School of computer science
Georgia Institute of Technology
CS6290HPCA Fall 2011
Paper survey and Project descriptions
Prof. Hyesoon Kim
Schedule
- Team members and tentative title: 11/8 (Tu)
Topics & 3~5-min individual group presentation: 11/10 (Th)
- 1st Progress report (combined paper survey and project): 11/17 (Th)
- 2nd Progress report (combined paper survey and project): 12/1 (Th)
- Paper survey report: 12/8 (Th) (extension can be granted upon a request)
- Project final report: 12/8 (Th) (extension can be granted upon a request)
Teams
Students may choose to work alone or in teams of up to three. The amount of work must be more ambitious as you have more people, and all students will get the same grade within a team.
Possible topics
You can choose any topic that is relevant to computer architecture. You can choose the same topic for your term project
and paper survey or you can choose different topics. However, I think the same topic would be easier to work.
Here are some suggestions of topics. One of the important goal of the paper survey is finding the relevant work and compare them. A collection of short summaries of papers will not receive a good credit.
The paper surveys can discuss the differences between different papers and also survey the latest of the
technology. For example, you can survey cache replacement policy. You can discuss what kind of
polices have been proposed and what the most challenging problems are.
Possible topics
- DRAM memory scheduling algorithms
- Cache partitioning problems in SMP
- PCM/Non-volatile memory
- Instruction fetch problems
- S/W and H/W Race detection
- Speculative threading
- Transactional memory
- GPUs
- Performance modeling
- Flash memory/SSD
- Reducing simulation time
- Data center architecture
- Architecture support for virtualization
- Prefetching
- Dark Silicon
- Energy-efficient compiler
- Reliability problems
- Hardware security
- 3D stacking
- Photonics interconnection
- Branch prediction
- Indirect branch prediction paper1 , paper2
- Branch prediction using machine learning paper
- Security
- AEGIS paper
- Dynamic information flow tracking paper
- Power
- DRAM power paper1 , paper2
- DVFS and resource managements : ISCA 2009 Session 7A papers
- Interconnections
- on-chip interconnections paper
CMP resource managements
- Dynamic cache partitioning paper
- Thread scheduling in heterogeneous architectures
Fault tolerant architecture
- Application-Level Correctness and its Impact on Fault Tolerance paper
- Fault screening paper
Additional Project ideas
For the project, you have to implement the ideas in the simulator (your project simulator) or other architecture simulators such as MacSim that is developed from Prof. Kim's research group. (If you are interested in using MacSim, please contact Prof. Kim directly.) Or you can implement the idea in other simulators or other systems. The project should have some level of originality, (even if you are comparing previously proposed techniques, it should include some aspects such as new workloads or some new architecture characteristics.)
Some additional potential topics:
Dynamic Simization:
Instead of compile time, at run-time, can we identify a trace of instructions that can be vectorized?
Scalable cache partitioning mechanisms:
there are many cache partitioning schemes but will they work in many-core systems?
High performance computing in dark-silicon area with reconfigurable architecture.
Project grading guidelines
- Topic: (5%)
- 1st progress report: (10%) (about 2 pages 11 pts double column)
- 2nd progress report (25%) (about 3-4 pages 11 pts double column)
- Project Final report (60%) (6-8 pages 11 pts double column) (IEEE or ACM conference style)
- Paper survey final report (60%) (2-4 pages 11 pts double column) (IEEE or ACM conference style)
- Progress report writing guideline
- What's the goal of the paper survey/project report?
- The descriptions of the problems that the papers (your project) are solving
- list of the paper that you have read so far and short summary
- outline of the final report
- updated paper list or changes in the list
- Initial results of the evaluations
- Infrastructure descriptions
For the format of the final report, please follow computer architecture
conference/workshop papers in IEEE and ACM.