FYI. KTH Forum building is at Isafjordsgatan 39, Kista, Stockholm.

________________________________________
From: Mats Brorsson
Sent: Saturday, June 05, 2010 7:33 AM
To: [email protected]; [email protected]; '[email protected]' ([email protected]); multicore-
[email protected]
Cc: Sebastien Bourdeauducq; Artur Podobas; Ananya Muddukrishna
Subject: Two master thesis presentations, June 8: VJ application on an FPGA 
and thermal-aware multicore scheduling

You are all cordially welcome to the following master thesis presentations 
next Tuesday afternoon, June 8. Both presentations will take place in room 
Gemini at KTH Forum building. Elevator C, level 8.

Number 1: 15.00

Title: A Performance-driven SoC for Video Synthesis
Author: Sébastien Bourdeauducq
Opponent: Ananya Muddukrishna

Abstract:
Commercial system-on-chips with advanced graphics acceleration capabilities 
are becoming ubiquitous today. However, in contradiction with the open source 
idea, little is known about the details of their architecture and 
implementation, as they are usually covered by trade secrets.

Fostered by the falling costs of high-density FPGAs, our thesis project 
encompasses researching, developing and implementing the key points of the 
architecture of an open source and comprehensive system-on-chip with 
competitive yet reasonable graphics capabilities. The chosen target 
application is the synthesis of visual effects similar to those produced by 
the popular MilkDrop visualization plug-in for Winamp.

Our system-on-chip design consists principally of a custom bus infrastructure, 
a custom DDR SDRAM memory controller, a microprocessor core, and custom 
graphics accelerators for texture mapping and floating point processing.

Our base microprocessor system is capable of running Linux (without MMU) and 
outperforms a Microblaze-based solution tested in similar conditions by a 15 
to 35% increase in speed of execution. For our video synthesis application, 
our texture mapping accelerator achieves an average fill rate of 44 megapixels 
per second and our floating point processing unit provides in excess of 70 
million floating point operations per second. Everything, including I/O 
peripherals (AC97 audio, Ethernet, RS232 UART, GPIO), is implemented on a 
Virtex-4 XC4VLX25 FPGA, where it utilizes about 80% of the resources.

Finally, we have successfully developed an embedded video synthesis program 
that leverages the possibilities of our hardware architecture to permit the 
live rendering of many MilkDrop effects in 640x480 resolution at 30 frames per 
second.

Number 2: 16:00

Title: Thermal-aware scheduling in OpenMP
Author: Artur Podobas
Opponent: Sébastien Bourdeauducq


Abstract:  Multi-core computer systems have widely been accepted as the future 
standard in computer architecture. Many simple processing cores bundled 
together on a single die have been shown more successful in terms of power 
consumption and execution performance compared to previous large and heavily 
pipelined uni-core systems. And as the chip dimensions decrease, temperature 
effects starts to become prominent. Elevated temperature gradients and hotspot 
on processors put an upper limit on both the execution performance and the 
life-time of the devices, leading to failures and eventually malfunction.



Most work involved in software-based temperature management in multi-core 
system has been in the kernel-space, hidden from the user. This work hopes to 
change this, and show that it is possible to make user-level schedulers 
account for temperature changes occurring in the system. OpenMP is the current 
standard in parallel programming and was used to implement a scheduling policy 
that uses hardware feedback to effectively try to eliminate elevated 
temperature in the chip. The system itself was simulated and modeled using 
well-established simulators and models.



The results were promising, showing a decrease in time spent above the 
critical temperature with up to 140 times in some benchmarks and a decreased 
power consumption in all the benchmarks as compared to the cilk and breadth-
first scheduler. This alone should be encourage to more research in this area, 
to hopefully give rise to a future standard of user-level temperature control 
in parallel based schedulers.




Mats Brorsson
--
Professor of Computer Architecture
http://web.ict.kth.se/~matsbror
_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkym...@freenode
Webchat: www.milkymist.org/irc.html
Wiki: www.milkymist.org/wiki

Reply via email to