Program Schedule
The 36th International Workshop on Languages and Compilers for Parallel Computing
October 11-13, 2023, Lexington, Kentucky, USA
The workshop schedule appears below.
Registration
is required for attendees,
but is free and includes the banquet.
The papers linked here are all to be considered preliminary versions for
attendees to access during the conference;
their URLs here might not be maintained in the long run and thus
should NOT be cited.
Click here for parking information and maps.
The space we will be using for the workshop is primarily the Hardymon Theater
on the first floor of the Davis Marksbury
Building. Professor Hank Dietz, the organizer for this year's LCPC, also
has his office, research lab, and cluster computing machine room in this
building. The research lab (room 108) and machine room (room 108A) are both
just down the hall from the Hardymon Theater; the lab has plenty of space for
attendees to do things ranging from temporarily storing luggage to having a
quiet space for a Zoom meeting that they couldn't miss while attending LCPC.
The workshop technical content includes:
-
13 talks on peer-reviewed submitted research papers.
Each talk will be given a 30-minute window;
it is recommended that at least 5 minutes be left for questions.
-
Two invited keynotes.
Each keynote will be given a 60-minute window;
it is recommended that at least 10 minutes be left for questions.
-
Three invited panels.
Each panel will be given a 60-minute window.
The moderator will briefly introduce the panel topic and each
panelist will present a 5-10 minute explanation of their position;
the remainder of the time will be used for questions and discussion.
-
Two invited papers.
Each talk will be given a 30-minute window;
it is recommended that at least 5 minutes be left for questions.
The tentative workshop schedule follows.
Wednesday, October 11, 2023
- 10:00-
-
Optionally visit UK Art Museum (1 block from Marksbury, takes about 1 hour)
- 12:00-12:50
-
Workshop check-in at Davis Marksbury Building
- 12:50-13:00
-
Workshop introductory remarks
- 13:00-14:00
-
Keynote: AB Siddique
- 14:00-15:30
-
Reviewed paper session: AI in Compilers
-
Comparing Llama-2 and GPT-3 LLMs for HPC kernels generation
Valero-Lara, Huante, Al Lail, Godoy, Teranishi, Balaprakash, Vetter
-
Tiling parameters prediction using Machine Learning techniques
Berezov, Ancourt, Kashchenko
Note: this paper will be presented live via Zoom
-
PEAK: Generating High-Performance Schedules in MLIR
Tavakkoli, Joshi, Singh, Xu, Sadayappan, Hall
-
- 15:30-16:00
-
Break
- 16:00-17:00
-
Panel: AI in Programming & Compilers
Moderator: Dietz, Panelists: Siddique, Shivam, Sadayappan
- 17:00-17:30
-
Break
- 17:30
-
Board bus to Banquet, to be held at
Merrick Inn;
LCPC will provide food,
there will be a cash bar for drinks
Thursday, October 12, 2023
- 08:30-09:00
-
Workshop check-in at Davis Marksbury Building
- 09:00-10:00
-
Keynote: Hironori Kasahara
OSCAR Codesigned Compiler and Multicore Architecture
Abstract:
This talk starts research on the Codesign of OSCAR
Multiprocessor to support compiler parallelization using
fine-grain to coarse-grain hierarchical parallelism, static and
dynamic scheduling, overlapping data transfer using DMA, and
variable group barrier synchronization that has started since
1984. The architecture and compiler have evolved, chasing the
progress of both technologies. Currently, codesign allows us
high-performance execution of various scientific, medical,
AI-robot, automotive, FA, multi-media applications, and so on,
using multigrain parallel processing with last-level cache,
distributed, and local memories for a total program. In
addition, the codesign has realized low-power green computing
with no-cooling or solar-powered execution for homogeneous and
heterogeneous multicores.
- 10:00-11:30
-
Reviewed paper session: Data Layout and Movement
- 11:30-13:00
-
Lunch, on your own
- 13:00-14:00
-
Panel: Evolution of Parallel Architecture Targets
Moderator: Dietz, Panelists: Kasahara, Sadeghi, Thakkar
- 14:00-15:30
-
Reviewed paper session: Target Architecture Details Matter
-
A Multiple Compiler Framework for Improved Performance
Shivam, Nicolau, Veidenbaum
-
Automatic Deep Learning Parallelization for Vector Multicore Chips with the OSCAR Parallelizing and the TVM Open-Source Deep Learning Compiler
Onishi, Otaka, Fujita, Suetsugu, Kawasumi, Kitamura, Kasahara, Kimura
-
Quantum Circuit Resizing via Serial Execution
Sadeghi, Khadirsharbiyani, Kandemir
- 15:30-16:00
-
Break
- 16:00-17:00
-
Invited paper session: Programming Languages
-
Declaring Independence from Incidental Dependencies
Marron
-
Answer-set programming: an introduction based on Sudoku
and (Thomas G.) Room arrangements
Finkel
- 17:00-18:00
-
Panel: Parallel Programming and Languages
Moderator: Dietz, Panelists: Gressman, Marron, Eigenmann
- 18:00-18:30
-
Break
- 18:30
-
Walk to Kentucky Native Cafe;
LCPC will provide apetizers,
you would pay for your own food and drink
Friday, October 13, 2023
- 08:30-09:00
-
Workshop check-in at Davis Marksbury Building
- 09:00-11:00
-
Reviewed paper session: Application Support and Portable Performance
-
A Parallel Hybrid Genetic Search for the Capacitated VRP with Pickup and Delivery
Stadler, Nita, Dunnweber
-
Efficiently Exploiting Irregular Parallelism Using Keys at Scale
Wang, Rajasukumar, Su, Nourian, Monsalve Diaz, Pervaiz, Ding, Colley, Wang, Li, F. Gleich, Hoffmann, A. Chien
-
SEER: An on-node Performance-Portable C++ Library for Heterogeneous Systems
Valero-Lara, Vetter
-
Benchmarking Operators in Deep Neural Networks for Improving Performance Portability of SYCL
Jin, Vetter
- 11:00-11:30
-
Workshop closing remarks
- 11:30-
-
At your option, there are many things to do, including:
LCPC2023 Languages and Compilers for Parallel Computing