High-performance computing (HPC) demos since I was a freshmen.

My attempt of making a GEMM kernel...

The fastest Tropical number matrix multiplication on GPU

CUDA version GEMM

Pytorch Operation for distributed gemm in nvidia blackwell gpus

GEMM on Apple Silicon

SM89 Optimized CUDA Kernels

Yet another rust BLAS

Development of deep learning inference code by OpenCL kerenl function.

My gemm optimization on RPi (ARM) achieved a 170x performance boost, showing speeds faster than Eigen and close to OpenBLAS.

Low Precision Arithmetic for Convolutional Neural Network Inference

Generate optimized MatMul cuda kernel automatically using tvm auto schedule.

Performance comparison of naive, AVX2-optimized, and cBLAS matrix multiplication implementations in C.

Safety-hardened GEMM (matrix multiply) implementation achieving 169.8 GFLOPS on Intel i9-14900. Built for embedded systems and safety-critical applications where reliability matters as much as speed. 162× faster than naive, zero UB, fully validated.

Repo for the SPOGA Accelerator - Scaling Analog Photonic Accelerators for Byte-Size, Integer General Matrix Multiply (GEMM) Kernels

Fast Convoluion Implementation via CUDA

Learning notes of "flame/how-to-optimize-gemm".

mixed-precision GEMM library

Implementing LLM from scratch. (Developing...)