<?xml version="1.0" encoding="UTF-8"?> 
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
<entry>
  <name>p2pBandwidthLatencyTest</name>
  <cuda_api_list>
    <toolkit>cudaSetDevice</toolkit>
    <toolkit>cudaEventDestroy</toolkit>
    <toolkit>cudaOccupancyMaxPotentialBlockSize</toolkit>
    <toolkit>cudaCheckError</toolkit>
    <toolkit>cudaFreeHost</toolkit>
    <toolkit>cudaGetDeviceCount</toolkit>
    <toolkit>cudaDeviceCanAccessPeer</toolkit>
    <toolkit>cudaStreamCreateWithFlags</toolkit>
    <toolkit>cudaStreamDestroy</toolkit>
    <toolkit>cudaGetLastError</toolkit>
    <toolkit>cudaMemset</toolkit>
    <toolkit>cudaStreamWaitEvent</toolkit>
    <toolkit>cudaEventElapsedTime</toolkit>
    <toolkit>cudaEventCreate</toolkit>
    <toolkit>cudaHostAlloc</toolkit>
    <toolkit>cudaFree</toolkit>
    <toolkit>cudaGetErrorString</toolkit>
    <toolkit>cudaMemcpyPeerAsync</toolkit>
    <toolkit>cudaDeviceDisablePeerAccess</toolkit>
    <toolkit>cudaEventRecord</toolkit>
    <toolkit>cudaStreamSynchronize</toolkit>
    <toolkit>cudaDeviceEnablePeerAccess</toolkit>
    <toolkit>cudaMalloc</toolkit>
    <toolkit>cudaGetDeviceProperties</toolkit>
  </cuda_api_list>
  <description><![CDATA[This application demonstrates the CUDA Peer-To-Peer (P2P) data transfers between pairs of GPUs and computes latency and bandwidth.  Tests on GPU pairs using P2P and without P2P are tested.]]></description>
  <devicecompilation>whole</devicecompilation>
  <includepaths>
    <path>./</path>
    <path>../</path>
    <path>../../../Common</path>
  </includepaths>
  <keyconcepts>
    <concept level="basic">Performance Strategies</concept>
    <concept level="basic">Asynchronous Data Transfers</concept>
    <concept level="basic">Unified Virtual Address Space</concept>
    <concept level="basic">Peer to Peer Data Transfers</concept>
    <concept level="basic">Multi-GPU</concept>
  </keyconcepts>
  <keywords>
    <keyword>CUDA</keyword>
    <keyword>Performance</keyword>
    <keyword>multi-GPU support</keyword>
    <keyword>peer to peer</keyword>
  </keywords>
  <libraries>
  </libraries>
  <librarypaths>
  </librarypaths>
  <nsight_eclipse>true</nsight_eclipse>
  <primary_file>p2pBandwidthLatencyTest.cu</primary_file>
  <scopes>
    <scope>1:CUDA Basic Topics</scope>
    <scope>1:Performance Strategies</scope>
  </scopes>
  <sm-arch>sm50</sm-arch>
  <sm-arch>sm52</sm-arch>
  <sm-arch>sm53</sm-arch>
  <sm-arch>sm60</sm-arch>
  <sm-arch>sm61</sm-arch>
  <sm-arch>sm70</sm-arch>
  <sm-arch>sm72</sm-arch>
  <sm-arch>sm75</sm-arch>
  <sm-arch>sm80</sm-arch>
  <sm-arch>sm86</sm-arch>
  <sm-arch>sm87</sm-arch>
  <sm-arch>sm89</sm-arch>
  <sm-arch>sm90</sm-arch>
  <supported_envs>
    <env>
      <arch>x86_64</arch>
      <platform>linux</platform>
    </env>
    <env>
      <platform>windows7</platform>
    </env>
    <env>
      <arch>x86_64</arch>
      <platform>macosx</platform>
    </env>
    <env>
      <arch>arm</arch>
    </env>
    <env>
      <arch>sbsa</arch>
    </env>
    <env>
      <arch>ppc64le</arch>
      <platform>linux</platform>
    </env>
  </supported_envs>
  <supported_sm_architectures>
    <include>all</include>
  </supported_sm_architectures>
  <title>Peer-to-Peer Bandwidth Latency Test with Multi-GPUs</title>
  <type>exe</type>
</entry>