[Cryptech Tech] fyi: Automatic Generation of Efficient Accelerator Designs for Reconfigurable, Hardware

[=JeffH]

of possible background interest wrt FPGAs...


Subject: [netseminar] Stanford Platform Lab Seminar,
  *Tuesday FEB / 28* @ 4:30pm, Raghu Prabhakar (Stanford University)
From: Alex Pinedo <asandra at cs.stanford.edu>
Date: Thu, 23 Feb 2017 11:49:10 -0800
To: platformlab at lists.stanford.edu, netseminar at lists.stanford.edu,
  christos_students at mailman.stanford.edu, "ON.Lab All" <all at onlab.us>,
  cs-seminars at lists.stanford.edu
Cc: "Pinedo, Alex Sandra" <apinedo at stanford.edu>

*Speaker: *Raghu Prabhakar, Stanford University

*Title:*
Automatic Generation of Efficient Accelerator Designs for Reconfigurable
Hardware



*Abstract: *
Acceleration in the form of customized datapaths offers large performance
and energy improvements over general purpose processors. Reconfigurable
fabrics such as FPGAs are gaining popularity for use in implementing
application-specific accelerators. However, current tools for targeting
FPGAs offer inadequate support for high-level programming, resource
estimation, and rapid and automatic design space exploration. In this talk,
I will describe a design framework we have developed that addresses these
challenges. We introduce a new parallel patterns-based representation of
hardware using parameterized templates that capture locality and
parallelism information at multiple levels of nesting. We describe an area
estimation technique using high-level models to account for low-level effects
from hardware place-and-route tools. We use our estimation capabilities to
rapidly explore a large space of designs across tile sizes, parallelization
factors, and optional coarse-grained pipelining, all at multiple loop levels.
We show that estimates average 4.8% error for logic resources, 6.1% error
for runtimes, and are 279 to 6533 times faster than a commercial high-level
synthesis tool. We compare the best-performing designs to optimized CPU
code running on a server-grade 6 core processor and show speedups of up to
16.7×


*About the speaker*:
Raghu Prabhakar is a fourth-year PhD candidate in the Computer Science
department at Stanford University, co-advised by Prof. Christos Kozyrakis
and Prof. Kunle Olukotun. He is interested in exploring new programming
models, compiler techniques, and architectures for spatially reconfigurable
hardware.