/* ////////////////////////////////////////////////////////////////////////////
-- Testing zprint
*/
int main( int argc, char** argv)
{
TESTING_INIT();
magmaDoubleComplex *hA, *dA;
//magma_int_t ione = 1;
//magma_int_t ISEED[4] = {0,0,0,1};
magma_int_t M, N, lda, ldda; //size
magma_int_t status = 0;
magma_opts opts;
parse_opts( argc, argv, &opts );
for( int itest = 0; itest < opts.ntest; ++itest ) {
for( int iter = 0; iter < opts.niter; ++iter ) {
M = opts.msize[itest];
N = opts.nsize[itest];
lda = M;
ldda = ((M + 31)/32)*32;
//size = lda*N;
/* Allocate host memory for the matrix */
TESTING_MALLOC_CPU( hA, magmaDoubleComplex, lda *N );
TESTING_MALLOC_DEV( dA, magmaDoubleComplex, ldda*N );
//lapackf77_zlarnv( &ione, ISEED, &size, hA );
for( int j = 0; j < N; ++j ) {
for( int i = 0; i < M; ++i ) {
hA[i + j*lda] = MAGMA_Z_MAKE( i + j*0.01, 0. );
}
}
magma_zsetmatrix( M, N, hA, lda, dA, ldda );
printf( "A=" );
magma_zprint( M, N, hA, lda );
printf( "dA=" );
magma_zprint_gpu( M, N, dA, ldda );
TESTING_FREE_CPU( hA );
TESTING_FREE_DEV( dA );
}
}
TESTING_FINALIZE();
return status;
}
/* ////////////////////////////////////////////////////////////////////////////
-- Testing zlaset
Code is very similar to testing_zlacpy.cpp
*/
int main( int argc, char** argv)
{
TESTING_INIT();
real_Double_t gbytes, gpu_perf, gpu_time, cpu_perf, cpu_time;
double error, work[1];
magmaDoubleComplex c_neg_one = MAGMA_Z_NEG_ONE;
magmaDoubleComplex *h_A, *h_R;
magmaDoubleComplex_ptr d_A;
magmaDoubleComplex offdiag, diag;
magma_int_t M, N, size, lda, ldda;
magma_int_t ione = 1;
magma_int_t status = 0;
magma_opts opts;
opts.parse_opts( argc, argv );
magma_uplo_t uplo[] = { MagmaLower, MagmaUpper, MagmaFull };
printf("%% uplo M N offdiag diag CPU GByte/s (ms) GPU GByte/s (ms) check\n");
printf("%%===================================================================================\n");
for( int iuplo = 0; iuplo < 3; ++iuplo ) {
for( int itest = 0; itest < opts.ntest; ++itest ) {
for( int iter = 0; iter < opts.niter; ++iter ) {
for( int ival = 0; ival < 4; ++ival ) {
// test combinations of zero & non-zero:
// ival offdiag diag
// 0 0 0
// 1 0 3.14
// 2 1.23 0
// 3 1.23 3.14
offdiag = MAGMA_Z_MAKE( 1.2345, 6.7890 ) * (ival / 2);
diag = MAGMA_Z_MAKE( 3.1415, 2.7183 ) * (ival % 2);
M = opts.msize[itest];
N = opts.nsize[itest];
//M += 2; // space for insets
//N += 2;
lda = M;
ldda = magma_roundup( M, opts.align );
size = lda*N;
if ( uplo[iuplo] == MagmaLower ) {
// save lower trapezoid (with diagonal)
if ( M > N ) {
gbytes = sizeof(magmaDoubleComplex) * (1.*M*N - 0.5*N*(N-1)) / 1e9;
} else {
gbytes = sizeof(magmaDoubleComplex) * 0.5*M*(M+1) / 1e9;
}
}
else if ( uplo[iuplo] == MagmaUpper ) {
// save upper trapezoid (with diagonal)
if ( N > M ) {
gbytes = sizeof(magmaDoubleComplex) * (1.*M*N - 0.5*M*(M-1)) / 1e9;
} else {
gbytes = sizeof(magmaDoubleComplex) * 0.5*N*(N+1) / 1e9;
}
}
else {
// save entire matrix
gbytes = sizeof(magmaDoubleComplex) * 1.*M*N / 1e9;
}
TESTING_MALLOC_CPU( h_A, magmaDoubleComplex, size );
TESTING_MALLOC_CPU( h_R, magmaDoubleComplex, size );
TESTING_MALLOC_DEV( d_A, magmaDoubleComplex, ldda*N );
/* Initialize the matrix */
for( int j = 0; j < N; ++j ) {
for( int i = 0; i < M; ++i ) {
h_A[i + j*lda] = MAGMA_Z_MAKE( i + j/10000., j );
}
}
/* ====================================================================
Performs operation using MAGMA
=================================================================== */
magma_zsetmatrix( M, N, h_A, lda, d_A, ldda );
magmablasSetKernelStream( opts.queue );
gpu_time = magma_sync_wtime( opts.queue );
//magmablas_zlaset( uplo[iuplo], M-2, N-2, offdiag, diag, d_A+1+ldda, ldda ); // inset by 1 row & col
magmablas_zlaset( uplo[iuplo], M, N, offdiag, diag, d_A, ldda );
gpu_time = magma_sync_wtime( opts.queue ) - gpu_time;
gpu_perf = gbytes / gpu_time;
/* =====================================================================
Performs operation using LAPACK
=================================================================== */
cpu_time = magma_wtime();
//magma_int_t M2 = M-2; // inset by 1 row & col
//magma_int_t N2 = N-2;
//lapackf77_zlaset( lapack_uplo_const( uplo[iuplo] ), &M2, &N2, &offdiag, &diag, h_A+1+lda, &lda );
lapackf77_zlaset( lapack_uplo_const( uplo[iuplo] ), &M, &N, &offdiag, &diag, h_A, &lda );
cpu_time = magma_wtime() - cpu_time;
cpu_perf = gbytes / cpu_time;
if ( opts.verbose ) {
printf( "A= " ); magma_zprint( M, N, h_A, lda );
printf( "dA=" ); magma_zprint_gpu( M, N, d_A, ldda );
}
/* =====================================================================
Check the result
=================================================================== */
magma_zgetmatrix( M, N, d_A, ldda, h_R, lda );
blasf77_zaxpy(&size, &c_neg_one, h_A, &ione, h_R, &ione);
error = lapackf77_zlange("f", &M, &N, h_R, &lda, work);
bool okay = (error == 0);
status += ! okay;
printf("%5s %5d %5d %9.4f %6.4f %7.2f (%7.2f) %7.2f (%7.2f) %s\n",
lapack_uplo_const( uplo[iuplo] ), (int) M, (int) N,
real(offdiag), real(diag),
cpu_perf, cpu_time*1000., gpu_perf, gpu_time*1000.,
(okay ? "ok" : "failed") );
TESTING_FREE_CPU( h_A );
TESTING_FREE_CPU( h_R );
TESTING_FREE_DEV( d_A );
fflush( stdout );
}
}
if ( opts.niter > 1 ) {
printf( "\n" );
}
}
printf( "\n" );
}
opts.cleanup();
TESTING_FINALIZE();
return status;
}