Tutorial Overview
ESESC is a fast multiprocessor simulator with detailed power, thermal, and performance models for modern out-of-order multicores. A full-day Saturday tutorial about how to use and develop ESESC will be presented at ISCA 2014 on June 14, 2014.
The tutorial will provide a comprehensive overview of ESESC while also providing detailed coverage of improvements in the memory and cache coherence modeling, and a new GPU simulation methodology that combines co-execution using native hardware with simulation using the ESESC timing and power models.
Compared with previous MICRO tutorial, this ISCA tutorial presents new topics like GPU modeling, validation, and the new memory hierarchy.
Schedule
- 8:00 - 8:30 Breakfast
- 8:30 - 9:00 Overview
- 9:00 - 9:30 Building and Running
- 9:30 - 10:00 Validation
- 10:00 - 10:30 Break
- 10:30 - 11:30 Memory Hierarchy and Cache Coherence
- 11:30 - 12:00 Power
- 12:00 - 1:00 Lunch
- 1:00 - 2:00 GPU modeling
- 2:00 - 3:00 Sampling
- 3:00 - 3:30 Break
- 3:30 - 4:00 Thermal
- 4:00 - 5:00 Recent Projects and Wrap-up / Q&A
Overview
A high level overview of how ESESC is designed, how it has been used, and how to use ESESC for your own research.
Building and Running
A hands-on session to quickly get ESESC built and running so that attendees can follow along during demos throughout the tutorial
Validation
A summary of recent work to validate the simulation results generated by ESESC with ones from real hardware (ARM Samsung chromebook).
Memory Hierarchy and Cache Coherence
An in-depth look at how to configure the simulated memory hiearchy in ESESC and how ESESC models local caches and cache coherence.
Power Model
Detailed explanation of how ESESC integrates the use of McPAT for calculating power consumption of simulated systems.
GPU Modeling
The first ever presentation of how ESESC is able to model GPU workloads by combining co-execution of CUDA applications on native hardware with a software performance timing model running in ESESC.
Thermal Model
An overview of the thermal modeling that is integrated in ESESC and an explanation of the associated configuration parameters.
Sampling Methods
A detailed explanation of how ESESC implements statistical sampling to improve simulation speed, with particular focus on the Time Based Sampling method that was presented at HPCA 2013.
Wrap-up
Conclusion that will address topics of interest raised by attendees and also survey recent work on ESESC and how ESESC is being used by the research community.