{"id":1068,"date":"2014-10-04T15:44:00","date_gmt":"2014-10-04T07:44:00","guid":{"rendered":"http:\/\/www.brofive.org\/?p=1068"},"modified":"2018-04-29T22:38:55","modified_gmt":"2018-04-29T14:38:55","slug":"%e4%b8%8b%e4%b8%80%e4%bb%a3%e8%ae%a1%e7%ae%97%e8%ae%be%e5%a4%87cpu-gpu-fpga%ef%bc%9afpga%e8%a7%84%e6%a8%a1%e5%95%86%e7%94%a8%e8%bf%98%e9%9c%80%e6%97%b6%e6%97%a5","status":"publish","type":"post","link":"http:\/\/www.brofive.net\/?p=1068","title":{"rendered":"\u4e0b\u4e00\u4ee3\u8ba1\u7b97\u8bbe\u5907CPU\/GPU\/FPGA\uff1aFPGA\u89c4\u6a21\u5546\u7528\u8fd8\u9700\u65f6\u65e5"},"content":{"rendered":"

\u6700\u8fd1\u548c\u4e00\u4e2a\u5355\u4f4d\u6280\u672f\u4ea4\u6d41\uff0c\u5f97\u77e5\u6700\u8fd1FPGA\u5ffd\u7136\u975e\u5e38\u706b\uff0c\u636e\u8bf4\u53bb\u5e74\u7684\u51fa\u8d27\u91cf\u589e\u52a0\u4e00\u500d\uff0c\u5fae\u8f6f\u5728Bing\u5185\u90e8\u4e5f\u90e8\u7f72\u4e861000\u591a\u8282\u70b9\u7684\u6d4b\u8bd5\u73af\u5883\uff0c\u636e\u8bf4\u6d4b\u8bd5\u7ed3\u679c\u8868\u660e\u53ef\u4ee5\u51cf\u5c1150%\u7684\u8ba1\u7b97\u8282\u70b9\uff0c\u5355\u673a\u6210\u672c\u589e\u52a030%\u5de6\u53f3\uff0c\u56e0\u6b64\u8fd8\u662f\u7701\u94b1\u4e86\u3002\u540c\u65f6\u767e\u5ea6\u3001\u817e\u8baf\u3001\u8c37\u6b4c\u4e5f\u90fd\u6709\u5c1d\u8bd5FPGA\u7684\u8ba1\u5212\u3002\u636e\u8bf4\u5438\u5f15\u6295\u8d44\u65b9\u7684\u4e24\u4e2a\u91cd\u8981\u95ee\u9898\u662f\uff1a1.CPU\u6536\u5230\u91cf\u5b50\u6548\u5e94\u7684\u5f71\u54cd\uff0c\u96c6\u6210\u5ea6\u96be\u4ee5\u63d0\u9ad82.GPU\u80fd\u8017\u592a\u9ad8\u3002<\/font> <\/p>\n

\u5728\u5b66\u6821\u671f\u95f4\u66fe\u7ecf\u505a\u8fc78\u4f4d\u673a\uff0c\u5f53\u65f6\u627f\u62c5\u89d2\u8272\u662f\u8bbe\u8ba1\u8fd0\u7b97\u903b\u8f91\uff0c\u57fa\u672c\u4e0a\u5c31\u662f\u95e8\u7535\u8def\u8bbe\u8ba1\uff0c\u4e3a\u4e86\u653e\u5230\u91d1\u949fGAL\u91cc\u9762\uff0c\u8d39\u4e86\u8001\u9f3b\u5b50\u52b2\uff0c\u540e\u6765\u8fd8\u70e7\u574f\u4e86\u4e00\u6279GAL\u3002GAL\u4e4b\u540e\u662fCPLD\uff0c\u652f\u6301\u66f4\u591a\u7684\u53ef\u7f16\u7a0b\u95e8\u7535\u8def\u3002<\/font>\u540c\u65f6\u53e6\u5916\u4e00\u4e2a\u6280\u672f\u5206\u652f\u5c31\u662fFPGA\uff0c\u8fd9\u4e2a\u73a9\u610f\u5de5\u4f5c\u539f\u7406\u7c7b\u4f3c\uff0c\u548cCPLD\u7684\u4e3b\u8981\u533a\u522b\u662fFPGA\u57fa\u4e8e\u67e5\u627e\u8868\uff08LUT\uff09\u5b9e\u73b0\u4e86\u5927\u89c4\u6a21\u72b6\u6001\u673a\u7535\u8def\uff0c\u800cCPLD\u7531\u5927\u91cf\u95e8\u7535\u8def\u6784\u6210\u3002\u56e0\u6b64CPLD\u9002\u7528\u4e8e\u7ec4\u5408\u903b\u8f91\u5e94\u7528\uff0c\u800cFPGA\u66f4\u9002\u5408\u4e8e\u7c7b\u4f3cCPU\u7684\u5de5\u4f5c<\/font><\/a>\u3002<\/font> <\/p>\n

\u5b66\u672f\u754c\u57282010\u5e74\u524d\u540e\u53d1\u8868\u4e86\u4e0d\u5c11FPGA\u7684\u7814\u7a76\u8bba\u6587\uff0c\u603b\u4f53\u7ed3\u8bba\u662fGPU\u7684\u6027\u80fd\u5360\u4f18\uff0c\u5f00\u53d1\u5bb9\u6613\uff0cFPGA\u5f00\u53d1\u590d\u6742\uff0c\u5728\u90e8\u5206\u9886\u57df\u53ef\u80fd\u5360\u4f18\uff0c\u4f46\u662f\u9700\u8981\u793e\u533a\u6765\u5927\u529b\u63a8\u52a8\u3002\u5de5\u4e1a\u754c\u8fd1\u671f\u5f00\u59cb\u5229\u7528FPGA\u505a\u4e00\u4e9b\u4ea7\u54c1\u5316\u5de5\u4f5c\uff0c\u4f46\u662f\u4e3b\u8981\u8fd8\u662f\u96c6\u4e2d\u5728\u7f51\u7edc\u8bbe\u5907\uff0c\u5728\u8ba1\u7b97\u8bbe\u5907\u65b9\u9762\uff0c\u6709\u4e9b\u5382\u5546\u5728\u7814\u7a76FPGA\u5728\u4ea7\u54c1\u548c\u670d\u52a1\u4e2d\u7684\u5e94\u7528\uff0c\u4f46\u662f\u5e76\u672a\u6709\u5927\u89c4\u6a21\u5546\u7528\u6848\u4f8b\u3002\u4e2a\u4eba\u9884\u8ba1\u5728\u56fa\u5b9a\u8ba1\u7b97\u6a21\u5f0f\u7684\u4ea7\u54c1\u65b9\u9762\u4f1a\u6709\u8f83\u591a\u4e00\u4f53\u5316\u8bbe\u5907\u4f7f\u7528FPGA\uff0c\u4f46\u662f\u5728\u670d\u52a1\u65b9\u9762\uff0cFPGA\u76ee\u524d\u8fd8\u4e0d\u6210\u719f\u3002<\/font> <\/p>\n

\u5b66\u672f\u754c\uff1a\u6709\u4e00\u4e9b\u7814\u7a76\u548c\u57fa\u51c6\u6d4b\u8bd5<\/font> <\/p>\n

<\/font> <\/p>\n

\u4e00\u30012013\u5e74\uff0cEgil Fykse\u7855\u58eb\u8bba\u6587\u6d4b\u8bd5\u6bd4\u8f83DSP\u3001GPU\u548cFPGA\uff08\u6ce8\uff1a\u6d4b\u8bd5\u7684\u5185\u5bb9\u4e0d\u662f\u5f88\u5145\u5206\uff09<\/font><\/strong><\/p>\n

\"1\"<\/a> <\/font><\/strong><\/p>\n

\u7efc\u5408\u6bd4\u8f83\u4e86\u6027\u4ef7\u6bd4\uff0c\u603b\u4f53\u6765\u8bf4\u4ef7\u683c\u4e0d\u597d\u6bd4\u8f83\uff0c\u540c\u6837\u4f5c\u4e1aFPGA\u80fd\u8017\u5927\u5927\u4f4e\u4e8eGPU\uff0c\u5f00\u53d1\u80fd\u529b\u65b9\u9762GPU\u5927\u5927\u4f18\u4e8eFPGA\uff0c\u6027\u80fd\u65b9\u9762GPU\u5360\u4f18<\/font> <\/p>\n

\u4e8c\u30012013\u5e74Mathworks\u7684Christopher Cullinan\u7b49\u4eba\u5bf9CPU\u3001GPU\u548cFPGA\u8fdb\u884c\u6027\u80fd\u6d4b\u8bd5<\/font><\/strong> <\/p>\n

\"2\"<\/a> <\/font><\/strong> <\/p>\n

FFT\u7684\u6027\u80fd\u6d4b\u8bd5\uff0cGPU\u67091000\u500d\u5de6\u53f3\u7684\u6027\u80fd\u4f18\u52bf\uff0c\u4f46\u662f\u5982\u679c\u8003\u8651\u6570\u636e\u4f20\u8f93\u65f6\u95f4\uff0c\u5219GPU\u53d8\u621069ms\uff0c\u53d8\u6210\u4e86\u6700\u6162\u7684\uff0c\u800cFPGA\u53d8\u6210\u6700\u5feb\u7684\u3002\u5176\u4ed6\u7684\u6d4b\u8bd5\u4e5f\u8868\u660e\uff0cGPU\u7684\u6027\u80fd\u4e3b\u8981\u5728\u6570\u636e\u4f20\u8f93\u65b9\u9762<\/font> <\/p>\n

\"3\"<\/a> <\/font><\/strong> <\/p>\n

\u5176\u4ed6\u4e09\u79cd\u7b97\u6cd5\u7684\u6d4b\u8bd5\uff0cGPU\u5bf9\u6bd4CPU\u6709\u4e00\u4e9b\u4f18\u52bf\uff0c\u6700\u59273~100\u500d\u5de6\u53f3<\/font> <\/p>\n

<\/font><\/strong>  <\/p>\n

\u4e09\u30012010\u5e74David H. Jones\u7b49\u4eba\u53d1\u8868\u8bba\u6587\u6bd4\u8f83GPU\u548cFPGA\u7684\u6027\u80fd\uff0cHPC\u9886\u57df<\/font><\/strong> <\/p>\n

\"4\"<\/a> <\/font><\/strong> <\/p>\n

\u4f2a\u968f\u673a\u6570\u751f\u6210\uff0cGPU\u548cFPGA\u76f8\u5f53\uff08\u552f\u4e00\u4e00\u4e2a\u76f8\u5f53\u7684\u4f8b\u5b50\uff09\uff0c64\u4f4d\u77e9\u9635\u76f8\u4e58\uff0cGPU\u4f18\u4e8eFPGA<\/font> <\/p>\n

\"5\"<\/a> <\/font><\/strong> <\/p>\n

N-bady\u4eff\u771f\uff0cGPU\u663e\u8457\u4f18\u4e8eFPGA\u3002\u6700\u7ec8\u7684\u7ed3\u8bba\u662f\uff0c\u5982\u679c\u4e0d\u80fd\u901a\u8fc7\u5f00\u6e90\u4f18\u5316\u7b97\u6cd5\uff0cFPGA\u53ea\u80fd\u7528\u4e8e\u7279\u5b9a\u7684\u8ba1\u7b97\u9886\u57df\u3002\u4e00\u4e2a\u8bc1\u636e\u662fCray\u7cfb\u7edf\u5df2\u7ecf\u653e\u5f03FPGA\u800c\u662f\u7528GPU<\/font> <\/p>\n

\u5de5\u4e1a\u754c\uff1a\u5728FPGA\u65b9\u9762\u7684\u8fdb\u5c55\u5e76\u6ca1\u6709\u4ec0\u4e48\u5546\u7528\u6848\u4f8b\uff0c\u5fae\u8f6fBing\u6848\u4f8b\u662f\u4e00\u4e2a\u6700\u5927\u7684\u8bd5\u70b9\u7cfb\u7edf\uff0c\u5176\u4ed6\u5c31\u5f88\u4e0d\u6e05\u695a\u4e86<\/font> <\/p>\n

\u4e00\u3001\u5fae\u8f6fBing\u6848\u4f8b<\/font><\/strong> <\/p>\n

\"6\"<\/a> <\/font><\/p>\n

<\/font><\/strong> <\/p>\n

Putnam\u4ecb\u7ecd\u4e86\u5728Bing\u505a\u7684\u539f\u578b\u9879\u76ee\u4e2d\uff08\u79bb\u7ebf\u8ba1\u7b97\u641c\u7d22\u8bcd\u6392\u540d\uff0c\u5e76\u884c\u7cfb\u7edf\uff09\uff0c\u4f7f\u75281632\u4e2a\u670d\u52a1\u5668\uff0c\u6bcf\u53f0\u670d\u52a1\u5668\u4f7f\u7528\u4e00\u4e2aPCIe\u7684FPGA\u5361\uff0c\u589e\u52a0\u5355\u53f0\u670d\u52a1\u5668\u6210\u672c\u4e0d\u523030%\uff0c\u4f46\u662f\u63d0\u5347\u541e\u5410\u7387100%\u3002\u5373\u4f20\u8bf4\u53ef\u4ee5\u8282\u7ea650%\u670d\u52a1\u5668\u7684\u7531\u6765\u3002\u4e3b\u8981\u95ee\u9898\u662f\u96be\u4ee5\u5f00\u53d1\u5e94\u7528\u7a0b\u5e8f\uff0c\u9700\u8981\u9ad8\u7ea7\u8bed\u8a00\u7684\u652f\u6301\u3002\u540c\u65f6\u5f00\u53d1\u6d41\u7a0b\u4e5f\u975e\u5e38\u6162\u3002\u8c8c\u4f3c\u7531\u4e8e\u7cfb\u7edf\u57fa\u4e8eWindows\uff0c\u800c\u8bbe\u5907\u7684Windows\u9a71\u52a8\u4e5f\u662f\u4e2a\u95ee\u9898\u3002<\/font><\/p>\n

\"7\"<\/a> <\/font><\/strong><\/p>\n

\u4e8c\u3001\u767e\u5ea6\u6848\u4f8b<\/font><\/strong> <\/p>\n

\u5f00\u53d1\u4e86FPGA\u7684\u677f\u5b50\uff0c\u636e\u8bf4\u7528\u4e8e\u81f3\u5c1110\u4e2a\u5e95\u5c42\u4efb\u52a1\u7684\u4f18\u5316\uff0c\u636e\u8bf4\u5728\u6027\u80fd\/\u529f\u8017\/\u6210\u672c\u65b9\u9762\u7531\u4e8eGPU\u548cCPU\u3002\u767e\u5ea6\u7684FPGA\u67b6\u6784\u5e08\uff08\u6b27\u9633\u5065\uff09\u539f\u6765\u5728\u5fae\u8f6f\u7814\u7a76\u9662\u505aFPGA\u65b9\u9762\u7684\u7814\u7a76\u5de5\u4f5c\uff0cPutnam\u7684\u56e2\u961f\u3002<\/font> <\/p>\n

\"8\"<\/a> <\/font><\/p>\n

<\/font><\/strong> <\/p>\n

\u4e09\u3001Google\u6848\u4f8b<\/font> <\/p>\n

Jonathan Ross\u5728FPGA 2014\u5927\u4f1a\u4e0a\u4e3b\u8981\u8bb2\u4e86Google\u5bf9FPGA\u7684\u4e00\u4e9b\u6001\u5ea6\uff0c\u4e00\u65b9\u9762\u5b58\u5728\u8ba1\u7b97\u9700\u6c42\uff0c\u4f46\u662fFPGA\u76ee\u524d\u5b58\u5728\u5f88\u591a\u95ee\u9898\uff0c\u5982\u4e0d\u5f00\u6e90\u3001\u4e0d\u6210\u719f\u3001\u8bb8\u53ef\u8d39\u3001\u652f\u6301\u5dee\u7b49\u95ee\u9898\u3002\u770b\u6765Google\u5e76\u672a\u5f00\u59cb\u4ec0\u4e48\u5546\u7528\u3002<\/font> <\/p>\n

\"9\"<\/a> <\/font><\/p>\n

<\/font><\/strong> <\/p>\n

\u56db\u3001Oracle\u548cIBM\u6848\u4f8b<\/font> <\/p>\n

Oracle\u6253\u7b97\u5229\u7528FPGA\u4f18\u5316\u6570\u636e\u5e93\uff0c\u76ee\u524d\u505c\u7559\u5728Lab\u72b6\u6001\u3002IBM\u8fd8\u5728\u5ffd\u60a0<\/font> <\/p>\n

\u76f8\u5173\u4fe1\u606f\uff1a<\/font> <\/p>\n

    \n
  1. Programmable logic device<\/font><\/a>\n
  2. Performance Comparison of GPU, DSP and FPGA implementations of image processing and computer vision  algorithms in embedded systems<\/font><\/a>\n
  3. Computing Performance Benchmarks among CPU, GPU, and FPGA<\/font><\/a>\n
  4. GPU versus FPGA for high productivity computing<\/font><\/a>\n
  5. Lattice\u2019s iCE40 FPGA Shipments Reach 200 Million Unit Milestone<\/font><\/a>\n
  6. Xilinx Begins Shipping TSMC 20nm FPGAs<\/font><\/a><\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"

    \u6700\u8fd1\u548c\u4e00\u4e2a\u5355\u4f4d\u6280\u672f\u4ea4\u6d41\uff0c\u5f97\u77e5\u6700\u8fd1FPGA\u5ffd\u7136\u975e\u5e38\u706b\uff0c\u636e\u8bf4\u53bb\u5e74\u7684\u51fa\u8d27\u91cf\u589e\u52a0\u4e00\u500d\uff0c\u5fae\u8f6f\u5728Bing\u5185\u90e8\u4e5f\u90e8\u7f72\u4e86100…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[39,128],"tags":[191,192],"views":2925,"_links":{"self":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts\/1068"}],"collection":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1068"}],"version-history":[{"count":1,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts\/1068\/revisions"}],"predecessor-version":[{"id":1069,"href":"http:\/\/www.brofive.net\/index.php?rest_route=\/wp\/v2\/posts\/1068\/revisions\/1069"}],"wp:attachment":[{"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1068"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1068"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.brofive.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1068"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}