Compare commits

..

12 Commits

Author SHA1 Message Date
Andrew Kane
191bef7e35 Started ARM support [skip ci] 2024-03-30 09:43:18 -07:00
Andrew Kane
582e6ad821 Keep -march=native for Mac x86-64 [skip ci] 2024-03-28 01:30:56 -07:00
Andrew Kane
d0028ae769 Restored comment [skip ci] 2024-03-27 20:40:39 -07:00
Andrew Kane
81dc35b62b Check x86-64 [skip ci] 2024-03-27 19:16:49 -07:00
Andrew Kane
c82d15acad Test runtime dispatching [skip ci] 2024-03-27 18:30:39 -07:00
Andrew Kane
ba18942fcf Removed normvec from IVFFlat for simplicity (no difference in performance) 2024-03-27 16:41:17 -07:00
Andrew Kane
8e59455c3c Removed normvec for simplicity (no difference in performance) 2024-03-27 16:33:11 -07:00
Andrew Kane
bd50e3067d Updated readme [skip ci] 2024-03-27 14:14:49 -07:00
Andrew Kane
af9d4ad659 Updated readme [skip ci] 2024-03-27 14:12:08 -07:00
Andrew Kane
08abb63cbe Added notes about NULL vectors [skip ci] 2024-03-27 11:50:37 -07:00
Andrew Kane
06b8556a49 Revert "Updated readme [skip ci]"
This reverts commit 3f674c9994.
2024-03-25 23:33:46 -07:00
Andrew Kane
3f674c9994 Updated readme [skip ci] 2024-03-25 23:33:17 -07:00
28 changed files with 60 additions and 593 deletions

View File

@@ -1,10 +1,3 @@
## 0.7.0 (unreleased)
- Added support for binary vectors to HNSW
- Added `hamming_distance` function
- Added `jaccard_distance` function
- Added `quantize_binary` function
## 0.6.2 (2024-03-18) ## 0.6.2 (2024-03-18)
- Reduced lock contention with parallel HNSW index builds - Reduced lock contention with parallel HNSW index builds

View File

@@ -3,28 +3,22 @@ EXTVERSION = 0.6.2
MODULE_big = vector MODULE_big = vector
DATA = $(wildcard sql/*--*.sql) DATA = $(wildcard sql/*--*.sql)
OBJS = src/bitvector.o 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 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/vector.h HEADERS = src/vector.h
TESTS = $(wildcard test/sql/*.sql) TESTS = $(wildcard test/sql/*.sql)
REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS)) REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))
REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION) REGRESS_OPTS = --inputdir=test --load-extension=$(EXTENSION)
OPTFLAGS = -march=native OPTFLAGS =
# Mac ARM doesn't support -march=native # Since runtime dispatch not supported
ifeq ($(shell uname -s), Darwin) ifeq ($(shell uname -s), Darwin)
ifeq ($(shell uname -p), arm) ifeq ($(shell uname -m), x86_64)
# no difference with -march=armv8.5-a OPTFLAGS = -march=native
OPTFLAGS =
endif endif
endif endif
# PowerPC doesn't support -march=native
ifneq ($(filter ppc64%, $(shell uname -m)), )
OPTFLAGS =
endif
# For auto-vectorization: # For auto-vectorization:
# - GCC (needs -ftree-vectorize OR -O3) - https://gcc.gnu.org/projects/tree-ssa/vectorization.html # - 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 # - Clang (could use pragma instead) - https://llvm.org/docs/Vectorizers.html

View File

@@ -1,7 +1,7 @@
EXTENSION = vector EXTENSION = vector
EXTVERSION = 0.6.2 EXTVERSION = 0.6.2
OBJS = src\bitvector.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\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\vector.h HEADERS = src\vector.h
REGRESS = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged REGRESS = btree cast copy functions input ivfflat_cosine ivfflat_ip ivfflat_l2 ivfflat_options ivfflat_unlogged

View File

@@ -5,7 +5,7 @@ Open-source vector similarity search for Postgres
Store your vectors with the rest of your data. Supports: Store your vectors with the rest of your data. Supports:
- exact and approximate nearest neighbor search - exact and approximate nearest neighbor search
- L2 distance, inner product, cosine distance, and more - L2 distance, inner product, and cosine distance
- any [language](#languages) with a Postgres client - any [language](#languages) with a Postgres client
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 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
@@ -221,19 +221,7 @@ Cosine distance
CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops); CREATE INDEX ON items USING hnsw (embedding vector_cosine_ops);
``` ```
Hamming distance - added in 0.7.0 Vectors with up to 2,000 dimensions can be indexed.
```sql
CREATE INDEX ON items USING hnsw (embedding bit_hamming_ops);
```
Jaccard distance - added in 0.7.0
```sql
CREATE INDEX ON items USING hnsw (embedding bit_jaccard_ops);
```
Vectors with up to 2,000 dimensions can be indexed, or bit vectors with up to 64,000 dimensions.
### Index Options ### Index Options
@@ -683,6 +671,8 @@ ALTER TABLE items ALTER COLUMN embedding SET STORAGE PLAIN;
Results are limited by the size of the dynamic candidate list (`hnsw.ef_search`). There may be even less results due to dead tuples or filtering conditions in the query. We recommend setting `hnsw.ef_search` to at least twice the `LIMIT` of the query. If you need more than 500 results, use an IVFFlat index instead. Results are limited by the size of the dynamic candidate list (`hnsw.ef_search`). There may be even less results due to dead tuples or filtering conditions in the query. We recommend setting `hnsw.ef_search` to at least twice the `LIMIT` of the query. If you need more than 500 results, use an IVFFlat index instead.
Also, note that `NULL` vectors are not indexed (as well as zero vectors for cosine distance).
#### Why are there less results for a query after adding an IVFFlat index? #### Why are there less results for a query after adding an IVFFlat index?
The index was likely created with too little data for the number of lists. Drop the index until the table has more data. The index was likely created with too little data for the number of lists. Drop the index until the table has more data.
@@ -693,11 +683,13 @@ DROP INDEX index_name;
Results can also be limited by the number of probes (`ivfflat.probes`). Results can also be limited by the number of probes (`ivfflat.probes`).
Also, note that `NULL` vectors are not indexed (as well as zero vectors for cosine distance).
## Reference ## Reference
### Vector Type ### Vector Type
Each vector takes `4 * dimensions + 8` bytes of storage. Each element is a single precision floating-point number (like the `real` type in Postgres), and all elements must be finite (no `NaN`, `Infinity` or `-Infinity`). Vectors can have up to 16,000 dimensions. Each vector takes `4 * dimensions + 8` bytes of storage. Each element is a single-precision floating-point number (like the `real` type in Postgres), and all elements must be finite (no `NaN`, `Infinity` or `-Infinity`). Vectors can have up to 16,000 dimensions.
### Vector Operators ### Vector Operators
@@ -718,31 +710,16 @@ cosine_distance(vector, vector) → double precision | cosine distance |
inner_product(vector, vector) → double precision | inner product | inner_product(vector, vector) → double precision | inner product |
l2_distance(vector, vector) → double precision | Euclidean distance | l2_distance(vector, vector) → double precision | Euclidean distance |
l1_distance(vector, vector) → double precision | taxicab distance | 0.5.0 l1_distance(vector, vector) → double precision | taxicab distance | 0.5.0
quantize_binary(vector) → bit | quantize | 0.7.0
vector_dims(vector) → integer | number of dimensions | vector_dims(vector) → integer | number of dimensions |
vector_norm(vector) → double precision | Euclidean norm | vector_norm(vector) → double precision | Euclidean norm |
### Aggregate Functions ### Vector Aggregate Functions
Function | Description | Added Function | Description | Added
--- | --- | --- --- | --- | ---
avg(vector) → vector | average | avg(vector) → vector | average |
sum(vector) → vector | sum | 0.5.0 sum(vector) → vector | sum | 0.5.0
### Bit Operators
Operator | Description | Added
--- | --- | ---
<~> | Hamming distance | 0.7.0
<%> | Jaccard distance | 0.7.0
### Bit Functions
Function | Description | Added
--- | --- | ---
hamming_distance(bit, bit) → double precision | Hamming distance | 0.7.0
jaccard_distance(bit, bit) → double precision | Jaccard distance | 0.7.0
## Installation Notes - Linux and Mac ## Installation Notes - Linux and Mac
### Postgres Location ### Postgres Location

View File

@@ -1,31 +0,0 @@
-- complain if script is sourced in psql, rather than via CREATE EXTENSION
\echo Use "ALTER EXTENSION vector UPDATE TO '0.7.0'" to load this file. \quit
CREATE FUNCTION quantize_binary(vector) RETURNS bit
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION jaccard_distance(bit, bit) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE OPERATOR <~> (
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = hamming_distance,
COMMUTATOR = '<~>'
);
CREATE OPERATOR <%> (
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = jaccard_distance,
COMMUTATOR = '<%>'
);
CREATE OPERATOR CLASS bit_hamming_ops
FOR TYPE bit USING hnsw AS
OPERATOR 1 <~> (bit, bit) FOR ORDER BY float_ops,
FUNCTION 1 hamming_distance(bit, bit);
CREATE OPERATOR CLASS bit_jaccard_ops
FOR TYPE bit USING hnsw AS
OPERATOR 1 <%> (bit, bit) FOR ORDER BY float_ops,
FUNCTION 1 jaccard_distance(bit, bit);

View File

@@ -58,9 +58,6 @@ CREATE FUNCTION vector_sub(vector, vector) RETURNS vector
CREATE FUNCTION vector_mul(vector, vector) RETURNS vector CREATE FUNCTION vector_mul(vector, vector) RETURNS vector
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE; AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION quantize_binary(vector) RETURNS bit
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
-- private functions -- private functions
CREATE FUNCTION vector_lt(vector, vector) RETURNS bool CREATE FUNCTION vector_lt(vector, vector) RETURNS bool
@@ -290,31 +287,3 @@ CREATE OPERATOR CLASS vector_cosine_ops
OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops, OPERATOR 1 <=> (vector, vector) FOR ORDER BY float_ops,
FUNCTION 1 vector_negative_inner_product(vector, vector), FUNCTION 1 vector_negative_inner_product(vector, vector),
FUNCTION 2 vector_norm(vector); FUNCTION 2 vector_norm(vector);
-- bit functions
CREATE FUNCTION hamming_distance(bit, bit) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE FUNCTION jaccard_distance(bit, bit) RETURNS float8
AS 'MODULE_PATHNAME' LANGUAGE C IMMUTABLE STRICT PARALLEL SAFE;
CREATE OPERATOR <~> (
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = hamming_distance,
COMMUTATOR = '<~>'
);
CREATE OPERATOR <%> (
LEFTARG = bit, RIGHTARG = bit, PROCEDURE = jaccard_distance,
COMMUTATOR = '<%>'
);
CREATE OPERATOR CLASS bit_hamming_ops
FOR TYPE bit USING hnsw AS
OPERATOR 1 <~> (bit, bit) FOR ORDER BY float_ops,
FUNCTION 1 hamming_distance(bit, bit);
CREATE OPERATOR CLASS bit_jaccard_ops
FOR TYPE bit USING hnsw AS
OPERATOR 1 <%> (bit, bit) FOR ORDER BY float_ops,
FUNCTION 1 jaccard_distance(bit, bit);

View File

@@ -1,90 +0,0 @@
#include "postgres.h"
#include "bitvector.h"
#include "port/pg_bitutils.h"
#include "utils/varbit.h"
#if PG_VERSION_NUM >= 160000
#include "varatt.h"
#endif
/*
* Allocate and initialize a new bit vector
*/
VarBit *
InitBitVector(int dim)
{
VarBit *result;
int size;
size = VARBITTOTALLEN(dim);
result = (VarBit *) palloc0(size);
SET_VARSIZE(result, size);
VARBITLEN(result) = dim;
return result;
}
/*
* Ensure same number of bits
*/
static inline void
CheckBitLengths(uint32 aLen, uint32 bLen)
{
if (aLen != bLen)
ereport(ERROR,
(errcode(ERRCODE_DATA_EXCEPTION),
errmsg("different bit lengths %u and %u", aLen, bLen)));
}
/*
* Get the Hamming distance between two bit strings
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(hamming_distance);
Datum
hamming_distance(PG_FUNCTION_ARGS)
{
VarBit *a = PG_GETARG_VARBIT_P(0);
VarBit *b = PG_GETARG_VARBIT_P(1);
unsigned char *ax = VARBITS(a);
unsigned char *bx = VARBITS(b);
uint64 distance = 0;
CheckBitLengths(VARBITLEN(a), VARBITLEN(b));
/* TODO Improve performance */
for (uint32 i = 0; i < VARBITBYTES(a); i++)
distance += pg_number_of_ones[ax[i] ^ bx[i]];
PG_RETURN_FLOAT8((double) distance);
}
/*
* Get the Jaccard distance between two bit strings
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(jaccard_distance);
Datum
jaccard_distance(PG_FUNCTION_ARGS)
{
VarBit *a = PG_GETARG_VARBIT_P(0);
VarBit *b = PG_GETARG_VARBIT_P(1);
unsigned char *ax = VARBITS(a);
unsigned char *bx = VARBITS(b);
uint64 ab = 0;
uint64 aa;
uint64 bb;
CheckBitLengths(VARBITLEN(a), VARBITLEN(b));
/* TODO Improve performance */
for (uint32 i = 0; i < VARBITBYTES(a); i++)
ab += pg_number_of_ones[ax[i] & bx[i]];
if (ab == 0)
PG_RETURN_FLOAT8(1);
aa = pg_popcount((char *) ax, VARBITBYTES(a));
bb = pg_popcount((char *) bx, VARBITBYTES(b));
PG_RETURN_FLOAT8(1 - (ab / ((double) (aa + bb - ab))));
}

View File

@@ -1,8 +0,0 @@
#ifndef BITVECTOR_H
#define BITVECTOR_H
#include "utils/varbit.h"
VarBit *InitBitVector(int dim);
#endif

View File

@@ -262,7 +262,6 @@ typedef struct HnswBuildState
HnswGraph *graph; HnswGraph *graph;
double ml; double ml;
int maxLevel; int maxLevel;
Vector *normvec;
/* Memory */ /* Memory */
MemoryContext graphCtx; MemoryContext graphCtx;
@@ -367,7 +366,7 @@ typedef struct HnswVacuumState
int HnswGetM(Relation index); int HnswGetM(Relation index);
int HnswGetEfConstruction(Relation index); int HnswGetEfConstruction(Relation index);
FmgrInfo *HnswOptionalProcInfo(Relation index, uint16 procnum); FmgrInfo *HnswOptionalProcInfo(Relation index, uint16 procnum);
bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result); bool HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value);
Buffer HnswNewBuffer(Relation index, ForkNumber forkNum); Buffer HnswNewBuffer(Relation index, ForkNumber forkNum);
void HnswInitPage(Buffer buf, Page page); void HnswInitPage(Buffer buf, Page page);
void HnswInit(void); void HnswInit(void);

View File

@@ -44,7 +44,6 @@
#include "access/xact.h" #include "access/xact.h"
#include "access/xloginsert.h" #include "access/xloginsert.h"
#include "catalog/index.h" #include "catalog/index.h"
#include "catalog/pg_type_d.h"
#include "commands/progress.h" #include "commands/progress.h"
#include "hnsw.h" #include "hnsw.h"
#include "miscadmin.h" #include "miscadmin.h"
@@ -490,7 +489,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heaptid, Hn
/* Normalize if needed */ /* Normalize if needed */
if (buildstate->normprocinfo != NULL) if (buildstate->normprocinfo != NULL)
{ {
if (!HnswNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec)) if (!HnswNormValue(buildstate->normprocinfo, buildstate->collation, &value))
return false; return false;
} }
@@ -672,12 +671,6 @@ HnswSharedMemoryAlloc(Size size, void *state)
static void static void
InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, IndexInfo *indexInfo, ForkNumber forkNum) InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, IndexInfo *indexInfo, ForkNumber forkNum)
{ {
int maxDimensions = HNSW_MAX_DIM;
Oid typid = TupleDescAttr(index->rd_att, 0)->atttypid;
if (typid == BITOID || typid == VARBITOID)
maxDimensions *= 32;
buildstate->heap = heap; buildstate->heap = heap;
buildstate->index = index; buildstate->index = index;
buildstate->indexInfo = indexInfo; buildstate->indexInfo = indexInfo;
@@ -691,8 +684,8 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
if (buildstate->dimensions < 0) if (buildstate->dimensions < 0)
elog(ERROR, "column does not have dimensions"); elog(ERROR, "column does not have dimensions");
if (buildstate->dimensions > maxDimensions) if (buildstate->dimensions > HNSW_MAX_DIM)
elog(ERROR, "column cannot have more than %d dimensions for hnsw index", maxDimensions); elog(ERROR, "column cannot have more than %d dimensions for hnsw index", HNSW_MAX_DIM);
if (buildstate->efConstruction < 2 * buildstate->m) if (buildstate->efConstruction < 2 * buildstate->m)
elog(ERROR, "ef_construction must be greater than or equal to 2 * m"); elog(ERROR, "ef_construction must be greater than or equal to 2 * m");
@@ -710,9 +703,6 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
buildstate->ml = HnswGetMl(buildstate->m); buildstate->ml = HnswGetMl(buildstate->m);
buildstate->maxLevel = HnswGetMaxLevel(buildstate->m); buildstate->maxLevel = HnswGetMaxLevel(buildstate->m);
/* Reuse for each tuple */
buildstate->normvec = InitVector(buildstate->dimensions);
buildstate->graphCtx = GenerationContextCreate(CurrentMemoryContext, buildstate->graphCtx = GenerationContextCreate(CurrentMemoryContext,
"Hnsw build graph context", "Hnsw build graph context",
#if PG_VERSION_NUM >= 150000 #if PG_VERSION_NUM >= 150000
@@ -736,7 +726,6 @@ InitBuildState(HnswBuildState * buildstate, Relation heap, Relation index, Index
static void static void
FreeBuildState(HnswBuildState * buildstate) FreeBuildState(HnswBuildState * buildstate)
{ {
pfree(buildstate->normvec);
MemoryContextDelete(buildstate->graphCtx); MemoryContextDelete(buildstate->graphCtx);
MemoryContextDelete(buildstate->tmpCtx); MemoryContextDelete(buildstate->tmpCtx);
} }

View File

@@ -622,7 +622,7 @@ HnswInsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_ti
normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC); normprocinfo = HnswOptionalProcInfo(index, HNSW_NORM_PROC);
if (normprocinfo != NULL) if (normprocinfo != NULL)
{ {
if (!HnswNormValue(normprocinfo, collation, &value, NULL)) if (!HnswNormValue(normprocinfo, collation, &value))
return; return;
} }

View File

@@ -1,8 +1,6 @@
#include "postgres.h" #include "postgres.h"
#include "access/relscan.h" #include "access/relscan.h"
#include "bitvector.h"
#include "catalog/pg_type_d.h"
#include "hnsw.h" #include "hnsw.h"
#include "pgstat.h" #include "pgstat.h"
#include "storage/bufmgr.h" #include "storage/bufmgr.h"
@@ -75,15 +73,7 @@ GetScanValue(IndexScanDesc scan)
Datum value; Datum value;
if (scan->orderByData->sk_flags & SK_ISNULL) if (scan->orderByData->sk_flags & SK_ISNULL)
{ value = PointerGetDatum(InitVector(GetDimensions(scan->indexRelation)));
Oid typid = TupleDescAttr(scan->indexRelation->rd_att, 0)->atttypid;
int dimensions = GetDimensions(scan->indexRelation);
if (typid == BITOID || typid == VARBITOID)
value = PointerGetDatum(InitBitVector(dimensions));
else
value = PointerGetDatum(InitVector(dimensions));
}
else else
{ {
value = scan->orderByData->sk_argument; value = scan->orderByData->sk_argument;
@@ -94,7 +84,7 @@ GetScanValue(IndexScanDesc scan)
/* Fine if normalization fails */ /* Fine if normalization fails */
if (so->normprocinfo != NULL) if (so->normprocinfo != NULL)
HnswNormValue(so->normprocinfo, so->collation, &value, NULL); HnswNormValue(so->normprocinfo, so->collation, &value);
} }
return value; return value;

View File

@@ -158,16 +158,14 @@ HnswOptionalProcInfo(Relation index, uint16 procnum)
* if it's different than the original value * if it's different than the original value
*/ */
bool bool
HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result) HnswNormValue(FmgrInfo *procinfo, Oid collation, Datum *value)
{ {
double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value)); double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
if (norm > 0) if (norm > 0)
{ {
Vector *v = DatumGetVector(*value); Vector *v = DatumGetVector(*value);
Vector *result = InitVector(v->dim);
if (result == NULL)
result = InitVector(v->dim);
for (int i = 0; i < v->dim; i++) for (int i = 0; i < v->dim; i++)
result->x[i] = v->x[i] / norm; result->x[i] = v->x[i] / norm;

View File

@@ -57,7 +57,7 @@ AddSample(Datum *values, IvfflatBuildState * buildstate)
*/ */
if (buildstate->kmeansnormprocinfo != NULL) if (buildstate->kmeansnormprocinfo != NULL)
{ {
if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value, buildstate->normvec)) if (!IvfflatNormValue(buildstate->kmeansnormprocinfo, buildstate->collation, &value))
return; return;
} }
@@ -153,7 +153,7 @@ AddTupleToSort(Relation index, ItemPointer tid, Datum *values, IvfflatBuildState
/* Normalize if needed */ /* Normalize if needed */
if (buildstate->normprocinfo != NULL) if (buildstate->normprocinfo != NULL)
{ {
if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value, buildstate->normvec)) if (!IvfflatNormValue(buildstate->normprocinfo, buildstate->collation, &value))
return; return;
} }
@@ -356,9 +356,6 @@ InitBuildState(IvfflatBuildState * buildstate, Relation heap, Relation index, In
buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions); buildstate->centers = VectorArrayInit(buildstate->lists, buildstate->dimensions);
buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists); buildstate->listInfo = palloc(sizeof(ListInfo) * buildstate->lists);
/* Reuse for each tuple */
buildstate->normvec = InitVector(buildstate->dimensions);
buildstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext, buildstate->tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Ivfflat build temporary context", "Ivfflat build temporary context",
ALLOCSET_DEFAULT_SIZES); ALLOCSET_DEFAULT_SIZES);
@@ -380,7 +377,6 @@ FreeBuildState(IvfflatBuildState * buildstate)
{ {
VectorArrayFree(buildstate->centers); VectorArrayFree(buildstate->centers);
pfree(buildstate->listInfo); pfree(buildstate->listInfo);
pfree(buildstate->normvec);
#ifdef IVFFLAT_KMEANS_DEBUG #ifdef IVFFLAT_KMEANS_DEBUG
pfree(buildstate->listSums); pfree(buildstate->listSums);

View File

@@ -172,7 +172,6 @@ typedef struct IvfflatBuildState
VectorArray samples; VectorArray samples;
VectorArray centers; VectorArray centers;
ListInfo *listInfo; ListInfo *listInfo;
Vector *normvec;
#ifdef IVFFLAT_KMEANS_DEBUG #ifdef IVFFLAT_KMEANS_DEBUG
double inertia; double inertia;
@@ -267,7 +266,7 @@ void VectorArrayFree(VectorArray arr);
void PrintVectorArray(char *msg, VectorArray arr); void PrintVectorArray(char *msg, VectorArray arr);
void IvfflatKmeans(Relation index, VectorArray samples, VectorArray centers); void IvfflatKmeans(Relation index, VectorArray samples, VectorArray centers);
FmgrInfo *IvfflatOptionalProcInfo(Relation index, uint16 procnum); FmgrInfo *IvfflatOptionalProcInfo(Relation index, uint16 procnum);
bool IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result); bool IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value);
int IvfflatGetLists(Relation index); int IvfflatGetLists(Relation index);
void IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions); void IvfflatGetMetaPageInfo(Relation index, int *lists, int *dimensions);
void IvfflatUpdateList(Relation index, ListInfo listInfo, BlockNumber insertPage, BlockNumber originalInsertPage, BlockNumber startPage, ForkNumber forkNum); void IvfflatUpdateList(Relation index, ListInfo listInfo, BlockNumber insertPage, BlockNumber originalInsertPage, BlockNumber startPage, ForkNumber forkNum);

View File

@@ -85,7 +85,7 @@ InsertTuple(Relation index, Datum *values, bool *isnull, ItemPointer heap_tid, R
normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC); normprocinfo = IvfflatOptionalProcInfo(index, IVFFLAT_NORM_PROC);
if (normprocinfo != NULL) if (normprocinfo != NULL)
{ {
if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value, NULL)) if (!IvfflatNormValue(normprocinfo, index->rd_indcollation[0], &value))
return; return;
} }

View File

@@ -293,7 +293,7 @@ ivfflatgettuple(IndexScanDesc scan, ScanDirection dir)
/* Fine if normalization fails */ /* Fine if normalization fails */
if (so->normprocinfo != NULL) if (so->normprocinfo != NULL)
IvfflatNormValue(so->normprocinfo, so->collation, &value, NULL); IvfflatNormValue(so->normprocinfo, so->collation, &value);
} }
IvfflatBench("GetScanLists", GetScanLists(scan, value)); IvfflatBench("GetScanLists", GetScanLists(scan, value));

View File

@@ -75,16 +75,14 @@ IvfflatOptionalProcInfo(Relation index, uint16 procnum)
* if it's different than the original value * if it's different than the original value
*/ */
bool bool
IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value, Vector * result) IvfflatNormValue(FmgrInfo *procinfo, Oid collation, Datum *value)
{ {
double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value)); double norm = DatumGetFloat8(FunctionCall1Coll(procinfo, collation, *value));
if (norm > 0) if (norm > 0)
{ {
Vector *v = DatumGetVector(*value); Vector *v = DatumGetVector(*value);
Vector *result = InitVector(v->dim);
if (result == NULL)
result = InitVector(v->dim);
for (int i = 0; i < v->dim; i++) for (int i = 0; i < v->dim; i++)
result->x[i] = v->x[i] / norm; result->x[i] = v->x[i] / norm;

View File

@@ -2,7 +2,6 @@
#include <math.h> #include <math.h>
#include "bitvector.h"
#include "catalog/pg_type.h" #include "catalog/pg_type.h"
#include "common/shortest_dec.h" #include "common/shortest_dec.h"
#include "fmgr.h" #include "fmgr.h"
@@ -30,6 +29,15 @@
#define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1) #define STATE_DIMS(x) (ARR_DIMS(x)[0] - 1)
#define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1)) #define CreateStateDatums(dim) palloc(sizeof(Datum) * (dim + 1))
#if defined(__x86_64__) && defined(__gnu_linux__) && defined(__has_attribute) && __has_attribute(target_clones)
#define RUNTIME_DISPATCH __attribute__((target_clones("default", "avx", "fma", "avx512f")))
#elif defined(__aarch64__) && defined(__gnu_linux__) && defined(__has_attribute) && __has_attribute(target_clones)
/* TODO Fix error: target does not support function version dispatcher */
#define RUNTIME_DISPATCH __attribute__((target_clones("default", "arch=armv8.5-a")))
#else
#define RUNTIME_DISPATCH
#endif
PG_MODULE_MAGIC; PG_MODULE_MAGIC;
/* /*
@@ -533,6 +541,23 @@ vector_to_float4(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result); PG_RETURN_POINTER(result);
} }
RUNTIME_DISPATCH
static float
l2_squared_distance_impl(int16 dim, float *ax, float *bx)
{
float distance = 0.0;
/* Auto-vectorized */
for (int16 i = 0; i < dim; i++)
{
float diff = ax[i] - bx[i];
distance += diff * diff;
}
return distance;
}
/* /*
* Get the L2 distance between vectors * Get the L2 distance between vectors
*/ */
@@ -542,19 +567,11 @@ l2_distance(PG_FUNCTION_ARGS)
{ {
Vector *a = PG_GETARG_VECTOR_P(0); Vector *a = PG_GETARG_VECTOR_P(0);
Vector *b = PG_GETARG_VECTOR_P(1); Vector *b = PG_GETARG_VECTOR_P(1);
float *ax = a->x; float distance;
float *bx = b->x;
float distance = 0.0;
float diff;
CheckDims(a, b); CheckDims(a, b);
/* Auto-vectorized */ distance = l2_squared_distance_impl(a->dim, a->x, b->x);
for (int i = 0; i < a->dim; i++)
{
diff = ax[i] - bx[i];
distance += diff * diff;
}
PG_RETURN_FLOAT8(sqrt((double) distance)); PG_RETURN_FLOAT8(sqrt((double) distance));
} }
@@ -569,19 +586,11 @@ vector_l2_squared_distance(PG_FUNCTION_ARGS)
{ {
Vector *a = PG_GETARG_VECTOR_P(0); Vector *a = PG_GETARG_VECTOR_P(0);
Vector *b = PG_GETARG_VECTOR_P(1); Vector *b = PG_GETARG_VECTOR_P(1);
float *ax = a->x; float distance;
float *bx = b->x;
float distance = 0.0;
float diff;
CheckDims(a, b); CheckDims(a, b);
/* Auto-vectorized */ distance = l2_squared_distance_impl(a->dim, a->x, b->x);
for (int i = 0; i < a->dim; i++)
{
diff = ax[i] - bx[i];
distance += diff * diff;
}
PG_RETURN_FLOAT8((double) distance); PG_RETURN_FLOAT8((double) distance);
} }
@@ -861,26 +870,6 @@ vector_mul(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result); PG_RETURN_POINTER(result);
} }
/*
* Quantize a vector
*/
PGDLLEXPORT PG_FUNCTION_INFO_V1(quantize_binary);
Datum
quantize_binary(PG_FUNCTION_ARGS)
{
Vector *a = PG_GETARG_VECTOR_P(0);
float *ax = a->x;
VarBit *result = InitBitVector(a->dim);
unsigned char *rx = VARBITS(result);
/* TODO Improve */
for (int i = 0; i < a->dim; i++)
rx[i / 8] |= (ax[i] > 0) << (7 - (i % 8));
PG_RETURN_VARBIT_P(result);
}
/* /*
* Internal helper to compare vectors * Internal helper to compare vectors
*/ */

View File

@@ -1,64 +0,0 @@
SELECT hamming_distance(B'111', B'111');
hamming_distance
------------------
0
(1 row)
SELECT hamming_distance(B'111', B'110');
hamming_distance
------------------
1
(1 row)
SELECT hamming_distance(B'111', B'100');
hamming_distance
------------------
2
(1 row)
SELECT hamming_distance(B'111', B'000');
hamming_distance
------------------
3
(1 row)
SELECT hamming_distance(B'111', B'00');
ERROR: different bit lengths 3 and 2
SELECT jaccard_distance(B'1111', B'1111');
jaccard_distance
------------------
0
(1 row)
SELECT jaccard_distance(B'1111', B'1110');
jaccard_distance
------------------
0.25
(1 row)
SELECT jaccard_distance(B'1111', B'1100');
jaccard_distance
------------------
0.5
(1 row)
SELECT jaccard_distance(B'1111', B'1000');
jaccard_distance
------------------
0.75
(1 row)
SELECT jaccard_distance(B'1111', B'0000');
jaccard_distance
------------------
1
(1 row)
SELECT jaccard_distance(B'1100', B'1000');
jaccard_distance
------------------
0.5
(1 row)
SELECT jaccard_distance(B'1111', B'000');
ERROR: different bit lengths 4 and 3

View File

@@ -208,18 +208,6 @@ SELECT l1_distance('[3e38]', '[-3e38]');
Infinity Infinity
(1 row) (1 row)
SELECT quantize_binary('[1,0,-1]');
quantize_binary
-----------------
100
(1 row)
SELECT quantize_binary('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]');
quantize_binary
-----------------
01001110101
(1 row)
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v; SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
avg avg
----------- -----------

View File

@@ -1,21 +0,0 @@
SET enable_seqscan = off;
CREATE TABLE t (val bit(3));
INSERT INTO t (val) VALUES (B'000'), (B'100'), (B'111'), (NULL);
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
INSERT INTO t (val) VALUES (B'110');
SELECT * FROM t ORDER BY val <~> B'111';
val
-----
111
110
100
000
(4 rows)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <~> (SELECT NULL::bit)) t2;
count
-------
4
(1 row)
DROP TABLE t;

View File

@@ -1,21 +0,0 @@
SET enable_seqscan = off;
CREATE TABLE t (val bit(4));
INSERT INTO t (val) VALUES (B'0000'), (B'1100'), (B'1111'), (NULL);
CREATE INDEX ON t USING hnsw (val bit_jaccard_ops);
INSERT INTO t (val) VALUES (B'1110');
SELECT * FROM t ORDER BY val <%> B'1111';
val
------
1111
1110
1100
0000
(4 rows)
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <%> (SELECT NULL::bit)) t2;
count
-------
4
(1 row)
DROP TABLE t;

View File

@@ -1,13 +0,0 @@
SELECT hamming_distance(B'111', B'111');
SELECT hamming_distance(B'111', B'110');
SELECT hamming_distance(B'111', B'100');
SELECT hamming_distance(B'111', B'000');
SELECT hamming_distance(B'111', B'00');
SELECT jaccard_distance(B'1111', B'1111');
SELECT jaccard_distance(B'1111', B'1110');
SELECT jaccard_distance(B'1111', B'1100');
SELECT jaccard_distance(B'1111', B'1000');
SELECT jaccard_distance(B'1111', B'0000');
SELECT jaccard_distance(B'1100', B'1000');
SELECT jaccard_distance(B'1111', B'000');

View File

@@ -48,9 +48,6 @@ SELECT l1_distance('[0,0]', '[0,1]');
SELECT l1_distance('[1,2]', '[3]'); SELECT l1_distance('[1,2]', '[3]');
SELECT l1_distance('[3e38]', '[-3e38]'); SELECT l1_distance('[3e38]', '[-3e38]');
SELECT quantize_binary('[1,0,-1]');
SELECT quantize_binary('[0,0.1,-0.2,-0.3,0.4,0.5,0.6,-0.7,0.8,-0.9,1]');
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v; SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]']) v;
SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v; SELECT avg(v) FROM unnest(ARRAY['[1,2,3]'::vector, '[3,5,7]', NULL]) v;
SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v; SELECT avg(v) FROM unnest(ARRAY[]::vector[]) v;

View File

@@ -1,12 +0,0 @@
SET enable_seqscan = off;
CREATE TABLE t (val bit(3));
INSERT INTO t (val) VALUES (B'000'), (B'100'), (B'111'), (NULL);
CREATE INDEX ON t USING hnsw (val bit_hamming_ops);
INSERT INTO t (val) VALUES (B'110');
SELECT * FROM t ORDER BY val <~> B'111';
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <~> (SELECT NULL::bit)) t2;
DROP TABLE t;

View File

@@ -1,12 +0,0 @@
SET enable_seqscan = off;
CREATE TABLE t (val bit(4));
INSERT INTO t (val) VALUES (B'0000'), (B'1100'), (B'1111'), (NULL);
CREATE INDEX ON t USING hnsw (val bit_jaccard_ops);
INSERT INTO t (val) VALUES (B'1110');
SELECT * FROM t ORDER BY val <%> B'1111';
SELECT COUNT(*) FROM (SELECT * FROM t ORDER BY val <%> (SELECT NULL::bit)) t2;
DROP TABLE t;

View File

@@ -1,137 +0,0 @@
use strict;
use warnings;
use PostgresNode;
use TestLib;
use Test::More;
my $node;
my @queries = ();
my @expected;
my $limit = 20;
my $dim = 52;
my $max = 2**$dim;
sub test_recall
{
my ($min, $operator) = @_;
my $correct = 0;
my $total = 0;
my $explain = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.ef_search = 100;
EXPLAIN ANALYZE SELECT i FROM tst ORDER BY v $operator $queries[0] LIMIT $limit;
));
like($explain, qr/Index Scan/);
for my $i (0 .. $#queries)
{
my $actual = $node->safe_psql("postgres", qq(
SET enable_seqscan = off;
SET hnsw.ef_search = 100;
SELECT i FROM tst ORDER BY v $operator $queries[$i] LIMIT $limit;
));
my @actual_ids = split("\n", $actual);
my @expected_ids = split("\n", $expected[$i]);
my %expected_set = map { $_ => 1 } @expected_ids;
foreach (@actual_ids)
{
if (exists($expected_set{$_}))
{
$correct++;
}
}
$total += $limit;
}
cmp_ok($correct / $total, ">=", $min, $operator);
}
# Initialize node
$node = get_new_node('node');
$node->init;
$node->start;
# Create table
$node->safe_psql("postgres", "CREATE EXTENSION vector;");
$node->safe_psql("postgres", "CREATE TABLE tst (i int4, v bit($dim));");
$node->safe_psql("postgres",
"INSERT INTO tst SELECT i, (random() * $max)::bigint::bit($dim) FROM generate_series(1, 10000) i;"
);
# Generate queries
for (1 .. 20)
{
my $r = int(rand() * $max);
push(@queries, "${r}::bigint::bit($dim)");
}
# Check each index type
my @operators = ("<~>", "<\%>");
my @opclasses = ("bit_hamming_ops", "bit_jaccard_ops");
for my $i (0 .. $#operators)
{
my $operator = $operators[$i];
my $opclass = $opclasses[$i];
# Get exact results
@expected = ();
foreach (@queries)
{
# Handle ties
my $res = $node->safe_psql("postgres", qq(
WITH top AS (
SELECT v $operator $_ AS distance FROM tst ORDER BY v $operator $_ LIMIT $limit
)
SELECT i FROM tst WHERE (v $operator $_) <= (SELECT MAX(distance) FROM top)
));
push(@expected, $res);
}
# Build index serially
$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 = $operator eq "<\%>" ? 0.96 : 0.99;
test_recall($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();