BLIS Retreat 2019 Program

A preliminary list of talks! Under construction.

Thursday Sept. 19

08:30-09:30 Breakfast (pastries, coffee)
If you need something more solid, you are on your own!
09:30-09:40 Welcome Robert van de Geijn, UT-Austin
09:40-10:10 The BLIS Approach to Skinny Matrix Multiplication
[ slides ] [ video ]
Field Van Zee, UT-Austin
10:10-10:25 Issues with fat by thin matrix multiplication
[ slides ]
Thomas Hines, Tennessee Tech
10:25-10:40 I/O Lower Bounds for Small MMM Tyler Smith, ETH-Zurich
10:40-11:45 Discussion/Coffee
11:45-12:00 What BLIS brings to the table--performance, portability, and productivity Devangi Parikh, UT-Austin
12:00-12:15 Programming GPUs for speed-of-light linear algebra Cris Cecka, NVIDIA
12:15-12:30 BLIS optimizations and results on AMD Rome Kiran Varaganti, AMD
12:30-2:00 Discussion/Lunch
Please buy your own lunch at O's Cafe and bring it back to the room.
02:00-02:15 Analytical models for MMM-like problems on GPUs Tze Meng Low, CMU
02:15-02:30 FBGEMM: High-Performance Low-Precision Library for Deep Learning Inference Jianyu Huang, Facebook
02:30-2:45 Indirect GEMM and Indirect Convolution Algorithm
[ Abstract ]
Marat Dukhan, Google Research
02:45-03:45 Discussion/Coffee
03:45-04:00 The Linear Algebra Mapping Problem Christos Psarras, RWTH-Aachen
04:00-04:15 TBA Greg Henry, Intel
04:15-04:30 GEMM-Based Kernels for Tensor Hypercontraction Devin Matthews
05:30- Beer and food at Scholz Garten for those who want to gather some more.

Friday Sept. 20

08:30-09:30 Breakfast (pastries, coffee)
If you need something more solid, you are on your own!
09:30-09:45 What you don't know can hurt performance — Snoop filters in Intel Xeon Scalable processors John McCalpin, TACC
09:45-10:00 Mapping BLIS to the IBM Power9 architecture Nicholai Tukanov, UT-Austin
10:00-10:15 What MLIR has to offer as a toolkit for building and leveraging numerical libraries in ML and HPC
[ Link ]
Albert Cohen, Google
10:15-11:15 Discussion/Coffee
11:15-11:30 Ensuring the future of BLIS through education Maggie Myers, UT-Austin
11:30-11:45 Formal derivation: FLAME's secret sauce
[ Paper ]
Robert van de Geijn, UT-Austin
11:45-12:00 FLAME in Machine Learning (ML) Applications Martin Schatz, Facebook
12:00-12:10 Wrap up Robert van de Geijn, UT-Austin