{"id":188,"date":"2013-11-12T22:40:48","date_gmt":"2013-11-12T14:40:48","guid":{"rendered":"http:\/\/blog.stlover.org\/?p=188"},"modified":"2013-11-12T23:33:37","modified_gmt":"2013-11-12T15:33:37","slug":"cuda%ef%bc%9a%e8%92%99%e7%89%b9%e5%8d%a1%e6%b4%9b%e4%bb%a5%e5%8f%8a%e9%9a%8f%e6%9c%ba%e6%95%b0%e6%94%af%e6%8c%81","status":"publish","type":"post","link":"http:\/\/blog.xuhao1.me\/?p=188","title":{"rendered":"CUDA\uff1a\u8499\u7279\u5361\u6d1b\u4ee5\u53ca\u968f\u673a\u6570\u652f\u6301"},"content":{"rendered":"

\u7531\u4e8e\u968f\u673a\u6570\u662f\u6574\u4e2a\u8ba1\u7b97\u7269\u7406\u7684\u57fa\u77f3\uff0c\u5b9e\u5728\u592a\u8fc7\u4e8e\u91cd\u8981\uff0c\u5e72\u8106\u7528\u4e00\u7247\u6587\u7ae0\u6765\u53d9\u8ff0CUDA\u6709\u5173\u4e8e\u968f\u673a\u6570\u4ea7\u751f\u5668\u7684\u79cd\u79cd\u3002<\/p>\n

<\/p>\n

\u76ee\u5f55<\/a><\/p>\n

\u8fd9\u4e00\u7bc7\u6587\u7ae0\u6682\u65f6\u66f4\u65b0\u4e00\u4e9b\u968f\u673a\u6570\u76f8\u5173\u7684\u4e1c\u897f\uff0c\u7b2c\u4e00\u6b65\u7684\u7b80\u5355\u5e94\u7528\u8bf7\u53c2\u8003\u6211\u7684\u53e6\u4e00\u7bc7\u6587\u7ae0\uff0c\u5bf9\u4e8e\u4f7f\u7528CUDA\u505a\u8499\u7279\u5361\u6d1b\u79ef\u5206\u7684\u4e00\u4e9b\u63a2\u8ba8<\/a><\/p>\n

\u9996\u5148\uff1a\u5b66\u4e60\u8d44\u6599<\/p>\n

http:\/\/docs.nvidia.com\/cuda\/curand\/introduction.html<\/p>\n

\u6587\u7ae0\u4e2d\u5f88\u591a\u5730\u65b9\u6211\u5c31\u6458\u6284guide\u7684\u539f\u8bdd\u4e86\uff0c\u907f\u514d\u56e0\u4e3a\u7ffb\u8bd1\u9020\u6210\u7684\u7406\u89e3\u8bef\u5dee\u3002<\/p>\n

\u4f5c\u4e3a\u4e00\u4e2a\u4e25\u8c28\u7684\u7f16\u7a0b\u8005\uff0c\u5f88\u591a\u4e8b\u60c5\u4e0d\u4ec5\u8981\u8003\u8651\u7cfb\u7edfAPI\u7684\u901f\u5ea6\uff0c\u4e5f\u8981\u5bf9\u5176\u4e25\u8c28\u6027\u5ba1\u67e5\uff0c\u5c24\u5176\u8499\u7279\u5361\u6d1b\u8fd9\u79cd\u9760\u968f\u673a\u6570\u8d77\u5bb6\u7684\u73a9\u610f\u3002\u597d\u5728CUDA\u7ed9\u6211\u4eec\u63d0\u4f9b\u4e86\u4e00\u5957\u975e\u5e38\u9760\u8c31\u7684\u968f\u673a\u6570\u4ea7\u751f\u5668<\/p>\n

The CURAND library provides facilities that focus on the simple and efficient generation of high-quality pseudorandom and quasirandom numbers.<\/p><\/blockquote>\n

–\u6458\u81eaintroduction<\/p>\n

\u8fd9\u4e2aquasirandom\u8bf4\u7684\u53e3\u6c14\u5f88\u5927\u561b\uff01\u4e01\u6cfd\u519b\u8001\u5e08\u7684\u8ba1\u7b97\u7269\u7406\u8bfe\u4e5f\u5c31\u6562\u8bf4\u4f2a\u968f\u673a\u800c\u5df2\u3002<\/p>\n

\u90a3\u6211\u4eec\u6765\u4e00\u63a2\u7a76\u7adf\u5427\uff01<\/p>\n

\u9996\u5148\uff0cCUDA\u7684\u968f\u673a\u6570\u63d0\u4f9b\u5206\u4e3a\u4e24\u79cd\u7c7b\u578b\uff0c\u7b2c\u4e00\u662f\u7531GPU\u8fd0\u7b97\u5e76\u4e14\u4ea4\u7ed9\u4e3b\u673a\u7684\u968f\u673a\uff0c\u8ba9\u6211\u4e00\u4e0b\u5b50\u5c31\u60f3\u5230\u4e86\u8ba1\u7b97\u7269\u7406\u8bfe\u4e0a\u7684\u90a3\u4e2a\u201c\u968f\u673a\u6570\u786c\u4ef6\u4ea7\u751f\u5668\u201d\uff0c\u5927\u62b5\u5c31\u662f\u5982\u6b64\u4e86\u3002<\/p>\n

\u7528\u8fc7\u4e2d\u56fd\u94f6\u884c\u7f51\u94f6\u7684\u5b69\u5b50\u4f30\u8ba1\u89c1\u8fc7\u90a3\u79cd\u9ad8\u5927\u4e0a\u7684\u5bc6\u7801\u751f\u6210\u5668\uff0c\u5b83\u91c7\u7528\u6bcf\u5206\u949f\u7ed9\u51fa\u4e00\u4e2a6\u4f4d\u6570\u5b57\u7684\u5e8f\u5217\uff0c\u5e76\u548c\u7f51\u7edc\u670d\u52a1\u5668\u540c\u6b65\uff0c\u6709\u6548\u671f\u6709\u51e0\u5e74\uff0c\u7b97\u4e0b\u6765\u8fd9\u4e2a\u7ed9\u51fa\u968f\u673a\u6570\u7684\u5c0f\u73a9\u610f\u8981\u5728\u5176\u4e00\u751f\u4e2d\u628a6\u4f4d\u6570\u7684\u4e00\u534a\u4f5c\u4e3a\u6392\u5e8f\u7ed9\u51fa\uff0c\u5e76\u4e14\u8981\u548c\u7f51\u7edc\u670d\u52a1\u5668\u8fbe\u5230\u5206\u949f\u7684\u7cbe\u5ea6\uff0c\u90a3\u4e48\u5fc5\u7136\u662f\u5b58\u50a8\u597d\u7684\u4e86\u3002\u4e14\u4e0d\u8bba\u5176\u5b89\u5168\u6027\uff08\u4ee5\u540e\u6709\u673a\u4f1a\u53ef\u4ee5\u7814\u7a76\u4e0b\uff09\uff0c\u4f46\u5c31\u8fd9\u6837\u4e00\u4e2a\u8d85\u5927\u89c4\u6a21\u53c8\u6d89\u53ca\u91d1\u878d\u5b89\u5168\u7684\u73a9\u610f\uff0c\u5982\u679c\u4f60\u62ff16807\u7b97\u6cd5\u6765\u751f\u6210\u968f\u673a\u5e8f\u5217\u8fd8\u771f\u4e0d\u600e\u4e48\u9760\u8c31\u3002<\/p>\n

\u53e6\u4e00\u79cd\u5c31\u662f\u5728device\u4e0a\u76f4\u63a5\u8fd0\u884c\u7684\u968f\u673a\u6570\u4ea7\u751f\u5668\uff0c\u7531\u4e8e\u4e34\u8fd1\u8003\u8bd5\u5728\u6562\u4f5c\u4e1a\uff0c\u5c31\u5148\u4ecedevice\u7684\u968f\u673a\u6570\u4ea7\u751f\u5668\u8bb2\u8d77\u5427\u3002<\/p>\n

\u5173\u4e8edevice api\uff0cNVIDIA\u8fd9\u4e48\u53d9\u8ff0<\/p>\n

To use the device API, include the file\u00a0curand_kernel.h\u00a0in files that define kernels that use CURAND device functions.The device API includes functions\u00a0pseudorandom generation<\/a>\u00a0for and\u00a0quasirandom generation<\/a>\u00a0.<\/p><\/blockquote>\n

CUDA\u91c7\u7528\u51e0\u79cd\u968f\u673a\u6570\u751f\u6210\u7b97\u6cd5\uff0c\u5305\u62ec\uff1a<\/p>\n

\u00a0XORWOW\uff0c\u00a0MRG32k3a\uff0c\u00a0MTGP32<\/p><\/blockquote>\n

\u5177\u4f53\u7684\u7b97\u6cd5Google\u5427\u3002\u4ee5\u540e\u518d\u5206\u6790\u3002\uff08\u8c01\u521a\u521a\u8fd8\u5728\u8bf4\u81ea\u5df1\u662f\u4e25\u8c28\u7684\u7f16\u7a0b\u8005\u5462\u3002\u3002\u53e3\u4ea8\u3002\u3002\u3002\uff09<\/p>\n

\u4e8e\u662f\uff0c\u968f\u673a\u6570\u4ea7\u751f\u5668\u4e00\u822c\u8981\u505a\u4e14\u53ea\u9700\u8981\u4e24\u4ef6\u4e8b\u60c5\uff0c\u521d\u59cb\u5316\u4e0e\u83b7\u5f97\u503c\uff0c\u5927\u7ea6\u8fd8\u8bb0\u5f97Pascal\u7684\u968f\u673a\u6570\u4ea7\u751f\u5668\u662f\u8fd9\u4e48\u5199\u7684<\/p>\n

    \n
  1. …<\/span><\/li>\n
  2. randomize<\/span>();<\/span><\/li>\n
  3. a<\/span>:=<\/span>random<\/span>();<\/span><\/li>\n
  4. …<\/span><\/li>\n<\/ol>\n

    \u975e\u5e38\u7701\u65f6\u7701\u5fc3\uff0c\u53ef\u5bf9\u4e8eCUDA\u6211\u4eec\u5c31\u5f97\u4e00\u6b65\u6b65\u6765\u54af\u3002<\/p>\n

    \u521d\u59cb\u5316\u51fd\u6570\u7684\u5b9a\u4e49\u5982\u4e0b\uff1a<\/p>\n

      \n
    1. __device__\u00a0<\/span>void<\/span><\/li>\n
    2. curand_init\u00a0<\/span>(<\/span><\/li>\n
    3. \u00a0\u00a0\u00a0\u00a0<\/span>unsigned<\/span>\u00a0<\/span>long<\/span>\u00a0<\/span>long<\/span>\u00a0seed<\/span>,<\/span>\u00a0<\/span>unsigned<\/span>\u00a0<\/span>long<\/span>\u00a0<\/span>long<\/span>\u00a0sequence<\/span>,<\/span><\/li>\n
    4. \u00a0\u00a0\u00a0\u00a0<\/span>unsigned<\/span>\u00a0<\/span>long<\/span>\u00a0<\/span>long<\/span>\u00a0offset<\/span>,<\/span>\u00a0<\/span>curandState_t<\/span>\u00a0<\/span>*<\/span>state<\/span>)<\/span><\/li>\n<\/ol>\n

      \u539f\u6587\u89e3\u91ca\u5982\u4e0b<\/p>\n

      The\u00a0curand_init()<\/samp>\u00a0function sets up an initial state allocated by the caller using the given seed, sequence number, and offset within the sequence. Different seeds are guaranteed to produce different starting states and different sequences. The same seed always produces the same state and the same sequence. The state set up will be the state after<\/p>\n2<\/math>\n

      67\u00a0\u22c5\u00a0sequence<\/samp>\u00a0+\u00a0offset<\/samp>\u00a0calls to\u00a0curand()<\/samp>\u00a0from the seed state.<\/p><\/blockquote>\n

      CUDA\u53ef\u4ee5\u4ea7\u751f\u82e5\u5e72\u968f\u673a\u5e8f\u5217\uff0c\u6bcf\u4e00\u4e2a\u968f\u673a\u5e8f\u5217\u4ea7\u751f\u6709\u4e24\u4e2a\u5c5e\u6027\uff1a\u7b2c\u4e00\u662f\u4ea7\u751f\u5176\u7684\u79cd\u5b50\uff0c\u4e5f\u5c31\u662fseed\uff0c\u53e6\u4e00\u662f\u968f\u673a\u6570\u4ea7\u751f\u7684\u504f\u79fb\u91cf\uff0c\u56e0\u4e3a\u6211\u4eec\u77e5\u9053\u968f\u673a\u6570\u4ea7\u751f\u90fd\u662f\u4e00\u4e2a\u8fed\u4ee3\u8fc7\u7a0b\uff0cCUDA\u63d0\u4f9b\u4e86\u5728\u521d\u59cb\u5316\u65f6\u5019\u5c31\u8fed\u4ee3\u82e5\u5e72\u6b21\u7684\u7b97\u6cd5\u6765\u89e3\u51b3\u79cd\u5b50\u8fc7\u5c11\u7684\u95ee\u9898\u3002<\/p>\n

      \u4f20\u7edf\u800c\u5e38\u89c1\u7684\u79cd\u5b50\u63d0\u4f9b\u505a\u6cd5\u662f\u4f7f\u7528\u5f53\u524d\u7cfb\u7edf\u65f6\u95f4\uff08\u5fae\u79d2\u6570\uff09\uff0c\u6765\u4f5c\u4e3a\u79cd\u5b50\u7ed9\u51fa\uff0c\u5bf9\u4e8e\u4e32\u884c\u7a0b\u5e8f\u8fd9\u7b80\u5355\u800c\u6709\u6548\uff0c\u4f46\u662f\u663e\u7136\u5bf9\u4e8e\u5e76\u884c\u7a0b\u5e8f\uff0c\u8fd9\u5e76\u975e\u660e\u667a\u7684\u505a\u6cd5\uff0c\u56e0\u4e3a\u6211\u4eec\u5728\u4e3b\u673a\u542f\u52a8GPU\u5185\u6838\u7684\u65f6\u5019\u65f6\u95f4\u662f\u4e00\u5b9a\u7684\u4e14\u53ea\u80fd\u7528\u8fd9\u4e2a\u503c\u4f5c\u4e3a\u79cd\u5b50\uff0c\u5373\u4f7f\u8fdb\u884c\u4e00\u4e9b\u5904\u7406\u4e5f\u672a\u5fc5\u53ef\u4ee5\u8fbe\u5230\u8981\u6c42\uff0c\u4e8e\u662f\u4e4e\uff0cCUDA\u63d0\u4f9b\u4e86\u8fd9\u6837\u4e00\u4e2a\u89e3\u51b3\u65b9\u6848\uff0c\u5bf9\u4e8e\u975e\u4e25\u8c28\u7684\u4f7f\u7528\u8005\u53cd\u800c\u66f4\u52a0\u4f7f\u7528\u3002<\/p>\n

      \u5bf9\u4e8eCUDA\u7684\u6bcf\u4e2a\u968f\u673a\u5e8f\u5217\uff0c\u4f7f\u7528\u4e00\u4e2aState\u7ed9\u51fa\u6807\u5fd7\uff0c\u7136\u540e\u5728\u8c03\u7528\u67d0\u4e00\u5e8f\u5217\u7684\u4e0b\u4e00\u4e2a\u503c\u7684\u65f6\u5019\u4f7f\u7528<\/p>\n

        \n
      1. __device__\u00a0<\/span>unsigned<\/span>\u00a0<\/span>int<\/span><\/li>\n
      2. curand\u00a0<\/span>(<\/span>curandState_t<\/span>\u00a0<\/span>*<\/span>state<\/span>);<\/span><\/li>\n<\/ol>\n

        \u5373\u53ef\u3002\u7ed9\u51fa\u4f8b\u5b50\uff0c\u5b8c\u6574\u4ee3\u7801\u8bf7\u53c2\u89c1\u6211\u7684GitHub<\/a><\/p>\n

        \u5bf9\u4e8e\u5185\u6838\u4ee3\u7801<\/p>\n

          \n
        1. __global__\u00a0<\/span>void<\/span>\u00a0random_gpu<\/span>(<\/span>long<\/span>*<\/span>\u00a0C<\/span>,<\/span>long<\/span>*<\/span>\u00a0time<\/span>,<\/span>curandState<\/span>*<\/span>state<\/span>)<\/span><\/li>\n
        2. {<\/span><\/li>\n
        3. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0i\u00a0<\/span>=<\/span>\u00a0threadIdx<\/span>.<\/span>x<\/span>;<\/span><\/li>\n
        4. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0seed<\/span>=(*<\/span>time<\/span>)*(<\/span>i<\/span>+<\/span>1<\/span>);<\/span>\/\/\u56e0\u4e3a\u6240\u6709\u7ed9\u5b9a\u65f6\u95f4\u4e00\u5b9a\uff0c\u6240\u4ee5\u6211\u4eec\u53ea\u80fd\u901a\u8fc7\u5bf9\u65f6\u95f4\u8fdb\u884c\u7b80\u5355\u5904\u7406<\/span><\/li>\n
        5. \u00a0\u00a0\u00a0\u00a0<\/span>int<\/span>\u00a0offset<\/span>=<\/span>0<\/span>;<\/span>\/\/\u5b8c\u5168\u72ec\u7acb\u7684\u5e8f\u5217\uff0c\u6240\u4ee5offset\u5168\u90e8\u4e3a\u96f6\u6765\u8282\u7ea6\u65f6\u95f4<\/span><\/li>\n
        6. \u00a0\u00a0\u00a0\u00a0curand_init\u00a0<\/span>(<\/span>seed<\/span>,<\/span>i<\/span>,<\/span>offset<\/span>,><\/span>state<\/span>[<\/span>i<\/span>]);<\/span>\/\/\u8bbe\u7f6e\u7b2ci\u4e2a\u968f\u673a\u5e8f\u5217<\/span><\/li>\n
        7. \u00a0\u00a0\u00a0\u00a0C<\/span>[<\/span>i<\/span>]=<\/span>curand<\/span>(><\/span>state<\/span>[<\/span>i<\/span>]);<\/span>\/\/\u83b7\u5f97\u7b2ci\u4e2a\u968f\u673a\u5e8f\u5217\u7684\u968f\u673a\u503c<\/span><\/li>\n
        8. }<\/span><\/li>\n<\/ol>\n

          \u5bf9\u4e8e\u4e3b\u51fd\u6570<\/p>\n

            \n
          1. int<\/span>\u00a0main<\/span>()<\/span><\/li>\n
          2. {<\/span><\/li>\n
          3. \u00a0\u00a0\u00a0\u00a0<\/span>size_t<\/span>\u00a0size\u00a0<\/span>=<\/span>\u00a0N\u00a0<\/span>*<\/span>\u00a0<\/span>sizeof<\/span>(<\/span>float<\/span>);<\/span><\/li>\n
          4. \u00a0<\/span><\/li>\n
          5. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>*<\/span>\u00a0C<\/span>=<\/span>new<\/span>\u00a0<\/span>long<\/span>[<\/span>N<\/span>];<\/span><\/li>\n
          6. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0st<\/span>=<\/span>getCurrentTime<\/span>();<\/span><\/li>\n
          7. \u00a0\u00a0\u00a0\u00a0curandState\u00a0<\/span>*<\/span>state<\/span>;<\/span><\/li>\n
          8. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0<\/span>*<\/span>d_C<\/span>;<\/span><\/li>\n
          9. \u00a0\u00a0\u00a0\u00a0cudaMalloc<\/span>(><\/span>state<\/span>,<\/span>sizeof<\/span>(<\/span>curandState<\/span>)*<\/span>N<\/span>);<\/span>\/\/\u8bbe\u7acb\u968f\u673a\u72b6\u6001\u5217<\/span><\/li>\n
          10. \u00a0\u00a0\u00a0\u00a0cudaMalloc<\/span>(><\/span>d_C<\/span>,<\/span>\u00a0size<\/span>);<\/span><\/li>\n
          11. \u00a0\u00a0\u00a0\u00a0random_gpu<\/span><<<<\/span>1<\/span>,<\/span>N<\/span>>>>(<\/span>d_C<\/span>,<\/span>getCurrentTimeForDev<\/span>(),<\/span>state<\/span>);<\/span><\/li>\n
          12. \u00a0\u00a0\u00a0\u00a0cudaMemcpy<\/span>(<\/span>C<\/span>,<\/span>\u00a0d_C<\/span>,<\/span>\u00a0size<\/span>,<\/span>\u00a0cudaMemcpyDeviceToHost<\/span>);<\/span><\/li>\n
          13. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0ed<\/span>=<\/span>getCurrentTime<\/span>();<\/span><\/li>\n
          14. \u00a0\u00a0\u00a0\u00a0printf<\/span>(<\/span>“gpu\u00a0running\u00a0time:%ld\\n”<\/span>,<\/span>ed<\/span>–<\/span>st<\/span>);<\/span><\/li>\n
          15. \u00a0\u00a0\u00a0\u00a0cudaFree<\/span>(<\/span>d_C<\/span>);<\/span><\/li>\n
          16. \u00a0\u00a0\u00a0\u00a0<\/span>for<\/span>(<\/span>int<\/span>\u00a0i<\/span>=<\/span>0<\/span>;<\/span>i<\/span><<\/span>10<\/span>;<\/span>i<\/span>++)<\/span><\/li>\n
          17. \u00a0\u00a0\u00a0\u00a0<\/span>{<\/span><\/li>\n
          18. \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0printf<\/span>(<\/span>“%ld\u00a0“<\/span>,<\/span>C<\/span>[<\/span>i<\/span>]);<\/span><\/li>\n
          19. \u00a0\u00a0\u00a0\u00a0<\/span>}<\/span><\/li>\n
          20. \u00a0\u00a0\u00a0\u00a0<\/span>delete<\/span>[]<\/span>\u00a0C<\/span>;<\/span><\/li>\n
          21. \u00a0\u00a0\u00a0\u00a0printf<\/span>(<\/span>“\\n”<\/span>);<\/span><\/li>\n
          22. }<\/span><\/li>\n<\/ol>\n

            \u4e8e\u662f\u5c31\u5b8c\u6210\u4e86\uff0c\u8fd8\u6709\u51e0\u4e2a\u7ec6\u8282\u9700\u8981\u5904\u7406\uff0c\u56e0\u4e3aCUDA\u7565\u5fae\u7e41\u7410\u7684\u5185\u5b58\u7ba1\u7406\uff0c\u6240\u6709\u8f93\u5165\u503c\u4ec5\u4ec5\u63d0\u4f9b\u6307\u9488\u7684\u5f62\u5f0f\u7ed9\u51fa\uff0c\u6240\u4ee5\u6211\u4eec\u7528getCurrentTimeForDev()\u6765\u76f4\u63a5\u7ed9\u51fadevice\u6307\u9488\u7684\u65f6\u95f4<\/p>\n

              \n
            1. long<\/span>\u00a0getCurrentTime<\/span>()<\/span>\u00a0\u00a0<\/span><\/li>\n
            2. {<\/span>\u00a0\u00a0<\/span><\/li>\n
            3. \u00a0\u00a0\u00a0\u00a0<\/span>struct<\/span>\u00a0timeval\u00a0tv<\/span>;<\/span>\u00a0\u00a0<\/span><\/li>\n
            4. \u00a0\u00a0\u00a0\u00a0gettimeofday<\/span>(><\/span>tv<\/span>,<\/span>NULL<\/span>);<\/span>\u00a0\u00a0<\/span><\/li>\n
            5. \u00a0\u00a0\u00a0\u00a0<\/span>return<\/span>\u00a0tv<\/span>.<\/span>tv_sec\u00a0<\/span>*<\/span>\u00a0<\/span>1000000<\/span>\u00a0<\/span>+<\/span>\u00a0tv<\/span>.<\/span>tv_usec<\/span>;<\/span>\u00a0\u00a0<\/span><\/li>\n
            6. }<\/span><\/li>\n
            7. long<\/span>*<\/span>getCurrentTimeForDev<\/span>()<\/span><\/li>\n
            8. {<\/span>\u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0<\/span>*<\/span>time<\/span>;<\/span><\/li>\n
            9. \u00a0\u00a0\u00a0\u00a0cudaMalloc<\/span>(><\/span>time<\/span>,<\/span>sizeof<\/span>(<\/span>long<\/span>));<\/span><\/li>\n
            10. \u00a0\u00a0\u00a0\u00a0<\/span>long<\/span>\u00a0<\/span>*<\/span>timenow<\/span>=<\/span>new<\/span>\u00a0<\/span>long<\/span>;<\/span><\/li>\n
            11. \u00a0\u00a0\u00a0\u00a0<\/span>*<\/span>timenow<\/span>=<\/span>getCurrentTime<\/span>();<\/span><\/li>\n
            12. \u00a0\u00a0\u00a0\u00a0cudaMemcpy<\/span>(<\/span>time<\/span>,<\/span>timenow<\/span>,<\/span>sizeof<\/span>(<\/span>long<\/span>),<\/span>cudaMemcpyHostToDevice<\/span>);<\/span><\/li>\n
            13. \u00a0\u00a0\u00a0\u00a0<\/span>return<\/span>\u00a0time<\/span>;<\/span><\/li>\n
            14. }<\/span><\/li>\n<\/ol>\n

              \u7136\u540e\u8fd0\u884c\u4e0b\uff1a<\/p>\n

                \n
              1. CUDA_Learning$\u00a0<\/span>.\/<\/span>rand<\/span><\/li>\n
              2. gpu\u00a0running\u00a0time<\/span>:<\/span>490969<\/span><\/li>\n
              3. 3500780692<\/span>\u00a0<\/span>144000514<\/span>\u00a0<\/span>797145604<\/span>\u00a0<\/span>4199381721<\/span>\u00a0<\/span>2915119786<\/span>\u00a0<\/span>2557417372<\/span>\u00a0<\/span>4151631408<\/span>\u00a0<\/span>1974695633<\/span>\u00a0<\/span>2578972200<\/span>\u00a0<\/span>2865224429<\/span>\u00a0<\/span><\/li>\n<\/ol>\n

                \u968f\u673a\u6570\u751f\u6210\u8fd9\u4e8b\u60c5\u786e\u5b9e\u6162\u7684\u60ca\u4eba\u3002\u3002\u3002\u5c45\u7136\u8dd1\u4e86\u534a\u79d2\u3002<\/p>\n

                \u4e0b\u9762\u7684\u6837\u4f8b\u76f4\u63a5\u5207\u5230\u7528\u8fd9\u4e2a\u968f\u673a\u6570\u5e8f\u5217\u6765\u505a\u8499\u7279\u5361\u6d1b\u79ef\u5206\u4e86<\/a>\u3002\u8bf7\u6233\u76f8\u5e94\u94fe\u63a5\u3002<\/p>\n

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

                \u7531\u4e8e\u968f\u673a\u6570\u662f\u6574\u4e2a\u8ba1\u7b97\u7269\u7406\u7684\u57fa\u77f3\uff0c\u5b9e\u5728\u592a\u8fc7\u4e8e\u91cd\u8981\uff0c\u5e72\u8106\u7528\u4e00\u7247\u6587\u7ae0\u6765\u53d9\u8ff0CUDA\u6709\u5173\u4e8e\u968f\u673a\u6570\u4ea7\u751f\u5668\u7684\u79cd\u79cd\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[5,7],"tags":[],"_links":{"self":[{"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=\/wp\/v2\/posts\/188"}],"collection":[{"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=188"}],"version-history":[{"count":4,"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=\/wp\/v2\/posts\/188\/revisions"}],"predecessor-version":[{"id":199,"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=\/wp\/v2\/posts\/188\/revisions\/199"}],"wp:attachment":[{"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=188"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=188"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/blog.xuhao1.me\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=188"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}