STE||AR Group Summer Mini Symposium

The STE||AR Group held an online Mini Symposium July 7 – 9, 2020 beginning each day 1:00 pm CDT showcasing our students’ and researchers’ presentations on their various research and topics of interest.

Below are the topics, presenters, and links to presentation slides, as well as the video recordings, which have been posted to YouTube.

Tuesday 7/7

Nanmiao WuDistributed Alternating Least Square Algorithm in Phylanx

In this presentation, I will introduce distributed alternating least square algorithm and how we implement it in Phylanx. 


Shahrzad ShirzadSplittable Tasks

In this talk I will briefly talk about Splittable Tasks and their implementation in HPX. Splittable tasks are tasks that could be partitioned into smaller tasks, when sufficient parallelism is available and are used as an adaptive runtime method to control task granularity. I will discuss strategies for splitting the tasks, which would result in creating different number of tasks, with different sizes, and will show the obtained results for a simple benchmark.


Steve Brandt Jetlag

The Phylanx Project combines a number of tools: HPX, Python, Apex, and Traveler. The unified workspace is called JetLag. In this talk, I will provide an overview of how it works and what it’s capable of.


Avah Banerjee – Moderator


Wednesday 7/8

Avah Banerjee – ModeratorThe Tiling Problem: An Update

In this talk I will revisit the tiling problem in the context of dense distributed matrices. We give a greedy algorithm based on a preference order. This is determined by a layered decomposition of the program expression graph. I will also discuss the work remains for improving and testing the algorithm using Phylanx.


Rod TohidJIT Code Generation for PhySL Expression-Trees

Introduce the Phyfleaux [1] framework, demonstrating how Phyfleaux’s high-productivity Python API is used to transform and generate “optimized” code for PhySL expression-trees.

[1] https://github.com/stellar-group/phyfleaux


Thursday 7/9

Weile WeiIntegrating Hazard Pointers of LibCDS with HPX threading

This talk will introduce part of my work in Google Summer of Code: how to integrate HPX user-level threading to support hazard pointers, a safe memory reclamation (SMR) algorithm, provided by LibCDS (Library for Concurrent Data Structures). 


Bita HasheminezhadDistributed Convolutions

In this talk, I will introduce Data and Spatial parallelization methods of 1D Convolutions in Phylanx using tiled arrays.


Alireza KheirkhahanZFS Experience in Rostam

In this talk we describe some of shortcomings of conventional storage solutions and discuss how ZFS tries to address those. Also we talk about practical challenges of adopting ZFS in Linux environment.


Avah Banerjee – Moderator


Tuesday 7/7

Wednesday 7/8

Thursday 7/9