Squashed 'vendor/ruvector/' content from commit b64c2172

git-subtree-dir: vendor/ruvector
git-subtree-split: b64c21726f2bb37286d9ee36a7869fef60cc6900
This commit is contained in:
ruv
2026-02-28 14:39:40 -05:00
commit d803bfe2b1
7854 changed files with 3522914 additions and 0 deletions

View File

@@ -0,0 +1,102 @@
#!/bin/bash
# Download ANN benchmark datasets (SIFT1M, GIST1M, Deep1M)
set -e
DATASETS_DIR="datasets"
mkdir -p "$DATASETS_DIR"
echo "╔════════════════════════════════════════╗"
echo "║ ANN Benchmark Dataset Downloader ║"
echo "╚════════════════════════════════════════╝"
echo ""
# Function to download and extract dataset
download_dataset() {
local name=$1
local url=$2
local file=$(basename "$url")
echo "Downloading $name..."
if [ -f "$DATASETS_DIR/$file" ]; then
echo " ✓ Already downloaded: $file"
else
wget -q --show-progress -O "$DATASETS_DIR/$file" "$url"
echo " ✓ Downloaded: $file"
fi
echo "Extracting $name..."
if [[ $file == *.tar.gz ]]; then
tar -xzf "$DATASETS_DIR/$file" -C "$DATASETS_DIR"
elif [[ $file == *.gz ]]; then
gunzip -k "$DATASETS_DIR/$file"
fi
echo " ✓ Extracted successfully"
echo ""
}
# SIFT1M Dataset (128D, 1M vectors)
# http://corpus-texmex.irisa.fr/
echo "1. SIFT1M Dataset (128 dimensions, 1M vectors)"
echo " Download from: http://corpus-texmex.irisa.fr/"
echo " Note: Direct download requires manual intervention due to terms of service"
echo " Please visit the website and download sift.tar.gz manually to datasets/"
echo ""
# GIST1M Dataset (960D, 1M vectors)
echo "2. GIST1M Dataset (960 dimensions, 1M vectors)"
echo " Download from: http://corpus-texmex.irisa.fr/"
echo " Note: Direct download requires manual intervention due to terms of service"
echo " Please visit the website and download gist.tar.gz manually to datasets/"
echo ""
# Deep1M Dataset (96D, 1M vectors)
echo "3. Deep1M Dataset (96 dimensions, 1M vectors)"
echo " Download from: http://sites.skoltech.ru/compvision/noimi/"
echo " Note: This dataset may require registration"
echo ""
# Alternative: Generate synthetic datasets
echo "═══════════════════════════════════════════════════════════════"
echo "ALTERNATIVE: Generate Synthetic Datasets"
echo "═══════════════════════════════════════════════════════════════"
echo ""
echo "If you prefer to use synthetic data for benchmarking, the"
echo "benchmark tools will automatically generate appropriate datasets."
echo ""
echo "To run with synthetic data:"
echo " cargo run --release --bin ann-benchmark -- --dataset synthetic"
echo ""
# Check for HDF5 support
echo "Checking dependencies..."
if command -v h5dump &> /dev/null; then
echo " ✓ HDF5 tools installed"
else
echo " ⚠ HDF5 tools not found. Install with:"
echo " Ubuntu/Debian: sudo apt-get install hdf5-tools"
echo " macOS: brew install hdf5"
echo " Note: HDF5 is optional for synthetic benchmarks"
fi
echo ""
echo "════════════════════════════════════════"
echo "Setup Instructions:"
echo "════════════════════════════════════════"
echo ""
echo "1. Manual Download (for real datasets):"
echo " - Visit http://corpus-texmex.irisa.fr/"
echo " - Download sift.tar.gz, gist.tar.gz"
echo " - Place in: $DATASETS_DIR/"
echo " - Extract: tar -xzf $DATASETS_DIR/sift.tar.gz -C $DATASETS_DIR/"
echo ""
echo "2. Synthetic Datasets (recommended for testing):"
echo " - No download required"
echo " - Generated automatically by benchmark tools"
echo " - Suitable for performance testing and profiling"
echo ""
echo "3. Run Benchmarks:"
echo " cd crates/ruvector-bench"
echo " cargo run --release --bin ann-benchmark"
echo ""
echo "✓ Setup guide complete!"

View File

@@ -0,0 +1,246 @@
#!/bin/bash
# Run complete Ruvector benchmark suite
set -e
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
BENCH_DIR="$(dirname "$SCRIPT_DIR")"
OUTPUT_DIR="${BENCH_DIR}/bench_results"
echo "╔════════════════════════════════════════╗"
echo "║ Ruvector Benchmark Suite Runner ║"
echo "╚════════════════════════════════════════╝"
echo ""
# Create output directory
mkdir -p "$OUTPUT_DIR"
# Parse arguments
QUICK_MODE=false
PROFILE=false
while [[ $# -gt 0 ]]; do
case $1 in
--quick)
QUICK_MODE=true
shift
;;
--profile)
PROFILE=true
shift
;;
*)
echo "Unknown option: $1"
echo "Usage: $0 [--quick] [--profile]"
exit 1
;;
esac
done
# Set benchmark parameters based on mode
if [ "$QUICK_MODE" = true ]; then
echo "Running in QUICK mode (reduced dataset sizes)..."
VECTORS=10000
QUERIES=500
else
echo "Running in FULL mode (standard dataset sizes)..."
VECTORS=100000
QUERIES=1000
fi
echo "Output directory: $OUTPUT_DIR"
echo ""
# Build benchmarks
echo "═══════════════════════════════════════════════════════════════"
echo "Building benchmark suite..."
echo "═══════════════════════════════════════════════════════════════"
cd "$BENCH_DIR"
cargo build --release
echo "✓ Build complete"
echo ""
# Run ANN Benchmarks
echo "═══════════════════════════════════════════════════════════════"
echo "1. ANN Benchmarks (SIFT/GIST/Deep1M compatibility)"
echo "═══════════════════════════════════════════════════════════════"
cargo run --release --bin ann-benchmark -- \
--dataset synthetic \
--num-vectors $VECTORS \
--queries $QUERIES \
--dimensions 128 \
--output "$OUTPUT_DIR"
echo ""
# Run AgenticDB Benchmarks
echo "═══════════════════════════════════════════════════════════════"
echo "2. AgenticDB Workload Benchmarks"
echo "═══════════════════════════════════════════════════════════════"
cargo run --release --bin agenticdb-benchmark -- \
--episodes $VECTORS \
--skills $(($VECTORS / 10)) \
--queries $QUERIES \
--output "$OUTPUT_DIR"
echo ""
# Run Latency Benchmarks
echo "═══════════════════════════════════════════════════════════════"
echo "3. Latency Profiling"
echo "═══════════════════════════════════════════════════════════════"
cargo run --release --bin latency-benchmark -- \
--num-vectors $(($VECTORS / 2)) \
--queries $QUERIES \
--dimensions 384 \
--threads "1,4,8" \
--output "$OUTPUT_DIR"
echo ""
# Run Memory Benchmarks
echo "═══════════════════════════════════════════════════════════════"
echo "4. Memory Profiling"
echo "═══════════════════════════════════════════════════════════════"
if [ "$QUICK_MODE" = true ]; then
SCALES="1000,10000"
else
SCALES="1000,10000,100000"
fi
cargo run --release --bin memory-benchmark -- \
--dimensions 384 \
--scales "$SCALES" \
--output "$OUTPUT_DIR"
echo ""
# Run Comparison Benchmarks
echo "═══════════════════════════════════════════════════════════════"
echo "5. Cross-System Comparison"
echo "═══════════════════════════════════════════════════════════════"
cargo run --release --bin comparison-benchmark -- \
--num-vectors $(($VECTORS / 2)) \
--queries $QUERIES \
--dimensions 384 \
--output "$OUTPUT_DIR"
echo ""
# Run Profiling (optional)
if [ "$PROFILE" = true ]; then
echo "═══════════════════════════════════════════════════════════════"
echo "6. Performance Profiling with Flamegraph"
echo "═══════════════════════════════════════════════════════════════"
cargo run --release --features profiling --bin profiling-benchmark -- \
--num-vectors $(($VECTORS / 2)) \
--queries $QUERIES \
--dimensions 384 \
--flamegraph \
--output "$OUTPUT_DIR/profiling"
echo ""
fi
# Generate summary report
echo "═══════════════════════════════════════════════════════════════"
echo "Generating Summary Report"
echo "═══════════════════════════════════════════════════════════════"
SUMMARY_FILE="$OUTPUT_DIR/SUMMARY.md"
cat > "$SUMMARY_FILE" << EOF
# Ruvector Benchmark Results Summary
**Generated:** $(date)
**Mode:** $([ "$QUICK_MODE" = true ] && echo "Quick" || echo "Full")
## Configuration
- Vectors: $VECTORS
- Queries: $QUERIES
- Profiling: $([ "$PROFILE" = true ] && echo "Enabled" || echo "Disabled")
## Results Location
All benchmark results are saved in: \`$OUTPUT_DIR\`
## Available Reports
### 1. ANN Benchmarks
- JSON: \`ann_benchmark.json\`
- CSV: \`ann_benchmark.csv\`
- Report: \`ann_benchmark.md\`
### 2. AgenticDB Workloads
- JSON: \`agenticdb_benchmark.json\`
- CSV: \`agenticdb_benchmark.csv\`
- Report: \`agenticdb_benchmark.md\`
### 3. Latency Profiling
- JSON: \`latency_benchmark.json\`
- CSV: \`latency_benchmark.csv\`
- Report: \`latency_benchmark.md\`
### 4. Memory Profiling
- JSON: \`memory_benchmark.json\`
- CSV: \`memory_benchmark.csv\`
- Report: \`memory_benchmark.md\`
### 5. System Comparison
- JSON: \`comparison_benchmark.json\`
- CSV: \`comparison_benchmark.csv\`
- Report: \`comparison_benchmark.md\`
EOF
if [ "$PROFILE" = true ]; then
cat >> "$SUMMARY_FILE" << EOF
### 6. Performance Profiling
- Flamegraph: \`profiling/flamegraph.svg\`
- Profile: \`profiling/profile.txt\`
EOF
fi
cat >> "$SUMMARY_FILE" << EOF
## Quick Analysis
To view individual benchmark reports, use:
\`\`\`bash
cat $OUTPUT_DIR/ann_benchmark.md
cat $OUTPUT_DIR/agenticdb_benchmark.md
cat $OUTPUT_DIR/latency_benchmark.md
cat $OUTPUT_DIR/memory_benchmark.md
cat $OUTPUT_DIR/comparison_benchmark.md
\`\`\`
To view CSV data for analysis:
\`\`\`bash
column -t -s, $OUTPUT_DIR/ann_benchmark.csv | less -S
\`\`\`
EOF
echo "✓ Summary report generated: $SUMMARY_FILE"
echo ""
echo "════════════════════════════════════════════════════════════════"
echo "✓ All benchmarks complete!"
echo "════════════════════════════════════════════════════════════════"
echo ""
echo "Results saved to: $OUTPUT_DIR"
echo "Summary report: $SUMMARY_FILE"
echo ""
echo "View results:"
echo " cat $SUMMARY_FILE"
echo " cat $OUTPUT_DIR/*.md"
echo ""
# Display quick stats if available
if [ -f "$OUTPUT_DIR/comparison_benchmark.csv" ]; then
echo "Quick Performance Summary:"
echo "─────────────────────────────────────────"
grep "ruvector_optimized" "$OUTPUT_DIR/comparison_benchmark.csv" | \
awk -F',' '{printf " Optimized QPS: %s\n Latency p50: %sms\n Latency p99: %sms\n", $7, $8, $10}'
echo ""
fi
echo "To run again:"
echo " ./scripts/run_all_benchmarks.sh # Full benchmarks"
echo " ./scripts/run_all_benchmarks.sh --quick # Quick mode"
echo " ./scripts/run_all_benchmarks.sh --profile # With profiling"