{"id":1930,"date":"2022-08-08T13:39:06","date_gmt":"2022-08-08T13:39:06","guid":{"rendered":"http:\/\/81.70.49.155\/?p=1930"},"modified":"2023-01-12T10:12:08","modified_gmt":"2023-01-12T10:12:08","slug":"%e5%8c%97%e4%ba%ac%e5%b9%b6%e8%a1%8c%e7%a7%91%e6%8a%80gpu-cuda-fortran%e8%ae%a1%e7%ae%97%e5%b9%b3%e5%8f%b0%e9%85%8d%e7%bd%ae","status":"publish","type":"post","link":"http:\/\/81.70.49.155\/?p=1930","title":{"rendered":"\u5317\u4eac\u5e76\u884c\u79d1\u6280GPU-CUDA Fortran\u8ba1\u7b97\u5e73\u53f0\u914d\u7f6e"},"content":{"rendered":"

\n\t\u542f\u7528MPI OpenMP
\n<\/h3>\n

\n\tmodule load nvhpc\/nvhpc\/21.7_cuda11.0_openmpi4.0.5_gcc7.5_qd
\n<\/h3>\n

\n\t
\n<\/h3>\n

\n\t\u8c03\u7528ifort \u7f16\u8bd1\u5668
\n<\/h3>\n

\n\t module load intel\/parallelstudio\/2017.1.5<\/span><\/b>
\n<\/h3>\n

\n\t
\n<\/b>
\n<\/h3>\n

\n\tStep1 <\/b>\u62a2\u5360\u4e00\u4e2a<\/b>GPU<\/b>\u8ba1\u7b97\u8282\u70b9<\/b>
\n<\/h3>\n

\n\t<\/b>\n<\/p>\n

\n\t[scv1172@ln01 xyqtest2022]$ salloc --gpus=1<\/span>\n<\/p>\n

\n\tsalloc:Pending job allocation 308347\n<\/p>\n

\n\tsalloc:job 308347 queued and waiting for resources\n<\/p>\n

\n\tsalloc:job 308347 has been allocated resources\n<\/p>\n

\n\tsalloc:Granted job allocation 308347\n<\/p>\n

\n\tsalloc:Waiting for resource configuration\n<\/p>\n

\n\tsalloc:Nodes g0001 are ready for job  (\u627e\u5230\u7684\u8ba1\u7b97\u8282\u70b9\u4e3ag0001)\n<\/p>\n

\n\t \n<\/p>\n

\n\tStep2 <\/b>\u7528<\/b>ssh<\/b>\u547d\u4ee4\u52a0\u8f7d\u627e\u5230\u7684\u8282\u70b9<\/b> g0001<\/span><\/b>
\n<\/h3>\n

\n\t[scv1172@ln01 xyqtest2022]$ ssh g0001<\/span>\n<\/p>\n

\n\tWarning:Permanently added 'g0001,192.168.11.1' (ECDSA) to the list of known hosts.\n<\/p>\n

\n\t[scv1172@g0001 ~]$\n<\/p>\n

\n\t\u4ee5\u4e0a\u7ed3\u679c\u8868\u660e\u52a0\u8f7dg0001\u8282\u70b9\u6210\u529f\n<\/p>\n

\n\t \n<\/p>\n

\n\tStep3 <\/b>\u67e5\u8be2<\/b>NVHPC<\/b>\uff1a\u7528<\/b> module ava<\/span><\/b>\u547d\u4ee4\u67e5\u627e<\/b> <\/b>\u662f\u5426\u5b89\u88c5<\/b>nvhpc<\/b>\uff0c\u4ee5\u53ca\u7248\u672c<\/b>
\n<\/h3>\n

\n\t[scv1172@g0001 ~]$ module ava nvhpc<\/span>\n<\/p>\n

\n\t-------------------
\n\/data\/apps\/modulefiles --------------------------\n<\/p>\n

\n\tnvhpc\/nvhpc\/20.7_cuda10.1_openmpi3.1.5\n<\/p>\n

\n\tnvhpc\/nvhpc\/20.7_cuda10.2_openmpi3.1.5\n<\/p>\n

\n\tnvhpc\/nvhpc\/21.7_cuda11.0_openmpi4.0.5_gcc7.5_qd\n<\/p>\n

\n\tnvhpc\/nvhpc_byo_compiler\/20.7_cuda10.1\n<\/p>\n

\n\tnvhpc\/nvhpc_byo_compiler\/20.7_cuda10.2\n<\/p>\n

\n\tnvhpc\/nvhpc_byo_compiler\/21.7_cuda11.0_gcc7.5\n<\/p>\n

\n\tnvhpc\/nvhpc_nompi\/20.7_cuda10.1\n<\/p>\n

\n\tnvhpc\/nvhpc_nompi\/20.7_cuda10.2\n<\/p>\n

\n\tnvhpc\/nvhpc_nompi\/21.7_cuda11.0_gcc7.5_qd\n<\/p>\n

\n\t \n<\/p>\n

\n\tStep 4 <\/b>\u7528<\/b>module load<\/b>\u52a0\u8f7d\u76f8\u5e94\u7248\u672c\u7684<\/b>nvhpc<\/b>
\n<\/h3>\n

\n\t[scv1172@g0001 ~]$ module load nvhpc\/nvhpc_nompi\/21.7_cuda11.0_gcc7.5_qd<\/span>\n<\/p>\n

\n\t \n<\/p>\n

\n\tStep 5 <\/b>\u7528<\/b>which<\/b>\u547d\u4ee4\u68c0\u67e5<\/b>nvhpc<\/b>\u7684\u52a0\u8f7d\u662f\u5426\u6210\u529f<\/b>
\n<\/h3>\n

\n\t[scv1172@g0001 xyqtest2022]$ which nvfortran<\/span>\n<\/p>\n

\n\t\/data\/apps\/nvhpc\/nvhpc_21.7\/Linux_x86_64\/21.7\/compilers\/bin\/nvfortran\n<\/p>\n

\n\t[scv1172@g0001
\nxyqtest2022]$\n<\/p>\n

\n\t\u4ee5\u4e0a\u7ed3\u679c\u8868\u660e\u76f8\u5e94\u7248\u672c\u7684nvhpc\u52a0\u8f7d\u6210\u529f\n<\/p>\n

\n\t\u5229\u7528\u4ee5\u4e0a5\u6b65\u5219\u53ef\u4ee5\u5bf9Cuda-GPU\u7684\u8ba1\u7b97\u8282\u70b9\u8fdb\u6027\u914d\u7f6e\n<\/p>\n

\n\t\n<\/p>\n

\n\t~~~~~~\u6d4b\u8bd5\u4ee3\u7801(\u547d\u540d\u4e3atest.cuf)~~~~~~~~~~~~<\/span>
\n<\/h3>\n

\n\tprogram deviceQuery
\n use cudafor
\n type (cudaDeviceProp) :: prop
\n integer :: nDevices=0, i, ierr
\n ! Number of CUDA -capable devices
\nierr = cudaGetDeviceCount(nDevices)
\n if (nDevices == 0) then
\nwrite(*,\"(\/,'No CUDA devices found ',\/)\")
\n stop
\n else if (nDevices == 1) then
\n write(*,\"(\/,'One CUDA device found ',\/)\")
\n else
\n write(*,\"(\/,i0,' CUDA devices found ',\/)\") nDevices
\n end if<\/p>\n

! Loop over devices
\n do i = 0, nDevices -1
\n write(*,\"('Device Number: ',i0)\") i
\n ierr = cudaGetDeviceProperties(prop , i)
\n ! General device info
\n write(*,\"(' Device Name: ',a)\") trim(prop%name)
\n write(*,\"(' Compute Capability: ',i0,'.',i0)\") &
\n prop%major , prop%minor
\n write(*,\"(' Number of Multiprocessors: ',i0)\") &
\n prop%multiProcessorCount
\nwrite(*,\"(' Max Threads per Multiprocessor: ',i0)\") &
\n prop%maxThreadsPerMultiprocessor
\n write(*,\"(' Global Memory (GB): ',f9.3,\/)\") &
\n prop%totalGlobalMem \/1024.0**3<\/p>\n

 ! Execution Configuration
\n write(*,\"(' Execution Configuration Limits ')\")
\n write(*,\"(' Max Grid Dims: ',2(i0,' x '),i0)\") &
\n prop%maxGridSize
\n write(*,\"(' Max Block Dims: ',2(i0,' x '),i0)\") &
\n prop%maxThreadsDim
\n write(*,\"(' Max Threads per Block: ',i0 ,\/)\") &
\n prop%maxThreadsPerBlock
\n enddo
\nend program deviceQuery\n<\/p>\n

\n\t\n<\/p>\n

\n\t~~~~~~~~~~~~~\u7f16\u8bd1\u8fd0\u884c~~~~~~~~~~~~~~~~~<\/span>
\n<\/h3>\n

\n\t\n<\/p>\n

\n\t <\/span>[scv1172@g0001 xyqtest2022]$ nvfortran *.cuf -o ou<\/span>t<\/span>\n<\/p>\n

\n\t <\/span>[scv1172@g0001 xyqtest2022]$ .\/out<\/span>
\nOne CUDA device found 
\nDevice Number: 0
\n Device Name: Tesla V100-SXM2-32GB
\n Compute Capability: 7.0
\n Number of Multiprocessors: 80
\n Max Threads per Multiprocessor: 2048
\n Global Memory (GB):    31.749
\n Execution Configuration Limits 
\n Max Grid Dims: 2147483647 x 65535 x 65535
\n Max Block Dims: 1024 x 1024 x 64
\n Max Threads per Block: 1024\n<\/p>\n

\n\t<\/p>\n","protected":false},"excerpt":{"rendered":"

\u542f\u7528MPI OpenMP module load nvhpc\/nvhpc\/21.7_cuda11.0_open … \u7ee7\u7eed\u9605\u8bfb\u201c\u5317\u4eac\u5e76\u884c\u79d1\u6280GPU-CUDA Fortran\u8ba1\u7b97\u5e73\u53f0\u914d\u7f6e\u201d<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[12],"tags":[],"views":1449,"_links":{"self":[{"href":"http:\/\/81.70.49.155\/index.php?rest_route=\/wp\/v2\/posts\/1930"}],"collection":[{"href":"http:\/\/81.70.49.155\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/81.70.49.155\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/81.70.49.155\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/81.70.49.155\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1930"}],"version-history":[{"count":6,"href":"http:\/\/81.70.49.155\/index.php?rest_route=\/wp\/v2\/posts\/1930\/revisions"}],"predecessor-version":[{"id":1963,"href":"http:\/\/81.70.49.155\/index.php?rest_route=\/wp\/v2\/posts\/1930\/revisions\/1963"}],"wp:attachment":[{"href":"http:\/\/81.70.49.155\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1930"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/81.70.49.155\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1930"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/81.70.49.155\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1930"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}