Lifan Xu

TOP-PIM: throughput-oriented programmable processing in memory

As computation becomes increasingly limited by data movement and energy consumption, exploiting locality throughout the memory hierarchy becomes critical to continued performance scaling. Moving computation closer to memory presents an opportunity to reduce both energy and data movement overheads. We explore the use of 3D die stacking to move memory-intensive computations closer to memory. This approach to processing in memory addresses some drawbacks of prior research on in-memory computing and is commercially viable in the foreseeable future. Because 3D stacking provides increased bandwidth, we study throughput-oriented computing using programmable GPU compute units across a broad range of benchmarks, including graph and HPC applications. We also introduce a methodology for rapid design space exploration by analytically predicting performance and energy of in-memory processors based on metrics obtained from execution on todays GPU hardware. Our results show that, on average, viable PIM configurations show moderate performance losses (27%) in return for significant energy efficiency improvements (76\% reduction in EDP) relative to a representative mainstream GPU at 22nm technology. At 16nm technology, on average, viable PIM configurations are performance competitive with a representative mainstream GPU (7% speedup) and provide even greater energy efficiency improvements (85\% reduction in EDP).

Hierarchical fractional-step approximations and parallel kinetic Monte Carlo algorithms

We present a mathematical framework for constructing and analyzing parallel algorithms for lattice kinetic Monte Carlo (KMC) simulations. The resulting algorithms have the capacity to simulate a wide range of spatio-temporal scales in spatially distributed, non-equilibrium physiochemical processes with complex chemistry and transport micro-mechanisms. Rather than focusing on constructing exactly the stochastic trajectories, our approach relies on approximating the evolution of observables, such as density, coverage, correlations and so on. More specifically, we develop a spatial domain decomposition of the Markov operator (generator) that describes the evolution of all observables according to the kinetic Monte Carlo algorithm. This domain decomposition corresponds to a decomposition of the Markov generator into a hierarchy of operators and can be tailored to specific hierarchical parallel architectures such as multi-core processors or clusters of Graphical Processing Units (GPUs). Based on this operator decomposition, we formulate parallel Fractional step kinetic Monte Carlo algorithms by employing the Trotter Theorem and its randomized variants; these schemes, (a) are partially asynchronous on each fractional step time-window, and (b) are characterized by their communication schedule between processors. The proposed mathematical framework allows us to rigorously justify the numerical and statistical consistency of the proposed algorithms, showing the convergence of our approximating schemes to the original serial KMC. The approach also provides a systematic evaluation of different processor communicating schedules. We carry out a detailed benchmarking of the parallel KMC schemes using available exact solutions, for example, in Ising-type systems and we demonstrate the capabilities of the method to simulate complex spatially distributed reactions at very large scales on GPUs. Finally, we discuss work load balancing between processors and propose a re-balancing scheme based on probabilistic mass transport methods.

Parallelization of tau-leap coarse-grained Monte Carlo simulations on GPUs

The Coarse-Grained Monte Carlo (CGMC) method is a multi-scale stochastic mathematical and simulation framework for spatially distributed systems. CGMC simulations are important tools for studying phenomena such as catalysis, crystal growth, surface diffusion, phase transitions on single crystals, and cell membrane receptor dynamics. In parallel CGMC, the tau-leap method is used for parallel simulations that are executed on traditional CPU clusters in a master-slave setting. Unfortunately the communications between master and slaves negatively impact speedup and scalability. In this paper, we explore the potentials of GPUs for the tau-leap method and we present an extensive performance evaluation that leads to the most suitable degree of parallelism for this method under different simulation profiles. We show how the efficient parallelization of the tau-leap method for GPUs includes (1) the redefinition of its data structures, (2) the redesign of its algorithm, and (3) the selection of the most appropriate degree of parallelism (i.e., fine-grained or course-gained) on a single GPU or multiple GPUs. Exceptional performance improvements can thus be achieved for this method.

Fast acceleration of 2D wave propagation simulations using modern computational accelerators.

Recent developments in modern computational accelerators like Graphics Processing Units (GPUs) and coprocessors provide great opportunities for making scientific applications run faster than ever before. However, efficient parallelization of scientific code using new programming tools like CUDA requires a high level of expertise that is not available to many scientists. This, plus the fact that parallelized code is usually not portable to different architectures, creates major challenges for exploiting the full capabilities of modern computational accelerators. In this work, we sought to overcome these challenges by studying how to achieve both automated parallelization using OpenACC and enhanced portability using OpenCL. We applied our parallelization schemes using GPUs as well as Intel Many Integrated Core (MIC) coprocessor to reduce the run time of wave propagation simulations. We used a well-established 2D cardiac action potential model as a specific case-study. To the best of our knowledge, we are the first to study auto-parallelization of 2D cardiac wave propagation simulations using OpenACC. Our results identify several approaches that provide substantial speedups. The OpenACC-generated GPU code achieved more than speedup above the sequential implementation and required the addition of only a few OpenACC pragmas to the code. An OpenCL implementation provided speedups on GPUs of at least faster than the sequential implementation and faster than a parallelized OpenMP implementation. An implementation of OpenMP on Intel MIC coprocessor provided speedups of with only a few code changes to the sequential implementation. We highlight that OpenACC provides an automatic, efficient, and portable approach to achieve parallelization of 2D cardiac wave simulations on GPUs. Our approach of using OpenACC, OpenCL, and OpenMP to parallelize this particular model on modern computational accelerators should be applicable to other computational models of wave propagation in multi-dimensional media.

HADM: Hybrid Analysis for Detection of Malware

Android is the most popular mobile operating system with a market share of over 80% [1]. Due to its popularity and also its open source nature, Android is now the platform most targeted by malware, creating an urgent need for effective defense mechanisms to protect Android-enabled devices.

Parallelization of Machine Learning Applied to Call Graphs of Binaries for Malware Detection

Malicious applications have become increasingly numerous. This demands adaptive, learning-based techniques for constructing malware detection engines, instead of the traditional manual-based strategies. Prior work in learning-based malware detection engines primarily focuses on dynamic trace analysis and byte-level n-grams. Our approach in this paper differs in that we use compiler intermediate representations, i.e., the callgraph representation of binaries. Using graph-based program representations for learning provides structure of the program, which can be used to learn more advanced patterns. We use the Shortest Path Graph Kernel (SPGK) to identify similarities between call graphs extracted from binaries. The output similarity matrix is fed into a Support Vector Machine (SVM) algorithm to construct highly-accurate models to predict whether a binary is malicious or not. However, SPGK is computationally expensive due to the size of the input graphs. Therefore, we evaluate different parallelization methods for CPUs and GPUs to speed up this kernel, allowing us to continuously construct up-to-date models in a timely manner. Our hybrid implementation, which leverages both CPU and GPU, yields the best performance, achieving up to a 14.2x improvement over our already optimized OpenMP version. We compared our generated graph-based models to previously state-of-the-art feature vector 2-gram and 3-gram models on a dataset consisting of over 22,000 binaries. We show that our classification accuracy using graphs is over 19% higher than either n-gram model and gives a false positive rate (FPR) of less than 0.1%. We are also able to consider large call graphs and dataset sizes because of the reduced execution time of our parallelized SPGK implementation.

Dynamic Android Malware Classification Using Graph-Based Representations

Malware classification for the Android ecosystem can be performed using a range of techniques. One major technique that has been gaining ground recently is dynamic analysis based on system call invocations recorded during the executions of Android applications. Dynamic analysis has traditionally been based on converting system calls into flat feature vectors and feeding the vectors into machine learning algorithms for classification. In this paper, we implement three traditional feature-vector-based representations for Android system calls. For each feature vector representation, we also propose a novel graph-based representation. We then use graph kernels to compute pair-wise similarities and feed these similarity measures into a Support Vector Machine (SVM) for classification. To speed up the graph kernel computation, we compress the graphs using the Compressed Row Storage format, and then we apply OpenMP to parallelize the computation. Experiments show that the graph-based representations are able to improve the classification accuracy over the corresponding feature-vector-based representations from the same input. Finally we show that different representations can be combined together to further improve classification accuracy.