Difference between revisions of "CSE591 Low Power Architecture"
From esoterum.org
Line 1: | Line 1: | ||
− | ==29 Jan 07 Presentation I== | + | == 29 Jan 07 Presentation I == |
[http://www.esoterum.org/mw/images/0/04/Instruction-level-power-dissipation.pdf Instruction Level Power Dissipation in the Intel XScale Embedded Microprocessor] | [http://www.esoterum.org/mw/images/0/04/Instruction-level-power-dissipation.pdf Instruction Level Power Dissipation in the Intel XScale Embedded Microprocessor] | ||
Line 10: | Line 10: | ||
:*[http://tab.computer.org/tcca/NEWS/jan2001/irwin.pdf SimplePower "A cycle accurate energy simulator"] | :*[http://tab.computer.org/tcca/NEWS/jan2001/irwin.pdf SimplePower "A cycle accurate energy simulator"] | ||
− | ==5 Feb 07 Project I Topic due== | + | == 5 Feb 07 Project I Topic due == |
− | ==12 Feb 07 Project I due== | + | == 12 Feb 07 Project I due == |
*[http://www.intel.com/design/PentiumD/documentation.htm Pentium D specifications (65nm, up to 5GHz)] | *[http://www.intel.com/design/PentiumD/documentation.htm Pentium D specifications (65nm, up to 5GHz)] | ||
:-[http://www.intel.com/design/packtech/ch_04.pdf Thermal/Failure Rate relationships, Fig. 4-27] | :-[http://www.intel.com/design/packtech/ch_04.pdf Thermal/Failure Rate relationships, Fig. 4-27] | ||
− | ===[[Simulation Software]]=== | + | === [[Simulation Software]] === |
− | ===DVFS Project Resources=== | + | === DVFS Project Resources === |
*[http://www.eecg.toronto.edu/~tamda/csc2228/ Dynamic Voltage Scaling in Mobile Devices] | *[http://www.eecg.toronto.edu/~tamda/csc2228/ Dynamic Voltage Scaling in Mobile Devices] | ||
::Concise references for DVFS project from 2003 using speedstep on an IBM Thinkpad | ::Concise references for DVFS project from 2003 using speedstep on an IBM Thinkpad | ||
Line 28: | Line 28: | ||
− | ==19 Feb 07 Presentation II== | + | == 19 Feb 07 Presentation II == |
*[http://www.public.asu.edu/~ashriva6/teaching/LPCA/Readings/05B-01.pdf Mitigating Amdahl’s Law Through EPI Throttling] | *[http://www.public.asu.edu/~ashriva6/teaching/LPCA/Readings/05B-01.pdf Mitigating Amdahl’s Law Through EPI Throttling] | ||
Line 42: | Line 42: | ||
[http://en.wikipedia.org/wiki/Wall_clock_time Wall clock time] | [http://en.wikipedia.org/wiki/Wall_clock_time Wall clock time] | ||
− | == Project I Take II == | + | == 5 March 07 Project I Take II == |
'''Considerations:''' | '''Considerations:''' |
Revision as of 15:42, 2 March 2007
Contents
29 Jan 07 Presentation I
Instruction Level Power Dissipation in the Intel XScale Embedded Microprocessor
Tools for power analysis of microprocessors at the microarchitectural level:
- This software is based on the SimpleScalar software
- Download Wattch
5 Feb 07 Project I Topic due
12 Feb 07 Project I due
Simulation Software
DVFS Project Resources
- Concise references for DVFS project from 2003 using speedstep on an IBM Thinkpad
- Good linux kernel reference
Project II
- XScale project using
PXA25080200 already in Dr. Shrivastava's possession.
19 Feb 07 Presentation II
- -Energy Efficient Co-Adaptive Instruction Fetch and Issue (Throttling paper)
- -Instruction Flow-Based Front-end Throttling for Power-Aware High-Performance Processors (Throttling paper)
- Amdahl's Law (Wikipedia)
- Intel article on EPI throttling
- Best of Both Latency and Throughput (introduces EPI Throttling)
Vocabulary
5 March 07 Project I Take II
Considerations:
- Existing power reduction schemes for LCDs and thier applicability to Electrophoretic Displays (EPDs)
- Dynamic Backlight and Contrast Scaling
- The advantages of bi-stability of EPD over constant refresh of LCD
- Possibility of redraw circuitry for bi-stable display which only addressed pixels which need to be updated
- Relevance of the backlight in LCD in comparison with "front-light" in EPD
- Best-case/Worst-case for each display and in what circumstances are they opposed, in what circumstances are they matched
- Power characterization of power consumption of the LCD vs. the EPD
- Power Modeling for LCD and EPD
- Traditionally pixel switching has been ignored because it is insignificant compared to backlight and control circuitry, is the EPD a different case
- LCD switching power modelled as capacitors (1.16, 1.5, 1.10)
- EPD switching power modelled as resistor (1.12)