BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163557Z
LOCATION:505
DTSTART;TZID=America/Denver:20191118T090000
DTEND;TZID=America/Denver:20191118T173000
UID:submissions.supercomputing.org_SC19_sess128@linklings.com
SUMMARY:IA^3 2019: 9th Workshop on Irregular Applications: Architectures a
nd Algorithms
DESCRIPTION:Workshop\n\nDue to the heterogeneous data sets they process, d
ata intensive applications employ a diverse set of methods and data struct
ures, exhibiting irregular memory accesses, control flows, and communicati
on patterns. Current supercomputing systems are organized around component
s optimized for data locality and bulk synchronous computations. Managing
any form of irregularity on them demands substantial programming effort, a
nd often leads to poor performance. Holistic solutions to these challenges
emerge only by considering the problem from multiple perspectives: from m
icro- to system-architectures, from compilers to languages, from libraries
to runtimes, and from algorithm design to data characteristics. Only stro
ng collaborative efforts among researchers with different expertise, inclu
ding domain experts and end users, can lead to significant breakthroughs.
This workshop brings together scientists with different backgrounds to dis
cuss methods and technologies for efficiently supporting irregular applica
tions on current and future architectures.\n\nhttps://hpc.pnl.gov/IA3/\n\
nMetall: A Persistent Memory Allocator Enabling Graph Processing\n\nIwabuc
hi, Lebanoff, Gokhale, Pearce\n\nWe present Metall, a persistent memory al
locator designed to provide developers with an API to allocate custom C++
data structures in both block-storage and byte-addressable persistent memo
ries (e.g., NVMe and Intel Optane DC Persistent Memory). Metall incorpora
tes state-of-the-art allocation algor...\n\n---------------------\nIA^3 20
19: 9th Workshop on Irregular Applications: Architectures and Algorithms\n
\nTumeo, Feo, Castellana\n\nDue to the heterogeneous data sets they proces
s, data intensive applications employ a diverse set of methods and data st
ructures, exhibiting irregular memory accesses, control flows, and communi
cation patterns. Current supercomputing systems are organized around compo
nents optimized for data localit...\n\n---------------------\nKeynote 2: S
parse Linear Algebra in Facebook's Deep Learning Models\n\nPark\n\nSparse
linear algebra plays surprisingly important roles in deep learning models
used in social network. The use cases share similar challenges as scientif
ic computing applications using sparse linear algebra. I believe both comm
unities can benefit from each other by working together on basic SW bui...
\n\n---------------------\nKeynote 1\n\nWong\n\n---------------------\nRDM
A vs. RPC for Implementing Distributed Data Structures\n\nBrock, Chen, Yan
, Owens, BuluĂ§...\n\nDistributed data structures are key to implementing s
calable applications for scientific simulations and data analysis. In thi
s paper we look at two implementation styles for distributed data structur
es: remote direct memory access (RDMA) and remote procedure call (RPC). W
e focus on operations tha...\n\n---------------------\nCascaded DMA Contro
ller for Speedup of Indirect Memory Access in Irregular Applications\n\nKa
shimata, Kitamura, Kimura, Kasahara\n\nIndirect memory accesses caused by
sparse linear algebra calculations are widely used in important real appli
cations. However, they also cause serious inefficient memory accesses and
pipeline stalls resulting in low execution efficiency even with high memo
ry bandwidth and much computational resourc...\n\n---------------------\nP
erformance Impact of Memory Channels on Sparse and Irregular Algorithms\n\
nGreen, Fox, Young, Shirako, Bader\n\nGraph processing is typically consid
ered to be a memory-bound rather than compute-bound problem. One common li
ne of thought is that more available memory bandwidth corresponds to bette
r graph processing performance. However, in this work we demonstrate that
the key factor in the utilization of the m...\n\n---------------------\nCo
nveyors for Streaming Many-to-Many Communication\n\nMaley, DeVinney\n\nWe
report on a software package that offers high-bandwidth and memory-efficie
nt ways for a parallel application to transmit numerous small data items a
mong its processes. The package provides a standalone library that can in
tegrated into any SHMEM, UPC, or MPI application. It defines a simple int
e...\n\n---------------------\nStretching Jacobi: A Two-Stage Pivoting App
roach for Block-Based Factorization\n\nThuerck\n\nGiven enough parallel co
mpute units, the processing time for sparse (in-)complete factorization is
determined by the number of level sets found in the pattern of the matrix
. For positive definite or diagonal dominant matrices, the iterative Jacob
i-method presents itself as a scalable alternative on ...\n\n-------------
--------\nA Mixed Precision Multicolor Point-Implicit Solver for Unstructu
red Grids on GPUs\n\nWalden, Nielsen, Diskin, Zubair\n\nThis paper present
s a new mixed-precision implementation of a linear-solver kernel used in p
ractical large-scale CFD simulations to improve GPU performance. The new i
mplementation reduces memory traffic by using the half-precision format fo
r some critical computations while maintaining double-precis...\n\n-------
--------------\nIA^3 2019 Morning Break\n\n\n\n---------------------\nIA^3
2019 Lunch Break\n\n\n\n---------------------\nDebate\n\nRaugas, Sorensen
, Ahmed, Beard, Verbanescu\n\n---------------------\niPregel: Strategies t
o Deal with an Extreme Form of Irregularity in Vertex-Centric Graph Proces
sing\n\nCapelli, Brown, Bull\n\nOver the last decade, the vertex-centric p
rogramming model has attracted significant attention in the world of graph
processing, resulting in the emergence of a number of vertex-centric fra
meworks. Its simple programming interface, where computation is expressed
from a vertex point of view, offers ...\n\n---------------------\nA Hardwa
re Prefetching Mechanism for Vector Gather Instructions\n\nTakayashiki, Sa
to, Komatsu, Kobayashi\n\nVector gather instructions are responsible for h
andling indirect memory accesses in vector processing. Since the indirect
memory accesses usually express irregular access patterns, they have relat
ively low spatial and temporal locality compared with regular access patte
rns. As a result, an applicati...\n\n---------------------\nExtending a Wo
rk-Stealing Framework with Priorities and Weights\n\nNakashima, Yoritaka,
Yasugi, Hiraishi, Umatani\n\nThis paper proposes priority- and weight-base
d steal strategies for an idle worker (thief) to select a victim worker in
work-stealing frameworks. Typical work-stealing frameworks employ uniform
ly random victim selection. We implemented the proposed strategies on a wo
rk-stealing framework called Tasc...\n\n---------------------\nIA^3 2019 A
fternoon Break\n\n\n\n---------------------\nMixed-Precision Tomographic R
econstructor Computations on Hardware Accelerators\n\nDoucet, Ltaief, Grat
adour, Keyes\n\nThe computation of tomographic reconstructors (ToR) is at
the core of a simulation framework to design the next generation of adapti
ve optics (AO) systems to be installed on future Extremely Large Telescope
s (ELT). In fact, it is also a critical component for their operation on s
ky. The goals of th...\n\n\nTag: Workshop Reg Pass, Algorithms, Architect
ures, Irregular Applications\n\nRegistration Category: Workshop Reg Pass,
Algorithms, Architectures, Irregular Applications
END:VEVENT
END:VCALENDAR