What problem does it solve? When optimizing z88dk library code for 8085, 8080, or gbz80 targets, developers need a rigorous way to prove what is slow, what changed, and what is correct. This Skill provides the measurement methodology: cycle-accurate timing with z88dk-ticks, hotspot profiling, map/nm link verification, A/B library swaps, and correctness gates via the math test suites. ## Core Features & Use Cases - Cycle-Accurate Timing: Use z88dk-ticks with TIMER_START/TIMER_STOP labels and map files to get exact T-state counts for timed regions, avoiding wall-clock noise. - Hotspot Profiling: Capture PC histograms via the ticks debugger (hotspot on) and aggregate cycles by symbol, source file, or library category to find where time is actually spent. - A/B Library Validation: Swap a single .asm file, rebuild the affected libraries (math32, math16, crt0), verify binary identity with cmp/md5sum, and confirm symbol linkage via map files and z88dk-z80nm. - Correctness Gates: Run test/suites/math targets (test_math32_8085.bin, test_math16.bin, etc.) before publishing benchmark numbers. - Use Case: After rewriting a 32-bit division routine for the 8085, rebuild math32_8085.lib, run the math suite to 16/16 passing, then remeasure the fasta benchmark ticks and confirm via the map that l_long_div_0 is actually linked before claiming a 5% improvement. ## Quick Start Ask the AI to measure the cycle count of a z88dk benchmark using z88dk-ticks with TIMER labels and an 8085 CPU model, then show the top hotspots by symbol.
