Stack and Heap Allocator with Alignment

Alignment Is a Hardware Contract

malloc on glibc returns 16-byte aligned memory because SSE and AVX loads fault or slow down on misaligned addresses. An int wants 4-byte alignment; a double wants 8. The check is simple: (uintptr_t)ptr % align == 0.

Two Allocation Strategies

Aligned heap block: over-allocate, bump the pointer up to the next aligned boundary, stash the original malloc pointer just before the aligned region so free can find it:

cCopy
Loading…

Stack bump allocator: advance a current pointer through a fixed buffer; pop resets it. LLVM and game engines use this pattern for per-pass temporaries.

• No per-object free on the stack side, only reset
• LIFO discipline: pop must mirror push order
• Alignment rounding uses the same bitmask trick as arena allocators

For this exercise, you will implement aligned_malloc, aligned_free, and a small stack allocator with push/pop. This task asks you to verify alignment with uintptr_t checks, because mysterious SIMD crashes in production are almost always an alignment bug.

Keep the relevant man page, ABI doc, or Rust reference chapter open while you work. When your output disagrees with the reference implementation on the same machine, the mismatch is usually an alignment rule, an off-by-one terminator, or a register slot you misread in GDB.