Zachary DeVito

Liszt: a domain specific language for building portable mesh-based PDE solvers

Heterogeneous computers with processors and accelerators are becoming widespread in scientific computing. However, it is difficult to program hybrid architectures and there is no commonly accepted programming model. Ideally, applications should be written in a way that is portable to many platforms, but providing this portability for general programs is a hard problem. By restricting the class of programs considered, we can make this portability feasible. We present Liszt, a domain- specific language for constructing mesh-based PDE solvers. We introduce language statements for interacting with an unstructured mesh, and storing data at its elements. Pro- gram analysis of these statements enables our compiler to expose the parallelism, locality, and synchronization of Liszt programs. Using this analysis, we generate applications for multiple platforms: a cluster, an SMP, and a GPU. This approach allows Liszt applications to perform within 12% of hand-written C++, scale to large clusters, and experience order-of-magnitude speedups on GPUs.

Exploring Traditional and Emerging Parallel Programming Models Using a Proxy Application

Parallel machines are becoming more complex with increasing core counts and more heterogeneous architectures. However, the commonly used parallel programming models, C/C++ with MPI and/or OpenMP, make it difficult to write source code that is easily tuned for many targets. Newer language approaches attempt to ease this burden by providing optimization features such as automatic load balancing, overlap of computation and communication, message-driven execution, and implicit data layout optimizations. In this paper, we compare several implementations of LULESH, a proxy application for shock hydrodynamics, to determine strengths and weaknesses of different programming models for parallel computation. We focus on four traditional (OpenMP, MPI, MPI+OpenMP, CUDA) and four emerging (Chapel, Charm++, Liszt, Loci) programming models. In evaluating these models, we focus on programmer productivity, performance and ease of applying optimizations.

Terra: a multi-stage language for high-performance computing

High-performance computing applications, such as auto-tuners and domain-specific languages, rely on generative programming techniques to achieve high performance and portability. However, these systems are often implemented in multiple disparate languages and perform code generation in a separate process from program execution, making certain optimizations difficult to engineer. We leverage a popular scripting language, Lua, to stage the execution of a novel low-level language, Terra. Users can implement optimizations in the high-level language, and use built-in constructs to generate and execute high-performance Terra code. To simplify meta-programming, Lua and Terra share the same lexical environment, but, to ensure performance, Terra code can execute independently of Luas runtime. We evaluate our design by reimplementing existing multi-language systems entirely in Terra. Our Terra-based auto-tuner for BLAS routines performs within 20% of ATLAS, and our DSL for stencil computations runs 2.3x faster than hand-written C.

Darkroom: compiling high-level image processing code into hardware pipelines

Specialized image signal processors (ISPs) exploit the structure of image processing pipelines to minimize memory bandwidth using the architectural pattern of line-buffering, where all intermediate data between each stage is stored in small on-chip buffers. This provides high energy efficiency, allowing long pipelines with tera-op/sec. image processing in battery-powered devices, but traditionally requires painstaking manual design in hardware. Based on this pattern, we present Darkroom, a language and compiler for image processing. The semantics of the Darkroom language allow it to compile programs directly into line-buffered pipelines, with all intermediate values in local line-buffer storage, eliminating unnecessary communication with off-chip DRAM. We formulate the problem of optimally scheduling line-buffered pipelines to minimize buffering as an integer linear program. Finally, given an optimally scheduled pipeline, Darkroom synthesizes hardware descriptions for ASIC or FPGA, or fast CPU code. We evaluate Darkroom implementations of a range of applications, including a camera pipeline, low-level feature detection algorithms, and deblurring. For many applications, we demonstrate gigapixel/sec. performance in under 0.5mm2 of ASIC silicon at 250 mW (simulated on a 45nm foundry process), real-time 1080p/60 video processing using a fraction of the resources of a modern FPGA, and tens of megapixels/sec. of throughput on a quad-core x86 processor.Specialized image signal processors (ISPs) exploit the structure of image processing pipelines to minimize memory bandwidth using the architectural pattern of line-buffering, where all intermediate data between each stage is stored in small on-chip buffers. This provides high energy efficiency, allowing long pipelines with tera-op/sec. image processing in battery-powered devices, but traditionally requires painstaking manual design in hardware. Based on this pattern, we present Darkroom, a language and compiler for image processing. The semantics of the Darkroom language allow it to compile programs directly into line-buffered pipelines, with all intermediate values in local line-buffer storage, eliminating unnecessary communication with off-chip DRAM. We formulate the problem of optimally scheduling line-buffered pipelines to minimize buffering as an integer linear program. Finally, given an optimally scheduled pipeline, Darkroom synthesizes hardware descriptions for ASIC or FPGA, or fast CPU code. We evaluate Darkroom implementations of a range of applications, including a camera pipeline, low-level feature detection algorithms, and deblurring. For many applications, we demonstrate gigapixel/sec. performance in under 0.5mm2 of ASIC silicon at 250 mW (simulated on a 45nm foundry process), real-time 1080p/60 video processing using a fraction of the resources of a modern FPGA, and tens of megapixels/sec. of throughput on a quad-core x86 processor.

Riposte: a trace-driven compiler and parallel VM for vector code in R

There is a growing utilization gap between modern hardware and modern programming languages for data analysis. Due to power and other constraints, recent processor design has sought improved performance through increased SIMD and multi-core parallelism. At the same time, high-level, dynamically typed languages for data analysis have become popular. These languages emphasize ease of use and high productivity, but have, in general, low performance and limited support for exploiting hardware parallelism. In this paper, we describe Riposte, a new runtime for the R language, which bridges this gap. Riposte uses tracing, a technique commonly used to accelerate scalar code, to dynamically discover and extract sequences of vector operations from arbitrary R code. Once extracted, we can fuse traces to eliminate unnecessary memory traffic, compile them to use hardware SIMD units, and schedule them to run across multiple cores, allowing us to fully utilize the available parallelism on modern shared-memory machines. Our evaluation shows that Riposte can run vector R code near the speed of hand-optimized C, 5–50× faster than the open source implementation of R, and can also linearly scale to 32 cores for some tasks. Across 12 different workloads we achieve an overall average speed-up of over 150× without explicit programmer parallelization.

First-class runtime generation of high-performance types using exotypes

We introduce exotypes, user-defined types that combine the flexibility of meta-object protocols in dynamically-typed languages with the performance control of low-level languages. Like objects in dynamic languages, exotypes are defined programmatically at run-time, allowing behavior based on external data such as a database schema. To achieve high performance, we use staged programming to define the behavior of an exotype during a runtime compilation step and implement exotypes in Terra, a low-level staged programming language. We show how exotype constructors compose, and use exotypes to implement high-performance libraries for serialization, dynamic assembly, automatic differentiation, and probabilistic programming. Each exotype achieves expressiveness similar to libraries written in dynamically-typed languages but implements optimizations that exceed the performance of existing libraries written in low-level statically-typed languages. Though each implementation is significantly shorter, our serialization library is 11 times faster than Kryo, and our dynamic assembler is 3--20 times faster than Googles Chrome assembler.

Ebb: A DSL for Physical Simulation on CPUs and GPUs

Designing programming environments for physical simulation is challenging because simulations rely on diverse algorithms and geometric domains. These challenges are compounded when we try to run efficiently on heterogeneous parallel architectures. We present Ebb, a Domain-Specific Language (DSL) for simulation, that runs efficiently on both CPUs and GPUs. Unlike previous DSLs, Ebb uses a three-layer architecture to separate (1) simulation code, (2) definition of data structures for geometric domains, and (3) runtimes supporting parallel architectures. Different geometric domains are implemented as libraries that use a common, unified, relational data model. By structuring the simulation framework in this way, programmers implementing simulations can focus on the physics and algorithms for each simulation without worrying about their implementation on parallel computers. Because the geometric domain libraries are all implemented using a common runtime based on relations, new geometric domains can be added as needed, without specifying the details of memory management, mapping to different parallel architectures, or having to expand the runtime’s interface. We evaluate Ebb by comparing it to several widely used simulations, demonstrating comparable performance to handwritten GPU code where available, and surpassing existing CPU performance optimizations by up to 9 × when no GPU code exists.Designing programming environments for physical simulation is challenging because simulations rely on diverse algorithms and geometric domains. These challenges are compounded when we try to run efficiently on heterogeneous parallel architectures. We present Ebb, a Domain-Specific Language (DSL) for simulation, that runs efficiently on both CPUs and GPUs. Unlike previous DSLs, Ebb uses a three-layer architecture to separate (1) simulation code, (2) definition of data structures for geometric domains, and (3) runtimes supporting parallel architectures. Different geometric domains are implemented as libraries that use a common, unified, relational data model. By structuring the simulation framework in this way, programmers implementing simulations can focus on the physics and algorithms for each simulation without worrying about their implementation on parallel computers. Because the geometric domain libraries are all implemented using a common runtime based on relations, new geometric domains can be added as needed, without specifying the details of memory management, mapping to different parallel architectures, or having to expand the runtime’s interface. We evaluate Ebb by comparing it to several widely used simulations, demonstrating comparable performance to handwritten GPU code where available, and surpassing existing CPU performance optimizations by up to 9 × when no GPU code exists.

Rigel: flexible multi-rate image processing hardware

Image processing algorithms implemented using custom hardware or FPGAs of can be orders-of-magnitude more energy efficient and performant than software. Unfortunately, converting an algorithm by hand to a hardware description language suitable for compilation on these platforms is frequently too time consuming to be practical. Recent work on hardware synthesis of high-level image processing languages demonstrated that a single-rate pipeline of stencil kernels can be synthesized into hardware with provably minimal buffering. Unfortunately, few advanced image processing or vision algorithms fit into this highly-restricted programming model. In this paper, we present Rigel, which takes pipelines specified in our new multi-rate architecture and lowers them to FPGA implementations. Our flexible multi-rate architecture supports pyramid image processing, sparse computations, and space-time implementation tradeoffs. We demonstrate depth from stereo, Lucas-Kanade, the SIFT descriptor, and a Gaussian pyramid running on two FPGA boards. Our system can synthesize hardware for FPGAs with up to 436 Megapixels/second throughput, and up to 297x faster runtime than a tablet-class ARM CPU.

Opt: A Domain Specific Language for Non-Linear Least Squares Optimization in Graphics and Imaging

Many graphics and vision problems can be expressed as non-linear least squares optimizations of objective functions over visual data, such as images and meshes. The mathematical descriptions of these functions are extremely concise, but their implementation in real code is tedious, especially when optimized for real-time performance on modern GPUs in interactive applications. In this work, we propose a new language, Opt (available under this http URL), for writing these objective functions over image- or graph-structured unknowns concisely and at a high level. Our compiler automatically transforms these specifications into state-of-the-art GPU solvers based on Gauss-Newton or Levenberg-Marquardt methods. Opt can generate different variations of the solver, so users can easily explore tradeoffs in numerical precision, matrix-free methods, and solver approaches. In our results, we implement a variety of real-world graphics and vision applications. Their energy functions are expressible in tens of lines of code, and produce highly-optimized GPU solver implementations. These solver have performance competitive with the best published hand-tuned, application-specific GPU solvers, and orders of magnitude beyond a general-purpose auto-generated solver.Many graphics and vision problems can be expressed as non-linear least squares optimizations of objective functions over visual data, such as images and meshes. The mathematical descriptions of these functions are extremely concise, but their implementation in real code is tedious, especially when optimized for real-time performance on modern GPUs in interactive applications. In this work, we propose a new language, Opt,1 for writing these objective functions over image- or graph-structured unknowns concisely and at a high level. Our compiler automatically transforms these specifications into state-of-the-art GPU solvers based on Gauss-Newton or Levenberg-Marquardt methods. Opt can generate different variations of the solver, so users can easily explore tradeoffs in numerical precision, matrix-free methods, and solver approaches. In our results, we implement a variety of real-world graphics and vision applications. Their energy functions are expressible in tens of lines of code and produce highly optimized GPU solver implementations. These solvers are competitive in performance with the best published hand-tuned, application-specific GPU solvers, and orders of magnitude beyond a general-purpose auto-generated solver.

Just-in-time Length Specialization of Dynamic Vector Code

Dynamically typed vector languages are popular in data analytics and statistical computing. In these languages, vectors have both dynamic type and dynamic length, making static generation of efficient machine code difficult. In this paper, we describe a trace-based just-in-time compilation strategy that performs partial length specialization of dynamically typed vector code. This selective specialization is designed to avoid excessive compilation overhead while still enabling the generation of efficient machine code through length-based optimizations such as vector fusion, vector copy elimination, and the use of hardware SIMD units. We have implemented our approach in a virtual machine for a subset of R, a vector-based statistical computing language. In a variety of workloads, containing both scalar and vector code, we show near autovectorized C performance over a large range of vector sizes.