Section 4.1 Program at a glance
All times are Central Daylight Time (CDT). In other words, the time in Texas.
It is likely that some "talks" do not draw a full 10 minutes of discussion. In that case, we will move on to the next "talk." Thus, the start of a session is where we synchronize, but within a session we may get ahead of the schedule.
Eventually we hope to add recordings of the discussions to each "talk," with permission of those who participated in the discussion. So, you should feel free to say what you want in front of the workshop audience. If you say something you later regret, then we will work with you to edit that out.
| 10:00- | Zoom session and Breakout rooms are open for mingling. (Click on "Breakout room" at bottom of zoom window.) |
|
| 10:55-11:00 | Robert van de Geijn | Welcome |
| Session 1 | Moderator: Tze Meng Low | |
| 11:00-11:20 | Field Van Zee | The State of BLIS + Ask Me Anything! |
| 11:20-11:30 | Gianluca Frison | Small/skinny matrix-matrix multiplication: to pack or not to pack? |
| 11:30-11:40 | Nicholai Tukanov | Power10: MMA and GEMM in BLIS |
| 11:40-11:50 | Stepan Nassyr | Porting BLIS to Arm SVE |
| Session 2 | Moderator: Robert van de Geijn | |
| 12:00-12:10 | Minh Quan Ho | clBLIS: Accelerating BLIS with BLIS |
| 12:10-12:20 | Devangi Parikh and Greg Henry | Toward Fast FP128 GEMM Accuracy without higher precision GEMM kernels |
| 12:20-12:30 | Jeff Diamond | How slow can it go? Detuning BLIS for fun and profit |
| 12:30-12:40 | Mesut Meterelliyoz | Intel® oneAPI Math Kernel Library for Intel CPUs and GPUs |
| 12:40-12:50 | Marat Dukhan | Accelerating Convolutional Neural Networks with Sparsity |
| 12:50-13:00 | Pablo San Juan | High Performance and Portable Convolution Operators through BLIS integration |
| Session 3 | Moderator: Devin Matthews | |
| 13:10-13:20 | Henrik Barthels | Linnea: Automatic Generation of Efficient Linear Algebra Programs |
| 13:20-13:30 | Uday R. Bondhugula | Using MLIR for BLIS like Code Generation |
| 13:30-13:40 | Tim Davis | Suitesparse: GraphBLAS |
| 13:40-13:50 | Tze Meng Low | Small Prime DFTs as Specialized Matrix Vector Multiplies |
| 13:50-14:00 | Richard Veras | A BLIS for Stencil Computations |
| Session 4 | Moderator: Devangi Parikh | |
| 14:10-14:20 | Koby Hayashi | Distributed-Memory Parallel Symmetric Nonnegative Matrix Factorization |
| 14:20-14:30 | Christos Psarras | Achieving the compute bound with CP-CALS |
| 14:30-14:40 | Devin Matthews | The BLAS are Not Enough |
| 14:40-14:50 | Maggie Myers and Robert van de Geijn | Sharing BLIS with A LAFF |
| 14:50-15:00 | Stefan Robila | Data and Software Cyberinfrastructure Research and Development |
| Panel | Moderator: Tim Mattson | |
| 15:10-15:40 | Perspectives on Moving a Research Project to Successful Open Source Software | |
| 15:40-15:45 | Robert van de Geijn | Wrapup |
| 15:45- | Zoom session and breakout rooms continue to be open for mingling | |