Scalable parallel programming

By John R. Nickolls, Ian Buck, Michael Garland, Kevin Skadron. 2008

2,068 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Hardware and Architecture

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

In Hardware and Architecture

Citations in the corpus

AMC: AutoML for Model Compression and Acceleration on Mobile Devices (2018, background): “ also gain gracefully inference time speed up on GPU. (Inference time is tested on GeForce GTX TITAN Xp with a mini batch size of 50 (the reported latency is amortized latency for each image), CUDA 8 [46] and cuDNN 6 [10].) On an Android phone, the speed-up is even better than on GPU. We achieved 1:”
Channel Pruning for Accelerating Very Deep Neural Networks (2017, methodology): “Results in Table 3 are obtained under Caffe [23], CUDA8 [38] and cuDNN5 [6], with a mini-batch of 32 on a GPU (GeForce GTX TITAN X).”
Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches (2015, methodology): “The post-processing steps of the stereo method were implemented in CUDA (Nickolls et al., 2008), the network training was done with the Torch7 environment (Collobert et al.”
Recognition of prokaryotic and eukaryotic promoters using convolutional deep learning neural networks (2016, methodology): “It uses Theano library [33, 34] as a backend and utilizes GPU [35] for fast neural network training.”
Accelerating Compute-Intensive Applications with GPUs and FPGAs (2008, background): “This is a clear-cut issue for GPUs, as G80 GPUs implement IEEE-754 single-precision [14] (with double-precision coming in the next generation), and AMD’s FireStream 9710 GPUs already support double-precision.”
An adaptive performance modeling tool for GPU architectures (2010, methodology): “With the introduction of CUDA [17] and OpenCL [15], developers now have the programming and architectural features to quickly port programs to a platform with a massively parallel, GPU-based coprocessor [16].”
Designing efficient sorting algorithms for manycore GPUs (2009, methodology): “CUDA [10], [11] provides the means for developers to execute parallel programs on the GPU.”
AutoTM: Automatic Tensor Movement in Heterogeneous Memory Systems using Integer Linear Programming (2020, methodology): “Asynchronous movement was implemented using two CUDA [31] streams: one for computation and the other for data movement via cudaMemcpyAsync .”
ADC: Automated Deep Compression and Acceleration with Reinforcement Learning (2018, methodology): “(Inference time is tested on GeForce GTX TITAN Xp with a mini batch size of 50 (the reported latency is amortized latency for each image), CUDA 8 [46] and cuDNN 6 [10].”
Deep Learning in Mobile and Wireless Networking: A Survey (2018, methodology): “GPUs were originally designed for high performance video games and graphical rendering, but new techniques such as Compute Unified Device Architecture (CUDA) [115] and the CUDA Deep Neural Network library (cuDNN) [116] developed by NVIDIA add flexibility to this type of hardware, allowing users to c”
High-order finite-element seismic wave propagation modeling with MPI on a large GPU cluster (2010, background): “It constitutes both a hardware architecture for throughput-oriented computing and an associated programming model [8, 9, 10, 4].”
An analytical model for a GPU architecture with memory-level and thread-level parallelism awareness (2009, methodology): “We evaluate our analytical model based on the CUDA [20, 22]”
Medical image processing on the GPU - Past, present and future (2013, methodology): “The release of the compute unified device architecture (CUDA) programming language made using Nvidia GPUs for more general purpose calculations much easier (Nickolls et al., 2008; Garland et al., 2008; Che et al., 2008).”