Liv(f)e

Family reunion

31-May-2016

15-Aug-2016

My duty is to conduct research related to programming systems and applications on NVIDIA's GPU. To be specific, I developed high-performance general kernel matrix primitives, which can be instantiated to accelerate machine learning and deep learning applications. The whole development requires C/C++, CUDA, PTX, and Python programming, advanced knowledge in machine learning and GPU architecture. The testing and deployment platform involves CPU clusters and NVIDIA Kepler GPUs.

27-May-2014

15-Aug-2014

My duty is to conduct research on one-way seismic image reconstruction with Fourier finite-difference (FFD) migration. To be specific, I develop parallel FFD algorithms that employ a distributed memory hybrid CPU-GPU cluster (Cray-XC30) using OpenACC, a user-driven directive-based performance-portable parallel programming model. The whole development requires Fortran90, CUDA, PTX, and MPI programming, advanced knowledge in partial differential equations, image reconstruction, CPU, GPU, and network architecture. The testing and deployment platform is Cray-XC30, a hybrid CPU-GPU clusters with NVIDIA K40 GPUs.

My duty is to develop algorithms and infrastructure related to deep learning and artificial intelligence. This position requires C/C++, CUDA, PTX, Python and MPI programming, advanced knowledge in machine learning, deep learning, computer, and network architecture. The whole development requires a set of profiling, debugging, and visualization tools from CUDA Toolkits, deep learning library and runtime such as cuDNN and TensorRT, frameworks such as pyTorch and TensorFlow. The experiment and production are performed on hybrid CPU-GPU clusters (e.g. DGX, NGC compute cloud, and supercomputers) and mobile (or embedded) systems.

After 7 years, finally I got into SC. It has been so long. -- June 30, 2015

事隔 7 年. I finally got accepted by SC (Super Computing). 第一次聽到 SC 這個詞是在 08 年參與 BCX-EXT 計畫得時候. 相中 GPU 潛力的 Dan'l 打算搶在這潮流的前端首先投入 Sparse matrix computation 的領域攻城掠地. 可惜在 Fortran 77 和可怕的 Tesla Beta (non IEEE-754) 測試機的折騰下, 最後結果不盡理想. SC at Las Vegas 的念頭也就打消了. 如今去美國開會這件事已經不怎麼 exciting 了, 尤其是 SC15 還在 Austin, 我哪也去不了. 不過人生有種圓滿的感覺. (茶) -- June 30, 2015

We got snow at Texas!

Received student paper prize of the Mathematics Institution of Taiwan at Hsinchu, 2013.

When I was notified by the institution that I won the prize, I was still serving my country at Kinmen island (金門島). My company commender granted me a 4-day vacation to return to Taiwan for the ceremony.

I was quite fit at that moment. LOL ...