MAGMA Forum

Code: Select all

#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>
#include <cuda_runtime.h>
#include <cuda_runtime_api.h>
#include <cublas.h>
#include <cublas_v2.h>

#include "flops.h"
#include "magma.h"
#include "magma_lapack.h"
#include "testings.h"


int main( int argc, char** argv)
{
  //TESTING_CUDA_INIT();

  FILE *fp;

  magma_timestr_t  start, end;
  magma_int_t i,j,k;
  float       flops, magma_perf, cuda_perf, error, work[1];
  char        transA;
  char        transB;
  
  magma_int_t M, M0 = 0;
  magma_int_t N, N0 = 0;
  magma_int_t K, K0 = 0;

  const char  *uplo = MagmaLowerStr;
  magma_int_t info;


  
  float alpha = 1.0f;
  float beta = 0.0f;


  //  float *h_A, *h_B, *h_C;
  double *h_C;
  double *d_A, *d_B, *d_C;

  float c_neg_one = MAGMA_S_NEG_ONE;

  
  transA = MagmaTrans;
  transB = MagmaNoTrans;
  // A[MxK]
  M = 3;
  K = 1;

  //B [KxN]
  N = 4;

  //C [MxN]
  TESTING_MALLOC( h_C,  double, M*N );

  TESTING_DEVALLOC( d_A, double, M*K );
  TESTING_DEVALLOC( d_B, double, K*N );
  TESTING_DEVALLOC( d_C, double, M*N );


  double h_A[3] = {-0.222225,0.05397, 0.456780};
  double h_B[4] = {0.340188, -0.105617, 0.283099, 0.29844};   

  for (i = 0; i < M; i++)
    for (j = 0; j < K; j++){
      printf("A[%d] = %g\n",i*K+j,h_A[i*K+j] );
    }

  printf("\n");   
  
  for (i = 0; i < K; i++)
    for (j = 0; j< N; j++){
      printf("B[%d] = %g\n",i*N+j, h_B[i*N+j] );
    }
  
  cudaMemcpy(d_A, h_A, M*K*sizeof(double),cudaMemcpyHostToDevice);
  cudaMemcpy(d_B, h_B, K*N*sizeof(double),cudaMemcpyHostToDevice);


  //h_A[MxK]
  int lda = K;
  // h_B[KxN]
  int ldb = K;
  // h_C[MxN]
  int ldc = M;
  

  magmablas_dgemm( transA, transB, M, N, K, alpha, d_A, lda,
 		   d_B, ldb, beta,  d_C, ldc);
 
  magma_dgetmatrix( M, N, d_C, M, h_C, M );

  printf("\n");
  for (i = 0; i< M;i++)
    for (j = 0; j < N;j++)
      printf("C[%d] = %g\n",i*N+j,h_C[i*N+j] );

  magma_dpotrf_gpu(uplo[0], 2, d_B, 2, &info);
  if (info != 0)
    printf("Argument %d of magma_dpotrf had an illegal value.\n", (int) -info);

  magma_dpotri_gpu(uplo[0], 2, d_B, 2, &info);
  if (info != 0)
    printf("Argument %d of magma_dpotri had an illegal value.\n", (int) -info);
  
  magma_dgetmatrix( 2, 2, d_B, 2, h_B, 2 );

  for (i = 0; i< 2;i++)
    for (j = 0; j < 2;j++)
       printf("B_inv[%d] = %g\n",i*2+j,h_B[i*2+j] );



  TESTING_FREE( h_C );
  
  TESTING_DEVFREE( d_A );
  TESTING_DEVFREE( d_C );
  TESTING_DEVFREE( d_B );

  
  TESTING_CUDA_FINALIZE();
}

Code: Select all

#//////////////////////////////////////////////////////////////////////////////
#   -- MAGMA (version 1.2.1) --
#      Univ. of Tennessee, Knoxville
#      Univ. of California, Berkeley
#      Univ. of Colorado, Denver
#      June 2012
#//////////////////////////////////////////////////////////////////////////////


# GPU_TARGET specifies for which GPU you want to compile MAGMA:
#     "Tesla" (NVIDIA compute capability 1.x cards)
#     "Fermi" (NVIDIA compute capability 2.x cards)
# See http://developer.nvidia.com/cuda-gpus


GPU_TARGET = Fermi

CC        = g++ 
NVCC      = nvcc
FORT      = gfortran

ARCH      = ar
ARCHFLAGS = cr
RANLIB    = ranlib

OPTS      = -O3 -DADD_ -fopenmp -DUNIX -w -DMKL_ILP64 -m64 -I$(MKLROOT)/include
F77OPTS   = -O3 -DADD_ -fdefault-integer-8
FOPTS     = -O3 -DADD_ -x f95-cpp-input -fdefault-integer-8
NVOPTS    = -O3 -DADD_  -Xcompiler -fopenmp -lgomp --compiler-options -fno-strict-aliasing -DUNIX -Xptxas="-v"
LDOPTS    = -fPIC

# Multi-threaded version
LIB = -Wl,--start-group  $(MKLROOT)/lib/intel64/libmkl_intel_ilp64.a $(MKLROOT)/lib/intel64/libmkl_gnu_thread.a $(MKLROOT)/lib/intel64/libmkl_core.a -Wl,--end-group -ldl -lpthread -lm -lcublas -lcudart

CUDADIR   = /usr/local/cuda

LIBDIR    = -L /opt/intel/mkl/lib/intel64 \
            -L$(CUDADIR)/lib64

INC       = -I$(CUDADIR)/include