cudaMemcpyAsync 的奇怪行为: 1. cudaMemcpyKind 没有区别。 2.复制失败,但无声

Odd behavior of cudaMemcpyAsync: 1. cudaMemcpyKind makes no difference. 2. Copy fails, but silently

我正在熟悉配备 Pascal P100 GPUs+Nvlink 的新集群。我写了一个乒乓程序来测试 gpu<->gpu 和 gpu<->cpu 带宽和点对点访问。 (我知道 cuda 样本包含这样的程序,但我想自己做,以便更好地理解。)Nvlink 带宽似乎合理(~35 GB/s 双向,理论最大值为 40)。然而,在调试 ping-pong 时,我发现了一些奇怪的行为。

首先,无论我指定什么 cudaMemcpyKind,cudaMemcpyAsync 都会成功,例如,如果 cudaMemcpyAsync 正在将内存从主机复制到设备,即使我将 cudaMemcpyDeviceToHost 作为种类传递,它也会成功。

其次,当主机内存未被页面锁定时,cudaMemcpyAsync 执行以下操作:

这种行为是否符合预期?我已经包含了一个在我的系统上演示它的最小工作示例代码(该示例不是乒乓应用程序,它所做的只是使用各种参数测试 cudaMemcpyAsync)。

P100 启用了 UVA,因此我认为 cudaMemcpyAsync 只是推断 src 和 dst 指针的位置并忽略 cudaMemcpyKind 参数是合理的。但是,我不确定为什么 cudaMemcpyAsync 无法为非页锁定主机内存抛出错误。我的印象是这是一个严格的禁忌。

#include <stdio.h>
#include <cuda_runtime.h>
#include <stdlib.h>

#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true)
{
   if (code != cudaSuccess)
   {
      fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line);
      if (abort) exit(code);
   }
}

__global__ void checkDataDevice( int* current, int* next, int expected_current_val, int n )
{
  int tid = threadIdx.x + blockIdx.x*blockDim.x;
  for( int i = tid; i < n; i += blockDim.x*gridDim.x )
  {
    if( current[i] != expected_current_val )
      printf( "Error on device:  expected = %d, current[%d] = %d\n"
          , expected_current_val
          , i
          , current[i] );
    // Increment the data so the next copy is properly tested
    next[i] = current[i] + 1;
  }
}

void checkDataHost( int* current, int* next, int expected_current_val, int n )
{
  for( int i = 0; i < n; i++ )
  {
    if( current[i] != expected_current_val )
      printf( "Error on host:  expected = %d, current[%d] = %d\n"
          , expected_current_val
          , i
          , current[i] );
    // Increment the data so the next copy is properly tested
    next[i] = current[i] + 1;
  }
}

int main( int argc, char** argv )
{
  bool pagelocked = true;
  // invoking the executable with any additional argument(s) will turn off page locked memory, i.e.,
  // Run with pagelocked memory:  ./a.out
  // Run with ordinary malloc'd memory: ./a.out jkfdlsja
  if( argc > 1 )
    pagelocked = false;

  int copybytes = 1e8; // Ok to use int instead of size_t for 1e8.

  cudaStream_t* stream = (cudaStream_t*)malloc( sizeof(cudaStream_t) );
  cudaStreamCreate( stream );

  int* srcHost;
  int* dstHost;
  int* srcDevice;
  int* dstDevice;

  cudaMalloc( (void**)&srcDevice, copybytes );
  cudaMalloc( (void**)&dstDevice, copybytes );
  if( pagelocked )
  {
    printf( "Using page locked memory\n" );
    cudaMallocHost( (void**)&srcHost, copybytes );
    cudaMallocHost( (void**)&dstHost, copybytes );
  }
  else
  {
    printf( "Using non page locked memory\n" );
    srcHost = (int*)malloc( copybytes );
    dstHost = (int*)malloc( copybytes );
  }

  for( int i = 0; i < copybytes/sizeof(int); i++ )
    srcHost[i] = 1;

  cudaMemcpyKind kinds[4];
  kinds[0] = cudaMemcpyHostToDevice;
  kinds[1] = cudaMemcpyDeviceToHost;
  kinds[2] = cudaMemcpyHostToHost;
  kinds[3] = cudaMemcpyDeviceToDevice;

  // Test cudaMemcpyAsync in both directions,
  // iterating through all "cudaMemcpyKinds" to verify
  // that they don't matter.
  int expected_current_val = 1;
  for( int kind = 0; kind<4; kind++ )
  {
    // Host to device copy 
    cudaMemcpyAsync( dstDevice
        , srcHost
        , copybytes
        , kinds[kind]
        , *stream );
    gpuErrchk( cudaDeviceSynchronize() );

    checkDataDevice<<<56*8,256>>>( dstDevice
        , srcDevice
        , expected_current_val
        , copybytes/sizeof(int) );
    expected_current_val++;

    // Device to host copy
    cudaMemcpyAsync( dstHost
        , srcDevice
        , copybytes
        , kinds[kind]
        , *stream );
    gpuErrchk( cudaDeviceSynchronize() );

    checkDataHost( dstHost
        , srcHost
        , expected_current_val
        , copybytes/sizeof(int) );
    expected_current_val++;
  }

  free( stream );

  cudaFree( srcDevice );
  cudaFree( dstDevice );
  if( pagelocked )
  {
    cudaFreeHost( srcHost );
    cudaFreeHost( dstHost );
  }
  else
  {
    free( srcHost );
    free( dstHost );
  }

  return 0;
}

当 CUDA 代码出现问题时,我 强烈 建议使用 rigorous(== 每次调用 return 代码已选中)proper CUDA error checking.

您的错误检查存在缺陷,这些缺陷导致您有些困惑。

首先,在页面锁定的情况下,给定的(映射的)指针在主机和设备上都是accessible/valid。因此,每个可能的方向枚举(H2D、D2H、D2D、H2H)都是合法有效的。因此,不会 return 编辑任何错误并且 复制操作成功。

在非页面锁定的情况下,上述情况并非如此,因此一般来说,指示的传输方向最好与隐含的传输方向相匹配,如指针所检查的那样。如果没有,cudaMemcpyAsync 将 return 错误代码 (cudaErrorInvalidValue == 11)。在您的情况下,您忽略了此错误结果。如果你有足够的耐心,你可以自己证明这一点(如果你只是标记第一个错误会更好,而不是打印出 10M+ 元素中的每个不匹配),通过 运行 你的代码 cuda-memcheck (当您在使用 CUDA 代码时遇到问题时可以做的另一件好事)或者进行适当、严格的错误检查。

cudaMemcpyAsync 操作指示失败时,操作未成功完成,因此未复制数据,并且您的数据检查指示不匹配。希望现在这并不奇怪,因为预期的复制操作实际上并没有发生(也没有失败 "silently")。

也许您很困惑,认为捕获任何类型的异步操作错误的方法是执行 cudaDeviceSynchronize 然后检查错误。

这对 cudaMemcpyAsync 不正确。在调用 cudaMemcpyAsync 操作时可以检测到的错误将由调用本身立即 returned,并且 不会 returned作为后续 CUDA 调用的结果(很明显),因为这种类型的错误是非粘性的。

故事的寓意:

  1. 进行适当的 CUDA 错误检查。严格。
  2. 运行 您的代码 cuda-memcheck.

这是一个完整的示例,对您的代码进行了微不足道的修改以在失败案例中生成输出 "sane",证明在失败案例中指示了错误:

$ cat t153.cu
#include <stdio.h>
#include <stdlib.h>

#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true)
{
   if (code != cudaSuccess)
   {
      fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line);
      if (abort) exit(code);
   }
}

__global__ void checkDataDevice( int* current, int* next, int expected_current_val, int n )
{
  int tid = threadIdx.x + blockIdx.x*blockDim.x;
  for( int i = tid; i < n; i += blockDim.x*gridDim.x )
  {
    if( current[i] != expected_current_val )
      printf( "Error on device:  expected = %d, current[%d] = %d\n"
          , expected_current_val
          , i
          , current[i] );
    // Increment the data so the next copy is properly tested
    next[i] = current[i] + 1;
  }
}

void checkDataHost( int* current, int* next, int expected_current_val, int n )
{
  for( int i = 0; i < n; i++ )
  {
    if( current[i] != expected_current_val ){
      printf( "Error on host:  expected = %d, current[%d] = %d\n"
          , expected_current_val
          , i
          , current[i] );
      exit(0);}
    // Increment the data so the next copy is properly tested
    next[i] = current[i] + 1;
  }
}

int main( int argc, char** argv )
{
  bool pagelocked = true;
  // invoking the executable with any additional argument(s) will turn off page locked memory, i.e.,
  // Run with pagelocked memory:  ./a.out
  // Run with ordinary malloc'd memory: ./a.out jkfdlsja
  if( argc > 1 )
    pagelocked = false;

  int copybytes = 1e8; // Ok to use int instead of size_t for 1e8.

  cudaStream_t* stream = (cudaStream_t*)malloc( sizeof(cudaStream_t) );
  cudaStreamCreate( stream );

  int* srcHost;
  int* dstHost;
  int* srcDevice;
  int* dstDevice;

  cudaMalloc( (void**)&srcDevice, copybytes );
  cudaMalloc( (void**)&dstDevice, copybytes );
  if( pagelocked )
  {
    printf( "Using page locked memory\n" );
    cudaMallocHost( (void**)&srcHost, copybytes );
    cudaMallocHost( (void**)&dstHost, copybytes );
  }
  else
  {
    printf( "Using non page locked memory\n" );
    srcHost = (int*)malloc( copybytes );
    dstHost = (int*)malloc( copybytes );
  }

  for( int i = 0; i < copybytes/sizeof(int); i++ )
    srcHost[i] = 1;

  cudaMemcpyKind kinds[4];
  kinds[0] = cudaMemcpyHostToDevice;
  kinds[1] = cudaMemcpyDeviceToHost;
  kinds[2] = cudaMemcpyHostToHost;
  kinds[3] = cudaMemcpyDeviceToDevice;

  // Test cudaMemcpyAsync in both directions,
  // iterating through all "cudaMemcpyKinds" to verify
  // that they don't matter.
  int expected_current_val = 1;
  for( int kind = 0; kind<4; kind++ )
  {
    // Host to device copy
    cudaMemcpyAsync( dstDevice
        , srcHost
        , copybytes
        , kinds[kind]
        , *stream );
    gpuErrchk( cudaDeviceSynchronize() );

    checkDataDevice<<<56*8,256>>>( dstDevice
        , srcDevice
        , expected_current_val
        , copybytes/sizeof(int) );
    expected_current_val++;

    // Device to host copy
    cudaMemcpyAsync( dstHost
        , srcDevice
        , copybytes
        , kinds[kind]
        , *stream );
    gpuErrchk( cudaDeviceSynchronize() );

    checkDataHost( dstHost
        , srcHost
        , expected_current_val
        , copybytes/sizeof(int) );
    expected_current_val++;
  }

  free( stream );

  cudaFree( srcDevice );
  cudaFree( dstDevice );
  if( pagelocked )
  {
    cudaFreeHost( srcHost );
    cudaFreeHost( dstHost );
  }
  else
  {
    free( srcHost );
    free( dstHost );
  }

  return 0;
}
$ nvcc -arch=sm_61 -o t153 t153.cu
$ cuda-memcheck ./t153 a
========= CUDA-MEMCHECK
Using non page locked memory
========= Program hit cudaErrorInvalidValue (error 11) due to "invalid argument" on CUDA API call to cudaMemcpyAsync.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:/usr/lib/x86_64-linux-gnu/libcuda.so.1 [0x2ef423]
=========     Host Frame:./t153 [0x489a3]
=========     Host Frame:./t153 [0x2e11]
=========     Host Frame:/lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main + 0xf5) [0x21ec5]
=========     Host Frame:./t153 [0x2a49]
=========
Error on host:  expected = 2, current[0] = 0
========= ERROR SUMMARY: 1 error
$