From 47710bc9914f2a4fefef940ba53bbb5ecb60fc86 Mon Sep 17 00:00:00 2001 From: Arne Juul Date: Tue, 11 Feb 2020 10:10:10 +0000 Subject: * add "remove" benchmark * redo ops tracking * use std::aligned_alloc * more stats - measure reach This reverts commit 37fd87978ab1c3abfa840403e4e8f289d5ea4a20. --- eval/src/tests/ann/CMakeLists.txt | 10 + eval/src/tests/ann/remove-bm.cpp | 514 +++++++++++++++++++++++++++++++++ eval/src/tests/ann/sift_benchmark.cpp | 14 +- eval/src/tests/ann/xp-annoy-nns.cpp | 20 +- eval/src/tests/ann/xp-hnsw-wrap.cpp | 5 +- eval/src/tests/ann/xp-hnswlike-nns.cpp | 161 +++++++---- 6 files changed, 652 insertions(+), 72 deletions(-) create mode 100644 eval/src/tests/ann/remove-bm.cpp (limited to 'eval') diff --git a/eval/src/tests/ann/CMakeLists.txt b/eval/src/tests/ann/CMakeLists.txt index 05256d19f00..52b4d675d9c 100644 --- a/eval/src/tests/ann/CMakeLists.txt +++ b/eval/src/tests/ann/CMakeLists.txt @@ -9,3 +9,13 @@ vespa_add_executable(eval_sift_benchmark_app DEPENDS vespaeval ) + +vespa_add_executable(eval_remove_bm_app + SOURCES + remove-bm.cpp + xp-annoy-nns.cpp + xp-hnswlike-nns.cpp + xp-lsh-nns.cpp + DEPENDS + vespaeval +) diff --git a/eval/src/tests/ann/remove-bm.cpp b/eval/src/tests/ann/remove-bm.cpp new file mode 100644 index 00000000000..2da735f1929 --- /dev/null +++ b/eval/src/tests/ann/remove-bm.cpp @@ -0,0 +1,514 @@ +// Copyright 2020 Oath Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root. + +#include +#include +#include +#include +#include +#include +#include +#include + +#define NUM_DIMS 960 +#define NUM_DOCS 250000 +#define NUM_DOCS_REMOVE 50000 +#define EFFECTIVE_DOCS (NUM_DOCS - NUM_DOCS_REMOVE) +#define NUM_Q 1000 + +#include "doc_vector_access.h" +#include "nns.h" +#include "for-sift-hit.h" +#include "for-sift-top-k.h" + +std::vector bruteforceResults; +std::vector tmp_v(NUM_DIMS); + +struct PointVector { + float v[NUM_DIMS]; + using ConstArr = vespalib::ConstArrayRef; + operator ConstArr() const { return ConstArr(v, NUM_DIMS); } +}; + +static PointVector *aligned_alloc(size_t num) { + size_t num_bytes = num * sizeof(PointVector); + double mega_bytes = num_bytes / (1024.0*1024.0); + fprintf(stderr, "allocate %.2f MB of vectors\n", mega_bytes); + char *mem = (char *)malloc(num_bytes + 512); + mem += 512; + size_t val = (size_t)mem; + size_t unalign = val % 512; + mem -= unalign; + return (PointVector *)mem; +} + +static PointVector *generatedQueries = aligned_alloc(NUM_Q); +static PointVector *generatedDocs = aligned_alloc(NUM_DOCS); + +struct DocVectorAdapter : public DocVectorAccess +{ + vespalib::ConstArrayRef get(uint32_t docid) const override { + ASSERT_TRUE(docid < NUM_DOCS); + return generatedDocs[docid]; + } +}; + +double computeDistance(const PointVector &query, uint32_t docid) { + const PointVector &docvector = generatedDocs[docid]; + return l2distCalc.l2sq_dist(query, docvector, tmp_v); +} + +void read_queries(std::string fn) { + int fd = open(fn.c_str(), O_RDONLY); + ASSERT_TRUE(fd > 0); + int d; + size_t rv; + fprintf(stderr, "reading %u queries from %s\n", NUM_Q, fn.c_str()); + for (uint32_t qid = 0; qid < NUM_Q; ++qid) { + rv = read(fd, &d, 4); + ASSERT_EQUAL(rv, 4u); + ASSERT_EQUAL(d, NUM_DIMS); + rv = read(fd, &generatedQueries[qid].v, NUM_DIMS*sizeof(float)); + ASSERT_EQUAL(rv, sizeof(PointVector)); + } + close(fd); +} + +void read_docs(std::string fn) { + int fd = open(fn.c_str(), O_RDONLY); + ASSERT_TRUE(fd > 0); + int d; + size_t rv; + fprintf(stderr, "reading %u doc vectors from %s\n", NUM_DOCS, fn.c_str()); + for (uint32_t docid = 0; docid < NUM_DOCS; ++docid) { + rv = read(fd, &d, 4); + ASSERT_EQUAL(rv, 4u); + ASSERT_EQUAL(d, NUM_DIMS); + rv = read(fd, &generatedDocs[docid].v, NUM_DIMS*sizeof(float)); + ASSERT_EQUAL(rv, sizeof(PointVector)); + } + close(fd); +} + +using TimePoint = std::chrono::steady_clock::time_point; +using Duration = std::chrono::steady_clock::duration; + +double to_ms(Duration elapsed) { + std::chrono::duration ms(elapsed); + return ms.count(); +} + +void read_data(std::string dir) { + TimePoint bef = std::chrono::steady_clock::now(); + read_queries(dir + "/gist_query.fvecs"); + TimePoint aft = std::chrono::steady_clock::now(); + fprintf(stderr, "read queries: %.3f ms\n", to_ms(aft - bef)); + bef = std::chrono::steady_clock::now(); + read_docs(dir + "/gist_base.fvecs"); + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "read docs: %.3f ms\n", to_ms(aft - bef)); +} + + +struct BfHitComparator { + bool operator() (const Hit &lhs, const Hit& rhs) const { + if (lhs.distance < rhs.distance) return false; + if (lhs.distance > rhs.distance) return true; + return (lhs.docid > rhs.docid); + } +}; + +class BfHitHeap { +private: + size_t _size; + vespalib::PriorityQueue _priQ; +public: + explicit BfHitHeap(size_t maxSize) : _size(maxSize), _priQ() { + _priQ.reserve(maxSize); + } + ~BfHitHeap() {} + void maybe_use(const Hit &hit) { + if (_priQ.size() < _size) { + _priQ.push(hit); + } else if (hit.distance < _priQ.front().distance) { + _priQ.front() = hit; + _priQ.adjust(); + } + } + std::vector bestHits() { + std::vector result; + size_t i = _priQ.size(); + result.resize(i); + while (i-- > 0) { + result[i] = _priQ.front(); + _priQ.pop_front(); + } + return result; + } +}; + +TopK bruteforce_nns(const PointVector &query) { + TopK result; + BfHitHeap heap(result.K); + for (uint32_t docid = 0; docid < EFFECTIVE_DOCS; ++docid) { + const PointVector &docvector = generatedDocs[docid]; + double d = l2distCalc.l2sq_dist(query, docvector, tmp_v); + Hit h(docid, d); + heap.maybe_use(h); + } + std::vector best = heap.bestHits(); + for (size_t i = 0; i < result.K; ++i) { + result.hits[i] = best[i]; + } + return result; +} + +void verifyBF(uint32_t qid) { + const PointVector &query = generatedQueries[qid]; + TopK &result = bruteforceResults[qid]; + double min_distance = result.hits[0].distance; + std::vector all_c2; + for (uint32_t i = 0; i < EFFECTIVE_DOCS; ++i) { + double dist = computeDistance(query, i); + if (dist < min_distance) { + fprintf(stderr, "WARN dist %.9g < mindist %.9g\n", dist, min_distance); + } + EXPECT_FALSE(dist+0.000001 < min_distance); + if (min_distance > 0.0) all_c2.push_back(dist / min_distance); + } + if (all_c2.size() != EFFECTIVE_DOCS) return; + std::sort(all_c2.begin(), all_c2.end()); + for (uint32_t idx : { 1, 3, 10, 30, 100, 300, 1000, 3000, EFFECTIVE_DOCS/2, EFFECTIVE_DOCS-1}) { + fprintf(stderr, "c2-factor[%u] = %.3f\n", idx, all_c2[idx]); + } +} + +using NNS_API = NNS; + +#if 1 +TEST("require that HNSW via NNS api remove all works") { + DocVectorAdapter adapter; + std::unique_ptr nns = make_hnsw_nns(NUM_DIMS, adapter); + fprintf(stderr, "adding and removing all docs forward...\n"); + for (uint32_t i = 0; i < 1000; ++i) { + nns->addDoc(i); + } + for (uint32_t i = 0; i < 1000; ++i) { + nns->removeDoc(i); + } + fprintf(stderr, "adding and removing all docs reverse...\n"); + for (uint32_t i = 1000; i < 2000; ++i) { + nns->addDoc(i); + } + for (uint32_t i = 2000; i-- > 1000; ) { + nns->removeDoc(i); + } +} +#endif + +TEST("require that brute force works") { + TimePoint bef = std::chrono::steady_clock::now(); + fprintf(stderr, "generating %u brute force results\n", NUM_Q); + bruteforceResults.reserve(NUM_Q); + for (uint32_t cnt = 0; cnt < NUM_Q; ++cnt) { + const PointVector &query = generatedQueries[cnt]; + bruteforceResults.emplace_back(bruteforce_nns(query)); + } + TimePoint aft = std::chrono::steady_clock::now(); + fprintf(stderr, "timing for brute force: %.3f ms = %.3f ms per query\n", + to_ms(aft - bef), to_ms(aft - bef)/NUM_Q); + for (int cnt = 0; cnt < NUM_Q; cnt = (cnt+1)*2) { + verifyBF(cnt); + } +} + +bool reach_with_nns_1(NNS_API &nns, uint32_t docid) { + const PointVector &qv = generatedDocs[docid]; + vespalib::ConstArrayRef query(qv.v, NUM_DIMS); + auto rv = nns.topK(1, query, 1); + if (rv.size() != 1) { + fprintf(stderr, "Result/A from query for %u is %zu hits\n", docid, rv.size()); + return false; + } + if (rv[0].docid != docid) { + if (rv[0].sq.distance != 0.0) + fprintf(stderr, "Expected/A to find %u but got %u with sq distance %.3f\n", + docid, rv[0].docid, rv[0].sq.distance); + } + return (rv[0].docid == docid || rv[0].sq.distance == 0.0); +} + +bool reach_with_nns_100(NNS_API &nns, uint32_t docid) { + const PointVector &qv = generatedDocs[docid]; + vespalib::ConstArrayRef query(qv.v, NUM_DIMS); + auto rv = nns.topK(10, query, 100); + if (rv.size() != 10) { + fprintf(stderr, "Result/B from query for %u is %zu hits\n", docid, rv.size()); + } + if (rv[0].docid != docid) { + if (rv[0].sq.distance != 0.0) + fprintf(stderr, "Expected/B to find %u but got %u with sq distance %.3f\n", + docid, rv[0].docid, rv[0].sq.distance); + } + return (rv[0].docid == docid || rv[0].sq.distance == 0.0); +} + +bool reach_with_nns_1k(NNS_API &nns, uint32_t docid) { + const PointVector &qv = generatedDocs[docid]; + vespalib::ConstArrayRef query(qv.v, NUM_DIMS); + auto rv = nns.topK(10, query, 1000); + if (rv.size() != 10) { + fprintf(stderr, "Result/C from query for %u is %zu hits\n", docid, rv.size()); + } + if (rv[0].docid != docid) { + if (rv[0].sq.distance != 0.0) + fprintf(stderr, "Expected/C to find %u but got %u with sq distance %.3f\n", + docid, rv[0].docid, rv[0].sq.distance); + } + return (rv[0].docid == docid || rv[0].sq.distance == 0.0); +} + +TopK find_with_nns(uint32_t sk, NNS_API &nns, uint32_t qid) { + TopK result; + const PointVector &qv = generatedQueries[qid]; + vespalib::ConstArrayRef query(qv.v, NUM_DIMS); + auto rv = nns.topK(result.K, query, sk); + for (size_t i = 0; i < result.K; ++i) { + result.hits[i] = Hit(rv[i].docid, rv[i].sq.distance); + } + return result; +} + +void verify_nns_quality(uint32_t sk, NNS_API &nns, uint32_t qid) { + TopK perfect = bruteforceResults[qid]; + TopK result = find_with_nns(sk, nns, qid); + int recall = perfect.recall(result); + EXPECT_TRUE(recall > 40); + double sum_error = 0.0; + double c_factor = 1.0; + for (size_t i = 0; i < result.K; ++i) { + double factor = (result.hits[i].distance / perfect.hits[i].distance); + if (factor < 0.99 || factor > 25) { + fprintf(stderr, "hit[%zu] got distance %.3f, expected %.3f\n", + i, result.hits[i].distance, perfect.hits[i].distance); + } + sum_error += factor; + c_factor = std::max(c_factor, factor); + } + EXPECT_TRUE(c_factor < 1.5); + fprintf(stderr, "quality sk=%u: query %u: recall %d c2-factor %.3f avg c2: %.3f\n", + sk, qid, recall, c_factor, sum_error / result.K); +} + +void timing_nns(const char *name, NNS_API &nns, std::vector sk_list) { + for (uint32_t search_k : sk_list) { + TimePoint bef = std::chrono::steady_clock::now(); + for (int cnt = 0; cnt < NUM_Q; ++cnt) { + find_with_nns(search_k, nns, cnt); + } + TimePoint aft = std::chrono::steady_clock::now(); + fprintf(stderr, "timing for %s search_k=%u: %.3f ms = %.3f ms/q\n", + name, search_k, to_ms(aft - bef), to_ms(aft - bef)/NUM_Q); + } +} + +void quality_nns(NNS_API &nns, std::vector sk_list) { + for (uint32_t search_k : sk_list) { + for (int cnt = 0; cnt < NUM_Q; ++cnt) { + verify_nns_quality(search_k, nns, cnt); + } + } + uint32_t reached = 0; + for (uint32_t i = 0; i < 20000; ++i) { + if (reach_with_nns_1(nns, i)) ++reached; + } + fprintf(stderr, "Could reach %u of 20000 first documents with k=1\n", reached); + reached = 0; + for (uint32_t i = 0; i < 20000; ++i) { + if (reach_with_nns_100(nns, i)) ++reached; + } + fprintf(stderr, "Could reach %u of 20000 first documents with k=100\n", reached); + reached = 0; + for (uint32_t i = 0; i < 20000; ++i) { + if (reach_with_nns_1k(nns, i)) ++reached; + } + fprintf(stderr, "Could reach %u of 20000 first documents with k=1000\n", reached); +} + +void benchmark_nns(const char *name, NNS_API &nns, std::vector sk_list) { + fprintf(stderr, "trying %s indexing...\n", name); + +#if 0 + TimePoint bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < NUM_DOCS_REMOVE; ++i) { + nns.addDoc(EFFECTIVE_DOCS + i); + } + for (uint32_t i = 0; i < EFFECTIVE_DOCS - NUM_DOCS_REMOVE; ++i) { + nns.addDoc(i); + } + for (uint32_t i = 0; i < NUM_DOCS_REMOVE; ++i) { + nns.removeDoc(EFFECTIVE_DOCS + i); + nns.addDoc(EFFECTIVE_DOCS - NUM_DOCS_REMOVE + i); + } + TimePoint aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index with %u docs: %.3f ms\n", name, EFFECTIVE_DOCS, to_ms(aft - bef)); + + timing_nns(name, nns, sk_list); + fprintf(stderr, "Quality for %s realistic build with %u documents:\n", name, EFFECTIVE_DOCS); + quality_nns(nns, sk_list); +#endif + +#if 1 + TimePoint bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < EFFECTIVE_DOCS; ++i) { + nns.addDoc(i); + } + TimePoint aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index with %u docs: %.3f ms\n", name, EFFECTIVE_DOCS, to_ms(aft - bef)); + + timing_nns(name, nns, sk_list); + fprintf(stderr, "Quality for %s clean build with %u documents:\n", name, EFFECTIVE_DOCS); + quality_nns(nns, sk_list); + + bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < NUM_DOCS_REMOVE; ++i) { + nns.addDoc(EFFECTIVE_DOCS + i); + } + for (uint32_t i = 0; i < NUM_DOCS_REMOVE; ++i) { + nns.removeDoc(EFFECTIVE_DOCS + i); + } + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index add then remove %u docs: %.3f ms\n", + name, NUM_DOCS_REMOVE, to_ms(aft - bef)); + + timing_nns(name, nns, sk_list); + fprintf(stderr, "Quality for %s remove-damaged build with %u documents:\n", name, EFFECTIVE_DOCS); + quality_nns(nns, sk_list); +#endif + +#if 0 + TimePoint bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < EFFECTIVE_DOCS; ++i) { + nns.addDoc(i); + } + TimePoint aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index with %u docs: %.3f ms\n", name, EFFECTIVE_DOCS, to_ms(aft - bef)); + + timing_nns(name, nns, sk_list); + fprintf(stderr, "Quality for %s clean build with %u documents:\n", name, EFFECTIVE_DOCS); + quality_nns(nns, sk_list); + + bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < EFFECTIVE_DOCS; ++i) { + nns.removeDoc(i); + } + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index removed %u docs: %.3f ms\n", name, EFFECTIVE_DOCS, to_ms(aft - bef)); + + const uint32_t addFirst = NUM_DOCS - (NUM_DOCS_REMOVE * 3); + const uint32_t addSecond = NUM_DOCS - (NUM_DOCS_REMOVE * 2); + + bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < addFirst; ++i) { + nns.addDoc(i); + } + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index with %u docs: %.3f ms\n", name, addFirst, to_ms(aft - bef)); + + bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < NUM_DOCS_REMOVE; ++i) { + nns.addDoc(EFFECTIVE_DOCS + i); + nns.addDoc(addFirst + i); + } + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index added %u docs: %.3f ms\n", + name, 2 * NUM_DOCS_REMOVE, to_ms(aft - bef)); + + bef = std::chrono::steady_clock::now(); + for (uint32_t i = 0; i < NUM_DOCS_REMOVE; ++i) { + nns.removeDoc(EFFECTIVE_DOCS + i); + nns.addDoc(addSecond + i); + } + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index added %u and removed %u docs: %.3f ms\n", + name, NUM_DOCS_REMOVE, NUM_DOCS_REMOVE, to_ms(aft - bef)); + + timing_nns(name, nns, sk_list); + fprintf(stderr, "Quality for %s with %u documents some churn:\n", name, EFFECTIVE_DOCS); + quality_nns(nns, sk_list); + +#endif + +#if 0 + bef = std::chrono::steady_clock::now(); + fprintf(stderr, "removing and adding %u documents...\n", EFFECTIVE_DOCS); + for (uint32_t i = 0; i < EFFECTIVE_DOCS; ++i) { + nns.removeDoc(i); + nns.addDoc(i); + } + aft = std::chrono::steady_clock::now(); + fprintf(stderr, "build %s index rem/add %u docs: %.3f ms\n", + name, EFFECTIVE_DOCS, to_ms(aft - bef)); + + timing_nns(name, nns, sk_list); + fprintf(stderr, "Quality for %s with %u documents full churn:\n", name, EFFECTIVE_DOCS); + quality_nns(nns, sk_list); +#endif +} + +#if 0 +TEST("require that Locality Sensitive Hashing mostly works") { + DocVectorAdapter adapter; + std::unique_ptr nns = make_rplsh_nns(NUM_DIMS, adapter); + benchmark_nns("RPLSH", *nns, { 200, 1000 }); +} +#endif + +#if 0 +TEST("require that Annoy via NNS api mostly works") { + DocVectorAdapter adapter; + std::unique_ptr nns = make_annoy_nns(NUM_DIMS, adapter); + benchmark_nns("Annoy", *nns, { 8000, 10000 }); +} +#endif + +#if 1 +TEST("require that HNSW via NNS api mostly works") { + DocVectorAdapter adapter; + std::unique_ptr nns = make_hnsw_nns(NUM_DIMS, adapter); + benchmark_nns("HNSW-like", *nns, { 100, 150, 200 }); +} +#endif + +#if 0 +TEST("require that HNSW wrapped api mostly works") { + DocVectorAdapter adapter; + std::unique_ptr nns = make_hnsw_wrap(NUM_DIMS, adapter); + benchmark_nns("HNSW-wrap", *nns, { 100, 150, 200 }); +} +#endif + +/** + * Before running the benchmark the ANN_GIST1M data set must be downloaded and extracted: + * wget ftp://ftp.irisa.fr/local/texmex/corpus/gist.tar.gz + * tar -xf gist.tar.gz + * + * The benchmark program will load the data set from $HOME/gist if no directory is specified. + * + * More information about the dataset is found here: http://corpus-texmex.irisa.fr/. + */ +int main(int argc, char **argv) { + TEST_MASTER.init(__FILE__); + std::string gist_dir = "."; + if (argc > 1) { + gist_dir = argv[1]; + } else { + char *home = getenv("HOME"); + if (home) { + gist_dir = home; + gist_dir += "/gist"; + } + } + read_data(gist_dir); + TEST_RUN_ALL(); + return (TEST_MASTER.fini() ? 0 : 1); +} diff --git a/eval/src/tests/ann/sift_benchmark.cpp b/eval/src/tests/ann/sift_benchmark.cpp index f3570eb9247..7c060d86371 100644 --- a/eval/src/tests/ann/sift_benchmark.cpp +++ b/eval/src/tests/ann/sift_benchmark.cpp @@ -8,6 +8,7 @@ #include #include #include +#include #define NUM_DIMS 128 #define NUM_DOCS 1000000 @@ -28,11 +29,12 @@ struct PointVector { }; static PointVector *aligned_alloc(size_t num) { - char *mem = (char *)malloc(num * sizeof(PointVector) + 512); - mem += 512; - size_t val = (size_t)mem; - size_t unalign = val % 512; - mem -= unalign; + size_t sz = num * sizeof(PointVector); + size_t align = 512; + while ((sz % align) != 0) { align /= 2; } + double mega_bytes = sz / (1024.0*1024.0); + fprintf(stderr, "allocate %.2f MB of vectors (align %zu)\n", mega_bytes, align); + void *mem = std::aligned_alloc(align, sz); return reinterpret_cast(mem); } @@ -169,7 +171,7 @@ void verifyBF(uint32_t qid) { fprintf(stderr, "WARN dist %.9g < mindist %.9g\n", dist, min_distance); } EXPECT_FALSE(dist+0.000001 < min_distance); - if (qid == 6) all_c2.push_back(dist / min_distance); + if (min_distance > 0) all_c2.push_back(dist / min_distance); } if (all_c2.size() != NUM_DOCS) return; std::sort(all_c2.begin(), all_c2.end()); diff --git a/eval/src/tests/ann/xp-annoy-nns.cpp b/eval/src/tests/ann/xp-annoy-nns.cpp index c34f9f6eb36..f022aae5974 100644 --- a/eval/src/tests/ann/xp-annoy-nns.cpp +++ b/eval/src/tests/ann/xp-annoy-nns.cpp @@ -12,11 +12,11 @@ using V = vespalib::ConstArrayRef; class AnnoyLikeNns; struct Node; -static uint64_t plane_dist_cnt = 0; -static uint64_t w_cen_dist_cnt = 0; -static uint64_t leaf_split_cnt = 0; -static uint64_t find_top_k_cnt = 0; -static uint64_t find_cand_cnt = 0; +static size_t plane_dist_cnt = 0; +static size_t w_cen_dist_cnt = 0; +static size_t leaf_split_cnt = 0; +static size_t find_top_k_cnt = 0; +static size_t find_cand_cnt = 0; using QueueNode = std::pair; using NodeQueue = std::priority_queue; @@ -390,11 +390,11 @@ AnnoyLikeNns::topK(uint32_t k, Vector vector, uint32_t search_k) void AnnoyLikeNns::dumpStats() { fprintf(stderr, "stats for AnnoyLikeNns:\n"); - fprintf(stderr, "planeDistance() calls: %" PRIu64 "\n", plane_dist_cnt); - fprintf(stderr, "weightedDistance() calls: %" PRIu64 "\n", w_cen_dist_cnt); - fprintf(stderr, "leaf split() calls: %" PRIu64 "\n", leaf_split_cnt); - fprintf(stderr, "topK() calls: %" PRIu64 "\n", find_top_k_cnt); - fprintf(stderr, "findCandidates() calls: %" PRIu64 "\n", find_cand_cnt); + fprintf(stderr, "planeDistance() calls: %zu\n", plane_dist_cnt); + fprintf(stderr, "weightedDistance() calls: %zu\n", w_cen_dist_cnt); + fprintf(stderr, "leaf split() calls: %zu\n", leaf_split_cnt); + fprintf(stderr, "topK() calls: %zu\n", find_top_k_cnt); + fprintf(stderr, "findCandidates() calls: %zu\n", find_cand_cnt); std::vector depths; _roots[0]->stats(depths); std::vector counts; diff --git a/eval/src/tests/ann/xp-hnsw-wrap.cpp b/eval/src/tests/ann/xp-hnsw-wrap.cpp index 33895b2bd7c..3eb01142dcd 100644 --- a/eval/src/tests/ann/xp-hnsw-wrap.cpp +++ b/eval/src/tests/ann/xp-hnsw-wrap.cpp @@ -15,7 +15,7 @@ public: HnswWrapNns(uint32_t numDims, const DocVectorAccess &dva) : NNS(numDims, dva), _l2space(numDims), - _hnsw(&_l2space, 1000000, 16, 200) + _hnsw(&_l2space, 2500000, 16, 200) { } @@ -32,7 +32,8 @@ public: std::vector topK(uint32_t k, Vector vector, uint32_t search_k) override { std::vector reversed; - auto priQ = _hnsw.searchKnn(vector.cbegin(), std::max(k, search_k)); + _hnsw.setEf(search_k); + auto priQ = _hnsw.searchKnn(vector.cbegin(), k); while (! priQ.empty()) { auto pair = priQ.top(); reversed.emplace_back(pair.second, SqDist(pair.first)); diff --git a/eval/src/tests/ann/xp-hnswlike-nns.cpp b/eval/src/tests/ann/xp-hnswlike-nns.cpp index 72b3fdb21f9..5cdbdd8efa3 100644 --- a/eval/src/tests/ann/xp-hnswlike-nns.cpp +++ b/eval/src/tests/ann/xp-hnswlike-nns.cpp @@ -7,13 +7,31 @@ #include "std-random.h" #include "nns.h" -static uint64_t distcalls_simple; -static uint64_t distcalls_search_layer; -static uint64_t distcalls_other; -static uint64_t distcalls_heuristic; -static uint64_t distcalls_shrink; -static uint64_t distcalls_refill; -static uint64_t refill_needed_calls; +/* + Todo: + + measure effect of: + 1) removing leftover backlinks during "shrink" operation + 2) refilling to low-watermark after 1) happens + 3) refilling to mid-watermark after 1) happens + 4) adding then removing 20% extra documents + 5) removing 20% first-added documents + 6) removing first-added documents while inserting new ones + + 7) auto-tune search_k to ensure >= 50% recall on 1000 Q with k=100 + 8) auto-tune search_k to ensure avg 90% recall on 1000 Q with k=100 + 9) auto-tune search_k to ensure >= 90% reachability of 10000 docids + + 10) timings for SIFT, GIST, and DEEP data (100k, 200k, 300k, 500k, 700k, 1000k) + */ + +static size_t distcalls_simple; +static size_t distcalls_search_layer; +static size_t distcalls_other; +static size_t distcalls_heuristic; +static size_t distcalls_shrink; +static size_t distcalls_refill; +static size_t refill_needed_calls; struct LinkList : std::vector { @@ -31,6 +49,7 @@ struct LinkList : std::vector } } fprintf(stderr, "BAD missing link to remove: %u\n", id); + abort(); } }; @@ -130,6 +149,7 @@ private: double _levelMultiplier; RndGen _rndGen; VisitedSetPool _visitedSetPool; + size_t _ops_counter; double distance(Vector v, uint32_t id) const; @@ -144,6 +164,7 @@ private: return (int) r; } + uint32_t count_reachable() const; void dumpStats() const; public: @@ -155,9 +176,9 @@ public: _M(16), _efConstruction(200), _levelMultiplier(1.0 / log(1.0 * _M)), - _rndGen() + _rndGen(), + _ops_counter(0) { - _nodes.reserve(1234567); } ~HnswLikeNns() { dumpStats(); } @@ -217,6 +238,7 @@ public: if (_entryLevel < 0) { _entryId = docid; _entryLevel = level; + track_ops(); return; } int searchLevel = _entryLevel; @@ -241,18 +263,23 @@ public: _entryLevel = level; _entryId = docid; } - if (_nodes.size() % 10000 == 0) { - double div = _nodes.size(); - fprintf(stderr, "added docs: %d\n", (int)div); - fprintf(stderr, "distance calls for layer: %" PRIu64 " is %.3f per doc\n", distcalls_search_layer, distcalls_search_layer/ div); - fprintf(stderr, "distance calls for heuristic: %" PRIu64 " is %.3f per doc\n", distcalls_heuristic, distcalls_heuristic / div); - fprintf(stderr, "distance calls for simple: %" PRIu64 " is %.3f per doc\n", distcalls_simple, distcalls_simple / div); - fprintf(stderr, "distance calls for shrink: %" PRIu64 " is %.3f per doc\n", distcalls_shrink, distcalls_shrink / div); - fprintf(stderr, "distance calls for refill: %" PRIu64 " is %.3f per doc\n", distcalls_refill, distcalls_refill / div); - fprintf(stderr, "distance calls for other: %" PRIu64 " is %.3f per doc\n", distcalls_other, distcalls_other / div); - fprintf(stderr, "refill needed calls: %" PRIu64 " is %.3f per doc\n", refill_needed_calls, refill_needed_calls / div); + track_ops(); + } + + void track_ops() { + _ops_counter++; + if ((_ops_counter % 10000) == 0) { + double div = _ops_counter; + fprintf(stderr, "add / remove ops: %zu\n", _ops_counter); + fprintf(stderr, "distance calls for layer: %zu is %.3f per op\n", distcalls_search_layer, distcalls_search_layer/ div); + fprintf(stderr, "distance calls for heuristic: %zu is %.3f per op\n", distcalls_heuristic, distcalls_heuristic / div); + fprintf(stderr, "distance calls for simple: %zu is %.3f per op\n", distcalls_simple, distcalls_simple / div); + fprintf(stderr, "distance calls for shrink: %zu is %.3f per op\n", distcalls_shrink, distcalls_shrink / div); + fprintf(stderr, "distance calls for refill: %zu is %.3f per op\n", distcalls_refill, distcalls_refill / div); + fprintf(stderr, "distance calls for other: %zu is %.3f per op\n", distcalls_other, distcalls_other / div); + fprintf(stderr, "refill needed calls: %zu is %.3f per op\n", refill_needed_calls, refill_needed_calls / div); } - } + } void remove_link_from(uint32_t from_id, uint32_t remove_id, uint32_t level) { LinkList &links = getLinkList(from_id, level); @@ -267,9 +294,10 @@ public: if (repl_id == my_id) continue; if (my_links.has_link_to(repl_id)) continue; LinkList &other_links = getLinkList(repl_id, level); - if (other_links.size() >= _M) continue; + if (other_links.size() + 1 >= _M) continue; other_links.push_back(my_id); my_links.push_back(repl_id); + if (my_links.size() >= _M) return; } } } @@ -299,14 +327,17 @@ public: Node &node = _nodes[docid]; bool need_new_entrypoint = (docid == _entryId); for (int level = node._links.size(); level-- > 0; ) { - const LinkList &my_links = node._links[level]; + LinkList my_links; + my_links.swap(node._links[level]); for (uint32_t n_id : my_links) { if (need_new_entrypoint) { _entryId = n_id; _entryLevel = level; - need_new_entrypoint = false; + need_new_entrypoint = false; } remove_link_from(n_id, docid, level); + } + for (uint32_t n_id : my_links) { refill_ifneeded(n_id, my_links, level); } } @@ -322,6 +353,7 @@ public: } } } + track_ops(); } std::vector topK(uint32_t k, Vector vector, uint32_t search_k) override { @@ -331,12 +363,12 @@ public: ++distcalls_other; HnswHit entryPoint(_entryId, SqDist(entryDist)); int searchLevel = _entryLevel; + FurthestPriQ w; + w.push(entryPoint); while (searchLevel > 0) { - entryPoint = search_layer_simple(vector, entryPoint, searchLevel); + search_layer(vector, w, std::min(k, search_k), searchLevel); --searchLevel; } - FurthestPriQ w; - w.push(entryPoint); search_layer(vector, w, std::max(k, search_k), 0); while (w.size() > k) { w.pop(); @@ -490,66 +522,87 @@ HnswLikeNns::connect_new_node(uint32_t id, const LinkList &neighbors, uint32_t l } } +uint32_t +HnswLikeNns::count_reachable() const { + VisitedSet visited(_nodes.size()); + int level = _entryLevel; + LinkList curList; + curList.push_back(_entryId); + visited.mark(_entryId); + uint32_t idx = 0; + while (level >= 0) { + while (idx < curList.size()) { + uint32_t id = curList[idx++]; + const LinkList &links = getLinkList(id, level); + for (uint32_t n_id : links) { + if (visited.isMarked(n_id)) continue; + visited.mark(n_id); + curList.push_back(n_id); + } + } + --level; + idx = 0; + } + return curList.size(); +} + void HnswLikeNns::dumpStats() const { - std::vector inLinkCounters; - inLinkCounters.resize(_nodes.size()); - std::vector outLinkCounters; - outLinkCounters.resize(_nodes.size()); std::vector levelCounts; levelCounts.resize(_entryLevel + 2); std::vector outLinkHist; outLinkHist.resize(2 * _M + 2); + uint32_t symmetrics = 0; + uint32_t level1links = 0; + uint32_t both_l_links = 0; fprintf(stderr, "stats for HnswLikeNns with %zu nodes, entry level = %d, entry id = %u\n", _nodes.size(), _entryLevel, _entryId); + for (uint32_t id = 0; id < _nodes.size(); ++id) { const auto &node = _nodes[id]; uint32_t levels = node._links.size(); levelCounts[levels]++; if (levels < 1) { - outLinkCounters[id] = 0; outLinkHist[0]++; continue; } const LinkList &link_list = getLinkList(id, 0); uint32_t numlinks = link_list.size(); - outLinkCounters[id] = numlinks; outLinkHist[numlinks]++; - if (numlinks < 2) { + if (numlinks < 1) { fprintf(stderr, "node with %u links: id %u\n", numlinks, id); - for (uint32_t n_id : link_list) { - const LinkList &neigh_list = getLinkList(n_id, 0); - fprintf(stderr, "neighbor id %u has %zu links\n", n_id, neigh_list.size()); - if (! neigh_list.has_link_to(id)) { - fprintf(stderr, "BAD neighbor %u is missing backlink\n", n_id); - } - } } + bool all_sym = true; for (uint32_t n_id : link_list) { - inLinkCounters[n_id]++; + const LinkList &neigh_list = getLinkList(n_id, 0); + if (! neigh_list.has_link_to(id)) { + fprintf(stderr, "BAD: %u has link to neighbor %u, but backlink is missing\n", id, n_id); + all_sym = false; + } + } + if (all_sym) ++symmetrics; + if (levels < 2) continue; + const LinkList &link_list_1 = getLinkList(id, 1); + for (uint32_t n_id : link_list_1) { + ++level1links; + if (link_list.has_link_to(n_id)) ++both_l_links; } } for (uint32_t l = 0; l < levelCounts.size(); ++l) { fprintf(stderr, "Nodes on %u levels: %u\n", l, levelCounts[l]); } + fprintf(stderr, "reachable nodes %u / %zu\n", + count_reachable(), _nodes.size() - levelCounts[0]); + fprintf(stderr, "level 1 links overlapping on l0: %u / total: %u\n", + both_l_links, level1links); for (uint32_t l = 0; l < outLinkHist.size(); ++l) { - fprintf(stderr, "Nodes with %u outward links on L0: %u\n", l, outLinkHist[l]); - } - uint32_t symmetrics = 0; - std::vector inLinkHist; - for (uint32_t id = 0; id < _nodes.size(); ++id) { - uint32_t cnt = inLinkCounters[id]; - while (cnt >= inLinkHist.size()) inLinkHist.push_back(0); - inLinkHist[cnt]++; - if (cnt == outLinkCounters[id]) ++symmetrics; - } - for (uint32_t l = 0; l < inLinkHist.size(); ++l) { - fprintf(stderr, "Nodes with %u inward links on L0: %u\n", l, inLinkHist[l]); + if (outLinkHist[l] != 0) { + fprintf(stderr, "Nodes with %u outward links on L0: %u\n", l, outLinkHist[l]); + } } fprintf(stderr, "Symmetric in-out nodes: %u\n", symmetrics); } - std::unique_ptr> make_hnsw_nns(uint32_t numDims, const DocVectorAccess &dva) { -- cgit v1.2.3