ResearchBenchmarksGeneral Matrix-Vector Multiplication on GEMV (M=1, N=28672, K=8192) (kernel-level evaluation)Follow6.12TFLOPSCodeGEMM0.53521.98513.4354.8849Dec 19, 2025Evaluation ResultsMethodMethodLinksTFLOPSPower (W)GFLOPS/WGPU Util (%)Mem Util (%)CodeGEMMconfig_string=m1v4g128config_string=m1v4g1282025.126.12316.3819.3684.4749.8CodeGEMMconfig_string=m2v8g128config_string=m2v8g1282025.125.43304.6917.8385.3243.75AQLMkernel_config=2x8kernel_config=2x82025.122.59254.210.1892.8419.96cuBLASprecision=FP16precision=FP162025.121.58318.554.9596.8796.94AQLMkernel_config=1x16kernel_config=1x162025.120.75126.545.93996