[DecJA]

Significant reduction in simulation time by optimizing events writing

• imprint_wins shorts JSON text instead of nested objects
• Deepcopy is replcaced with custom clone function in several places
• get_clusters uses optimized recursive search path
• json_ready_sym makes one getattr call instead of hasattr and getattr
• create_board looks up per-reel values once per reel
• cache reelstrips and distribution conditions
• temp force files use pickle instead of ast.literal_eval
• multi thread final book compression

Results

Run	Before	After	Speedup
Single thread, 6 sample games	—	—	1.97–2.26× faster, about 2.5× lower peak memory
cluster, 8 threads, 48k sims/mode	184.2 s	53.3 s	3.46×
lines, 8 threads, 96k sims/mode	93.8 s	28.3 s	3.31×
scatter, 8 threads, 48k sims/mode	143.5 s	60.3 s	2.38×
merge_luts, 200k base rows	21 s	0.02 s	—