diff --git a/compute/pzgebrd_ge2gb.c b/compute/pzgebrd_ge2gb.c
index ea49d57ea15f34c33f2e8203f84fe77a05687eb8..0d27a9c089ed5b01ad96a6a755eed89c7fd21116 100644
--- a/compute/pzgebrd_ge2gb.c
+++ b/compute/pzgebrd_ge2gb.c
@@ -29,7 +29,6 @@ void morse_pzgebrd_ge2gb(MORSE_desc_t *A, MORSE_desc_t *T, MORSE_desc_t *D,
if (A->m >= A->n){
for (k = 0; k < A->nt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
A1 = morse_desc_submatrix(A, k*A->mb, k*A->nb, A->m-k*A->mb, tempkn);
@@ -47,7 +46,7 @@ void morse_pzgebrd_ge2gb(MORSE_desc_t *A, MORSE_desc_t *T, MORSE_desc_t *D,
sequence, request);
if (k+1 < A->nt){
- tempkn = k+1 == A->nt-1 ? A->n-(k+1)*A->nb : A->nb;
+ tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
A1 = morse_desc_submatrix(A, k*A->mb, (k+1)*A->nb, tempkm, A->n-(k+1)*A->nb);
A2 = morse_desc_submatrix(A, (k+1)*A->mb, (k+1)*A->nb, A->m-(k+1)*A->mb, A->n-(k+1)*A->nb);
@@ -68,7 +67,6 @@ void morse_pzgebrd_ge2gb(MORSE_desc_t *A, MORSE_desc_t *T, MORSE_desc_t *D,
else{
for (k = 0; k < A->mt; k++) {
tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
A1 = morse_desc_submatrix(A, k*A->mb, k*A->nb, tempkm, A->n-k*A->nb);
A2 = morse_desc_submatrix(A, (k+1)*A->mb, k*A->nb, A->m-(k+1)*A->mb, A->n-k*A->nb);
@@ -84,7 +82,7 @@ void morse_pzgebrd_ge2gb(MORSE_desc_t *A, MORSE_desc_t *T, MORSE_desc_t *D,
sequence, request);
if (k+1 < A->mt){
- tempkm = k+1 == A->mt-1 ? A->m-(k+1)*A->mb : A->mb;
+ tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
A1 = morse_desc_submatrix(A, (k+1)*A->mb, k*A->nb, A->m-(k+1)*A->mb, tempkn);
A2 = morse_desc_submatrix(A, (k+1)*A->mb, (k+1)*A->nb, A->m-(k+1)*A->mb, A->n-(k+1)*A->nb);
diff --git a/compute/pzgelqf_param.c b/compute/pzgelqf_param.c
index c6c5ceeaa3e7a422700bfa9b4f0fc488df9990c6..5acbf7c208b896ff0919fc1a3e13653b85d73d37 100644
--- a/compute/pzgelqf_param.c
+++ b/compute/pzgelqf_param.c
@@ -147,7 +147,6 @@ void morse_pzgelqf_param( const libhqr_tree_t *qrtree, MORSE_desc_t *A,
p = qrtree->currpiv(qrtree, k, n);
tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
- temppn = p == A->nt-1 ? A->n-p*A->nb : A->nb;
if (qrtree->gettype(qrtree, k, n) == 0) {
/* TS kernel */
diff --git a/compute/pztradd.c b/compute/pztradd.c
index 524136171756c19c79d4238e81b7be090e6f03d4..3db356b6eecafc0bccb2e7adeee4ad8a5f762157 100644
--- a/compute/pztradd.c
+++ b/compute/pztradd.c
@@ -172,7 +172,6 @@ void morse_pztradd(MORSE_enum uplo, MORSE_enum trans,
else {
for (m = 0; m < B->mt; m++) {
tempmm = m == B->mt-1 ? B->m-B->mb*m : B->nb;
- ldam = BLKLDD(A, m);
ldbm = BLKLDD(B, m);
for (n = 0; n < B->nt; n++) {
diff --git a/compute/pztrsm.c b/compute/pztrsm.c
index a621ad29eddda6f1bfbe6cf45fc323dc4b4e2c4a..366d08bbccaba46cc2e4b41cb02e7c7e74abcc06 100644
--- a/compute/pztrsm.c
+++ b/compute/pztrsm.c
@@ -193,7 +193,6 @@ void morse_pztrsm(MORSE_enum side, MORSE_enum uplo, MORSE_enum trans, MORSE_enum
}
RUNTIME_data_flush( sequence, A(B->mt-1-k, B->mt-1-k) );
for (m = k+1; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
ldbm = BLKLDD(B, B->mt-1-m);
for (n = 0; n < B->nt; n++) {
tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
diff --git a/compute/pzunmqr.c b/compute/pzunmqr.c
index c1897dcc1db6d8ac82d911afc274b1082f310074..d893bc3c18168d2166e3669913276d2e65bec072 100644
--- a/compute/pzunmqr.c
+++ b/compute/pzunmqr.c
@@ -245,7 +245,6 @@ void morse_pzunmqr(MORSE_enum side, MORSE_enum trans,
tempkn = k == B->nt - 1 ? B->n - k * B->nb : B->nb;
tempkmin = k == minMT - 1 ? minM - k * A->nb : A->nb;
ldak = BLKLDD(A, k);
- ldbk = BLKLDD(B, k);
for (n = B->nt-1; n > k; n--) {
tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
ldan = BLKLDD(A, n);
diff --git a/compute/pzunmqrrh.c b/compute/pzunmqrrh.c
index c4d6e3e34367dad7d10634662935ac042d32e280..af3fe285681745e04fd9bafc6c7834a9444bfd10 100644
--- a/compute/pzunmqrrh.c
+++ b/compute/pzunmqrrh.c
@@ -330,7 +330,6 @@ void morse_pzunmqrrh( MORSE_enum side, MORSE_enum trans,
tempMm = M == A->mt-1 ? A->m-M*A->mb : A->mb;
tempkmin = chameleon_min(tempMm, tempkn);
ldaM = BLKLDD(A, M);
- ldbM = BLKLDD(B, M);
for (n = chameleon_min(M+BS, A->mt)-1; n > M; n--) {
ldan = BLKLDD(A, n);
tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
diff --git a/compute/zgesvd.c b/compute/zgesvd.c
index 235bda79b6f0036a05d24d39bee241009bc6b78d..be88952272af125224d03b30bc4050d3e5f2be26 100644
--- a/compute/zgesvd.c
+++ b/compute/zgesvd.c
@@ -400,8 +400,8 @@ int MORSE_zgesvd_Tile_Async( MORSE_enum jobu, MORSE_enum jobvt,
{
MORSE_desc_t descA;
MORSE_desc_t descT;
- MORSE_desc_t descU;
- MORSE_desc_t descVT;
+ MORSE_desc_t descUl, descUt;
+ MORSE_desc_t descVTl, descVTt;
MORSE_desc_t descAB;
MORSE_desc_t D, *Dptr = NULL;
MORSE_desc_t *subA, *subT, *subUVT;
@@ -555,65 +555,61 @@ int MORSE_zgesvd_Tile_Async( MORSE_enum jobu, MORSE_enum jobvt,
morse_desc_mat_free( &descAB );
- /* Transform U and Vt into tile format */
- if ( jobu != MorseNoVec ) {
- //morse_zooplap2tile( descU, U, NB, NB, LDU, M, 0, 0, M, M, sequence, request, morse_desc_mat_free(&(descU)) );
- }
- if ( jobvt != MorseNoVec ) {
- //morse_zooplap2tile( descVT, VT, NB, NB, LDVT, N, 0, 0, N, N, sequence, request, morse_desc_mat_free(&(descVT)) );
- }
-
- morse_sequence_wait( morse, sequence );
-
subA = NULL;
subT = NULL;
subUVT = NULL;
+
if ( jobu != MorseNoVec ) {
+ morse_zlap2tile( morse, &descUl, &descUt, MorseDescInout, MorseUpperLower,
+ U, NB, NB, LDU, M, M, M, sequence, request );
+
if ( M < N ){
- subA = morse_desc_submatrix(&descA, descA.mb, 0, descA.m-descA.mb, descA.n-descA.nb);
- subUVT = morse_desc_submatrix(&descU, descU.mb, 0, descU.m-descU.mb, descU.n);
- subT = morse_desc_submatrix(&descT, descT.mb, 0, descT.m-descT.mb, descT.n-descT.nb);
+ subA = morse_desc_submatrix(&descA, descA.mb, 0, descA.m -descA.mb, descA.n-descA.nb);
+ subUVT = morse_desc_submatrix(&descUt, descUt.mb, 0, descUt.m-descUt.mb, descUt.n);
+ subT = morse_desc_submatrix(&descT, descT.mb, 0, descT.m -descT.mb, descT.n-descT.nb);
+
morse_pzunmqr( MorseLeft, MorseNoTrans,
subA, subUVT, subT, Dptr,
sequence, request );
+
+ free(subA); free(subUVT); free(subT);
}
else {
morse_pzunmqr( MorseLeft, MorseNoTrans,
- &descA, &descU, &descT, Dptr,
+ &descA, &descUt, &descT, Dptr,
sequence, request );
}
+
+ morse_ztile2lap( morse, &descUl, &descUt,
+ MorseDescInout, MorseUpperLower, sequence, request );
}
if ( jobvt != MorseNoVec ) {
+ morse_zlap2tile( morse, &descVTl, &descVTt, MorseDescInout, MorseUpperLower,
+ VT, NB, NB, LDVT, N, N, N, sequence, request );
+
if ( M < N ){
morse_pzunmlq( MorseRight, MorseNoTrans,
- &descA, &descVT, &descT, Dptr,
+ &descA, &descVTt, &descT, Dptr,
sequence, request );
}
else {
- subA = morse_desc_submatrix(&descA, 0, descA.nb, descA.m-descA.mb, descA.n -descA.nb);
- subUVT = morse_desc_submatrix(&descVT, 0, descVT.nb, descVT.m, descVT.n-descVT.nb);
- subT = morse_desc_submatrix(&descT, 0, descT.nb, descT.m-descT.mb, descT.n -descT.nb);
+ subA = morse_desc_submatrix(&descA, 0, descA.nb, descA.m-descA.mb, descA.n -descA.nb );
+ subUVT = morse_desc_submatrix(&descVTt, 0, descVTt.nb, descVTt.m, descVTt.n-descVTt.nb);
+ subT = morse_desc_submatrix(&descT, 0, descT.nb, descT.m-descT.mb, descT.n -descT.nb );
+
morse_pzunmlq( MorseRight, MorseNoTrans,
subA, subUVT, subT, Dptr,
sequence, request );
+
+ free(subA); free(subUVT); free(subT);
}
- }
- /* Transform U and VT into lapack layout */
- if ( jobu != MorseNoVec ) {
- //morse_zooptile2lap( descU, U, NB, NB, LDU, M, sequence, request );
- }
- if ( jobvt != MorseNoVec ) {
- //morse_zooptile2lap( descVT, VT, NB, NB, LDVT, N, sequence, request );
+ morse_ztile2lap( morse, &descVTl, &descVTt,
+ MorseDescInout, MorseUpperLower, sequence, request );
}
-
morse_sequence_wait( morse, sequence );
- if (subA) {
- free(subA); free(subUVT); free(subT);
- }
-
/* Solve the bidiagonal SVD problem */
/* On exit, U and VT are updated with bidiagonal matrix singular vectors */
#if !defined(CHAMELEON_SIMULATION)
@@ -626,12 +622,6 @@ int MORSE_zgesvd_Tile_Async( MORSE_enum jobu, MORSE_enum jobvt,
}
#endif /* !defined(CHAMELEON_SIMULATION) */
- if (jobu != MorseNoVec) {
- morse_desc_mat_free( &descU );
- }
- if (jobvt != MorseNoVec) {
- morse_desc_mat_free( &descVT );
- }
free(E);
if ( Dptr ) {
morse_desc_mat_free( Dptr );
diff --git a/compute/zlange.c b/compute/zlange.c
index e0d4658470dc43728270a453372c2fd164a362ce..a7b6ffaf38a4b9dff46aadac6d2b4d8409797154 100644
--- a/compute/zlange.c
+++ b/compute/zlange.c
@@ -83,7 +83,7 @@ double MORSE_zlange(MORSE_enum norm, int M, int N,
{
int NB;
int status;
- double value;
+ double value = -1.;
MORSE_context_t *morse;
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
@@ -125,7 +125,7 @@ double MORSE_zlange(MORSE_enum norm, int M, int N,
}
/* Set NT */
- NB = MORSE_NB;
+ NB = MORSE_NB;
morse_sequence_create( morse, &sequence );
@@ -194,7 +194,7 @@ double MORSE_zlange_Tile(MORSE_enum norm, MORSE_desc_t *A )
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
int status;
- double value;
+ double value = -1.;
morse = morse_context_self();
if (morse == NULL) {
diff --git a/compute/zlanhe.c b/compute/zlanhe.c
index 20d6d2873527213c6d8575ca435151edfdf77460..fb2e7f300990bc7ff424e86793e6f46a76a787f0 100644
--- a/compute/zlanhe.c
+++ b/compute/zlanhe.c
@@ -83,7 +83,7 @@ double MORSE_zlanhe(MORSE_enum norm, MORSE_enum uplo, int N,
{
int NB;
int status;
- double value;
+ double value = -1.;
MORSE_context_t *morse;
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
@@ -198,7 +198,7 @@ double MORSE_zlanhe_Tile( MORSE_enum norm, MORSE_enum uplo, MORSE_desc_t *A )
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
int status;
- double value;
+ double value = -1.;
morse = morse_context_self();
if (morse == NULL) {
diff --git a/compute/zlansy.c b/compute/zlansy.c
index 10a2d3557283c1815585184480852bf62eb41d6c..673532d77d95d763a18d34b8279907da0581c936 100644
--- a/compute/zlansy.c
+++ b/compute/zlansy.c
@@ -83,7 +83,7 @@ double MORSE_zlansy(MORSE_enum norm, MORSE_enum uplo, int N,
{
int NB;
int status;
- double value;
+ double value = -1.;
MORSE_context_t *morse;
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
@@ -198,7 +198,7 @@ double MORSE_zlansy_Tile( MORSE_enum norm, MORSE_enum uplo, MORSE_desc_t *A )
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
int status;
- double value;
+ double value = -1.;
morse = morse_context_self();
if (morse == NULL) {
diff --git a/compute/zlantr.c b/compute/zlantr.c
index 42127a8b32fa5e669e886a61c1158906d9188b1e..224faca65f8f75400e08ad786be2d77a102697cb 100644
--- a/compute/zlantr.c
+++ b/compute/zlantr.c
@@ -95,7 +95,7 @@ double MORSE_zlantr(MORSE_enum norm, MORSE_enum uplo, MORSE_enum diag,
{
int NB;
int status;
- double value;
+ double value = -1.;
MORSE_context_t *morse;
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
@@ -220,7 +220,7 @@ double MORSE_zlantr_Tile(MORSE_enum norm, MORSE_enum uplo, MORSE_enum diag, MORS
MORSE_sequence_t *sequence = NULL;
MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
int status;
- double value;
+ double value = -1.;
morse = morse_context_self();
if (morse == NULL) {
diff --git a/example/lapack_to_morse/step0.c b/example/lapack_to_morse/step0.c
index 6879a1ef1e3ab1f0fde0ddb067255e842bb91128..0ae46a070a70d62fa6540d9257c26bc18ea2f4ad 100644
--- a/example/lapack_to_morse/step0.c
+++ b/example/lapack_to_morse/step0.c
@@ -36,7 +36,8 @@ int main(int argc, char *argv[]) {
size_t NRHS; // number of RHS vectors
/* declarations to time the program and evaluate performances */
- double fmuls, fadds, flops, gflops, cpu_time;
+ double fmuls, fadds, flops, gflops, cpu_time, eps, res;
+ int hres;
/* initialize some parameters with default values */
int iparam[IPARAM_SIZEOF];
@@ -52,8 +53,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -137,26 +136,27 @@ int main(int argc, char *argv[]) {
/* compute A*X-B, store the result in B */
cblas_dgemm(CblasColMajor, CblasNoTrans, CblasNoTrans,
N, NRHS, N, 1.0, Acpy, N, X, N, -1.0, B, N);
- double res = LAPACKE_dlange( LAPACK_COL_MAJOR, 'I', N, NRHS, B, N);
+ res = LAPACKE_dlange( LAPACK_COL_MAJOR, 'I', N, NRHS, B, N);
/* check residual and print a message */
- double eps = LAPACKE_dlamch_work( 'e' );
+ eps = LAPACKE_dlamch_work( 'e' );
/*
* if hres = 0 then the test succeed
* else the test failed
*/
- int hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
- else
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* deallocate data */
free(A);
diff --git a/example/lapack_to_morse/step1.c b/example/lapack_to_morse/step1.c
index 6099a6f9a673f2df637a93b3f28d02f9cd16c1cb..025f6a53494a829039b2eed6185095e0a0544aed 100644
--- a/example/lapack_to_morse/step1.c
+++ b/example/lapack_to_morse/step1.c
@@ -38,7 +38,6 @@
* The precision is: double
*/
int main(int argc, char *argv[]) {
-
size_t N; // matrix order
size_t NRHS; // number of RHS vectors
int NCPU; // number of cores to use
@@ -66,8 +65,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -88,7 +85,7 @@ int main(int argc, char *argv[]) {
}
/*
- * allocate memory for our data using a C macro (see step1.h)
+ * Allocate memory for our data using a C macro (see step1.h)
* - matrix A : size N x N
* - set of RHS vectors B : size N x NRHS
* - set of solutions vectors X : size N x NRHS
@@ -154,17 +151,18 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
- else
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* deallocate data */
free(A);
diff --git a/example/lapack_to_morse/step2.c b/example/lapack_to_morse/step2.c
index fbb13949987b0d2379a3e5fabe0b1943f5e59fb2..e0195479c27d63a73647042164ab92f72ec90a70 100644
--- a/example/lapack_to_morse/step2.c
+++ b/example/lapack_to_morse/step2.c
@@ -29,8 +29,7 @@
* The goal is also to show examples of calling MORSE tile interface.
*/
int main(int argc, char *argv[]) {
-
- size_t N; // matrix order
+ size_t N; // matrix order
int NB; // number of rows and columns in tiles
int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
@@ -61,8 +60,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -76,7 +73,7 @@ int main(int argc, char *argv[]) {
/* print informations to user */
print_header( argv[0], iparam);
- /* Initialize MORSE with main parameters */
+ /* Initialize MORSE with main parameters */
if ( MORSE_Init( NCPU, NGPU ) != MORSE_SUCCESS ) {
fprintf(stderr, "Error initializing MORSE library\n");
return EXIT_FAILURE;
@@ -198,17 +195,18 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
- else
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* get back results in LAPACK format */
MORSE_dTile_to_Lapack(descA, A, N);
diff --git a/example/lapack_to_morse/step3.c b/example/lapack_to_morse/step3.c
index fdad78747b24d0704f8be9e8d7be4bcc2bbf848a..eb312137b9fc7e11c031472ea751c147b4f1462e 100644
--- a/example/lapack_to_morse/step3.c
+++ b/example/lapack_to_morse/step3.c
@@ -27,7 +27,6 @@
* During this step we do not use classical LAPACK matrices (1D array) anymore.
*/
int main(int argc, char *argv[]) {
-
size_t N; // matrix order
int NB; // number of rows and columns in tiles
int NRHS; // number of RHS vectors
@@ -62,8 +61,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -77,7 +74,7 @@ int main(int argc, char *argv[]) {
/* print informations to user */
print_header( argv[0], iparam);
- /* Initialize MORSE with main parameters */
+ /* Initialize MORSE with main parameters */
if ( MORSE_Init( NCPU, NGPU ) != MORSE_SUCCESS ) {
fprintf(stderr, "Error initializing MORSE library\n");
return EXIT_FAILURE;
@@ -181,17 +178,18 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
- else
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* free the matrix of tiles */
deallocate_tile_matrix(matA, N, N, NB);
diff --git a/example/lapack_to_morse/step4.c b/example/lapack_to_morse/step4.c
index 6eafc33b1b87f95a5822799e8068351de29a9e4a..32abb06f164fd6e18e1fdf7749ef610ae25269d4 100644
--- a/example/lapack_to_morse/step4.c
+++ b/example/lapack_to_morse/step4.c
@@ -31,8 +31,7 @@
* subsequent algorithms so that less synchronisations are done.
*/
int main(int argc, char *argv[]) {
-
- size_t N; // matrix order
+ size_t N; // matrix order
int NB; // number of rows and columns in tiles
int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
@@ -70,8 +69,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -85,7 +82,7 @@ int main(int argc, char *argv[]) {
/* print informations to user */
print_header( argv[0], iparam);
- /* Initialize MORSE with main parameters */
+ /* Initialize MORSE with main parameters */
if ( MORSE_Init( NCPU, NGPU ) != MORSE_SUCCESS ) {
fprintf(stderr, "Error initializing MORSE library\n");
return EXIT_FAILURE;
@@ -179,17 +176,18 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
- else
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* deallocate A, B, X, Acpy and associated descriptors descA, ... */
MORSE_Desc_Destroy( &descA );
diff --git a/example/lapack_to_morse/step5.c b/example/lapack_to_morse/step5.c
index 0b7c03a9b27ae942a2f2da93d786fa721fdd7f4f..a3ef2994dced8993bda8d9041f02302bf9cc5034 100644
--- a/example/lapack_to_morse/step5.c
+++ b/example/lapack_to_morse/step5.c
@@ -31,8 +31,7 @@
* - number of right-hand sides
*/
int main(int argc, char *argv[]) {
-
- size_t N; // matrix order
+ size_t N; // matrix order
int NB; // number of rows and columns in tiles
int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
@@ -71,8 +70,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -183,17 +180,18 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
- res, anorm, xnorm, bnorm,
- res / N / eps / (anorm * xnorm + bnorm ));
- else
+ res, anorm, xnorm, bnorm,
+ res / N / eps / (anorm * xnorm + bnorm ));
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
- res, anorm, xnorm, bnorm,
- res / N / eps / (anorm * xnorm + bnorm ));
+ res, anorm, xnorm, bnorm,
+ res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* deallocate A, B, X, Acpy and associated descriptors descA, ... */
MORSE_Desc_Destroy( &descA );
diff --git a/example/lapack_to_morse/step6.c b/example/lapack_to_morse/step6.c
index 49e02c1b00f429c3dfa887a021d7c035f3ed76ff..3397942336eb891c2e32a2e92ba8abd69f3977fc 100644
--- a/example/lapack_to_morse/step6.c
+++ b/example/lapack_to_morse/step6.c
@@ -29,8 +29,7 @@
* --p= and --q=
*/
int main(int argc, char *argv[]) {
-
- size_t N; // matrix order
+ size_t N; // matrix order
int NB; // number of rows and columns in tiles
int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
@@ -72,8 +71,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -181,8 +178,9 @@ int main(int argc, char *argv[]) {
/* print informations to user */
gflops = flops / cpu_time;
- if ( MORSE_My_Mpi_Rank() == 0 )
+ if ( MORSE_My_Mpi_Rank() == 0 ) {
printf( "%9.3f %9.2f\n", cpu_time, gflops);
+ }
fflush( stdout );
/************************************************************/
@@ -206,18 +204,19 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
if ( MORSE_My_Mpi_Rank() == 0 ){
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
- res, anorm, xnorm, bnorm,
- res / N / eps / (anorm * xnorm + bnorm ));
- else
+ res, anorm, xnorm, bnorm,
+ res / N / eps / (anorm * xnorm + bnorm ));
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
- res, anorm, xnorm, bnorm,
- res / N / eps / (anorm * xnorm + bnorm ));
+ res, anorm, xnorm, bnorm,
+ res / N / eps / (anorm * xnorm + bnorm ));
+ }
}
/* deallocate A, B, X, Acpy and associated descriptors descA, ... */
diff --git a/example/lapack_to_morse/step7.c b/example/lapack_to_morse/step7.c
index 4aa0b0001f3a8dfd4f71ec4b4892f47c62ab2016..2c5f408725edc0c78a604997ab3ec94ece680184 100644
--- a/example/lapack_to_morse/step7.c
+++ b/example/lapack_to_morse/step7.c
@@ -33,8 +33,7 @@
* and therefore this approach is much more flexible.
*/
int main(int argc, char *argv[]) {
-
- size_t N; // matrix order
+ size_t N; // matrix order
int NB; // number of rows and columns in tiles
int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
@@ -76,8 +75,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv
* It makes sense only if this program is linked with pthread and
@@ -189,8 +186,9 @@ int main(int argc, char *argv[]) {
/* print informations to user */
gflops = flops / cpu_time;
- if ( MORSE_My_Mpi_Rank() == 0 )
+ if ( MORSE_My_Mpi_Rank() == 0 ) {
printf( "%9.3f %9.2f\n", cpu_time, gflops);
+ }
fflush( stdout );
/************************************************************/
@@ -214,18 +212,19 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
if ( MORSE_My_Mpi_Rank() == 0 ){
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
- res, anorm, xnorm, bnorm,
- res / N / eps / (anorm * xnorm + bnorm ));
- else
+ res, anorm, xnorm, bnorm,
+ res / N / eps / (anorm * xnorm + bnorm ));
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
- res, anorm, xnorm, bnorm,
- res / N / eps / (anorm * xnorm + bnorm ));
+ res, anorm, xnorm, bnorm,
+ res / N / eps / (anorm * xnorm + bnorm ));
+ }
}
/* deallocate A, B, X, Acpy and associated descriptors descA, ... */
diff --git a/example/out_of_core/out_of_core.c b/example/out_of_core/out_of_core.c
index 0db7f3af70c30d48d13f027c7c8fcd4731d34844..90af9ca34e5a59fed75f872d262552863b61457c 100644
--- a/example/out_of_core/out_of_core.c
+++ b/example/out_of_core/out_of_core.c
@@ -56,8 +56,6 @@ int main(int argc, char *argv[]) {
fadds = (double)( FADDS_POTRF(N) + 2 * FADDS_TRSM(N,NRHS) );
fmuls = (double)( FMULS_POTRF(N) + 2 * FMULS_TRSM(N,NRHS) );
flops = 1e-9 * (fmuls + fadds);
- gflops = 0.0;
- cpu_time = 0.0;
/* initialize the number of thread if not given by the user in argv */
if ( iparam[IPARAM_THRDNBR] == -1 ) {
@@ -166,17 +164,18 @@ int main(int argc, char *argv[]) {
* if hres = 0 then the test succeed
* else the test failed
*/
- hres = 0;
- hres = ( res / N / eps / (anorm * xnorm + bnorm ) > 100.0 );
+ hres = ( (res / N / eps / (anorm * xnorm + bnorm )) > 100.0 );
printf( " ||Ax-b|| ||A|| ||x|| ||b|| ||Ax-b||/N/eps/(||A||||x||+||b||) RETURN\n");
- if (hres)
+ if (hres) {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e FAILURE \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
- else
+ }
+ else {
printf( "%8.5e %8.5e %8.5e %8.5e %8.5e SUCCESS \n",
res, anorm, xnorm, bnorm,
res / N / eps / (anorm * xnorm + bnorm ));
+ }
/* free descriptors descA, descB, descX, descAC */
MORSE_Desc_Destroy( &descA );
diff --git a/testing/testing_zgels.c b/testing/testing_zgels.c
index 1259502af39c374e83d1e28936c80cf804b9cb0f..46da823c103411ec88f9c4afabf1b4ae9228cae0 100644
--- a/testing/testing_zgels.c
+++ b/testing/testing_zgels.c
@@ -34,170 +34,6 @@
#include <coreblas.h>
#include "testing_zauxiliary.h"
-enum blas_order_type {
- blas_rowmajor = 101,
- blas_colmajor = 102 };
-
-enum blas_uplo_type {
- blas_upper = 121,
- blas_lower = 122 };
-
-enum blas_cmach_type {
- blas_base = 151,
- blas_t = 152,
- blas_rnd = 153,
- blas_ieee = 154,
- blas_emin = 155,
- blas_emax = 156,
- blas_eps = 157,
- blas_prec = 158,
- blas_underflow = 159,
- blas_overflow = 160,
- blas_sfmin = 161};
-
-enum blas_norm_type {
- blas_one_norm = 171,
- blas_real_one_norm = 172,
- blas_two_norm = 173,
- blas_frobenius_norm = 174,
- blas_inf_norm = 175,
- blas_real_inf_norm = 176,
- blas_max_norm = 177,
- blas_real_max_norm = 178 };
-
-static void
-BLAS_error(char *rname, int err, int val, int x) {
- fprintf( stderr, "%s %d %d %d\n", rname, err, val, x );
- abort();
-}
-
-static
-void
-BLAS_zsy_norm(enum blas_order_type order, enum blas_norm_type norm,
- enum blas_uplo_type uplo, int n, const MORSE_Complex64_t *a, int lda, double *res) {
- int i, j; double anorm, v;
- char rname[] = "BLAS_zsy_norm";
-
- if (order != blas_colmajor) BLAS_error( rname, -1, order, 0 );
-
- if (norm == blas_inf_norm) {
- anorm = 0.0;
- if (blas_upper == uplo) {
- for (i = 0; i < n; ++i) {
- v = 0.0;
- for (j = 0; j < i; ++j) {
- v += cabs( a[j + i * lda] );
- }
- for (j = i; j < n; ++j) {
- v += cabs( a[i + j * lda] );
- }
- if (v > anorm)
- anorm = v;
- }
- } else {
- BLAS_error( rname, -3, norm, 0 );
- return;
- }
- } else {
- BLAS_error( rname, -2, norm, 0 );
- return;
- }
-
- if (res) *res = anorm;
-}
-
-static
-void
-BLAS_zge_norm(enum blas_order_type order, enum blas_norm_type norm,
- int m, int n, const MORSE_Complex64_t *a, int lda, double *res) {
- int i, j; float anorm, v;
- char rname[] = "BLAS_zge_norm";
-
- if (order != blas_colmajor) BLAS_error( rname, -1, order, 0 );
-
- if (norm == blas_frobenius_norm) {
- anorm = 0.0f;
- for (j = n; j; --j) {
- for (i = m; i; --i) {
- v = a[0];
- anorm += v * v;
- a++;
- }
- a += lda - m;
- }
- anorm = sqrt( anorm );
- } else if (norm == blas_inf_norm) {
- anorm = 0.0f;
- for (i = 0; i < m; ++i) {
- v = 0.0f;
- for (j = 0; j < n; ++j) {
- v += cabs( a[i + j * lda] );
- }
- if (v > anorm)
- anorm = v;
- }
- } else {
- BLAS_error( rname, -2, norm, 0 );
- return;
- }
-
- if (res) *res = anorm;
-}
-
-static
-double
-BLAS_dpow_di(double x, int n) {
- double rv = 1.0;
-
- if (n < 0) {
- n = -n;
- x = 1.0 / x;
- }
-
- for (; n; n >>= 1, x *= x) {
- if (n & 1)
- rv *= x;
- }
-
- return rv;
-}
-
-static
-double
-BLAS_dfpinfo(enum blas_cmach_type cmach) {
- double eps = 1.0, r = 1.0, o = 1.0, b = 2.0;
- int t = 53, l = 1024, m = -1021;
- char rname[] = "BLAS_dfpinfo";
-
- if ((sizeof eps) == sizeof(float)) {
- t = 24;
- l = 128;
- m = -125;
- } else {
- t = 53;
- l = 1024;
- m = -1021;
- }
-
- /* for (i = 0; i < t; ++i) eps *= half; */
- eps = BLAS_dpow_di( b, -t );
- /* for (i = 0; i >= m; --i) r *= half; */
- r = BLAS_dpow_di( b, m-1 );
-
- o -= eps;
- /* for (i = 0; i < l; ++i) o *= b; */
- o = (o * BLAS_dpow_di( b, l-1 )) * b;
-
- switch (cmach) {
- case blas_eps: return eps;
- case blas_sfmin: return r;
- default:
- BLAS_error( rname, -1, cmach, 0 );
- break;
- }
- return 0.0;
-}
-
static int check_orthogonality(int, int, int, MORSE_Complex64_t*, double);
static int check_factorization(int, int, MORSE_Complex64_t*, MORSE_Complex64_t*, int, MORSE_Complex64_t*, double);
static int check_solution(int, int, int, MORSE_Complex64_t*, int, MORSE_Complex64_t*, MORSE_Complex64_t*, int, double);
@@ -268,7 +104,7 @@ int testing_zgels(int argc, char **argv)
MORSE_Alloc_Workspace_zgels(M, N, &T, 1, 1);
memset(T->mat, 0, (T->llm*T->lln)*sizeof(MORSE_Complex64_t));
- eps = BLAS_dfpinfo( blas_eps );
+ eps = LAPACKE_dlamch_work('e');
/*----------------------------------------------------------
* TESTING ZGELS
@@ -508,16 +344,18 @@ static int check_orthogonality(int M, int N, int LDQ, MORSE_Complex64_t *Q, doub
/* Build the idendity matrix USE DLASET?*/
MORSE_Complex64_t *Id = (MORSE_Complex64_t *) malloc(minMN*minMN*sizeof(MORSE_Complex64_t));
memset((void*)Id, 0, minMN*minMN*sizeof(MORSE_Complex64_t));
- for (i = 0; i < minMN; i++)
+ for (i = 0; i < minMN; i++) {
Id[i*minMN+i] = (MORSE_Complex64_t)1.0;
+ }
/* Perform Id - Q'Q */
- if (M >= N)
+ if (M >= N) {
cblas_zherk(CblasColMajor, CblasUpper, CblasConjTrans, N, M, alpha, Q, LDQ, beta, Id, N);
- else
+ }
+ else {
cblas_zherk(CblasColMajor, CblasUpper, CblasNoTrans, M, N, alpha, Q, LDQ, beta, Id, M);
-
- BLAS_zsy_norm( blas_colmajor, blas_inf_norm, blas_upper, minMN, Id, minMN, &normQ );
+ }
+ normQ = LAPACKE_zlansy_work( LAPACK_COL_MAJOR, 'i', 'u', minMN, Id, minMN, work );
printf("============\n");
printf("Checking the orthogonality of Q \n");
@@ -583,8 +421,8 @@ static int check_factorization(int M, int N, MORSE_Complex64_t *A1, MORSE_Comple
for (j = 0 ; j < N; j++)
Residual[j*M+i] = A1[j*LDA+i]-Ql[j*M+i];
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, M, N, Residual, M, &Rnorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, M, N, A2, LDA, &Anorm );
+ Rnorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', M, N, Residual, M, work );
+ Anorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', M, N, A2, LDA, work );
if (M >= N) {
printf("============\n");
@@ -626,9 +464,9 @@ static int check_solution(int M, int N, int NRHS, MORSE_Complex64_t *A, int LDA,
alpha = 1.0;
beta = -1.0;
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, M, N, A, LDA, &Anorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B, LDB, &Bnorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, M, NRHS, X, LDB, &Xnorm );
+ Anorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', M, N, A, LDA, work );
+ Bnorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', N, NRHS, B, LDB, work );
+ Xnorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', M, NRHS, X, LDB, work );
cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, M, NRHS, N, CBLAS_SADDR(alpha), A, LDA, X, LDB, CBLAS_SADDR(beta), B, LDB);
@@ -636,14 +474,14 @@ static int check_solution(int M, int N, int NRHS, MORSE_Complex64_t *A, int LDA,
MORSE_Complex64_t *Residual = (MORSE_Complex64_t *)malloc(M*NRHS*sizeof(MORSE_Complex64_t));
memset((void*)Residual, 0, M*NRHS*sizeof(MORSE_Complex64_t));
cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, N, NRHS, M, CBLAS_SADDR(alpha), A, LDA, B, LDB, CBLAS_SADDR(beta), Residual, M);
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, M, NRHS, Residual, M, &Rnorm );
+ Rnorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', M, NRHS, Residual, M, work );
free(Residual);
}
else {
MORSE_Complex64_t *Residual = (MORSE_Complex64_t *)malloc(N*NRHS*sizeof(MORSE_Complex64_t));
memset((void*)Residual, 0, N*NRHS*sizeof(MORSE_Complex64_t));
cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, N, NRHS, M, CBLAS_SADDR(alpha), A, LDA, B, LDB, CBLAS_SADDR(beta), Residual, N);
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, Residual, N, &Rnorm );
+ Rnorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', N, NRHS, Residual, N, work );
free(Residual);
}
diff --git a/testing/testing_zgesv_incpiv.c b/testing/testing_zgesv_incpiv.c
index dcc2a25d2687bae71baff909b0f192ad9930fe4e..b7702f153efe5c24c8a0f702e0ed8aa22e66e3a1 100644
--- a/testing/testing_zgesv_incpiv.c
+++ b/testing/testing_zgesv_incpiv.c
@@ -33,131 +33,6 @@
#include <coreblas.h>
#include "testing_zauxiliary.h"
-enum blas_order_type {
- blas_rowmajor = 101,
- blas_colmajor = 102 };
-
-enum blas_cmach_type {
- blas_base = 151,
- blas_t = 152,
- blas_rnd = 153,
- blas_ieee = 154,
- blas_emin = 155,
- blas_emax = 156,
- blas_eps = 157,
- blas_prec = 158,
- blas_underflow = 159,
- blas_overflow = 160,
- blas_sfmin = 161};
-
-enum blas_norm_type {
- blas_one_norm = 171,
- blas_real_one_norm = 172,
- blas_two_norm = 173,
- blas_frobenius_norm = 174,
- blas_inf_norm = 175,
- blas_real_inf_norm = 176,
- blas_max_norm = 177,
- blas_real_max_norm = 178 };
-
-static void
-BLAS_error(char *rname, int err, int val, int x) {
- fprintf( stderr, "%s %d %d %d\n", rname, err, val, x );
- abort();
-}
-
-static
-void
-BLAS_zge_norm(enum blas_order_type order, enum blas_norm_type norm,
- int m, int n, const MORSE_Complex64_t *a, int lda, double *res) {
- int i, j; float anorm, v;
- char rname[] = "BLAS_zge_norm";
-
- if (order != blas_colmajor) BLAS_error( rname, -1, order, 0 );
-
- if (norm == blas_frobenius_norm) {
- anorm = 0.0f;
- for (j = n; j; --j) {
- for (i = m; i; --i) {
- v = a[0];
- anorm += v * v;
- a++;
- }
- a += lda - m;
- }
- anorm = sqrt( anorm );
- } else if (norm == blas_inf_norm) {
- anorm = 0.0f;
- for (i = 0; i < m; ++i) {
- v = 0.0f;
- for (j = 0; j < n; ++j) {
- v += cabs( a[i + j * lda] );
- }
- if (v > anorm)
- anorm = v;
- }
- } else {
- BLAS_error( rname, -2, norm, 0 );
- return;
- }
-
- if (res) *res = anorm;
-}
-
-static
-double
-BLAS_dpow_di(double x, int n) {
- double rv = 1.0;
-
- if (n < 0) {
- n = -n;
- x = 1.0 / x;
- }
-
- for (; n; n >>= 1, x *= x) {
- if (n & 1)
- rv *= x;
- }
-
- return rv;
-}
-
-static
-double
-BLAS_dfpinfo(enum blas_cmach_type cmach) {
- double eps = 1.0, r = 1.0, o = 1.0, b = 2.0;
- int t = 53, l = 1024, m = -1021;
- char rname[] = "BLAS_dfpinfo";
-
- if ((sizeof eps) == sizeof(float)) {
- t = 24;
- l = 128;
- m = -125;
- } else {
- t = 53;
- l = 1024;
- m = -1021;
- }
-
- /* for (i = 0; i < t; ++i) eps *= half; */
- eps = BLAS_dpow_di( b, -t );
- /* for (i = 0; i >= m; --i) r *= half; */
- r = BLAS_dpow_di( b, m-1 );
-
- o -= eps;
- /* for (i = 0; i < l; ++i) o *= b; */
- o = (o * BLAS_dpow_di( b, l-1 )) * b;
-
- switch (cmach) {
- case blas_eps: return eps;
- case blas_sfmin: return r;
- default:
- BLAS_error( rname, -1, cmach, 0 );
- break;
- }
- return 0.0;
-}
-
static int check_solution(int, int , MORSE_Complex64_t *, int, MORSE_Complex64_t *, MORSE_Complex64_t *, int, double);
int testing_zgesv_incpiv(int argc, char **argv)
@@ -199,7 +74,7 @@ int testing_zgesv_incpiv(int argc, char **argv)
return -2;
}
- eps = BLAS_dfpinfo(blas_eps);
+ eps = LAPACKE_dlamch_work( 'e' );
/*----------------------------------------------------------
* TESTING ZGESV
@@ -352,12 +227,12 @@ static int check_solution(int N, int NRHS, MORSE_Complex64_t *A1, int LDA, MORSE
alpha = 1.0;
beta = -1.0;
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B2, LDB, &Xnorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, N, A1, LDA, &Anorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B1, LDB, &Bnorm );
+ Xnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, NRHS, B2, LDB );
+ Anorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, N, A1, LDA );
+ Bnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, NRHS, B1, LDB );
cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, N, NRHS, N, CBLAS_SADDR(alpha), A1, LDA, B2, LDB, CBLAS_SADDR(beta), B1, LDB);
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B1, LDB, &Rnorm );
+ Rnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, NRHS, B1, LDB );
if (getenv("MORSE_TESTING_VERBOSE"))
printf( "||A||_oo=%f\n||X||_oo=%f\n||B||_oo=%f\n||A X - B||_oo=%e\n", Anorm, Xnorm, Bnorm, Rnorm );
diff --git a/testing/testing_zposv.c b/testing/testing_zposv.c
index 82e0d5e26ffa7eade5aacb343e587078c3a66d56..f2e1ead1759acce639fef1d56e95310263d542f0 100644
--- a/testing/testing_zposv.c
+++ b/testing/testing_zposv.c
@@ -33,131 +33,6 @@
#include <coreblas.h>
#include "testing_zauxiliary.h"
-enum blas_order_type {
- blas_rowmajor = 101,
- blas_colmajor = 102 };
-
-enum blas_cmach_type {
- blas_base = 151,
- blas_t = 152,
- blas_rnd = 153,
- blas_ieee = 154,
- blas_emin = 155,
- blas_emax = 156,
- blas_eps = 157,
- blas_prec = 158,
- blas_underflow = 159,
- blas_overflow = 160,
- blas_sfmin = 161};
-
-enum blas_norm_type {
- blas_one_norm = 171,
- blas_real_one_norm = 172,
- blas_two_norm = 173,
- blas_frobenius_norm = 174,
- blas_inf_norm = 175,
- blas_real_inf_norm = 176,
- blas_max_norm = 177,
- blas_real_max_norm = 178 };
-
-static void
-BLAS_error(char *rname, int err, int val, int x) {
- fprintf( stderr, "%s %d %d %d\n", rname, err, val, x );
- abort();
-}
-
-static
-void
-BLAS_zge_norm(enum blas_order_type order, enum blas_norm_type norm,
- int m, int n, const MORSE_Complex64_t *a, int lda, double *res) {
- int i, j; float anorm, v;
- char rname[] = "BLAS_zge_norm";
-
- if (order != blas_colmajor) BLAS_error( rname, -1, order, 0 );
-
- if (norm == blas_frobenius_norm) {
- anorm = 0.0f;
- for (j = n; j; --j) {
- for (i = m; i; --i) {
- v = a[0];
- anorm += v * v;
- a++;
- }
- a += lda - m;
- }
- anorm = sqrt( anorm );
- } else if (norm == blas_inf_norm) {
- anorm = 0.0f;
- for (i = 0; i < m; ++i) {
- v = 0.0f;
- for (j = 0; j < n; ++j) {
- v += cabs( a[i + j * lda] );
- }
- if (v > anorm)
- anorm = v;
- }
- } else {
- BLAS_error( rname, -2, norm, 0 );
- return;
- }
-
- if (res) *res = anorm;
-}
-
-static
-double
-BLAS_dpow_di(double x, int n) {
- double rv = 1.0;
-
- if (n < 0) {
- n = -n;
- x = 1.0 / x;
- }
-
- for (; n; n >>= 1, x *= x) {
- if (n & 1)
- rv *= x;
- }
-
- return rv;
-}
-
-static
-double
-BLAS_dfpinfo(enum blas_cmach_type cmach) {
- double eps = 1.0, r = 1.0, o = 1.0, b = 2.0;
- int t = 53, l = 1024, m = -1021;
- char rname[] = "BLAS_dfpinfo";
-
- if ((sizeof eps) == sizeof(float)) {
- t = 24;
- l = 128;
- m = -125;
- } else {
- t = 53;
- l = 1024;
- m = -1021;
- }
-
- /* for (i = 0; i < t; ++i) eps *= half; */
- eps = BLAS_dpow_di( b, -t );
- /* for (i = 0; i >= m; --i) r *= half; */
- r = BLAS_dpow_di( b, m-1 );
-
- o -= eps;
- /* for (i = 0; i < l; ++i) o *= b; */
- o = (o * BLAS_dpow_di( b, l-1 )) * b;
-
- switch (cmach) {
- case blas_eps: return eps;
- case blas_sfmin: return r;
- default:
- BLAS_error( rname, -1, cmach, 0 );
- break;
- }
- return 0.0;
-}
-
static int check_factorization(int, MORSE_Complex64_t*, MORSE_Complex64_t*, int, int , double);
static int check_solution(int, int, MORSE_Complex64_t*, int, MORSE_Complex64_t*, MORSE_Complex64_t*, int, double);
@@ -198,7 +73,7 @@ int testing_zposv(int argc, char **argv)
return -2;
}
- eps = BLAS_dfpinfo( blas_eps );
+ eps = LAPACKE_dlamch_work( 'e' );
uplo = MorseUpper;
trans1 = uplo == MorseUpper ? MorseConjTrans : MorseNoTrans;
@@ -371,8 +246,8 @@ static int check_factorization(int N, MORSE_Complex64_t *A1, MORSE_Complex64_t *
for (j = 0; j < N; j++)
Residual[j*N+i] = L2[j*N+i] - Residual[j*N+i];
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, N, Residual, N, &Rnorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, N, A1, LDA, &Anorm );
+ Rnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, N, Residual, N );
+ Anorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, N, A1, LDA );
printf("============\n");
printf("Checking the Cholesky Factorization \n");
@@ -407,12 +282,12 @@ static int check_solution(int N, int NRHS, MORSE_Complex64_t *A1, int LDA, MORSE
alpha = 1.0;
beta = -1.0;
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B2, LDB, &Xnorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, N, A1, LDA, &Anorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B1, LDB, &Bnorm );
+ Xnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, NRHS, B2, LDB );
+ Anorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, N, A1, LDA );
+ Bnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, NRHS, B1, LDB );
cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, N, NRHS, N, CBLAS_SADDR(alpha), A1, LDA, B2, LDB, CBLAS_SADDR(beta), B1, LDB);
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, NRHS, B1, LDB, &Rnorm );
+ Rnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'i', N, NRHS, B1, LDB );
if (getenv("MORSE_TESTING_VERBOSE"))
printf( "||A||_oo=%f\n||X||_oo=%f\n||B||_oo=%f\n||A X - B||_oo=%e\n", Anorm, Xnorm, Bnorm, Rnorm );
diff --git a/testing/testing_zpotri.c b/testing/testing_zpotri.c
index f16dccb77b569e3f9d6e8db2508dd685d0f3e526..379db6631896321a20e1dc95e47cdee1c8d003f6 100644
--- a/testing/testing_zpotri.c
+++ b/testing/testing_zpotri.c
@@ -33,141 +33,6 @@
#include <coreblas.h>
#include "testing_zauxiliary.h"
-enum blas_order_type {
- blas_rowmajor = 101,
- blas_colmajor = 102 };
-
-enum blas_cmach_type {
- blas_base = 151,
- blas_t = 152,
- blas_rnd = 153,
- blas_ieee = 154,
- blas_emin = 155,
- blas_emax = 156,
- blas_eps = 157,
- blas_prec = 158,
- blas_underflow = 159,
- blas_overflow = 160,
- blas_sfmin = 161};
-
-enum blas_norm_type {
- blas_one_norm = 171,
- blas_real_one_norm = 172,
- blas_two_norm = 173,
- blas_frobenius_norm = 174,
- blas_inf_norm = 175,
- blas_real_inf_norm = 176,
- blas_max_norm = 177,
- blas_real_max_norm = 178 };
-
-static void
-BLAS_error(char *rname, int err, int val, int x) {
- fprintf( stderr, "%s %d %d %d\n", rname, err, val, x );
- abort();
-}
-
-static
-void
-BLAS_zge_norm(enum blas_order_type order, enum blas_norm_type norm,
- int m, int n, const MORSE_Complex64_t *a, int lda, double *res) {
- int i, j; float anorm, v;
- char rname[] = "BLAS_zge_norm";
-
- if (order != blas_colmajor) BLAS_error( rname, -1, order, 0 );
-
- if (norm == blas_frobenius_norm) {
- anorm = 0.0f;
- for (j = n; j; --j) {
- for (i = m; i; --i) {
- v = a[0];
- anorm += v * v;
- a++;
- }
- a += lda - m;
- }
- anorm = sqrt( anorm );
- } else if (norm == blas_inf_norm) {
- anorm = 0.0f;
- for (i = 0; i < m; ++i) {
- v = 0.0f;
- for (j = 0; j < n; ++j) {
- v += cabs( a[i + j * lda] );
- }
- if (v > anorm)
- anorm = v;
- }
- } else if (norm == blas_one_norm) {
- anorm = 0.0f;
- for (i = 0; i < m; ++i) {
- v = 0.0f;
- for (j = 0; j < n; ++j) {
- v += cabs( a[i + j * lda] );
- }
- if (v > anorm)
- anorm = v;
- }
- } else {
- BLAS_error( rname, -2, norm, 0 );
- return;
- }
-
- if (res) *res = anorm;
-}
-
-static
-double
-BLAS_dpow_di(double x, int n) {
- double rv = 1.0;
-
- if (n < 0) {
- n = -n;
- x = 1.0 / x;
- }
-
- for (; n; n >>= 1, x *= x) {
- if (n & 1)
- rv *= x;
- }
-
- return rv;
-}
-
-static
-double
-BLAS_dfpinfo(enum blas_cmach_type cmach) {
- double eps = 1.0, r = 1.0, o = 1.0, b = 2.0;
- int t = 53, l = 1024, m = -1021;
- char rname[] = "BLAS_dfpinfo";
-
- if ((sizeof eps) == sizeof(float)) {
- t = 24;
- l = 128;
- m = -125;
- } else {
- t = 53;
- l = 1024;
- m = -1021;
- }
-
- /* for (i = 0; i < t; ++i) eps *= half; */
- eps = BLAS_dpow_di( b, -t );
- /* for (i = 0; i >= m; --i) r *= half; */
- r = BLAS_dpow_di( b, m-1 );
-
- o -= eps;
- /* for (i = 0; i < l; ++i) o *= b; */
- o = (o * BLAS_dpow_di( b, l-1 )) * b;
-
- switch (cmach) {
- case blas_eps: return eps;
- case blas_sfmin: return r;
- default:
- BLAS_error( rname, -1, cmach, 0 );
- break;
- }
- return 0.0;
-}
-
static int check_factorization(int, MORSE_Complex64_t*, MORSE_Complex64_t*, int, int, double);
static int check_inverse(int, MORSE_Complex64_t *, MORSE_Complex64_t *, int, int, double);
@@ -202,13 +67,13 @@ int testing_zpotri(int argc, char **argv)
return -2;
}
- eps = BLAS_dfpinfo( blas_eps );
+ eps = LAPACKE_dlamch_work( 'e' );
uplo = MorseUpper;
/*-------------------------------------------------------------
- * TESTING ZPOTRI
- */
+ * TESTING ZPOTRI
+ */
/* Initialize A1 and A2 for Symmetric Positif Matrix */
MORSE_zplghe( (double)N, MorseUpperLower, N, A1, LDA, 51 );
@@ -289,10 +154,10 @@ static int check_factorization(int N, MORSE_Complex64_t *A1, MORSE_Complex64_t *
/* Compute the Residual || A -L'L|| */
for (i = 0; i < N; i++)
for (j = 0; j < N; j++)
- Residual[j*N+i] = L2[j*N+i] - Residual[j*N+i];
+ Residual[j*N+i] = L2[j*N+i] - Residual[j*N+i];
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, N, Residual, N, &Rnorm );
- BLAS_zge_norm( blas_colmajor, blas_inf_norm, N, N, A1, LDA, &Anorm );
+ Rnorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', N, N, Residual, N, work );
+ Anorm = LAPACKE_zlange_work( LAPACK_COL_MAJOR, 'i', N, N, A1, LDA, work );
printf("============\n");
printf("Checking the Cholesky Factorization \n");
@@ -331,29 +196,34 @@ static int check_inverse(int N, MORSE_Complex64_t *A1, MORSE_Complex64_t *A2, in
/* Rebuild the other part of the inverse matrix */
if(uplo == MorseUpper){
- for(j=0; j<N; j++)
- for(i=0; i<j; i++)
- *(A2+j+i*LDA) = *(A2+i+j*LDA);
- cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, N, N, CBLAS_SADDR(alpha), A2, LDA, A1, LDA, CBLAS_SADDR(beta), work, N);
+ for(j=0; j<N; j++) {
+ for(i=0; i<j; i++) {
+ *(A2+j+i*LDA) = *(A2+i+j*LDA);
+ }
+ }
+ cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, N, N, CBLAS_SADDR(alpha), A2, LDA, A1, LDA, CBLAS_SADDR(beta), work, N);
}
else {
- for(j=0; j<N; j++)
- for(i=j; i<N; i++)
- *(A2+j+i*LDA) = *(A2+i+j*LDA);
- cblas_zhemm(CblasColMajor, CblasLeft, CblasLower, N, N, CBLAS_SADDR(alpha), A2, LDA, A1, LDA, CBLAS_SADDR(beta), work, N);
+ for(j=0; j<N; j++) {
+ for(i=j; i<N; i++) {
+ *(A2+j+i*LDA) = *(A2+i+j*LDA);
+ }
+ }
+ cblas_zhemm(CblasColMajor, CblasLeft, CblasLower, N, N, CBLAS_SADDR(alpha), A2, LDA, A1, LDA, CBLAS_SADDR(beta), work, N);
}
/* Add the identity matrix to work */
for(i=0; i<N; i++)
*(work+i+i*N) = *(work+i+i*N) + zone;
- BLAS_zge_norm( blas_colmajor, blas_one_norm, N, N, work, N, &Rnorm );
- BLAS_zge_norm( blas_colmajor, blas_one_norm, N, N, A1, LDA, &Anorm );
- BLAS_zge_norm( blas_colmajor, blas_one_norm, N, N, A2, LDA, &Ainvnorm );
+ Rnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'o', N, N, work, N );
+ Anorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'o', N, N, A1, LDA );
+ Ainvnorm = LAPACKE_zlange( LAPACK_COL_MAJOR, 'o', N, N, A2, LDA );
- if (getenv("MORSE_TESTING_VERBOSE"))
- printf( "||A||_1=%f\n||Ainv||_1=%f\n||Id - A*Ainv||_1=%e\n", Anorm, Ainvnorm, Rnorm );
+ if (getenv("MORSE_TESTING_VERBOSE")) {
+ printf( "||A||_1=%f\n||Ainv||_1=%f\n||Id - A*Ainv||_1=%e\n", Anorm, Ainvnorm, Rnorm );
+ }
result = Rnorm / ( (Anorm*Ainvnorm)*N*eps ) ;
printf("============\n");
@@ -363,7 +233,7 @@ static int check_inverse(int N, MORSE_Complex64_t *A1, MORSE_Complex64_t *A2, in
if ( isnan(Ainvnorm) || isinf(Ainvnorm) || isnan(result) || isinf(result) || (result > 60.0) ) {
printf("-- The inverse is suspicious ! \n");
info_inverse = 1;
- }
+ }
else{
printf("-- The inverse is CORRECT ! \n");
info_inverse = 0;