MKLab ITI CUDA Versions Save

GPU-accelerated LIBSVM is a modification of the original LIBSVM that exploits the CUDA framework to significantly reduce processing time while producing identical results. The functionality and interface of LIBSVM remains the same. The modifications were done in the kernel computation, that is now performed using the GPU.

v1.2.2

10 years ago
  • Deleted unneeded library "cufft"

v1.2.1

10 years ago
  • Added "train_set" file. A large dataset to observe the speedups between CPU and GPU execution times.

v1.2

10 years ago

Using LibSVM 3.17 & CUDA SDK 5.5

  • Exchanged the header file “cublas.h” for “cublas_v2.h”.
  • Exchanged the type cublasStatus for cublasStatus_t.
  • Exchanged the function cublasAlloc() and cublasFree() for cudaMalloc() and cudaFree(), respectively.
  • Declared the cublasHandle_t CUBLAS library handle.
  • Initialized the handle using cublasCreate(). Also, released the handle once finished using cublasDestroy().
  • Added the handle as the first parameter to all the CUBLAS library function calls.
  • Changed the scalar parameters to be passed by reference, instead of by value (usually simply adding “&” symbol in C/C++ is enough, because the parameters are passed by reference on the host by default).
  • Changed the parameter characters 'N' or 'n' (non-transpose operation), to CUBLAS_OP_N.
  • Checked the function prototypes in the header files “cublas.h” and “cublas_v2.h” for code correctness.