The PARSEC Benchmark Suite: Characterization and Architectural Implications (2008)
A benchmark suite for studies of CMPs (Chip-Multiprocessors)
Diverse in working set, locality, data sharing, synchronization, and off-chip traffic.
existing benchmark suites cannot be considered adequate to describe future CMP applications.
■ Motivation
□ Requirements for a Benchmarks Suite
Multi-threaded Applications
Emerging Workloads
Diverse
Employ State-of-Art Techniques
Support Research
□ Limitations of Existing Benchmark Suites
SPLASH-2
Program collection is skewed towards HPC and graphics programs.
Does not include parallelization models such as the pipeline model.
SPEC CPU2006 and OMP2001
Provide a snapshot of current scientific and engineering applications
Workloads such as systems programs and parallelization models which employ the producer-consumer model are not included.
SPEC CPU2006 is a suite of serial programs.
Other Benchmark Suites
Designed to study a specific program area and limited to a single application domain.
■ The PARSEC Benchmark Suite
9 applications and 3 kernels (+PARSEC 2.0 includes RayTrace)
□ Input Sets
test, simdev, simsmall, simmedium, simlarge, native
□ Workloads
Blackscholes (Financial Analysis)
This application is an Intel RMS benchmark. It calculates the prices for a portfolio of European options analytically with the Black-Scholes partial differential equation (PDE). There is no closed-form expression for the Black-Scholes equation and as such it must be computed numerically.
Bodytrack (Computer Vision)
This computer vision application is an Intel RMS workload which tracks a human body with multiple cameras through an image sequence. This benchmark was included due to the increasing significance of computer vision algorithms in areas such as video surveillance, character animation and computer interfaces.
Canneal (Engineering)
This kernel was developed by Princeton University. It uses cache-aware simulated annealing (SA) to minimize the routing cost of a chip design. Canneal uses fine-grained parallelism with a lock-free algorithm and a very aggressive synchronization strategy that is based on data race recovery instead of avoidance.
Dedup (Enterprise Storage)
This kernel was developed by Princeton University. It compresses a data stream with a combination of global and local compression that is called 'deduplication'. The kernel uses a pipelined programming model to mimic real-world implementations. The reason for the inclusion of this kernel is that deduplication has become a mainstream method for new-generation backup storage systems.
Facesim (Animation)
This Intel RMS application was originally developed by Stanford University. It computes a visually realistic animation of the modeled face by simulating the underlying physics. The workload was included in the benchmark suite because an increasing number of animations employ physical simulation to create more realistic effects.
Ferret (Similarity Search)
This application is based on the Ferret toolkit which is used for content-based similarity search. It was developed by Princeton University. The reason for the inclusion in the benchmark suite is that it represents emerging next-generation search engines for non-text document data types. In the benchmark, we have configured the Ferret toolkit for image similarity search. Ferret is parallelized using the pipeline model.
Fluidanimate (Animation)
This Intel RMS application uses an extension of the Smoothed Particle Hydrodynamics (SPH) method to simulate an incompressible fluid for interactive animation purposes. It was included in the PARSEC benchmark suite because of the increasing significance of physics simulations for animations.
Freqmine (Data Mining)
This application employs an array-based version of the FP-growth (Frequent Pattern-growth) method for Frequent Itemset Mining (FIMI). It is an Intel RMS benchmark which was originally developed by Concordia University. Freqmine was included in the PARSEC benchmark suite because of the increasing use of data mining techniques.
Raytrace (PARSEC 2.0 추가됨, Graphics?)
The Intel RMS application uses a version of the raytracing method that would typically be employed for real-time animations such as computer games. It is optimized for speed rather than realism. The computational complexity of the algorithm depends on the resolution of the output image and the scene.
Streamcluster (Data Mining)
This RMS kernel was developed by Princeton University and solves the online clustering problem. Streamcluster was included in the PARSEC benchmark suite because of the importance of data mining algorithms and the prevalence of problems with streaming characteristics.
Swaptions (Financial Analysis)
The application is an Intel RMS workload which uses the Heath-Jarrow-Morton (HJM) framework to price a portfolio of swaptions. Swaptions employs Monte Carlo (MC) simulation to compute the prices.
Vips (Media Processing)
This application is based on the VASARI Image Processing System (VIPS) which was originally developed through several projects funded by European Union (EU) grants. The benchmark version is derived from a print on demand service that is offered at the National Gallery of London, which is also the current maintainer of the system. The benchmark includes fundamental image operations such as an affine transformation and a convolution.
X264 (Media Processing)
This application is an H.264/AVC (Advanced Video Coding) video encoder. H.264 describes the lossy compression of a video stream and is also part of ISO/IEC MPEG-4. The flexibility and wide range of application of the H.264 standard and its ubiquity in next-generation video systems are the reasons for the inclusion of x264 in the PARSEC benchmark suite.
■ Methodology
Parallelization
Working sets and locality
Communication to computation ratio and sharing
Off-chip traffic
'Architecture' 카테고리의 다른 글
| Interconnection Network Topologies (0) | 2009.07.01 |
|---|---|
| PARSEC vs. SPLASH-2 (0) | 2009.06.16 |
| The SPLASH-2 Programs (0) | 2009.06.05 |
| CMP vs. SMP (0) | 2009.05.26 |
| Evaluating MapReduce for Multi-core and Multiprocessor Systems (0) | 2009.05.25 |
민둥