 	pfree(vacuumstate->ntup);
 	MemoryContextDelete(vacuumstate->tmpCtx);

src/ivfbuild.c

@@ -543,10 +543,10 @@ CreateListPages(Relation index, VectorArray centers, int dimensions,
 	pfree(list);
 }
 
+#ifdef IVFFLAT_KMEANS_DEBUG
 /*
  * Print k-means metrics
  */
-#ifdef IVFFLAT_KMEANS_DEBUG
 static void
 PrintKmeansMetrics(IvfflatBuildState * buildstate)
 {

src/ivfkmeans.c

@@ -6,6 +6,10 @@
 #include "ivfflat.h"
 #include "miscadmin.h"
 
+#ifdef IVFFLAT_MEMORY
+#include "utils/memutils.h"
+#endif
+
 /*
  * Initialize with kmeans++
  *
@@ -151,6 +155,23 @@ QuickCenters(Relation index, VectorArray samples, VectorArray centers)
 	}
 }
 
+#ifdef IVFFLAT_MEMORY
+/*
+ * Show memory usage
+ */
+static void
+ShowMemoryUsage(Size estimatedSize)
+{
+#if PG_VERSION_NUM >= 130000
+	elog(INFO, "total memory: %zu MB",
+		 MemoryContextMemAllocated(CurrentMemoryContext, true) / (1024 * 1024));
+#else
+	MemoryContextStats(CurrentMemoryContext);
+#endif
+	elog(INFO, "estimated memory: %zu MB", estimatedSize / (1024 * 1024));
+}
+#endif
+
 /*
  * Use Elkan for performance. This requires distance function to satisfy triangle inequality.
  *
@@ -231,6 +252,10 @@ ElkanKmeans(Relation index, VectorArray samples, VectorArray centers)
 		vec->dim = dimensions;
 	}
 
+#ifdef IVFFLAT_MEMORY
+	ShowMemoryUsage(totalSize);
+#endif
+
 	/* Pick initial centers */
 	InitCenters(index, samples, centers, lowerBound);
 

src/svector.c

@@ -1,705 +0,0 @@
-#include "postgres.h"
-
-#include <math.h>
-
-#include "fmgr.h"
-#include "libpq/pqformat.h"
-#include "svector.h"
-#include "utils/array.h"
-#include "vector.h"
-
-#if PG_VERSION_NUM >= 120000
-#include "common/shortest_dec.h"
-#include "utils/float.h"
-#else
-#include <float.h>
-#include "utils/builtins.h"
-#endif
-
-/*
- * Ensure same dimensions
- */
-static inline void
-CheckDims(SVector * a, SVector * b)
-{
-	if (a->dim != b->dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("different svector dimensions %d and %d", a->dim, b->dim)));
-}
-
-/*
- * Ensure expected dimensions
- */
-static inline void
-CheckExpectedDim(int32 typmod, int dim)
-{
-	if (typmod != -1 && typmod != dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("expected %d dimensions, not %d", typmod, dim)));
-}
-
-/*
- * Ensure valid dimensions
- */
-static inline void
-CheckDim(int dim)
-{
-	if (dim < 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("svector must have at least 1 dimension")));
-
-	if (dim > SVECTOR_MAX_DIM)
-		ereport(ERROR,
-				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
-				 errmsg("svector cannot have more than %d dimensions", SVECTOR_MAX_DIM)));
-}
-
-/*
- * Ensure valid nnz
- */
-static inline void
-CheckNnz(int nnz, int dim)
-{
-	if (nnz < 0)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("svector must have at least one element")));
-
-	if (nnz > dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
-				 errmsg("svector cannot have more elements than dimensions")));
-}
-
-/*
- * Ensure valid index
- */
-static inline void
-CheckIndex(int32 *indices, int i, int dim)
-{
-	int32		index = indices[i];
-
-	if (index < 0)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("index must not be negative")));
-
-	if (index >= dim)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("index must be less than dimensions")));
-
-	if (i > 0)
-	{
-		if (index < indices[i - 1])
-			ereport(ERROR,
-					(errcode(ERRCODE_DATA_EXCEPTION),
-					 errmsg("indexes must be in ascending order")));
-
-		if (index == indices[i - 1])
-			ereport(ERROR,
-					(errcode(ERRCODE_DATA_EXCEPTION),
-					 errmsg("indexes must not contain duplicates")));
-	}
-}
-
-/*
- * Ensure finite element
- */
-static inline void
-CheckElement(float value)
-{
-	if (isnan(value))
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("NaN not allowed in svector")));
-
-	if (isinf(value))
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("infinite value not allowed in svector")));
-}
-
-/*
- * Allocate and initialize a new sparse vector
- */
-SVector *
-InitSVector(int dim, int nnz)
-{
-	SVector    *result;
-	int			size;
-
-	size = SVECTOR_SIZE(nnz);
-	result = (SVector *) palloc0(size);
-	SET_VARSIZE(result, size);
-	result->dim = dim;
-	result->nnz = nnz;
-
-	return result;
-}
-
-/*
- * Convert textual representation to internal representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_in);
-Datum
-svector_in(PG_FUNCTION_ARGS)
-{
-	char	   *str = PG_GETARG_CSTRING(0);
-	int32		typmod = PG_GETARG_INT32(2);
-	int			dim;
-	char	   *pt;
-	SVector    *result;
-	float	   *rvalues;
-	char	   *lit = pstrdup(str);
-	int			n;
-	int32	   *indices;
-	float	   *values;
-	int			index;
-	float		value;
-	int			maxNnz;
-	int			nnz = 0;
-
-	/* TODO Improve code and checks after deciding on format */
-
-	maxNnz = 1;
-	pt = str;
-	while (*pt != '\0')
-	{
-		if (*pt == ',')
-			maxNnz++;
-
-		pt++;
-	}
-	maxNnz /= 2;
-
-	indices = palloc(maxNnz * sizeof(int32));
-	values = palloc(maxNnz * sizeof(float));
-
-	while (sscanf(str, "(%d,%f)%n", &index, &value, &n) == 2)
-	{
-		/* TODO Better error */
-		if (nnz == maxNnz)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("ran out of buffer: \"%s\"", lit)));
-
-		/* TODO Decide whether to store zero values */
-		indices[nnz] = index;
-		values[nnz] = value;
-		nnz++;
-
-		str += n;
-
-		if (*str == ',')
-			str++;
-		else if (*str == '|')
-			break;
-		else
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-					 errmsg("malformed svector literal: \"%s\"", lit)));
-	}
-
-	if (sscanf(str, "|%d|%n", &dim, &n) != 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed svector literal: \"%s\"", lit)));
-
-	str += n;
-
-	if (*str != '\0')
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TEXT_REPRESENTATION),
-				 errmsg("malformed svector literal: \"%s\"", lit),
-				 errdetail("Junk after closing pipe.")));
-
-	pfree(lit);
-
-	CheckDim(dim);
-	CheckExpectedDim(typmod, dim);
-
-	result = InitSVector(dim, nnz);
-	rvalues = SVECTOR_VALUES(result);
-	for (int i = 0; i < nnz; i++)
-	{
-		result->indices[i] = indices[i];
-		rvalues[i] = values[i];
-
-		CheckIndex(result->indices, i, dim);
-		CheckElement(rvalues[i]);
-	}
-
-	PG_RETURN_POINTER(result);
-}
-
-/*
- * Convert internal representation to textual representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_out);
-Datum
-svector_out(PG_FUNCTION_ARGS)
-{
-	SVector    *svector = PG_GETARG_SVECTOR_P(0);
-	float	   *values = SVECTOR_VALUES(svector);
-	char	   *buf;
-	char	   *ptr;
-	int			n;
-
-	/* TODO Improve code after deciding on format */
-
-#if PG_VERSION_NUM < 120000
-	int			ndig = FLT_DIG + extra_float_digits;
-
-	if (ndig < 1)
-		ndig = 1;
-
-#define FLOAT_SHORTEST_DECIMAL_LEN (ndig + 10)
-#endif
-
-	/* TODO Move */
-#define APPEND_CHAR(ptr, ch) (*(ptr)++ = (ch))
-
-	/* TODO Improve */
-	buf = (char *) palloc((FLOAT_SHORTEST_DECIMAL_LEN + 20) * svector->nnz + 20);
-	ptr = buf;
-
-	for (int i = 0; i < svector->nnz; i++)
-	{
-		if (i > 0)
-			APPEND_CHAR(ptr, ',');
-
-		n = sprintf(ptr, "(%d,", svector->indices[i]);
-		ptr += n;
-
-#if PG_VERSION_NUM >= 120000
-		n = float_to_shortest_decimal_bufn(values[i], ptr);
-#else
-		n = sprintf(ptr, "%.*g", ndig, values[i]);
-#endif
-		ptr += n;
-
-		APPEND_CHAR(ptr, ')');
-	}
-
-	n = sprintf(ptr, "|%d|", svector->dim);
-	ptr += n;
-
-	APPEND_CHAR(ptr, '\0');
-
-	PG_FREE_IF_COPY(svector, 0);
-	PG_RETURN_CSTRING(buf);
-}
-
-/*
- * Convert type modifier
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_typmod_in);
-Datum
-svector_typmod_in(PG_FUNCTION_ARGS)
-{
-	ArrayType  *ta = PG_GETARG_ARRAYTYPE_P(0);
-	int32	   *tl;
-	int			n;
-
-	tl = ArrayGetIntegerTypmods(ta, &n);
-
-	if (n != 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("invalid type modifier")));
-
-	if (*tl < 1)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("dimensions for type svector must be at least 1")));
-
-	if (*tl > SVECTOR_MAX_DIM)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
-				 errmsg("dimensions for type svector cannot exceed %d", SVECTOR_MAX_DIM)));
-
-	PG_RETURN_INT32(*tl);
-}
-
-/*
- * Convert external binary representation to internal representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_recv);
-Datum
-svector_recv(PG_FUNCTION_ARGS)
-{
-	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
-	int32		typmod = PG_GETARG_INT32(2);
-	SVector    *result;
-	int32		dim;
-	int32		nnz;
-	int32		unused;
-	float	   *values;
-
-	dim = pq_getmsgint(buf, sizeof(int32));
-	nnz = pq_getmsgint(buf, sizeof(int32));
-	unused = pq_getmsgint(buf, sizeof(int32));
-
-	CheckDim(dim);
-	CheckNnz(nnz, dim);
-	CheckExpectedDim(typmod, dim);
-
-	if (unused != 0)
-		ereport(ERROR,
-				(errcode(ERRCODE_DATA_EXCEPTION),
-				 errmsg("expected unused to be 0, not %d", unused)));
-
-	result = InitSVector(dim, nnz);
-	values = SVECTOR_VALUES(result);
-
-	for (int i = 0; i < nnz; i++)
-	{
-		result->indices[i] = pq_getmsgint(buf, sizeof(int32));
-		CheckIndex(result->indices, i, dim);
-	}
-
-	for (int i = 0; i < nnz; i++)
-	{
-		values[i] = pq_getmsgfloat4(buf);
-		CheckElement(values[i]);
-	}
-
-	PG_RETURN_POINTER(result);
-}
-
-/*
- * Convert internal representation to the external binary representation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_send);
-Datum
-svector_send(PG_FUNCTION_ARGS)
-{
-	SVector    *svec = PG_GETARG_SVECTOR_P(0);
-	float	   *values = SVECTOR_VALUES(svec);
-	StringInfoData buf;
-
-	pq_begintypsend(&buf);
-	pq_sendint(&buf, svec->dim, sizeof(int32));
-	pq_sendint(&buf, svec->nnz, sizeof(int32));
-	pq_sendint(&buf, svec->unused, sizeof(int32));
-	for (int i = 0; i < svec->nnz; i++)
-		pq_sendint(&buf, svec->indices[i], sizeof(int32));
-	for (int i = 0; i < svec->nnz; i++)
-		pq_sendfloat4(&buf, values[i]);
-
-	PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
-}
-
-/*
- * Convert sparse vector to sparse vector
- * This is needed to check the type modifier
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector);
-Datum
-svector(PG_FUNCTION_ARGS)
-{
-	SVector    *svec = PG_GETARG_SVECTOR_P(0);
-	int32		typmod = PG_GETARG_INT32(1);
-
-	CheckExpectedDim(typmod, svec->dim);
-
-	PG_RETURN_POINTER(svec);
-}
-
-/*
- * Convert dense vector to sparse vector
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_to_svector);
-Datum
-vector_to_svector(PG_FUNCTION_ARGS)
-{
-	Vector	   *vec = PG_GETARG_VECTOR_P(0);
-	int32		typmod = PG_GETARG_INT32(1);
-	SVector    *result;
-	int			dim = vec->dim;
-	int			nnz = 0;
-	float	   *values;
-	int			j = 0;
-
-	CheckDim(dim);
-	CheckExpectedDim(typmod, dim);
-
-	for (int i = 0; i < dim; i++)
-	{
-		if (vec->x[i] != 0)
-			nnz++;
-	}
-
-	result = InitSVector(dim, nnz);
-	values = SVECTOR_VALUES(result);
-	for (int i = 0; i < dim; i++)
-	{
-		if (vec->x[i] != 0)
-		{
-			/* Safety check */
-			if (j == nnz)
-				elog(ERROR, "safety check failed");
-
-			result->indices[j] = i;
-			values[j] = vec->x[i];
-			j++;
-		}
-	}
-
-	PG_RETURN_POINTER(result);
-}
-
-/*
- *  Get the L2 squared distance between sparse vectors
- */
-static double
-l2_distance_squared_internal(SVector * a, SVector * b)
-{
-	float	   *ax = SVECTOR_VALUES(a);
-	float	   *bx = SVECTOR_VALUES(b);
-	double		distance = 0.0;
-	int			bpos = 0;
-
-	for (int i = 0; i < a->nnz; i++)
-	{
-		int			ai = a->indices[i];
-		int			bi = -1;
-
-		for (int j = bpos; j < b->nnz; j++)
-		{
-			bi = b->indices[j];
-
-			if (ai == bi)
-			{
-				double		diff = ax[i] - bx[j];
-
-				distance += diff * diff;
-			}
-			else if (ai > bi)
-				distance += bx[j] * bx[j];
-
-			/* Update start for next iteration */
-			if (ai >= bi)
-				bpos = j + 1;
-
-			/* Found or passed it */
-			if (bi >= ai)
-				break;
-		}
-
-		if (ai != bi)
-			distance += ax[i] * ax[i];
-	}
-
-	for (int j = bpos; j < b->nnz; j++)
-		distance += bx[j] * bx[j];
-
-	return distance;
-}
-
-/*
- * Get the L2 distance between sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_l2_distance);
-Datum
-svector_l2_distance(PG_FUNCTION_ARGS)
-{
-	SVector    *a = PG_GETARG_SVECTOR_P(0);
-	SVector    *b = PG_GETARG_SVECTOR_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(sqrt(l2_distance_squared_internal(a, b)));
-}
-
-/*
- * Get the L2 squared distance between sparse vectors
- * This saves a sqrt calculation
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_l2_squared_distance);
-Datum
-svector_l2_squared_distance(PG_FUNCTION_ARGS)
-{
-	SVector    *a = PG_GETARG_SVECTOR_P(0);
-	SVector    *b = PG_GETARG_SVECTOR_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(l2_distance_squared_internal(a, b));
-}
-
-/*
- * Get the inner product of two sparse vectors
- */
-static double
-inner_product_internal(SVector * a, SVector * b)
-{
-	float	   *ax = SVECTOR_VALUES(a);
-	float	   *bx = SVECTOR_VALUES(b);
-	double		distance = 0.0;
-	int			bpos = 0;
-
-	for (int i = 0; i < a->nnz; i++)
-	{
-		int			ai = a->indices[i];
-
-		for (int j = bpos; j < b->nnz; j++)
-		{
-			int			bi = b->indices[j];
-
-			/* Only update when the same index */
-			if (ai == bi)
-				distance += ax[i] * bx[j];
-
-			/* Update start for next iteration */
-			if (ai >= bi)
-				bpos = j + 1;
-
-			/* Found or passed it */
-			if (bi >= ai)
-				break;
-		}
-	}
-
-	return distance;
-}
-
-/*
- * Get the inner product of two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_inner_product);
-Datum
-svector_inner_product(PG_FUNCTION_ARGS)
-{
-	SVector    *a = PG_GETARG_SVECTOR_P(0);
-	SVector    *b = PG_GETARG_SVECTOR_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(inner_product_internal(a, b));
-}
-
-/*
- * Get the negative inner product of two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_negative_inner_product);
-Datum
-svector_negative_inner_product(PG_FUNCTION_ARGS)
-{
-	SVector    *a = PG_GETARG_SVECTOR_P(0);
-	SVector    *b = PG_GETARG_SVECTOR_P(1);
-
-	CheckDims(a, b);
-
-	PG_RETURN_FLOAT8(-inner_product_internal(a, b));
-}
-
-/*
- * Get the cosine distance between two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_cosine_distance);
-Datum
-svector_cosine_distance(PG_FUNCTION_ARGS)
-{
-	SVector    *a = PG_GETARG_SVECTOR_P(0);
-	SVector    *b = PG_GETARG_SVECTOR_P(1);
-	float	   *ax = SVECTOR_VALUES(a);
-	float	   *bx = SVECTOR_VALUES(b);
-	float		norma = 0.0;
-	float		normb = 0.0;
-	double		similarity;
-
-	CheckDims(a, b);
-
-	similarity = inner_product_internal(a, b);
-
-	/* Auto-vectorized */
-	for (int i = 0; i < a->nnz; i++)
-		norma += ax[i] * ax[i];
-
-	/* Auto-vectorized */
-	for (int i = 0; i < b->nnz; i++)
-		normb += bx[i] * bx[i];
-
-	/* Use sqrt(a * b) over sqrt(a) * sqrt(b) */
-	similarity /= sqrt((double) norma * (double) normb);
-
-#ifdef _MSC_VER
-	/* /fp:fast may not propagate NaN */
-	if (isnan(similarity))
-		PG_RETURN_FLOAT8(NAN);
-#endif
-
-	/* Keep in range */
-	if (similarity > 1)
-		similarity = 1.0;
-	else if (similarity < -1)
-		similarity = -1.0;
-
-	PG_RETURN_FLOAT8(1.0 - similarity);
-}
-
-/*
- * Get the weighted Jaccard distance between two sparse vectors
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_jaccard_distance);
-Datum
-svector_jaccard_distance(PG_FUNCTION_ARGS)
-{
-	SVector    *a = PG_GETARG_SVECTOR_P(0);
-	SVector    *b = PG_GETARG_SVECTOR_P(1);
-	float	   *ax = SVECTOR_VALUES(a);
-	float	   *bx = SVECTOR_VALUES(b);
-	double		num = 0.0;
-	double		denom = 0.0;
-	int			bpos = 0;
-
-	CheckDims(a, b);
-
-	/*
-	 * Weighted Jaccard distance is not defined for vectors with negative
-	 * values. Could check and return NaN if minimal impact on performance.
-	 */
-
-	for (int i = 0; i < a->nnz; i++)
-	{
-		int			ai = a->indices[i];
-		int			bi = -1;
-
-		for (int j = bpos; j < b->nnz; j++)
-		{
-			bi = b->indices[j];
-
-			if (ai == bi)
-			{
-				num += ax[i] < bx[j] ? ax[i] : bx[j];
-				denom += ax[i] > bx[j] ? ax[i] : bx[j];
-			}
-			else if (ai > bi)
-				denom += bx[j];
-
-			/* Update start for next iteration */
-			if (ai >= bi)
-				bpos = j + 1;
-
-			/* Found or passed it */
-			if (bi >= ai)
-				break;
-		}
-
-		if (ai != bi)
-			denom += ax[i];
-	}
-
-	for (int j = bpos; j < b->nnz; j++)
-		denom += bx[j];
-
-	if (denom > 0)
-		PG_RETURN_FLOAT8(1.0 - (num / denom));
-	else
-		PG_RETURN_FLOAT8(NAN);
-}

src/svector.h

@@ -1,23 +0,0 @@
-#ifndef SVECTOR_H
-#define SVECTOR_H
-
-#define SVECTOR_MAX_DIM 100000
-
-#define SVECTOR_SIZE(_nnz)		(offsetof(SVector, indices) + (_nnz) * sizeof(int32) + (_nnz * sizeof(float)))
-#define SVECTOR_VALUES(x)		((float *) (((char *) (x)) + offsetof(SVector, indices) + (x)->nnz * sizeof(int32)))
-#define DatumGetSVector(x)		((SVector *) PG_DETOAST_DATUM(x))
-#define PG_GETARG_SVECTOR_P(x)	DatumGetSVector(PG_GETARG_DATUM(x))
-#define PG_RETURN_SVECTOR_P(x)	PG_RETURN_POINTER(x)
-
-typedef struct SVector
-{
-	int32		vl_len_;		/* varlena header (do not touch directly!) */
-	int32		dim;			/* number of dimensions */
-	int32		nnz;
-	int32		unused;
-	int32		indices[FLEXIBLE_ARRAY_MEMBER];
-}			SVector;
-
-SVector    *InitSVector(int dim, int nnz);
-
-#endif

src/vector.c

@@ -9,7 +9,6 @@
 #include "lib/stringinfo.h"
 #include "libpq/pqformat.h"
 #include "port.h"				/* for strtof() */
-#include "svector.h"
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/lsyscache.h"
@@ -725,7 +724,7 @@ vector_spherical_distance(PG_FUNCTION_ARGS)
 }
 
 /*
- * Get the L1 distance between vectors
+ * Get the L1 distance between two vectors
  */
 PGDLLEXPORT PG_FUNCTION_INFO_V1(l1_distance);
 Datum
@@ -904,8 +903,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_lt);
 Datum
 vector_lt(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) < 0);
 }
@@ -917,8 +916,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_le);
 Datum
 vector_le(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) <= 0);
 }
@@ -930,8 +929,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_eq);
 Datum
 vector_eq(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) == 0);
 }
@@ -943,8 +942,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_ne);
 Datum
 vector_ne(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) != 0);
 }
@@ -956,8 +955,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_ge);
 Datum
 vector_ge(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) >= 0);
 }
@@ -969,8 +968,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_gt);
 Datum
 vector_gt(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_BOOL(vector_cmp_internal(a, b) > 0);
 }
@@ -982,8 +981,8 @@ PGDLLEXPORT PG_FUNCTION_INFO_V1(vector_cmp);
 Datum
 vector_cmp(PG_FUNCTION_ARGS)
 {
-	Vector	   *a = (Vector *) PG_GETARG_VECTOR_P(0);
+	Vector	   *a = PG_GETARG_VECTOR_P(0);
-	Vector	   *b = (Vector *) PG_GETARG_VECTOR_P(1);
+	Vector	   *b = PG_GETARG_VECTOR_P(1);
 
 	PG_RETURN_INT32(vector_cmp_internal(a, b));
 }
@@ -1152,26 +1151,3 @@ vector_avg(PG_FUNCTION_ARGS)
 
 	PG_RETURN_POINTER(result);
 }
-
-/*
- * Convert sparse vector to dense vector
- */
-PGDLLEXPORT PG_FUNCTION_INFO_V1(svector_to_vector);
-Datum
-svector_to_vector(PG_FUNCTION_ARGS)
-{
-	SVector    *svec = PG_GETARG_SVECTOR_P(0);
-	int32		typmod = PG_GETARG_INT32(1);
-	Vector	   *result;
-	int			dim = svec->dim;
-	float	   *values = SVECTOR_VALUES(svec);
-
-	CheckDim(dim);
-	CheckExpectedDim(typmod, dim);
-
-	result = InitVector(dim);
-	for (int i = 0; i < svec->nnz; i++)
-		result->x[svec->indices[i]] = values[i];
-
-	PG_RETURN_POINTER(result);
-}

test/expected/functions.out

@@ -54,85 +54,85 @@ SELECT vector_norm('[3e37,4e37]')::real;
        5e+37
 (1 row)
 
-SELECT l2_distance('[0,0]'::vector, '[3,4]');
+SELECT l2_distance('[0,0]', '[3,4]');
  l2_distance 
 -------------
            5
 (1 row)
 
-SELECT l2_distance('[0,0]'::vector, '[0,1]');
+SELECT l2_distance('[0,0]', '[0,1]');
  l2_distance 
 -------------
            1
 (1 row)
 
-SELECT l2_distance('[1,2]'::vector, '[3]');
+SELECT l2_distance('[1,2]', '[3]');
 ERROR:  different vector dimensions 2 and 1
-SELECT l2_distance('[3e38]'::vector, '[-3e38]');
+SELECT l2_distance('[3e38]', '[-3e38]');
  l2_distance 
 -------------
     Infinity
 (1 row)
 
-SELECT inner_product('[1,2]'::vector, '[3,4]');
+SELECT inner_product('[1,2]', '[3,4]');
  inner_product 
 ---------------
             11
 (1 row)
 
-SELECT inner_product('[1,2]'::vector, '[3]');
+SELECT inner_product('[1,2]', '[3]');
 ERROR:  different vector dimensions 2 and 1
-SELECT inner_product('[3e38]'::vector, '[3e38]');
+SELECT inner_product('[3e38]', '[3e38]');
  inner_product 
 ---------------
       Infinity
 (1 row)
 
-SELECT cosine_distance('[1,2]'::vector, '[2,4]');
+SELECT cosine_distance('[1,2]', '[2,4]');
  cosine_distance 
 -----------------
                0
 (1 row)
 
-SELECT cosine_distance('[1,2]'::vector, '[0,0]');
+SELECT cosine_distance('[1,2]', '[0,0]');
  cosine_distance 
 -----------------
              NaN
 (1 row)
 
-SELECT cosine_distance('[1,1]'::vector, '[1,1]');
+SELECT cosine_distance('[1,1]', '[1,1]');
  cosine_distance 
 -----------------
                0
 (1 row)
 
-SELECT cosine_distance('[1,0]'::vector, '[0,2]');
+SELECT cosine_distance('[1,0]', '[0,2]');
  cosine_distance 
 -----------------
                1
 (1 row)
 
-SELECT cosine_distance('[1,1]'::vector, '[-1,-1]');
+SELECT cosine_distance('[1,1]', '[-1,-1]');
  cosine_distance 
 -----------------
                2
 (1 row)
 
-SELECT cosine_distance('[1,2]'::vector, '[3]');
+SELECT cosine_distance('[1,2]', '[3]');
 ERROR:  different vector dimensions 2 and 1
-SELECT cosine_distance('[1,1]'::vector, '[1.1,1.1]');
+SELECT cosine_distance('[1,1]', '[1.1,1.1]');
  cosine_distance 
 -----------------
                0
 (1 row)
 
-SELECT cosine_distance('[1,1]'::vector, '[-1.1,-1.1]');
+SELECT cosine_distance('[1,1]', '[-1.1,-1.1]');
  cosine_distance 
 -----------------
                2
 (1 row)
 
-SELECT cosine_distance('[3e38]'::vector, '[3e38]');
+SELECT cosine_distance('[3e38]', '[3e38]');
  cosine_distance 
 -----------------
              NaN

test/expected/svector.out

@@ -1,140 +0,0 @@
-SELECT '(0,1.5),(2,3.5)|5|'::svector;
-      svector       
---------------------
- (0,1.5),(2,3.5)|5|
-(1 row)
-
-SELECT '(0,1.5),(2,3.5)|5|'::svector::vector;
-     vector      
------------------
- [1.5,0,3.5,0,0]
-(1 row)
-
-SELECT '(0,1.5),(2,3.5)|5|'::svector::vector(5);
-     vector      
------------------
- [1.5,0,3.5,0,0]
-(1 row)
-
-SELECT '(0,1.5),(2,3.5)|5|'::svector::vector(4);
-ERROR:  expected 4 dimensions, not 5
-SELECT '[0,1.5,0,3.5,0]'::vector::svector;
-      svector       
---------------------
- (1,1.5),(3,3.5)|5|
-(1 row)
-
-SELECT '(0,0),(1,1),(2,0)|3|'::svector;
-       svector        
-----------------------
- (0,0),(1,1),(2,0)|3|
-(1 row)
-
-SELECT '|5|'::svector;
- svector 
----------
- |5|
-(1 row)
-
-SELECT '|-1|'::svector;
-ERROR:  svector must have at least 1 dimension
-LINE 1: SELECT '|-1|'::svector;
-               ^
-SELECT '|100001|'::svector;
-ERROR:  svector cannot have more than 100000 dimensions
-LINE 1: SELECT '|100001|'::svector;
-               ^
-SELECT '|16001|'::svector::vector;
-ERROR:  vector cannot have more than 16000 dimensions
-SELECT '(-1,1)|1|'::svector;
-ERROR:  index must not be negative
-LINE 1: SELECT '(-1,1)|1|'::svector;
-               ^
-SELECT '(1,1)|1|'::svector;
-ERROR:  index must be less than dimensions
-LINE 1: SELECT '(1,1)|1|'::svector;
-               ^
-SELECT '|1|'::svector(2);
-ERROR:  expected 2 dimensions, not 1
-SELECT l2_distance('|2|'::svector, '(0,3),(1,4)|2|');
- l2_distance 
--------------
-           5
-(1 row)
-
-SELECT l2_distance('|2|'::svector, '(1,1)|2|');
- l2_distance 
--------------
-           1
-(1 row)
-
-SELECT '|2|'::svector <-> '(0,3),(1,4)|2|';
- ?column? 
-----------
-        5
-(1 row)
-
-SELECT inner_product('(0,1),(1,2)|2|'::svector, '(0,2),(1,4)|2|');
- inner_product 
----------------
-            10
-(1 row)
-
-SELECT svector_negative_inner_product('(0,1),(1,2)|2|', '(0,2),(1,4)|2|');
- svector_negative_inner_product 
---------------------------------
-                            -10
-(1 row)
-
-SELECT cosine_distance('(0,1),(1,2)|2|'::svector, '(0,2),(1,4)|2|');
- cosine_distance 
------------------
-               0
-(1 row)
-
-SELECT cosine_distance('(0,1),(1,2)|2|'::svector, '|2|');
- cosine_distance 
------------------
-             NaN
-(1 row)
-
-SELECT cosine_distance('(0,1),(1,1)|2|'::svector, '(0,-1),(1,-1)|2|');
- cosine_distance 
------------------
-               2
-(1 row)
-
-SELECT cosine_distance('(0,1)|2|'::svector, '(1,2)|2|');
- cosine_distance 
------------------
-               1
-(1 row)
-
-SELECT cosine_distance('|1|'::svector, '|1|');
- cosine_distance 
------------------
-             NaN
-(1 row)
-
-SELECT cosine_distance('(0,1)|2|'::svector, '(0,1)|3|');
-ERROR:  different svector dimensions 2 and 3
-SELECT jaccard_distance('(0,1)|2|', '(0,1)|2|');
- jaccard_distance 
-------------------
-                0
-(1 row)
-
-SELECT jaccard_distance('(0,1)|2|', '(1,1)|2|');
- jaccard_distance 
-------------------
-                1
-(1 row)
-
-SELECT jaccard_distance('|1|', '|1|');
- jaccard_distance 
-------------------
-              NaN
-(1 row)
-
-SELECT jaccard_distance('(0,1)|2|', '(0,1)|3|');
-ERROR:  different svector dimensions 2 and 3

test/sql/functions.sql

@@ -13,24 +13,24 @@ SELECT vector_norm('[3,4]');
 SELECT vector_norm('[0,1]');
 SELECT vector_norm('[3e37,4e37]')::real;
 
-SELECT l2_distance('[0,0]'::vector, '[3,4]');
+SELECT l2_distance('[0,0]', '[3,4]');
-SELECT l2_distance('[0,0]'::vector, '[0,1]');
+SELECT l2_distance('[0,0]', '[0,1]');
-SELECT l2_distance('[1,2]'::vector, '[3]');
+SELECT l2_distance('[1,2]', '[3]');
-SELECT l2_distance('[3e38]'::vector, '[-3e38]');
+SELECT l2_distance('[3e38]', '[-3e38]');
 
-SELECT inner_product('[1,2]'::vector, '[3,4]');
+SELECT inner_product('[1,2]', '[3,4]');
-SELECT inner_product('[1,2]'::vector, '[3]');
+SELECT inner_product('[1,2]', '[3]');
-SELECT inner_product('[3e38]'::vector, '[3e38]');
+SELECT inner_product('[3e38]', '[3e38]');
 
-SELECT cosine_distance('[1,2]'::vector, '[2,4]');
+SELECT cosine_distance('[1,2]', '[2,4]');
-SELECT cosine_distance('[1,2]'::vector, '[0,0]');
+SELECT cosine_distance('[1,2]', '[0,0]');
-SELECT cosine_distance('[1,1]'::vector, '[1,1]');
+SELECT cosine_distance('[1,1]', '[1,1]');
-SELECT cosine_distance('[1,0]'::vector, '[0,2]');
+SELECT cosine_distance('[1,0]', '[0,2]');
-SELECT cosine_distance('[1,1]'::vector, '[-1,-1]');
+SELECT cosine_distance('[1,1]', '[-1,-1]');
-SELECT cosine_distance('[1,2]'::vector, '[3]');
+SELECT cosine_distance('[1,2]', '[3]');
-SELECT cosine_distance('[1,1]'::vector, '[1.1,1.1]');
+SELECT cosine_distance('[1,1]', '[1.1,1.1]');
-SELECT cosine_distance('[1,1]'::vector, '[-1.1,-1.1]');
+SELECT cosine_distance('[1,1]', '[-1.1,-1.1]');
-SELECT cosine_distance('[3e38]'::vector, '[3e38]');
+SELECT cosine_distance('[3e38]', '[3e38]');
 
 SELECT l1_distance('[0,0]', '[3,4]');
 SELECT l1_distance('[0,0]', '[0,1]');

test/sql/svector.sql

@@ -1,36 +0,0 @@
-SELECT '(0,1.5),(2,3.5)|5|'::svector;
-SELECT '(0,1.5),(2,3.5)|5|'::svector::vector;
-SELECT '(0,1.5),(2,3.5)|5|'::svector::vector(5);
-SELECT '(0,1.5),(2,3.5)|5|'::svector::vector(4);
-SELECT '[0,1.5,0,3.5,0]'::vector::svector;
-
-SELECT '(0,0),(1,1),(2,0)|3|'::svector;
-
-SELECT '|5|'::svector;
-SELECT '|-1|'::svector;
-SELECT '|100001|'::svector;
-SELECT '|16001|'::svector::vector;
-
-SELECT '(-1,1)|1|'::svector;
-SELECT '(1,1)|1|'::svector;
-
-SELECT '|1|'::svector(2);
-
-SELECT l2_distance('|2|'::svector, '(0,3),(1,4)|2|');
-SELECT l2_distance('|2|'::svector, '(1,1)|2|');
-SELECT '|2|'::svector <-> '(0,3),(1,4)|2|';
-
-SELECT inner_product('(0,1),(1,2)|2|'::svector, '(0,2),(1,4)|2|');
-SELECT svector_negative_inner_product('(0,1),(1,2)|2|', '(0,2),(1,4)|2|');
-
-SELECT cosine_distance('(0,1),(1,2)|2|'::svector, '(0,2),(1,4)|2|');
-SELECT cosine_distance('(0,1),(1,2)|2|'::svector, '|2|');
-SELECT cosine_distance('(0,1),(1,1)|2|'::svector, '(0,-1),(1,-1)|2|');
-SELECT cosine_distance('(0,1)|2|'::svector, '(1,2)|2|');
-SELECT cosine_distance('|1|'::svector, '|1|');
-SELECT cosine_distance('(0,1)|2|'::svector, '(0,1)|3|');
-
-SELECT jaccard_distance('(0,1)|2|', '(0,1)|2|');
-SELECT jaccard_distance('(0,1)|2|', '(1,1)|2|');
-SELECT jaccard_distance('|1|', '|1|');
-SELECT jaccard_distance('(0,1)|2|', '(0,1)|3|');

test/t/012_hnsw_build_recall.pl

@@ -83,11 +83,46 @@ for my $i (0 .. $#operators)
 		push(@expected, $res);
 	}
 
-	# Add index
+	# Build index serially
-	$node->safe_psql("postgres", "CREATE INDEX ON tst USING hnsw (v $opclass);");
+	$node->safe_psql("postgres", qq(
+		SET max_parallel_maintenance_workers = 0;
+		CREATE INDEX idx ON tst USING hnsw (v $opclass);
+	));
 
+	# Test approximate results
 	my $min = $operator eq "<#>" ? 0.80 : 0.99;
 	test_recall($min, $operator);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+
+	# Build index in parallel in memory
+	my ($ret, $stdout, $stderr) = $node->psql("postgres", qq(
+		SET client_min_messages = DEBUG;
+		SET min_parallel_table_scan_size = 1;
+		CREATE INDEX idx ON tst USING hnsw (v $opclass);
+	));
+	is($ret, 0, $stderr);
+	like($stderr, qr/using \d+ parallel workers/);
+
+	# Test approximate results
+	test_recall($min, $operator);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
+
+	# Build index in parallel on disk
+	# Set parallel_workers on table to use workers with low maintenance_work_mem
+	($ret, $stdout, $stderr) = $node->psql("postgres", qq(
+		ALTER TABLE tst SET (parallel_workers = 2);
+		SET client_min_messages = DEBUG;
+		SET maintenance_work_mem = '4MB';
+		CREATE INDEX idx ON tst USING hnsw (v $opclass);
+		ALTER TABLE tst RESET (parallel_workers);
+	));
+	is($ret, 0, $stderr);
+	like($stderr, qr/using \d+ parallel workers/);
+	like($stderr, qr/hnsw graph no longer fits into maintenance_work_mem/);
+
+	$node->safe_psql("postgres", "DROP INDEX idx;");
 }
 
 done_testing();