ããã¾ã§ããã¼ã¿ã並åã«å¦çãããã¨ã§CPUã«æ¯ã¹ã¦åªããæ§è½ãGPUã示ããã¨ãç´¹ä»ãã¦ããã GPUã«ã¯ä»ã®ä¸¦åå¦çããã£ã¦ãããã¯CPUã§ã®ã¿ã¹ã¯ä¸¦åã«ä¼¼ããã®ã GPUã§ã®ã¿ã¹ã¯å¦çã¯ãCPUã§ã®ããã«æ¯ã¹ã¦æéãããã/èªåã§é½ã«è¨è¿°ãã¦ããå¿ è¦ããã ãã®ç« ã§ã¯ãstreamã®åºç¤ã¨ããã«ããã¿ã¹ã¯ä¸¦åã«ã¤ãã¦å¦ã¶ã 10.1 10ç« ã®ç®ç ãã¼ã¸ããã¯ã¡ã¢ãªã«ã¤ãã¦å¦ã¶ CUDA streamã«ã¤ãã¦å¦ã¶ CUDA streamãã¤ãã£ã¦ãã¢ããªãé«éåããæ¹æ³ 10.2 ãã¼ã¸ããã¯ãããããã¹ãã¡ã¢ãª ãã¹ãã«ã¡ã¢ãªã確ä¿ããã«ã¯ãmallocã¨cudaHostAllocããããéãã¯ï¼ mallocã¯æ¨æºãªãpageable host memoryãå²ãå½ã¦ã cudaHostAllocã¯ãã¼ã¸ããã¯ã¡ã¢ãªï¼pinned memoryã¨ãï¼ãå²ãå½ã¦ã ãã¼ã¸ããã¯ã¡ã¢
{{#tags}}- {{label}}
{{/tags}}