CompilerResearchCon 2025 — Virtual Event on 30 October at 15:00 CET/07:00 PT
Vassil Vassilev
Dear colleagues and community members,
We are pleased to invite you to the upcoming CompilerResearchCon 2025, taking place virtually on 30 October 2025 at 15:00 CET/07:00 PT.
Join us for a series of exciting talks showcasing cutting-edge research at the intersection of compiler technology, high-performance computing, and computational biology.
We look forward to your participation in this exciting session celebrating innovation at the frontiers of compiler research and computational science.
Warm regards,
The CompilerResearchCon
2025 Organizing Committee
Agenda
15:00 – 15:20
Salvador de la Torre
Gonzalez
“CARTopiaX: an Agent-Based
Simulation of CAR-T Cell Therapy built on BioDynaMo”
Abstract:
CAR-T-cell therapy is a transformative cancer immunotherapy, yet
its success in solid tumors remains limited by tumor complexity. CARTopiaX, built on the
BioDynaMo simulation platform, offers a high-performance
agent-based model that replicates and extends published results in
CAR-T dynamics. The project achieved major milestones in
simulation speed, code quality, and modular design—empowering
researchers to explore tumor–immune interactions at unprecedented
scales.
15:20 – 15:40
Rohan Timmaraju
“Efficient LLM Training in
C++ via Compiler-Level Autodiff with Clad”
Abstract:
To overcome the computational bottlenecks of Python-based
frameworks, this project introduces cladtorch, a C++ tensor library with a
PyTorch-style API integrated with Clad’s compiler-level automatic
differentiation. Two LLM training implementations demonstrate the
performance gains achievable through compiler-based AD—showcasing
significant efficiency improvements over PyTorch and paving the
way toward future hardware acceleration.
15:40 – 16:00
Aditi Milind Joshi
“Implement and Improve an
Efficient, Layered Tape with Prefetching Capabilities”
Abstract:
This work enhances Clad’s tape data structure for reverse-mode
differentiation by introducing slab allocation, small buffer
optimization, and thread safety. The ongoing development of a
multilayer, offloadable tape system promises improved scalability
and memory efficiency, enabling faster and more concurrent
computation.
16:00 – 16:20
Abdelrhman Elrawy
“Support Usage of Thrust API
in Clad”
Abstract:
Integrating NVIDIA’s Thrust
library into Clad enables automatic differentiation of
GPU-accelerated algorithms such as reductions, transformations,
and scans. This extension marks an important step toward robust,
compiler-based AD for parallel and scientific computing.
Event Details
Date: Thursday, 30 October 2025
Time:
15:00–16:20 (Geneva, CH Time)
Location:
Virtual https://compiler-research.org/crcon2025/#crcon25_part_1
Vassil Vassilev
Dear colleagues and community members,
We are pleased to invite you to the upcoming second part of CompilerResearchCon 2025, taking place virtually on 13 November 2025 at 15:00 CET/06:00 PT.
Join us for a series of exciting talks showcasing cutting-edge research at the intersection of compiler technology, high-performance computing, and computational biology.
We look forward to your participation!
Warm regards,
The
CompilerResearchCon 2025 Organizing Committee
Agenda
15:00 – 15:20
Abhinav Kumar
“Implementing Debugging
Support for xeus-cpp”
Abstract:
This project integrates debugging support into the
xeus-cpp kernel for Jupyter, leveraging LLDB and the Debug
Adapter Protocol (lldb-dap). Mirroring the approach used in
xeus-python, it enables breakpoints, variable inspection, and
step-through debugging for interactive C++ development. Key
achievements include DAP integration, full debugging
functionality in JupyterLab, and a major refactor of
Out-of-Process JIT execution to support the new debugge
15:20 – 15:40
Maksym
Andriichuk
“Activity Analysis for
Reverse-Mode Differentiation of (CUDA) GPU Kernels”
Abstract:
This work enhances Clad,
the Clang plugin for automatic differentiation, with advanced
activity analysis for GPU kernels. By implementing static
thread-safety analysis, the project detects and optimizes away
potential data-race conditions—significantly reducing atomic
operations and improving performance in differentiated CUDA
code.
15:40 – 16:00
Jiayang Li
“Enable Automatic
Differentiation of OpenMP Programs with Clad”
Abstract:
Extending Clad to support OpenMP programs, this project enables
gradient computation in multi-threaded environments. The
implementation adds differentiation for forward and reverse
modes across common OpenMP directives and clauses, complete
with robust AST parsing, differentiation strategies, and
comprehensive testing.
16:00 – 16:20
Aditya Pandey
“Using ROOT in the Field
of Genome Sequencing”
Abstract:
This project adapts ROOT,
CERN’s large-scale data framework, to the domain of genome sequencing.
Using ROOT’s next-generation RNTuple format, it delivers 23.5%
smaller genomic data files and up to 3.2× faster full-chromosome queries
compared to CRAM. Additional improvements include
chromosome-based file splitting and optimized compression,
enabling efficient storage and analysis of large genomic
datasets.
Event Details
Date: Thursday, 13 November 2025
Time:
15:00–16:20 (Geneva, CH Time)