Section 5.1 Program at a glance
All times are Central Daylight Time (CDT), in other words, the time in Texas.
Talks are scheduled in sessions. Speakers should plan for a 20 minute talk plus time for questions. Once discussion comes to an end, we move on to the next talk. Thus, the schedule is elastic.
Eventually Robert hopes to add recordings of the talks to these pages, with permission of the speaker.
For now, all times are preliminary.
Thursday September 25 | ||
8:30- | Zoom session, breakout rooms, and Discord are open for mingling. Coffee and muffins for in-person participants. |
|
(Click on "Breakout room" at bottom of zoom window.) | ||
8:55-9:00 | Welcome | |
Robert van de Geijn | ||
9:00am - 10:45am | Session 1 | Moderator: |
Devin Matthews SMU |
BLIS Day the 13th: BLIS out of BLAS (talks gets extra time) More Info 5.2.5 |
|
Harsh Dave AMD |
Decomposition-Aware GEMM Kernels for High-Performance Skinny Matrix Computation More Info 5.2.7 |
|
Sameer Ahmadl AMD |
Accelerating Eigenvalue and Singular Value Decompositions: Techniques for Improved Performance More Info 5.2.9 |
|
Additional discussion | ||
10:45am - 11:00am | Break | |
11:00am -12:30pm | Session 2 | Moderator: |
Stepan Nassyr Juelich Supercomputing Center |
Continued experiments in microkernel generation More Info 5.2.2 |
|
Nima Sahraneshan Universitat Jaume I |
No-I-Meant-Another QR” (NIMA-QR): Two-Stage Newton–Schulz-Refined Mixed-Precision QR Factorisation More Info 5.2.12 |
|
Participants |
Discussions of how you use and contribute to BLIS |
|
Additional discussion | ||
12:30 - 1:30pm | Lunch | |
1:30pm - 3:00pm | Session 3 | Moderator: |
John Gunnels Nvidia |
A Framework for Productizing Ozaki-Style Emulation More Info 5.2.1 |
|
Chao Yin SMU |
Low-Precision GEMM on ARM Architecture More Info 5.2.4 |
|
Cem Bassoy DeepL |
Fast and layout-oblivious tensor-matrix multiplication with BLAS More Info 5.2.13 |
|
Additional discussion | ||
3:00pm | End of Day | |
Friday Sept. 26 | ||
8:00- | Zoom session, breakout rooms, and Discord are open for mingling. Coffee and muffins for in-person participants. |
|
(Click on "Breakout room" at bottom of zoom window.) | ||
9:00am - 10:30 | Session 4 | Moderator: |
Harsh Dave AMD |
Memory Allocator Impact on AOCL-BLAS: Jemalloc More Info 5.2.8 |
|
Bhaskar Nallani AMD |
AOCL-DLP Overview More Info 5.2.10 |
|
Shiv Sundram Stanford University |
REPTILE: Performant Tiling of Recurrences and Linear Solvers More Info 5.2.11 |
|
Additional discussion | ||
10:30am - 10:45am | Break | |
10:45am -12:15pm | Session 5 | Moderator: |
Tze Meng Low CMU |
Lessons learned from modeling GEMM in the face of changing architectures More Info 5.2.6 |
|
Marco Barbone Flatiron Institute |
Fast NUFFT and SIMD More Info 5.2.3 |
|
Jackson Vanover University of California, Davis |
EXCVATE: Spoofing Exceptions and Solving Constraints to Test Exception Handling in Numerical Libraries More Info 5.2.14 |
|
Additional discussion | ||