This paper presents and characterizes Rodinia, a benchmark suite for heterogeneous computing. To help architects study emerging platforms such as GPUs (Graphics Processing Units), Rodinia includes applications and kernels which target multi-core CPU and GPU platforms. The choice of applications is inspired by Berkeley's dwarf taxonomy. Our characterization shows that the Rodinia benchmarks cover a wide range of parallel communication patterns, synchronization techniques and power consumption, and has led to some important architectural insight, such as the growing importance of memory-bandwidth limitations and the consequent importance of data layout.

/pdf/rodinia-a-benchmark-suite-for-heterogeneous-computing-gib3ik78jh.pdf

Rodinia: A benchmark suite for heterogeneous computing

The advent of multicore CPUs and manycore GPUs means that mainstream processor chips are now parallel systems. Furthermore, their parallelism continues to scale with Moore's law. The challenge is to develop mainstream application software that transparently scales its parallelism to leverage the increasing number of processor cores, much as 3D graphics applications transparently scale their parallelism to manycore GPUs with widely varying numbers of cores.

/pdf/scalable-parallel-programming-with-cuda-fx7qj9io5g.pdf

Scalable parallel programming with CUDA

The rapid increase in the performance of graphics hardware, coupled with recent improvements in its programmability, have made graphics hardware a compelling platform for computationally demanding tasks in a wide variety of application domains. In this report, we describe, summarize, and analyze the latest research in mapping general-purpose computation to graphics hardware. We begin with the technical motivations that underlie general-purpose computation on graphics processors (GPGPU) and describe the hardware and software developments that have led to the recent interest in this field. We then aim the main body of this report at two separate audiences. First, we describe the techniques used in mapping general-purpose computation to graphics hardware. We believe these techniques will be generally useful for researchers who plan to develop the next generation of GPGPU algorithms and techniques. Second, we survey and categorize the latest developments in general-purpose application development on graphics hardware. This survey should be of particular interest to researchers who are interested in using the latest GPGPU applications in their systems of interest.

/pdf/a-survey-of-general-purpose-computation-on-graphics-hardware-2phx1q9xxr.pdf

A Survey of General-Purpose Computation on Graphics Hardware

A Survey of General-Purpose Computation on Graphics Hardware.

The advent of multicore CPUs and manycore GPUs means that mainstream processor chips are now parallel systems. Furthermore, their parallelism continues to scale with Moore’s law. The challenge is to develop mainstream application software that transparently scales its parallelism to leverage the increasing number of processor cores, much as 3D graphics applications transparently scale their parallelism to manycore GPUs with widely varying numbers of cores.

https://dl.acm.org/doi/pdf/10.1145/1365490.1365500

Scalable Parallel Programming with CUDA: Is CUDA the parallel programming model that application developers have been waiting for?

Embodiments of end-to-end deep learning systems and methods are disclosed to recognize speech of vastly different languages, such as English or Mandarin Chinese. In embodiments, the entire pipelines of hand-engineered components are replaced with neural networks, and the end-to-end learning allows handling a diverse variety of speech including noisy environments, accents, and different languages. Using a trained embodiment and an embodiment of a batch dispatch technique with GPUs in a data center, an end-to-end deep learning system can be inexpensively deployed in an online setting, delivering low latency when serving users at scale.

End-to-end speech recognition

Author(s): Sengupta, Shubhabrata; Harris, Mark; Garland, Michael; Owens, John D. | Editor(s): Kurzak, Jakub; Bader, David A.; Dongarra, Jack

/pdf/efficient-parallel-scan-algorithms-for-manycore-gpus-2m01x7n8nh.pdf

Efficient Parallel Scan Algorithms for Manycore GPUs

Author(s): Lefohn, Aaron; Sengupta, Shubhabrata; Kniss, Joe M.; Strzodka, Robert; Owens, John D. | Abstract: We present a novel implementation of adaptive shadow maps (ASMs) that performs all shadow lookups and scene analysis on the GPU, enabling interactive rendering with ASMs while moving both the light and camera. Adaptive shadow maps offer a rigorous solution to projective and perspective shadow map aliasing while maintaining the simplicity of a purely image-based technique. The complexity of the ASM data structure, however, has prevented full GPU-based implementations until now. Our approach uses an entirely GPU-based data structure and a blend of graphics and GPU stream programming. We support shadow map effective resolutions up to 131,072 x 131,072 and, unlike previous implementations, provide smooth transitions between resolution levels by trilinearly filtering (mipmapping) the shadow lookups.

/pdf/dynamic-adaptive-shadow-maps-on-graphics-hardware-2yaz0r8bay.pdf

Dynamic adaptive shadow maps on graphics hardware

Author(s): Kniss, Joe M.; Lefohn, Aaron; Strzodka, Robert; Sengupta, Shubhabrata; Owens, John D. | Abstract: We implement an interactive 3D painting application that stores paint in an octree-like GPU-based adaptive data structure. Interactive painting of complex or unparameterized surfaces is an important problem in the digital film community. Many models used in production environments are either difficult to parameterize or are unparameterized implicit surfaces. We address this problem with a system that allows interactive 3D painting of complex, unparameterized models. The included movie demonstrates interactive painting of a 817k polygon model with effective paint resolutions varying between 64^3 to 2048^3. Our implementation differs from previous work in two important ways: first, it uses an adaptive data structure implemented entirely on the GPU, and second, it enables interactive performance with high quality by supporting quadlinear (mipmapped) filtering and fast, constant-time data accesses.

/pdf/octree-textures-on-graphics-hardware-2vghmxeltp.pdf

Octree textures on graphics hardware

Author(s): Owens, John D.; Sengupta, Shubhabrata; Horn, Daniel | Abstract: In this report we analyze the performance of the fast Fourier transform (FFT) on graphics hardware (the GPU), comparing it to the best-of-class CPU implementation FFTW. We describe the FFT, the architecture of the GPU, and how general-purpose computation is structured on the GPU. We then identify the factors that influence FFT performance and describe several experiments that compare these factors between the CPU and the GPU. We conclude that the overhead of transferring data and initiating GPU computation are substantially higher than on the CPU, and thus for latency-critical applications, the CPU is a superior choice. We show that the CPU implementation is limited by computation and the GPU implementation by GPU memory bandwidth and its lack of a writable cache. The GPU is comparatively better suited for larger FFTs with many FFTs computed in parallel in applications where FFT throughput is most important; on these applications GPU and CPU performance is roughly on par. We also demonstrate that adding additional computation to an application that includes the FFT, particularly computation that is GPU-friendly, puts the GPU at an advantage compared to the CPU.

/pdf/assessment-of-graphic-processing-units-gpus-for-department-24a4xi20nx.pdf

Shubhabrata Sengupta

Papers

End-to-end speech recognition

Efficient Parallel Scan Algorithms for Manycore GPUs

Dynamic adaptive shadow maps on graphics hardware

Octree textures on graphics hardware

Assessment of Graphic Processing Units (GPUs) for Department of Defense (DoD) Digital Signal Processing (DSP) Applications