1、Cuda 平行運算機制報告者:林威辰 Slides MPI基本定理 各種平行運算的簡介 CUDA簡介 使用 VS 建置 CUDA CUDA基本知識 CUDA硬體架構 CUDA缺點 My ResearchSlides http:/courses.ece.uiuc.edu/ece498/al/Syllabus.html National Center for High-Performance Computing http:/ http:/pccluster.nchc.org.tw/main/tutor/09nctu/ http:/.tw/object/cuda_home_tw.html NCHC教
2、育訓練網 https:/edu.nchc.org.tw/ http:/ http:/ Computing Introduction Flynns Taxonomy Amdahls Law Moores Law Finding ConcurrencyFlynns TaxonomyAmdahls Law Expected speedup from partial improvement P: proportion of program that is parallel S: speedup of parallel portionMoores Law The number of transistors on ICs doubles every 18 months In the past, CPUs have been taking advantage of Moore,s Law to: Increase clock frequency Increase exploitation of ILP ILP: Instruction-Level Parallelism Result: CPUs gets faster Increasingly difficult to get faster各種平行運算的簡介平行運算示意圖