| Operation | STL (std::vector) | Bfd3 core library | Improvement | |------------------------------------|-------------------|------------------|-------------| | 1M int insert at back | 12.3 ms | 11.1 ms | 9% | | 100k small string push (FixedString)| 45.2 ms (string) | 8.4 ms | 438% | | Multi-producer queue throughput | 8.2M ops/sec (mutex) | 24.5M ops/sec | 199% | | Arena allocation (1M blocks) | 345 ms (new/delete) | 87 ms | 296% |
bfd3::BinaryWriter writer(bfd3::Endian::Little); writer.write<uint32_t>(0x12345678); writer.writeString("hello"); auto bytes = writer.data(); In a controlled benchmark (x86_64, GCC 12, O3 optimization), the Bfd3 core library often outperforms equivalent STL constructs in specific metrics.
If your project demands the absolute best from every cycle and every byte, it's time to explore what the Bfd3 core library can do for you. Have you used the Bfd3 core library in a project? Share your experience or performance metrics in the comments below. For further reading, check out other articles on custom memory management and lock-free programming. Bfd3 core library
🚫 Pitfall 1: Forgetting Intrusive Container Node Lifetime If an object with an intrusive list node is destroyed while still in a list, the list's next/prev pointers become dangling. Always remove before destruction. 🚫 Pitfall 2: Mixing Allocators Memory allocated from bfd3::MemoryArena must not be freed with delete . Use the arena's own reset mechanism. ✅ Best Practice: Use RAII Wrappers Even with custom memory, encapsulate allocations in small scope-bound objects.
struct Task : public bfd3::IntrusiveListNode<Task> int priority; void execute(); ; bfd3::IntrusiveList<Task> pendingTasks; Task t1, t2; pendingTasks.push_back(t1); pendingTasks.push_back(t2); For multi-threaded producer-consumer scenarios, a lock-free ring buffer (multi-producer, single-consumer or multi-consumer) is essential. The Bfd3 core library often includes a highly tuned implementation based on atomic operations, avoiding mutex overhead entirely. | Operation | STL (std::vector) | Bfd3 core
#include <bfd3/MemoryArena.h> bfd3::MemoryArena arena(1024 * 1024); // 1 MB arena void* buffer = arena.alloc(256); // allocate 256 bytes arena.reset(); // free all allocations at once
bfd3::MCRingBuffer<int, 1024> queue; queue.push(42); // lock-free, safe from multiple threads int value; if (queue.pop(value)) ... Heap-allocated strings are a common source of fragmentation and performance issues. The Bfd3 core library provides a fixed-capacity string that lives entirely on the stack (or inside any other object). Share your experience or performance metrics in the
bfd3::FixedString<64> filename = "config_"; filename.append("data.bin"); const char* cstr = filename.c_str(); // null-terminated For network protocols or file I/O, endianness and padding matter. The core library offers binary streams with explicit byte ordering.