The llvm-ptx-samples from jholewinski

32/64 bit OS/GPU issues with cuParamSetv

I admit I'm pretty new to the CUDA&PTX arena, so maybe I've done something stupid, but the examples refused to work for me. After much debugging and Googling, I found http://visionexperts.blogspot.com/2010/07/cuda-parameter-alignment.html

I'm running 64-bit Linux and a Quadro 4000, with a pretty recent top-of-tree LLVM build.

Sure enough, the examples started working for me after I changed the way array pointers were submitted. Does this seem reasonable? Is there a more elegant solution?

I also found http://developer.download.nvidia.com/compute/cuda/2_3/toolkit/docs/NVIDIA_CUDA_BestPracticesGuide_2.3.pdf which, in Listing 1.2, has an even slightly-more-complicated approach.

FROM:

  // Bind kernel paramters
  status = cuParamSetv(function, 0, &deviceA, sizeof(CUdeviceptr));
  checkSuccess(status, "cuParamSetv");
  status = cuParamSetv(function, sizeof(CUdeviceptr), &deviceB, sizeof(CUdeviceptr));
  checkSuccess(status, "cuParamSetv");
  status = cuParamSetv(function, 2*sizeof(CUdeviceptr), &deviceC, sizeof(CUdeviceptr));
  checkSuccess(status, "cuParamSetv");
  status = cuParamSeti(function, 3*sizeof(CUdeviceptr), problemSize);
  checkSuccess(status, "cuParamSeti");

  status = cuParamSetSize(function, 3*sizeof(CUdeviceptr) + sizeof(int));
  checkSuccess(status, "cuParamSetSize");

TO:

  // Bind kernel paramters
  void *ptr;
  ptr = (void*)(deviceA);
  status = cuParamSetv(function, 0, &ptr, sizeof(void*));
  checkSuccess(status, "cuParamSetv");
  ptr = (void*)(deviceB);
  status = cuParamSetv(function, sizeof(ptr), &ptr, sizeof(ptr));
  checkSuccess(status, "cuParamSetv");
  ptr = (void*)(deviceC);
  status = cuParamSetv(function, 2*sizeof(ptr), &ptr, sizeof(ptr));
  checkSuccess(status, "cuParamSetv");
  status = cuParamSeti(function, 3*sizeof(ptr), problemSize);
  checkSuccess(status, "cuParamSeti");

  status = cuParamSetSize(function, 3*sizeof(ptr) + sizeof(int));
  checkSuccess(status, "cuParamSetSize");

Bring Back Driver API Samples

Re-integrate the Driver API samples into the build.

[blur-2d] Actually output something cool

Manipulate real image data in the blur-2d sample.

New Build System

Create a new build system based on make, or maybe ninja, to prevent the dependency on CMake.

ran into trouble when running the samples

Hello,

I'm running into trouble trying to create a kernel object from ptx:

Assertion failed: (result == CL_SUCCESS && "Failed to extract kernel"), function initialize, file /Users/cos/Projects/cs526/test2/llvm-ptx-samples/opencl/matmul/matmul.cpp, line 75.
Abort trap: 6

So it works for the lc but not for the ptx...

I'm using LLVM+Clang 3.2 on OS X.

Any ideas?

Thanks!

jholewinski / llvm-ptx-samples Goto Github PK

llvm-ptx-samples's People

Contributors

Stargazers

Watchers

Forkers

llvm-ptx-samples's Issues

32/64 bit OS/GPU issues with cuParamSetv

Bring Back Driver API Samples

[blur-2d] Actually output something cool

New Build System

ran into trouble when running the samples

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent