pop_l123_popcorn2_st_gpt · variant overview
trajectories
285
finished
285
correct
50
compiled
133
avg speedup (correct)
245.83x
sweep-wide eval metricsaveraged over correct kernels
outcomescompile_fail 103 · incorrect 83 · correct 50 · error 49
avg speedup245.83x (over 50 correct)
avg memory ratio (kernel / ref)0.90x (50 samples)
avg fusion ratio (ref launches / kernel launches)927.75x (50 samples)
avg energy ratio (ref / kernel)15.77x (10 samples)
avg SOL score (0–1)0.260 (47 samples)
avg DRAM util %20.9%
avg occupancy %131.5%
avg max abs error9.13e-05
modelncorrectcompiled avg speedup
FW-GLM-5-1 95 17 41 366.56x
gpt-5.5-priority 95 20 50 272.24x
grok-4.3 95 13 42 47.32x