SIMD2: A Generalized Matrix Instruction Set for Accelerating Tensor Computation beyond GEMM

Matrix-multiplication units (MXUs) are now prevalent in every computing platform. The key attribute that makes MXUs so successful is the semiring structure, which allows tiling for both parallelism and data reuse. Nonetheless, matrix-multiplication is not the only algorithm with such attributes. We find that many algorithms share the same structure and differ in only the core operation; for example, using add-minimum instead of multiply-add. Algorithms with a semiring-like structure therefore have potential to be accelerated by a general-purpose matrix operation architecture, instead of common MXUs. In this paper, we propose SIMD2, a new programming paradigm to support generalized matrix operations with a semiring-like structure. SIMD2 instructions accelerate eight more types of matrix operations, in addition to matrix multiplications. Since SIMD2 instructions resemble a matrix-multiplication instruction, we are able to build SIMD2 architecture on top of any MXU architecture with minimal modifications. We developed a framework that emulates and validates SIMD2 using NVIDIA GPUs with Tensor Cores. Across 8 applications, SIMD2 provides up to 38.59× speedup and more than 10.63× on average over optimized CUDA programs, with only 5% of full-chip area overhead.