My research interests broadly include computer architecture
and computer systems with a focus on caching. I am particularly
interested in problems at the intersection of hardware and
software.
I was also a member of the
CALCM and
PDL research
organizations at CMU.
Research Projects
I developed
täkō, a polymorphic cache hierarchy which enables software to
fully observe and control data movement. täkō programs can register
callbacks on certain address ranges so that software is invoked when
data in this range moves. täkō's general architecture and programming
interface enable many application-specific optimizations which
previously each required their own dedicated hardware.
I developed
Jumanji, a data-placement technique for dynamic NUCA last-level
caches in chip-multiprocessors. Whereas previous D-NUCA techniques
focused on only throughput-oriented applications, Jumanji provides
quality-of-service for datacenter workloads where multiple
high-priority, latency-critical applications run alongside
low-priority, batch applications and where applications across VMs
require security isolation from each other.
I spent multiple internships at Google optimizing in-memory caches for
miss ratio, I/O cost, and memory usage. This involved developing a
production-accurate simulation infrastructure from scratch.
The Tyr Dataflow Architecture: Improving Locality by Taming Parallelism
Nikhil Agarwal, Mitchell Fream, Souradip Ghosh, Brian C. Schwedock, Nathan Beckmann
MICRO 2024
[pdf]
Leviathan: A Unified System for General-Purpose Near-Data Computing Brian C. Schwedock, Nathan Beckmann
MICRO 2024
[pdf]
UDIR: Towards a Unified Compiler Framework for Reconfigurable Dataflow Architectures
Nikhil Agarwal, Mitchell Fream, Souradip Ghosh, Brian C. Schwedock, Nathan Beckmann
IEEE CAL 2024
[doi]
UDIR: Towards a Unified Compiler Framework for Reconfigurable Dataflow Architectures
Nikhil Agarwal, Mitchell Fream, Souradip Ghosh, Brian C. Schwedock, Nathan Beckmann
WDDSA @ MICRO 2023
[pdf]
Kobold: Simplified Cache Coherence for Cache-Attached Accelerators
Jennifer Brana, Brian C. Schwedock, Yatin A. Manerkar, Nathan Beckmann
IEEE CAL 2023
[doi]
[pdf]
[talk]
Kobold: Simplified Cache Coherence for Cache-Attached Accelerators
Jennifer Brana, Brian C. Schwedock, Yatin A. Manerkar, Nathan Beckmann
WDDSA @ MICRO 2022
[pdf]
täkō: A Polymorphic Cache Hierarchy for General-Purpose Optimization of Data Movement Brian C. Schwedock, Piratach Yoovidhya, Jennifer Seibert, Nathan Beckmann
ISCA 2022 (Best Paper nominee)
[doi]
[pdf]
[slides]
[talk - live]
[talk - recorded]
Jumanji: The Case for Dynamic NUCA in the Datacenter Brian C. Schwedock, Nathan Beckmann
MICRO 2020
[doi]
[pdf]
[slides]
[talk]
[code]
PAWS – A Deployed Game-Theoretic Application to Combat Poaching
Fei Fang, Thanh H. Nguyen, Rob Pickles, Wai Y. Lam,
Gopalasamy R. Clements, Bo An, Amandeep Singh, Brian C. Schwedock,
Milind Tambe, Andrew Lemieux
AI Magazine 2017
[doi]
[pdf]
[AI Magazine version]