pop_l123_popcorn2_bench_default_gpt · variant overview
trajectories
102
finished
99
correct
33
compiled
76
avg speedup (correct)
5388.87x
sweep-wide eval metricsaveraged over correct kernels
outcomesincorrect 43 · correct 33 · error 17 · compile_fail 6 · in_progress 3
avg speedup5388.87x (over 33 correct)
avg memory ratio (kernel / ref)0.96x (33 samples)
avg fusion ratio (ref launches / kernel launches)3623.82x (33 samples)
avg energy ratio (ref / kernel)62.29x (7 samples)
avg SOL score (0–1)0.220 (32 samples)
avg DRAM util %15.4%
avg occupancy %202.9%
avg max abs error2.03e-06
modelncorrectcompiled avg speedup
FW-GLM-5-1 34 8 23 15111.91x
gpt-5.5-priority 34 13 26 2118.71x
grok-4.3 34 12 27 1855.90x