mleku
e0a95ca1cd
- Replaced all instances of p256k1signer with the new p8k.Signer across various modules, including event creation, policy handling, and database interactions. - Updated related test cases and benchmarks to ensure compatibility with the new signer interface. - Bumped version to v0.25.0 to reflect these significant changes and improvements in cryptographic operations.
98 lines
3.8 KiB
Markdown
98 lines
3.8 KiB
Markdown
# Performance Benchmark Results
|
|
|
|
## Test Environment
|
|
|
|
- **CPU**: AMD Ryzen 5 PRO 4650G with Radeon Graphics
|
|
- **OS**: Linux (amd64)
|
|
- **Date**: November 4, 2025
|
|
- **Benchmark Time**: 1 second per test
|
|
|
|
## Implementations Compared
|
|
|
|
1. **BTCEC** - btcsuite/btcd/btcec/v2 (Pure Go)
|
|
2. **P256K1** - p256k1.mleku.dev v1.0.2 (Pure Go)
|
|
3. **P8K** - p8k.mleku.dev (Purego + libsecp256k1 v5.0.0)
|
|
|
|
## Results Summary
|
|
|
|
| Operation | BTCEC (ns/op) | P256K1 (ns/op) | **P8K (ns/op)** | P8K Speedup vs BTCEC | P8K Speedup vs P256K1 |
|
|
|---------------------|---------------|----------------|-----------------|----------------------|-----------------------|
|
|
| **Pubkey Derivation** | 32,226 | 28,098 | **19,329** | **1.67x faster** ✨ | 1.45x faster |
|
|
| **Schnorr Sign** | 225,536 | 28,855 | **19,982** | **11.3x faster** 🚀 | 1.44x faster |
|
|
| **Schnorr Verify** | 153,205 | 133,235 | **36,541** | **4.19x faster** ⚡ | 3.65x faster |
|
|
| **ECDH** | 125,679 | 97,435 | **41,087** | **3.06x faster** 💨 | 2.37x faster |
|
|
|
|
## Memory Allocations
|
|
|
|
| Operation | BTCEC | P256K1 | P8K |
|
|
|---------------------|---------------|-------------|-------------|
|
|
| Pubkey Derivation | 80 B / 1 alloc | 0 B / 0 alloc | 160 B / 4 allocs |
|
|
| Schnorr Sign | 1408 B / 26 allocs | 640 B / 12 allocs | 304 B / 5 allocs |
|
|
| Schnorr Verify | 240 B / 5 allocs | 96 B / 3 allocs | 216 B / 5 allocs |
|
|
| ECDH | 32 B / 1 alloc | 0 B / 0 alloc | 208 B / 6 allocs |
|
|
|
|
## Key Findings
|
|
|
|
### 🏆 P8K Wins All Categories
|
|
|
|
**P8K consistently outperforms both pure Go implementations:**
|
|
|
|
- **Schnorr Signing**: 11.3x faster than BTCEC, making it ideal for high-throughput signing operations
|
|
- **Schnorr Verification**: 4.2x faster than BTCEC, excellent for validation-heavy workloads
|
|
- **ECDH**: 3x faster than BTCEC, great for key exchange protocols
|
|
- **Pubkey Derivation**: 1.67x faster than BTCEC
|
|
|
|
### Memory Efficiency
|
|
|
|
- **P256K1** has the best memory efficiency with zero allocations for pubkey derivation and ECDH
|
|
- **P8K** has reasonable memory usage with more allocations due to the FFI boundary
|
|
- **BTCEC** has higher memory overhead, especially for Schnorr operations (1408 B/op)
|
|
|
|
### Trade-offs
|
|
|
|
**P8K (This Package)**
|
|
- ✅ Best performance across all operations
|
|
- ✅ Uses battle-tested C implementation
|
|
- ✅ Bundled library for Linux AMD64 (zero installation)
|
|
- ⚠️ Requires libsecp256k1 on other platforms
|
|
- ⚠️ Slightly more memory allocations (FFI overhead)
|
|
|
|
**P256K1**
|
|
- ✅ Pure Go (no dependencies)
|
|
- ✅ Zero allocations for some operations
|
|
- ✅ Good performance overall
|
|
- ⚠️ ~1.5x slower than P8K
|
|
|
|
**BTCEC**
|
|
- ✅ Pure Go (no dependencies)
|
|
- ✅ Well-tested in Bitcoin ecosystem
|
|
- ✅ Reasonable performance for most use cases
|
|
- ⚠️ Significantly slower for Schnorr operations
|
|
- ⚠️ Higher memory usage
|
|
|
|
## Recommendations
|
|
|
|
**Choose P8K if:**
|
|
- You need maximum performance
|
|
- You're on Linux AMD64 (bundled library)
|
|
- You can install libsecp256k1 on other platforms
|
|
- You're building high-throughput systems
|
|
|
|
**Choose P256K1 if:**
|
|
- You need pure Go (no external dependencies)
|
|
- Memory efficiency is critical
|
|
- Performance is good enough for your use case
|
|
|
|
**Choose BTCEC if:**
|
|
- You're already using btcsuite packages
|
|
- You need Bitcoin-specific features
|
|
- Performance is not critical
|
|
|
|
## Conclusion
|
|
|
|
**P8K delivers exceptional performance** by leveraging the highly optimized C implementation of libsecp256k1 through CGO-free dynamic loading. The 11x speedup for Schnorr signing makes it ideal for applications requiring high-throughput cryptographic operations.
|
|
|
|
The bundled library for Linux AMD64 provides **zero-installation convenience** while maintaining the performance benefits of the native C library.
|
|
|
|
|