Blogs Map

Paper Reading

Recovering single precision accuracy from Tensor Cores while surpassing the FP32 theoretical peak performance -- Hiroyuki Ootomo, Rio Yokota

Study on matrix units of AMD and NVIDIA GPUs

Investigation on NVIDIA Tensor cores v.s. AMD Matrix cores

• AMD matrix cores
• NVIDIA tensor cores
  Matrix Multiplication Background
• Tiled Matrix Multiplication -- CUDA implementation
• Programming tensor cores using nvcuda-wmma

GPU performance optimization

NVIDIA GPU Performance Background

FPChecker

FPChecker

• FPChecker Installation
  • FPchecker - CMakeLists.txt
• FPChecker issue -- cannot link the openMP lib
• FPChecker exploration -- mixed precision

Conjugate Gradient

Conjugate Gradient

• CG -- Gram-Schmidt Conjugation
• CG -- The Method of Steepest Descent
• CG -- The Method of Conjugate Directions
• CG -- Eigenvalues (Eigenvectors) and convergency