diff --git a/compute/pzcesca.c b/compute/pzcesca.c
index d09505c2ff68796825105f20093e9f2d4d783a15..ff464ad222563abee098b108bf4a68bb6df83161 100644
--- a/compute/pzcesca.c
+++ b/compute/pzcesca.c
@@ -44,9 +44,9 @@ chameleon_pzcesca_internal( int center,
* 1) compute sums and sum-square (scl,ssq) in each tile
*/
for(n = 0; n < NT; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
for(m = 0; m < MT; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
if ( (center == 1) && ( (axis == ChamColumnwise) || (axis == ChamEltwise) ) ) {
INSERT_TASK_zgesum(
options, ChamColumnwise, tempmm, tempnn,
@@ -71,7 +71,7 @@ chameleon_pzcesca_internal( int center,
}
for(n = 0; n < NT; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( (center == 1) && ( (axis == ChamColumnwise) || (axis == ChamEltwise) ) ) {
/**
@@ -126,7 +126,7 @@ chameleon_pzcesca_internal( int center,
}
for(m = 0; m < MT; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
if ( (center == 1) && ( (axis == ChamRowwise) || (axis == ChamEltwise) ) ) {
/**
@@ -193,10 +193,10 @@ chameleon_pzcesca_internal( int center,
/* Finally compute Centered-Scaled matrix coefficients inplace */
for(n = 0; n < NT; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
for(m = 0; m < MT; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
INSERT_TASK_zcesca(
options,
@@ -234,9 +234,9 @@ void chameleon_pzcesca( struct chameleon_pzcesca_s *ws, int center, int scale, c
/* Initialize Wgcol */
for(m = 0; m < Wgcol->mt; m++) {
- tempmm = m == Wgcol->mt-1 ? Wgcol->m-m*Wgcol->mb : Wgcol->mb;
+ tempmm = Wgcol->get_blkdim( Wgcol, m, DIM_m, Wgcol->m );
for(n = 0; n < Wgcol->nt; n++) {
- tempnn = n == Wgcol->nt-1 ? Wgcol->n-n*Wgcol->nb : Wgcol->nb;
+ tempnn = Wgcol->get_blkdim( Wgcol, n, DIM_n, Wgcol->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
@@ -246,9 +246,9 @@ void chameleon_pzcesca( struct chameleon_pzcesca_s *ws, int center, int scale, c
}
/* Initialize Wgrow */
for(m = 0; m < Wgrow->mt; m++) {
- tempmm = m == Wgrow->mt-1 ? Wgrow->m-m*Wgrow->mb : Wgrow->mb;
+ tempmm = Wgrow->get_blkdim( Wgrow, m, DIM_m, Wgrow->m );
for(n = 0; n < Wgrow->nt; n++) {
- tempnn = n == Wgrow->nt-1 ? Wgrow->n-n*Wgrow->nb : Wgrow->nb;
+ tempnn = Wgrow->get_blkdim( Wgrow, n, DIM_n, Wgrow->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
@@ -258,9 +258,9 @@ void chameleon_pzcesca( struct chameleon_pzcesca_s *ws, int center, int scale, c
}
/* Initialize Wgelt */
for(m = 0; m < Wgelt->mt; m++) {
- tempmm = m == Wgelt->mt-1 ? Wgelt->m-m*Wgelt->mb : Wgelt->mb;
+ tempmm = Wgelt->get_blkdim( Wgelt, m, DIM_m, Wgelt->m );
for(n = 0; n < Wgelt->nt; n++) {
- tempnn = n == Wgelt->nt-1 ? Wgelt->n-n*Wgelt->nb : Wgelt->nb;
+ tempnn = Wgelt->get_blkdim( Wgelt, n, DIM_n, Wgelt->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
@@ -270,9 +270,9 @@ void chameleon_pzcesca( struct chameleon_pzcesca_s *ws, int center, int scale, c
}
/* Initialize Wdcol */
for(m = 0; m < Wdcol->mt; m++) {
- tempmm = m == Wdcol->mt-1 ? Wdcol->m-m*Wdcol->mb : Wdcol->mb;
+ tempmm = Wdcol->get_blkdim( Wdcol, m, DIM_m, Wdcol->m );
for(n = 0; n < Wdcol->nt; n++) {
- tempnn = n == Wdcol->nt-1 ? Wdcol->n-n*Wdcol->nb : Wdcol->nb;
+ tempnn = Wdcol->get_blkdim( Wdcol, n, DIM_n, Wdcol->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
@@ -282,9 +282,9 @@ void chameleon_pzcesca( struct chameleon_pzcesca_s *ws, int center, int scale, c
}
/* Initialize Wdrow */
for(m = 0; m < Wdrow->mt; m++) {
- tempmm = m == Wdrow->mt-1 ? Wdrow->m-m*Wdrow->mb : Wdrow->mb;
+ tempmm = Wdrow->get_blkdim( Wdrow, m, DIM_m, Wdrow->m );
for(n = 0; n < Wdrow->nt; n++) {
- tempnn = n == Wdrow->nt-1 ? Wdrow->n-n*Wdrow->nb : Wdrow->nb;
+ tempnn = Wdrow->get_blkdim( Wdrow, n, DIM_n, Wdrow->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
diff --git a/compute/pzgelqf.c b/compute/pzgelqf.c
index 19fec4446eda909e39cdfca22dc4d9993750ddac..0ef55f468f4205c6c25768aa5fc3460cb39ff76f 100644
--- a/compute/pzgelqf.c
+++ b/compute/pzgelqf.c
@@ -46,8 +46,8 @@ int chameleon_pzgelqf_step( int genD, int k, int ib,
int m, n;
int tempkm, tempkn, tempmm, tempnn;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgelqt(
options,
tempkm, tempkn, ib, T->nb,
@@ -55,8 +55,8 @@ int chameleon_pzgelqf_step( int genD, int k, int ib,
T(k, k));
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
options,
ChamUpper, tempDkm, tempDkn,
@@ -72,7 +72,7 @@ int chameleon_pzgelqf_step( int genD, int k, int ib,
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zunmlq(
options,
ChamRight, ChamConjTrans,
@@ -85,7 +85,7 @@ int chameleon_pzgelqf_step( int genD, int k, int ib,
RUNTIME_data_flush( sequence, T(k, k) );
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
RUNTIME_data_migrate( sequence, A(k, k),
A->get_rankof( A, k, n ) );
@@ -98,7 +98,7 @@ int chameleon_pzgelqf_step( int genD, int k, int ib,
A(k, n),
T(k, n));
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
RUNTIME_data_migrate( sequence, A(m, k),
A->get_rankof( A, m, n ) );
diff --git a/compute/pzgelqf_param.c b/compute/pzgelqf_param.c
index 60921a28b040d02b77b8431e097d8c51544bba35..e97d99f1ca3af1e70488b0bb9d3bddb1cdaee007 100644
--- a/compute/pzgelqf_param.c
+++ b/compute/pzgelqf_param.c
@@ -49,7 +49,7 @@ int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
int tempkmin, tempkm, tempnn, tempmm, temppn;
int node, nbtiles;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
/* The number of geqrt to apply */
nbgelqt = qrtree->getnbgeqrf( qrtree, k );
@@ -63,7 +63,7 @@ int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
continue;
}
- temppn = p == A->nt-1 ? A->n-p*A->nb : A->nb;
+ temppn = A->get_blkdim( A, p, DIM_n, A->n );
tempkmin = chameleon_min(tempkm, temppn);
INSERT_TASK_zgelqt(
@@ -72,8 +72,8 @@ int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
A(k, p), T(k, p));
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
options,
@@ -89,7 +89,7 @@ int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zunmlq(
options,
ChamRight, ChamConjTrans,
@@ -112,7 +112,7 @@ int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
n = tiles[i];
p = qrtree->currpiv( qrtree, k, n );
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
if ( qrtree->gettype( qrtree, k, n ) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -142,7 +142,7 @@ int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
T(k, n));
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
node = A->get_rankof( A, m, n );
RUNTIME_data_migrate( sequence, A(m, p), node );
diff --git a/compute/pzgelqfrh.c b/compute/pzgelqfrh.c
index 09d9bab3e0c459232e2b2dbf5f369a30ee2f8cda..ae9afdf58647e23f32b166468b88a93a19ce914c 100644
--- a/compute/pzgelqfrh.c
+++ b/compute/pzgelqfrh.c
@@ -92,10 +92,10 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
for (k = 0; k < K; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for (N = k; N < A->nt; N += BS) {
- tempNn = N == A->nt-1 ? A->n-N*A->nb : A->nb;
+ tempNn = A->get_blkdim( A, N, DIM_n, A->n );
tempkmin = chameleon_min(tempkm, tempNn);
INSERT_TASK_zgelqt(
&options,
@@ -103,8 +103,8 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
A(k, N),
T(k, N));
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
- int tempDNn = N == D->nt-1 ? D->n-N*D->nb : D->nb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
+ int tempDNn = D->get_blkdim( D, N, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -120,7 +120,7 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
#endif
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zunmlq(
&options,
ChamRight, ChamConjTrans,
@@ -133,7 +133,7 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
RUNTIME_data_flush( sequence, T(k, N) );
for (n = N+1; n < chameleon_min(N+BS, A->nt); n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
RUNTIME_data_migrate( sequence, A(k, N),
A->get_rankof( A, k, n ) );
@@ -147,7 +147,7 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
T(k, n));
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
RUNTIME_data_migrate( sequence, A(m, N),
A->get_rankof( A, m, n ) );
@@ -167,7 +167,7 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
}
for (RD = BS; RD < A->nt-k; RD *= 2) {
for (N = k; N+RD < A->nt; N += 2*RD) {
- tempNRDn = N+RD == A->nt-1 ? A->n-(N+RD)*A->nb : A->nb;
+ tempNRDn = A->get_blkdim( A, N+RD, DIM_n, A->n );
node = A->get_rankof( A, k, N+RD );
RUNTIME_data_migrate( sequence, A(k, N), node );
@@ -182,7 +182,7 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
T2(k, N+RD));
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
node = A->get_rankof( A, m, N+RD );
RUNTIME_data_migrate( sequence, A(m, N), node );
diff --git a/compute/pzgemm.c b/compute/pzgemm.c
index a4a12eca743aa1e9ef69a15ddfec7bc1564edc92..1c8b05aaf43de10bb0b734a48e9e352887ab7434 100644
--- a/compute/pzgemm.c
+++ b/compute/pzgemm.c
@@ -84,9 +84,9 @@ chameleon_pzgemm_Astat( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
/* Scale C */
options->forcesub = 0;
@@ -100,7 +100,7 @@ chameleon_pzgemm_Astat( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
if (transA == ChamNoTrans) {
if (transB == ChamNoTrans) {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgemm_Astat(
options,
@@ -116,7 +116,7 @@ chameleon_pzgemm_Astat( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
*/
else {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgemm_Astat(
options,
@@ -134,7 +134,7 @@ chameleon_pzgemm_Astat( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
else {
if (transB == ChamNoTrans) {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
INSERT_TASK_zgemm_Astat(
options,
@@ -150,7 +150,7 @@ chameleon_pzgemm_Astat( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
INSERT_TASK_zgemm_Astat(
options,
@@ -186,17 +186,25 @@ chameleon_pzgemm_summa( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
RUNTIME_sequence_t *sequence = options->sequence;
int m, n, k, p, q, KT, K, lp, lq;
int tempmm, tempnn, tempkk;
- int lookahead, myp, myq;
+ int lookahead, myp, myq, DIM_k;
CHAMELEON_Complex64_t zbeta;
CHAMELEON_Complex64_t zone = (CHAMELEON_Complex64_t)1.0;
lookahead = chamctxt->lookahead;
- KT = transA == ChamNoTrans ? A->nt : A->mt;
- K = transA == ChamNoTrans ? A->n : A->m;
- myp = C->myrank / chameleon_desc_datadist_get_iparam(C, 1);
- myq = C->myrank % chameleon_desc_datadist_get_iparam(C, 1);
+ if ( transA == ChamNoTrans ) {
+ KT = A->nt;
+ K = A->n;
+ DIM_k = DIM_n;
+ }
+ else {
+ KT = A->mt;
+ K = A->m;
+ DIM_k = DIM_m;
+ }
+ myp = C->myrank / chameleon_desc_datadist_get_iparam(C, 1);
+ myq = C->myrank % chameleon_desc_datadist_get_iparam(C, 1);
/*
* A: ChamNoTrans / B: ChamNoTrans
@@ -204,12 +212,13 @@ chameleon_pzgemm_summa( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
for (k = 0; k < KT; k++ ) {
lp = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 0);
lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 1);
- tempkk = k == KT - 1 ? K - k * A->nb : A->nb;
+
+ tempkk = A->get_blkdim( A, k, DIM_k, K );
zbeta = k == 0 ? beta : zone;
/* Transfert ownership of the k column of A */
for (m = 0; m < C->mt; m ++ ) {
- tempmm = m == C->mt-1 ? C->m - m * C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if ( transA == ChamNoTrans ) {
INSERT_TASK_zlacpy(
@@ -249,7 +258,7 @@ chameleon_pzgemm_summa( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
/* Transfert ownership of the k row of B */
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if ( transB == ChamNoTrans ) {
INSERT_TASK_zlacpy(
@@ -288,10 +297,10 @@ chameleon_pzgemm_summa( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tran
}
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zgemm(
options,
@@ -327,16 +336,16 @@ chameleon_pzgemm_generic( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tr
CHAMELEON_Complex64_t zone = (CHAMELEON_Complex64_t)1.0;
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
/*
* A: ChamNoTrans / B: ChamNoTrans
*/
if (transA == ChamNoTrans) {
if (transB == ChamNoTrans) {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
options,
@@ -352,7 +361,7 @@ chameleon_pzgemm_generic( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tr
*/
else {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
options,
@@ -370,7 +379,7 @@ chameleon_pzgemm_generic( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tr
else {
if (transB == ChamNoTrans) {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
options,
@@ -386,7 +395,7 @@ chameleon_pzgemm_generic( CHAM_context_t *chamctxt, cham_trans_t transA, cham_tr
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
options,
diff --git a/compute/pzgenm2.c b/compute/pzgenm2.c
index 99455940aae73822a1e7dfa937873b30715dce9d..922559a5824a81525e4d2b86dc5e05ff3b531ab2 100644
--- a/compute/pzgenm2.c
+++ b/compute/pzgenm2.c
@@ -92,7 +92,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
*
*/
for(n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1)) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
/* Zeroes the local intermediate vector */
INSERT_TASK_dlaset(
@@ -103,7 +103,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
/* Computes the sums of the local tiles into the local vector */
for(m = myp; m < A->mt; m += chameleon_desc_datadist_get_iparam(A, 0)) {
- tempmm = m == A->mt-1 ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_dzasum(
&options,
ChamColumnwise, ChamUpperLower, tempmm, tempnn,
@@ -131,7 +131,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
NRMX( myp, myq ) );
for( n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1) ) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_dgessq(
&options, ChamEltwise, 1, tempnn,
DROW( myp, n ),
@@ -207,7 +207,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
if ( cnt == 0 )
{
for (n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1)) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
if ( myp == 0 ) {
#if defined(PRECISION_z) || defined(PRECISION_c)
@@ -245,7 +245,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
*/
scl = 1. / e0;
for (n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1)) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlascal(
&options,
@@ -257,10 +257,10 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
* Compute Sx = S * x
*/
for(m = myp; m < A->mt; m+=chameleon_desc_datadist_get_iparam(A, 0)) {
- tempmm = m == A->mt-1 ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1) ) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
beta = n == myq ? 0. : 1.;
INSERT_TASK_zgemv(
@@ -292,10 +292,10 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
* Compute x = S' * S * x = S' * Sx
*/
for ( n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1) ) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
for( m = myp; m < A->mt; m += chameleon_desc_datadist_get_iparam(A, 0) ) {
- tempmm = m == A->mt-1 ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
beta = m == myp ? 0. : 1.;
INSERT_TASK_zgemv(
@@ -336,7 +336,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
NRMX( myp, myq ) );
for( n = myq; n < A->nt; n += chameleon_desc_datadist_get_iparam(A, 1) ) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zgessq(
&options, ChamEltwise, 1, tempnn,
@@ -378,7 +378,7 @@ chameleon_pzgenm2( double tol, const CHAM_desc_t *A, double *result,
NRMSX( myp, myq ) );
for( m = myp; m < A->mt; m += chameleon_desc_datadist_get_iparam(A, 0) ) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgessq(
&options, ChamEltwise, tempmm, 1,
SX( m, myq ),
diff --git a/compute/pzgeqrf.c b/compute/pzgeqrf.c
index 450f3fb968d8a759e84474ffde3b8c6774c96147..a827177fa9074d42e0487eaad856e545877ba0d6 100644
--- a/compute/pzgeqrf.c
+++ b/compute/pzgeqrf.c
@@ -45,8 +45,8 @@ int chameleon_pzgeqrf_step( int genD, int k, int ib,
int m, n;
int tempkm, tempkn, tempnn, tempmm;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgeqrt(
options,
@@ -55,8 +55,8 @@ int chameleon_pzgeqrf_step( int genD, int k, int ib,
T(k, k));
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
options,
ChamLower, tempDkm, tempDkn,
@@ -71,7 +71,7 @@ int chameleon_pzgeqrf_step( int genD, int k, int ib,
#endif
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zunmqr(
options,
ChamLeft, ChamConjTrans,
@@ -84,7 +84,7 @@ int chameleon_pzgeqrf_step( int genD, int k, int ib,
RUNTIME_data_flush( sequence, T(k, k) );
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
RUNTIME_data_migrate( sequence, A(k, k),
A->get_rankof( A, m, k ) );
@@ -98,7 +98,7 @@ int chameleon_pzgeqrf_step( int genD, int k, int ib,
T(m, k));
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
RUNTIME_data_migrate( sequence, A(k, n),
A->get_rankof( A, m, n ) );
diff --git a/compute/pzgeqrf_param.c b/compute/pzgeqrf_param.c
index ad75e652eb79cd26b01d1f4ae7ed1253140f8e23..f8c9020ffc9b461d94f199bbf77079a87635df1b 100644
--- a/compute/pzgeqrf_param.c
+++ b/compute/pzgeqrf_param.c
@@ -50,7 +50,7 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
int tempkmin, tempkn, tempnn, tempmm, temppm;
int node, nbtiles;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
/* The number of geqrt to apply */
nbgeqrt = qrtree->getnbgeqrf( qrtree, k );
@@ -64,7 +64,7 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
continue;
}
- temppm = p == A->mt-1 ? A->m-p*A->mb : A->mb;
+ temppm = A->get_blkdim( A, p, DIM_m, A->m );
tempkmin = chameleon_min(temppm, tempkn);
INSERT_TASK_zgeqrt(
@@ -73,8 +73,8 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
A(p, k), T(p, k) );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
options,
@@ -90,7 +90,7 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zunmqr(
options,
ChamLeft, ChamConjTrans,
@@ -113,7 +113,7 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
m = tiles[i];
p = qrtree->currpiv( qrtree, k, m );
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
if ( qrtree->gettype( qrtree, k, m ) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -143,7 +143,7 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
T(m, k));
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
node = A->get_rankof( A, m, n );
RUNTIME_data_migrate( sequence, A(p, n), node );
diff --git a/compute/pzgeqrfrh.c b/compute/pzgeqrfrh.c
index 546f802d9c8161bce7734a1ce50bb612ed4a5970..83f93fc3d8e3ef46ec65a8b8bc6eebf495a1d929 100644
--- a/compute/pzgeqrfrh.c
+++ b/compute/pzgeqrfrh.c
@@ -89,9 +89,9 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
for (k = 0; k < K; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (M = k; M < A->mt; M += BS) {
- tempMm = M == A->mt-1 ? A->m-M*A->mb : A->mb;
+ tempMm = A->get_blkdim( A, M, DIM_m, A->m );
tempkmin = chameleon_min(tempMm, tempkn);
INSERT_TASK_zgeqrt(
@@ -100,8 +100,8 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
A(M, k),
T(M, k));
if ( genD ) {
- int tempDMm = M == D->mt-1 ? D->m-M*D->mb : D->mb;
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDMm = D->get_blkdim( D, M, DIM_m, D->m );
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -117,7 +117,7 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
#endif
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zunmqr(
&options,
ChamLeft, ChamConjTrans,
@@ -130,7 +130,7 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
RUNTIME_data_flush( sequence, T(M, k) );
for (m = M+1; m < chameleon_min(M+BS, A->mt); m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
RUNTIME_data_migrate( sequence, A(M, k),
A->get_rankof( A, m, k ) );
@@ -144,7 +144,7 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
T(m, k));
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
RUNTIME_data_migrate( sequence, A(M, n),
A->get_rankof( A, m, n ) );
@@ -164,7 +164,7 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
}
for (RD = BS; RD < A->mt-k; RD *= 2) {
for (M = k; M+RD < A->mt; M += 2*RD) {
- tempMRDm = M+RD == A->mt-1 ? A->m-(M+RD)*A->mb : A->mb;
+ tempMRDm = A->get_blkdim( A, M+RD, DIM_m, A->m );
node = A->get_rankof( A, M+RD, k );
RUNTIME_data_migrate( sequence, A(M, k), node );
@@ -179,7 +179,7 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
T2(M+RD, k));
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
node = A->get_rankof( A, M+RD, n );
RUNTIME_data_migrate( sequence, A(M, n), node );
diff --git a/compute/pzgered.c b/compute/pzgered.c
index dd2c4a597c063ae33f468c529a6302808e3f7356..a0c38cb8ea320a90863842b8c92639b7e3cfe0be 100644
--- a/compute/pzgered.c
+++ b/compute/pzgered.c
@@ -80,10 +80,10 @@ chameleon_pzgered_frb( cham_uplo_t uplo,
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( (n == m) && (uplo != ChamUpperLower) ) {
INSERT_TASK_ztrssq(
@@ -235,13 +235,13 @@ void chameleon_pzgered( cham_uplo_t uplo,
for(m = 0; m < A->mt; m++)
{
- int tempmm = ( m == (A->mt-1) ) ? A->m - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, A->m );
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, A->nt) : A->nt;
for(n = nmin; n < nmax; n++)
{
- int tempnn = ( n == (A->nt-1) ) ? A->n - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, A->n );
/*
* u_{high} = 1e-16 (later should be application accuracy)
diff --git a/compute/pzgerst.c b/compute/pzgerst.c
index 52c801cbe0945bb3e3249b602c97e92fe6c310d5..df031662ffcfd7397f5d9f4a49dfdbc4aef6f1fc 100644
--- a/compute/pzgerst.c
+++ b/compute/pzgerst.c
@@ -36,7 +36,7 @@ void chameleon_pzgerst( cham_uplo_t uplo,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for(m = 0; m < A->mt; m++) {
- int tempmm = ( m == (A->mt-1) ) ? A->m - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, A->m );
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, A->nt) : A->nt;
@@ -46,7 +46,7 @@ void chameleon_pzgerst( cham_uplo_t uplo,
if (( tile->rank == A->myrank ) &&
( tile->flttype != ChamComplexDouble ) )
{
- int tempnn = ( n == (A->nt-1) ) ? A->n - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zgerst( &options,
tempmm, tempnn, A( m, n ) );
diff --git a/compute/pzgetrf.c b/compute/pzgetrf.c
index 8c41f81d550fa77e1c5bb0fd6d6dc59af3b0dea4..6db9a8a40148a67fcaeda74f9a718c949b12e59d 100644
--- a/compute/pzgetrf.c
+++ b/compute/pzgetrf.c
@@ -58,8 +58,8 @@ chameleon_pzgetrf_panel_facto_nopiv( struct chameleon_pzgetrf_s *ws,
const CHAMELEON_Complex64_t zone = (CHAMELEON_Complex64_t) 1.0;
int m, tempkm, tempkn, tempmm;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
/*
* Algorithm per block without pivoting
@@ -70,7 +70,7 @@ chameleon_pzgetrf_panel_facto_nopiv( struct chameleon_pzgetrf_s *ws,
A(k, k), 0);
for (m = k+1; m < A->mt; m++) {
- tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztrsm(
options,
ChamRight, ChamUpper, ChamNoTrans, ChamNonUnit,
@@ -90,8 +90,8 @@ chameleon_pzgetrf_panel_facto_nopiv_percol( struct chameleon_pzgetrf_s *ws,
int m, h;
int tempkm, tempkn, tempmm, minmn;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/*
@@ -103,7 +103,7 @@ chameleon_pzgetrf_panel_facto_nopiv_percol( struct chameleon_pzgetrf_s *ws,
A( k, k ), U( k, k ), A->mb * k );
for (m = k+1; m < A->mt; m++) {
- tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgetrf_nopiv_percol_trsm(
options, tempmm, tempkn, h,
A( m, k ), U( k, k ) );
@@ -123,8 +123,8 @@ chameleon_pzgetrf_panel_facto_percol( struct chameleon_pzgetrf_s *ws,
int m, h;
int tempkm, tempkn, tempmm, minmn;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/* Update the number of column */
@@ -141,7 +141,7 @@ chameleon_pzgetrf_panel_facto_percol( struct chameleon_pzgetrf_s *ws,
ipiv );
for (m = k+1; m < A->mt; m++) {
- tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgetrf_percol_offdiag(
options,
tempmm, tempkn, h, m * A->mb,
@@ -173,8 +173,8 @@ chameleon_pzgetrf_panel_facto_percol_batched( struct chameleon_pzgetrf_s *ws,
void **clargs = malloc( sizeof(char *) );
memset( clargs, 0, sizeof(char *) );
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/* Update the number of column */
@@ -190,7 +190,7 @@ chameleon_pzgetrf_panel_facto_percol_batched( struct chameleon_pzgetrf_s *ws,
INSERT_TASK_zgetrf_percol_diag( options, tempkm, tempkn, h, k * A->mb, A(k, k), ipiv );
for ( m = k+1; m < A->mt; m++ ) {
- tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgetrf_panel_offdiag_batched( options, tempmm, tempkn, h, m * A->mb,
(void *)ws, A(m, k), clargs, ipiv );
}
@@ -216,8 +216,8 @@ chameleon_pzgetrf_panel_facto_blocked( struct chameleon_pzgetrf_s *ws,
int m, h, b, nbblock;
int tempkm, tempkn, tempmm, minmn;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/* Update the number of column */
@@ -240,7 +240,7 @@ chameleon_pzgetrf_panel_facto_blocked( struct chameleon_pzgetrf_s *ws,
ipiv );
for (m = k+1; m < A->mt; m++) {
- tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgetrf_blocked_offdiag(
options,
tempmm, tempkn, j, m * A->mb, ws->ib,
@@ -283,8 +283,8 @@ chameleon_pzgetrf_panel_facto_blocked_batched( struct chameleon_pzgetrf_s *ws,
void **clargs = malloc( sizeof(char *) );
memset( clargs, 0, sizeof(char *) );
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/* Update the number of column */
@@ -303,7 +303,7 @@ chameleon_pzgetrf_panel_facto_blocked_batched( struct chameleon_pzgetrf_s *ws,
j = h + b * ws->ib;
for ( m = k; m < A->mt; m++ ) {
- tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgetrf_panel_blocked_batched( options, tempmm, tempkn, j, m * A->mb,
(void *)ws, A(m, k), Up(k, k), clargs, ipiv );
}
@@ -405,9 +405,9 @@ chameleon_pzgetrf_panel_permute( struct chameleon_pzgetrf_s *ws,
return;
}
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/* Extract selected rows into U */
@@ -474,9 +474,9 @@ chameleon_pzgetrf_panel_permute_batched( struct chameleon_pzgetrf_s *ws,
void **clargs = malloc( sizeof(char *) );
*clargs = NULL;
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
minmn = chameleon_min( tempkm, tempkn );
/* Extract selected rows into U */
@@ -523,13 +523,13 @@ chameleon_pzgetrf_panel_update_ws( struct chameleon_pzgetrf_s *ws,
int lq = (k % lookahead) * Q;
int myp = A->myrank / Q;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
if ( k >= ws->ringswitch ) {
for ( m = k+1; m < A->mt; m++ ) {
if ( ( m % P ) != myp ) continue;
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zlacpy(
options,
ChamUpperLower, tempmm, tempkn,
@@ -550,7 +550,7 @@ chameleon_pzgetrf_panel_update_ws( struct chameleon_pzgetrf_s *ws,
for ( m = k+1; m < A->mt; m++ ) {
if ( ( m % P ) != myp ) continue;
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for ( q = 0; q < Q; q++ ) {
INSERT_TASK_zlacpy(
options,
@@ -581,8 +581,8 @@ chameleon_pzgetrf_panel_update( struct chameleon_pzgetrf_s *ws,
int myq = A->myrank % chameleon_desc_datadist_get_iparam(A, 1);
int lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(A, 1);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
if ( ws->batch_size > 0 ) {
chameleon_pzgetrf_panel_permute_batched( ws, A, ipiv, k, n, options );
@@ -612,7 +612,7 @@ chameleon_pzgetrf_panel_update( struct chameleon_pzgetrf_s *ws,
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
rankAmn = A->get_rankof( A, m, n );
if ( A->myrank == rankAmn ) {
@@ -700,8 +700,8 @@ void chameleon_pzgetrf( struct chameleon_pzgetrf_s *ws,
{
chameleon_pzgetrf_panel_permute_batched( ws, A, IPIV, k, n, &options );
if ( A->myrank == chameleon_getrankof_2d( A, k, n ) ) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlacpy( &options, ChamUpperLower, tempkm, tempnn,
Wu(A->myrank, n), A(k, n) );
RUNTIME_data_flush( sequence, A(k, n) );
@@ -720,8 +720,8 @@ void chameleon_pzgetrf( struct chameleon_pzgetrf_s *ws,
{
chameleon_pzgetrf_panel_permute( ws, A, IPIV, k, n, &options );
if ( A->myrank == chameleon_getrankof_2d( A, k, n ) ) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlacpy( &options, ChamUpperLower, tempkm, tempnn,
Wu(A->myrank, n), A(k, n) );
RUNTIME_data_flush( sequence, A(k, n) );
diff --git a/compute/pzgetrf_incpiv.c b/compute/pzgetrf_incpiv.c
index 090bb03fd611de7c15689511190b9a6fba8f654f..932a0f6678b43bade54ca8e920dce93207b4371b 100644
--- a/compute/pzgetrf_incpiv.c
+++ b/compute/pzgetrf_incpiv.c
@@ -84,8 +84,8 @@ void chameleon_pzgetrf_incpiv( CHAM_desc_t *A, CHAM_desc_t *L, CHAM_desc_t *D, i
for (k = 0; k < minMNT; k++) {
RUNTIME_iteration_push(chamctxt, 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;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgetrf_incpiv(
&options,
tempkm, tempkn, ib, L->nb,
@@ -105,7 +105,7 @@ void chameleon_pzgetrf_incpiv( CHAM_desc_t *A, CHAM_desc_t *L, CHAM_desc_t *D, i
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zgessm(
&options,
tempkm, tempnn, tempkm, ib, L->nb,
@@ -115,7 +115,7 @@ void chameleon_pzgetrf_incpiv( CHAM_desc_t *A, CHAM_desc_t *L, CHAM_desc_t *D, i
A(k, n));
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztstrf(
&options,
tempmm, tempkn, ib, L->nb,
@@ -126,7 +126,7 @@ void chameleon_pzgetrf_incpiv( CHAM_desc_t *A, CHAM_desc_t *L, CHAM_desc_t *D, i
m == A->mt-1, A->nb*k);
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zssssm(
&options,
A->nb, tempnn, tempmm, tempnn, A->nb, ib, L->nb,
diff --git a/compute/pzgetrf_nopiv.c b/compute/pzgetrf_nopiv.c
index 424842bc06c9639442124d13509e1f2308b11820..5e7f8423e0c4b1d5f4de257768db4599cfe114cc 100644
--- a/compute/pzgetrf_nopiv.c
+++ b/compute/pzgetrf_nopiv.c
@@ -71,8 +71,8 @@ void chameleon_pzgetrf_nopiv_generic( CHAM_desc_t *A,
for (k = 0; k < chameleon_min(A->mt, A->nt); k++) {
RUNTIME_iteration_push(chamctxt, 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;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
options.priority = 2*A->nt - 2*k;
INSERT_TASK_zgetrf_nopiv(
@@ -82,7 +82,7 @@ void chameleon_pzgetrf_nopiv_generic( CHAM_desc_t *A,
for (m = k+1; m < A->mt; m++) {
options.priority = 2*A->nt - 2*k - m;
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztrsm(
&options,
ChamRight, ChamUpper, ChamNoTrans, ChamNonUnit,
@@ -91,7 +91,7 @@ void chameleon_pzgetrf_nopiv_generic( CHAM_desc_t *A,
A(m, k));
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = 2*A->nt - 2*k - n;
INSERT_TASK_ztrsm(
&options,
@@ -101,7 +101,7 @@ void chameleon_pzgetrf_nopiv_generic( CHAM_desc_t *A,
A(k, n));
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = 2*A->nt - 2*k - n - m;
INSERT_TASK_zgemm(
&options,
@@ -157,8 +157,8 @@ void chameleon_pzgetrf_nopiv_ws( CHAM_desc_t *A,
lp = (k % lookahead) * chameleon_desc_datadist_get_iparam(A, 0);
lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(A, 1);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
options.priority = 2*A->nt - 2*k;
INSERT_TASK_zgetrf_nopiv(
@@ -207,7 +207,7 @@ void chameleon_pzgetrf_nopiv_ws( CHAM_desc_t *A,
}
options.priority = 2*A->nt - 2*k - m;
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
assert( A->get_rankof( A, m, k ) == WU->get_rankof( WU, myp + lp, k) );
INSERT_TASK_ztrsm(
@@ -244,7 +244,7 @@ void chameleon_pzgetrf_nopiv_ws( CHAM_desc_t *A,
continue;
}
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = 2*A->nt - 2*k - n;
assert( A->get_rankof( A, k, n ) == WL->get_rankof( WL, k, myq+lq) );
@@ -281,7 +281,7 @@ void chameleon_pzgetrf_nopiv_ws( CHAM_desc_t *A,
continue;
}
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = 2*A->nt - 2*k - n - m;
assert( A->get_rankof( A, m, n ) == WL->get_rankof( WL, m, myq + lq) );
diff --git a/compute/pzgram.c b/compute/pzgram.c
index 04f849c6ba3a0092ec846fd31a56fdc259e7bc59..788b83998c013345bbbf623a286947e93f06d476 100644
--- a/compute/pzgram.c
+++ b/compute/pzgram.c
@@ -42,10 +42,10 @@ chameleon_pzgram_internal( cham_uplo_t uplo,
for(n = 0; n < NT; n++) {
int mmin = ( uplo == ChamLower ) ? n : 0;
int mmax = ( uplo == ChamUpper ) ? chameleon_min(n+1, MT) : MT;
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
for(m = mmin; m < mmax; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
if ( n == m ) {
INSERT_TASK_dsyssq(
@@ -66,7 +66,7 @@ chameleon_pzgram_internal( cham_uplo_t uplo,
}
for(n = 0; n < NT; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
/**
* 2) reduce columns (scl,ssq) tiles per processus (between lines)
@@ -116,10 +116,10 @@ chameleon_pzgram_internal( cham_uplo_t uplo,
for(n = 0; n < NT; n++) {
int mmin = ( uplo == ChamLower ) ? n : 0;
int mmax = ( uplo == ChamUpper ) ? chameleon_min(n+1, MT) : MT;
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
for(m = mmin; m < mmax; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
INSERT_TASK_zgram(
options,
@@ -152,9 +152,9 @@ void chameleon_pzgram( struct chameleon_pzgram_s *ws, cham_uplo_t uplo, CHAM_des
/* Initialize Wcol */
for(m = 0; m < Wcol->mt; m++) {
- tempmm = m == Wcol->mt-1 ? Wcol->m-m*Wcol->mb : Wcol->mb;
+ tempmm = Wcol->get_blkdim( Wcol, m, DIM_m, Wcol->m );
for(n = 0; n < Wcol->nt; n++) {
- tempnn = n == Wcol->nt-1 ? Wcol->n-n*Wcol->nb : Wcol->nb;
+ tempnn = Wcol->get_blkdim( Wcol, n, DIM_n, Wcol->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
@@ -164,9 +164,9 @@ void chameleon_pzgram( struct chameleon_pzgram_s *ws, cham_uplo_t uplo, CHAM_des
}
/* Initialize Welt */
for(m = 0; m < Welt->mt; m++) {
- tempmm = m == Welt->mt-1 ? Welt->m-m*Welt->mb : Welt->mb;
+ tempmm = Welt->get_blkdim( Welt, m, DIM_m, Welt->m );
for(n = 0; n < Welt->nt; n++) {
- tempnn = n == Welt->nt-1 ? Welt->n-n*Welt->nb : Welt->nb;
+ tempnn = Welt->get_blkdim( Welt, n, DIM_n, Welt->n );
INSERT_TASK_dlaset(
&options,
ChamUpperLower, tempmm, tempnn,
diff --git a/compute/pzhemm.c b/compute/pzhemm.c
index 1d9205311c3e108a0fe4162a5d2f32ef9ed528e0..2a6b69f429cfb19b8b2a2da3d98cb9690e52ff45 100644
--- a/compute/pzhemm.c
+++ b/compute/pzhemm.c
@@ -109,9 +109,9 @@ chameleon_pzhemm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
}
for(n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for(m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
/* Scale C */
options->forcesub = 0;
@@ -125,7 +125,7 @@ chameleon_pzhemm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
if (side == ChamLeft) {
if (uplo == ChamLower) {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
if (k < m) {
INSERT_TASK_zgemm_Astat(
@@ -161,7 +161,7 @@ chameleon_pzhemm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
*/
else {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
if (k < m) {
INSERT_TASK_zgemm_Astat(
@@ -199,7 +199,7 @@ chameleon_pzhemm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
else {
if (uplo == ChamLower) {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
if (k < n) {
INSERT_TASK_zgemm_Astat(
@@ -235,7 +235,7 @@ chameleon_pzhemm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
*/
else {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
if (k < n) {
INSERT_TASK_zgemm_Astat(
@@ -292,7 +292,7 @@ chameleon_pzhemm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
{
RUNTIME_sequence_t *sequence = options->sequence;
cham_trans_t transA;
- int m, n, k, p, q, KT, K, lp, lq;
+ int m, n, k, p, q, KT, lp, lq;
int tempmm, tempnn, tempkk;
int lookahead, myp, myq;
@@ -301,14 +301,13 @@ chameleon_pzhemm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
lookahead = chamctxt->lookahead;
KT = A->nt;
- K = A->n;
myp = C->myrank / chameleon_desc_datadist_get_iparam(C, 1);
myq = C->myrank % chameleon_desc_datadist_get_iparam(C, 1);
for (k = 0; k < KT; k++ ) {
lp = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 0);
lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 1);
- tempkk = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkk = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
/* Transfert ownership of the k column of A or B */
@@ -316,7 +315,7 @@ chameleon_pzhemm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
int Am, Ak;
int tempam, tempak;
- tempmm = m == C->mt-1 ? C->m - m * C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if ( (( uplo == ChamUpper ) && ( m > k )) ||
(( uplo == ChamLower ) && ( m < k )) )
@@ -355,7 +354,7 @@ chameleon_pzhemm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
/* Transfert ownership of the k row of B, or A */
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zlacpy(
options,
@@ -376,11 +375,11 @@ chameleon_pzhemm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
/* Perform the update of this iteration */
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if ( k == m ) {
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zhemm(
options, ChamLeft, uplo,
@@ -401,7 +400,7 @@ chameleon_pzhemm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
}
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zgemm(
options, transA, ChamNoTrans,
@@ -428,7 +427,7 @@ chameleon_pzhemm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
{
RUNTIME_sequence_t *sequence = options->sequence;
cham_trans_t transA;
- int m, n, k, p, q, KT, K, lp, lq;
+ int m, n, k, p, q, KT, lp, lq;
int tempmm, tempnn, tempkk;
int lookahead, myp, myq;
@@ -437,20 +436,19 @@ chameleon_pzhemm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
lookahead = chamctxt->lookahead;
KT = A->mt;
- K = A->m;
myp = C->myrank / chameleon_desc_datadist_get_iparam(C, 1);
myq = C->myrank % chameleon_desc_datadist_get_iparam(C, 1);
for (k = 0; k < KT; k++ ) {
lp = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 0);
lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 1);
- tempkk = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkk = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
/* Transfert ownership of the k column of A or B */
for (m = 0; m < C->mt; m++ ) {
- tempmm = m == C->mt-1 ? C->m - m * C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zlacpy(
options,
@@ -474,7 +472,7 @@ chameleon_pzhemm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
int Ak, An;
int tempak, tempan;
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if ( (( uplo == ChamUpper ) && ( n < k )) ||
(( uplo == ChamLower ) && ( n > k )) )
@@ -511,11 +509,11 @@ chameleon_pzhemm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
/* Perform the update of this iteration */
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if ( k == n ) {
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
/* A has been stored in WA or WB for the summa ring */
INSERT_TASK_zhemm(
@@ -537,7 +535,7 @@ chameleon_pzhemm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
}
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zgemm(
options, ChamNoTrans, transA,
@@ -594,16 +592,16 @@ chameleon_pzhemm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
CHAMELEON_Complex64_t zone = (CHAMELEON_Complex64_t)1.0;
for(m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for(n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
/*
* ChamLeft / ChamLower
*/
if (side == ChamLeft) {
if (uplo == ChamLower) {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
zbeta = k == 0 ? beta : zone;
if (k < m) {
INSERT_TASK_zgemm(
@@ -641,7 +639,7 @@ chameleon_pzhemm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
*/
else {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
zbeta = k == 0 ? beta : zone;
if (k < m) {
INSERT_TASK_zgemm(
@@ -681,7 +679,7 @@ chameleon_pzhemm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
else {
if (uplo == ChamLower) {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
zbeta = k == 0 ? beta : zone;
if (k < n) {
INSERT_TASK_zgemm(
@@ -719,7 +717,7 @@ chameleon_pzhemm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
*/
else {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
zbeta = k == 0 ? beta : zone;
if (k < n) {
INSERT_TASK_zgemm(
diff --git a/compute/pzher2k.c b/compute/pzher2k.c
index 5a1be1c13a13b6f172897f39198e54ea3f6c032e..060926d88ba863248d14838c134167f5ea1398ba 100644
--- a/compute/pzher2k.c
+++ b/compute/pzher2k.c
@@ -52,7 +52,7 @@ void chameleon_pzher2k( cham_uplo_t uplo, cham_trans_t trans,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if (uplo == ChamLower) {
mmin = n+1;
@@ -68,7 +68,7 @@ void chameleon_pzher2k( cham_uplo_t uplo, cham_trans_t trans,
*/
if (trans == ChamNoTrans) {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
dbeta = k == 0 ? beta : 1.0;
INSERT_TASK_zher2k(
&options,
@@ -79,9 +79,9 @@ void chameleon_pzher2k( cham_uplo_t uplo, cham_trans_t trans,
dbeta, C(n, n)); /* ldc * N */
}
for (m = mmin; m < mmax; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? (CHAMELEON_Complex64_t)beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -106,7 +106,7 @@ void chameleon_pzher2k( cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
dbeta = k == 0 ? beta : 1.0;
INSERT_TASK_zher2k(
&options,
@@ -117,9 +117,9 @@ void chameleon_pzher2k( cham_uplo_t uplo, cham_trans_t trans,
dbeta, C(n, n)); /* ldc * N */
}
for (m = mmin; m < mmax; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? (CHAMELEON_Complex64_t)beta : zone;
INSERT_TASK_zgemm(
&options,
diff --git a/compute/pzhered.c b/compute/pzhered.c
index 621cb87ae8a3e3136cf95b9acc12502ee6b7543f..c6821d3638501008710155a2d6ab1ac7137d0dda 100644
--- a/compute/pzhered.c
+++ b/compute/pzhered.c
@@ -81,11 +81,11 @@ chameleon_pzhered_frb( cham_trans_t trans,
int nmin = (uplo == ChamUpper) ? m : 0;
int nmax = (uplo == ChamLower) ? chameleon_min(m + 1, NT) : NT;
- int tempmm = (m == (MT - 1)) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for (n = nmin; n < nmax; n++)
{
- int tempnn = (n == (NT - 1)) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( n == m ) {
if ( trans == ChamConjTrans ) {
@@ -257,13 +257,13 @@ void chameleon_pzhered( cham_trans_t trans,
for (m = 0; m < A->mt; m++)
{
- int tempmm = (m == (A->mt - 1)) ? A->m - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, A->m );
int nmin = (uplo == ChamUpper) ? m : 0;
int nmax = (uplo == ChamLower) ? chameleon_min(m + 1, A->nt) : A->nt;
for (n = nmin; n < nmax; n++)
{
- int tempnn = (n == (A->nt - 1)) ? A->n - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, A->n );
/*
* u_{high} = 1e-16 (later should be application accuracy)
diff --git a/compute/pzherk.c b/compute/pzherk.c
index 4fec1fd779e9acda3c9fe2de5cdb4e5a996cff4c..d1e88310c00e1392cdb668b201ca89445b4723c9 100644
--- a/compute/pzherk.c
+++ b/compute/pzherk.c
@@ -52,13 +52,13 @@ void chameleon_pzherk(cham_uplo_t uplo, cham_trans_t trans,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
/*
* ChamNoTrans
*/
if (trans == ChamNoTrans) {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
dbeta = k == 0 ? beta : 1.0;
INSERT_TASK_zherk(
&options,
@@ -72,9 +72,9 @@ void chameleon_pzherk(cham_uplo_t uplo, cham_trans_t trans,
*/
if (uplo == ChamLower) {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? (CHAMELEON_Complex64_t)beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -91,9 +91,9 @@ void chameleon_pzherk(cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? (CHAMELEON_Complex64_t)beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -111,7 +111,7 @@ void chameleon_pzherk(cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
dbeta = k == 0 ? beta : 1.0;
INSERT_TASK_zherk(
&options,
@@ -125,9 +125,9 @@ void chameleon_pzherk(cham_uplo_t uplo, cham_trans_t trans,
*/
if (uplo == ChamLower) {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? (CHAMELEON_Complex64_t)beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -144,9 +144,9 @@ void chameleon_pzherk(cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? (CHAMELEON_Complex64_t)beta : zone;
INSERT_TASK_zgemm(
&options,
diff --git a/compute/pzhetrd_he2hb.c b/compute/pzhetrd_he2hb.c
index 706740891ade6bc8d3cd10ac1ff1be2338ce0c55..ae4906dd0c360d5b2c265c42ecdf98539122a80d 100644
--- a/compute/pzhetrd_he2hb.c
+++ b/compute/pzhetrd_he2hb.c
@@ -101,7 +101,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
/* Let's extract the diagonal in a temporary copy that contains A and A' */
for (k = 1; k < A->nt; k++){
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zhe2ge( &options,
uplo, tempkn, tempkn, A->mb,
@@ -112,8 +112,8 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
for (k = 0; k < A->nt-1; k++){
RUNTIME_iteration_push(chamctxt, k);
- 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;
+ tempkm = A->get_blkdim( A, k+1, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgeqrt(
&options,
@@ -147,7 +147,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
/* RIGHT on the remaining tiles until the bottom */
for (m = k+2; m < A->mt ; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zunmqr(
&options,
ChamRight, ChamNoTrans,
@@ -158,7 +158,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
}
for (m = k+2; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = 1;
INSERT_TASK_ztsqrt(
@@ -183,7 +183,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
/* RIGHT */
for (j = m+1; j < A->mt ; j++) {
- tempjj = j == A->mt-1 ? A->m-j*A->mb : A->mb;
+ tempjj = A->get_blkdim( A, j, DIM_m, A->m );
INSERT_TASK_ztsmqr(
&options,
ChamRight, ChamNoTrans,
@@ -264,8 +264,9 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
for (k = 0; k < A->nt-1; k++){
RUNTIME_iteration_push(chamctxt, k);
- 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;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k+1, DIM_n, A->n );
+
INSERT_TASK_zgelqt(
&options,
tempkm, tempkn, ib, A->nb,
@@ -298,7 +299,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
/* LEFT on the remaining tiles until the left side */
for (n = k+2; n < A->nt ; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zunmlq(
&options,
ChamLeft, ChamNoTrans,
@@ -309,7 +310,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
}
for (n = k+2; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = 1;
INSERT_TASK_ztslqt(
&options,
@@ -333,7 +334,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
/* LEFT */
for (j = n+1; j < A->nt ; j++) {
- tempjj = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjj = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_ztsmlq(
&options,
ChamLeft, ChamNoTrans,
@@ -413,7 +414,7 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
/* Copy-back into A */
for (k = 1; k < A->nt; k++){
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zlacpy( &options,
uplo, tempkn, tempkn,
D(k), A(k, k));
diff --git a/compute/pzlag2c.c b/compute/pzlag2c.c
index fb45797c676de7a50a3012e1df1997728abfc416..4113e49138e186c67fdefe58ea1d739d5e971f74 100644
--- a/compute/pzlag2c.c
+++ b/compute/pzlag2c.c
@@ -46,10 +46,10 @@ void chameleon_pclag2z( CHAM_desc_t *A, CHAM_desc_t *B,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for(m = 0; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for(n = 0; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_clag2z(
&options,
@@ -81,10 +81,10 @@ void chameleon_pzlag2c( CHAM_desc_t *A, CHAM_desc_t *B,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for(m = 0; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m * A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for(n = 0; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n - n * A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlag2c(
&options,
diff --git a/compute/pzlange.c b/compute/pzlange.c
index abfb27a366655537e0fa0e6e4a48d7e3626aa2bf..100b44059da604131a3f32751741e6cdc5c27f8a 100644
--- a/compute/pzlange.c
+++ b/compute/pzlange.c
@@ -54,10 +54,10 @@ chameleon_pzlange_one( cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A,
int mmin = ( uplo == ChamLower ) ? n : 0;
int mmax = ( uplo == ChamUpper ) ? chameleon_min(n+1, MT) : MT;
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
for(m = mmin; m < mmax; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
if ( (n == m) && (uplo != ChamUpperLower) ) {
INSERT_TASK_ztrasm(
@@ -144,10 +144,10 @@ chameleon_pzlange_inf( cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A,
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( (n == m) && (uplo != ChamUpperLower) ) {
INSERT_TASK_ztrasm(
@@ -230,10 +230,10 @@ chameleon_pzlange_max( cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A, CHAM_
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( (n == m) && (uplo != ChamUpperLower) ) {
INSERT_TASK_zlantr(
@@ -313,10 +313,10 @@ chameleon_pzlange_frb( cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A, CHAM_
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( (n == m) && (uplo != ChamUpperLower) ) {
INSERT_TASK_ztrssq(
diff --git a/compute/pzlansy.c b/compute/pzlansy.c
index ab709ac9e396a578ad3e5db31c7577661505848b..928e26964c6674f67b46c07c6be1b4f5db2a717c 100644
--- a/compute/pzlansy.c
+++ b/compute/pzlansy.c
@@ -51,10 +51,10 @@ chameleon_pzlansy_inf( cham_uplo_t uplo, CHAM_desc_t *A,
int nmin = ( uplo == ChamUpper ) ? m : 0;
int nmax = ( uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( n == m ) {
INSERT_TASK_dzasum(
@@ -77,7 +77,7 @@ chameleon_pzlansy_inf( cham_uplo_t uplo, CHAM_desc_t *A,
}
for(m = 0; m < MT; m++) {
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = Q; n < NT; n++) {
INSERT_TASK_daxpy(
@@ -144,10 +144,10 @@ chameleon_pzlansy_max( cham_trans_t trans, cham_uplo_t uplo, CHAM_desc_t *A,
int nmin = (uplo == ChamUpper ) ? m : 0;
int nmax = (uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( n == m ) {
if ( trans == ChamConjTrans) {
@@ -230,10 +230,10 @@ chameleon_pzlansy_frb( cham_trans_t trans, cham_uplo_t uplo,
int nmin = (uplo == ChamUpper ) ? m : 0;
int nmax = (uplo == ChamLower ) ? chameleon_min(m+1, NT) : NT;
- int tempmm = ( m == (MT-1) ) ? M - m * A->mb : A->mb;
+ int tempmm = A->get_blkdim( A, m, DIM_m, M );
for(n = nmin; n < nmax; n++) {
- int tempnn = ( n == (NT-1) ) ? N - n * A->nb : A->nb;
+ int tempnn = A->get_blkdim( A, n, DIM_n, N );
if ( n == m ) {
if ( trans == ChamConjTrans) {
diff --git a/compute/pzlascal.c b/compute/pzlascal.c
index 5334730b3cae41d140c6f452fefe2e1986bb3d01..09ac36ca9b44aae41bb64b3f49ba04532e2e6f28 100644
--- a/compute/pzlascal.c
+++ b/compute/pzlascal.c
@@ -44,8 +44,8 @@ void chameleon_pzlascal(cham_uplo_t uplo, CHAMELEON_Complex64_t alpha, CHAM_desc
switch(uplo) {
case ChamLower:
for (n = 0; n < minmnt; n++) {
- tempnm = n == A->mt-1 ? A->m-n*A->mb : A->mb;
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnm = A->get_blkdim( A, n, DIM_m, A->m );
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlascal(
&options,
@@ -53,7 +53,7 @@ void chameleon_pzlascal(cham_uplo_t uplo, CHAMELEON_Complex64_t alpha, CHAM_desc
alpha, A(n, n));
for (m = n+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-A->mb*m : A->nb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zlascal(
&options,
@@ -65,8 +65,8 @@ void chameleon_pzlascal(cham_uplo_t uplo, CHAMELEON_Complex64_t alpha, CHAM_desc
case ChamUpper:
for (m = 0; m < minmnt; m++) {
- tempmm = m == A->mt-1 ? A->m-A->mb*m : A->nb;
- tempmn = m == A->nt-1 ? A->n-m*A->nb : A->nb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
+ tempmn = A->get_blkdim( A, m, DIM_n, A->n );
INSERT_TASK_zlascal(
&options,
@@ -74,7 +74,7 @@ void chameleon_pzlascal(cham_uplo_t uplo, CHAMELEON_Complex64_t alpha, CHAM_desc
alpha, A(m, m));
for (n = m+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlascal(
&options,
@@ -87,10 +87,10 @@ void chameleon_pzlascal(cham_uplo_t uplo, CHAMELEON_Complex64_t alpha, CHAM_desc
case ChamUpperLower:
default:
for (m = 0; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-A->mb*m : A->nb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zlascal(
&options,
diff --git a/compute/pzlaset.c b/compute/pzlaset.c
index 3eec0b7c67c9fd9fc17142747c50921c1f4006de..94ce79ce7f133fd744ff735bb51b7c47d51d579d 100644
--- a/compute/pzlaset.c
+++ b/compute/pzlaset.c
@@ -52,15 +52,15 @@ void chameleon_pzlaset( cham_uplo_t uplo,
if (uplo == ChamLower) {
for (j = 0; j < minmn; j++){
- tempjm = j == A->mt-1 ? A->m-j*A->mb : A->mb;
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjm = A->get_blkdim( A, j, DIM_m, A->m );
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset(
&options,
ChamLower, tempjm, tempjn, alpha, beta,
A(j, j));
for (i = j+1; i < A->mt; i++){
- tempim = i == A->mt-1 ? A->m-i*A->mb : A->mb;
+ tempim = A->get_blkdim( A, i, DIM_m, A->m );
INSERT_TASK_zlaset(
&options,
ChamUpperLower, tempim, tempjn, alpha, alpha,
@@ -70,11 +70,11 @@ void chameleon_pzlaset( cham_uplo_t uplo,
}
else if (uplo == ChamUpper) {
for (i = 0; i < A->mt; i++) {
- tempim = i == A->mt-1 ? A->m-i*A->mb : A->mb;
+ tempim = A->get_blkdim( A, i, DIM_m, A->m );
if ( i < A->nt ) {
j = i;
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset(
&options,
@@ -82,7 +82,7 @@ void chameleon_pzlaset( cham_uplo_t uplo,
alpha, beta, A(i, j));
}
for (j = i+1; j < A->nt; j++) {
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset(
&options,
@@ -93,9 +93,9 @@ void chameleon_pzlaset( cham_uplo_t uplo,
}
else {
for (i = 0; i < A->mt; i++){
- tempim = i == A->mt-1 ? A->m-i*A->mb : A->mb;
+ tempim = A->get_blkdim( A, i, DIM_m, A->m );
for (j = 0; j < A->nt; j++){
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset(
&options,
ChamUpperLower, tempim, tempjn,
diff --git a/compute/pzlaset2.c b/compute/pzlaset2.c
index 1819b70b79db0f3406619682ecd10b4e4d1bbf7a..7efc84208b76f992eb96fb41b5ea8b060d424975 100644
--- a/compute/pzlaset2.c
+++ b/compute/pzlaset2.c
@@ -51,15 +51,15 @@ void chameleon_pzlaset2( cham_uplo_t uplo,
if (uplo == ChamLower) {
for (j = 0; j < minmn; j++){
- tempjm = j == A->mt-1 ? A->m-j*A->mb : A->mb;
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjm = A->get_blkdim( A, j, DIM_m, A->m );
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset2(
&options,
ChamLower, tempjm, tempjn, alpha,
A(j, j));
for (i = j+1; i < A->mt; i++){
- tempim = i == A->mt-1 ? A->m-i*A->mb : A->mb;
+ tempim = A->get_blkdim( A, i, DIM_m, A->m );
INSERT_TASK_zlaset2(
&options,
ChamUpperLower, tempim, tempjn, alpha,
@@ -69,9 +69,9 @@ void chameleon_pzlaset2( cham_uplo_t uplo,
}
else if (uplo == ChamUpper) {
for (j = 1; j < A->nt; j++){
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
for (i = 0; i < chameleon_min(j, A->mt); i++){
- tempim = i == A->mt-1 ? A->m-i*A->mb : A->mb;
+ tempim = A->get_blkdim( A, i, DIM_m, A->m );
INSERT_TASK_zlaset2(
&options,
ChamUpperLower, tempim, tempjn, alpha,
@@ -79,8 +79,8 @@ void chameleon_pzlaset2( cham_uplo_t uplo,
}
}
for (j = 0; j < minmn; j++){
- tempjm = j == A->mt-1 ? A->m-j*A->mb : A->mb;
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjm = A->get_blkdim( A, j, DIM_m, A->m );
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset2(
&options,
ChamUpper, tempjm, tempjn, alpha,
@@ -89,9 +89,9 @@ void chameleon_pzlaset2( cham_uplo_t uplo,
}
else {
for (i = 0; i < A->mt; i++){
- tempim = i == A->mt-1 ? A->m-i*A->mb : A->mb;
+ tempim = A->get_blkdim( A, i, DIM_m, A->m );
for (j = 0; j < A->nt; j++){
- tempjn = j == A->nt-1 ? A->n-j*A->nb : A->nb;
+ tempjn = A->get_blkdim( A, j, DIM_n, A->n );
INSERT_TASK_zlaset2(
&options,
ChamUpperLower, tempim, tempjn, alpha,
diff --git a/compute/pzlatms.c b/compute/pzlatms.c
index 483ce2a1f7afa2b79623bba80bb22b86afc51fec..7e6fd24e3eb7f053ef7a6efe0de702316773b394 100644
--- a/compute/pzlatms.c
+++ b/compute/pzlatms.c
@@ -34,8 +34,8 @@ zlaset_diag_cpu( void *op_args,
const double *D = (const double *)op_args;
CHAMELEON_Complex64_t *A = CHAM_tile_get_ptr( tileA );
- int tempmm = m == descA->mt-1 ? descA->m-m*descA->mb : descA->mb;
- int tempnn = n == descA->nt-1 ? descA->n-n*descA->nb : descA->nb;
+ int tempmm = descA->get_blkdim( descA, m, DIM_m, descA->m );
+ int tempnn = descA->get_blkdim( descA, n, DIM_n, descA->n );
int minmn = chameleon_min( tempmm, tempnn );
int lda = tileA->ld;
int i;
diff --git a/compute/pzlauum.c b/compute/pzlauum.c
index 9a948e50a969743cc650e4d10bb6602c2ebba8b4..b462e65d92e64bde05ae08e4963279464d1b1cc6 100644
--- a/compute/pzlauum.c
+++ b/compute/pzlauum.c
@@ -49,7 +49,7 @@ void chameleon_pzlauum(cham_uplo_t uplo, CHAM_desc_t *A,
*/
if (uplo == ChamLower) {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for(n = 0; n < k; n++) {
INSERT_TASK_zherk(
&options,
@@ -89,7 +89,7 @@ void chameleon_pzlauum(cham_uplo_t uplo, CHAM_desc_t *A,
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (m = 0; m < k; m++) {
INSERT_TASK_zherk(
diff --git a/compute/pzplghe.c b/compute/pzplghe.c
index b224835a07e5afe4b0cff8f5a6f2248c4afe8a09..5add769a300d06979e3512b7b7a3f4628024ab4a 100644
--- a/compute/pzplghe.c
+++ b/compute/pzplghe.c
@@ -47,10 +47,10 @@ void chameleon_pzplghe( double bump, cham_uplo_t uplo, CHAM_desc_t *A,
switch ( uplo ) {
case ChamLower:
for (n = 0; n < minmn; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
for (m = n; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = m + n;
INSERT_TASK_zplghe(
@@ -63,10 +63,10 @@ void chameleon_pzplghe( double bump, cham_uplo_t uplo, CHAM_desc_t *A,
case ChamUpper:
for (m = 0; m < minmn; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = m; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = m + n;
INSERT_TASK_zplghe(
@@ -80,10 +80,10 @@ void chameleon_pzplghe( double bump, cham_uplo_t uplo, CHAM_desc_t *A,
case ChamUpperLower:
default:
for (m = 0; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = m + n;
INSERT_TASK_zplghe(
diff --git a/compute/pzplgsy.c b/compute/pzplgsy.c
index bc64fadaf68bad0f778f4fd1c5097b085043ae5f..a1d0eedcde2a00e9a108eda50b6d1e0154bebfd9 100644
--- a/compute/pzplgsy.c
+++ b/compute/pzplgsy.c
@@ -47,10 +47,10 @@ void chameleon_pzplgsy( CHAMELEON_Complex64_t bump, cham_uplo_t uplo, CHAM_desc_
switch ( uplo ) {
case ChamLower:
for (n = 0; n < minmn; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
for (m = n; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = m + n;
INSERT_TASK_zplgsy(
@@ -63,10 +63,10 @@ void chameleon_pzplgsy( CHAMELEON_Complex64_t bump, cham_uplo_t uplo, CHAM_desc_
case ChamUpper:
for (m = 0; m < minmn; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = m; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = m + n;
INSERT_TASK_zplgsy(
@@ -80,10 +80,10 @@ void chameleon_pzplgsy( CHAMELEON_Complex64_t bump, cham_uplo_t uplo, CHAM_desc_
case ChamUpperLower:
default:
for (m = 0; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = m + n;
INSERT_TASK_zplgsy(
diff --git a/compute/pzplrnk.c b/compute/pzplrnk.c
index bab04f16ad8af1bc45021f8ab0dbf1ac504ae7eb..97e7c8bdccb2f45b323717391d7ccf7c892e5a06 100644
--- a/compute/pzplrnk.c
+++ b/compute/pzplrnk.c
@@ -53,12 +53,12 @@ chameleon_pzplrnk_generic( CHAM_context_t *chamctxt,
memset( initB, 0, C->nt * sizeof(int) );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
initA = 0;
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if ( C->get_rankof( C(m, n) ) == myrank ) {
if ( !initA ) {
@@ -129,7 +129,7 @@ chameleon_pzplrnk_2dbc( CHAM_context_t *chamctxt,
zbeta = k == 0 ? 0. : 1.;
for (n = myq; n < C->nt; n += q) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zplrnt(
options,
@@ -138,7 +138,7 @@ chameleon_pzplrnk_2dbc( CHAM_context_t *chamctxt,
}
for (m = myp; m < C->mt; m += p) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zplrnt(
options,
@@ -146,7 +146,7 @@ chameleon_pzplrnk_2dbc( CHAM_context_t *chamctxt,
WA->m, m * WA->mb, k * WA->nb, seedA );
for (n = myq; n < C->nt; n+=q) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zgemm(
options,
@@ -195,13 +195,13 @@ chameleon_pzplrnk( int K,
{
chameleon_desc_init( &WA, CHAMELEON_MAT_ALLOC_TILE,
ChamComplexDouble, C->mb, C->nb, (C->mb * C->nb),
- C->mt * C->mb, C->nb * q, 0, 0,
- C->mt * C->mb, C->nb * q, p, q,
+ C->m, C->nb * q, 0, 0,
+ C->m, C->nb * q, p, q,
NULL, NULL, NULL, NULL );
chameleon_desc_init( &WB, CHAMELEON_MAT_ALLOC_TILE,
ChamComplexDouble, C->mb, C->nb, (C->mb * C->nb),
- C->mb * p, C->nt * C->nb, 0, 0,
- C->mb * p, C->nt * C->nb, p, q,
+ C->mb * p, C->n, 0, 0,
+ C->mb * p, C->n, p, q,
NULL, NULL, NULL, NULL );
chameleon_pzplrnk_2dbc( chamctxt, K, &WA, &WB, C, seedA, seedB, &options );
@@ -210,13 +210,13 @@ chameleon_pzplrnk( int K,
int np = p * q;
chameleon_desc_init( &WA, CHAMELEON_MAT_ALLOC_TILE,
ChamComplexDouble, C->mb, C->nb, (C->mb * C->nb),
- C->mt * C->mb, C->nb * np, 0, 0,
- C->mt * C->mb, C->nb * np, 1, np,
+ C->m, C->nb * np, 0, 0,
+ C->m, C->nb * np, 1, np,
NULL, NULL, NULL, NULL );
chameleon_desc_init( &WB, CHAMELEON_MAT_ALLOC_TILE,
ChamComplexDouble, C->mb, C->nb, (C->mb * C->nb),
- C->mb * np, C->nt * C->nb, 0, 0,
- C->mb * np, C->nt * C->nb, np, 1,
+ C->mb * np, C->n, 0, 0,
+ C->mb * np, C->n, np, 1,
NULL, NULL, NULL, NULL );
chameleon_pzplrnk_generic( chamctxt, K, &WA, &WB, C, seedA, seedB, &options );
diff --git a/compute/pzplrnt.c b/compute/pzplrnt.c
index 571a2fa05c1f33596ae9a29773e91555a2fe9b65..93d4d7e830b1f11d92207ded0ab70f58099919c5 100644
--- a/compute/pzplrnt.c
+++ b/compute/pzplrnt.c
@@ -46,10 +46,10 @@ void chameleon_pzplrnt( CHAM_desc_t *A,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for (m = 0; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zplrnt(
&options,
diff --git a/compute/pzpotrf.c b/compute/pzpotrf.c
index 43656e9d38e4cbdc312221ce81be2edd3f2977b7..0b613229ed33ed4bae99cfec0d91bb4c903fabda 100644
--- a/compute/pzpotrf.c
+++ b/compute/pzpotrf.c
@@ -60,7 +60,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
options.priority = 2*A->mt - 2*k;
INSERT_TASK_zpotrf(
@@ -69,7 +69,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
A(k, k), A->nb*k);
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = 2*A->mt - 2*k - m;
INSERT_TASK_ztrsm(
@@ -82,7 +82,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
RUNTIME_data_flush( sequence, A(k, k) );
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = 2*A->mt - 2*k - n;
INSERT_TASK_zherk(
@@ -93,7 +93,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
1.0, A(n, n));
for (m = n+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = 2*A->mt - 2*k - n - m;
INSERT_TASK_zgemm(
@@ -116,7 +116,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
for (k = 0; k < A->nt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_n, A->n );
options.priority = 2*A->nt - 2*k;
INSERT_TASK_zpotrf(
@@ -126,7 +126,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
A(k, k), A->nb*k);
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n - n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = 2*A->nt - 2*k - n;
INSERT_TASK_ztrsm(
@@ -139,7 +139,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
RUNTIME_data_flush( sequence, A(k, k) );
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
options.priority = 2*A->nt - 2*k - m;
INSERT_TASK_zherk(
@@ -150,7 +150,7 @@ void chameleon_pzpotrf(cham_uplo_t uplo, CHAM_desc_t *A,
1.0, A(m, m));
for (n = m+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
options.priority = 2*A->nt - 2*k - n - m;
INSERT_TASK_zgemm(
diff --git a/compute/pzpotrimm.c b/compute/pzpotrimm.c
index c6d301836b8517b60dba1f62babd69cb9377869f..e3638491cb73a07b9eb291043c9cfd6d24cf6ff3 100644
--- a/compute/pzpotrimm.c
+++ b/compute/pzpotrimm.c
@@ -63,7 +63,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
INSERT_TASK_zpotrf(
&options,
@@ -71,7 +71,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
A(k, k), A->nb*k);
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztrsm(
&options,
ChamRight, ChamLower, ChamConjTrans, ChamNonUnit,
@@ -82,7 +82,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
RUNTIME_data_flush( sequence, A(k, k) );
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zherk(
&options,
ChamLower, ChamNoTrans,
@@ -91,7 +91,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
1.0, A(n, n));
for (m = n+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, ChamConjTrans,
@@ -111,9 +111,9 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < A->nt; k++) {
RUNTIME_iteration_push(chamctxt, A->nt + k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztrsm(
&options,
ChamRight, uplo, ChamNoTrans, ChamNonUnit,
@@ -122,7 +122,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
A(m, k));
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < k; n++) {
INSERT_TASK_zgemm(
&options,
@@ -158,7 +158,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, 2*A->nt + k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for(n = 0; n < k; n++) {
INSERT_TASK_zherk(
&options,
@@ -200,14 +200,14 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < C->nt; k++) {
RUNTIME_iteration_push(chamctxt, 3*A->nt + k);
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
zbeta = k == 0 ? beta : zone;
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if (k < n) {
INSERT_TASK_zgemm(
@@ -258,7 +258,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < A->nt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zpotrf(
&options,
ChamUpper,
@@ -266,7 +266,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
A(k, k), A->nb*k);
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n - n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_ztrsm(
&options,
ChamLeft, ChamUpper, ChamConjTrans, ChamNonUnit,
@@ -277,7 +277,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
RUNTIME_data_flush( sequence, A(k, k) );
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zherk(
&options,
@@ -287,7 +287,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
1.0, A(m, m));
for (n = m+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zgemm(
&options,
@@ -308,9 +308,9 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, A->nt + k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_ztrsm(
&options,
ChamLeft, uplo, ChamNoTrans, ChamNonUnit,
@@ -319,7 +319,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
A(k, n));
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
for (m = 0; m < k; m++) {
INSERT_TASK_zgemm(
&options,
@@ -355,7 +355,7 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, 2*A->nt + k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (m = 0; m < k; m++) {
INSERT_TASK_zherk(
@@ -398,14 +398,14 @@ void chameleon_pzpotrimm(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_
for (k = 0; k < C->nt; k++) {
RUNTIME_iteration_push(chamctxt, 3*A->nt + k);
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
zbeta = k == 0 ? beta : zone;
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if (k < n) {
INSERT_TASK_zgemm(
diff --git a/compute/pzsymm.c b/compute/pzsymm.c
index 8f16daf6fa1a1ed601ea31e6425f14c949891e3c..bcde1927be003743fe8c041bdb258aeca91c6d29 100644
--- a/compute/pzsymm.c
+++ b/compute/pzsymm.c
@@ -110,9 +110,9 @@ chameleon_pzsymm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
}
for(n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for(m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
/* Scale C */
options->forcesub = 0;
@@ -126,7 +126,7 @@ chameleon_pzsymm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
if (side == ChamLeft) {
if (uplo == ChamLower) {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
if (k < m) {
INSERT_TASK_zgemm_Astat(
@@ -162,7 +162,7 @@ chameleon_pzsymm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
*/
else {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
if (k < m) {
INSERT_TASK_zgemm_Astat(
@@ -200,7 +200,7 @@ chameleon_pzsymm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
else {
if (uplo == ChamLower) {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
if (k < n) {
INSERT_TASK_zgemm_Astat(
@@ -236,7 +236,7 @@ chameleon_pzsymm_Astat( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_t
*/
else {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
if (k < n) {
INSERT_TASK_zgemm_Astat(
@@ -293,7 +293,7 @@ chameleon_pzsymm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
{
RUNTIME_sequence_t *sequence = options->sequence;
cham_trans_t transA;
- int m, n, k, p, q, KT, K, lp, lq;
+ int m, n, k, p, q, KT, lp, lq;
int tempmm, tempnn, tempkk;
int lookahead, myp, myq;
@@ -302,14 +302,13 @@ chameleon_pzsymm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
lookahead = chamctxt->lookahead;
KT = A->nt;
- K = A->n;
myp = C->myrank / chameleon_desc_datadist_get_iparam(C, 1);
myq = C->myrank % chameleon_desc_datadist_get_iparam(C, 1);
for (k = 0; k < KT; k++ ) {
lp = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 0);
lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 1);
- tempkk = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkk = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
/* Transfert ownership of the k column of A or B */
@@ -317,7 +316,7 @@ chameleon_pzsymm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
int Am, Ak;
int tempam, tempak;
- tempmm = m == C->mt-1 ? C->m - m * C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if ( (( uplo == ChamUpper ) && ( m > k )) ||
(( uplo == ChamLower ) && ( m < k )) )
@@ -356,7 +355,7 @@ chameleon_pzsymm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
/* Transfert ownership of the k row of B, or A */
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zlacpy(
options,
@@ -377,11 +376,11 @@ chameleon_pzsymm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
/* Perform the update of this iteration */
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if ( k == m ) {
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zsymm(
options, ChamLeft, uplo,
@@ -402,7 +401,7 @@ chameleon_pzsymm_summa_left( CHAM_context_t *chamctxt, cham_uplo_t uplo,
}
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zgemm(
options, transA, ChamNoTrans,
@@ -429,7 +428,7 @@ chameleon_pzsymm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
{
RUNTIME_sequence_t *sequence = options->sequence;
cham_trans_t transA;
- int m, n, k, p, q, KT, K, lp, lq;
+ int m, n, k, p, q, KT, lp, lq;
int tempmm, tempnn, tempkk;
int lookahead, myp, myq;
@@ -438,20 +437,19 @@ chameleon_pzsymm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
lookahead = chamctxt->lookahead;
KT = A->mt;
- K = A->m;
myp = C->myrank / chameleon_desc_datadist_get_iparam(C, 1);
myq = C->myrank % chameleon_desc_datadist_get_iparam(C, 1);
for (k = 0; k < KT; k++ ) {
lp = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 0);
lq = (k % lookahead) * chameleon_desc_datadist_get_iparam(C, 1);
- tempkk = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkk = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
/* Transfert ownership of the k column of A or B */
for (m = 0; m < C->mt; m++ ) {
- tempmm = m == C->mt-1 ? C->m - m * C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zlacpy(
options,
@@ -475,7 +473,7 @@ chameleon_pzsymm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
int Ak, An;
int tempak, tempan;
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if ( (( uplo == ChamUpper ) && ( n < k )) ||
(( uplo == ChamLower ) && ( n > k )) )
@@ -512,11 +510,11 @@ chameleon_pzsymm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
/* Perform the update of this iteration */
for (n = myq; n < C->nt; n+=chameleon_desc_datadist_get_iparam(C, 1)) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if ( k == n ) {
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
/* A has been stored in WA or WB for the summa ring */
INSERT_TASK_zsymm(
@@ -538,7 +536,7 @@ chameleon_pzsymm_summa_right( CHAM_context_t *chamctxt, cham_uplo_t uplo,
}
for (m = myp; m < C->mt; m+=chameleon_desc_datadist_get_iparam(C, 0)) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zgemm(
options, ChamNoTrans, transA,
@@ -595,16 +593,16 @@ chameleon_pzsymm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
CHAMELEON_Complex64_t zone = (CHAMELEON_Complex64_t)1.0;
for(m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for(n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
/*
* ChamLeft / ChamLower
*/
if (side == ChamLeft) {
if (uplo == ChamLower) {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
zbeta = k == 0 ? beta : zone;
if (k < m) {
INSERT_TASK_zgemm(
@@ -642,7 +640,7 @@ chameleon_pzsymm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
*/
else {
for (k = 0; k < C->mt; k++) {
- tempkm = k == C->mt-1 ? C->m-k*C->mb : C->mb;
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
zbeta = k == 0 ? beta : zone;
if (k < m) {
INSERT_TASK_zgemm(
@@ -682,7 +680,7 @@ chameleon_pzsymm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
else {
if (uplo == ChamLower) {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
zbeta = k == 0 ? beta : zone;
if (k < n) {
INSERT_TASK_zgemm(
@@ -720,7 +718,7 @@ chameleon_pzsymm_generic( CHAM_context_t *chamctxt, cham_side_t side, cham_uplo_
*/
else {
for (k = 0; k < C->nt; k++) {
- tempkn = k == C->nt-1 ? C->n-k*C->nb : C->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
zbeta = k == 0 ? beta : zone;
if (k < n) {
INSERT_TASK_zgemm(
diff --git a/compute/pzsyr2k.c b/compute/pzsyr2k.c
index 74b386cd1d1e97a12054ee362017ecdc38855aad..945b97f30c91932b7e6be8f8afeca08e79073a5d 100644
--- a/compute/pzsyr2k.c
+++ b/compute/pzsyr2k.c
@@ -51,7 +51,7 @@ void chameleon_pzsyr2k( cham_uplo_t uplo, cham_trans_t trans,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if (uplo == ChamLower) {
mmin = n+1;
@@ -67,7 +67,7 @@ void chameleon_pzsyr2k( cham_uplo_t uplo, cham_trans_t trans,
*/
if (trans == ChamNoTrans) {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zsyr2k(
&options,
@@ -78,9 +78,9 @@ void chameleon_pzsyr2k( cham_uplo_t uplo, cham_trans_t trans,
zbeta, C(n, n)); /* ldc * N */
}
for (m = mmin; m < mmax; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -105,7 +105,7 @@ void chameleon_pzsyr2k( cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zsyr2k(
&options,
@@ -116,9 +116,9 @@ void chameleon_pzsyr2k( cham_uplo_t uplo, cham_trans_t trans,
zbeta, C(n, n)); /* ldc * N */
}
for (m = mmin; m < mmax; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
&options,
diff --git a/compute/pzsyrk.c b/compute/pzsyrk.c
index aa4a5b29741941fa9706a273ea7afe388679c868..9b4ab5e117af9ceb79ab57dbbc34605325e0e4e1 100644
--- a/compute/pzsyrk.c
+++ b/compute/pzsyrk.c
@@ -51,13 +51,13 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans,
RUNTIME_options_init(&options, chamctxt, sequence, request);
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
/*
* ChamNoTrans
*/
if (trans == ChamNoTrans) {
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zsyrk(
&options,
@@ -71,9 +71,9 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans,
*/
if (uplo == ChamLower) {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -90,9 +90,9 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->nt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -110,7 +110,7 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zsyrk(
&options,
@@ -124,9 +124,9 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans,
*/
if (uplo == ChamLower) {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
&options,
@@ -143,9 +143,9 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans,
*/
else {
for (m = n+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (k = 0; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
zbeta = k == 0 ? beta : zone;
INSERT_TASK_zgemm(
&options,
diff --git a/compute/pzsytrf.c b/compute/pzsytrf.c
index 6a9c41bedca23e8e8ecf493c61745d6b39c8e241..1cf7ce198cd3bd2fad7758d7604c03f7bac9238a 100644
--- a/compute/pzsytrf.c
+++ b/compute/pzsytrf.c
@@ -58,7 +58,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
INSERT_TASK_zsytrf_nopiv(
&options,
@@ -66,7 +66,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
A(k, k), A->nb*k);
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztrsm(
&options,
ChamRight, ChamLower, ChamTrans, ChamNonUnit,
@@ -77,7 +77,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
RUNTIME_data_flush( sequence, A(k, k) );
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zsyrk(
&options,
ChamLower, ChamNoTrans,
@@ -86,7 +86,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
1.0, A(n, n));
for (m = n+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, ChamTrans,
@@ -108,7 +108,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
for (k = 0; k < A->nt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkm = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zsytrf_nopiv(
&options,
ChamUpper,
@@ -116,7 +116,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
A(k, k), A->nb*k);
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n - n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_ztrsm(
&options,
ChamLeft, ChamUpper, ChamTrans, ChamNonUnit,
@@ -127,7 +127,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
RUNTIME_data_flush( sequence, A(k, k) );
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m - m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_zsyrk(
&options,
@@ -137,7 +137,7 @@ void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A,
1.0, A(m, m));
for (n = m+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_zgemm(
&options,
diff --git a/compute/pztile2band.c b/compute/pztile2band.c
index 3ec2c09edaffbfd26fa2513b4500d291490b68fd..cabfe55d383530d3e4c1738f1319279d77964ad8 100644
--- a/compute/pztile2band.c
+++ b/compute/pztile2band.c
@@ -38,9 +38,7 @@ void chameleon_pztile2band( cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B,
int k;
int tempkm, tempkn;
int minmnt = chameleon_min(A->mt, A->nt);
- int Bnb = B->nb;
int Bmb = B->mb;
- int Amb = A->mb;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -59,8 +57,8 @@ void chameleon_pztile2band( cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B,
assert( A->i == B->j );
assert( A->j >= B->j );
- tempkm = ( k == A->mt-1 ) ? A->m - k * Amb : Amb;
- tempkn = ( k == B->nt-1 ) ? B->n - k * Bnb : Bnb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = B->get_blkdim( B, k, DIM_n, B->n );
INSERT_TASK_zlaset( &options, ChamUpperLower, Bmb, tempkn,
0., 0., B, 0, k );
@@ -72,7 +70,7 @@ void chameleon_pztile2band( cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B,
if ( k < minmnt-1 ) {
tileA = A->get_blktile( A, k+1, k );
- tempkm = ( (k+1) == A->mt-1 ) ? A->m - (k+1) * Amb : Amb;
+ tempkm = A->get_blkdim( A, k+1, DIM_m, A->m );
INSERT_TASK_zlacpyx( &options, ChamUpper, tempkm, tempkn,
0, A, k+1, k, tileA->ld,
@@ -88,8 +86,8 @@ void chameleon_pztile2band( cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B,
assert( A->i == B->i );
assert( A->i >= B->j );
- tempkm = ( k == A->mt-1 ) ? A->m - k * Amb : Amb;
- tempkn = ( k == B->nt-1 ) ? B->n - k * Bnb : Bnb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = B->get_blkdim( B, k, DIM_n, B->n );
INSERT_TASK_zlaset( &options, ChamUpperLower, Bmb, tempkn,
0., 0., B, 0, k );
@@ -101,7 +99,7 @@ void chameleon_pztile2band( cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *B,
if ( k > 0 ) {
tileA = A->get_blktile( A, k-1, k );
- tempkm = ( (k-1) == A->mt-1 ) ? A->m - (k-1) * Amb : Amb;
+ tempkm = A->get_blkdim( A, k-1, DIM_m, A->m );
INSERT_TASK_zlacpyx( &options, ChamLower, tempkm, tempkn,
0, A, k-1, k, tileA->ld,
diff --git a/compute/pztpgqrt.c b/compute/pztpgqrt.c
index 7dccc6b2db6f97576a69c6eda3e76525471c1def..7d2ee5712fc391c39c41e019b59d703053a290e8 100644
--- a/compute/pztpgqrt.c
+++ b/compute/pztpgqrt.c
@@ -80,16 +80,16 @@ void chameleon_pztpgqrt( int KT, int L,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == Q1->nt-1 ? Q1->n-k*Q1->nb : Q1->nb;
+ tempkn = Q1->get_blkdim( Q1, k, DIM_n, Q1->n );
/* Equivalent to the tsmqr step on Q1,Q2 */
maxmtk = chameleon_min( Q2->mt, maxmt+k ) - 1;
for (m = maxmtk; m > -1; m--) {
- tempmm = m == Q2->mt-1 ? Q2->m-m*Q2->mb : Q2->mb;
+ tempmm = Q2->get_blkdim( Q2, m, DIM_m, Q2->m );
templm = ((L > 0) && (m == maxmtk)) ? tempmm : 0;
for (n = k; n < Q2->nt; n++) {
- tempnn = n == Q2->nt-1 ? Q2->n-n*Q2->nb : Q2->nb;
+ tempnn = Q2->get_blkdim( Q2, n, DIM_n, Q2->n );
/* TT kernel */
INSERT_TASK_ztpmqrt(
&options,
diff --git a/compute/pztpgqrt_param.c b/compute/pztpgqrt_param.c
index 6da22b663ffc5d7cea537bda9c9129fb0e70759e..b667cc968c5d9c8f74e8e01ad07082c84d057b77 100644
--- a/compute/pztpgqrt_param.c
+++ b/compute/pztpgqrt_param.c
@@ -98,13 +98,13 @@ void chameleon_pztpgqrt_param( int genD, cham_uplo_t uplo, int K,
/* Combine Bottom and Top matrices by merging last pivot with ATop(k,*) */
{
CHAM_desc_t *T = TT;
- int temppm = p == Q->mt-1 ? Q->m - p * Q->mb : Q->mb;
- int tempkn = k == A->nt-1 ? A->n - k * A->nb : A->nb;
+ int temppm = Q->get_blkdim( Q, p, DIM_m, Q->m );
+ int tempkn = A->get_blkdim( A, k, DIM_n, A->n );
int tempnn;
int L = temppm;
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
node = Q->get_rankof( Q, p, n );
RUNTIME_data_migrate( sequence, QTop(k, n), node );
diff --git a/compute/pztpqrt.c b/compute/pztpqrt.c
index 658c75c687e1661ae2c310b2ac1dde00c5477a15..5d595035d3defeaf9145daf381031ac0757293aa 100644
--- a/compute/pztpqrt.c
+++ b/compute/pztpqrt.c
@@ -74,11 +74,11 @@ void chameleon_pztpqrt( int L, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *T,
for (k = 0; k < A->nt; k++) {
RUNTIME_iteration_push(chamctxt, 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;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (m = 0; m < maxmt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
templm = ((L > 0) && (m == maxmt-1)) ? tempmm : 0;
/* TT kernel */
INSERT_TASK_ztpqrt(
@@ -89,7 +89,7 @@ void chameleon_pztpqrt( int L, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *T,
T(m, k) );
for (n = k+1; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztpmqrt(
&options,
ChamLeft, ChamConjTrans,
diff --git a/compute/pztpqrt_param.c b/compute/pztpqrt_param.c
index 4ef4d9770a152037e0510fee009198786a6607d7..f01ed10832492f2440ebb3a6b75d4e578c1a5b44 100644
--- a/compute/pztpqrt_param.c
+++ b/compute/pztpqrt_param.c
@@ -99,8 +99,8 @@ void chameleon_pztpqrt_param( int genD, cham_uplo_t uplo, int K,
/* Combine with ATop and A by merging last pivot with A(k,k) */
{
CHAM_desc_t *T;
- int temppm = p == ATop->mt-1 ? ATop->m - p * ATop->mb : ATop->mb;
- int tempkn = k == ATop->nt-1 ? ATop->n - k * ATop->nb : ATop->nb;
+ int temppm = ATop->get_blkdim( ATop, p, DIM_m, ATop->m );
+ int tempkn = ATop->get_blkdim( ATop, k, DIM_n, ATop->n );
int L, node, tempnn;
T = TT;
@@ -118,7 +118,7 @@ void chameleon_pztpqrt_param( int genD, cham_uplo_t uplo, int K,
T(p, k));
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
node = A->get_rankof( A, p, n );
RUNTIME_data_migrate( sequence, ATop(k, n), node );
diff --git a/compute/pztradd.c b/compute/pztradd.c
index 78d7a3d282e8a4d08bd57b91ffe2935fe337210e..370ccc0923e56f1aacb18973990057e51ad17f84 100644
--- a/compute/pztradd.c
+++ b/compute/pztradd.c
@@ -53,8 +53,8 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
case ChamLower:
if (trans == ChamNoTrans) {
for (n = 0; n < minmn; n++) {
- tempnm = n == B->mt-1 ? B->m-n*B->mb : B->mb;
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnm = B->get_blkdim( B, n, DIM_m, B->m );
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztradd(
&options,
@@ -63,7 +63,7 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
beta, B(n, n));
for (m = n+1; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-B->mb*m : B->nb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_zgeadd(
&options,
@@ -75,8 +75,8 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
}
else {
for (n = 0; n < chameleon_min(B->mt,B->nt); n++) {
- tempnm = n == B->mt-1 ? B->m - n * B->mb : B->mb;
- tempnn = n == B->nt-1 ? B->n - n * B->nb : B->nb;
+ tempnm = B->get_blkdim( B, n, DIM_m, B->m );
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztradd(
&options,
@@ -85,7 +85,7 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
beta, B(n, n));
for (m = n+1; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-B->mb*m : B->nb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_zgeadd(
&options,
@@ -99,8 +99,8 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
case ChamUpper:
if (trans == ChamNoTrans) {
for (m = 0; m < minmn; m++) {
- tempmm = m == B->mt-1 ? B->m - m * B->mb : B->nb;
- tempmn = m == B->nt-1 ? B->n - m * B->nb : B->nb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
+ tempmn = B->get_blkdim( B, m, DIM_n, B->n );
INSERT_TASK_ztradd(
&options,
@@ -109,7 +109,7 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
beta, B(m, m));
for (n = m+1; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n - n * B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgeadd(
&options,
@@ -121,8 +121,8 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
}
else {
for (m = 0; m < chameleon_min(B->mt,B->nt); m++) {
- tempmm = m == B->mt-1 ? B->m-B->mb*m : B->nb;
- tempmn = m == B->nt-1 ? B->n-m*B->nb : B->nb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
+ tempmn = B->get_blkdim( B, m, DIM_n, B->n );
INSERT_TASK_ztradd(
&options,
@@ -131,7 +131,7 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
beta, B(m, m));
for (n = m+1; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgeadd(
&options,
@@ -146,10 +146,10 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
default:
if (trans == ChamNoTrans) {
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-B->mb*m : B->nb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgeadd(
&options,
@@ -161,10 +161,10 @@ void chameleon_pztradd( cham_uplo_t uplo, cham_trans_t trans,
}
else {
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-B->mb*m : B->nb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgeadd(
&options,
diff --git a/compute/pztrmm.c b/compute/pztrmm.c
index 40b0a6f0f856c2c858eb0ee620e157dd484bb6d7..9eeb07dfa0de17c119d54339abd28bc05498d67e 100644
--- a/compute/pztrmm.c
+++ b/compute/pztrmm.c
@@ -57,9 +57,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
if (uplo == ChamUpper) {
if (trans == ChamNoTrans) {
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -68,7 +68,7 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
B(m, n)); /* ldb * tempnn */
for (k = m+1; k < A->mt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgemm(
&options,
trans, ChamNoTrans,
@@ -85,9 +85,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
*/
else {
for (m = B->mt-1; m > -1; m--) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -114,9 +114,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
else {
if (trans == ChamNoTrans) {
for (m = B->mt-1; m > -1; m--) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -141,9 +141,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
*/
else {
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -152,7 +152,7 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
B(m, n)); /* ldb * tempnn */
for (k = m+1; k < A->mt; k++) {
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
INSERT_TASK_zgemm(
&options,
trans, ChamNoTrans,
@@ -173,9 +173,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
if (uplo == ChamUpper) {
if (trans == ChamNoTrans) {
for (n = B->nt-1; n > -1; n--) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -200,9 +200,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
*/
else {
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -211,7 +211,7 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
B(m, n)); /* ldb * tempnn */
for (k = n+1; k < A->mt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, trans,
@@ -230,9 +230,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
else {
if (trans == ChamNoTrans) {
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
@@ -241,7 +241,7 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
B(m, n)); /* ldb * tempnn */
for (k = n+1; k < A->mt; k++) {
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, trans,
@@ -258,9 +258,9 @@ void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo,
*/
else {
for (n = B->nt-1; n > -1; n--) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrmm(
&options,
side, uplo, trans, diag,
diff --git a/compute/pztrsm.c b/compute/pztrsm.c
index 6084fa5a162728bd566f148146c9bbf10b44f6c9..01976543372569da9415ebb9f3bdee2214889b14 100644
--- a/compute/pztrsm.c
+++ b/compute/pztrsm.c
@@ -58,10 +58,10 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
if (uplo == ChamUpper) {
if (trans == ChamNoTrans) {
for (k = 0; k < B->mt; k++) {
- tempkm = k == 0 ? B->m-(B->mt-1)*B->mb : B->mb;
+ tempkm = B->get_blkdim( B, B->mt-1-k, DIM_m, B->m );
lalpha = k == 0 ? alpha : zone;
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -72,7 +72,7 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
RUNTIME_data_flush( sequence, A(B->mt-1-k, B->mt-1-k) );
for (m = k+1; m < B->mt; m++) {
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, ChamNoTrans,
@@ -93,10 +93,10 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
*/
else {
for (k = 0; k < B->mt; k++) {
- tempkm = k == B->mt-1 ? B->m-k*B->mb : B->mb;
+ tempkm = B->get_blkdim( B, k, DIM_m, B->m );
lalpha = k == 0 ? alpha : zone;
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -106,9 +106,9 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
}
RUNTIME_data_flush( sequence, A(k, k) );
for (m = k+1; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgemm(
&options,
trans, ChamNoTrans,
@@ -132,10 +132,10 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
else {
if (trans == ChamNoTrans) {
for (k = 0; k < B->mt; k++) {
- tempkm = k == B->mt-1 ? B->m-k*B->mb : B->mb;
+ tempkm = B->get_blkdim( B, k, DIM_m, B->m );
lalpha = k == 0 ? alpha : zone;
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -145,9 +145,9 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
}
RUNTIME_data_flush( sequence, A(k, k) );
for (m = k+1; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, ChamNoTrans,
@@ -168,10 +168,10 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
*/
else {
for (k = 0; k < B->mt; k++) {
- tempkm = k == 0 ? B->m-(B->mt-1)*B->mb : B->mb;
+ tempkm = B->get_blkdim( B, B->mt-1-k, DIM_m, B->m );
lalpha = k == 0 ? alpha : zone;
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -182,7 +182,7 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
RUNTIME_data_flush( sequence, A(B->mt-1-k, B->mt-1-k) );
for (m = k+1; m < B->mt; m++) {
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgemm(
&options,
trans, ChamNoTrans,
@@ -207,10 +207,10 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
if (uplo == ChamUpper) {
if (trans == ChamNoTrans) {
for (k = 0; k < B->nt; k++) {
- tempkn = k == B->nt-1 ? B->n-k*B->nb : B->nb;
+ tempkn = B->get_blkdim( B, k, DIM_n, B->n );
lalpha = k == 0 ? alpha : zone;
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -220,9 +220,9 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
}
RUNTIME_data_flush( sequence, A(k, k) );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
for (n = k+1; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, ChamNoTrans,
@@ -243,9 +243,9 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
*/
else {
for (k = 0; k < B->nt; k++) {
- tempkn = k == 0 ? B->n-(B->nt-1)*B->nb : B->nb;
+ tempkn = B->get_blkdim( B, B->nt-1-k, DIM_n, B->n );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -277,10 +277,10 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
else {
if (trans == ChamNoTrans) {
for (k = 0; k < B->nt; k++) {
- tempkn = k == 0 ? B->n-(B->nt-1)*B->nb : B->nb;
+ tempkn = B->get_blkdim( B, B->nt-1-k, DIM_n, B->n );
lalpha = k == 0 ? alpha : zone;
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -310,9 +310,9 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
*/
else {
for (k = 0; k < B->nt; k++) {
- tempkn = k == B->nt-1 ? B->n-k*B->nb : B->nb;
+ tempkn = B->get_blkdim( B, k, DIM_n, B->n );
for (m = 0; m < B->mt; m++) {
- tempmm = m == B->mt-1 ? B->m-m*B->mb : B->mb;
+ tempmm = B->get_blkdim( B, m, DIM_m, B->m );
INSERT_TASK_ztrsm(
&options,
side, uplo, trans, diag,
@@ -322,7 +322,7 @@ void chameleon_pztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t trans, c
RUNTIME_data_flush( sequence, A(k, k) );
for (n = k+1; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgemm(
&options,
ChamNoTrans, trans,
diff --git a/compute/pztrsmpl.c b/compute/pztrsmpl.c
index 0b794cd9e74e5391bb81bc36ecaa937922f5352a..27dcb64fe1adcffe13e8ef6c62700129500c628c 100644
--- a/compute/pztrsmpl.c
+++ b/compute/pztrsmpl.c
@@ -41,7 +41,7 @@ void chameleon_pztrsmpl( CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *L, int *IP
int k, m, n;
int tempkm, tempnn, tempkmin, tempmm, tempkn;
- int ib;
+ int ib, K, DIM_k;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -49,13 +49,23 @@ void chameleon_pztrsmpl( CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *L, int *IP
}
RUNTIME_options_init(&options, chamctxt, sequence, request);
+ if ( A->m <= A->n ) {
+ K = A->m;
+ DIM_k = DIM_m;
+ }
+ else {
+ K = A->n;
+ DIM_k = DIM_n;
+ }
+
ib = CHAMELEON_IB;
for (k = 0; k < chameleon_min(A->mt, 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;
- tempkmin = k == chameleon_min(A->mt, A->nt)-1 ? chameleon_min(A->m, A->n)-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
+
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zgessm(
&options,
tempkm, tempnn, tempkmin, ib, L->nb,
@@ -65,9 +75,9 @@ void chameleon_pztrsmpl( CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *L, int *IP
B(k, n));
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < B->nt; n++) {
- tempnn = n == B->nt-1 ? B->n-n*B->nb : B->nb;
+ tempnn = B->get_blkdim( B, n, DIM_n, B->n );
INSERT_TASK_zssssm(
&options,
A->nb, tempnn, tempmm, tempnn, tempkn, ib, L->nb,
diff --git a/compute/pztrtri.c b/compute/pztrtri.c
index 1d6a223ebb1c88449eb72ceb57a7c7fa1e38b7b0..87dd22c336216506a3f3d232af5b4ec4a20a8289 100644
--- a/compute/pztrtri.c
+++ b/compute/pztrtri.c
@@ -55,9 +55,9 @@ void chameleon_pztrtri(cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A,
for (k = 0; k < A->nt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
INSERT_TASK_ztrsm(
&options,
ChamRight, uplo, ChamNoTrans, diag,
@@ -66,7 +66,7 @@ void chameleon_pztrtri(cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A,
A(m, k));
}
for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = 0; n < k; n++) {
INSERT_TASK_zgemm(
&options,
@@ -104,9 +104,9 @@ void chameleon_pztrtri(cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A,
for (k = 0; k < A->mt; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
INSERT_TASK_ztrsm(
&options,
ChamLeft, uplo, ChamNoTrans, diag,
@@ -115,7 +115,7 @@ void chameleon_pztrtri(cham_uplo_t uplo, cham_diag_t diag, CHAM_desc_t *A,
A(k, n));
}
for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ tempnn = A->get_blkdim( A, n, DIM_n, A->n );
for (m = 0; m < k; m++) {
INSERT_TASK_zgemm(
&options,
diff --git a/compute/pzunglq.c b/compute/pzunglq.c
index d0da04302ab14f63c86f12179e4cf836720f268c..f10ae39f2839336bb37946678e20ed9a4e902e53 100644
--- a/compute/pzunglq.c
+++ b/compute/pzunglq.c
@@ -93,15 +93,15 @@ void chameleon_pzunglq( int genD, CHAM_desc_t *A, CHAM_desc_t *Q, CHAM_desc_t *T
for (k = minMT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempAkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempAkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempAkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempAkn = A->get_blkdim( A, k, DIM_n, A->n );
tempkmin = chameleon_min( tempAkn, tempAkm );
- tempkn = k == Q->nt-1 ? Q->n-k*Q->nb : Q->nb;
+ tempkn = Q->get_blkdim( Q, k, DIM_n, Q->n );
for (n = Q->nt-1; n > k; n--) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
RUNTIME_data_migrate( sequence, Q(m, k),
Q->get_rankof( Q, m, n ) );
@@ -121,7 +121,7 @@ void chameleon_pzunglq( int genD, CHAM_desc_t *A, CHAM_desc_t *Q, CHAM_desc_t *T
}
if ( genD ) {
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
ChamUpper, tempkmin, tempDkn,
@@ -136,7 +136,7 @@ void chameleon_pzunglq( int genD, CHAM_desc_t *A, CHAM_desc_t *Q, CHAM_desc_t *T
#endif
}
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
/* Restore the original location of the tiles */
RUNTIME_data_migrate( sequence, Q(m, k),
diff --git a/compute/pzunglq_param.c b/compute/pzunglq_param.c
index 8bd49dbeff65f872de70fb988b7032383610abd8..36f5b58ecdbc7c3c04d77ed7363edd743f0c245d 100644
--- a/compute/pzunglq_param.c
+++ b/compute/pzunglq_param.c
@@ -86,7 +86,7 @@ void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
for (k = K-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
/* Setting the order of the tiles*/
nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
@@ -95,7 +95,7 @@ void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
n = tiles[i];
p = qrtree->currpiv(qrtree, k, n);
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
if( qrtree->gettype(qrtree, k, n) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -108,7 +108,7 @@ void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
T = TT;
}
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
node = Q->get_rankof( Q, m, n );
RUNTIME_data_migrate( sequence, Q(m, p), node );
@@ -131,11 +131,11 @@ void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppn = p == A->nt-1 ? A->n-p*A->nb : A->nb;
+ temppn = A->get_blkdim( A, p, DIM_n, A->n );
tempkmin = chameleon_min(tempkm, temppn);
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
ChamUpper, tempkmin, tempDpn,
@@ -150,7 +150,7 @@ void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
#endif
}
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
RUNTIME_data_migrate( sequence, Q(m, p),
Q->get_rankof( Q, m, p ) );
diff --git a/compute/pzunglqrh.c b/compute/pzunglqrh.c
index ec04ee0e79c0b3b70ffbb1b651154a6c392a16fa..85d184ae8fb4efd829481682183fc1bea40cf7e4 100644
--- a/compute/pzunglqrh.c
+++ b/compute/pzunglqrh.c
@@ -89,15 +89,15 @@ void chameleon_pzunglqrh( int genD, int BS,
for (k = K-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
lastRD = 0;
for (RD = BS; RD < A->nt-k; RD *= 2)
lastRD = RD;
for (RD = lastRD; RD >= BS; RD /= 2) {
for (N = k; N+RD < A->nt; N += 2*RD) {
- tempNRDn = N+RD == A->nt-1 ? A->n-(N+RD)*A->nb : A->nb;
+ tempNRDn = A->get_blkdim( A, N+RD, DIM_n, A->n );
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
node = Q->get_rankof( Q, m, N+RD );
RUNTIME_data_migrate( sequence, Q(m, N), node );
@@ -119,13 +119,13 @@ void chameleon_pzunglqrh( int genD, int BS,
}
}
for (N = k; N < A->nt; N += BS) {
- tempNn = N == A->nt-1 ? A->n-N*A->nb : A->nb;
+ tempNn = A->get_blkdim( A, N, DIM_n, A->n );
tempkmin = chameleon_min(tempkm, tempNn);
for (n = chameleon_min(N+BS, A->nt)-1; n > N; n--) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
node = Q->get_rankof( Q, m, n );
RUNTIME_data_migrate( sequence, Q(m, N), node );
@@ -147,7 +147,7 @@ void chameleon_pzunglqrh( int genD, int BS,
}
if ( genD ) {
- int tempDNn = N == D->nt-1 ? D->n-N*D->nb : D->nb;
+ int tempDNn = D->get_blkdim( D, N, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -163,7 +163,7 @@ void chameleon_pzunglqrh( int genD, int BS,
#endif
}
for (m = k; m < Q->mt; m++) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
RUNTIME_data_migrate( sequence, Q(m, N),
Q->get_rankof( Q, m, N ) );
diff --git a/compute/pzungqr.c b/compute/pzungqr.c
index ac5261230302338bc31233ebf677ee2c24c8bcba..f8f41adc10c6b709ad57ba135771f0f352a9bedc 100644
--- a/compute/pzungqr.c
+++ b/compute/pzungqr.c
@@ -93,14 +93,14 @@ void chameleon_pzungqr( int genD, CHAM_desc_t *A, CHAM_desc_t *Q,
for (k = minMT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempAkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
- tempAkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempAkm = A->get_blkdim( A, k, DIM_m, A->m );
+ tempAkn = A->get_blkdim( A, k, DIM_n, A->n );
tempkmin = chameleon_min( tempAkn, tempAkm );
- tempkm = k == Q->mt-1 ? Q->m-k*Q->mb : Q->mb;
+ tempkm = Q->get_blkdim( Q, k, DIM_m, Q->m );
for (m = Q->mt - 1; m > k; m--) {
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
RUNTIME_data_migrate( sequence, Q(k, n),
Q->get_rankof( Q, m, n ) );
@@ -120,7 +120,7 @@ void chameleon_pzungqr( int genD, CHAM_desc_t *A, CHAM_desc_t *Q,
}
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -136,7 +136,7 @@ void chameleon_pzungqr( int genD, CHAM_desc_t *A, CHAM_desc_t *Q,
#endif
}
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
/* Restore the original location of the tiles */
RUNTIME_data_migrate( sequence, Q(k, n),
diff --git a/compute/pzungqr_param.c b/compute/pzungqr_param.c
index 5bf1a8f566509fc91b13d0b84b0d938659400392..9149ea3b8b41bd6b81747bcf460691b2a749887e 100644
--- a/compute/pzungqr_param.c
+++ b/compute/pzungqr_param.c
@@ -52,13 +52,13 @@ void chameleon_pzungqr_param_step( int genD, cham_uplo_t uplo, int k, int ib,
int tempmm, tempnn, tempkmin, tempkn;
int nbgeqrt, node;
- tempkn = k == A->nt-1 ? A->n - k * A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (i = nbtiles-1; i >= 0; i--) {
m = tiles[i];
p = qrtree->currpiv( qrtree, k, m );
- tempmm = m == Q->mt-1 ? Q->m-m*Q->mb : Q->mb;
+ tempmm = Q->get_blkdim( Q, m, DIM_m, Q->m );
if( qrtree->gettype( qrtree, k, m ) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -77,7 +77,7 @@ void chameleon_pzungqr_param_step( int genD, cham_uplo_t uplo, int k, int ib,
}
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
node = Q->get_rankof( Q, m, n );
RUNTIME_data_migrate( sequence, Q(p, n), node );
@@ -108,11 +108,11 @@ void chameleon_pzungqr_param_step( int genD, cham_uplo_t uplo, int k, int ib,
continue;
}
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
tempkmin = chameleon_min( tempmm, tempkn );
if ( genD ) {
- int tempDmm = m == D->mt-1 ? D->m - m * D->mb : D->mb;
+ int tempDmm = D->get_blkdim( D, m, DIM_m, D->m );
INSERT_TASK_zlacpy(
options,
ChamLower, tempDmm, tempkmin,
@@ -128,7 +128,7 @@ void chameleon_pzungqr_param_step( int genD, cham_uplo_t uplo, int k, int ib,
}
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
/* Restore the original location of the tiles */
RUNTIME_data_migrate( sequence, Q(m, n),
diff --git a/compute/pzungqrrh.c b/compute/pzungqrrh.c
index 3eaac991f5beba28ef97bb4f4a65293082905cf6..5d13e544d5e05b6b617dbf9f21c820812f693b8d 100644
--- a/compute/pzungqrrh.c
+++ b/compute/pzungqrrh.c
@@ -91,15 +91,15 @@ void chameleon_pzungqrrh( int genD, int BS,
for (k = K-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
lastRD = 0;
for (RD = BS; RD < A->mt-k; RD *= 2)
lastRD = RD;
for (RD = lastRD; RD >= BS; RD /= 2) {
for (M = k; M+RD < A->mt; M += 2*RD) {
- tempMRDm = M+RD == A->mt-1 ? A->m-(M+RD)*A->mb : A->mb;
+ tempMRDm = A->get_blkdim( A, M+RD, DIM_m, A->m );
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
node = Q->get_rankof( Q, M+RD, n );
RUNTIME_data_migrate( sequence, Q(M, n), node );
@@ -121,13 +121,13 @@ void chameleon_pzungqrrh( int genD, int BS,
}
}
for (M = k; M < A->mt; M += BS) {
- tempMm = M == A->mt-1 ? A->m-M*A->mb : A->mb;
+ tempMm = A->get_blkdim( A, M, DIM_m, A->m );
tempkmin = chameleon_min(tempMm, tempkn);
for (m = chameleon_min(M+BS, A->mt)-1; m > M; m--) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ tempmm = A->get_blkdim( A, m, DIM_m, A->m );
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
node = Q->get_rankof( Q, m, n );
RUNTIME_data_migrate( sequence, Q(M, n), node );
@@ -148,7 +148,7 @@ void chameleon_pzungqrrh( int genD, int BS,
}
if ( genD ) {
- int tempDMm = M == D->mt-1 ? D->m-M*D->mb : D->mb;
+ int tempDMm = D->get_blkdim( D, M, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
ChamLower, tempDMm, tempkmin,
@@ -163,7 +163,7 @@ void chameleon_pzungqrrh( int genD, int BS,
#endif
}
for (n = k; n < Q->nt; n++) {
- tempnn = n == Q->nt-1 ? Q->n-n*Q->nb : Q->nb;
+ tempnn = Q->get_blkdim( Q, n, DIM_n, Q->n );
/* Restore the original location of the tiles */
RUNTIME_data_migrate( sequence, Q(M, n),
diff --git a/compute/pzunmlq.c b/compute/pzunmlq.c
index c1593bdab0f9a7c35bf89e0e30fa1e7209cf6989..d7043b08b03b5aeafe02362aebd89e35067b22ff 100644
--- a/compute/pzunmlq.c
+++ b/compute/pzunmlq.c
@@ -49,7 +49,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
int k, m, n;
int tempkm, tempkn, tempkmin, tempmm, tempnn;
- int ib, KT, K;
+ int ib, KT, K, DIM_k;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -60,11 +60,13 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
ib = CHAMELEON_IB;
if (A->m > A->n) {
- KT = A->nt;
- K = A->n;
+ KT = A->nt;
+ K = A->n;
+ DIM_k = DIM_n;
} else {
- KT = A->mt;
- K = A->m;
+ KT = A->mt;
+ K = A->m;
+ DIM_k = DIM_m;
}
if ( D == NULL ) {
@@ -100,12 +102,11 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == C->mt - 1 ? C->m - k * C->mb : C->mb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
-
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
if ( genD ) {
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
ChamUpper, tempkmin, tempDkn,
@@ -120,7 +121,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zunmlq(
&options,
side, trans,
@@ -134,9 +135,9 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
RUNTIME_data_flush( sequence, T(k, k) );
for (m = k+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(k, n),
C->get_rankof( C, m, n ) );
@@ -172,14 +173,13 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == C->mt - 1 ? C->m - k * C->mb : C->mb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
-
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
for (m = C->mt-1; m > k; m--) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(k, n),
C->get_rankof( C, m, n ) );
@@ -200,7 +200,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
}
if ( genD ) {
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
ChamUpper, tempkmin, tempDkn,
@@ -215,7 +215,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(k, n),
C->get_rankof( C, k, n ) );
@@ -242,13 +242,13 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == C->nt - 1 ? C->n - k * C->nb : C->nb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
for (n = C->nt-1; n > k; n--) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, k),
C->get_rankof( C, m, n ) );
@@ -269,7 +269,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
}
if ( genD ) {
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
ChamUpper, tempkmin, tempDkn,
@@ -284,7 +284,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, k),
C->get_rankof( C, m, k ) );
@@ -311,11 +311,11 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == C->nt - 1 ? C->n - k * C->nb : C->nb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
if ( genD ) {
- int tempDkn = k == D->nt-1 ? D->n-k*D->nb : D->nb;
+ int tempDkn = D->get_blkdim( D, k, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -331,7 +331,7 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zunmlq(
&options,
side, trans,
@@ -345,9 +345,9 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
RUNTIME_data_flush( sequence, T(k, k) );
for (n = k+1; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, k),
C->get_rankof( C, m, n ) );
diff --git a/compute/pzunmlq_param.c b/compute/pzunmlq_param.c
index 94f73255fae1919344f054212a1c6c2b79c2d99b..80d9303fe6442e25a38f2909fb1e5660db833751 100644
--- a/compute/pzunmlq_param.c
+++ b/compute/pzunmlq_param.c
@@ -92,18 +92,18 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt - 1 ? A->m - k * A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
T = TS;
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppm = p == C->mt-1 ? C->m - p * C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -119,7 +119,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zunmlq(
&options, side, trans,
temppm, tempnn, tempkmin, ib, T->nb,
@@ -138,7 +138,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
m = tiles[i];
p = qrtree->currpiv(qrtree, k, m);
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if( qrtree->gettype(qrtree, k, m) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -151,7 +151,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
T = TT;
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -185,7 +185,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
/* Setting the order of the tiles*/
nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
@@ -194,7 +194,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
m = tiles[i];
p = qrtree->currpiv(qrtree, k, m);
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if( qrtree->gettype(qrtree, k, m) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -207,7 +207,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
T = TT;
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -229,12 +229,12 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppm = p == C->mt-1 ? C->m-p*C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -251,7 +251,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(p, n),
C->get_rankof( C, p, n ) );
@@ -279,7 +279,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
/* Setting the order of the tiles*/
nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
@@ -288,7 +288,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
n = tiles[i];
p = qrtree->currpiv(qrtree, k, n);
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if( qrtree->gettype(qrtree, k, n) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -302,7 +302,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -324,11 +324,11 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppn = p == C->nt-1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -345,7 +345,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, p),
C->get_rankof( C, m, p ) );
@@ -370,17 +370,17 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
T = TS;
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppn = p == C->nt - 1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -397,7 +397,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zunmlq(
&options, side, trans,
tempmm, temppn, tempkmin, ib, T->nb,
@@ -416,7 +416,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
n = tiles[i];
p = qrtree->currpiv(qrtree, k, n);
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if( qrtree->gettype(qrtree, k, n) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -430,7 +430,7 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
diff --git a/compute/pzunmlqrh.c b/compute/pzunmlqrh.c
index cdf7dd9b8ee0dfb2c0489b607bda2de8c901c6f8..0d8608963e1767180d160356dbdf1cafdb35b20b 100644
--- a/compute/pzunmlqrh.c
+++ b/compute/pzunmlqrh.c
@@ -95,17 +95,17 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt - 1 ? A->m - k * A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for (p = k; p < C->mt; p += BS) {
- temppm = p == C->mt-1 ? C->m - p * C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -121,7 +121,7 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zunmlq(
&options,
side, trans,
@@ -134,10 +134,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(k, p) );
for (m = p+1; m < chameleon_min(p+BS, C->mt); m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -160,10 +160,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p+RD < C->mt; p += 2*RD) {
m = p+RD;
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -200,7 +200,7 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
lastRD = 0;
for (RD = BS; RD < C->mt-k; RD *= 2)
@@ -209,10 +209,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p+RD < C->mt; p += 2*RD) {
m = p+RD;
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -234,10 +234,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p < C->mt; p += BS) {
for (m = chameleon_min(p+BS, C->mt)-1; m > p; m--) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -256,11 +256,11 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(k, m) );
}
- temppm = p == C->mt-1 ? C->m-p*C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -277,7 +277,7 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(p, n),
C->get_rankof( C, p, n ) );
@@ -304,7 +304,7 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
lastRD = 0;
for (RD = BS; RD < C->nt-k; RD *= 2)
lastRD = RD;
@@ -312,10 +312,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p+RD < C->nt; p += 2*RD) {
n = p+RD;
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -338,10 +338,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (n = chameleon_min(p+BS, C->nt)-1; n > p; n--) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -360,11 +360,11 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(k, n) );
}
- temppn = p == C->nt-1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -381,7 +381,7 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, p),
C->get_rankof( C, m, p ) );
@@ -406,14 +406,14 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+ tempkm = A->get_blkdim( A, k, DIM_m, A->m );
for (p = k; p < C->nt; p += BS) {
- temppn = p == C->nt - 1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkm );
if ( genD ) {
- int tempDpn = p == D->nt-1 ? D->n-p*D->nb : D->nb;
+ int tempDpn = D->get_blkdim( D, p, DIM_n, D->n );
INSERT_TASK_zlacpy(
&options,
@@ -430,7 +430,7 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zunmlq(
&options, side, trans,
tempmm, temppn, tempkmin, ib, T->nb,
@@ -442,9 +442,9 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(k, p) );
for (n = p+1; n < chameleon_min(p+BS, C->nt); n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -466,10 +466,10 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (RD = BS; RD < C->nt-k; RD *= 2) {
for (p = k; p+RD < C->nt; p += 2*RD) {
n = p + RD;
- tempnn = n == C->mt-1 ? C->m-n*C->mb : C->mb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
diff --git a/compute/pzunmqr.c b/compute/pzunmqr.c
index 3275a87dbbdbeb37a0b82db7abf7cf0a11d86674..821b11d264fcd61966a49fca78c32e195108c219 100644
--- a/compute/pzunmqr.c
+++ b/compute/pzunmqr.c
@@ -49,7 +49,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
int k, m, n;
int tempkm, tempkn, tempkmin, tempmm, tempnn;
- int ib, KT, K;
+ int ib, KT, K, DIM_k;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -60,11 +60,13 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
ib = CHAMELEON_IB;
if (A->m > A->n) {
- KT = A->nt;
- K = A->n;
+ KT = A->nt;
+ K = A->n;
+ DIM_k = DIM_n;
} else {
- KT = A->mt;
- K = A->m;
+ KT = A->mt;
+ K = A->m;
+ DIM_k = DIM_m;
}
if ( D == NULL ) {
@@ -100,12 +102,11 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == C->mt - 1 ? C->m - k * C->mb : C->mb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
-
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -121,7 +122,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zunmqr(
&options,
side, trans,
@@ -135,9 +136,9 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
RUNTIME_data_flush( sequence, T(k, k) );
for (m = k+1; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(k, n),
C->get_rankof( C, m, n ) );
@@ -173,14 +174,13 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == C->mt - 1 ? C->m - k * C->mb : C->mb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
-
+ tempkm = C->get_blkdim( C, k, DIM_m, C->m );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
for (m = C->mt-1; m > k; m--) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(k, n),
C->get_rankof( C, m, n ) );
@@ -200,7 +200,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
}
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -216,7 +216,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(k, n),
C->get_rankof( C, k, n ) );
@@ -243,13 +243,13 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == C->nt - 1 ? C->n - k * C->nb : C->nb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
for (n = C->nt-1; n > k; n--) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, k),
C->get_rankof( C, m, n ) );
@@ -270,7 +270,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
}
if ( genD ) {
- int tempDkm = k == D->mt-1 ? D->m-k*D->mb : D->mb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -286,7 +286,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, k),
C->get_rankof( C, m, k ) );
@@ -313,11 +313,11 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == C->nt - 1 ? C->n - k * C->nb : C->nb;
- tempkmin = k == KT - 1 ? K - k * A->nb : A->nb;
+ tempkn = C->get_blkdim( C, k, DIM_n, C->n );
+ tempkmin = A->get_blkdim( A, k, DIM_k, K );
if ( genD ) {
- int tempDkm = k == D->mt - 1 ? D->m - k * D->mb : D->mb;
+ int tempDkm = D->get_blkdim( D, k, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -333,7 +333,7 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
#endif
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zunmqr(
&options,
side, trans,
@@ -347,9 +347,9 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
RUNTIME_data_flush( sequence, T(k, k) );
for (n = k+1; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, k),
C->get_rankof( C, m, n ) );
diff --git a/compute/pzunmqr_param.c b/compute/pzunmqr_param.c
index 47904aaaf165c8ed104b86b30b27b710531e19f4..72dd950a1e5e9b8848f9f88ceabfce4c08a6673a 100644
--- a/compute/pzunmqr_param.c
+++ b/compute/pzunmqr_param.c
@@ -92,18 +92,18 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt - 1 ? A->n - k * A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
T = TS;
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppm = p == C->mt-1 ? C->m - p * C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -119,7 +119,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zunmqr(
&options, side, trans,
temppm, tempnn, tempkmin, ib, T->nb,
@@ -138,7 +138,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
m = tiles[i];
p = qrtree->currpiv(qrtree, k, m);
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if( qrtree->gettype(qrtree, k, m) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -151,7 +151,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
T = TT;
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -185,7 +185,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
/* Setting the order of the tiles*/
nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
@@ -194,7 +194,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
m = tiles[i];
p = qrtree->currpiv(qrtree, k, m);
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
if( qrtree->gettype(qrtree, k, m) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -207,7 +207,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
T = TT;
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -229,12 +229,12 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppm = p == C->mt-1 ? C->m-p*C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -251,7 +251,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(p, n),
C->get_rankof( C, p, n ) );
@@ -278,7 +278,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n - k * A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
/* Setting the order of the tiles*/
nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
@@ -287,7 +287,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
n = tiles[i];
p = qrtree->currpiv(qrtree, k, n);
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if( qrtree->gettype(qrtree, k, n) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -301,7 +301,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -323,11 +323,11 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppn = p == C->nt-1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min(temppn, tempkn);
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -344,7 +344,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, p),
C->get_rankof( C, m, p ) );
@@ -369,17 +369,17 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
T = TS;
for (i = 0; i < qrtree->getnbgeqrf(qrtree, k); i++) {
p = qrtree->getm(qrtree, k, i);
- temppn = p == C->nt - 1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -396,7 +396,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zunmqr(
&options, side, trans,
tempmm, temppn, tempkmin, ib, T->nb,
@@ -415,7 +415,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
n = tiles[i];
p = qrtree->currpiv(qrtree, k, n);
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
if( qrtree->gettype(qrtree, k, n) == LIBHQR_KILLED_BY_TS ) {
/* TS kernel */
@@ -429,7 +429,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
diff --git a/compute/pzunmqrrh.c b/compute/pzunmqrrh.c
index 8eda28189e83810f173e2634160905c8f5101661..be0f0e5bffd285cb5d03c33970b565475102ddfd 100644
--- a/compute/pzunmqrrh.c
+++ b/compute/pzunmqrrh.c
@@ -95,16 +95,16 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt - 1 ? A->n - k * A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (p = k; p < C->mt; p += BS) {
- temppm = p == C->mt-1 ? C->m - p * C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -120,7 +120,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
#endif
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
INSERT_TASK_zunmqr(
&options,
side, trans,
@@ -133,10 +133,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(p, k) );
for (m = p+1; m < chameleon_min(p+BS, C->mt); m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -159,10 +159,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p+RD < C->mt; p += 2*RD) {
m = p+RD;
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -198,7 +198,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
lastRD = 0;
for (RD = BS; RD < C->mt-k; RD *= 2)
lastRD = RD;
@@ -206,10 +206,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p+RD < C->mt; p += 2*RD) {
m = p+RD;
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -231,10 +231,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p < C->mt; p += BS) {
for (m = chameleon_min(p+BS, C->mt)-1; m > p; m--) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(p, n), node );
@@ -253,11 +253,11 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(m, k) );
}
- temppm = p == C->mt-1 ? C->m-p*C->mb : C->mb;
+ temppm = C->get_blkdim( C, p, DIM_m, C->m );
tempkmin = chameleon_min( temppm, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -274,7 +274,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
for (n = 0; n < C->nt; n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
RUNTIME_data_migrate( sequence, C(p, n),
C->get_rankof( C, p, n ) );
@@ -301,7 +301,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = KT-1; k >= 0; k--) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n - k * A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
lastRD = 0;
for (RD = BS; RD < C->nt-k; RD *= 2)
@@ -310,10 +310,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (p = k; p+RD < C->nt; p += 2*RD) {
n = p+RD;
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -336,10 +336,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (n = chameleon_min(p+BS, C->nt)-1; n > p; n--) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -358,11 +358,11 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(n, k) );
}
- temppn = p == C->nt-1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -379,7 +379,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
RUNTIME_data_migrate( sequence, C(m, p),
C->get_rankof( C, m, p ) );
@@ -404,15 +404,15 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (k = 0; k < KT; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkn = k == A->nt-1 ? A->n-k*A->nb : A->nb;
+ tempkn = A->get_blkdim( A, k, DIM_n, A->n );
for (p = k; p < C->nt; p += BS) {
- temppn = p == C->nt - 1 ? C->n - p * C->nb : C->nb;
+ temppn = C->get_blkdim( C, p, DIM_n, C->n );
tempkmin = chameleon_min( temppn, tempkn );
if ( genD ) {
- int tempDpm = p == D->mt-1 ? D->m-p*D->mb : D->mb;
+ int tempDpm = D->get_blkdim( D, p, DIM_m, D->m );
INSERT_TASK_zlacpy(
&options,
@@ -429,7 +429,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
INSERT_TASK_zunmqr(
&options, side, trans,
tempmm, temppn, tempkmin, ib, T->nb,
@@ -441,9 +441,9 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
RUNTIME_data_flush( sequence, T(p, k) );
for (n = p+1; n < chameleon_min(p+BS, C->nt); n++) {
- tempnn = n == C->nt-1 ? C->n-n*C->nb : C->nb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
@@ -465,10 +465,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
for (RD = BS; RD < C->nt-k; RD *= 2) {
for (p = k; p+RD < C->nt; p += 2*RD) {
n = p + RD;
- tempnn = n == C->mt-1 ? C->m-n*C->mb : C->mb;
+ tempnn = C->get_blkdim( C, n, DIM_n, C->n );
for (m = 0; m < C->mt; m++) {
- tempmm = m == C->mt-1 ? C->m-m*C->mb : C->mb;
+ tempmm = C->get_blkdim( C, m, DIM_m, C->m );
node = C->get_rankof( C, m, n );
RUNTIME_data_migrate( sequence, C(m, p), node );
diff --git a/compute/zgemm_batch.c b/compute/zgemm_batch.c
index ab1ed4062062172c2d63b517485246763ac1e821..36d77960a50ee789516f82db85c672a4652a4ddb 100644
--- a/compute/zgemm_batch.c
+++ b/compute/zgemm_batch.c
@@ -61,13 +61,13 @@ zgemm_batch_cpu( void *op_args,
tileC = va_arg(ap, CHAM_tile_t *);
va_end(ap);
- tempmm = m == descC->mt-1 ? descC->m - m * descC->mb : descC->mb;
- tempnn = n == descC->nt-1 ? descC->n - n * descC->nb : descC->nb;
+ tempmm = descC->get_blkdim( descC, m, DIM_m, descC->m );
+ tempnn = descC->get_blkdim( descC, n, DIM_n, descC->n );
if ( args->transA == ChamNoTrans ) {
- tempkk = n == descA->nt-1 ? descA->n - n * descA->nb : descA->nb;
+ tempkk = descA->get_blkdim( descA, n, DIM_n, descA->n );
}
else {
- tempkk = m == descA->mt-1 ? descA->m - m * descA->mb : descA->mb;
+ tempkk = descA->get_blkdim( descA, m, DIM_m, descA->m );
}
TCORE_zgemm(
@@ -112,13 +112,13 @@ zgemm_batch_cuda( cublasHandle_t handle, void *op_args,
tileC = va_arg(ap, CHAM_tile_t *);
va_end(ap);
- tempmm = m == descC->mt-1 ? descC->m - m * descC->mb : descC->mb;
- tempnn = n == descC->nt-1 ? descC->n - n * descC->nb : descC->nb;
+ tempmm = descC->get_blkdim( descC, m, DIM_m, descC->m );
+ tempnn = descC->get_blkdim( descC, n, DIM_n, descC->n );
if ( args->transA == ChamNoTrans ) {
- tempkk = n == descA->nt-1 ? descA->n - n * descA->nb : descA->nb;
+ tempkk = descA->get_blkdim( descA, n, DIM_n, descA->n );
}
else {
- tempkk = m == descA->mt-1 ? descA->m - m * descA->mb : descA->mb;
+ tempkk = descA->get_blkdim( descA, m, DIM_m, descA->m );
}
CUDA_zgemm( args->transA, args->transB, tempmm, tempnn, tempkk,
diff --git a/compute/zherk_batch.c b/compute/zherk_batch.c
index bff967e9f0afe0b4d2242ef8184a0574f3f96501..12eeda0823406811792995d49e0cd3fd842ce1cd 100644
--- a/compute/zherk_batch.c
+++ b/compute/zherk_batch.c
@@ -57,12 +57,12 @@ zherk_batch_cpu( void *op_args,
tileC = va_arg(ap, CHAM_tile_t *);
va_end(ap);
- tempnn = n == descC->nt-1 ? descC->n - n * descC->nb : descC->nb;
+ tempnn = descC->get_blkdim( descC, n, DIM_n, descC->n );
if ( args->trans == ChamNoTrans ) {
- tempkk = n == descA->nt-1 ? descA->n - n * descA->nb : descA->nb;
+ tempkk = descA->get_blkdim( descA, n, DIM_n, descA->n );
}
else {
- tempkk = m == descA->mt-1 ? descA->m - m * descA->mb : descA->mb;
+ tempkk = descA->get_blkdim( descA, m, DIM_m, descA->m );
}
TCORE_zherk(
@@ -102,12 +102,12 @@ zherk_batch_cuda( cublasHandle_t handle, void *op_args,
tileC = va_arg(ap, CHAM_tile_t *);
va_end(ap);
- tempnn = n == descC->nt-1 ? descC->n - n * descC->nb : descC->nb;
+ tempnn = descC->get_blkdim( descC, n, DIM_n, descC->n );
if ( args->trans == ChamNoTrans ) {
- tempkk = n == descA->nt-1 ? descA->n - n * descA->nb : descA->nb;
+ tempkk = descA->get_blkdim( descA, n, DIM_n, descA->n );
}
else {
- tempkk = m == descA->mt-1 ? descA->m - m * descA->mb : descA->mb;
+ tempkk = descA->get_blkdim( descA, m, DIM_m, descA->m );
}
CUDA_zherk( args->uplo, args->trans, tempnn, tempkk,
diff --git a/compute/zplghe_batch.c b/compute/zplghe_batch.c
index 92a0a071ee77dbe136e91241a0226fe0e4d21f2f..e876df684909931eacd188099b862a3120b790cd 100644
--- a/compute/zplghe_batch.c
+++ b/compute/zplghe_batch.c
@@ -43,7 +43,7 @@ zplghe_batch_cpu( void *op_args,
fprintf( stderr, "zplghe_batch_cpu: requires two pieces of data and %d have been given\n", ndata );
}
- tempnn = n == descA->nt-1 ? descA->n - n * descA->nb : descA->nb;
+ tempnn = descA->get_blkdim( descA, n, DIM_n, descA->n );
/**
* Let's fo the math to give the right bigM:
diff --git a/compute/zpotrf_batch.c b/compute/zpotrf_batch.c
index 983a9d92fa3ba72645b47d0dbe78dd1346ad59f2..dd6a3f440801d2efe6ceb867323aa1bdac2692dc 100644
--- a/compute/zpotrf_batch.c
+++ b/compute/zpotrf_batch.c
@@ -38,7 +38,7 @@ zpotrf_batch_cpu( void *op_args,
fprintf( stderr, "zpotrf_batch_cpu: requires two pieces of data and %d have been given\n", ndata );
}
- tempnn = n == descA->nt-1 ? descA->n - n * descA->nb : descA->nb;
+ tempnn = descA->get_blkdim( descA, n, DIM_n, descA->n );
TCORE_zpotrf(
luplo, tempnn, tileA, &info );
diff --git a/compute/zprint.c b/compute/zprint.c
index e17a406a1b5adcdf80ff90e2d49a87ec62c104eb..065beac235802f252590db1289b210e293da3bf4 100644
--- a/compute/zprint.c
+++ b/compute/zprint.c
@@ -34,8 +34,8 @@ zprint_cpu( void *op_args,
struct zprint_args_s *options = (struct zprint_args_s *)op_args;
CHAMELEON_Complex64_t *A = CHAM_tile_get_ptr( tileA );
- int tempmm = m == descA->mt-1 ? descA->m-m*descA->mb : descA->mb;
- int tempnn = n == descA->nt-1 ? descA->n-n*descA->nb : descA->nb;
+ int tempmm = descA->get_blkdim( descA, m, DIM_m, descA->m );
+ int tempnn = descA->get_blkdim( descA, n, DIM_n, descA->n );
int lda = tileA->ld;
if ( ndata > 1 ) {
diff --git a/compute/ztrsm_batch.c b/compute/ztrsm_batch.c
index 9d426737b3bc3678892e452173ae05cfe5a26298..ad8e855c1137ef69095f4638ada5dc5b66abd7b0 100644
--- a/compute/ztrsm_batch.c
+++ b/compute/ztrsm_batch.c
@@ -58,8 +58,8 @@ ztrsm_batch_cpu( void *op_args,
tileB = va_arg(ap, CHAM_tile_t *);
va_end(ap);
- tempmm = m == descB->mt-1 ? descB->m - m * descB->mb : descB->mb;
- tempnn = n == descB->nt-1 ? descB->n - n * descB->nb : descB->nb;
+ tempmm = descB->get_blkdim( descB, m, DIM_m, descB->m );
+ tempnn = descB->get_blkdim( descB, n, DIM_n, descB->n );
TCORE_ztrsm(
args->side, args->uplo, args->transA, args->diag,
@@ -103,8 +103,8 @@ ztrsm_batch_cuda( cublasHandle_t handle, void *op_args,
assert( tileA->format & CHAMELEON_TILE_FULLRANK );
assert( tileB->format & CHAMELEON_TILE_FULLRANK );
- tempmm = m == descB->mt-1 ? descB->m - m * descB->mb : descB->mb;
- tempnn = n == descB->nt-1 ? descB->n - n * descB->nb : descB->nb;
+ tempmm = descB->get_blkdim( descB, m, DIM_m, descB->m );
+ tempnn = descB->get_blkdim( descB, n, DIM_n, descB->n );
CUDA_ztrsm(
args->side, args->uplo, args->transA, args->diag,
diff --git a/control/descriptor.c b/control/descriptor.c
index ff2732b2ccd670b250039ec9bdc899043964c589..4a1f2ef0073d1d8b8ece092ac856809b9031a4ed 100644
--- a/control/descriptor.c
+++ b/control/descriptor.c
@@ -222,6 +222,7 @@ int chameleon_desc_init_internal( CHAM_desc_t *desc, const char *name, void *mat
/* If one of the function get_* is NULL, we switch back to the default */
desc->get_blktile = chameleon_desc_gettile;
+ desc->get_blkdim = chameleon_getblkdim;
/* Data addresses */
if ( get_blkaddr ) {
diff --git a/control/descriptor_rec.c b/control/descriptor_rec.c
index 65d4f6148ccd279daa50178c4ea5a7b70ded3a5e..fed12a95ba7cd59aebd80063525dcb7e8be5c344 100644
--- a/control/descriptor_rec.c
+++ b/control/descriptor_rec.c
@@ -73,8 +73,8 @@ chameleon_recdesc_create( const char *name, CHAM_desc_t **descptr, void *mat, ch
char *subname;
tile = desc->get_blktile( desc, m, n );
- tempmm = m == desc->mt-1 ? desc->m - m * desc->mb : desc->mb;
- tempnn = n == desc->nt-1 ? desc->n - n * desc->nb : desc->nb;
+ tempmm = desc->get_blkdim( desc, m, DIM_m, desc->m );
+ tempnn = desc->get_blkdim( desc, n, DIM_n, desc->n );
chameleon_asprintf( &subname, "%s[%d,%d]", name, m, n );
diff --git a/example/lapack_to_chameleon/step7.h b/example/lapack_to_chameleon/step7.h
index 7622f5ddbbe669e4cdf2810d761c0aa74181f2fb..53331515b35d2ca9caa65ba202d1c063439ea322 100644
--- a/example/lapack_to_chameleon/step7.h
+++ b/example/lapack_to_chameleon/step7.h
@@ -83,8 +83,8 @@ static int Cham_build_plgsy_cpu( void *op_args, cham_uplo_t uplo, int m, int n,
int tempmm, tempnn;
/* Get the dimension of the tile */
- tempmm = (m == (descA->mt-1)) ? (descA->m - m * descA->mb) : descA->mb;
- tempnn = (n == (descA->nt-1)) ? (descA->n - n * descA->nb) : descA->nb;
+ tempmm = descA->get_blkdim( descA, m, DIM_m, descA->m );
+ tempnn = descA->get_blkdim( descA, n, DIM_n, descA->n );
/* fill the tile with the coreblas function plgsy = random SPD matrix generator */
TCORE_dplgsy( data->bump, tempmm, tempnn, tileA,
@@ -101,8 +101,8 @@ static int Cham_build_plrnt_cpu( void *op_args, cham_uplo_t uplo, int m, int n,
int tempmm, tempnn;
/* Get the dimension of the tile */
- tempmm = (m == (descA->mt-1)) ? (descA->m - m * descA->mb) : descA->mb;
- tempnn = (n == (descA->nt-1)) ? (descA->n - n * descA->nb) : descA->nb;
+ tempmm = descA->get_blkdim( descA, m, DIM_m, descA->m );
+ tempnn = descA->get_blkdim( descA, n, DIM_n, descA->n );
/* fill the tile with the coreblas function plrnt = random general matrix generator */
TCORE_dplrnt( tempmm, tempnn, tileA,
diff --git a/include/chameleon/constants.h b/include/chameleon/constants.h
index d3cfb96208b9478c878d20b987dca3305e4bd7c6..09bfc942f5fc1d2e829616b61b79b878324bf61c 100644
--- a/include/chameleon/constants.h
+++ b/include/chameleon/constants.h
@@ -113,6 +113,14 @@ typedef enum chameleon_flttype_e {
#define ChamConvertRealHalfToSingle ChamConvert( ChamRealHalf, ChamRealSingle )
#define ChamConvertRealHalfToHalf ChamConvert( ChamRealHalf, ChamRealHalf )
+/**
+ * @brief Matrix dimensions naming
+ */
+typedef enum chameleon_dim_e {
+ DIM_m = 0,
+ DIM_n = 1,
+} cham_dim_t;
+
/**
* @brief Matrix tile storage
*/
diff --git a/include/chameleon/descriptor_helpers.h b/include/chameleon/descriptor_helpers.h
index 7bfdeb77ba565c2da0f9668f5d7d347e6e44112c..9e60ef27dda1d76e3af3c8f5fe3a2b427ae719c5 100644
--- a/include/chameleon/descriptor_helpers.h
+++ b/include/chameleon/descriptor_helpers.h
@@ -88,8 +88,92 @@ int chameleon_getblkldd_cm ( const CHAM_desc_t *A, int m );
int chameleon_getblkldd_ccrb( const CHAM_desc_t *A, int m );
/**
* @}
+ * @name Tile dimensions computation in algorithms
+ * @{
*/
+/**
+ *
+ * @ingroup Descriptor
+ *
+ * @brief Return tile dimension along the m dimension with regular tile sizes
+ *
+ * @param[in] A
+ * The chameleon descriptor for which to compute the size
+ *
+ * @param[in] m
+ * The row index of the tile
+ *
+ * @param[in] lm
+ * The matrix row dimension against which to compute the size
+ *
+ * @retval The dimension of the tile along the row/first dimension with a limit on lm
+ *
+ */
+static inline int
+chameleon_getblkdim_m( const CHAM_desc_t *A, int m, int lm )
+{
+ return (((m + 1) * A->mb) > lm ) ? lm - m * A->mb : A->mb;
+}
+
+/**
+ *
+ * @ingroup Descriptor
+ *
+ * @brief Return tile dimension along the n dimension with regular tile sizes
+ *
+ * @param[in] A
+ * The chameleon descriptor for which to compute the size
+ *
+ * @param[in] n
+ * The column index of the tile
+ *
+ * @param[in] ln
+ * The matrix column dimension against which to compute the size
+ *
+ * @retval The dimension of the tile along the column/second dimension with a limit on ln
+ *
+ */
+static inline int
+chameleon_getblkdim_n( const CHAM_desc_t *A, int n, int ln )
+{
+ return (((n + 1) * A->nb) > ln ) ? ln - n * A->nb : A->nb;
+}
+
+/**
+ *
+ * @ingroup Descriptor
+ *
+ * @brief Return tile dimension along the dim dimension with regular tile sizes
+ *
+ * @param[in] A
+ * The chameleon descriptor for which to compute the size
+ *
+ * @param[in] m
+ * The index of the tile in the given dimension
+ *
+ * @param[in] dim
+ * The dimension on which to compute the size
+ *
+ * @param[in] lm
+ * The matrix dimension along the chosen dim.
+ *
+ * @retval The dimension of the tile along the dim dimension with a limit on lm
+ *
+ */
+static inline int
+chameleon_getblkdim( const CHAM_desc_t *A, int m, cham_dim_t dim, int lm )
+{
+ if ( dim == 0 ) {
+ return chameleon_getblkdim_m( A, m, lm );
+ }
+ else {
+ return chameleon_getblkdim_n( A, m, lm );
+ }
+}
+/**
+ * @}
+ */
#ifdef __cplusplus
}
#endif
diff --git a/include/chameleon/struct.h b/include/chameleon/struct.h
index e61f95b18ba8a9e2f5c204bc1aff1428e24e6620..d369325007052f3d65ea4ed33033f49f8aac0e45 100644
--- a/include/chameleon/struct.h
+++ b/include/chameleon/struct.h
@@ -80,6 +80,7 @@ typedef struct chameleon_desc_s CHAM_desc_t;
typedef void* (*blkaddr_fct_t) ( const CHAM_desc_t*, int, int );
typedef int (*blkldd_fct_t) ( const CHAM_desc_t*, int );
+typedef int (*blkdim_fct_t) ( const CHAM_desc_t*, int, cham_dim_t, int );
typedef int (*blkrankof_fct_t) ( const CHAM_desc_t*, int, int );
typedef int (*datadist_access_fct_t)( const CHAM_desc_t*, int, ... );
typedef CHAM_tile_t* (*blktile_fct_t) ( const CHAM_desc_t*, int, int );
@@ -116,11 +117,12 @@ void chameleon_desc_set_datadist( CHAM_desc_t *to, cham_data_dist_t *from );
struct chameleon_desc_s {
const char *name;
- blktile_fct_t get_blktile; /**> function to get chameleon tiles address */
- blkaddr_fct_t get_blkaddr; /**> function to get chameleon tiles address */
- blkldd_fct_t get_blkldd; /**> function to get chameleon tiles leading dimension */
- blkrankof_fct_t get_rankof; /**> function to get chameleon tiles MPI rank */
- blkrankof_fct_t get_rankof_init; /**> function to get chameleon tiles MPI rank */
+ blktile_fct_t get_blktile; /**> function to get chameleon tiles address */
+ blkaddr_fct_t get_blkaddr; /**> function to get chameleon tiles address */
+ blkldd_fct_t get_blkldd; /**> function to get chameleon tiles leading dimension */
+ blkdim_fct_t get_blkdim; /**> function to get chameleon tiles dimension within algorithms */
+ blkrankof_fct_t get_rankof; /**> function to get chameleon tiles MPI rank */
+ blkrankof_fct_t get_rankof_init; /**> function to get chameleon tiles MPI rank */
void* get_rankof_init_arg;
CHAM_tile_t *tiles; /**> pointer to the array of tiles descriptors */