var bibbase_data = {"data":"<img src=\"//bibbase.org/img/ajax-loader.gif\" id=\"spinner\"\n  style=\"display: none;\" alt=\"Loading..\" />\n\n<div id=\"bibbase\">\n\n  <script type=\"text/javascript\">\n    var bibbase = {\n      params: {\"bib\":\"https://awan-10.github.io/ammar.bib\",\"jsonp\":\"1\",\"host\":\"bibbase.org\"},\n      data: [{\"bibtype\":\"article\",\"type\":\"article\",\"title\":\"Optimized Large-message Broadcast for Deep Learning Workloads: MPI, MPI+NCCL, or NCCL2?\",\"journal\":\"Parallel Computing\",\"volume\":\"85\",\"pages\":\"141 - 152\",\"year\":\"2019\",\"issn\":\"0167-8191\",\"doi\":\"10.1016/j.parco.2019.03.005\",\"url\":\"http://www.sciencedirect.com/science/article/pii/S0167819118303284\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan\"],\"suffixes\":[]},{\"firstnames\":[\"KV\"],\"propositions\":[],\"lastnames\":[\"Manian\"],\"suffixes\":[]},{\"firstnames\":[\"C-H\"],\"propositions\":[],\"lastnames\":[\"Chu\"],\"suffixes\":[]},{\"firstnames\":[\"H.\"],\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"suffixes\":[]},{\"firstnames\":[\"DK\"],\"propositions\":[],\"lastnames\":[\"Panda\"],\"suffixes\":[]}],\"keywords\":\"HPC, Distributed deep learning, MPI_Bcast, NCCL, CUDA-Aware MPI\",\"bibtex\":\"@article{awan-parco19,\\ntitle = {{Optimized Large-message Broadcast for Deep Learning Workloads: MPI, MPI+NCCL, or NCCL2?}},\\njournal = \\\"Parallel Computing\\\",\\nvolume = \\\"85\\\",\\npages = \\\"141 - 152\\\",\\nyear = \\\"2019\\\",\\nissn = \\\"0167-8191\\\",\\ndoi = \\\"10.1016/j.parco.2019.03.005\\\",\\nurl = \\\"http://www.sciencedirect.com/science/article/pii/S0167819118303284\\\",\\nauthor = \\\"{A. A. Awan} and KV Manian and C-H Chu and H. Subramoni and DK Panda\\\",\\nkeywords = \\\"HPC, Distributed deep learning, MPI_Bcast, NCCL, CUDA-Aware MPI\\\",\\n}\\n\\n\",\"author_short\":[\"A. A. Awan\",\"Manian, K.\",\"Chu, C.\",\"Subramoni, H.\",\"Panda, D.\"],\"key\":\"awan-parco19\",\"id\":\"awan-parco19\",\"bibbaseid\":\"aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019\",\"role\":\"author\",\"urls\":{\"Paper\":\"http://www.sciencedirect.com/science/article/pii/S0167819118303284\"},\"keyword\":[\"HPC\",\"Distributed deep learning\",\"MPI_Bcast\",\"NCCL\",\"CUDA-Aware MPI\"],\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"},{\"bibtype\":\"article\",\"type\":\"article\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan\"],\"suffixes\":[]},{\"firstnames\":[\"A.\"],\"propositions\":[],\"lastnames\":[\"Jain\"],\"suffixes\":[]},{\"firstnames\":[\"C-H\"],\"propositions\":[],\"lastnames\":[\"Chu\"],\"suffixes\":[]},{\"firstnames\":[\"H.\"],\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"suffixes\":[]},{\"firstnames\":[\"DK\"],\"propositions\":[],\"lastnames\":[\"Panda\"],\"suffixes\":[]}],\"journal\":\"IEEE Micro\",\"title\":\"Communication Profiling and Characterization of Deep-Learning Workloads on Clusters With High-Performance Interconnects\",\"year\":\"2020\",\"volume\":\"40\",\"number\":\"1\",\"pages\":\"35-43\",\"keywords\":\"Training;Libraries;Measurement;Middleware;Deep learning;Graphics processing units;Performance analysis;InfiniBand;Omni-Path;NVLink;PCIe;TensorFlow;Horovod;MVAPICH2 MPI;Performance Analysis;Profiling;Communication Libraries\",\"doi\":\"10.1109/MM.2019.2949986\",\"issn\":\"1937-4143\",\"month\":\"Jan\",\"bibtex\":\"@ARTICLE{awan-micro19, \\nauthor={{A. A. Awan} and A. Jain and C-H Chu and H. Subramoni and DK Panda}, \\njournal={IEEE Micro}, \\ntitle={{Communication Profiling and Characterization of\\nDeep-Learning Workloads on Clusters With High-Performance\\nInterconnects}},\\nyear={2020}, \\nvolume={40}, \\nnumber={1}, \\npages={35-43}, \\nkeywords={Training;Libraries;Measurement;Middleware;Deep learning;Graphics processing units;Performance analysis;InfiniBand;Omni-Path;NVLink;PCIe;TensorFlow;Horovod;MVAPICH2 MPI;Performance Analysis;Profiling;Communication Libraries}, \\ndoi={10.1109/MM.2019.2949986}, \\nISSN={1937-4143}, \\nmonth={Jan},\\n}\\n\\n\",\"author_short\":[\"A. A. Awan\",\"Jain, A.\",\"Chu, C.\",\"Subramoni, H.\",\"Panda, D.\"],\"key\":\"awan-micro19\",\"id\":\"awan-micro19\",\"bibbaseid\":\"aaawan-jain-chu-subramoni-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2020\",\"role\":\"author\",\"urls\":{},\"keyword\":[\"Training;Libraries;Measurement;Middleware;Deep learning;Graphics processing units;Performance analysis;InfiniBand;Omni-Path;NVLink;PCIe;TensorFlow;Horovod;MVAPICH2 MPI;Performance Analysis;Profiling;Communication Libraries\"],\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":2,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan\"],\"suffixes\":[]},{\"firstnames\":[\"A.\"],\"propositions\":[],\"lastnames\":[\"Jain\"],\"suffixes\":[]},{\"firstnames\":[\"H.\"],\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"suffixes\":[]},{\"firstnames\":[\"C-H\"],\"propositions\":[],\"lastnames\":[\"Chu\"],\"suffixes\":[]},{\"firstnames\":[\"DK\"],\"propositions\":[],\"lastnames\":[\"Panda\"],\"suffixes\":[]}],\"booktitle\":\"26th Symposium on IEEE Hot Interconnects (HotI)\",\"month\":\"Aug\",\"title\":\"Communication Profiling and Characterization of Deep Learning Workloads on Clusters with High-Performance Interconnects\",\"year\":\"2019\",\"bibtex\":\"@inproceedings{awan-hoti19,\\nauthor = {{A. A. Awan} and A. Jain and H. Subramoni and C-H Chu and DK Panda},\\nbooktitle = {26th Symposium on IEEE Hot Interconnects (HotI)},\\nmonth = {Aug},\\ntitle = {{Communication Profiling and Characterization of Deep Learning Workloads on Clusters with High-Performance Interconnects}},\\nyear = {2019}\\n}\\n\\n\",\"author_short\":[\"A. A. Awan\",\"Jain, A.\",\"Subramoni, H.\",\"Chu, C.\",\"Panda, D.\"],\"key\":\"awan-hoti19\",\"id\":\"awan-hoti19\",\"bibbaseid\":\"aaawan-jain-subramoni-chu-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2019\",\"role\":\"author\",\"urls\":{},\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"propositions\":[],\"lastnames\":[\"Awan\"],\"firstnames\":[\"Ammar\",\"Ahmad\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Chu\"],\"firstnames\":[\"Ching-Hsiang\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"firstnames\":[\"Hari\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Panda\"],\"firstnames\":[\"Dhabaleswar\",\"K.\"],\"suffixes\":[]}],\"title\":\"Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?\",\"booktitle\":\"Proceedings of the 25th European MPI Users' Group Meeting\",\"series\":\"EuroMPI'18\",\"year\":\"2018\",\"isbn\":\"978-1-4503-6492-8\",\"location\":\"Barcelona, Spain\",\"pages\":\"2:1–2:9\",\"articleno\":\"2\",\"numpages\":\"9\",\"url\":\"http://doi.acm.org/10.1145/3236367.3236381\",\"doi\":\"10.1145/3236367.3236381\",\"acmid\":\"3236381\",\"publisher\":\"ACM\",\"address\":\"New York, NY, USA\",\"keywords\":\"CUDA-Aware MPI, Distributed Deep Learning, HPC, MPI_Bcast, NCCL\",\"bibtex\":\"@inproceedings{awan-eurompi18,\\n author = {Awan, Ammar Ahmad and Chu, Ching-Hsiang and Subramoni, Hari and Panda, Dhabaleswar K.},\\n title = {{Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?}},\\n booktitle = {Proceedings of the 25th European MPI Users' Group Meeting},\\n series = {EuroMPI'18},\\n year = {2018},\\n isbn = {978-1-4503-6492-8},\\n location = {Barcelona, Spain},\\n pages = {2:1--2:9},\\n articleno = {2},\\n numpages = {9},\\n url = {http://doi.acm.org/10.1145/3236367.3236381},\\n doi = {10.1145/3236367.3236381},\\n acmid = {3236381},\\n publisher = {ACM},\\n address = {New York, NY, USA},\\n keywords = {CUDA-Aware MPI, Distributed Deep Learning, HPC, MPI_Bcast, NCCL},\\n} \\n\\n\",\"author_short\":[\"Awan, A. A.\",\"Chu, C.\",\"Subramoni, H.\",\"Panda, D. K.\"],\"key\":\"awan-eurompi18\",\"id\":\"awan-eurompi18\",\"bibbaseid\":\"awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018\",\"role\":\"author\",\"urls\":{\"Paper\":\"http://doi.acm.org/10.1145/3236367.3236381\"},\"keyword\":[\"CUDA-Aware MPI\",\"Distributed Deep Learning\",\"HPC\",\"MPI_Bcast\",\"NCCL\"],\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan,\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Hamidouche\"],\"firstnames\":[\"K.\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Venkatesh\"],\"firstnames\":[\"A.\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Panda\"],\"firstnames\":[\"DK\"],\"suffixes\":[]}],\"title\":\"Efficient Large Message Broadcast Using NCCL and CUDA-Aware MPI for Deep Learning\",\"booktitle\":\"Proceedings of the 23rd European MPI Users' Group Meeting\",\"series\":\"EuroMPI 2016\",\"year\":\"2016\",\"isbn\":\"978-1-4503-4234-6\",\"location\":\"Edinburgh, United Kingdom\",\"pages\":\"15–22\",\"numpages\":\"8\",\"url\":\"http://doi.acm.org/10.1145/2966884.2966912\",\"doi\":\"10.1145/2966884.2966912\",\"acmid\":\"2966912\",\"publisher\":\"ACM\",\"address\":\"New York, NY, USA\",\"bibtex\":\"@inproceedings{awan-eurompi16,\\n author = {{A. A. Awan,} and Hamidouche, K. and Venkatesh, A. and Panda, DK},\\n title = {{Efficient Large Message Broadcast Using NCCL and CUDA-Aware MPI for\\nDeep Learning}},\\n booktitle = {Proceedings of the 23rd European MPI Users' Group Meeting},\\n series = {EuroMPI 2016},\\n year = {2016},\\n isbn = {978-1-4503-4234-6},\\n location = {Edinburgh, United Kingdom},\\n pages = {15--22},\\n numpages = {8},\\n url = {http://doi.acm.org/10.1145/2966884.2966912},\\n doi = {10.1145/2966884.2966912},\\n acmid = {2966912},\\n publisher = {ACM},\\n address = {New York, NY, USA},\\n}\\n\\n\",\"author_short\":[\"A. A. Awan,\",\"Hamidouche, K.\",\"Venkatesh, A.\",\"Panda, D.\"],\"key\":\"awan-eurompi16\",\"id\":\"awan-eurompi16\",\"bibbaseid\":\"aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016\",\"role\":\"author\",\"urls\":{\"Paper\":\"http://doi.acm.org/10.1145/2966884.2966912\"},\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"propositions\":[],\"lastnames\":[\"Awan\"],\"firstnames\":[\"Ammar\",\"Ahmad\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Hamidouche\"],\"firstnames\":[\"Khaled\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Hashmi\"],\"firstnames\":[\"Jahanzeb\",\"Maqbool\"],\"suffixes\":[]},{\"propositions\":[],\"lastnames\":[\"Panda\"],\"firstnames\":[\"Dhabaleswar\",\"K.\"],\"suffixes\":[]}],\"title\":\"S-Caffe: Co-designing MPI Runtimes and Caffe for Scalable Deep Learning on Modern GPU Clusters\",\"booktitle\":\"Proceedings of the 22Nd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming\",\"series\":\"PPoPP '17\",\"year\":\"2017\",\"isbn\":\"978-1-4503-4493-7\",\"location\":\"Austin, Texas, USA\",\"pages\":\"193–205\",\"numpages\":\"13\",\"url\":\"http://doi.acm.org/10.1145/3018743.3018769\",\"doi\":\"10.1145/3018743.3018769\",\"acmid\":\"3018769\",\"publisher\":\"ACM\",\"address\":\"New York, NY, USA\",\"keywords\":\"caffe, cuda-aware mpi, deep learning, distributed training, mpi_reduce\",\"bibtex\":\"@inproceedings{awan-ppopp17,\\n author = {Awan, Ammar Ahmad and Hamidouche, Khaled and Hashmi, Jahanzeb Maqbool and Panda, Dhabaleswar K.},\\n title = {{S-Caffe: Co-designing MPI Runtimes and Caffe for Scalable Deep Learning on Modern GPU Clusters}},\\n booktitle = {Proceedings of the 22Nd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming},\\n series = {PPoPP '17},\\n year = {2017},\\n isbn = {978-1-4503-4493-7},\\n location = {Austin, Texas, USA},\\n pages = {193--205},\\n numpages = {13},\\n url = {http://doi.acm.org/10.1145/3018743.3018769},\\n doi = {10.1145/3018743.3018769},\\n acmid = {3018769},\\n publisher = {ACM},\\n address = {New York, NY, USA},\\n keywords = {caffe, cuda-aware mpi, deep learning, distributed training, mpi\\\\_reduce},\\n} \\n\\n\",\"author_short\":[\"Awan, A. A.\",\"Hamidouche, K.\",\"Hashmi, J. M.\",\"Panda, D. K.\"],\"key\":\"awan-ppopp17\",\"id\":\"awan-ppopp17\",\"bibbaseid\":\"awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017\",\"role\":\"author\",\"urls\":{\"Paper\":\"http://doi.acm.org/10.1145/3018743.3018769\"},\"keyword\":[\"caffe\",\"cuda-aware mpi\",\"deep learning\",\"distributed training\",\"mpi_reduce\"],\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":1,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan\"],\"suffixes\":[]},{\"firstnames\":[\"J.\"],\"propositions\":[],\"lastnames\":[\"Bédorf\"],\"suffixes\":[]},{\"firstnames\":[\"C.\"],\"propositions\":[],\"lastnames\":[\"Chu\"],\"suffixes\":[]},{\"firstnames\":[\"H.\"],\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"suffixes\":[]},{\"firstnames\":[\"D.\",\"K.\"],\"propositions\":[],\"lastnames\":[\"Panda\"],\"suffixes\":[]}],\"booktitle\":\"2019 19th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)\",\"title\":\"Scalable Distributed DNN Training using TensorFlow and CUDA-Aware MPI: Characterization, Designs, and Performance Evaluation\",\"year\":\"2019\",\"volume\":\"\",\"number\":\"\",\"pages\":\"498-507\",\"doi\":\"10.1109/CCGRID.2019.00064\",\"issn\":\"\",\"month\":\"May\",\"bibtex\":\"@INPROCEEDINGS{awan-ccgrid19, \\nauthor={{A. A. Awan} and J. {Bédorf} and C. {Chu} and H. {Subramoni} and D. K. {Panda}}, \\nbooktitle={2019 19th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)}, \\ntitle={Scalable Distributed DNN Training using TensorFlow and CUDA-Aware MPI: Characterization, Designs, and Performance Evaluation}, \\nyear={2019}, \\nvolume={}, \\nnumber={}, \\npages={498-507}, \\ndoi={10.1109/CCGRID.2019.00064}, \\nISSN={}, \\nmonth={May},\\n}\\n\\n\",\"author_short\":[\"A. A. Awan\",\"Bédorf, J.\",\"Chu, C.\",\"Subramoni, H.\",\"Panda, D. K.\"],\"key\":\"awan-ccgrid19\",\"id\":\"awan-ccgrid19\",\"bibbaseid\":\"aaawan-bdorf-chu-subramoni-panda-scalabledistributeddnntrainingusingtensorflowandcudaawarempicharacterizationdesignsandperformanceevaluation-2019\",\"role\":\"author\",\"urls\":{},\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan\"],\"suffixes\":[]},{\"firstnames\":[\"H.\"],\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"suffixes\":[]},{\"firstnames\":[\"DK\"],\"propositions\":[],\"lastnames\":[\"Panda\"],\"suffixes\":[]}],\"title\":\"An In-depth Performance Characterization of CPU- and GPU-based DNN Training on Modern Architectures\",\"booktitle\":\"Proceedings of the Machine Learning on HPC Environments\",\"series\":\"MLHPC'17\",\"year\":\"2017\",\"isbn\":\"978-1-4503-5137-9\",\"location\":\"Denver, CO, USA\",\"pages\":\"8:1–8:8\",\"articleno\":\"8\",\"numpages\":\"8\",\"url\":\"http://doi.acm.org/10.1145/3146347.3146356\",\"doi\":\"10.1145/3146347.3146356\",\"acmid\":\"3146356\",\"publisher\":\"ACM\",\"address\":\"New York, NY, USA\",\"keywords\":\"Caffe, Deep Learning, High-Performance Computing, Pascal Architecture, Unified Memory\",\"bibtex\":\"@inproceedings{awan-mlhpc17, \\n author = {{A. A. Awan} and H. Subramoni and DK Panda},\\n title = {{An In-depth Performance Characterization of CPU- and GPU-based DNN\\n     Training on Modern Architectures}},\\n booktitle = {Proceedings of the Machine Learning on HPC Environments},\\n series = {MLHPC'17},\\n year = {2017},\\n isbn = {978-1-4503-5137-9},\\n location = {Denver, CO, USA},\\n pages = {8:1--8:8},\\n articleno = {8},\\n numpages = {8},\\n url = {http://doi.acm.org/10.1145/3146347.3146356},\\n doi = {10.1145/3146347.3146356},\\n acmid = {3146356},\\n publisher = {ACM},\\n address = {New York, NY, USA},\\n keywords = {Caffe, Deep Learning, High-Performance Computing, Pascal Architecture, Unified Memory},\\n} \\n\\n\",\"author_short\":[\"A. A. Awan\",\"Subramoni, H.\",\"Panda, D.\"],\"key\":\"awan-mlhpc17\",\"id\":\"awan-mlhpc17\",\"bibbaseid\":\"aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017\",\"role\":\"author\",\"urls\":{\"Paper\":\"http://doi.acm.org/10.1145/3146347.3146356\"},\"keyword\":[\"Caffe\",\"Deep Learning\",\"High-Performance Computing\",\"Pascal Architecture\",\"Unified Memory\"],\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"},{\"bibtype\":\"inproceedings\",\"type\":\"inproceedings\",\"author\":[{\"firstnames\":[],\"propositions\":[],\"lastnames\":[\"A. A. Awan\"],\"suffixes\":[]},{\"firstnames\":[\"C-H\"],\"propositions\":[],\"lastnames\":[\"Chu\"],\"suffixes\":[]},{\"firstnames\":[\"H.\"],\"propositions\":[],\"lastnames\":[\"Subramoni\"],\"suffixes\":[]},{\"firstnames\":[\"X.\"],\"propositions\":[],\"lastnames\":[\"Lu\"],\"suffixes\":[]},{\"firstnames\":[\"DK\"],\"propositions\":[],\"lastnames\":[\"Panda\"],\"suffixes\":[]}],\"booktitle\":\"2018 IEEE 25th International Conference on High Performance Computing (HiPC)\",\"doi\":\"10.1109/HiPC.2018.00024\",\"issn\":\"2640-0316\",\"keywords\":\"graphics processing units;learning (artificial int\",\"month\":\"dec\",\"pages\":\"143–152\",\"title\":\"OC-DNN: Exploiting Advanced Unified Memory Capabilities in CUDA 9 and Volta GPUs for Out-of-Core DNN Training\",\"year\":\"2018\",\"bibtex\":\"@inproceedings{awan-hipc18,\\nauthor = {{A. A. Awan} and C-H Chu and H. Subramoni and X. Lu and DK Panda},\\nbooktitle = {2018 IEEE 25th International Conference on High Performance Computing (HiPC)},\\ndoi = {10.1109/HiPC.2018.00024},\\nissn = {2640-0316},\\nkeywords = {graphics processing units;learning (artificial int},\\nmonth = {dec},\\npages = {143--152},\\ntitle = {{OC-DNN: Exploiting Advanced Unified Memory Capabilities in CUDA 9 and Volta GPUs for Out-of-Core DNN Training}},\\nyear = {2018}\\n}\\n\",\"author_short\":[\"A. A. Awan\",\"Chu, C.\",\"Subramoni, H.\",\"Lu, X.\",\"Panda, D.\"],\"key\":\"awan-hipc18\",\"id\":\"awan-hipc18\",\"bibbaseid\":\"aaawan-chu-subramoni-lu-panda-ocdnnexploitingadvancedunifiedmemorycapabilitiesincuda9andvoltagpusforoutofcorednntraining-2018\",\"role\":\"author\",\"urls\":{},\"keyword\":[\"graphics processing units;learning (artificial int\"],\"metadata\":{\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\"downloads\":0,\"html\":\"\"}],\n      metadata: {\"owner\":\"panda, d\",\"authorlinks\":{\"panda, d\":\"https://awan-10.github.io/research.html\"}},\n      ownerID: \"RupBLgDh3seizbcLL\"\n    };\n  </script>\n\n  <script type=\"text/javascript\">\n  var site$;\n  if (typeof $ != 'undefined') {\n    console.log('existing jquery: storing and then restoring');\n    site$ = $;\n    $ = null;\n  }\n  </script>\n\n  <script src=\"https://ajax.googleapis.com/ajax/libs/jquery/2.2.4/jquery.min.js\">\n  </script>\n  <script type=\"text/javascript\">\n  if (site$) {\n    console.log('existing jquery: avoiding conflict and restoring');\n    bb$ = $.noConflict(true);\n    $ = site$;\n  } else {\n    bb$ = $;\n  }\n  </script>\n\n  <script src=\"//bibbase.org/js/bibbase.min.js\"\n          type=\"text/javascript\">\n  </script>\n\n  <script type=\"text/javascript\"\n          src=\"https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.1/MathJax.js?config=TeX-AMS-MML_HTMLorMML\">\n  </script>\n  <script type=\"text/x-mathjax-config\">\n    MathJax.Hub.Config({tex2jax: {inlineMath: [['$','$'], ['\\\\(','\\\\)']]},\n                        skipTags: [\"body\"],\n                        processClass: \"bibbase_paper\"});\n  </script>\n\n  <style>\n  @media print {\n    .dontprint {\n        display: none !important;\n    }\n\n  .bibbase_group {\n    background-color: #E0E0E0;\n    font-style: italic;\n    font-size: larger;\n    text-shadow: 0px 0px 3px #FFFFFF;\n    border-radius: 4px 4px 4px 4px;\n    -moz-border-radius: 4px 4px 4px 4px;\n    -webkit-border-radius: 4px;\n    padding: 5px 40px 5px;\n    margin-top: 10px;\n    margin-bottom: 5px;\n    cursor: pointer;\n  }\n\n  .bibbase_paper {\n    margin-bottom: 5px;\n  }\n\n  }\n  </style>\n\n  \n  <div style=\"float: right; margin-right: 10px; margin-top: 10px; text-align: right; font-size: 12px\">\n    generated by\n    <a href=\"//bibbase.org\"\n       onclick=\"trackOutboundLink('bibbase', 'logo clicked', window.location.href, '//bibbase.org'); return false;\">\n      <img src=\"//bibbase.org/img/logo.svg\"\n           style=\"vertical-align: middle; margin-left: 3px; height: 16px;\"/\n           alt=\"bibbase.org\"\n           title=\"BibBase – The easiest way to maintain your publications page.\"\n        ></a>\n    \n  </div>\n  \n\n  <div id=\"bibbase_header\" class=\"dontprint\" style=\"\">\n\n    <ul class=\"nav nav-pills\" style=\"margin: 0px;\">\n\n      <li class=\"dropdown\" id=\"groupby_dropdown\">\n      <a class=\"dropdown-toggle\"\n         data-toggle=\"dropdown\"\n         href=\"#\">\n          Group by\n          <b class=\"caret\"></b>\n      </a>\n      <ul class=\"dropdown-menu\">\n        <li class=\"groupby year\"><a onclick=\"groupby('year')\">\n            Year</a>\n        </li>\n        <li class=\"groupby author\"><a onclick=\"groupby('author_short')\">\n            Author</a>\n        </li>\n        <li class=\"groupby type\"><a onclick=\"groupby('type')\">\n            Type</a>\n        </li>\n        <li class=\"groupby keyword\"><a onclick=\"groupby('keyword')\">\n            Keyword</a>\n        </li>\n        <li class=\"groupby downloads\"><a onclick=\"groupby('downloads')\">\n            Downloads</a>\n        </li>\n      </ul>\n      </li>\n\n\n      <li class=\"dropdown\" id=\"menu_dropdown\">\n      <a class=\"dropdown-toggle\"\n         data-toggle=\"dropdown\"\n         href=\"#\" alt=\"controls\">\n          <i class=\"fa fa-reorder\" alt=\"controls\"></i>\n          <b class=\"caret\"></b>\n      </a>\n      <ul class=\"dropdown-menu\">\n        <li>\n          <a onclick=\"toggleAll()\">\n            <i class=\"fa fa-chevron-down fa-fw\"></i> Expand/Collapse All\n          </a>\n        </li>\n        <li>\n          <a download=\"ammar.bib\" href=\"data:text/bibtex;base64,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\">\n            <i class=\"fa fa-download fa-fw\"></i> Download BibTeX\n          </a>\n        </li>\n        <li>\n          <a href=\"//bibbase.org/rss?bib=https://awan-10.github.io/ammar.bib\">\n            <i class=\"fa fa-rss fa-fw\"></i> RSS Feed\n          </a>\n          </li>\n      </ul>\n      </li>\n\n      \n\n\n      \n    </ul>\n\n\n    <div class=\"alert alert-success bibbase_msg\" id=\"bibbase_msg_embed\"\n         style=\"display: none;\">\n\n      Excellent! Next you can\n      <a href=\"/network/register?next=/network/newSiteFromTemplate%3Fbiburl=https://awan-10.github.io/ammar.bib\"\n      >create a new website with this list</a>, or\n      embed it in an existing web page by copying &amp; pasting\n      any of the following snippets.\n\n      <div style=\"text-align: left; margin-top: 1em;\">\n        <strong>JavaScript</strong>\n        <span class=\"comment recommended\">(easiest)</span>\n        <div class=\"well well-small\">\n          <code>\n            &lt;script src=\"https://bibbase.org/show?bib=https%3A%2F%2Fawan-10.github.io%2Fammar.bib&amp;jsonp=1&amp;jsonp=1\"&gt;&lt;/script&gt;\n          </code>\n        </div>\n\n        <strong>PHP</strong>\n        <div class=\"well well-small\">\n          <code>\n            &lt;?php\n            $contents = file_get_contents(\"https://bibbase.org/show?bib=https%3A%2F%2Fawan-10.github.io%2Fammar.bib&amp;jsonp=1\");\n            print_r($contents);\n            ?&gt;\n          </code>\n        </div>\n\n        <strong>iFrame</strong>\n        <span class=\"comment\">(not recommended)</span>\n        <div class=\"well well-small\">\n          <code>\n            &lt;iframe src=\"https://bibbase.org/show?bib=https%3A%2F%2Fawan-10.github.io%2Fammar.bib&amp;jsonp=1\"&gt;&lt;/iframe&gt;\n          </code>\n        </div>\n\n        <p>\n          For more details see the <a href=\"//bibbase.org/help\">documention</a>.\n        </p>\n      </div>\n    </div>\n\n    <div class=\"alert alert-success bibbase_msg\" id=\"bibbase_msg_preview\"\n         style=\"display: none;\">\n\n      This is a preview! To use this list on your own web site\n      or create a new web site from it,\n      <a href=\"/network/register?next=/network/newAccountWithFile%3FfileId=\"\n      >create a free account</a>. The file will be added\n      and you will be able to edit it in the File Manager.\n      We will show you instructions once you've created your account.\n    </div>\n\n    <div class=\"alert alert-warning bibbase_msg\" id=\"bibbase_msg_mendeley1\"\n         style=\"display: none;\">\n\n      <p>To the site owner:</p>\n\n      <p><strong>Action required!</strong> Mendeley is changing its\n        API. In order to keep using Mendeley with BibBase past April\n        14th, you need to:\n        <ol>\n          <li>renew the authorization for BibBase on Mendeley, and</li>\n          <li>update the BibBase URL\n            in your page the same way you did when you initially set up\n            this page.\n          </li>\n        </ol>\n      </p>\n\n      <p>\n        <a href=\"http://bibbase.org/cgi-bin/mendeley2/get.py\">\n          <i class=\"fa fa-angle-double-right\"></i>\n          Fix it now</a>\n      </p>\n    </div>\n\n  </div>\n\n\n  <div id=\"bibbase_body\">\n    \n  \n  <div class=\"bibbase_group_whole\" id=\"group_2020\">\n    <div class=\"bibbase_group\"\n      onclick=\"toggleGroup('group_2020')\"\n      onkeypress=\"toggleGroup('group_2020')\"\n      tabindex=\"0\">\n      <i class=\"fa fa-angle-down bibbase_group_head_icon\"></i>&nbsp;\n      <span>2020</span>\n      <span class=\"bibbase_group_count\">\n        \n        (1)\n        \n      </span>\n    </div>\n    <div class=\"bibbase_group_body\">\n      \n  \n  <div class=\"bibbase_paper\" id=\"aaawan-jain-chu-subramoni-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2020\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  Communication Profiling and Characterization of Deep-Learning Workloads on Clusters With High-Performance Interconnects.\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan; Jain, A.; Chu, C.; Subramoni, H.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D.</a>\n</span>\n\n  \n\n</span>\n\n  <i>IEEE Micro</i>, 40(1): 35-43. Jan 2020.\n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n\n  \n  <a href=\"http://doi.org/10.1109/MM.2019.2949986\"\n     onclick=\"log_download('aaawan-jain-chu-subramoni-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2020', 'http://doi.org/10.1109/MM.2019.2949986', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-jain-chu-subramoni-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2020\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n  &nbsp;\n  <span class=\"bibbase_stats_paper\" style=\"color: #777;\">\n    <span>2 downloads</span>\n  </span>\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-jain-chu-subramoni-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2020')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@ARTICLE{awan-micro19, \nauthor={{A. A. Awan} and A. Jain and C-H Chu and H. Subramoni and DK Panda}, \njournal={IEEE Micro}, \ntitle={{Communication Profiling and Characterization of\nDeep-Learning Workloads on Clusters With High-Performance\nInterconnects}},\nyear={2020}, \nvolume={40}, \nnumber={1}, \npages={35-43}, \nkeywords={Training;Libraries;Measurement;Middleware;Deep learning;Graphics processing units;Performance analysis;InfiniBand;Omni-Path;NVLink;PCIe;TensorFlow;Horovod;MVAPICH2 MPI;Performance Analysis;Profiling;Communication Libraries}, \ndoi={10.1109/MM.2019.2949986}, \nISSN={1937-4143}, \nmonth={Jan},\n}\n\n</pre>\n</div>\n\n\n\n</div>\n\n\n\n\n\n    </div>\n  </div>\n\n  <div class=\"bibbase_group_whole\" id=\"group_2019\">\n    <div class=\"bibbase_group\"\n      onclick=\"toggleGroup('group_2019')\"\n      onkeypress=\"toggleGroup('group_2019')\"\n      tabindex=\"0\">\n      <i class=\"fa fa-angle-down bibbase_group_head_icon\"></i>&nbsp;\n      <span>2019</span>\n      <span class=\"bibbase_group_count\">\n        \n        (3)\n        \n      </span>\n    </div>\n    <div class=\"bibbase_group_body\">\n      \n  \n  <div class=\"bibbase_paper\" id=\"aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  <a href=\"http://www.sciencedirect.com/science/article/pii/S0167819118303284\"\n     onclick=\"return trackOutboundLink('publications', 'click', 'aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019', 'http://www.sciencedirect.com/science/article/pii/S0167819118303284', event);\">\n    Optimized Large-message Broadcast for Deep Learning Workloads: MPI, MPI+NCCL, or NCCL2?.\n  </a>\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan; Manian, K.; Chu, C.; Subramoni, H.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D.</a>\n</span>\n\n  \n\n</span>\n\n  <i>Parallel Computing</i>, 85: 141 - 152.  2019.\n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n  <a href=\"http://www.sciencedirect.com/science/article/pii/S0167819118303284\"\n     onclick=\"log_download('aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019', 'http://www.sciencedirect.com/science/article/pii/S0167819118303284', event.ctrlKey, event)\">\n    <img src=\"//bibbase.org/img/filetypes/link.svg\"\n\t     alt=\"Optimized Large-message Broadcast for Deep Learning Workloads: MPI, MPI+NCCL, or NCCL2? [link]\"\n       title=\"Paper\"\n\t     class=\"bibbase_icon\"\n         ><span class=\"bibbase_icon_text\">Paper</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"http://doi.org/10.1016/j.parco.2019.03.005\"\n     onclick=\"log_download('aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019', 'http://doi.org/10.1016/j.parco.2019.03.005', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-manian-chu-subramoni-panda-optimizedlargemessagebroadcastfordeeplearningworkloadsmpimpincclornccl2-2019')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@article{awan-parco19,\ntitle = {{Optimized Large-message Broadcast for Deep Learning Workloads: MPI, MPI+NCCL, or NCCL2?}},\njournal = &quot;Parallel Computing&quot;,\nvolume = &quot;85&quot;,\npages = &quot;141 - 152&quot;,\nyear = &quot;2019&quot;,\nissn = &quot;0167-8191&quot;,\ndoi = &quot;10.1016/j.parco.2019.03.005&quot;,\nurl = &quot;http://www.sciencedirect.com/science/article/pii/S0167819118303284&quot;,\nauthor = &quot;{A. A. Awan} and KV Manian and C-H Chu and H. Subramoni and DK Panda&quot;,\nkeywords = &quot;HPC, Distributed deep learning, MPI_Bcast, NCCL, CUDA-Aware MPI&quot;,\n}\n\n</pre>\n</div>\n\n\n\n</div>\n\n\n  <div class=\"bibbase_paper\" id=\"aaawan-jain-subramoni-chu-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2019\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  Communication Profiling and Characterization of Deep Learning Workloads on Clusters with High-Performance Interconnects.\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan; Jain, A.; Subramoni, H.; Chu, C.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>26th Symposium on IEEE Hot Interconnects (HotI)</i>, Aug 2019. \n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-jain-subramoni-chu-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2019\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-jain-subramoni-chu-panda-communicationprofilingandcharacterizationofdeeplearningworkloadsonclusterswithhighperformanceinterconnects-2019')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@inproceedings{awan-hoti19,\nauthor = {{A. A. Awan} and A. Jain and H. Subramoni and C-H Chu and DK Panda},\nbooktitle = {26th Symposium on IEEE Hot Interconnects (HotI)},\nmonth = {Aug},\ntitle = {{Communication Profiling and Characterization of Deep Learning Workloads on Clusters with High-Performance Interconnects}},\nyear = {2019}\n}\n\n</pre>\n</div>\n\n\n\n</div>\n\n\n  <div class=\"bibbase_paper\" id=\"aaawan-bdorf-chu-subramoni-panda-scalabledistributeddnntrainingusingtensorflowandcudaawarempicharacterizationdesignsandperformanceevaluation-2019\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  Scalable Distributed DNN Training using TensorFlow and CUDA-Aware MPI: Characterization, Designs, and Performance Evaluation.\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan; Bédorf, J.; Chu, C.; Subramoni, H.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D. K.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>2019 19th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)</i>, pages 498-507, May 2019. \n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n\n  \n  <a href=\"http://doi.org/10.1109/CCGRID.2019.00064\"\n     onclick=\"log_download('aaawan-bdorf-chu-subramoni-panda-scalabledistributeddnntrainingusingtensorflowandcudaawarempicharacterizationdesignsandperformanceevaluation-2019', 'http://doi.org/10.1109/CCGRID.2019.00064', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-bdorf-chu-subramoni-panda-scalabledistributeddnntrainingusingtensorflowandcudaawarempicharacterizationdesignsandperformanceevaluation-2019\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-bdorf-chu-subramoni-panda-scalabledistributeddnntrainingusingtensorflowandcudaawarempicharacterizationdesignsandperformanceevaluation-2019')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@INPROCEEDINGS{awan-ccgrid19, \nauthor={{A. A. Awan} and J. {Bédorf} and C. {Chu} and H. {Subramoni} and D. K. {Panda}}, \nbooktitle={2019 19th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)}, \ntitle={Scalable Distributed DNN Training using TensorFlow and CUDA-Aware MPI: Characterization, Designs, and Performance Evaluation}, \nyear={2019}, \nvolume={}, \nnumber={}, \npages={498-507}, \ndoi={10.1109/CCGRID.2019.00064}, \nISSN={}, \nmonth={May},\n}\n\n</pre>\n</div>\n\n\n\n</div>\n\n\n\n\n\n    </div>\n  </div>\n\n  <div class=\"bibbase_group_whole\" id=\"group_2018\">\n    <div class=\"bibbase_group\"\n      onclick=\"toggleGroup('group_2018')\"\n      onkeypress=\"toggleGroup('group_2018')\"\n      tabindex=\"0\">\n      <i class=\"fa fa-angle-down bibbase_group_head_icon\"></i>&nbsp;\n      <span>2018</span>\n      <span class=\"bibbase_group_count\">\n        \n        (2)\n        \n      </span>\n    </div>\n    <div class=\"bibbase_group_body\">\n      \n  \n  <div class=\"bibbase_paper\" id=\"awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  <a href=\"http://doi.acm.org/10.1145/3236367.3236381\"\n     onclick=\"return trackOutboundLink('publications', 'click', 'awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018', 'http://doi.acm.org/10.1145/3236367.3236381', event);\">\n    Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?.\n  </a>\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  Awan, A. A.; Chu, C.; Subramoni, H.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D. K.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>Proceedings of the 25th European MPI Users' Group Meeting</i>, of <i>EuroMPI'18</i>, pages 2:1–2:9, New York, NY, USA,  2018. ACM\n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n  <a href=\"http://doi.acm.org/10.1145/3236367.3236381\"\n     onclick=\"log_download('awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018', 'http://doi.acm.org/10.1145/3236367.3236381', event.ctrlKey, event)\">\n    <img src=\"//bibbase.org/img/filetypes/link.svg\"\n\t     alt=\"Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL? [link]\"\n       title=\"Paper\"\n\t     class=\"bibbase_icon\"\n         ><span class=\"bibbase_icon_text\">Paper</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"http://doi.org/10.1145/3236367.3236381\"\n     onclick=\"log_download('awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018', 'http://doi.org/10.1145/3236367.3236381', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('awan-chu-subramoni-panda-optimizedbroadcastfordeeplearningworkloadsondensegpuinfinibandclustersmpiornccl-2018')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@inproceedings{awan-eurompi18,\n author = {Awan, Ammar Ahmad and Chu, Ching-Hsiang and Subramoni, Hari and Panda, Dhabaleswar K.},\n title = {{Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?}},\n booktitle = {Proceedings of the 25th European MPI Users&#x27; Group Meeting},\n series = {EuroMPI&#x27;18},\n year = {2018},\n isbn = {978-1-4503-6492-8},\n location = {Barcelona, Spain},\n pages = {2:1--2:9},\n articleno = {2},\n numpages = {9},\n url = {http://doi.acm.org/10.1145/3236367.3236381},\n doi = {10.1145/3236367.3236381},\n acmid = {3236381},\n publisher = {ACM},\n address = {New York, NY, USA},\n keywords = {CUDA-Aware MPI, Distributed Deep Learning, HPC, MPI_Bcast, NCCL},\n} \n\n</pre>\n</div>\n\n\n\n</div>\n\n\n  <div class=\"bibbase_paper\" id=\"aaawan-chu-subramoni-lu-panda-ocdnnexploitingadvancedunifiedmemorycapabilitiesincuda9andvoltagpusforoutofcorednntraining-2018\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  OC-DNN: Exploiting Advanced Unified Memory Capabilities in CUDA 9 and Volta GPUs for Out-of-Core DNN Training.\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan; Chu, C.; Subramoni, H.; Lu, X.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>2018 IEEE 25th International Conference on High Performance Computing (HiPC)</i>, pages 143–152, dec 2018. \n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n\n  \n  <a href=\"http://doi.org/10.1109/HiPC.2018.00024\"\n     onclick=\"log_download('aaawan-chu-subramoni-lu-panda-ocdnnexploitingadvancedunifiedmemorycapabilitiesincuda9andvoltagpusforoutofcorednntraining-2018', 'http://doi.org/10.1109/HiPC.2018.00024', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-chu-subramoni-lu-panda-ocdnnexploitingadvancedunifiedmemorycapabilitiesincuda9andvoltagpusforoutofcorednntraining-2018\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-chu-subramoni-lu-panda-ocdnnexploitingadvancedunifiedmemorycapabilitiesincuda9andvoltagpusforoutofcorednntraining-2018')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@inproceedings{awan-hipc18,\nauthor = {{A. A. Awan} and C-H Chu and H. Subramoni and X. Lu and DK Panda},\nbooktitle = {2018 IEEE 25th International Conference on High Performance Computing (HiPC)},\ndoi = {10.1109/HiPC.2018.00024},\nissn = {2640-0316},\nkeywords = {graphics processing units;learning (artificial int},\nmonth = {dec},\npages = {143--152},\ntitle = {{OC-DNN: Exploiting Advanced Unified Memory Capabilities in CUDA 9 and Volta GPUs for Out-of-Core DNN Training}},\nyear = {2018}\n}\n</pre>\n</div>\n\n\n\n</div>\n\n\n\n\n\n    </div>\n  </div>\n\n  <div class=\"bibbase_group_whole\" id=\"group_2017\">\n    <div class=\"bibbase_group\"\n      onclick=\"toggleGroup('group_2017')\"\n      onkeypress=\"toggleGroup('group_2017')\"\n      tabindex=\"0\">\n      <i class=\"fa fa-angle-down bibbase_group_head_icon\"></i>&nbsp;\n      <span>2017</span>\n      <span class=\"bibbase_group_count\">\n        \n        (2)\n        \n      </span>\n    </div>\n    <div class=\"bibbase_group_body\">\n      \n  \n  <div class=\"bibbase_paper\" id=\"awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  <a href=\"http://doi.acm.org/10.1145/3018743.3018769\"\n     onclick=\"return trackOutboundLink('publications', 'click', 'awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017', 'http://doi.acm.org/10.1145/3018743.3018769', event);\">\n    S-Caffe: Co-designing MPI Runtimes and Caffe for Scalable Deep Learning on Modern GPU Clusters.\n  </a>\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  Awan, A. A.; Hamidouche, K.; Hashmi, J. M.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D. K.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>Proceedings of the 22Nd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>, of <i>PPoPP '17</i>, pages 193–205, New York, NY, USA,  2017. ACM\n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n  <a href=\"http://doi.acm.org/10.1145/3018743.3018769\"\n     onclick=\"log_download('awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017', 'http://doi.acm.org/10.1145/3018743.3018769', event.ctrlKey, event)\">\n    <img src=\"//bibbase.org/img/filetypes/link.svg\"\n\t     alt=\"S-Caffe: Co-designing MPI Runtimes and Caffe for Scalable Deep Learning on Modern GPU Clusters [link]\"\n       title=\"Paper\"\n\t     class=\"bibbase_icon\"\n         ><span class=\"bibbase_icon_text\">Paper</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"http://doi.org/10.1145/3018743.3018769\"\n     onclick=\"log_download('awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017', 'http://doi.org/10.1145/3018743.3018769', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n  &nbsp;\n  <span class=\"bibbase_stats_paper\" style=\"color: #777;\">\n    <span>1 download</span>\n  </span>\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('awan-hamidouche-hashmi-panda-scaffecodesigningmpiruntimesandcaffeforscalabledeeplearningonmoderngpuclusters-2017')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@inproceedings{awan-ppopp17,\n author = {Awan, Ammar Ahmad and Hamidouche, Khaled and Hashmi, Jahanzeb Maqbool and Panda, Dhabaleswar K.},\n title = {{S-Caffe: Co-designing MPI Runtimes and Caffe for Scalable Deep Learning on Modern GPU Clusters}},\n booktitle = {Proceedings of the 22Nd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming},\n series = {PPoPP &#x27;17},\n year = {2017},\n isbn = {978-1-4503-4493-7},\n location = {Austin, Texas, USA},\n pages = {193--205},\n numpages = {13},\n url = {http://doi.acm.org/10.1145/3018743.3018769},\n doi = {10.1145/3018743.3018769},\n acmid = {3018769},\n publisher = {ACM},\n address = {New York, NY, USA},\n keywords = {caffe, cuda-aware mpi, deep learning, distributed training, mpi\\_reduce},\n} \n\n</pre>\n</div>\n\n\n\n</div>\n\n\n  <div class=\"bibbase_paper\" id=\"aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  <a href=\"http://doi.acm.org/10.1145/3146347.3146356\"\n     onclick=\"return trackOutboundLink('publications', 'click', 'aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017', 'http://doi.acm.org/10.1145/3146347.3146356', event);\">\n    An In-depth Performance Characterization of CPU- and GPU-based DNN Training on Modern Architectures.\n  </a>\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan; Subramoni, H.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>Proceedings of the Machine Learning on HPC Environments</i>, of <i>MLHPC'17</i>, pages 8:1–8:8, New York, NY, USA,  2017. ACM\n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n  <a href=\"http://doi.acm.org/10.1145/3146347.3146356\"\n     onclick=\"log_download('aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017', 'http://doi.acm.org/10.1145/3146347.3146356', event.ctrlKey, event)\">\n    <img src=\"//bibbase.org/img/filetypes/link.svg\"\n\t     alt=\"An In-depth Performance Characterization of CPU- and GPU-based DNN Training on Modern Architectures [link]\"\n       title=\"Paper\"\n\t     class=\"bibbase_icon\"\n         ><span class=\"bibbase_icon_text\">Paper</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"http://doi.org/10.1145/3146347.3146356\"\n     onclick=\"log_download('aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017', 'http://doi.org/10.1145/3146347.3146356', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-subramoni-panda-anindepthperformancecharacterizationofcpuandgpubaseddnntrainingonmodernarchitectures-2017')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@inproceedings{awan-mlhpc17, \n author = {{A. A. Awan} and H. Subramoni and DK Panda},\n title = {{An In-depth Performance Characterization of CPU- and GPU-based DNN\n     Training on Modern Architectures}},\n booktitle = {Proceedings of the Machine Learning on HPC Environments},\n series = {MLHPC&#x27;17},\n year = {2017},\n isbn = {978-1-4503-5137-9},\n location = {Denver, CO, USA},\n pages = {8:1--8:8},\n articleno = {8},\n numpages = {8},\n url = {http://doi.acm.org/10.1145/3146347.3146356},\n doi = {10.1145/3146347.3146356},\n acmid = {3146356},\n publisher = {ACM},\n address = {New York, NY, USA},\n keywords = {Caffe, Deep Learning, High-Performance Computing, Pascal Architecture, Unified Memory},\n} \n\n</pre>\n</div>\n\n\n\n</div>\n\n\n\n\n\n    </div>\n  </div>\n\n  <div class=\"bibbase_group_whole\" id=\"group_2016\">\n    <div class=\"bibbase_group\"\n      onclick=\"toggleGroup('group_2016')\"\n      onkeypress=\"toggleGroup('group_2016')\"\n      tabindex=\"0\">\n      <i class=\"fa fa-angle-down bibbase_group_head_icon\"></i>&nbsp;\n      <span>2016</span>\n      <span class=\"bibbase_group_count\">\n        \n        (1)\n        \n      </span>\n    </div>\n    <div class=\"bibbase_group_body\">\n      \n  \n  <div class=\"bibbase_paper\" id=\"aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016\">\n  <span class=\"bibbase_paper_titleauthoryear\">\n\n  \n  <span class=\"bibbase_paper_title\">\n  \n  \n  \n  <a href=\"http://doi.acm.org/10.1145/2966884.2966912\"\n     onclick=\"return trackOutboundLink('publications', 'click', 'aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016', 'http://doi.acm.org/10.1145/2966884.2966912', event);\">\n    Efficient Large Message Broadcast Using NCCL and CUDA-Aware MPI for Deep Learning.\n  </a>\n  \n  \n  \n</span>\n\n  <span class=\"bibbase_paper_author\">\n  A. A. Awan,; Hamidouche, K.; Venkatesh, A.;  and <a class=\"bibbase author link\" href=\"https://awan-10.github.io/research.html\">Panda, D.</a>\n</span>\n\n  \n\n</span>\n\n  In <i>Proceedings of the 23rd European MPI Users' Group Meeting</i>, of <i>EuroMPI 2016</i>, pages 15–22, New York, NY, USA,  2016. ACM\n  <span class=\"note\"></span>\n\n<span class=\"bibbase_note\"></span>\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content dontprint\">\n\n  \n  <a href=\"http://doi.acm.org/10.1145/2966884.2966912\"\n     onclick=\"log_download('aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016', 'http://doi.acm.org/10.1145/2966884.2966912', event.ctrlKey, event)\">\n    <img src=\"//bibbase.org/img/filetypes/link.svg\"\n\t     alt=\"Efficient Large Message Broadcast Using NCCL and CUDA-Aware MPI for Deep Learning [link]\"\n       title=\"Paper\"\n\t     class=\"bibbase_icon\"\n         ><span class=\"bibbase_icon_text\">Paper</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"http://doi.org/10.1145/2966884.2966912\"\n     onclick=\"log_download('aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016', 'http://doi.org/10.1145/2966884.2966912', event.ctrlKey)\"\n     class=\"bibbase doi link\">\n    <span>doi</span></a>\n  &nbsp;\n  \n\n  \n  <a href=\"//bibbase.org/network/publication/aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016\"\n     class=\"bibbase bibbase_page link\">link</a>\n  &nbsp;\n  \n\n  <a href=\"#\"\n     onclick=\"showBib(event); return false;\"\n     class=\"bibbase bibtex link\">bibtex\n    <i class=\"fa fa-caret-down\"></i></a>\n\n  \n\n  \n\n  <!-- <a class=\"bibbase read link hidden\" -->\n  <!--    href=\"javascript:toggleRead('aaawan-hamidouche-venkatesh-panda-efficientlargemessagebroadcastusingncclandcudaawarempifordeeplearning-2016')\" -->\n  <!--    title=\"Marking this paper as read adds it to your library on BibBase.\" -->\n  <!--    > -->\n  <!--   <i class=\"fa fa-check\"></i>&nbsp; -->\n  <!--   <span>mark as read</span> -->\n  <!-- </a> -->\n\n  &nbsp;\n  \n  \n\n</span>\n\n<div class=\"well well-small bibbase\" data-type=\"bibtex\"\n     style=\"display:none\">\n  <pre style=\"white-space: pre-wrap;\">@inproceedings{awan-eurompi16,\n author = {{A. A. Awan,} and Hamidouche, K. and Venkatesh, A. and Panda, DK},\n title = {{Efficient Large Message Broadcast Using NCCL and CUDA-Aware MPI for\nDeep Learning}},\n booktitle = {Proceedings of the 23rd European MPI Users&#x27; Group Meeting},\n series = {EuroMPI 2016},\n year = {2016},\n isbn = {978-1-4503-4234-6},\n location = {Edinburgh, United Kingdom},\n pages = {15--22},\n numpages = {8},\n url = {http://doi.acm.org/10.1145/2966884.2966912},\n doi = {10.1145/2966884.2966912},\n acmid = {2966912},\n publisher = {ACM},\n address = {New York, NY, USA},\n}\n\n</pre>\n</div>\n\n\n\n</div>\n\n\n\n\n\n    </div>\n  </div>\n\n\n\n\n  </div>\n\n\n  <!-- Modal -->\n\n  <!-- <iframe id=\"bibbase_network_frame\" -->\n  <!--         src=\"//bibbase.org/network/control\" -->\n  <!--         style=\"display: none;\"> -->\n  <!-- </iframe> -->\n\n</div>\n"}; document.write(bibbase_data.data);