From 6e31c9b960c999ed6cc10a89e184e22ef0c73f98 Mon Sep 17 00:00:00 2001
From: Mathieu Faverge <mathieu.faverge@inria.fr>
Date: Fri, 6 May 2022 15:49:45 +0200
Subject: [PATCH] qr/lq: create a step subfunction in all qr/lq factorization
function
---
compute/pzgelqf.c | 169 ++++++++++++++-----------
compute/pzgelqf_param.c | 267 ++++++++++++++++++++++------------------
compute/pzgeqrf.c | 180 ++++++++++++++++-----------
compute/pzgeqrf_param.c | 1 -
control/compute_z.h | 13 +-
5 files changed, 364 insertions(+), 266 deletions(-)
diff --git a/compute/pzgelqf.c b/compute/pzgelqf.c
index 32346a6e8..9a6a4f3da 100644
--- a/compute/pzgelqf.c
+++ b/compute/pzgelqf.c
@@ -33,7 +33,99 @@
#define D(k) D, k, k
/**
- * Parallel tile LQ factorization - dynamic scheduling
+ * Parallel tile QR factorization (reduction Householder) - dynamic scheduling
+ *
+ * @param[in] genD
+ * Indicate if copies of the geqrt tiles must be done to speedup
+ * computations in updates. genD is considered only if D is not NULL.
+ */
+int chameleon_pzgelqf_step( int genD, int k, int ib,
+ CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D,
+ RUNTIME_option_t *options, RUNTIME_sequence_t *sequence )
+{
+ 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;
+ INSERT_TASK_zgelqt(
+ options,
+ tempkm, tempkn, ib, T->nb,
+ A(k, k),
+ 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;
+ INSERT_TASK_zlacpy(
+ options,
+ ChamUpper, tempDkm, tempDkn,
+ A(k, k),
+ D(k) );
+#if defined(CHAMELEON_USE_CUDA)
+ INSERT_TASK_zlaset(
+ options,
+ ChamLower, tempDkm, tempDkn,
+ 0., 1.,
+ D(k) );
+#endif
+ }
+
+ for (m = k+1; m < A->mt; m++) {
+ tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ INSERT_TASK_zunmlq(
+ options,
+ ChamRight, ChamConjTrans,
+ tempmm, tempkn, tempkn, ib, T->nb,
+ D(k),
+ T(k, k),
+ A(m, k));
+ }
+ RUNTIME_data_flush( sequence, D(k) );
+ 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;
+
+ RUNTIME_data_migrate( sequence, A(k, k),
+ A->get_rankof( A, k, n ) );
+
+ /* TS kernel */
+ INSERT_TASK_ztplqt(
+ options,
+ tempkm, tempnn, 0, ib, T->nb,
+ A(k, k),
+ 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;
+
+ RUNTIME_data_migrate( sequence, A(m, k),
+ A->get_rankof( A, m, n ) );
+
+ INSERT_TASK_ztpmlqt(
+ options,
+ ChamRight, ChamConjTrans,
+ tempmm, tempnn, A->mb, 0, ib, T->nb,
+ A(k, n),
+ T(k, n),
+ A(m, k),
+ A(m, n));
+ }
+ RUNTIME_data_flush( sequence, A(k, n) );
+ RUNTIME_data_flush( sequence, T(k, n) );
+ }
+
+ return 1;
+}
+
+/**
+ * Parallel tile LQ factorization (reduction Householder) - dynamic scheduling
+ *
+ * @param[in] genD
+ * Indicate if copies of the gelqt tiles must be done to speedup
+ * computations in updates. genD is considered only if D is not NULL.
+ *
*/
void chameleon_pzgelqf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D,
RUNTIME_sequence_t *sequence, RUNTIME_request_t *request )
@@ -43,9 +135,7 @@ void chameleon_pzgelqf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
size_t ws_worker = 0;
size_t ws_host = 0;
- int k, m, n;
- int tempkm, tempkn, tempmm, tempnn;
- int ib, minMNT;
+ int k, m, ib, minMNT;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -61,7 +151,7 @@ void chameleon_pzgelqf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
minMNT = A->mt;
}
- if ( D == NULL ) {
+ if ( (genD == 0) || (D == NULL) ) {
D = A;
genD = 0;
}
@@ -92,73 +182,8 @@ void chameleon_pzgelqf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
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;
- INSERT_TASK_zgelqt(
- &options,
- tempkm, tempkn, ib, T->nb,
- A(k, k),
- 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;
- INSERT_TASK_zlacpy(
- &options,
- ChamUpper, tempDkm, tempDkn,
- A(k, k),
- D(k) );
-#if defined(CHAMELEON_USE_CUDA)
- INSERT_TASK_zlaset(
- &options,
- ChamLower, tempDkm, tempDkn,
- 0., 1.,
- D(k) );
-#endif
- }
- for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
- INSERT_TASK_zunmlq(
- &options,
- ChamRight, ChamConjTrans,
- tempmm, tempkn, tempkn, ib, T->nb,
- D(k),
- T(k, k),
- A(m, k));
- }
- RUNTIME_data_flush( sequence, D(k) );
- 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;
-
- RUNTIME_data_migrate( sequence, A(k, k),
- A->get_rankof( A, k, n ) );
-
- /* TS kernel */
- INSERT_TASK_ztplqt(
- &options,
- tempkm, tempnn, 0, ib, T->nb,
- A(k, k),
- 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;
-
- RUNTIME_data_migrate( sequence, A(m, k),
- A->get_rankof( A, m, n ) );
-
- INSERT_TASK_ztpmlqt(
- &options,
- ChamRight, ChamConjTrans,
- tempmm, tempnn, A->mb, 0, ib, T->nb,
- A(k, n),
- T(k, n),
- A(m, k),
- A(m, n));
- }
- RUNTIME_data_flush( sequence, A(k, n) );
- RUNTIME_data_flush( sequence, T(k, n) );
- }
+ chameleon_pzgelqf_step( genD, k, ib,
+ A, T, D, &options, sequence );
/* Restore the original location of the tiles */
for (m = k; m < A->mt; m++) {
diff --git a/compute/pzgelqf_param.c b/compute/pzgelqf_param.c
index b69b84b50..7bc06b159 100644
--- a/compute/pzgelqf_param.c
+++ b/compute/pzgelqf_param.c
@@ -28,21 +28,162 @@
/**
* Parallel tile LQ factorization (reduction Householder) - dynamic scheduling
+ *
+ * @param[in] genD
+ * Indicate if copies of the gelqt tiles must be done to speedup
+ * computations in updates. genD is considered only if D is not NULL.
+ *
+ * @param[in] uplo
+ * - ChamUpper: Classic LQ factorization of the matrix A.
+ * - ChamLower: LQ factorization of the TTLQT kernel.
+ * - ChamUpperLower: LQ factorization of the TSLQT kernel.
+ */
+int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
+ const libhqr_tree_t *qrtree, int *tiles,
+ CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
+ RUNTIME_option_t *options, RUNTIME_sequence_t *sequence )
+{
+ CHAM_desc_t *T;
+ int m, n, i, p;
+ int L, nbgelqt;
+ int tempkmin, tempkm, tempnn, tempmm, temppn;
+ int node, nbtiles;
+
+ tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
+
+ /* The number of geqrt to apply */
+ nbgelqt = qrtree->getnbgeqrf( qrtree, k );
+
+ T = TS;
+ for (i = 0; i < nbgelqt; i++) {
+ p = qrtree->getm( qrtree, k, i );
+
+ /* We skip the LQ factorization if this is the last diagonal tile */
+ if ( (uplo == ChamLower) && (p == k) ) {
+ continue;
+ }
+
+ temppn = p == A->nt-1 ? A->n-p*A->nb : A->nb;
+ tempkmin = chameleon_min(tempkm, temppn);
+
+ INSERT_TASK_zgelqt(
+ options,
+ tempkm, temppn, ib, T->nb,
+ 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;
+
+ INSERT_TASK_zlacpy(
+ options,
+ ChamUpper, tempDkm, tempDpn,
+ A(k, p), D(k, p) );
+#if defined(CHAMELEON_USE_CUDA)
+ INSERT_TASK_zlaset(
+ options,
+ ChamLower, tempDkm, tempDpn,
+ 0., 1.,
+ D(k, p) );
+#endif
+ }
+
+ for (m = k+1; m < A->mt; m++) {
+ tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
+ INSERT_TASK_zunmlq(
+ options,
+ ChamRight, ChamConjTrans,
+ tempmm, temppn, tempkmin, ib, T->nb,
+ D(k, p),
+ T(k, p),
+ A(m, p));
+ }
+
+ if ( genD || ((k+1) < A->mt)) {
+ RUNTIME_data_flush( sequence, D(k, p) );
+ }
+ RUNTIME_data_flush( sequence, T(k, p) );
+ }
+
+ /* Setting the order of the tiles */
+ nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
+
+ for (i = 0; i < nbtiles; i++) {
+ n = tiles[i];
+ p = qrtree->currpiv( qrtree, k, n );
+
+ tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+
+ if ( qrtree->gettype( qrtree, k, n ) == LIBHQR_KILLED_BY_TS ) {
+ /* TS kernel */
+ T = TS;
+ L = 0;
+
+ /* Force TT kernel if this is the last diagonal tile */
+ if ( (uplo == ChamLower) && (n == k) ) {
+ L = tempnn;
+ }
+ }
+ else {
+ /* TT kernel */
+ T = TT;
+ L = tempnn;
+ }
+
+ node = A->get_rankof( A, k, n );
+ RUNTIME_data_migrate( sequence, A(k, p), node );
+ RUNTIME_data_migrate( sequence, A(k, n), node );
+
+ INSERT_TASK_ztplqt(
+ options,
+ tempkm, tempnn, chameleon_min(L, tempkm), ib, T->nb,
+ A(k, p),
+ 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;
+
+ node = A->get_rankof( A, m, n );
+ RUNTIME_data_migrate( sequence, A(m, p), node );
+ RUNTIME_data_migrate( sequence, A(m, n), node );
+
+ INSERT_TASK_ztpmlqt(
+ options,
+ ChamRight, ChamConjTrans,
+ tempmm, tempnn, tempkm, L, ib, T->nb,
+ A(k, n),
+ T(k, n),
+ A(m, p),
+ A(m, n));
+ }
+ RUNTIME_data_flush( sequence, A(k, n) );
+ RUNTIME_data_flush( sequence, T(k, n) );
+ }
+
+ return tiles[nbtiles];
+}
+
+/**
+ * Parallel tile LQ factorization (reduction Householder) - dynamic scheduling
+ *
+ * @param[in] genD
+ * Indicate if copies of the gelqt tiles must be done to speedup
+ * computations in updates. genD is considered only if D is not NULL.
+ *
*/
-void chameleon_pzgelqf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t *A,
+void chameleon_pzgelqf_param( int genD, int K,
+ const libhqr_tree_t *qrtree, CHAM_desc_t *A,
CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
RUNTIME_sequence_t *sequence, RUNTIME_request_t *request )
{
CHAM_context_t *chamctxt;
RUNTIME_option_t options;
- CHAM_desc_t *T;
size_t ws_worker = 0;
size_t ws_host = 0;
- int k, m, n, i, p;
- int K, L, nbgeqrt;
- int tempkmin, tempkm, tempnn, tempmm, temppn;
- int ib, node, nbtiles, *tiles;
+ int k, m;
+ int ib, *tiles;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -52,7 +193,7 @@ void chameleon_pzgelqf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
ib = CHAMELEON_IB;
- if ( D == NULL ) {
+ if ( (genD == 0) || (D == NULL) ) {
D = A;
genD = 0;
}
@@ -82,119 +223,11 @@ void chameleon_pzgelqf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
/* Initialisation of temporary tiles array */
tiles = (int*)calloc(qrtree->mt, sizeof(int));
- K = chameleon_min(A->mt, A->nt);
-
for (k = 0; k < K; k++) {
RUNTIME_iteration_push(chamctxt, k);
- tempkm = k == A->mt-1 ? A->m-k*A->mb : A->mb;
-
- /* The number of geqrt to apply */
- nbgeqrt = qrtree->getnbgeqrf(qrtree, k);
-
- T = TS;
- for (i = 0; i < nbgeqrt; i++) {
- p = qrtree->getm(qrtree, k, i);
-
- /* /\* We skip the LQ factorization if this is the last diagonal tile *\/ */
- /* if ( (uplo == ChamLower) && (p == k) ) { */
- /* continue; */
- /* } */
-
- temppn = p == A->nt-1 ? A->n-p*A->nb : A->nb;
- tempkmin = chameleon_min(tempkm, temppn);
-
- INSERT_TASK_zgelqt(
- &options,
- tempkm, temppn, ib, T->nb,
- 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;
-
- INSERT_TASK_zlacpy(
- &options,
- ChamUpper, tempDkm, tempDpn,
- A(k, p), D(k, p) );
-#if defined(CHAMELEON_USE_CUDA)
- INSERT_TASK_zlaset(
- &options,
- ChamLower, tempDkm, tempDpn,
- 0., 1.,
- D(k, p) );
-#endif
- }
-
- for (m = k+1; m < A->mt; m++) {
- tempmm = m == A->mt-1 ? A->m-m*A->mb : A->mb;
- INSERT_TASK_zunmlq(
- &options,
- ChamRight, ChamConjTrans,
- tempmm, temppn, tempkmin, ib, T->nb,
- D(k, p),
- T(k, p),
- A(m, p));
- }
- RUNTIME_data_flush( sequence, D(k, p) );
- RUNTIME_data_flush( sequence, T(k, p) );
- }
-
- /* Setting the order of the tiles */
- nbtiles = libhqr_walk_stepk( qrtree, k, tiles );
-
- for (i = 0; i < nbtiles; i++) {
- n = tiles[i];
- p = qrtree->currpiv( qrtree, k, n );
-
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
-
- if ( qrtree->gettype( qrtree, k, n ) == LIBHQR_KILLED_BY_TS ) {
- /* TS kernel */
- T = TS;
- L = 0;
-
- /* /\* Force TT kernel if this is the last diagonal tile *\/ */
- /* if ( (uplo == ChamLower) && (n == k) ) { */
- /* L = tempnn; */
- /* } */
- }
- else {
- /* TT kernel */
- T = TT;
- L = tempnn;
- }
-
- node = A->get_rankof( A, k, n );
- RUNTIME_data_migrate( sequence, A(k, p), node );
- RUNTIME_data_migrate( sequence, A(k, n), node );
-
- INSERT_TASK_ztplqt(
- &options,
- tempkm, tempnn, chameleon_min(L, tempkm), ib, T->nb,
- A(k, p),
- 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;
-
- node = A->get_rankof( A, m, n );
- RUNTIME_data_migrate( sequence, A(m, p), node );
- RUNTIME_data_migrate( sequence, A(m, n), node );
-
- INSERT_TASK_ztpmlqt(
- &options,
- ChamRight, ChamConjTrans,
- tempmm, tempnn, tempkm, L, ib, T->nb,
- A(k, n),
- T(k, n),
- A(m, p),
- A(m, n));
- }
- RUNTIME_data_flush( sequence, A(k, n) );
- RUNTIME_data_flush( sequence, T(k, n) );
- }
+ chameleon_pzgelqf_param_step( genD, ChamUpper, k, ib, qrtree, tiles,
+ A, TS, TT, D, &options, sequence );
/* Restore the original location of the tiles */
for (m = k; m < A->mt; m++) {
diff --git a/compute/pzgeqrf.c b/compute/pzgeqrf.c
index a7637392d..49c270cf0 100644
--- a/compute/pzgeqrf.c
+++ b/compute/pzgeqrf.c
@@ -32,7 +32,101 @@
#define D(k) D, k, k
/**
- * Parallel tile QR factorization - dynamic scheduling
+ * Parallel tile QR factorization (reduction Householder) - dynamic scheduling
+ *
+ * @param[in] genD
+ * Indicate if copies of the geqrt tiles must be done to speedup
+ * computations in updates. genD is considered only if D is not NULL.
+ */
+int chameleon_pzgeqrf_step( int genD, int k, int ib,
+ CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D,
+ RUNTIME_option_t *options, RUNTIME_sequence_t *sequence )
+{
+ 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;
+
+ INSERT_TASK_zgeqrt(
+ options,
+ tempkm, tempkn, ib, T->nb,
+ A(k, k),
+ 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;
+ INSERT_TASK_zlacpy(
+ options,
+ ChamLower, tempDkm, tempDkn,
+ A(k, k),
+ D(k) );
+#if defined(CHAMELEON_USE_CUDA)
+ INSERT_TASK_zlaset(
+ options,
+ ChamUpper, tempDkm, tempDkn,
+ 0., 1.,
+ D(k) );
+#endif
+ }
+ for (n = k+1; n < A->nt; n++) {
+ tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+ INSERT_TASK_zunmqr(
+ options,
+ ChamLeft, ChamConjTrans,
+ tempkm, tempnn, tempkm, ib, T->nb,
+ D(k),
+ T(k, k),
+ A(k, n));
+ }
+ RUNTIME_data_flush( sequence, D(k) );
+ 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;
+
+ RUNTIME_data_migrate( sequence, A(k, k),
+ A->get_rankof( A, m, k ) );
+
+ /* TS kernel */
+ INSERT_TASK_ztpqrt(
+ options,
+ tempmm, tempkn, 0, ib, T->nb,
+ A(k, k),
+ A(m, k),
+ T(m, k));
+
+ for (n = k+1; n < A->nt; n++) {
+ tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
+
+ RUNTIME_data_migrate( sequence, A(k, n),
+ A->get_rankof( A, m, n ) );
+
+ /* TS kernel */
+ INSERT_TASK_ztpmqrt(
+ options,
+ ChamLeft, ChamConjTrans,
+ tempmm, tempnn, A->nb, 0, ib, T->nb,
+ A(m, k),
+ T(m, k),
+ A(k, n),
+ A(m, n));
+ }
+ RUNTIME_data_flush( sequence, A(m, k) );
+ RUNTIME_data_flush( sequence, T(m, k) );
+ }
+
+ return 1;
+}
+
+/**
+ * Parallel tile QR factorization (reduction Householder) - dynamic scheduling
+ *
+ * @param[in] genD
+ * Indicate if copies of the geqrt tiles must be done to speedup
+ * computations in updates. genD is considered only if D is not NULL.
+ *
*/
void chameleon_pzgeqrf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D,
RUNTIME_sequence_t *sequence, RUNTIME_request_t *request )
@@ -42,10 +136,7 @@ void chameleon_pzgeqrf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
size_t ws_worker = 0;
size_t ws_host = 0;
- int k, m, n;
- int tempkm, tempkn, tempnn, tempmm;
- int ib;
- int minMNT = chameleon_min(A->mt, A->nt);
+ int k, n, ib, minMNT;
chamctxt = chameleon_context_self();
if (sequence->status != CHAMELEON_SUCCESS) {
@@ -55,7 +146,13 @@ void chameleon_pzgeqrf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
ib = CHAMELEON_IB;
- if ( D == NULL ) {
+ if (A->m > A->n) {
+ minMNT = A->nt;
+ } else {
+ minMNT = A->mt;
+ }
+
+ if ( (genD == 0) || (D == NULL) ) {
D = A;
genD = 0;
}
@@ -86,75 +183,8 @@ void chameleon_pzgeqrf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
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;
- INSERT_TASK_zgeqrt(
- &options,
- tempkm, tempkn, ib, T->nb,
- A(k, k),
- 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;
- INSERT_TASK_zlacpy(
- &options,
- ChamLower, tempDkm, tempDkn,
- A(k, k),
- D(k) );
-#if defined(CHAMELEON_USE_CUDA)
- INSERT_TASK_zlaset(
- &options,
- ChamUpper, tempDkm, tempDkn,
- 0., 1.,
- D(k) );
-#endif
- }
- for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
- INSERT_TASK_zunmqr(
- &options,
- ChamLeft, ChamConjTrans,
- tempkm, tempnn, tempkm, ib, T->nb,
- D(k),
- T(k, k),
- A(k, n));
- }
- RUNTIME_data_flush( sequence, D(k) );
- 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;
-
- RUNTIME_data_migrate( sequence, A(k, k),
- A->get_rankof( A, m, k ) );
-
- /* TS kernel */
- INSERT_TASK_ztpqrt(
- &options,
- tempmm, tempkn, 0, ib, T->nb,
- A(k, k),
- A(m, k),
- T(m, k));
-
- for (n = k+1; n < A->nt; n++) {
- tempnn = n == A->nt-1 ? A->n-n*A->nb : A->nb;
-
- RUNTIME_data_migrate( sequence, A(k, n),
- A->get_rankof( A, m, n ) );
-
- /* TS kernel */
- INSERT_TASK_ztpmqrt(
- &options,
- ChamLeft, ChamConjTrans,
- tempmm, tempnn, A->nb, 0, ib, T->nb,
- A(m, k),
- T(m, k),
- A(k, n),
- A(m, n));
- }
- RUNTIME_data_flush( sequence, A(m, k) );
- RUNTIME_data_flush( sequence, T(m, k) );
- }
+ chameleon_pzgeqrf_step( genD, k, ib,
+ A, T, D, &options, sequence );
/* Restore the original location of the tiles */
for (n = k; n < A->nt; n++) {
diff --git a/compute/pzgeqrf_param.c b/compute/pzgeqrf_param.c
index a4e08e061..d2bd084d2 100644
--- a/compute/pzgeqrf_param.c
+++ b/compute/pzgeqrf_param.c
@@ -164,7 +164,6 @@ int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
return tiles[nbtiles];
}
-
/**
* Parallel tile QR factorization (reduction Householder) - dynamic scheduling
*
diff --git a/control/compute_z.h b/control/compute_z.h
index d4acf6976..295c64264 100644
--- a/control/compute_z.h
+++ b/control/compute_z.h
@@ -133,6 +133,16 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans, CHAM_des
void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *T, CHAM_desc_t *D, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
void chameleon_pzbuild( cham_uplo_t uplo, CHAM_desc_t *A, void *user_data, void* user_build_callback, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request );
+int chameleon_pzgelqf_step( int genD, int k, int ib,
+ CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D,
+ RUNTIME_option_t *options, RUNTIME_sequence_t *sequence );
+int chameleon_pzgeqrf_step( int genD, int k, int ib,
+ CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D,
+ RUNTIME_option_t *options, RUNTIME_sequence_t *sequence );
+int chameleon_pzgelqf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
+ const libhqr_tree_t *qrtree, int *tiles,
+ CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
+ RUNTIME_option_t *options, RUNTIME_sequence_t *sequence );
int chameleon_pzgeqrf_param_step( int genD, cham_uplo_t uplo, int k, int ib,
const libhqr_tree_t *qrtree, int *tiles,
CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
@@ -142,7 +152,8 @@ void chameleon_pzungqr_param_step( int genD, cham_uplo_t uplo, int k, int ib,
CHAM_desc_t *A, CHAM_desc_t *Q,
CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
RUNTIME_option_t *options, RUNTIME_sequence_t *sequence );
-void chameleon_pzgelqf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
+void chameleon_pzgelqf_param( int genD, int K, const libhqr_tree_t *qrtree,
+ CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
void chameleon_pzgeqrf_param( int genD, int K, const libhqr_tree_t *qrtree,
CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
--
GitLab