On a B200, the nvjet_tst_16x64_64x16_4x1_v_bz_TNN kernel is used, and it takes roughly 8.1 microseconds. On a H200, the nvjet_tst_64x8_64x16_4x1_v_bz_TNT kernel is ...

Abstract: A mixed-precision analog compute-in-memory (Mix-ACIM) is presented for mixed-precision vector-matrix multiplication (VMM). The design features an all-analog current-domain fixed-point (FxP) ...

This code accompanies the blog post Matrix Multiplication Faster Than Nvidia, Sometimes. It provides a CUDA kernel for single-precision matrix-matrix multiplication, with two notable features: use of ...

Abstract: The demand for high-speed matrix multiplication continues to grow due to recent developments in images processing, graphics processing, digital signal processing and communication via ...