{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,9,24]],"date-time":"2025-09-24T09:17:18Z","timestamp":1758705438698,"version":"3.37.3"},"publisher-location":"Berlin, Heidelberg","reference-count":18,"publisher":"Springer Berlin Heidelberg","isbn-type":[{"type":"print","value":"9783642400469"},{"type":"electronic","value":"9783642400476"}],"license":[{"start":{"date-parts":[[2013,1,1]],"date-time":"2013-01-01T00:00:00Z","timestamp":1356998400000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/www.springer.com\/tdm"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2013]]},"DOI":"10.1007\/978-3-642-40047-6_82","type":"book-chapter","created":{"date-parts":[[2013,7,20]],"date-time":"2013-07-20T12:18:02Z","timestamp":1374322682000},"page":"826-837","update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":5,"title":["Optimizing 3D Convolutions for Wavelet Transforms on CPUs with SSE Units and GPUs"],"prefix":"10.1007","author":[{"given":"Brice","family":"Videau","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Vania","family":"Marangozova-Martin","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Luigi","family":"Genovese","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Thierry","family":"Deutsch","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","reference":[{"key":"82_CR1","unstructured":"The BigDFT Scientific Application (2012), \n                    \n                      http:\/\/inac.cea.fr\/LSim\/BigDFT\/"},{"key":"82_CR2","unstructured":"The Mont-Blanc Project (2012), \n                    \n                      http:\/\/www.montblanc-project.eu"},{"key":"82_CR3","unstructured":"Danovich, Z.: 16bit 3D Convolution: SSE4+OpenMP implementation on Penryn CPU, \n                    \n                      http:\/\/software.intel.com\/en-us\/articles\/16bit-3d-convolution-sse4openmp-implementation-on-penryn-cpu\/"},{"key":"82_CR4","unstructured":"Fialka, O., Cadik, M.: FFT and Convolution Performance in Image Filtering on GPU. In: Tenth International Conference on Information Visualization, IV 2006, pp. 609\u2013614 (July 2006)"},{"key":"82_CR5","doi-asserted-by":"crossref","first-page":"014109","DOI":"10.1063\/1.2949547","volume":"129","author":"L. Genovese","year":"2008","unstructured":"Genovese, L., Neelov, A., Goedecker, S., Deutsch, T., Ghasemi, S., Willand, A., Caliste, D., Zilberberg, O., Rayson, M., Bergman, A., et al.: Daubechies Wavelets as a Basis Set for Density Functional Pseudopotential Calculations. The Journal of Chemical Physics\u00a0129, 014109 (2008)","journal-title":"The Journal of Chemical Physics"},{"key":"82_CR6","doi-asserted-by":"crossref","first-page":"034103","DOI":"10.1063\/1.3166140","volume":"131","author":"L. Genovese","year":"2009","unstructured":"Genovese, L., Ospici, M., Deutsch, T., M\u00e9haut, J., Neelov, A., Goedecker, S.: Density functional theory calculation on many-cores hybrid central processing unit-graphic processing unit architectures. The Journal of Chemical Physics\u00a0131, 034103 (2009)","journal-title":"The Journal of Chemical Physics"},{"issue":"3","key":"82_CR7","doi-asserted-by":"publisher","first-page":"294","DOI":"10.1016\/0010-4655(93)90057-J","volume":"76","author":"S. Goedecker","year":"1993","unstructured":"Goedecker, S.: Rotating a three-dimensional array in an optimal position for vector processing: case study for a three-dimensional fast fourier transform. Computer Physics Communications\u00a076(3), 294\u2013300 (1993)","journal-title":"Computer Physics Communications"},{"issue":"2","key":"82_CR8","doi-asserted-by":"publisher","first-page":"105","DOI":"10.1016\/S0010-4655(03)00287-X","volume":"154","author":"S. Goedecker","year":"2003","unstructured":"Goedecker, S., Boulet, M., Deutsch, T.: An efficient 3-dim fft for plane wave electronic structure calculations on massively parallel machines composed of multiprocessor nodes. Computer Physics Communications\u00a0154(2), 105\u2013110 (2003)","journal-title":"Computer Physics Communications"},{"key":"82_CR9","first-page":"471","volume-title":"Proceedings of the Conference on Visualization 1999: Celebrating Ten Years, VIS 1999","author":"M. Hopf","year":"1999","unstructured":"Hopf, M., Ertl, T.: Accelerating 3D Convolution Using Graphics Hardware (Case Study). In: Proceedings of the Conference on Visualization 1999: Celebrating Ten Years, VIS 1999, pp. 471\u2013474. IEEE Computer Society Press, Los Alamitos (1999), \n                    \n                      http:\/\/dl.acm.org\/citation.cfm?id=319351.319457"},{"key":"82_CR10","unstructured":"Khronos OpenCL consortium: OpenCL: Open Computing Language, \n                    \n                      http:\/\/www.khronos.org\/opencl\/"},{"key":"82_CR11","unstructured":"Mucci, P., Browne, S., Deane, C., Ho, G.: PAPI: A Portable Interface to Hardware Performance Counters. In: Proc. Dept. of Defense HPCMP Users Group Conference, pp. 7\u201310 (1999)"},{"key":"82_CR12","series-title":"Lecture Notes in Computer Science","doi-asserted-by":"publisher","first-page":"958","DOI":"10.1007\/978-3-540-74742-0_84","volume-title":"Parallel and Distributed Processing and Applications","author":"A. Nukada","year":"2007","unstructured":"Nukada, A., Hourai, Y., Nishida, A., Akiyama, Y.: High Performance 3D Convolution for Protein Docking on IBM Blue Gene. In: Stojmenovic, I., Thulasiram, R.K., Yang, L.T., Jia, W., Guo, M., de Mello, R.F. (eds.) ISPA 2007. LNCS, vol.\u00a04742, pp. 958\u2013969. Springer, Heidelberg (2007)"},{"key":"82_CR13","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-81897-4","volume-title":"Fast Fourier Transform and Convolution Algorithms","author":"H. Nussbaumer","year":"1982","unstructured":"Nussbaumer, H.: Fast Fourier Transform and Convolution Algorithms, vol.\u00a02. Springer, Berlin (1982)"},{"key":"82_CR14","unstructured":"nVidia, C.: OpenCL Best Practices Guide (2010), \n                    \n                      http:\/\/www.nvidia.com\/content\/cudazone\/CUDABrowser\/downloads\/papers\/NVIDIA_OpenCL_BestPracticesGuide.pdf"},{"key":"82_CR15","unstructured":"Podlozhnyuk, V.: Image Convolution with CUDA. NVIDIA Corporation White Paper 2097(3) (June 2007)"},{"issue":"12","key":"82_CR16","doi-asserted-by":"publisher","first-page":"26","DOI":"10.1109\/2.809248","volume":"32","author":"S. Thakkur","year":"1999","unstructured":"Thakkur, S., Huff, T.: Internet Streaming SIMD Extensions. Computer\u00a032(12), 26\u201334 (1999)","journal-title":"Computer"},{"issue":"1","key":"82_CR17","doi-asserted-by":"publisher","first-page":"100","DOI":"10.1002\/cpe.1620","volume":"23","author":"A. Va\u0161ko","year":"2011","unstructured":"Va\u0161ko, A., \u0160r\u00e1mek, M.: Optimizing Gaussian Filtering of Volumetric Data Using SSE. Concurrency and Computation: Practice and Experience\u00a023(1), 100\u2013116 (2011), \n                    \n                      http:\/\/dx.doi.org\/10.1002\/cpe.1620","journal-title":"Concurrency and Computation: Practice and Experience"},{"issue":"4","key":"82_CR18","doi-asserted-by":"publisher","first-page":"452","DOI":"10.1109\/71.97902","volume":"2","author":"M. Wolf","year":"1991","unstructured":"Wolf, M., Lam, M.: A Loop Transformation Theory and an Algorithm to Maximize Parallelism. IEEE Transactions on Parallel and Distributed Systems\u00a02(4), 452\u2013471 (1991)","journal-title":"IEEE Transactions on Parallel and Distributed Systems"}],"container-title":["Lecture Notes in Computer Science","Euro-Par 2013 Parallel Processing"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/link.springer.com\/content\/pdf\/10.1007\/978-3-642-40047-6_82","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2020,4,20]],"date-time":"2020-04-20T01:09:46Z","timestamp":1587344986000},"score":1,"resource":{"primary":{"URL":"http:\/\/link.springer.com\/10.1007\/978-3-642-40047-6_82"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2013]]},"ISBN":["9783642400469","9783642400476"],"references-count":18,"URL":"https:\/\/doi.org\/10.1007\/978-3-642-40047-6_82","relation":{},"ISSN":["0302-9743","1611-3349"],"issn-type":[{"type":"print","value":"0302-9743"},{"type":"electronic","value":"1611-3349"}],"subject":[],"published":{"date-parts":[[2013]]},"assertion":[{"value":"This content has been made available to all.","name":"free","label":"Free to read"}]}}