ResearchBenchmarksGEMM Kernel Dispatch Optimization on Generic GEMM KernelsGEMM Kernel Dispatch Optimization on Generic GEMM KernelsFollow1.12SpeedupLoKA Dispatch0.83920.91210.9851.0579May 11, 2026Evaluation resultsMethodMethodLinksSpeedupLoKA DispatchScaling Recipe=MixedScaling Recipe=Mixed2026.051.12TorchAOScaling Recipe=RW GW HPScaling Recipe=RW GW HP2026.051.08TorchAOScaling Recipe=TWScaling Recipe=TW2026.051.05TorchAOScaling Recipe=RWScaling Recipe=RW2026.051.01FBGEMMScaling Recipe=RWScaling Recipe=RW2026.050.98DeepGEMMScaling Recipe=BWScaling Recipe=BW2026.050.85