Hands-On GPU Programming with Python and CUDA
Hands-OnGPUProgrammingwithPythonandCUDAhitsthegroundrunning:you’llstartbylearninghowtoapplyAmdahl’sLaw,useacodeprofilertoidentifybottlenecksinyourPythoncode,andsetupanappropriateGPUprogrammingenvironment.You’llthenseehowto"query"theGPU’sfeaturesandcopyarraysofdatatoandfromtheGPU’sownmemory.Asyoumakeyourwaythroughthebook,you’lllaunchcodedirectlyontotheGPUandwritefullblownGPUkernelsanddevicefunctionsinCUDAC.You’llgettogripswithprofilingGPUcodeeffectivelyandfullytestanddebugyourcodeusingNsightIDE.Next,you’llexploresomeofthemorewell-knownNVIDIAlibraries,suchascuFFTandcuBLAS.Withasolidbackgroundinplace,youwillnowapplyyournew-foundknowledgetodevelopyourveryownGPU-baseddeepneuralnetworkfromscratch.You’llthenexploreadvancedtopics,suchaswarpshuffling,dynamicparallelism,andPTXassembly.Inthefinalchapter,you’llseesometopicsandapplicationsrelatedtoGPUprogrammingthatyoumaywishtopursue,includingAI,graphics,andblockchain.Bytheendofthisbook,youwillbeabletoapplyGPUprogrammingtoproblemsrelatedtodatascienceandhigh-performancecomputing.
·7.4萬字