Intel Math Kernel Library

By Endong Wang, Qing Zhang, Shen Bo, Guangyong Zhang, Xiaowei Lu, Qing Wu, Yajuan Wang. 2014

728 citations recorded by the source corpus.

Subfields: Computational Theory and Mathematics, Hardware and Architecture, Information Systems

View Paper: https://www.semanticscholar.org/paper/83cee2606be3ceda59ef28fe1a9f179cd4e1144d

How researchers see this paper

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

In Computational Theory and Mathematics

In Hardware and Architecture

In Information Systems

Citations in the corpus

Cache Telepathy: Leveraging Shared Resource Attacks to Learn DNN Architectures (2018, neutral): “Second, the frameworks’ backends are all supported by highperformance and popular linear algebra libraries, such as OpenBLAS [31], Eigen [32] and MKL [33].”
FP-DNN: An Automated Framework for Mapping Deep Neural Networks onto FPGAs with RTL-HLS Hybrid Templates (2017, neutral): “We show the performance (in GOP/S, giga operations per second) of our implementations, Intel MKL [24] and Altera example design [2] for matrix multiplication in Figure 5.”
MLPerf Inference Benchmark (2019, neutral): “…al., 2014), Intel MKL (Intel, 2018a), and FBGEMM (Khudia et al., 2018), supporting various inference run times, such as Apple CoreML (Apple, 2017), Intel OpenVino (Intel, 2018b), NVIDIA TensorRT (NVIDIA), ONNX Runtime (Bai et al., 2019), Qualcomm SNPE (Qual-comm), and TF-Lite (Lee et al., 2019a).”
Occlumency: Privacy-preserving Remote Deep-learning Inference Using SGX (2019, neutral): “Then, linear algebra libraries such as OpenBLAS [74] or Intel MKL [73] optimizes the matrix multiplication using CPU cache based on the fact that the matrix multiplication reads each element of matrix redundantly.”
GeniePath: Graph Neural Networks with Adaptive Receptive Paths (2018, neutral): “Our algorithms could be easily formulated in terms of linear algebra, thus could leverage the power of Intel MKL (Intel 2007) on CPUs or cuBLAS (Nvidia 2008) on GPUs.”
Inducing and Exploiting Activation Sparsity for Fast Neural Network Inference (2020, positive): “We make use of a variant of the Compressed Sparse Row (CSR) representation (e.g. as implemented in (Wang et al., 2014)).”
DeepRecSys: A System for Optimizing End-To-End At-Scale Neural Recommendation Inference (2020, positive): “As discussed in Section III, DeepRecIn-fra comprises three key components: • Model Implementation: We implement all the recommendation models (shown in Table I) in Caffe2 with Intel MKL as the backend library for CPUs [33] and CUDA/cuDNN 10.1 for GPUs [34].”
Generative Models and Model Criticism via Optimized Maximum Mean Discrepancy (2016, neutral): “This streaming estimator uses data very inefﬁciently, however, often requiring m 2 samples to achieve power comparable to tests based on (cid:91) MMD with m samples (Ramdas et al., 2015a).”
Hardware-Guided Symbiotic Training for Compact, Accurate, yet Execution-Efficient LSTM (2019, positive): “For CPU inference, we use Intel MKL [33] implementations for sparse matrix operation.”
Benchmarking State-of-the-Art Deep Learning Software Tools (2016, neutral): “TensorFlow uses Eigen [10] as accelerated matrix operation library, while Caffe, CNTK, MXNet and Torch employ OpenBLAS [11], Intel MKL [12] or cuBLAS [13] to speed up matrix related calculations.”
Pipe-SGD: A Decentralized Pipelined SGD Framework for Distributed Deep Net Training (2018, neutral): “The key idea of SSP: 1) each worker has its own iteration index, 2) the slowest and fastest worker must be within S iterations, otherwise, the fastest worker is forced to wait until the slowest worker catches up.”
Dandelion: a compiler and runtime for heterogeneous systems (2013, positive): “The primary use of Apply is to integrate existing CPU and GPU libraries such as CUBLAS [63] and MKL [43] into Dandelion, bringing the primitives deﬁned in those libraries to the same level of abstraction.”