Aetherial | Benchmark

Hard Serial Dependency: Defeats out-of-order execution by making every memory address dependent on the previous iteration's result. The CPU must finish the entire calculation before it can even calculate the address of the next memory read.

Pointer Chasing (16MB): Overwhelms L1, L2, and L3 caches using a massive interleaved dependency chain. This forces physical RAM latency (60ns+) into the core loop, grounding performance to the speed of the hardware memory controller.

FPU Bottleneck: Uses complex transcendental operations (`Math.atan2`, `Math.pow`) and megamorphic dispatch to saturate the CPU's floating-point units and break speculative JIT inlining.

Deterministic PRNG: Uses the Mulberry32 algorithm seeded at 0xDEADC0DE, now interlocked with the memory-dependent pointer state to prevent speculative execution.

Thermal Stability: Marathon mode calculates stability by dividing the lowest sustained floor by the peak burst. 95%+ indicates excellent cooling.

A simple Benchmark.

Technical Methodology