diff --git a/compute/pzgelqf.c b/compute/pzgelqf.c
index 6bf814a62bd42c1b1107eb759f89fdb69e2e808e..313b83eebd7e3815e2a35dcea4f7d9c8ab2f4727 100644
--- a/compute/pzgelqf.c
+++ b/compute/pzgelqf.c
@@ -19,7 +19,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -65,10 +65,10 @@ void chameleon_pzgelqf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
}
/*
- * zgelqt = A->nb * (ib+1)
- * zunmlq = A->nb * ib
- * ztslqt = A->nb * (ib+1)
- * ztsmlq = A->nb * ib
+ * zgelqt = A->nb * (ib+1)
+ * zunmlq = A->nb * ib
+ * ztplqt = A->nb * (ib+1)
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * (ib+1);
@@ -76,8 +76,8 @@ void chameleon_pzgelqf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 2 * A->nb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
#endif
diff --git a/compute/pzgelqf_param.c b/compute/pzgelqf_param.c
index c91cea637129374915d28d8a9ee0743e82c7acda..7cc655e6deef5bd566445956ecf19886095816f4 100644
--- a/compute/pzgelqf_param.c
+++ b/compute/pzgelqf_param.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Raphael Boucherie
- * @date 2017-05-17
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -69,10 +69,10 @@ void chameleon_pzgelqf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
/* Allocation of temporary (scratch) working space */
#if defined(CHAMELEON_USE_CUDA)
/*
- * zunmqr = A->nb * ib
- * ztpmqrt = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzgelqfrh.c b/compute/pzgelqfrh.c
index d9b77e5487e11f6c23e95bc347dc7943a5439ca5..7cec545bbf0318e47fe5920993876608e8c50cb7 100644
--- a/compute/pzgelqfrh.c
+++ b/compute/pzgelqfrh.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -62,10 +62,10 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
}
/*
- * zgelqt = A->nb * (ib+1)
- * zunmlq = A->nb * ib
- * ztplqt = A->nb * (ib+1)
- * ztpmlq = A->nb * ib
+ * zgelqt = A->nb * (ib+1)
+ * zunmlq = A->nb * ib
+ * ztplqt = A->nb * (ib+1)
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * (ib+1);
@@ -73,10 +73,10 @@ void chameleon_pzgelqfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztpmqr = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzgeqrf.c b/compute/pzgeqrf.c
index 650e769a022a38c2066aadd75dfa0d776aa62f78..4ee0c4dabb4d13afdb94ea51bdce23f6b64d9fe8 100644
--- a/compute/pzgeqrf.c
+++ b/compute/pzgeqrf.c
@@ -19,7 +19,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -60,10 +60,10 @@ void chameleon_pzgeqrf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
}
/*
- * zgeqrt = A->nb * (ib+1)
- * zunmqr = A->nb * ib
- * ztsqrt = A->nb * (ib+1)
- * ztsmqr = A->nb * ib
+ * zgeqrt = A->nb * (ib+1)
+ * zunmqr = A->nb * ib
+ * ztpqrt = A->nb * (ib+1)
+ * ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * (ib+1);
@@ -71,8 +71,8 @@ void chameleon_pzgeqrf( int genD, CHAM_desc_t *A, CHAM_desc_t *T, CHAM_desc_t *D
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 2 * A->nb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
#endif
diff --git a/compute/pzgeqrf_param.c b/compute/pzgeqrf_param.c
index b297e598ff0e0bf40899be877d4942ab172dd7c6..72dcc8e5775b315a2e22f383f57e153fd1ee186e 100644
--- a/compute/pzgeqrf_param.c
+++ b/compute/pzgeqrf_param.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Raphael Boucherie
- * @date 2017-05-17
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -74,10 +74,10 @@ void chameleon_pzgeqrf_param( int genD, int K,
/* Allocation of temporary (scratch) working space */
#if defined(CHAMELEON_USE_CUDA)
/*
- * zunmqr = A->nb * ib
- * ztpmqrt = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzgeqrfrh.c b/compute/pzgeqrfrh.c
index b075a0ef19c860c3952220cf58d0f826b8b394ab..74d153ae80f99c8f19bc0913470c69759a9d45f6 100644
--- a/compute/pzgeqrfrh.c
+++ b/compute/pzgeqrfrh.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -62,10 +62,10 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
}
/*
- * zgeqrt = A->nb * (ib+1)
- * zunmqr = A->nb * ib
- * ztpqrt = A->nb * (ib+1)
- * ztpmqr = A->nb * ib
+ * zgeqrt = A->nb * (ib+1)
+ * zunmqr = A->nb * ib
+ * ztpqrt = A->nb * (ib+1)
+ * ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * (ib+1);
@@ -73,10 +73,10 @@ void chameleon_pzgeqrfrh( int genD, int BS, CHAM_desc_t *A, CHAM_desc_t *T, CHAM
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztpmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzhetrd_he2hb.c b/compute/pzhetrd_he2hb.c
index 711651d15bfcc711bb46315a7d936f4c64c18c74..623a25517808356acaa7b61373e831a9b043452b 100644
--- a/compute/pzhetrd_he2hb.c
+++ b/compute/pzhetrd_he2hb.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Hatem Ltaief
* @author Azzam Haidar
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -74,9 +74,9 @@ void chameleon_pzhetrd_he2hb(cham_uplo_t uplo,
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
+ * zunmqr = A->nb * ib
* ztsmqr = 2 * A->nb * ib
- * zherfb = A->nb * ib
+ * zherfb = A->nb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
#endif
diff --git a/compute/pztpgqrt.c b/compute/pztpgqrt.c
index ec1880018e8b2b0f2089e7a9cc6364d135095b5b..52d99f42850f531dc8177d30f73c12169040272b 100644
--- a/compute/pztpgqrt.c
+++ b/compute/pztpgqrt.c
@@ -14,7 +14,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -66,9 +66,9 @@ void chameleon_pztpgqrt( int KT, int L,
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * ztpmqrt = 2 * Q1->nb * ib
+ * ztpmqrt = 3 * Q1->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * Q1->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * Q1->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pztpqrt.c b/compute/pztpqrt.c
index 37de659fee2f50ac169db02532dd00082b38bc1c..0783d14da4e55f294f3b784349039485952da66d 100644
--- a/compute/pztpqrt.c
+++ b/compute/pztpqrt.c
@@ -14,7 +14,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -61,9 +61,9 @@ void chameleon_pztpqrt( int L, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *T,
/* Allocation of temporary (scratch) working space */
#if defined(CHAMELEON_USE_CUDA)
/*
- * ztpmqrt = 2 * A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzunglq.c b/compute/pzunglq.c
index 05ad8cdef0ae7fb131376b631559f8f14dfb7f42..492843fc3710220d31abda8728dfbaf0903555b7 100644
--- a/compute/pzunglq.c
+++ b/compute/pzunglq.c
@@ -19,7 +19,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -67,8 +67,8 @@ void chameleon_pzunglq( int genD, CHAM_desc_t *A, CHAM_desc_t *Q, CHAM_desc_t *T
}
/*
- * zunmlq = A->nb * ib
- * ztpmlq = A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * ib;
@@ -76,8 +76,8 @@ void chameleon_pzunglq( int genD, CHAM_desc_t *A, CHAM_desc_t *Q, CHAM_desc_t *T
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmlq = A->nb * ib
- * ztpmlq = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 2 * A->nb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
#endif
diff --git a/compute/pzunglq_param.c b/compute/pzunglq_param.c
index d98a98ac70c4c7909f937e99bd588d1a95107eb2..b70080358fb49e66f3176097d98ecb9422ec44f2 100644
--- a/compute/pzunglq_param.c
+++ b/compute/pzunglq_param.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Raphael Boucherie
- * @date 2017-05-17
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -63,18 +63,18 @@ void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t
}
/*
- * zunmqr = A->nb * ib
- * ztpmqr = A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * ib;
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztpmqr = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzunglqrh.c b/compute/pzunglqrh.c
index 95d6a7df46b5125504e2986c54e15f95ea96f94f..d31203a9eec0ad3c51fe633a2469eb065fad568e 100644
--- a/compute/pzunglqrh.c
+++ b/compute/pzunglqrh.c
@@ -18,7 +18,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2011-05-24
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -64,18 +64,18 @@ void chameleon_pzunglqrh( int genD, int BS,
}
/*
- * zunmqr = A->nb * ib
- * ztpmqr = A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * ib;
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztpmqr = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzungqr.c b/compute/pzungqr.c
index f9cc9182224acb9d6b01366f863b3d1c54fa779c..a190027288631727f1a08a2e4b1bcbe2fb69af19 100644
--- a/compute/pzungqr.c
+++ b/compute/pzungqr.c
@@ -19,7 +19,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -68,8 +68,8 @@ void chameleon_pzungqr( int genD, CHAM_desc_t *A, CHAM_desc_t *Q,
}
/*
- * zunmqr = A->nb * ib
- * ztsmqr = A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * ib;
@@ -77,8 +77,8 @@ void chameleon_pzungqr( int genD, CHAM_desc_t *A, CHAM_desc_t *Q,
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 2 * A->nb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
#endif
diff --git a/compute/pzungqr_param.c b/compute/pzungqr_param.c
index 6f550160137776a091b9ed685ebf24dd64b1c992..8bcb901c13a962a4e199a8e0c2d64966e81ca3ea 100644
--- a/compute/pzungqr_param.c
+++ b/compute/pzungqr_param.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Raphael Boucherie
- * @date 2017-05-17
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -60,17 +60,17 @@ void chameleon_pzungqr_param( int genD, int K,
}
/*
- * zunmqr = A->nb * ib
- * ztpmqr = A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * ib;
/* Allocation of temporary (scratch) working space */
#if defined(CHAMELEON_USE_CUDA)
/*
- * ztpmqrt = 2 * A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzungqrrh.c b/compute/pzungqrrh.c
index 8fbfb9f489647bac8526a48b824380231cadb2ea..23d5e29db53f570516eee7231e4b326fde04e140 100644
--- a/compute/pzungqrrh.c
+++ b/compute/pzungqrrh.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -67,8 +67,7 @@ void chameleon_pzungqrrh( int genD, int BS,
}
/*
- * zunmqr = A->nb * ib
- * ztsmqr = A->nb * ib
+ * zunmqr = A->nb * ib
* ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * ib;
@@ -76,10 +75,10 @@ void chameleon_pzungqrrh( int genD, int BS,
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzunmlq.c b/compute/pzunmlq.c
index faff5895e38d393a431996019f483a0ee2a9a4ef..a5d23bf2b1db09789830af95372ffd67d46d76da 100644
--- a/compute/pzunmlq.c
+++ b/compute/pzunmlq.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -70,16 +70,16 @@ void chameleon_pzunmlq( int genD, cham_side_t side, cham_trans_t trans,
}
/*
- * zunmlq = A->mb * ib
- * ztsmlq = A->mb * ib
+ * zunmlq = A->mb * ib
+ * ztpmlqt = A->mb * ib
*/
ws_worker = A->mb * ib;
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmlq = A->mb * ib
- * ztsmlq = 2 * A->mb * ib
+ * zunmlq = A->mb * ib
+ * ztpmlqt = 2 * A->mb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->mb * 2 );
#endif
diff --git a/compute/pzunmlq_param.c b/compute/pzunmlq_param.c
index 01c34e52b5a9a2f61b322737125a53322394b7c8..48dbc13eaab3b75b2a75b01c567c68524041680e 100644
--- a/compute/pzunmlq_param.c
+++ b/compute/pzunmlq_param.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Raphael Boucherie
- * @date 2017-05-17
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -63,19 +63,18 @@ void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree,
}
/*
- * zunmlq = A->nb * ib
- * ztsmlq = A->nb * ib
- * zttmlq = A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * ib;
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmlq = A->nb * ib
- * ztsmlq = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzunmlqrh.c b/compute/pzunmlqrh.c
index 23d0381769f2878af18f28f992f446dce18e1b97..26130cb69a7071245bc341f2310b01067e35f5de 100644
--- a/compute/pzunmlqrh.c
+++ b/compute/pzunmlqrh.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -65,19 +65,18 @@ void chameleon_pzunmlqrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
/*
- * zunmlq = A->nb * ib
- * ztsmlq = A->nb * ib
- * zttmlq = A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = A->nb * ib
*/
ws_worker = A->nb * ib;
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmlq = A->nb * ib
- * ztsmlq = 2 * A->nb * ib
+ * zunmlq = A->nb * ib
+ * ztpmlqt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzunmqr.c b/compute/pzunmqr.c
index 9000c51582d83b89c351f94ac5533f88d792f0d6..4c15ba7fd0ac6ec475583f73c190c34caae6f1bb 100644
--- a/compute/pzunmqr.c
+++ b/compute/pzunmqr.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -70,16 +70,16 @@ void chameleon_pzunmqr( int genD, cham_side_t side, cham_trans_t trans,
}
/*
- * zunmqr = A->nb * ib
- * ztsmqr = A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * ib;
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 2 * A->nb * ib
*/
ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
#endif
diff --git a/compute/pzunmqr_param.c b/compute/pzunmqr_param.c
index 7a9b2b6e11765a47f3aaaad0e654765ca25b2b1a..772bfdf48f4310ef9a96a7de427fdc7c35a09f49 100644
--- a/compute/pzunmqr_param.c
+++ b/compute/pzunmqr_param.c
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Raphael Boucherie
- * @date 2017-05-17
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -63,8 +63,7 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
}
/*
- * zunmqr = A->nb * ib
- * ztsmqr = A->nb * ib
+ * zunmqr = A->nb * ib
* ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * ib;
@@ -72,10 +71,10 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree,
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/compute/pzunmqrrh.c b/compute/pzunmqrrh.c
index 1073db6ac13142e4e71f6bbb761de68d4402470a..e8429f8cfeb7ae92709228dd8fab734364149944 100644
--- a/compute/pzunmqrrh.c
+++ b/compute/pzunmqrrh.c
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> s d c
*
*/
@@ -66,8 +66,7 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
}
/*
- * zunmqr = A->nb * ib
- * ztsmqr = A->nb * ib
+ * zunmqr = A->nb * ib
* ztpmqrt = A->nb * ib
*/
ws_worker = A->nb * ib;
@@ -75,10 +74,10 @@ void chameleon_pzunmqrrh( int genD, int BS, cham_side_t side, cham_trans_t trans
#if defined(CHAMELEON_USE_CUDA)
/* Worker space
*
- * zunmqr = A->nb * ib
- * ztsmqr = 2 * A->nb * ib
+ * zunmqr = A->nb * ib
+ * ztpmqrt = 3 * A->nb * ib
*/
- ws_worker = chameleon_max( ws_worker, ib * A->nb * 2 );
+ ws_worker = chameleon_max( ws_worker, ib * A->nb * 3 );
#endif
ws_worker *= sizeof(CHAMELEON_Complex64_t);
diff --git a/coreblas/compute/core_zparfb.c b/coreblas/compute/core_zparfb.c
index 5199173acfadea031f5d656407638c7b0e9a8c18..a359402d6b90c8aa0484e1d1b587c413060561c4 100644
--- a/coreblas/compute/core_zparfb.c
+++ b/coreblas/compute/core_zparfb.c
@@ -18,7 +18,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2011-06-14
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
@@ -139,7 +139,8 @@
*/
/* This kernel is never traced so return type on previous line for convert2eztrace.pl script */
int
-CORE_zparfb(cham_side_t side, cham_trans_t trans, cham_dir_t direct, cham_store_t storev,
+CORE_zparfb(cham_side_t side, cham_trans_t trans,
+ cham_dir_t direct, cham_store_t storev,
int M1, int N1, int M2, int N2, int K, int L,
CHAMELEON_Complex64_t *A1, int LDA1,
CHAMELEON_Complex64_t *A2, int LDA2,
diff --git a/coreblas/compute/core_ztpmlqt.c b/coreblas/compute/core_ztpmlqt.c
index 72e54b1ad14c6590adb0a034f7d1c86d22df9737..bfc53b98dba91a9ca56fbf71e2b6b3faf1828c99 100644
--- a/coreblas/compute/core_ztpmlqt.c
+++ b/coreblas/compute/core_ztpmlqt.c
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
@@ -24,9 +24,11 @@
*
* @ingroup CORE_CHAMELEON_Complex64_t
*
- * CORE_ztpmlqt applies a complex orthogonal matrix Q obtained from a
- * "triangular-pentagonal" complex block reflector H to a general complex matrix
- * C, which consists of two blocks A and B.
+ * @brief Applies a complex orthogonal matrix Q.
+ *
+ * The matrix Q is obtained from a "triangular-pentagonal" complex block
+ * reflector H to a general complex matrix C, which consists of two blocks A and
+ * B.
*
*******************************************************************************
*
@@ -128,9 +130,8 @@
*
*******************************************************************************
*
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
+ * @retval CHAMELEON_SUCCESS successful exit
+ * @retval <0 if -i, the i-th argument had an illegal value
*
*/
diff --git a/coreblas/compute/core_zttmlq.c b/coreblas/compute/core_zttmlq.c
index 69a7004c6d9d9cd148766f822025fcdbe7f441a9..5b6ee0261ec8e920f4883847931526fc864ddf76 100644
--- a/coreblas/compute/core_zttmlq.c
+++ b/coreblas/compute/core_zttmlq.c
@@ -19,7 +19,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
@@ -127,9 +127,9 @@ int CORE_zttmlq(cham_side_t side, cham_trans_t trans,
const CHAMELEON_Complex64_t *T, int LDT,
CHAMELEON_Complex64_t *WORK, int LDWORK)
{
- int i, i1, i3, l;
+ int i, i1, i3;
int NW;
- int kb;
+ int kb, l;
int ic = 0;
int jc = 0;
int mi1 = M1;
@@ -205,11 +205,13 @@ int CORE_zttmlq(cham_side_t side, cham_trans_t trans,
}
/* Quick return */
- if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0))
+ if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0)) {
return CHAMELEON_SUCCESS;
+ }
- if (((side == ChamLeft) && (trans == ChamNoTrans))
- || ((side == ChamRight) && (trans != ChamNoTrans))) {
+ if ( ((side == ChamLeft ) && (trans == ChamNoTrans)) ||
+ ((side == ChamRight) && (trans != ChamNoTrans)) )
+ {
i1 = 0;
i3 = IB;
}
@@ -248,13 +250,11 @@ int CORE_zttmlq(cham_side_t side, cham_trans_t trans,
CORE_zparfb(
side, trans, ChamDirForward, ChamRowwise,
mi1, ni1, mi2, ni2, kb, l,
- &A1[LDA1*jc+ic], LDA1,
+ A1 + LDA1 * jc + ic, LDA1,
A2, LDA2,
- &V[i], LDV,
- &T[LDT*i], LDT,
+ V + i, LDV,
+ T + LDT * i, LDT,
WORK, LDWORK);
}
return CHAMELEON_SUCCESS;
}
-
-
diff --git a/cudablas/compute/CMakeLists.txt b/cudablas/compute/CMakeLists.txt
index d9859a604616b7b4c8e79f7bc8a32238b68eca40..4a1f8559a9c62bbad84f79c545c89075dd71229d 100644
--- a/cudablas/compute/CMakeLists.txt
+++ b/cudablas/compute/CMakeLists.txt
@@ -19,7 +19,7 @@
#
# @version 1.0.0
# @author Florent Pruvost
-# @date 2015-09-16
+# @date 2018-11-09
#
###
@@ -38,11 +38,13 @@ set(ZSRC
cuda_zsymm.c
cuda_zsyr2k.c
cuda_zsyrk.c
+ cuda_ztpmlqt.c
cuda_ztpmqrt.c
cuda_ztrmm.c
cuda_ztrsm.c
cuda_ztsmlq.c
cuda_ztsmqr.c
+ cuda_zttmlq.c
cuda_zttmqr.c
cuda_zunmlqt.c
cuda_zunmqrt.c
diff --git a/cudablas/compute/cuda_zlarfb.c b/cudablas/compute/cuda_zlarfb.c
index b44b22ca9225b87dae0bc7cacf0de17050a6a2ef..51fb0f3c70c33d84ea01d9499f51eeac053a0791 100644
--- a/cudablas/compute/cuda_zlarfb.c
+++ b/cudablas/compute/cuda_zlarfb.c
@@ -15,21 +15,21 @@
*
* @version 1.0.0
* @author Florent Pruvost
- * @date 2015-09-16
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
#include "cudablas.h"
int
-CUDA_zlarfb(cham_side_t side, cham_trans_t trans,
- cham_dir_t direct, cham_store_t storev,
- int M, int N, int K,
- const cuDoubleComplex *V, int LDV,
- const cuDoubleComplex *T, int LDT,
- cuDoubleComplex *C, int LDC,
- cuDoubleComplex *WORK, int LDWORK,
- CUBLAS_STREAM_PARAM )
+CUDA_zlarfb( cham_side_t side, cham_trans_t trans,
+ cham_dir_t direct, cham_store_t storev,
+ int M, int N, int K,
+ const cuDoubleComplex *V, int LDV,
+ const cuDoubleComplex *T, int LDT,
+ cuDoubleComplex *C, int LDC,
+ cuDoubleComplex *WORK, int LDWORK,
+ CUBLAS_STREAM_PARAM )
{
#if defined(PRECISION_z) || defined(PRECISION_c)
cuDoubleComplex zzero = make_cuDoubleComplex(0.0, 0.0);
@@ -67,20 +67,25 @@ CUDA_zlarfb(cham_side_t side, cham_trans_t trans,
}
/* Quick return */
- if ((M == 0) || (N == 0) || (K == 0))
+ if ((M == 0) || (N == 0) || (K == 0)) {
return CHAMELEON_SUCCESS;
+ }
// opposite of trans
- if (trans == ChamNoTrans)
+ if (trans == ChamNoTrans) {
transT = ChamConjTrans;
- else
+ }
+ else {
transT = ChamNoTrans;
+ }
// whether T is upper or lower triangular
- if (direct == ChamDirForward)
+ if (direct == ChamDirForward) {
uplo = ChamUpper;
- else
+ }
+ else {
uplo = ChamLower;
+ }
if (storev == ChamColumnwise) {
notransV = ChamNoTrans;
@@ -106,8 +111,8 @@ CUDA_zlarfb(cham_side_t side, cham_trans_t trans,
// W = W T^H = C^H V T^H
CUDA_ztrmm( ChamRight, uplo, transT, ChamNonUnit,
N, K,
- CUBLAS_SADDR(zone), T, LDT,
- WORK, LDWORK,
+ &zone, T, LDT,
+ WORK, LDWORK,
CUBLAS_STREAM_VALUE );
// C = C - V W^H = C - V T V^H C = (I - V T V^H) C = H C
@@ -133,8 +138,8 @@ CUDA_zlarfb(cham_side_t side, cham_trans_t trans,
// W = W T = C V T
CUDA_ztrmm( ChamRight, uplo, trans, ChamNonUnit,
M, K,
- CUBLAS_SADDR(zone), T, LDT,
- WORK, LDWORK,
+ &zone, T, LDT,
+ WORK, LDWORK,
CUBLAS_STREAM_VALUE );
// C = C - W V^H = C - C V T V^H = C (I - V T V^H) = C H
diff --git a/cudablas/compute/cuda_zparfb.c b/cudablas/compute/cuda_zparfb.c
index 292ac3b647cebbee27d15f7336ed80a21c8de9b0..bcef47797c5892accf0073aa9e2fd67b2cb6ab1b 100644
--- a/cudablas/compute/cuda_zparfb.c
+++ b/cudablas/compute/cuda_zparfb.c
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Florent Pruvost
- * @date 2015-09-16
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
@@ -120,40 +120,32 @@
* The leading dimension of the array T. LDT >= K.
*
* @param[in,out] WORK
- * Workspace of dimension LDWORK-by-N1 if side == ChamLeft, LDWORK-by-K
- * otherwise.
+ * Workspace of dimension at least:
+ * - K * (M2 + N2).
+ * If L > 0, it is recommended to extend it to
+ * - K * (2 * M2 + N2 ) if side == ChamLeft.
+ * - K * (M2 + 2 * N2 ) if side == ChamRight.
*
- * @param[in] LDWORK
- * The leading dimension of the array WORK: LDWORK >= K, if side ==
- * ChamLeft, LDWORK >= M1 otehrwise.
- *
- * @param[in,out] WORKC
- * Optionnal additional workspace to replace the TRMM operation by a GEMM kernel.
- * This workspace is of dimension LDWORK-by-K if side == ChamLeft, LDWORK-by-N2
- * otherwise.
- *
- * @param[in] LDWORKC
- * The leading dimension of the array WORKC: LDWORKC >= M2, if side ==
- * ChamLeft, LDWORK >= K otehrwise.
+ * @param[in] LWORK
+ * The dimension of the array WORK. If LWORK < 0, returns immediately
+ * the recommended workspace size.
*
*******************************************************************************
*
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
+ * @retval CHAMELEON_SUCCESS successful exit
+ * @retval <0 if -i, the i-th argument had an illegal value
+ * @retval The recommended LWORK value, if LWORK == -1 on entry.
*/
int
-CUDA_zparfb(cham_side_t side, cham_trans_t trans,
- cham_dir_t direct, cham_store_t storev,
- int M1, int N1, int M2, int N2, int K, int L,
- cuDoubleComplex *A1, int LDA1,
- cuDoubleComplex *A2, int LDA2,
- const cuDoubleComplex *V, int LDV,
- const cuDoubleComplex *T, int LDT,
- cuDoubleComplex *WORK, int LDWORK,
- cuDoubleComplex *WORKC, int LDWORKC,
- CUBLAS_STREAM_PARAM )
+CUDA_zparfb( cham_side_t side, cham_trans_t trans,
+ cham_dir_t direct, cham_store_t storev,
+ int M1, int N1, int M2, int N2, int K, int L,
+ cuDoubleComplex *A1, int LDA1,
+ cuDoubleComplex *A2, int LDA2,
+ const cuDoubleComplex *V, int LDV,
+ const cuDoubleComplex *T, int LDT,
+ cuDoubleComplex *WORK, int LWORK,
+ CUBLAS_STREAM_PARAM )
{
#if defined(PRECISION_z) || defined(PRECISION_c)
cuDoubleComplex zzero = make_cuDoubleComplex(0.0, 0.0);
@@ -165,9 +157,13 @@ CUDA_zparfb(cham_side_t side, cham_trans_t trans,
double mzone = -1.0;
#endif /* defined(PRECISION_z) || defined(PRECISION_c) */
+ cuDoubleComplex *workW, *workC, *workV;
+ int ldW, ldC, ldV;
int j;
cham_trans_t transW;
cham_trans_t transA2;
+ int wssize = 0;
+ int wrsize = 0;
CUBLAS_GET_STREAM;
@@ -201,19 +197,30 @@ CUDA_zparfb(cham_side_t side, cham_trans_t trans,
if (K < 0) {
return -9;
}
- if ( ((LDWORK < K ) && (side == ChamLeft )) ||
- ((LDWORK < M1) && (side == ChamRight)) ) {
+
+ if (direct == ChamDirForward) {
+ wssize = K * (M2 + N2);
+ wrsize = wssize;
+ if ( L > 0 ) {
+ wrsize += (side == ChamLeft) ? M2 * K : K * N2;
+ }
+ }
+
+ if ( LWORK < 0 ) {
+ return wrsize;
+ }
+ else if ( LWORK < wssize ) {
+ cudablas_error(20, "Illegal value of LWORK");
return -20;
}
- /* Quick return */
- if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0))
+ if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0)) {
return CHAMELEON_SUCCESS;
+ }
if (direct == ChamDirForward) {
if (side == ChamLeft) {
-
/*
* Column or Rowwise / Forward / Left
* ----------------------------------
@@ -222,76 +229,137 @@ CUDA_zparfb(cham_side_t side, cham_trans_t trans,
* ( A2 )
*/
+ /*
+ * Store in WORK (N1 == N2):
+ * - Workspace W for the copy of A1 + V' * A2 (K x N1)
+ * - Workspace C for the copy of V * T (M2 x K )
+ * - Workspace V for the copy of V (M2 x K )
+ */
+ workW = WORK;
+ ldW = K;
+
+ workC = workW + K * N1;
+ ldC = M2;
+
+ if ( L == 0 ) {
+ workV = (cuDoubleComplex*)V;
+ ldV = LDV;
+ }
+ else {
+ if ( LWORK < wrsize ) {
+ workC = NULL;
+ workV = workW + K * N1;
+ }
+ else {
+ workV = workC + M2 * K;
+ }
+
+ if ( storev == ChamColumnwise ) {
+ ldV = M2;
+
+ /*
+ * Backup V, and put 0 in the lower part
+ */
+ cudaMemcpy2DAsync( workV, ldV * sizeof(cuDoubleComplex),
+ V, LDV * sizeof(cuDoubleComplex),
+ M2 * sizeof(cuDoubleComplex), K,
+ cudaMemcpyDeviceToDevice, stream );
+
+ for(j = 1; j < K; j++) {
+ cudaMemsetAsync( workV + (j-1) * ldV + M2 - L + j,
+ 0,
+ (L - j) * sizeof(cuDoubleComplex),
+ stream );
+ }
+ }
+ else {
+ ldV = K;
+
+ /*
+ * Backup V, and put 0 in the lower part
+ */
+ cudaMemcpy2DAsync( workV, ldV * sizeof(cuDoubleComplex),
+ V, LDV * sizeof(cuDoubleComplex),
+ K * sizeof(cuDoubleComplex), M2,
+ cudaMemcpyDeviceToDevice, stream );
+
+ for(j = 1; j < K; j++) {
+ cudaMemsetAsync( workV + ldV * ( M2 - L + j ),
+ 0,
+ j * sizeof(cuDoubleComplex),
+ stream );
+ }
+ }
+ }
+
/*
* W = A1 + V' * A2:
* W = A1
* W = W + V' * A2
*
*/
- cudaMemcpy2DAsync( WORK, LDWORK * sizeof(cuDoubleComplex),
- A1, LDA1 * sizeof(cuDoubleComplex),
+ cudaMemcpy2DAsync( workW, ldW * sizeof(cuDoubleComplex),
+ A1, LDA1 * sizeof(cuDoubleComplex),
K * sizeof(cuDoubleComplex), N1,
cudaMemcpyDeviceToDevice, stream );
transW = storev == ChamColumnwise ? ChamConjTrans : ChamNoTrans;
transA2 = storev == ChamColumnwise ? ChamNoTrans : ChamConjTrans;
- cublasZgemm(CUBLAS_HANDLE
- chameleon_cublas_const(transW), chameleon_cublas_const(ChamNoTrans),
- K, N1, M2,
- CUBLAS_SADDR(zone),
- V /* K*M2 */, LDV,
- A2 /* M2*N1 */, LDA2,
- CUBLAS_SADDR(zone),
- WORK /* K*N1 */, LDWORK);
-
- if (WORKC == NULL) {
+ cublasZgemm( CUBLAS_HANDLE
+ chameleon_cublas_const(transW), chameleon_cublas_const(ChamNoTrans),
+ K, N1, M2,
+ CUBLAS_SADDR(zone), workV /* M2*K */, ldV,
+ A2 /* M2*N2 */, LDA2,
+ CUBLAS_SADDR(zone), workW /* K *N2 */, ldW );
+
+ if ( workC == NULL ) {
/* W = op(T) * W */
CUDA_ztrmm( ChamLeft, ChamUpper, trans, ChamNonUnit,
K, N2,
- CUBLAS_SADDR(zone), T, LDT,
- WORK, LDWORK,
+ &zone, T, LDT,
+ workW, ldW,
CUBLAS_STREAM_VALUE );
/* A1 = A1 - W = A1 - op(T) * W */
for(j = 0; j < N1; j++) {
- cublasZaxpy(CUBLAS_HANDLE
- K, CUBLAS_SADDR(mzone),
- (WORK + LDWORK*j), 1,
- (A1 + LDA1*j), 1);
+ cublasZaxpy( CUBLAS_HANDLE
+ K, CUBLAS_SADDR(mzone),
+ workW + ldW * j, 1,
+ A1 + LDA1 * j, 1 );
}
/* A2 = A2 - op(V) * W */
- cublasZgemm(CUBLAS_HANDLE
- chameleon_cublas_const(transA2), chameleon_cublas_const(ChamNoTrans),
- M2, N2, K,
- CUBLAS_SADDR(mzone), V /* M2*K */, LDV,
- WORK /* K*N2 */, LDWORK,
- CUBLAS_SADDR(zone), A2 /* m2*N2 */, LDA2);
+ cublasZgemm( CUBLAS_HANDLE
+ chameleon_cublas_const(transA2), chameleon_cublas_const(ChamNoTrans),
+ M2, N2, K,
+ CUBLAS_SADDR(mzone), workV /* M2 * K */, ldV,
+ workW /* K * N2 */, ldW,
+ CUBLAS_SADDR(zone), A2 /* M2 * N2 */, LDA2 );
} else {
/* Wc = V * op(T) */
cublasZgemm( CUBLAS_HANDLE
chameleon_cublas_const(transA2), chameleon_cublas_const(trans),
M2, K, K,
- CUBLAS_SADDR(zone), V, LDV,
- T, LDT,
- CUBLAS_SADDR(zzero), WORKC, LDWORKC );
+ CUBLAS_SADDR(zone), workV, ldV,
+ T, LDT,
+ CUBLAS_SADDR(zzero), workC, ldC );
/* A1 = A1 - opt(T) * W */
cublasZgemm( CUBLAS_HANDLE
chameleon_cublas_const(trans), chameleon_cublas_const(ChamNoTrans),
K, N1, K,
- CUBLAS_SADDR(mzone), T, LDT,
- WORK, LDWORK,
- CUBLAS_SADDR(zone), A1, LDA1 );
+ CUBLAS_SADDR(mzone), T, LDT,
+ workW, ldW,
+ CUBLAS_SADDR(zone), A1, LDA1 );
/* A2 = A2 - Wc * W */
cublasZgemm( CUBLAS_HANDLE
chameleon_cublas_const(ChamNoTrans), chameleon_cublas_const(ChamNoTrans),
M2, N2, K,
- CUBLAS_SADDR(mzone), WORKC, LDWORKC,
- WORK, LDWORK,
+ CUBLAS_SADDR(mzone), workC, ldC,
+ workW, ldW,
CUBLAS_SADDR(zone), A2, LDA2 );
}
}
@@ -304,14 +372,77 @@ CUDA_zparfb(cham_side_t side, cham_trans_t trans,
*
*/
+ /*
+ * Store in WORK (M1 == M2):
+ * - Workspace W for the copy of A1 + A2 * V' (M1 x K )
+ * - Workspace C for the copy of V * T (K x N2)
+ * - Workspace V for the copy of V (K x N2)
+ */
+ workW = WORK;
+ ldW = M1;
+
+ workC = workW + M1 * K;
+ ldC = K;
+
+ if ( L == 0 ) {
+ workV = (cuDoubleComplex*)V;
+ ldV = LDV;
+ }
+ else {
+ if ( LWORK < wrsize ) {
+ workC = NULL;
+ workV = workW + M2 * K;
+ }
+ else {
+ workV = workC + K * N2;
+ }
+
+ if ( storev == ChamColumnwise ) {
+ ldV = N2;
+
+ /*
+ * Backup V, and put 0 in the lower part
+ */
+ cudaMemcpy2DAsync( workV, ldV * sizeof(cuDoubleComplex),
+ V, LDV * sizeof(cuDoubleComplex),
+ N2 * sizeof(cuDoubleComplex), K,
+ cudaMemcpyDeviceToDevice, stream );
+
+ for(j = 1; j < K; j++) {
+ cudaMemsetAsync( workV + (j-1) * ldV + N2 - L + j,
+ 0,
+ (L - j) * sizeof(cuDoubleComplex),
+ stream );
+ }
+ }
+ else {
+ ldV = K;
+
+ /*
+ * Backup V, and put 0 in the upper part
+ */
+ cudaMemcpy2DAsync( workV, ldV * sizeof(cuDoubleComplex),
+ V, LDV * sizeof(cuDoubleComplex),
+ K * sizeof(cuDoubleComplex), N2,
+ cudaMemcpyDeviceToDevice, stream );
+
+ for(j = 1; j < K; j++) {
+ cudaMemsetAsync( workV + ldV * ( N2 - L + j ),
+ 0,
+ j * sizeof(cuDoubleComplex),
+ stream );
+ }
+ }
+ }
+
/*
* W = A1 + A2 * V':
* W = A1
* W = W + A2 * V'
*
*/
- cudaMemcpy2DAsync( WORK, LDWORK * sizeof(cuDoubleComplex),
- A1, LDA1 * sizeof(cuDoubleComplex),
+ cudaMemcpy2DAsync( workW, ldW * sizeof(cuDoubleComplex),
+ A1, LDA1 * sizeof(cuDoubleComplex),
M1 * sizeof(cuDoubleComplex), K,
cudaMemcpyDeviceToDevice, stream );
@@ -321,40 +452,40 @@ CUDA_zparfb(cham_side_t side, cham_trans_t trans,
cublasZgemm(CUBLAS_HANDLE
chameleon_cublas_const(ChamNoTrans), chameleon_cublas_const(transW),
M1, K, N2,
- CUBLAS_SADDR(zone), A2 /* M1*N2 */, LDA2,
- V /* N2*K */, LDV,
- CUBLAS_SADDR(zone), WORK /* M1*K */, LDWORK);
+ CUBLAS_SADDR(zone), A2 /* M1*N2 */, LDA2,
+ workV /* K *N2 */, ldV,
+ CUBLAS_SADDR(zone), workW /* M1*K */, ldW);
- if (WORKC == NULL) {
+ if ( workC == NULL ) {
/* W = W * op(T) */
CUDA_ztrmm( ChamRight, ChamUpper, trans, ChamNonUnit,
M2, K,
- CUBLAS_SADDR(zone), T, LDT,
- WORK, LDWORK,
+ &zone, T, LDT,
+ workW, ldW,
CUBLAS_STREAM_VALUE );
/* A1 = A1 - W = A1 - W * op(T) */
for(j = 0; j < K; j++) {
- cublasZaxpy(CUBLAS_HANDLE
- M1, CUBLAS_SADDR(mzone),
- (WORK + LDWORK*j), 1,
- (A1 + LDA1*j), 1);
+ cublasZaxpy( CUBLAS_HANDLE
+ M1, CUBLAS_SADDR(mzone),
+ workW + ldW * j, 1,
+ A1 + LDA1 * j, 1 );
}
/* A2 = A2 - W * op(V) */
cublasZgemm(CUBLAS_HANDLE
chameleon_cublas_const(ChamNoTrans), chameleon_cublas_const(transA2),
M2, N2, K,
- CUBLAS_SADDR(mzone), WORK /* M2*K */, LDWORK,
- V /* K*N2 */, LDV,
- CUBLAS_SADDR(zone), A2 /* M2*N2 */, LDA2);
+ CUBLAS_SADDR(mzone), workW /* M2*K */, ldW,
+ workV /* K *N2 */, ldV,
+ CUBLAS_SADDR(zone), A2 /* M2*N2 */, LDA2);
} else {
/* A1 = A1 - W * opt(T) */
cublasZgemm( CUBLAS_HANDLE
chameleon_cublas_const(ChamNoTrans), chameleon_cublas_const(trans),
M1, K, K,
- CUBLAS_SADDR(mzone), WORK, LDWORK,
+ CUBLAS_SADDR(mzone), workW, ldW,
T, LDT,
CUBLAS_SADDR(zone), A1, LDA1 );
@@ -363,15 +494,15 @@ CUDA_zparfb(cham_side_t side, cham_trans_t trans,
chameleon_cublas_const(trans), chameleon_cublas_const(transA2),
K, N2, K,
CUBLAS_SADDR(zone), T, LDT,
- V, LDV,
- CUBLAS_SADDR(zzero), WORKC, LDWORKC );
+ workV, ldV,
+ CUBLAS_SADDR(zzero), workC, ldC );
/* A2 = A2 - W * Wc */
cublasZgemm( CUBLAS_HANDLE
chameleon_cublas_const(ChamNoTrans), chameleon_cublas_const(ChamNoTrans),
M2, N2, K,
- CUBLAS_SADDR(mzone), WORK, LDWORK,
- WORKC, LDWORKC,
+ CUBLAS_SADDR(mzone), workW, ldW,
+ workC, ldC,
CUBLAS_SADDR(zone), A2, LDA2 );
}
}
diff --git a/cudablas/compute/cuda_ztpmlqt.c b/cudablas/compute/cuda_ztpmlqt.c
new file mode 100644
index 0000000000000000000000000000000000000000..4f01e0e28cf34690cc3e3daa1823c418e9557fac
--- /dev/null
+++ b/cudablas/compute/cuda_ztpmlqt.c
@@ -0,0 +1,184 @@
+/**
+ *
+ * @file cuda_ztpmlqt.c
+ *
+ * @copyright 2009-2016 The University of Tennessee and The University of
+ * Tennessee Research Foundation. All rights reserved.
+ * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
+ * Univ. Bordeaux. All rights reserved.
+ *
+ ***
+ *
+ * @brief Chameleon cuda_ztpmlqt GPU kernel
+ *
+ * @version 1.0.0
+ * @author Mathieu Faverge
+ * @date 2018-11-09
+ * @precisions normal z -> c d s
+ *
+ */
+#include "cudablas.h"
+
+/**
+ *******************************************************************************
+ *
+ * @ingroup CORE_CHAMELEON_Complex64_t
+ *
+ * @brief Applies a complex orthogonal matrix Q.
+ *
+ * The matrix Q is obtained from a "triangular-pentagonal" complex block
+ * reflector H to a general complex matrix C, which consists of two blocks A and
+ * B.
+ *
+ *******************************************************************************
+ *
+ * @param[in] side
+ * @arg ChamLeft : apply Q or Q**H from the Left;
+ * @arg ChamRight : apply Q or Q**H from the Right.
+ *
+ * @param[in] trans
+ * @arg ChamNoTrans : No transpose, apply Q;
+ * @arg ChamConjTrans : ConjTranspose, apply Q**H.
+ *
+ * @param[in] M
+ * The number of rows of the tile B. M >= 0.
+ *
+ * @param[in] N
+ * The number of columns of the tile B. N >= 0.
+ *
+ * @param[in] K
+ * The number of elementary reflectors whose product defines
+ * the matrix Q.
+ *
+ * @param[in] L
+ * The number of rows of the upper trapezoidal part of V.
+ * K >= L >= 0. See Further Details.
+ *
+ * @param[in] IB
+ * The inner-blocking size. IB >= 0.
+ *
+ * @param[in] V
+ * The i-th row must contain the vector which defines the
+ * elementary reflector H(i), for i = 1,2,...,k, as returned by
+ * CORE_ZTPQRT in the first k rows of its array argument V.
+ *
+ * @param[in] LDV
+ * The leading dimension of the array V. LDV >= max(1,K).
+ *
+ * @param[in] T
+ * The IB-by-N1 triangular factor T of the block reflector.
+ * T is upper triangular by block (economic storage);
+ * The rest of the array is not referenced.
+ *
+ * @param[in] LDT
+ * The leading dimension of the array T. LDT >= IB.
+ *
+ * @param[in,out] A
+ * A is COMPLEX*16 array, dimension (LDA,N) if side = ChamLeft
+ * or (LDA,K) if SIDE = ChamRight
+ * On entry, the K-by-N or M-by-K matrix A.
+ * On exit, A is overwritten by the corresponding block of
+ * Q*C or Q**H*C or C*Q or C*Q**H. See Further Details.
+ *
+ * @param[in] LDA
+ * The leading dimension of the array A. LDA >= max(1,M).
+ * If side = ChamLeft, LDA >= max(1,K);
+ * If side = Chamright, LDA >= max(1,M).
+ *
+ * @param[in,out] B
+ * On entry, the M-by-N tile B.
+ * On exit, B is overwritten by the corresponding block of
+ * Q*C or Q**H*C or C*Q or C*Q**H. See Further Details.
+ *
+ * @param[in] LDB
+ * The leading dimension of the tile B. LDB >= max(1,M).
+ *
+ * @param[out] WORK
+ * Workspace array of size LDWORK-by-NB.
+ * LDWORK = N if side = ChamLeft, or M if side = ChamRight.
+ *
+ *******************************************************************************
+ *
+ * @par Further Details:
+ * =====================
+ *
+ * The columns of the pentagonal matrix V contain the elementary reflectors
+ * H(1), H(2), ..., H(K); V is composed of a rectangular block V1 and a
+ * trapezoidal block V2:
+ *
+ * V = [V1] [V2].
+ *
+ * The size of the trapezoidal block V2 is determined by the parameter L,
+ * where 0 <= L <= K; V2 is lower trapezoidal, consisting of the first L
+ * rows of a K-by-K upper triangular matrix. If L=K, V2 is lower triangular;
+ * if L=0, there is no trapezoidal block, hence V = V1 is rectangular.
+ *
+ * If side = ChamLeft: C = [A] where A is K-by-N, B is M-by-N and V is K-by-M.
+ * [B]
+ *
+ * If side = ChamRight: C = [A B] where A is M-by-K, B is M-by-N and V is K-by-N.
+ *
+ * The complex orthogonal matrix Q is formed from V and T.
+ *
+ * If trans='N' and side='L', C is on exit replaced with Q * C.
+ *
+ * If trans='C' and side='L', C is on exit replaced with Q**H * C.
+ *
+ * If trans='N' and side='R', C is on exit replaced with C * Q.
+ *
+ * If trans='C' and side='R', C is on exit replaced with C * Q**H.
+ *
+ *******************************************************************************
+ *
+ * @retval CHAMELEON_SUCCESS successful exit
+ * @retval <0 if -i, the i-th argument had an illegal value
+ *
+ */
+int
+CUDA_ztpmlqt( cham_side_t side, cham_trans_t trans,
+ int M, int N, int K, int L, int IB,
+ const cuDoubleComplex *V, int LDV,
+ const cuDoubleComplex *T, int LDT,
+ cuDoubleComplex *A, int LDA,
+ cuDoubleComplex *B, int LDB,
+ cuDoubleComplex *WORK, int lwork,
+ CUBLAS_STREAM_PARAM )
+{
+ int m1, n1;
+
+ /* Check input arguments */
+ if ((side != ChamLeft) && (side != ChamRight)) {
+ cudablas_error(1, "Illegal value of side");
+ return -1;
+ }
+
+ if ( side == ChamLeft ) {
+ m1 = K;
+ n1 = N;
+ }
+ else {
+ m1 = M;
+ n1 = K;
+ }
+
+ /* TS case */
+ if (L == 0) {
+ CUDA_ztsmlq( side, trans, m1, n1, M, N, K, IB,
+ A, LDA, B, LDB, V, LDV, T, LDT,
+ WORK, lwork,
+ CUBLAS_STREAM_VALUE );
+ }
+ /* TT case */
+ else if( L == N ) {
+ CUDA_zttmlq( side, trans, m1, n1, M, N, K, IB,
+ A, LDA, B, LDB, V, LDV, T, LDT,
+ WORK, lwork,
+ CUBLAS_STREAM_VALUE );
+ }
+ else {
+ cudablas_error(-6, "TPMLQT not available on GPU for general cases yet\n" );
+ return -6;
+ }
+
+ return CHAMELEON_SUCCESS;
+}
diff --git a/cudablas/compute/cuda_ztpmqrt.c b/cudablas/compute/cuda_ztpmqrt.c
index 2719edd32b1fef478dd150f67b89ba0c8dc465e2..c7a19fd76726491b982c0da9be7c0ff444fd829c 100644
--- a/cudablas/compute/cuda_ztpmqrt.c
+++ b/cudablas/compute/cuda_ztpmqrt.c
@@ -13,12 +13,128 @@
*
* @version 1.0.0
* @author Florent Pruvost
- * @date 2015-09-16
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
#include "cudablas.h"
+/**
+ *******************************************************************************
+ *
+ * @ingroup CORE_CHAMELEON_Complex64_t
+ *
+ * @brief Applies a complex orthogonal matrix Q.
+ *
+ * The matrix Q is obtained from a "triangular-pentagonal" complex block
+ * reflector H to a general complex matrix C, which consists of two blocks A and
+ * B.
+ *
+ *******************************************************************************
+ *
+ * @param[in] side
+ * @arg ChamLeft : apply Q or Q**H from the Left;
+ * @arg ChamRight : apply Q or Q**H from the Right.
+ *
+ * @param[in] trans
+ * @arg ChamNoTrans : No transpose, apply Q;
+ * @arg ChamConjTrans : ConjTranspose, apply Q**H.
+ *
+ * @param[in] M
+ * The number of rows of the tile B. M >= 0.
+ *
+ * @param[in] N
+ * The number of columns of the tile B. N >= 0.
+ *
+ * @param[in] K
+ * The number of elementary reflectors whose product defines
+ * the matrix Q.
+ *
+ * @param[in] L
+ * The number of rows of the upper trapezoidal part of V.
+ * K >= L >= 0. See Further Details.
+ *
+ * @param[in] IB
+ * The inner-blocking size. IB >= 0.
+ *
+ * @param[in] V
+ * The i-th row must contain the vector which defines the
+ * elementary reflector H(i), for i = 1,2,...,k, as returned by
+ * CORE_ZTPQRT in the first k rows of its array argument V.
+ *
+ * @param[in] LDV
+ * The leading dimension of the array V. LDV >= max(1,K).
+ *
+ * @param[in] T
+ * The IB-by-N1 triangular factor T of the block reflector.
+ * T is upper triangular by block (economic storage);
+ * The rest of the array is not referenced.
+ *
+ * @param[in] LDT
+ * The leading dimension of the array T. LDT >= IB.
+ *
+ * @param[in,out] A
+ * A is COMPLEX*16 array, dimension (LDA,N) if side = ChamLeft
+ * or (LDA,K) if SIDE = ChamRight
+ * On entry, the K-by-N or M-by-K matrix A.
+ * On exit, A is overwritten by the corresponding block of
+ * Q*C or Q**H*C or C*Q or C*Q**H. See Further Details.
+ *
+ * @param[in] LDA
+ * The leading dimension of the array A. LDA >= max(1,M).
+ * If side = ChamLeft, LDA >= max(1,K);
+ * If side = Chamright, LDA >= max(1,M).
+ *
+ * @param[in,out] B
+ * On entry, the M-by-N tile B.
+ * On exit, B is overwritten by the corresponding block of
+ * Q*C or Q**H*C or C*Q or C*Q**H. See Further Details.
+ *
+ * @param[in] LDB
+ * The leading dimension of the tile B. LDB >= max(1,M).
+ *
+ * @param[out] WORK
+ * Workspace array of size LDWORK-by-NB.
+ * LDWORK = N if side = ChamLeft, or M if side = ChamRight.
+ *
+ *******************************************************************************
+ *
+ * @par Further Details:
+ * =====================
+ *
+ * The columns of the pentagonal matrix V contain the elementary reflectors
+ * H(1), H(2), ..., H(K); V is composed of a rectangular block V1 and a
+ * trapezoidal block V2:
+ *
+ * V = [V1]
+ * [V2].
+ *
+ * The size of the trapezoidal block V2 is determined by the parameter L,
+ * where 0 <= L <= K; V2 is upper trapezoidal, consisting of the first L
+ * rows of a K-by-K upper triangular matrix. If L=K, V2 is upper triangular;
+ * if L=0, there is no trapezoidal block, hence V = V1 is rectangular.
+ *
+ * If side = ChamLeft: C = [A] where A is K-by-N, B is M-by-N and V is M-by-K.
+ * [B]
+ *
+ * If side = ChamRight: C = [A B] where A is M-by-K, B is M-by-N and V is N-by-K.
+ *
+ * The complex orthogonal matrix Q is formed from V and T.
+ *
+ * If trans='N' and side='L', C is on exit replaced with Q * C.
+ *
+ * If trans='C' and side='L', C is on exit replaced with Q**H * C.
+ *
+ * If trans='N' and side='R', C is on exit replaced with C * Q.
+ *
+ * If trans='C' and side='R', C is on exit replaced with C * Q**H.
+ *
+ *******************************************************************************
+ *
+ * @retval CHAMELEON_SUCCESS successful exit
+ * @retval <0 if -i, the i-th argument had an illegal value
+ *
+ */
int
CUDA_ztpmqrt( cham_side_t side, cham_trans_t trans,
int M, int N, int K, int L, int IB,
@@ -26,10 +142,10 @@ CUDA_ztpmqrt( cham_side_t side, cham_trans_t trans,
const cuDoubleComplex *T, int LDT,
cuDoubleComplex *A, int LDA,
cuDoubleComplex *B, int LDB,
- cuDoubleComplex *WORK,
+ cuDoubleComplex *WORK, int lwork,
CUBLAS_STREAM_PARAM )
{
- int m1, n1, ldwork, ldworkc, ws;
+ int m1, n1;
/* Check input arguments */
if ((side != ChamLeft) && (side != ChamRight)) {
@@ -40,30 +156,24 @@ CUDA_ztpmqrt( cham_side_t side, cham_trans_t trans,
if ( side == ChamLeft ) {
m1 = K;
n1 = N;
- ldwork = IB;
- ldworkc = M;
- ws = ldwork * n1;
}
else {
m1 = M;
n1 = K;
- ldwork = chameleon_max( K, chameleon_max( M, N ) );
- ldworkc = IB;
- ws = ldwork * IB;
}
/* TS case */
if (L == 0) {
CUDA_ztsmqr( side, trans, m1, n1, M, N, K, IB,
A, LDA, B, LDB, V, LDV, T, LDT,
- WORK, ldwork, WORK + ws, ldworkc,
+ WORK, lwork,
CUBLAS_STREAM_VALUE );
}
/* TT case */
else if( L == M ) {
CUDA_zttmqr( side, trans, m1, n1, M, N, K, IB,
A, LDA, B, LDB, V, LDV, T, LDT,
- WORK, ldwork, WORK + ws, ldworkc,
+ WORK, lwork,
CUBLAS_STREAM_VALUE );
}
else {
diff --git a/cudablas/compute/cuda_ztrmm.c b/cudablas/compute/cuda_ztrmm.c
index 390311e0834fd5b126f912b033b2dc5748c6b2e9..b2809c779ce0e00003a3d5b22e26e1fefb63805f 100644
--- a/cudablas/compute/cuda_ztrmm.c
+++ b/cudablas/compute/cuda_ztrmm.c
@@ -20,13 +20,13 @@
#include "cudablas.h"
int CUDA_ztrmm(
- cham_side_t side, cham_uplo_t uplo,
- cham_trans_t transa, cham_diag_t diag,
- int m, int n,
- cuDoubleComplex *alpha,
- const cuDoubleComplex *A, int lda,
- cuDoubleComplex *B, int ldb,
- CUBLAS_STREAM_PARAM)
+ cham_side_t side, cham_uplo_t uplo,
+ cham_trans_t transa, cham_diag_t diag,
+ int m, int n,
+ cuDoubleComplex *alpha,
+ const cuDoubleComplex *A, int lda,
+ cuDoubleComplex *B, int ldb,
+ CUBLAS_STREAM_PARAM)
{
#if defined(CHAMELEON_USE_CUBLAS_V2)
diff --git a/cudablas/compute/cuda_ztsmlq.c b/cudablas/compute/cuda_ztsmlq.c
index 7d2e3c4a3fbc19d5ec319991a9d56a537608016b..2dcfc24158a77a0427ed00bd294a72b55a97f238 100644
--- a/cudablas/compute/cuda_ztsmlq.c
+++ b/cudablas/compute/cuda_ztsmlq.c
@@ -13,27 +13,26 @@
*
* @version 1.0.0
* @author Florent Pruvost
- * @date 2015-09-16
+ * @author Mathieu Faverge
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
#include "cudablas.h"
int CUDA_ztsmlq(
- cham_side_t side, cham_trans_t trans,
- int M1, int N1,
- int M2, int N2,
- int K, int IB,
- cuDoubleComplex *A1, int LDA1,
- cuDoubleComplex *A2, int LDA2,
- const cuDoubleComplex *V, int LDV,
- const cuDoubleComplex *T, int LDT,
- cuDoubleComplex *WORK, int LDWORK,
- cuDoubleComplex *WORKC, int LDWORKC,
- CUBLAS_STREAM_PARAM)
+ cham_side_t side, cham_trans_t trans,
+ int M1, int N1,
+ int M2, int N2,
+ int K, int IB,
+ cuDoubleComplex *A1, int LDA1,
+ cuDoubleComplex *A2, int LDA2,
+ const cuDoubleComplex *V, int LDV,
+ const cuDoubleComplex *T, int LDT,
+ cuDoubleComplex *WORK, int LWORK,
+ CUBLAS_STREAM_PARAM)
{
int i, i1, i3;
- int NW;
int kb;
int ic = 0;
int jc = 0;
@@ -45,14 +44,6 @@ int CUDA_ztsmlq(
return -1;
}
- /* NW is the minimum dimension of WORK */
- if (side == ChamLeft) {
- NW = IB;
- }
- else {
- NW = N1;
- }
-
if ((trans != ChamNoTrans) && (trans != ChamConjTrans)) {
return -2;
}
@@ -90,21 +81,20 @@ int CUDA_ztsmlq(
if (LDT < chameleon_max(1,IB)){
return -16;
}
- if (LDWORK < chameleon_max(1,NW)){
- return -18;
- }
/* Quick return */
- if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0))
+ if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0)) {
return CHAMELEON_SUCCESS;
+ }
- if (((side == ChamLeft) && (trans == ChamNoTrans))
- || ((side == ChamRight) && (trans != ChamNoTrans))) {
+ if ( ((side == ChamLeft ) && (trans == ChamNoTrans)) ||
+ ((side == ChamRight) && (trans != ChamNoTrans)) )
+ {
i1 = 0;
i3 = IB;
}
else {
- i1 = ((K-1) / IB)*IB;
+ i1 = ( ( K-1 ) / IB )*IB;
i3 = -IB;
}
@@ -115,7 +105,7 @@ int CUDA_ztsmlq(
trans = ChamNoTrans;
}
- for(i = i1; (i > -1) && (i < K); i += i3) {
+ for (i = i1; (i > -1) && (i < K); i+=i3) {
kb = chameleon_min(IB, K-i);
if (side == ChamLeft) {
@@ -137,13 +127,13 @@ int CUDA_ztsmlq(
* Apply H or H' (NOTE: CORE_zparfb used to be CORE_ztsrfb)
*/
CUDA_zparfb(
- side, trans, ChamDirForward, ChamRowwise,
- mi, ni, M2, N2, kb, 0,
- A1 + LDA1*jc+ic, LDA1,
- A2, LDA2,
- V + i, LDV,
- T + LDT*i, LDT,
- WORK, LDWORK, WORKC, LDWORKC, CUBLAS_STREAM_VALUE );
+ side, trans, ChamDirForward, ChamRowwise,
+ mi, ni, M2, N2, kb, 0,
+ A1 + LDA1*jc+ic, LDA1,
+ A2, LDA2,
+ V + i, LDV,
+ T + LDT*i, LDT,
+ WORK, LWORK, CUBLAS_STREAM_VALUE );
}
return CHAMELEON_SUCCESS;
}
diff --git a/cudablas/compute/cuda_ztsmqr.c b/cudablas/compute/cuda_ztsmqr.c
index 4b07b9b9ce16397eabf8e358d81b7499eeab2c39..e731dfbf451bb06ab0d9f519292b529c04bfe00a 100644
--- a/cudablas/compute/cuda_ztsmqr.c
+++ b/cudablas/compute/cuda_ztsmqr.c
@@ -13,27 +13,27 @@
*
* @version 1.0.0
* @author Florent Pruvost
- * @date 2015-09-16
+ * @author Mathieu Faverge
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
#include "cudablas.h"
int CUDA_ztsmqr(
- cham_side_t side, cham_trans_t trans,
- int M1, int N1,
- int M2, int N2,
- int K, int IB,
- cuDoubleComplex *A1, int LDA1,
- cuDoubleComplex *A2, int LDA2,
- const cuDoubleComplex *V, int LDV,
- const cuDoubleComplex *T, int LDT,
- cuDoubleComplex *WORK, int LDWORK,
- cuDoubleComplex *WORKC, int LDWORKC,
- CUBLAS_STREAM_PARAM)
+ cham_side_t side, cham_trans_t trans,
+ int M1, int N1,
+ int M2, int N2,
+ int K, int IB,
+ cuDoubleComplex *A1, int LDA1,
+ cuDoubleComplex *A2, int LDA2,
+ const cuDoubleComplex *V, int LDV,
+ const cuDoubleComplex *T, int LDT,
+ cuDoubleComplex *WORK, int LWORK,
+ CUBLAS_STREAM_PARAM)
{
int i, i1, i3;
- int NQ, NW;
+ int NQ;
int kb;
int ic = 0;
int jc = 0;
@@ -48,11 +48,9 @@ int CUDA_ztsmqr(
/* NQ is the order of Q */
if (side == ChamLeft) {
NQ = M2;
- NW = IB;
}
else {
NQ = N2;
- NW = M1;
}
if ((trans != ChamNoTrans) && (trans != ChamConjTrans)) {
@@ -92,25 +90,24 @@ int CUDA_ztsmqr(
if (LDT < chameleon_max(1,IB)){
return -16;
}
- if (LDWORK < chameleon_max(1,NW)){
- return -18;
- }
/* Quick return */
- if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0))
+ if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0)) {
return CHAMELEON_SUCCESS;
+ }
- if (((side == ChamLeft) && (trans != ChamNoTrans))
- || ((side == ChamRight) && (trans == ChamNoTrans))) {
+ if ( ((side == ChamLeft ) && (trans != ChamNoTrans)) ||
+ ((side == ChamRight) && (trans == ChamNoTrans)) )
+ {
i1 = 0;
i3 = IB;
}
else {
- i1 = ((K-1) / IB)*IB;
+ i1 = ( ( K-1 ) / IB )*IB;
i3 = -IB;
}
- for(i = i1; (i > -1) && (i < K); i += i3) {
+ for (i = i1; (i > -1) && (i < K); i+=i3) {
kb = chameleon_min(IB, K-i);
if (side == ChamLeft) {
@@ -127,17 +124,18 @@ int CUDA_ztsmqr(
ni = N1 - i;
jc = i;
}
+
/*
* Apply H or H' (NOTE: CORE_zparfb used to be CORE_ztsrfb)
*/
CUDA_zparfb(
- side, trans, ChamDirForward, ChamColumnwise,
- mi, ni, M2, N2, kb, 0,
- A1 + LDA1*jc+ic, LDA1,
- A2, LDA2,
- V + LDV*i, LDV,
- T + LDT*i, LDT,
- WORK, LDWORK, WORKC, LDWORKC, CUBLAS_STREAM_VALUE );
+ side, trans, ChamDirForward, ChamColumnwise,
+ mi, ni, M2, N2, kb, 0,
+ A1 + LDA1*jc+ic, LDA1,
+ A2, LDA2,
+ V + LDV*i, LDV,
+ T + LDT*i, LDT,
+ WORK, LWORK, CUBLAS_STREAM_VALUE );
}
return CHAMELEON_SUCCESS;
}
diff --git a/cudablas/compute/cuda_zttmlq.c b/cudablas/compute/cuda_zttmlq.c
new file mode 100644
index 0000000000000000000000000000000000000000..c194adfe7f7c76847a4b64feaa9b0d5b560eb78b
--- /dev/null
+++ b/cudablas/compute/cuda_zttmlq.c
@@ -0,0 +1,140 @@
+/**
+ *
+ * @file cuda_zttmlq.c
+ *
+ * @copyright 2009-2014 The University of Tennessee and The University of
+ * Tennessee Research Foundation. All rights reserved.
+ * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
+ * Univ. Bordeaux. All rights reserved.
+ *
+ ***
+ *
+ * @brief Chameleon cuda_zttmlq GPU kernel
+ *
+ * @version 1.0.0
+ * @author Florent Pruvost
+ * @author Mathieu Faverge
+ * @date 2018-11-09
+ * @precisions normal z -> c d s
+ *
+ */
+#include "cudablas.h"
+
+int CUDA_zttmlq(
+ cham_side_t side, cham_trans_t trans,
+ int M1, int N1,
+ int M2, int N2,
+ int K, int IB,
+ cuDoubleComplex *A1, int LDA1,
+ cuDoubleComplex *A2, int LDA2,
+ const cuDoubleComplex *V, int LDV,
+ const cuDoubleComplex *T, int LDT,
+ cuDoubleComplex *WORK, int LWORK,
+ CUBLAS_STREAM_PARAM)
+{
+ int i, i1, i3;
+ int kb, l;
+ int ic = 0;
+ int jc = 0;
+ int mi1 = M1;
+ int mi2 = M2;
+ int ni1 = N1;
+ int ni2 = N2;
+
+ /* Check input arguments */
+ if ((side != ChamLeft) && (side != ChamRight)) {
+ return -1;
+ }
+
+ if ((trans != ChamNoTrans) && (trans != ChamConjTrans)) {
+ return -2;
+ }
+ if (M1 < 0) {
+ return -3;
+ }
+ if (N1 < 0) {
+ return -4;
+ }
+ if ( (M2 < 0) ||
+ ( (M2 != M1) && (side == ChamRight) ) ){
+ return -5;
+ }
+ if ( (N2 < 0) ||
+ ( (N2 != N1) && (side == ChamLeft) ) ){
+ return -6;
+ }
+ if ((K < 0) ||
+ ( (side == ChamLeft) && (K > M1) ) ||
+ ( (side == ChamRight) && (K > N1) ) ) {
+ return -7;
+ }
+ if (IB < 0) {
+ return -8;
+ }
+ if (LDA1 < chameleon_max(1,M1)){
+ return -10;
+ }
+ if (LDA2 < chameleon_max(1,M2)){
+ return -12;
+ }
+ if (LDV < chameleon_max(1,K)){
+ return -14;
+ }
+ if (LDT < chameleon_max(1,IB)){
+ return -16;
+ }
+
+ /* Quick return */
+ if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0)) {
+ return CHAMELEON_SUCCESS;
+ }
+
+ if ( ((side == ChamLeft ) && (trans == ChamNoTrans)) ||
+ ((side == ChamRight) && (trans != ChamNoTrans)) )
+ {
+ i1 = 0;
+ i3 = IB;
+ }
+ else {
+ i1 = ( ( K-1 ) / IB )*IB;
+ i3 = -IB;
+ }
+
+ /* Transpose */
+ if (trans == ChamNoTrans) {
+ trans = ChamConjTrans;
+ }
+ else {
+ trans = ChamNoTrans;
+ }
+
+ for (i = i1; (i > -1) && (i < K); i+=i3) {
+ kb = chameleon_min(IB, K-i);
+
+ if (side == ChamLeft) {
+ mi1 = kb;
+ mi2 = chameleon_min(i+kb, M2);
+ l = chameleon_min(kb, chameleon_max(0, M2-i));
+ ic = i;
+ }
+ else {
+ ni1 = kb;
+ ni2 = chameleon_min(i+kb, N2);
+ l = chameleon_min(kb, chameleon_max(0, N2-i));
+ jc = i;
+ }
+
+ /*
+ * Apply H or H' (NOTE: CORE_zparfb used to be CORE_zttrfb)
+ */
+ CUDA_zparfb(
+ side, trans, ChamDirForward, ChamRowwise,
+ mi1, ni1, mi2, ni2, kb, l,
+ A1 + LDA1 * jc + ic, LDA1,
+ A2, LDA2,
+ V + i, LDV,
+ T + LDT * i, LDT,
+ WORK, LWORK, CUBLAS_STREAM_VALUE );
+ }
+ return CHAMELEON_SUCCESS;
+}
diff --git a/cudablas/compute/cuda_zttmqr.c b/cudablas/compute/cuda_zttmqr.c
index 236405cdf080fd3c7941aa62426a70cde2e5d8f7..8664d0675afdb82b60e92bf054caa45122c7ec89 100644
--- a/cudablas/compute/cuda_zttmqr.c
+++ b/cudablas/compute/cuda_zttmqr.c
@@ -6,6 +6,7 @@
* Tennessee Research Foundation. All rights reserved.
* @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
* Univ. Bordeaux. All rights reserved.
+ *
***
*
* @brief Chameleon cuda_zttmqr GPU kernel
@@ -13,7 +14,7 @@
* @version 1.0.0
* @author Florent Pruvost
* @author Mathieu Faverge
- * @date 2015-09-16
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
@@ -28,13 +29,12 @@ int CUDA_zttmqr(
cuDoubleComplex *A2, int LDA2,
const cuDoubleComplex *V, int LDV,
const cuDoubleComplex *T, int LDT,
- cuDoubleComplex *WORK, int LDWORK,
- cuDoubleComplex *WORKC, int LDWORKC,
+ cuDoubleComplex *WORK, int LWORK,
CUBLAS_STREAM_PARAM)
{
- int i, i1, i3, l;
- int NQ, NW;
- int kb;
+ int i, i1, i3;
+ int NQ;
+ int kb, l;
int ic = 0;
int jc = 0;
int mi1 = M1;
@@ -50,11 +50,9 @@ int CUDA_zttmqr(
/* NQ is the order of Q */
if (side == ChamLeft) {
NQ = M2;
- NW = IB;
}
else {
NQ = N2;
- NW = M1;
}
if ((trans != ChamNoTrans) && (trans != ChamConjTrans)) {
@@ -94,25 +92,24 @@ int CUDA_zttmqr(
if (LDT < chameleon_max(1,IB)){
return -16;
}
- if (LDWORK < chameleon_max(1,NW)){
- return -18;
- }
/* Quick return */
- if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0))
+ if ((M1 == 0) || (N1 == 0) || (M2 == 0) || (N2 == 0) || (K == 0) || (IB == 0)) {
return CHAMELEON_SUCCESS;
+ }
- if (((side == ChamLeft) && (trans != ChamNoTrans))
- || ((side == ChamRight) && (trans == ChamNoTrans))) {
+ if ( ((side == ChamLeft ) && (trans != ChamNoTrans)) ||
+ ((side == ChamRight) && (trans == ChamNoTrans)) )
+ {
i1 = 0;
i3 = IB;
}
else {
- i1 = ((K-1) / IB)*IB;
+ i1 = ( ( K-1 ) / IB )*IB;
i3 = -IB;
}
- for(i = i1; (i > -1) && (i < K); i += i3) {
+ for (i = i1; (i > -1) && (i < K); i+=i3) {
kb = chameleon_min(IB, K-i);
if (side == ChamLeft) {
@@ -138,8 +135,7 @@ int CUDA_zttmqr(
A2, LDA2,
V + LDV*i, LDV,
T + LDT*i, LDT,
- WORK, LDWORK,
- WORKC, LDWORKC, CUBLAS_STREAM_VALUE );
+ WORK, LWORK, CUBLAS_STREAM_VALUE );
}
return CHAMELEON_SUCCESS;
}
diff --git a/cudablas/eztrace_module/cudablas_eztrace_module b/cudablas/eztrace_module/cudablas_eztrace_module
index 7ec370d3f3ad568264dc17c045c1c1f90de71eec..c631193b1fcedd7597a645dad9d208b29c1faed8 100644
--- a/cudablas/eztrace_module/cudablas_eztrace_module
+++ b/cudablas/eztrace_module/cudablas_eztrace_module
@@ -273,6 +273,30 @@ int CUDA_ctsmqr(
void *WORK, void* LDWORK,
void *WORKC, void* LDWORKC,
void* stream);
+int CUDA_cttmlq(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ void *A1, void* LDA1,
+ void *A2, void* LDA2,
+ void *V, void* LDV,
+ void *T, void* LDT,
+ void *WORK, void* LDWORK,
+ void *WORKC, void* LDWORKC,
+ void* stream);
+int CUDA_cttmqr(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ void *A1, void* LDA1,
+ void *A2, void* LDA2,
+ void *V, void* LDV,
+ void *T, void* LDT,
+ void *WORK, void* LDWORK,
+ void *WORKC, void* LDWORKC,
+ void* stream);
int CUDA_ctsqrt(
void* m, void* n, void* nb,
void *da1, void* ldda1,
@@ -528,6 +552,30 @@ int CUDA_dtsmqr(
double *WORK, void* LDWORK,
double *WORKC, void* LDWORKC,
void* stream);
+int CUDA_dttmlq(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ double *A1, void* LDA1,
+ double *A2, void* LDA2,
+ const double *V, void* LDV,
+ const double *T, void* LDT,
+ double *WORK, void* LDWORK,
+ double *WORKC, void* LDWORKC,
+ void* stream);
+int CUDA_dttmqr(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ double *A1, void* LDA1,
+ double *A2, void* LDA2,
+ const double *V, void* LDV,
+ const double *T, void* LDT,
+ double *WORK, void* LDWORK,
+ double *WORKC, void* LDWORKC,
+ void* stream);
int CUDA_dtsqrt(
void* m, void* n, void* nb,
double *da1, void* ldda1,
@@ -783,6 +831,30 @@ int CUDA_stsmqr(
float *WORK, void* LDWORK,
float *WORKC, void* LDWORKC,
void* stream);
+int CUDA_sttmlq(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ float *A1, void* LDA1,
+ float *A2, void* LDA2,
+ const float *V, void* LDV,
+ const float *T, void* LDT,
+ float *WORK, void* LDWORK,
+ float *WORKC, void* LDWORKC,
+ void* stream);
+int CUDA_sttmqr(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ float *A1, void* LDA1,
+ float *A2, void* LDA2,
+ const float *V, void* LDV,
+ const float *T, void* LDT,
+ float *WORK, void* LDWORK,
+ float *WORKC, void* LDWORKC,
+ void* stream);
int CUDA_stsqrt(
void* m, void* n, void* nb,
float *da1, void* ldda1,
@@ -1090,6 +1162,30 @@ int CUDA_ztsmqr(
void *WORK, void* LDWORK,
void *WORKC, void* LDWORKC,
void* stream);
+int CUDA_zttmlq(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ void *A1, void* LDA1,
+ void *A2, void* LDA2,
+ void *V, void* LDV,
+ void *T, void* LDT,
+ void *WORK, void* LDWORK,
+ void *WORKC, void* LDWORKC,
+ void* stream);
+int CUDA_zttmqr(
+ void* side, void* trans,
+ void* M1, void* N1,
+ void* M2, void* N2,
+ void* K, void* IB,
+ void *A1, void* LDA1,
+ void *A2, void* LDA2,
+ void *V, void* LDV,
+ void *T, void* LDT,
+ void *WORK, void* LDWORK,
+ void *WORKC, void* LDWORKC,
+ void* stream);
int CUDA_ztsqrt(
void* m, void* n, void* nb,
void *da1, void* ldda1,
diff --git a/cudablas/include/cudablas/cudablas_z.h b/cudablas/include/cudablas/cudablas_z.h
index 8e96d463ca451c74b17f91c22c2b6c03a27a101e..8895ff6485b33c423bd80b5b99599883eec58ce7 100644
--- a/cudablas/include/cudablas/cudablas_z.h
+++ b/cudablas/include/cudablas/cudablas_z.h
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Florent Pruvost
- * @date 2015-09-16
+ * @date 2018-11-09
* @precisions normal z -> c d s
*
*/
@@ -31,16 +31,18 @@ int CUDA_zher2k( cham_uplo_t uplo, cham_trans_t trans, int n, int k, cuDoubleCom
int CUDA_zherfb( cham_uplo_t uplo, int n, int k, int ib, int nb, const cuDoubleComplex *A, int lda, const cuDoubleComplex *T, int ldt, cuDoubleComplex *C, int ldc, cuDoubleComplex *WORK, int ldwork, CUBLAS_STREAM_PARAM );
int CUDA_zherk( cham_uplo_t uplo, cham_trans_t trans, int n, int k, double *alpha, const cuDoubleComplex *A, int lda, double *beta, cuDoubleComplex *B, int ldb, CUBLAS_STREAM_PARAM );
int CUDA_zlarfb(cham_side_t side, cham_trans_t trans, cham_dir_t direct, cham_store_t storev, int M, int N, int K, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *C, int LDC, cuDoubleComplex *WORK, int LDWORK, CUBLAS_STREAM_PARAM );
-int CUDA_zparfb(cham_side_t side, cham_trans_t trans, cham_dir_t direct, cham_store_t storev, int M1, int N1, int M2, int N2, int K, int L, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LDWORK, cuDoubleComplex *WORKC, int LDWORKC, CUBLAS_STREAM_PARAM );
+int CUDA_zparfb(cham_side_t side, cham_trans_t trans, cham_dir_t direct, cham_store_t storev, int M1, int N1, int M2, int N2, int K, int L, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LWORK, CUBLAS_STREAM_PARAM );
int CUDA_zsymm( cham_side_t side, cham_uplo_t uplo, int m, int n, cuDoubleComplex *alpha, const cuDoubleComplex *A, int lda, const cuDoubleComplex *B, int ldb, cuDoubleComplex *beta, cuDoubleComplex *C, int ldc, CUBLAS_STREAM_PARAM );
int CUDA_zsyr2k( cham_uplo_t uplo, cham_trans_t trans, int n, int k, cuDoubleComplex *alpha, const cuDoubleComplex *A, int lda, const cuDoubleComplex *B, int ldb, cuDoubleComplex *beta, cuDoubleComplex *C, int ldc, CUBLAS_STREAM_PARAM );
int CUDA_zsyrk( cham_uplo_t uplo, cham_trans_t trans, int n, int k, cuDoubleComplex *alpha, const cuDoubleComplex *A, int lda, cuDoubleComplex *beta, cuDoubleComplex *C, int ldc, CUBLAS_STREAM_PARAM );
-int CUDA_ztpmqrt( cham_side_t side, cham_trans_t trans, int M, int N, int K, int L, int IB, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *A, int LDA, cuDoubleComplex *B, int LDB, cuDoubleComplex *WORK, CUBLAS_STREAM_PARAM );
+int CUDA_ztpmqrt( cham_side_t side, cham_trans_t trans, int M, int N, int K, int L, int IB, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *A, int LDA, cuDoubleComplex *B, int LDB, cuDoubleComplex *WORK, int lwork, CUBLAS_STREAM_PARAM );
+int CUDA_ztpmlqt( cham_side_t side, cham_trans_t trans, int M, int N, int K, int L, int IB, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *A, int LDA, cuDoubleComplex *B, int LDB, cuDoubleComplex *WORK, int lwork, CUBLAS_STREAM_PARAM );
int CUDA_ztrmm( cham_side_t side, cham_uplo_t uplo, cham_trans_t transa, cham_diag_t diag, int m, int n, cuDoubleComplex *alpha, const cuDoubleComplex *A, int lda, cuDoubleComplex *B, int ldb, CUBLAS_STREAM_PARAM );
int CUDA_ztrsm( cham_side_t side, cham_uplo_t uplo, cham_trans_t transa, cham_diag_t diag, int m, int n, cuDoubleComplex *alpha, const cuDoubleComplex *A, int lda, cuDoubleComplex *B, int ldb, CUBLAS_STREAM_PARAM );
-int CUDA_ztsmlq( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LDWORK, cuDoubleComplex *WORKC, int LDWORKC, CUBLAS_STREAM_PARAM );
-int CUDA_ztsmqr( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LDWORK, cuDoubleComplex *WORKC, int LDWORKC, CUBLAS_STREAM_PARAM );
-int CUDA_zttmqr( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LDWORK, cuDoubleComplex *WORKC, int LDWORKC, CUBLAS_STREAM_PARAM );
+int CUDA_ztsmlq( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LWORK, CUBLAS_STREAM_PARAM );
+int CUDA_zttmlq( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LWORK, CUBLAS_STREAM_PARAM );
+int CUDA_ztsmqr( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LWORK, CUBLAS_STREAM_PARAM );
+int CUDA_zttmqr( cham_side_t side, cham_trans_t trans, int M1, int N1, int M2, int N2, int K, int IB, cuDoubleComplex *A1, int LDA1, cuDoubleComplex *A2, int LDA2, const cuDoubleComplex *V, int LDV, const cuDoubleComplex *T, int LDT, cuDoubleComplex *WORK, int LWORK, CUBLAS_STREAM_PARAM );
int CUDA_zunmlqt(cham_side_t side, cham_trans_t trans, int M, int N, int K, int IB, const cuDoubleComplex *A, int LDA, const cuDoubleComplex *T, int LDT, cuDoubleComplex *C, int LDC, cuDoubleComplex *WORK, int LDWORK, CUBLAS_STREAM_PARAM );
int CUDA_zunmqrt(cham_side_t side, cham_trans_t trans, int M, int N, int K, int IB, const cuDoubleComplex *A, int LDA, const cuDoubleComplex *T, int LDT, cuDoubleComplex *C, int LDC, cuDoubleComplex *WORK, int LDWORK, CUBLAS_STREAM_PARAM );
diff --git a/include/chameleon/tasks.h b/include/chameleon/tasks.h
index bb1794316b74504f8629d2fb3e9a40ac50d18879..b8f91fb211f98a417a9c722794df6c45c441dcdc 100644
--- a/include/chameleon/tasks.h
+++ b/include/chameleon/tasks.h
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Cedric Augonnet
- * @date 2011-06-01
+ * @date 2018-11-08
*
*/
#ifndef _chameleon_tasks_h_
@@ -54,16 +54,12 @@ typedef enum chameleon_tasktype_e {
TASK_ORMQR,
TASK_POTRF,
TASK_SSSSM,
+ TASK_TPLQT,
+ TASK_TPMLQT,
+ TASK_TPMQRT,
+ TASK_TPQRT,
TASK_TRTRI,
- TASK_TSLQT,
- TASK_TSMLQ,
- TASK_TSMQR,
- TASK_TSQRT,
TASK_TSTRF,
- TASK_TTLQT,
- TASK_TTMLQ,
- TASK_TTMQR,
- TASK_TTQRT,
TASK_UNMLQ,
TASK_UNMQR,
@@ -86,6 +82,15 @@ typedef enum chameleon_tasktype_e {
TASK_NBKERNELS
} cham_tasktype_t;
+#define TASK_TSLQT TASK_TPLQT
+#define TASK_TSMLQ TASK_TPMLQT
+#define TASK_TSMQR TASK_TPMQRT
+#define TASK_TSQRT TASK_TPQRT
+#define TASK_TTLQT TASK_TPLQT
+#define TASK_TTMLQ TASK_TPMLQT
+#define TASK_TTMQR TASK_TPMQRT
+#define TASK_TTQRT TASK_TPQRT
+
typedef int (*cham_unary_operator_t)( const CHAM_desc_t *desc,
cham_uplo_t uplo, int m, int n,
void *data, void *op_args );
diff --git a/include/chameleon/tasks_z.h b/include/chameleon/tasks_z.h
index 28aeaa1b889e0eb5fe16618b02d9bebd73de905a..8265e990686f22639acc23d0919348d017947cb2 100644
--- a/include/chameleon/tasks_z.h
+++ b/include/chameleon/tasks_z.h
@@ -20,7 +20,7 @@
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-07
* @precisions normal z -> c d s
*
*/
@@ -31,435 +31,499 @@
* Declarations of QUARK wrappers (called by CHAMELEON) - alphabetical order
*/
void INSERT_TASK_dzasum( const RUNTIME_option_t *options,
- cham_store_t storev, cham_uplo_t uplo, int M, int N,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn );
+ cham_store_t storev, cham_uplo_t uplo, int M, int N,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_zgeadd( const RUNTIME_option_t *options,
- cham_trans_t trans, int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_trans_t trans, int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_zlascal( const RUNTIME_option_t *options,
- cham_uplo_t uplo,
- int m, int n, int nb,
- CHAMELEON_Complex64_t alpha,
- const CHAM_desc_t *A, int Am, int An, int lda );
+ cham_uplo_t uplo,
+ int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha,
+ const CHAM_desc_t *A, int Am, int An, int lda );
void INSERT_TASK_zbrdalg( const RUNTIME_option_t *options,
- cham_uplo_t uplo,
- int N, int NB,
- const CHAM_desc_t *A,
- const CHAM_desc_t *C, int Cm, int Cn,
- const CHAM_desc_t *S, int Sm, int Sn,
- int i, int j, int m, int grsiz, int BAND,
- int *PCOL, int *ACOL, int *MCOL );
+ cham_uplo_t uplo,
+ int N, int NB,
+ const CHAM_desc_t *A,
+ const CHAM_desc_t *C, int Cm, int Cn,
+ const CHAM_desc_t *S, int Sm, int Sn,
+ int i, int j, int m, int grsiz, int BAND,
+ int *PCOL, int *ACOL, int *MCOL );
void INSERT_TASK_zgelqt( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt );
void INSERT_TASK_zgemm( const RUNTIME_option_t *options,
- cham_trans_t transA, cham_trans_t transB,
- int m, int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
-void INSERT_TASK_zgemm2( const RUNTIME_option_t *options,
cham_trans_t transA, cham_trans_t transB,
int m, int n, int k, int nb,
CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
const CHAM_desc_t *B, int Bm, int Bn, int ldb,
CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+void INSERT_TASK_zgemm2( const RUNTIME_option_t *options,
+ cham_trans_t transA, cham_trans_t transB,
+ int m, int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zgemm_f2( const RUNTIME_option_t *options,
- cham_trans_t transA, cham_trans_t transB,
- int m, int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc,
- const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1,
- const CHAM_desc_t *fake2, int fake2m, int fake2n, int szefake2, int flag2 );
+ cham_trans_t transA, cham_trans_t transB,
+ int m, int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc,
+ const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1,
+ const CHAM_desc_t *fake2, int fake2m, int fake2n, int szefake2, int flag2 );
void INSERT_TASK_zgemm_p2( const RUNTIME_option_t *options,
- cham_trans_t transA, cham_trans_t transB,
- int m, int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAMELEON_Complex64_t **B, int ldb,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+ cham_trans_t transA, cham_trans_t transB,
+ int m, int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAMELEON_Complex64_t **B, int ldb,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zgemm_p2f1( const RUNTIME_option_t *options,
- cham_trans_t transA, cham_trans_t transB,
- int m, int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAMELEON_Complex64_t **B, int ldb,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc,
- const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1 );
+ cham_trans_t transA, cham_trans_t transB,
+ int m, int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAMELEON_Complex64_t **B, int ldb,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc,
+ const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1 );
void INSERT_TASK_zgemm_p3( const RUNTIME_option_t *options,
- cham_trans_t transA, cham_trans_t transB,
- int m, int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- CHAMELEON_Complex64_t beta, CHAMELEON_Complex64_t **C, int ldc );
+ cham_trans_t transA, cham_trans_t transB,
+ int m, int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ CHAMELEON_Complex64_t beta, CHAMELEON_Complex64_t **C, int ldc );
void INSERT_TASK_zgeqrt( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt );
void INSERT_TASK_zgessm( const RUNTIME_option_t *options,
- int m, int n, int k, int ib, int nb,
- int *IPIV,
- const CHAM_desc_t *L, int Lm, int Ln, int ldl,
- const CHAM_desc_t *D, int Dm, int Dn, int ldd,
- const CHAM_desc_t *A, int Am, int An, int lda );
+ int m, int n, int k, int ib, int nb,
+ int *IPIV,
+ const CHAM_desc_t *L, int Lm, int Ln, int ldl,
+ const CHAM_desc_t *D, int Dm, int Dn, int ldd,
+ const CHAM_desc_t *A, int Am, int An, int lda );
void INSERT_TASK_zgessq( const RUNTIME_option_t *options,
- int m, int n,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
+ int m, int n,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
void INSERT_TASK_zgetrf( const RUNTIME_option_t *options,
- int m, int n, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- int *IPIV,
- cham_bool_t check_info, int iinfo );
+ int m, int n, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ int *IPIV,
+ cham_bool_t check_info, int iinfo );
void INSERT_TASK_zgetrf_incpiv( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *L, int Lm, int Ln, int ldl,
- int *IPIV,
- cham_bool_t check_info, int iinfo );
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *L, int Lm, int Ln, int ldl,
+ int *IPIV,
+ cham_bool_t check_info, int iinfo );
void INSERT_TASK_zgetrf_nopiv( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda, int iinfo );
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda, int iinfo );
void INSERT_TASK_zgetrf_reclap( const RUNTIME_option_t *options,
- int m, int n, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- int *IPIV,
+ int m, int n, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ int *IPIV,
- cham_bool_t check_info, int iinfo,
- int nbthread );
+ cham_bool_t check_info, int iinfo,
+ int nbthread );
void INSERT_TASK_zgetrf_rectil( const RUNTIME_option_t *options,
- const CHAM_desc_t A, const CHAM_desc_t *Amn, int Amnm, int Amnn, int size,
- int *IPIV,
+ const CHAM_desc_t A, const CHAM_desc_t *Amn, int Amnm, int Amnn, int size,
+ int *IPIV,
- cham_bool_t check_info, int iinfo,
- int nbthread );
+ cham_bool_t check_info, int iinfo,
+ int nbthread );
void INSERT_TASK_zgetrip( const RUNTIME_option_t *options,
- int m, int n, const CHAM_desc_t *A, int Am, int An, int szeA );
+ int m, int n, const CHAM_desc_t *A, int Am, int An, int szeA );
void INSERT_TASK_zgetrip_f1( const RUNTIME_option_t *options,
- int m, int n, const CHAM_desc_t *A, int Am, int An, int szeA,
- const CHAM_desc_t *fake, int fakem, int faken, int szeF, int paramF );
+ int m, int n, const CHAM_desc_t *A, int Am, int An, int szeA,
+ const CHAM_desc_t *fake, int fakem, int faken, int szeF, int paramF );
void INSERT_TASK_zgetrip_f2( const RUNTIME_option_t *options,
- int m, int n, const CHAM_desc_t *A, int Am, int An, int szeA,
- const CHAM_desc_t *fake1, int fake1m, int fake1n, int szeF1, int paramF1,
- const CHAM_desc_t *fake2, int fake2m, int fake2n, int szeF2, int paramF2 );
+ int m, int n, const CHAM_desc_t *A, int Am, int An, int szeA,
+ const CHAM_desc_t *fake1, int fake1m, int fake1n, int szeF1, int paramF1,
+ const CHAM_desc_t *fake2, int fake2m, int fake2n, int szeF2, int paramF2 );
void INSERT_TASK_zhe2ge( const RUNTIME_option_t *options,
- cham_uplo_t uplo,
- int m, int n, int mb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_uplo_t uplo,
+ int m, int n, int mb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_zhemm( const RUNTIME_option_t *options,
- cham_side_t side, cham_uplo_t uplo,
- int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+ cham_side_t side, cham_uplo_t uplo,
+ int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zhegst( const RUNTIME_option_t *options,
- int itype, cham_uplo_t uplo, int N,
- const CHAM_desc_t *A, int Am, int An, int LDA,
- const CHAM_desc_t *B, int Bm, int Bn, int LDB,
- int iinfo );
+ int itype, cham_uplo_t uplo, int N,
+ const CHAM_desc_t *A, int Am, int An, int LDA,
+ const CHAM_desc_t *B, int Bm, int Bn, int LDB,
+ int iinfo );
void INSERT_TASK_zherk( const RUNTIME_option_t *options,
- cham_uplo_t uplo, cham_trans_t trans,
- int n, int k, int nb,
- double alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- double beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
-void INSERT_TASK_zher2k( const RUNTIME_option_t *options,
cham_uplo_t uplo, cham_trans_t trans,
int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int LDB,
+ double alpha, const CHAM_desc_t *A, int Am, int An, int lda,
double beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+void INSERT_TASK_zher2k( const RUNTIME_option_t *options,
+ cham_uplo_t uplo, cham_trans_t trans,
+ int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int LDB,
+ double beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zherfb( const RUNTIME_option_t *options,
- cham_uplo_t uplo,
- int n, int k, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt,
- const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+ cham_uplo_t uplo,
+ int n, int k, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zlacpy( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int m, int n, int mb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
-void INSERT_TASK_zlacpyx( const RUNTIME_option_t *options,
cham_uplo_t uplo, int m, int n, int mb,
- int displA, const CHAM_desc_t *A, int Am, int An, int lda,
- int displB, const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+void INSERT_TASK_zlacpyx( const RUNTIME_option_t *options,
+ cham_uplo_t uplo, int m, int n, int mb,
+ int displA, const CHAM_desc_t *A, int Am, int An, int lda,
+ int displB, const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_zlange( const RUNTIME_option_t *options,
- cham_normtype_t norm, int M, int N, int NB,
- const CHAM_desc_t *A, int Am, int An, int LDA,
- const CHAM_desc_t *B, int Bm, int Bn );
+ cham_normtype_t norm, int M, int N, int NB,
+ const CHAM_desc_t *A, int Am, int An, int LDA,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_zlange_max( const RUNTIME_option_t *options,
- const CHAM_desc_t *A, int Am, int An,
- const CHAM_desc_t *B, int Bm, int Bn );
+ const CHAM_desc_t *A, int Am, int An,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_zhessq( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int n,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
+ cham_uplo_t uplo, int n,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
void INSERT_TASK_zlanhe( const RUNTIME_option_t *options,
- cham_normtype_t norm, cham_uplo_t uplo, int N, int NB,
- const CHAM_desc_t *A, int Am, int An, int LDA,
- const CHAM_desc_t *B, int Bm, int Bn );
+ cham_normtype_t norm, cham_uplo_t uplo, int N, int NB,
+ const CHAM_desc_t *A, int Am, int An, int LDA,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_zlansy( const RUNTIME_option_t *options,
- cham_normtype_t norm, cham_uplo_t uplo, int N, int NB,
- const CHAM_desc_t *A, int Am, int An, int LDA,
- const CHAM_desc_t *B, int Bm, int Bn );
+ cham_normtype_t norm, cham_uplo_t uplo, int N, int NB,
+ const CHAM_desc_t *A, int Am, int An, int LDA,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_zlantr( const RUNTIME_option_t *options,
- cham_normtype_t norm, cham_uplo_t uplo, cham_diag_t diag,
- int M, int N, int NB,
- const CHAM_desc_t *A, int Am, int An, int LDA,
- const CHAM_desc_t *B, int Bm, int Bn );
+ cham_normtype_t norm, cham_uplo_t uplo, cham_diag_t diag,
+ int M, int N, int NB,
+ const CHAM_desc_t *A, int Am, int An, int LDA,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_zlaset( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int n1, int n2, CHAMELEON_Complex64_t alpha,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *tileA, int tileAm, int tileAn, int ldtilea );
-void INSERT_TASK_zlaset2( const RUNTIME_option_t *options,
cham_uplo_t uplo, int n1, int n2, CHAMELEON_Complex64_t alpha,
- const CHAM_desc_t *tileA, int tileAm, int tileAn, int ldtilea );
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *tileA, int tileAm, int tileAn, int ldtilea );
+void INSERT_TASK_zlaset2( const RUNTIME_option_t *options,
+ cham_uplo_t uplo, int n1, int n2, CHAMELEON_Complex64_t alpha,
+ const CHAM_desc_t *tileA, int tileAm, int tileAn, int ldtilea );
void INSERT_TASK_zlaswp( const RUNTIME_option_t *options,
- int n, const CHAM_desc_t *A, int Am, int An, int lda,
- int i1, int i2, int *ipiv, int inc );
+ int n, const CHAM_desc_t *A, int Am, int An, int lda,
+ int i1, int i2, int *ipiv, int inc );
void INSERT_TASK_zlaswp_f2( const RUNTIME_option_t *options,
- int n, const CHAM_desc_t *A, int Am, int An, int lda,
- int i1, int i2, int *ipiv, int inc,
- const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1,
- const CHAM_desc_t *fake2, int fake2m, int fake2n, int szefake2, int flag2 );
+ int n, const CHAM_desc_t *A, int Am, int An, int lda,
+ int i1, int i2, int *ipiv, int inc,
+ const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1,
+ const CHAM_desc_t *fake2, int fake2m, int fake2n, int szefake2, int flag2 );
void INSERT_TASK_zlaswp_ontile( const RUNTIME_option_t *options,
- const CHAM_desc_t descA, const CHAM_desc_t *A, int Am, int An,
- int i1, int i2, int *ipiv, int inc, CHAMELEON_Complex64_t *fakepanel );
-void INSERT_TASK_zlaswp_ontile_f2( const RUNTIME_option_t *options,
- const CHAM_desc_t descA, const CHAM_desc_t *A, int Am, int An,
- int i1, int i2, int *ipiv, int inc,
- const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1,
- const CHAM_desc_t *fake2, int fake2m, int fake2n, int szefake2, int flag2 );
-void INSERT_TASK_zlaswpc_ontile( const RUNTIME_option_t *options,
const CHAM_desc_t descA, const CHAM_desc_t *A, int Am, int An,
int i1, int i2, int *ipiv, int inc, CHAMELEON_Complex64_t *fakepanel );
+void INSERT_TASK_zlaswp_ontile_f2( const RUNTIME_option_t *options,
+ const CHAM_desc_t descA, const CHAM_desc_t *A, int Am, int An,
+ int i1, int i2, int *ipiv, int inc,
+ const CHAM_desc_t *fake1, int fake1m, int fake1n, int szefake1, int flag1,
+ const CHAM_desc_t *fake2, int fake2m, int fake2n, int szefake2, int flag2 );
+void INSERT_TASK_zlaswpc_ontile( const RUNTIME_option_t *options,
+ const CHAM_desc_t descA, const CHAM_desc_t *A, int Am, int An,
+ int i1, int i2, int *ipiv, int inc, CHAMELEON_Complex64_t *fakepanel );
void INSERT_TASK_zlatro( const RUNTIME_option_t *options,
- cham_uplo_t uplo, cham_trans_t trans, int m, int n, int mb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_uplo_t uplo, cham_trans_t trans, int m, int n, int mb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_zlauum( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int n, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda );
+ cham_uplo_t uplo, int n, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda );
void INSERT_TASK_zplghe( const RUNTIME_option_t *options,
- double bump, int m, int n, const CHAM_desc_t *A, int Am, int An, int lda,
- int bigM, int m0, int n0, unsigned long long int seed );
+ double bump, int m, int n, const CHAM_desc_t *A, int Am, int An, int lda,
+ int bigM, int m0, int n0, unsigned long long int seed );
void INSERT_TASK_zplgsy( const RUNTIME_option_t *options,
- CHAMELEON_Complex64_t bump, int m, int n, const CHAM_desc_t *A, int Am, int An, int lda,
- int bigM, int m0, int n0, unsigned long long int seed );
+ CHAMELEON_Complex64_t bump, int m, int n, const CHAM_desc_t *A, int Am, int An, int lda,
+ int bigM, int m0, int n0, unsigned long long int seed );
void INSERT_TASK_zplrnt( const RUNTIME_option_t *options,
- int m, int n, const CHAM_desc_t *A, int Am, int An, int lda,
- int bigM, int m0, int n0, unsigned long long int seed );
+ int m, int n, const CHAM_desc_t *A, int Am, int An, int lda,
+ int bigM, int m0, int n0, unsigned long long int seed );
void INSERT_TASK_zpotrf( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int n, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
+ cham_uplo_t uplo, int n, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
- int iinfo );
+ int iinfo );
void INSERT_TASK_zshift( const RUNTIME_option_t *options,
- int s, int m, int n, int L,
- CHAMELEON_Complex64_t *A );
+ int s, int m, int n, int L,
+ CHAMELEON_Complex64_t *A );
void INSERT_TASK_zshiftw( const RUNTIME_option_t *options,
- int s, int cl, int m, int n, int L,
- const CHAM_desc_t *A, int Am, int An, CHAMELEON_Complex64_t *W );
+ int s, int cl, int m, int n, int L,
+ const CHAM_desc_t *A, int Am, int An, CHAMELEON_Complex64_t *W );
void INSERT_TASK_zssssm( const RUNTIME_option_t *options,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *L1, int L1m, int L1n, int ldl1,
- const CHAM_desc_t *L2, int L2m, int L2n, int ldl2,
- const int *IPIV );
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *L1, int L1m, int L1n, int ldl1,
+ const CHAM_desc_t *L2, int L2m, int L2n, int ldl2,
+ const int *IPIV );
void INSERT_TASK_zsymm( const RUNTIME_option_t *options,
- cham_side_t side, cham_uplo_t uplo,
- int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+ cham_side_t side, cham_uplo_t uplo,
+ int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zsyrk( const RUNTIME_option_t *options,
- cham_uplo_t uplo, cham_trans_t trans,
- int n, int k, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
-void INSERT_TASK_zsyr2k( const RUNTIME_option_t *options,
cham_uplo_t uplo, cham_trans_t trans,
int n, int k, int nb,
CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int LDB,
CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+void INSERT_TASK_zsyr2k( const RUNTIME_option_t *options,
+ cham_uplo_t uplo, cham_trans_t trans,
+ int n, int k, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int LDB,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zsyssq( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int n,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
+ cham_uplo_t uplo, int n,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
void INSERT_TASK_zsytrf_nopiv( const RUNTIME_option_t *options,
- cham_uplo_t uplo, int n, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- int iinfo );
+ cham_uplo_t uplo, int n, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ int iinfo );
void INSERT_TASK_zswpab( const RUNTIME_option_t *options,
- int i, int n1, int n2,
- const CHAM_desc_t *A, int Am, int An, int szeA );
+ int i, int n1, int n2,
+ const CHAM_desc_t *A, int Am, int An, int szeA );
void INSERT_TASK_zswptr_ontile( const RUNTIME_option_t *options,
- const CHAM_desc_t descA, const CHAM_desc_t *Aij, int Aijm, int Aijn,
- int i1, int i2, int *ipiv, int inc,
- const CHAM_desc_t *Akk, int Akkm, int Akkn, int ldak );
+ const CHAM_desc_t descA, const CHAM_desc_t *Aij, int Aijm, int Aijn,
+ int i1, int i2, int *ipiv, int inc,
+ const CHAM_desc_t *Akk, int Akkm, int Akkn, int ldak );
void INSERT_TASK_ztplqt( const RUNTIME_option_t *options,
- int m, int n, int l, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_ztpmlqt( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int M, int N, int K, int L, int ib, int nb,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ int m, int n, int l, int ib, int nb,
const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+void INSERT_TASK_ztpmlqt( const RUNTIME_option_t *options,
+ cham_side_t side, cham_trans_t trans,
+ int M, int N, int K, int L, int ib, int nb,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_ztpmqrt( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m, int n, int k, int l, int ib, int nb,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_side_t side, cham_trans_t trans,
+ int m, int n, int k, int l, int ib, int nb,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_ztpqrt( const RUNTIME_option_t *options,
- int m, int n, int l, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+ int m, int n, int l, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt );
void INSERT_TASK_ztrdalg( const RUNTIME_option_t *options,
- cham_uplo_t uplo,
- int N, int NB,
- const CHAM_desc_t *A,
- const CHAM_desc_t *C, int Cm, int Cn,
- const CHAM_desc_t *S, int Sm, int Sn,
- int i, int j, int m, int grsiz, int BAND,
- int *PCOL, int *ACOL, int *MCOL );
+ cham_uplo_t uplo,
+ int N, int NB,
+ const CHAM_desc_t *A,
+ const CHAM_desc_t *C, int Cm, int Cn,
+ const CHAM_desc_t *S, int Sm, int Sn,
+ int i, int j, int m, int grsiz, int BAND,
+ int *PCOL, int *ACOL, int *MCOL );
void INSERT_TASK_ztradd( const RUNTIME_option_t *options,
- cham_uplo_t uplo, cham_trans_t trans, int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- CHAMELEON_Complex64_t beta, const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_uplo_t uplo, cham_trans_t trans, int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ CHAMELEON_Complex64_t beta, const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_ztrasm( const RUNTIME_option_t *options,
- cham_store_t storev, cham_uplo_t uplo, cham_diag_t diag, int M, int N,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn );
+ cham_store_t storev, cham_uplo_t uplo, cham_diag_t diag, int M, int N,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn );
void INSERT_TASK_ztrmm( const RUNTIME_option_t *options,
- cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag,
- int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag,
+ int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_ztrmm_p2( const RUNTIME_option_t *options,
- cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag,
- int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- CHAMELEON_Complex64_t **B, int ldb );
+ cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag,
+ int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ CHAMELEON_Complex64_t **B, int ldb );
void INSERT_TASK_ztrsm( const RUNTIME_option_t *options,
- cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag,
- int m, int n, int nb,
- CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb );
+ cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag,
+ int m, int n, int nb,
+ CHAMELEON_Complex64_t alpha, const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb );
void INSERT_TASK_ztrssq( const RUNTIME_option_t *options,
- cham_uplo_t uplo, cham_diag_t diag,
- int m, int n,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
+ cham_uplo_t uplo, cham_diag_t diag,
+ int m, int n,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn );
void INSERT_TASK_ztrtri( const RUNTIME_option_t *options,
- cham_uplo_t uplo, cham_diag_t diag, int n, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
+ cham_uplo_t uplo, cham_diag_t diag, int n, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
- int iinfo );
-void INSERT_TASK_ztslqt( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_ztsmlq( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+ int iinfo );
void INSERT_TASK_ztsmlq_hetra1( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_ztsmqr( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+ cham_side_t side, cham_trans_t trans,
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt );
void INSERT_TASK_ztsmqr_hetra1( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_ztsqrt( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
+ cham_side_t side, cham_trans_t trans,
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt );
void INSERT_TASK_ztstrf( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *U, int Um, int Un, int ldu,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *L, int Lm, int Ln, int ldl,
- int *IPIV,
- cham_bool_t check_info, int iinfo );
-void INSERT_TASK_zttmqr( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_zttqrt( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_zttmlq( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *U, int Um, int Un, int ldu,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *L, int Lm, int Ln, int ldl,
+ int *IPIV,
+ cham_bool_t check_info, int iinfo );
+void INSERT_TASK_zpamm( const RUNTIME_option_t *options,
+ int op, cham_side_t side, cham_store_t storev,
+ int m, int n, int k, int l,
const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_zttlqt( const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt );
-void INSERT_TASK_zpamm( const RUNTIME_option_t *options,
- int op, cham_side_t side, cham_store_t storev,
- int m, int n, int k, int l,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *W, int Wm, int Wn, int ldw );
+ const CHAM_desc_t *W, int Wm, int Wn, int ldw );
void INSERT_TASK_zplssq( const RUNTIME_option_t *options,
- const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn,
- const CHAM_desc_t *SCLSSQ, int SCLSSQm, int SCLSSQn );
+ const CHAM_desc_t *SCALESUMSQ, int SCALESUMSQm, int SCALESUMSQn,
+ const CHAM_desc_t *SCLSSQ, int SCLSSQm, int SCLSSQn );
void INSERT_TASK_zplssq2( const RUNTIME_option_t *options,
- const CHAM_desc_t *RESULT, int RESULTm, int RESULTn );
+ const CHAM_desc_t *RESULT, int RESULTm, int RESULTn );
void INSERT_TASK_zunmlq( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m, int n, int ib, int nb, int k,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt,
- const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+ cham_side_t side, cham_trans_t trans,
+ int m, int n, int ib, int nb, int k,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zunmqr( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m, int n, int k, int ib, int nb,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt,
- const CHAM_desc_t *C, int Cm, int Cn, int ldc );
+ cham_side_t side, cham_trans_t trans,
+ int m, int n, int k, int ib, int nb,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ const CHAM_desc_t *C, int Cm, int Cn, int ldc );
void INSERT_TASK_zbuild( const RUNTIME_option_t *options,
- const CHAM_desc_t *A, int Am, int An, int lda,
- void *user_data, void* user_build_callback );
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ void *user_data, void* user_build_callback );
+
+
+/**
+ * Keep these insert_task for retro-compatibility
+ */
+static inline void
+INSERT_TASK_ztslqt( const RUNTIME_option_t *options,
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztplqt( options, m, n, 0, ib, nb,
+ A1, A1m, A1n, lda1,
+ A2, A2m, A2n, lda2,
+ T, Tm, Tn, ldt );
+}
+
+static inline void
+INSERT_TASK_ztsqrt( const RUNTIME_option_t *options,
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztpqrt( options, m, n, 0, ib, nb,
+ A1, A1m, A1n, lda1,
+ A2, A2m, A2n, lda2,
+ T, Tm, Tn, ldt );
+}
+
+static inline void
+INSERT_TASK_zttlqt( const RUNTIME_option_t *options,
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztplqt( options, m, n, n, ib, nb,
+ A1, A1m, A1n, lda1,
+ A2, A2m, A2n, lda2,
+ T, Tm, Tn, ldt );
+}
+
+static inline void
+INSERT_TASK_zttqrt( const RUNTIME_option_t *options,
+ int m, int n, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztpqrt( options, m, n, m, ib, nb,
+ A1, A1m, A1n, lda1,
+ A2, A2m, A2n, lda2,
+ T, Tm, Tn, ldt );
+}
+
+static inline void
+INSERT_TASK_ztsmlq( const RUNTIME_option_t *options,
+ cham_side_t side, cham_trans_t trans,
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztpmlqt( options, side, trans, m2, n2, k, 0, ib, nb,
+ V, Vm, Vn, ldv, T, Tm, Tn, ldt,
+ A1, A1m, A1n, lda1, A2, A2m, A2n, lda2 );
+}
+
+static inline void
+INSERT_TASK_ztsmqr( const RUNTIME_option_t *options,
+ cham_side_t side, cham_trans_t trans,
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztpmqrt( options, side, trans, m2, n2, k, 0, ib, nb,
+ V, Vm, Vn, ldv, T, Tm, Tn, ldt,
+ A1, A1m, A1n, lda1, A2, A2m, A2n, lda2 );
+}
+
+static inline void
+INSERT_TASK_zttmlq( const RUNTIME_option_t *options,
+ cham_side_t side, cham_trans_t trans,
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztpmlqt( options, side, trans, m2, n2, k, n2, ib, nb,
+ V, Vm, Vn, ldv, T, Tm, Tn, ldt,
+ A1, A1m, A1n, lda1, A2, A2m, A2n, lda2 );
+}
+
+static inline void
+INSERT_TASK_zttmqr( const RUNTIME_option_t *options,
+ cham_side_t side, cham_trans_t trans,
+ int m1, int n1, int m2, int n2, int k, int ib, int nb,
+ const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
+ const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt )
+{
+ return INSERT_TASK_ztpmqrt( options, side, trans, m2, n2, k, m2, ib, nb,
+ V, Vm, Vn, ldv, T, Tm, Tn, ldt,
+ A1, A1m, A1n, lda1, A2, A2m, A2n, lda2 );
+}
#endif /* _chameleon_tasks_z_h_ */
diff --git a/runtime/CMakeLists.txt b/runtime/CMakeLists.txt
index f14ef4838888f5481b52774b5b63116825fc00d8..73503ee0ce89aff056eb5971f4904bdbe0787315 100644
--- a/runtime/CMakeLists.txt
+++ b/runtime/CMakeLists.txt
@@ -21,7 +21,7 @@
# @author Cedric Castagnede
# @author Emmanuel Agullo
# @author Mathieu Faverge
-# @date 2012-07-13
+# @date 2018-11-07
#
###
@@ -86,17 +86,9 @@ set(CODELETS_ZSRC
codelets/codelet_ztrasm.c
codelets/codelet_ztrssq.c
codelets/codelet_ztrtri.c
- codelets/codelet_ztslqt.c
- codelets/codelet_ztsmlq.c
- codelets/codelet_ztsmqr.c
codelets/codelet_ztsmlq_hetra1.c
codelets/codelet_ztsmqr_hetra1.c
- codelets/codelet_ztsqrt.c
codelets/codelet_ztstrf.c
- codelets/codelet_zttlqt.c
- codelets/codelet_zttmlq.c
- codelets/codelet_zttmqr.c
- codelets/codelet_zttqrt.c
codelets/codelet_zunmlq.c
codelets/codelet_zunmqr.c
##################
diff --git a/runtime/parsec/codelets/codelet_ztslqt.c b/runtime/parsec/codelets/codelet_ztslqt.c
deleted file mode 100644
index 89c8721131a948e63657ebbd08eea7c553dca5c2..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_ztslqt.c
+++ /dev/null
@@ -1,70 +0,0 @@
-/**
- *
- * @file parsec/codelet_ztslqt.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztslqt PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_ztslqt_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- parsec_dtd_unpack_args(
- this_task, &m, &n, &ib, &A1, &lda1, &A2, &lda2, &T, &ldt, &TAU, &WORK );
-
- CORE_ztslqt( m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK );
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_ztslqt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_ztslqt_parsec, options->priority, "tslqt",
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_ztsmlq.c b/runtime/parsec/codelets/codelet_ztsmlq.c
deleted file mode 100644
index 56b86887aec4bcf6c3ad6685d6f027a40b188491..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_ztsmlq.c
+++ /dev/null
@@ -1,89 +0,0 @@
-/**
- *
- * @file parsec/codelet_ztsmlq.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsmlq PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_ztsmlq_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- parsec_dtd_unpack_args(
- this_task, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib, &A1, &lda1, &A2, &lda2, &V, &ldv, &T, &ldt, &WORK, &ldwork );
-
- CORE_ztsmlq( side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_ztsmlq(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_ztsmlq_parsec, options->priority, "tsmlq",
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( V, CHAMELEON_Complex64_t, Vm, Vn ), chameleon_parsec_get_arena_index( V ) | INPUT,
- sizeof(int), &ldv, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_ztsmqr.c b/runtime/parsec/codelets/codelet_ztsmqr.c
deleted file mode 100644
index e8059bde8466f5957cd435dbc0e1a7b3b55714eb..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_ztsmqr.c
+++ /dev/null
@@ -1,89 +0,0 @@
-/**
- *
- * @file parsec/codelet_ztsmqr.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsmqr PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_ztsmqr_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- parsec_dtd_unpack_args(
- this_task, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib, &A1, &lda1, &A2, &lda2, &V, &ldv, &T, &ldt, &WORK, &ldwork );
-
- CORE_ztsmqr( side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_ztsmqr(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_ztsmqr_parsec, options->priority, "tsmqr",
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( V, CHAMELEON_Complex64_t, Vm, Vn ), chameleon_parsec_get_arena_index( V ) | INPUT,
- sizeof(int), &ldv, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_ztsqrt.c b/runtime/parsec/codelets/codelet_ztsqrt.c
deleted file mode 100644
index a8edb3c0fdf0562f9c374bdb2100919d6f825b97..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_ztsqrt.c
+++ /dev/null
@@ -1,70 +0,0 @@
-/**
- *
- * @file parsec/codelet_ztsqrt.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsqrt PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_ztsqrt_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- parsec_dtd_unpack_args(
- this_task, &m, &n, &ib, &A1, &lda1, &A2, &lda2, &T, &ldt, &TAU, &WORK );
-
- CORE_ztsqrt( m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK );
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_ztsqrt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_ztsqrt_parsec, options->priority, "tsqrt",
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_zttlqt.c b/runtime/parsec/codelets/codelet_zttlqt.c
deleted file mode 100644
index 1a72dd5cffefc3c4cc929558042fa10d5343aa33..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_zttlqt.c
+++ /dev/null
@@ -1,71 +0,0 @@
-/**
- *
- * @file parsec/codelet_zttlqt.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttlqt PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_zttlqt_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- parsec_dtd_unpack_args(
- this_task, &m, &n, &ib, &A1, &lda1, &A2, &lda2, &T, &ldt, &TAU, &WORK );
-
- CORE_zttlqt( m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK );
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_zttlqt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
-
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_zttlqt_parsec, options->priority, "ttlqt",
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_zttmlq.c b/runtime/parsec/codelets/codelet_zttmlq.c
deleted file mode 100644
index b0788876c79e5197cbc30ffa4880b77a68531ff5..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_zttmlq.c
+++ /dev/null
@@ -1,89 +0,0 @@
-/**
- *
- * @file parsec/codelet_zttmlq.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttmlq PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_zttmlq_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- parsec_dtd_unpack_args(
- this_task, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib, &A1, &lda1, &A2, &lda2, &V, &ldv, &T, &ldt, &WORK, &ldwork );
-
- CORE_zttmlq( side, trans, m1, n1, m2, n2, k, ib, A1, lda1,
- A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_zttmlq(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_zttmlq_parsec, options->priority, "ttmlq",
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( V, CHAMELEON_Complex64_t, Vm, Vn ), chameleon_parsec_get_arena_index( V ) | INPUT,
- sizeof(int), &ldv, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_zttmqr.c b/runtime/parsec/codelets/codelet_zttmqr.c
deleted file mode 100644
index f8a8b8f6bc9fb415323554d7eb8efea1dab0725d..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_zttmqr.c
+++ /dev/null
@@ -1,90 +0,0 @@
-/**
- *
- * @file parsec/codelet_zttmqr.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttmqr PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_zttmqr_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- parsec_dtd_unpack_args(
- this_task, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib, &A1, &lda1, &A2, &lda2, &V, &ldv, &T, &ldt, &WORK, &ldwork );
-
- CORE_zttmqr( side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-
-void INSERT_TASK_zttmqr(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_zttmqr_parsec, options->priority, "ttmqr",
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( V, CHAMELEON_Complex64_t, Vm, Vn ), chameleon_parsec_get_arena_index( V ) | INPUT,
- sizeof(int), &ldv, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/parsec/codelets/codelet_zttqrt.c b/runtime/parsec/codelets/codelet_zttqrt.c
deleted file mode 100644
index 6b22180076622b164425b06089774675fc873cfb..0000000000000000000000000000000000000000
--- a/runtime/parsec/codelets/codelet_zttqrt.c
+++ /dev/null
@@ -1,70 +0,0 @@
-/**
- *
- * @file parsec/codelet_zttqrt.c
- *
- * @copyright 2009-2015 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttqrt PaRSEC codelet
- *
- * @version 1.0.0
- * @author Reazul Hoque
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_parsec.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static inline int
-CORE_zttqrt_parsec( parsec_execution_stream_t *context,
- parsec_task_t *this_task )
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- parsec_dtd_unpack_args(
- this_task, &m, &n, &ib, &A1, &lda1, &A2, &lda2, &T, &ldt, &TAU, &WORK );
-
- CORE_zttqrt( m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK );
-
- (void)context;
- return PARSEC_HOOK_RETURN_DONE;
-}
-
-void INSERT_TASK_zttqrt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- parsec_taskpool_t* PARSEC_dtd_taskpool = (parsec_taskpool_t *)(options->sequence->schedopt);
-
- parsec_dtd_taskpool_insert_task(
- PARSEC_dtd_taskpool, CORE_zttqrt_parsec, options->priority, "ttqrt",
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- PASSED_BY_REF, RTBLKADDR( A1, CHAMELEON_Complex64_t, A1m, A1n ), chameleon_parsec_get_arena_index( A1 ) | INOUT,
- sizeof(int), &lda1, VALUE,
- PASSED_BY_REF, RTBLKADDR( A2, CHAMELEON_Complex64_t, A2m, A2n ), chameleon_parsec_get_arena_index( A2 ) | INOUT | AFFINITY,
- sizeof(int), &lda2, VALUE,
- PASSED_BY_REF, RTBLKADDR( T, CHAMELEON_Complex64_t, Tm, Tn ), chameleon_parsec_get_arena_index( T ) | OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- PARSEC_DTD_ARG_END );
-}
diff --git a/runtime/quark/codelets/codelet_ztplqt.c b/runtime/quark/codelets/codelet_ztplqt.c
index 5b3f15ab4527f3d753d3e74d1d65932f6ce84e34..f0e51b3754d6460ea9d3be4a8e9d4f826c8a997f 100644
--- a/runtime/quark/codelets/codelet_ztplqt.c
+++ b/runtime/quark/codelets/codelet_ztplqt.c
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -50,7 +50,7 @@ void INSERT_TASK_ztplqt( const RUNTIME_option_t *options,
const CHAM_desc_t *T, int Tm, int Tn, int ldt )
{
quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSLQT;
+ DAG_CORE_TPLQT;
int shapeB = ( L == 0 ) ? 0 : (QUARK_REGION_L | QUARK_REGION_D);
diff --git a/runtime/quark/codelets/codelet_ztpmlqt.c b/runtime/quark/codelets/codelet_ztpmlqt.c
index e82f40c4dbe58525926f2381eef3cea39fb6ccdc..fa435550dbb7cea349898f43b5e8fa9ec71c8ef4 100644
--- a/runtime/quark/codelets/codelet_ztpmlqt.c
+++ b/runtime/quark/codelets/codelet_ztpmlqt.c
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -57,7 +57,7 @@ void INSERT_TASK_ztpmlqt( const RUNTIME_option_t *options,
const CHAM_desc_t *B, int Bm, int Bn, int ldb )
{
quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSMQR;
+ DAG_CORE_TPMLQT;
int shapeV = ( L == 0 ) ? 0 : (QUARK_REGION_L | QUARK_REGION_D);
diff --git a/runtime/quark/codelets/codelet_ztpmqrt.c b/runtime/quark/codelets/codelet_ztpmqrt.c
index 933ca3f1327d999f671f786002e6f559eb0130f8..bdf6627273065b4bdf18d93f07384628d5b113f7 100644
--- a/runtime/quark/codelets/codelet_ztpmqrt.c
+++ b/runtime/quark/codelets/codelet_ztpmqrt.c
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -57,7 +57,7 @@ void INSERT_TASK_ztpmqrt( const RUNTIME_option_t *options,
const CHAM_desc_t *B, int Bm, int Bn, int ldb )
{
quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSMQR;
+ DAG_CORE_TPMQRT;
int shapeV = ( L == 0 ) ? 0 : (QUARK_REGION_U | QUARK_REGION_D);
diff --git a/runtime/quark/codelets/codelet_ztpqrt.c b/runtime/quark/codelets/codelet_ztpqrt.c
index 50470ac8e806aed4479679a577629091f573b4cd..24ce98e124023f90184379c211200d21693ed503 100644
--- a/runtime/quark/codelets/codelet_ztpqrt.c
+++ b/runtime/quark/codelets/codelet_ztpqrt.c
@@ -13,7 +13,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -50,7 +50,7 @@ void INSERT_TASK_ztpqrt( const RUNTIME_option_t *options,
const CHAM_desc_t *T, int Tm, int Tn, int ldt )
{
quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSQRT;
+ DAG_CORE_TPQRT;
int shapeB = ( L == 0 ) ? 0 : (QUARK_REGION_U | QUARK_REGION_D);
diff --git a/runtime/quark/codelets/codelet_ztslqt.c b/runtime/quark/codelets/codelet_ztslqt.c
deleted file mode 100644
index 4efb19be91de19cb3819cf674e4a08a9f1defaaf..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_ztslqt.c
+++ /dev/null
@@ -1,143 +0,0 @@
-/**
- *
- * @file quark/codelet_ztslqt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztslqt Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_ztslqt_quark(Quark *quark)
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- quark_unpack_args_11(quark, m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
- CORE_ztslqt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztslqt computes a LQ factorization of a rectangular matrix
- * formed by coupling side-by-side a complex M-by-M
- * lower triangular tile A1 and a complex M-by-N tile A2:
- *
- * | A1 A2 | = L * Q
- *
- * The tile Q is represented as a product of elementary reflectors
- *
- * Q = H(k)' . . . H(2)' H(1)', where k = min(M,N).
- *
- * Each H(i) has the form
- *
- * H(i) = I - tau * v * v'
- *
- * where tau is a complex scalar, and v is a complex vector with
- * v(1:i-1) = 0 and v(i) = 1; conjg(v(i+1:n)) is stored on exit in
- * A2(i,1:n), and tau in TAU(i).
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of rows of the tile A1 and A2. M >= 0.
- * The number of columns of the tile A1.
- *
- * @param[in] N
- * The number of columns of the tile A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M-by-M tile A1.
- * On exit, the elements on and below the diagonal of the array
- * contain the M-by-M lower trapezoidal tile L;
- * the elements above the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M).
- *
- * @param[in,out] A2
- * On entry, the M-by-N tile A2.
- * On exit, all the elements with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_ztslqt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSLQT;
- QUARK_Insert_Task(opt->quark, CORE_ztslqt_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT | QUARK_REGION_L | QUARK_REGION_D,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_ztsmlq.c b/runtime/quark/codelets/codelet_ztsmlq.c
deleted file mode 100644
index b3003d130f6c08c8e5b60130b13b1baf5653ed04..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_ztsmlq.c
+++ /dev/null
@@ -1,190 +0,0 @@
-/**
- *
- * @file quark/codelet_ztsmlq.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsmlq Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Azzam Haidar
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_ztsmlq_quark(Quark *quark)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- quark_unpack_args_18(quark, side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
- CORE_ztsmlq(side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztsmlq overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 A2 | * Q
- * | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 A2 | * Q**H
- * | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(k)' . . . H(2)' H(1)'
- *
- * as returned by CORE_ZTSLQT.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- * M2 = M1 if side == ChamRight.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- * N2 = N1 if side == ChamLeft.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTSLQT in the first k rows of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size
- * LDWORK-by-M1 if side == ChamLeft
- * LDWORK-by-IB if side == ChamRight
- *
- * @param[in] LDWORK
- * The leading dimension of the array WORK.
- * LDWORK >= max(1,IB) if side == ChamLeft
- * LDWORK >= max(1,N1) if side == ChamRight
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_ztsmlq(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSMLQ;
- QUARK_Insert_Task(opt->quark, CORE_ztsmlq_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn), INPUT,
- sizeof(int), &ldv, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_ztsmqr.c b/runtime/quark/codelets/codelet_ztsmqr.c
deleted file mode 100644
index afcde5dfa02d9f67e8363f57698b47bd256ea59c..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_ztsmqr.c
+++ /dev/null
@@ -1,190 +0,0 @@
-/**
- *
- * @file quark/codelet_ztsmqr.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsmqr Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Azzam Haidar
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_ztsmqr_quark(Quark *quark)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- quark_unpack_args_18(quark, side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
- CORE_ztsmqr(side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztsmqr overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 A2 | * Q
- * | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 A2 | * Q**H
- * | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k)
- *
- * as returned by CORE_ZTSQRT.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- * M2 = M1 if side == ChamRight.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- * N2 = N1 if side == ChamLeft.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTSQRT in the first k columns of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size
- * LDWORK-by-N1 if side == ChamLeft
- * LDWORK-by-IB if side == ChamRight
- *
- * @param[in] LDWORK
- * The leading dimension of the array WORK.
- * LDWORK >= max(1,IB) if side == ChamLeft
- * LDWORK >= max(1,M1) if side == ChamRight
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_ztsmqr(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSMQR;
- QUARK_Insert_Task(opt->quark, CORE_ztsmqr_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn), INPUT,
- sizeof(int), &ldv, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_ztsqrt.c b/runtime/quark/codelets/codelet_ztsqrt.c
deleted file mode 100644
index 44457debb9112a93fd00ad3a6f6585d7f9d49ef0..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_ztsqrt.c
+++ /dev/null
@@ -1,131 +0,0 @@
-/**
- *
- * @file quark/codelet_ztsqrt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsqrt Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_ztsqrt_quark(Quark *quark)
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- quark_unpack_args_11(quark, m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
- CORE_ztsqrt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztsqrt computes a QR factorization of a rectangular matrix
- * formed by coupling a complex N-by-N upper triangular tile A1
- * on top of a complex M-by-N tile A2:
- *
- * | A1 | = Q * R
- * | A2 |
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of columns of the tile A2. M >= 0.
- *
- * @param[in] N
- * The number of rows of the tile A1.
- * The number of columns of the tiles A1 and A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the N-by-N tile A1.
- * On exit, the elements on and above the diagonal of the array
- * contain the N-by-N upper trapezoidal tile R;
- * the elements below the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,N).
- *
- * @param[in,out] A2
- * On entry, the M-by-N tile A2.
- * On exit, all the elements with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_ztsqrt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TSQRT;
- QUARK_Insert_Task(opt->quark, CORE_ztsqrt_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT | QUARK_REGION_U | QUARK_REGION_D,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_zttlqt.c b/runtime/quark/codelets/codelet_zttlqt.c
deleted file mode 100644
index 85eb8e3d80859879501a3592eff0ec4628c88e13..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_zttlqt.c
+++ /dev/null
@@ -1,143 +0,0 @@
-/**
- *
- * @file quark/codelet_zttlqt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttlqt Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_zttlqt_quark(Quark *quark)
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- quark_unpack_args_11(quark, m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
- CORE_zttlqt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttlqt computes a LQ factorization of a rectangular matrix
- * formed by coupling side-by-side a complex M-by-M lower triangular tile A1
- * and a complex M-by-N lower triangular tile A2:
- *
- * | A1 A2 | = L * Q
- *
- * The tile Q is represented as a product of elementary reflectors
- *
- * Q = H(k)' . . . H(2)' H(1)', where k = min(M,N).
- *
- * Each H(i) has the form
- *
- * H(i) = I - tau * v * v'
- *
- * where tau is a complex scalar, and v is a complex vector with
- * v(1:i-1) = 0 and v(i) = 1; conjg(v(i+1:n)) is stored on exit in
- * A2(i,1:n), and tau in TAU(i).
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of rows of the tile A1 and A2. M >= 0.
- * The number of columns of the tile A1.
- *
- * @param[in] N
- * The number of columns of the tile A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M-by-M tile A1.
- * On exit, the elements on and below the diagonal of the array
- * contain the M-by-M lower trapezoidal tile L;
- * the elements above the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,N).
- *
- * @param[in,out] A2
- * On entry, the M-by-N lower triangular tile A2.
- * On exit, the elements on and below the diagonal of the array
- * with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the array A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[in,out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_zttlqt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TTLQT;
- QUARK_Insert_Task(opt->quark, CORE_zttlqt_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT | QUARK_REGION_L | QUARK_REGION_D,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | QUARK_REGION_L | QUARK_REGION_D | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_zttmlq.c b/runtime/quark/codelets/codelet_zttmlq.c
deleted file mode 100644
index f3701869ca32a8a1376a8cf04d500eb8049a36f0..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_zttmlq.c
+++ /dev/null
@@ -1,182 +0,0 @@
-/**
- *
- * @file quark/codelet_zttmlq.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttmlq Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_zttmlq_quark(Quark *quark)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- quark_unpack_args_18(quark, side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
- CORE_zttmlq(side, trans, m1, n1, m2, n2, k, ib, A1, lda1,
- A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttmlq overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 (N1 == N2) with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 | * Q
- * | A2 | | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 | * Q**H
- * | A2 | | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k)
- *
- * as returned by CORE_zttqrt.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTTQRT in the first k rows of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size LDWORK-by-N1.
- *
- * @param[in] LDWORK
- * The dimension of the array WORK. LDWORK >= max(1,IB).
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_zttmlq(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TTMLQ;
- QUARK_Insert_Task(opt->quark, CORE_zttmlq_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn), INPUT | QUARK_REGION_L | QUARK_REGION_D,
- sizeof(int), &ldv, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_zttmqr.c b/runtime/quark/codelets/codelet_zttmqr.c
deleted file mode 100644
index e106a34ce42948c3786f10f3af9930368299a781..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_zttmqr.c
+++ /dev/null
@@ -1,183 +0,0 @@
-/**
- *
- * @file quark/codelet_zttmqr.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttmqr Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-static void
-CORE_zttmqr_quark( Quark *quark )
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- quark_unpack_args_18(quark, side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
- CORE_zttmqr(side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttmqr overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 (N1 == N2) with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 | * Q
- * | A2 | | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 | * Q**H
- * | A2 | | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k)
- *
- * as returned by CORE_zttqrt.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTTQRT in the first k rows of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size LDWORK-by-N1.
- *
- * @param[in] LDWORK
- * The dimension of the array WORK. LDWORK >= max(1,IB).
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_zttmqr(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- int ldwork = side == ChamLeft ? ib : nb;
-
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TTMQR;
- QUARK_Insert_Task(opt->quark, CORE_zttmqr_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &side, VALUE,
- sizeof(int), &trans, VALUE,
- sizeof(int), &m1, VALUE,
- sizeof(int), &n1, VALUE,
- sizeof(int), &m2, VALUE,
- sizeof(int), &n2, VALUE,
- sizeof(int), &k, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn), INPUT | QUARK_REGION_U | QUARK_REGION_D,
- sizeof(int), &ldv, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), INPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- sizeof(int), &ldwork, VALUE,
- 0);
-}
diff --git a/runtime/quark/codelets/codelet_zttqrt.c b/runtime/quark/codelets/codelet_zttqrt.c
deleted file mode 100644
index d5f62c44110d6d8f39baecfd9ca2ba75c611f0b1..0000000000000000000000000000000000000000
--- a/runtime/quark/codelets/codelet_zttqrt.c
+++ /dev/null
@@ -1,143 +0,0 @@
-/**
- *
- * @file quark/codelet_zttqrt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttqrt Quark codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_quark.h"
-#include "chameleon/tasks_z.h"
-#include "coreblas/coreblas_z.h"
-
-void CORE_zttqrt_quark(Quark *quark)
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- quark_unpack_args_11(quark, m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
- CORE_zttqrt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttqrt computes a QR factorization of a rectangular matrix
- * formed by coupling a complex N-by-N upper triangular tile A1
- * on top of a complex M-by-N upper trapezoidal tile A2:
- *
- * | A1 | = Q * R
- * | A2 |
- *
- * The tile Q is represented as a product of elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k), where k = min(M,N).
- *
- * Each H(i) has the form
- *
- * H(i) = I - tau * v * v'
- *
- * where tau is a complex scalar, and v is a complex vector with
- * v(1:i-1) = 0 and v(i) = 1; v(i+1:m) is stored on exit in A2(1:m,i),
- * and tau in TAU(i).
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of rows of the tile A2. M >= 0.
- *
- * @param[in] N
- * The number of columns of the tile A1 and A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the N-by-N tile A1.
- * On exit, the elements on and above the diagonal of the array
- * contain the N-by-N upper trapezoidal tile R;
- * the elements below the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,N).
- *
- * @param[in,out] A2
- * On entry, the M-by-N upper triangular tile A2.
- * On exit, the elements on and above the diagonal of the array
- * with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the array A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[in,out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-void INSERT_TASK_zttqrt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- quark_option_t *opt = (quark_option_t*)(options->schedopt);
- DAG_CORE_TTQRT;
- QUARK_Insert_Task(opt->quark, CORE_zttqrt_quark, (Quark_Task_Flags*)opt,
- sizeof(int), &m, VALUE,
- sizeof(int), &n, VALUE,
- sizeof(int), &ib, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n), INOUT | QUARK_REGION_U | QUARK_REGION_D,
- sizeof(int), &lda1, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb*nb, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n), INOUT | QUARK_REGION_U | QUARK_REGION_D | LOCALITY,
- sizeof(int), &lda2, VALUE,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn), OUTPUT,
- sizeof(int), &ldt, VALUE,
- sizeof(CHAMELEON_Complex64_t)*nb, NULL, SCRATCH,
- sizeof(CHAMELEON_Complex64_t)*ib*nb, NULL, SCRATCH,
- 0);
-}
diff --git a/runtime/quark/include/core_blas_dag.h b/runtime/quark/include/core_blas_dag.h
index fbee7539cdfcaa3066a71b5b85eb243a903325dc..649330aa1871b6e60cd322ba34f49ba9d4839af8 100644
--- a/runtime/quark/include/core_blas_dag.h
+++ b/runtime/quark/include/core_blas_dag.h
@@ -14,7 +14,7 @@
* @version 1.0.0
* @author Mathieu Faverge
* @author Cedric Castagnede
- * @date 2010-11-15
+ * @date 2018-11-08
*
*/
#ifndef _core_blas_dag_h_
@@ -71,16 +71,21 @@
#define DAG_CORE_TRSM DAG_SET_PROPERTIES( "TRSM" , "cyan" )
#define DAG_CORE_TRSSQ DAG_SET_PROPERTIES( "TRSSQ" , "white" )
#define DAG_CORE_TRTRI DAG_SET_PROPERTIES( "TRTRI" , "white" )
-#define DAG_CORE_TSLQT DAG_SET_PROPERTIES( "TSLQT" , "red" )
-#define DAG_CORE_TSMLQ DAG_SET_PROPERTIES( "TSMLQ" , "yellow" )
-#define DAG_CORE_TSMQR DAG_SET_PROPERTIES( "TSMQR" , "yellow" )
-#define DAG_CORE_TSQRT DAG_SET_PROPERTIES( "TSQRT" , "red" )
+#define DAG_CORE_TPLQT DAG_SET_PROPERTIES( "TPLQT" , "red" )
+#define DAG_CORE_TPMLQT DAG_SET_PROPERTIES( "TPMLQT" , "yellow" )
+#define DAG_CORE_TPMQRT DAG_SET_PROPERTIES( "TPMQRT" , "yellow" )
+#define DAG_CORE_TPQRT DAG_SET_PROPERTIES( "TPQRT" , "red" )
#define DAG_CORE_TSTRF DAG_SET_PROPERTIES( "TSTRF" , "red" )
-#define DAG_CORE_TTLQT DAG_SET_PROPERTIES( "TTLQT" , "pink" )
-#define DAG_CORE_TTMLQ DAG_SET_PROPERTIES( "TTMLQ" , "magenta" )
-#define DAG_CORE_TTMQR DAG_SET_PROPERTIES( "TTMQR" , "magenta" )
-#define DAG_CORE_TTQRT DAG_SET_PROPERTIES( "TTQRT" , "pink" )
#define DAG_CORE_UNMLQ DAG_SET_PROPERTIES( "UNMLQ" , "cyan" )
#define DAG_CORE_UNMQR DAG_SET_PROPERTIES( "UNMQR" , "cyan" )
+#define DAG_CORE_TSLQT DAG_CORE_TPLQT
+#define DAG_CORE_TSMLQ DAG_CORE_TPMLQT
+#define DAG_CORE_TSMQR DAG_CORE_TPMQRT
+#define DAG_CORE_TSQRT DAG_CORE_TPQRT
+#define DAG_CORE_TTLQT DAG_CORE_TPLQT
+#define DAG_CORE_TTMLQ DAG_CORE_TPMLQT
+#define DAG_CORE_TTMQR DAG_CORE_TPMQRT
+#define DAG_CORE_TTQRT DAG_CORE_TPQRT
+
#endif /* _core_blas_dag_h_ */
diff --git a/runtime/starpu/codelets/codelet_zcallback.c b/runtime/starpu/codelets/codelet_zcallback.c
index c42035d967b99c896705c8341a70155e26678025..745b122e3f11a4c9d92a126369fb736d9b8bb4fe 100644
--- a/runtime/starpu/codelets/codelet_zcallback.c
+++ b/runtime/starpu/codelets/codelet_zcallback.c
@@ -15,7 +15,7 @@
* @author Mathieu Faverge
* @author Cedric Augonnet
* @author Florent Pruvost
- * @date 2015-09-16
+ * @date 2018-11-08
* @precisions normal z -> c d s
*
*/
@@ -69,16 +69,8 @@ CHAMELEON_CL_CB(ztrasm, starpu_matrix_get_nx(task->handles[0]), starpu_ma
CHAMELEON_CL_CB(ztrmm, starpu_matrix_get_nx(task->handles[1]), starpu_matrix_get_ny(task->handles[1]), 0, M*M*N)
CHAMELEON_CL_CB(ztrsm, starpu_matrix_get_nx(task->handles[1]), starpu_matrix_get_ny(task->handles[1]), 0, M*M*N)
CHAMELEON_CL_CB(ztrtri, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (1./3.)*M *M*M)
-CHAMELEON_CL_CB(ztslqt, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), 2. *M* M*M)
-CHAMELEON_CL_CB(ztsmlq, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (4.0*M+starpu_matrix_get_nx(task->handles[3]))*M*M)
-CHAMELEON_CL_CB(ztsmqr, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (4.0*M+starpu_matrix_get_nx(task->handles[3]))*M*M)
CHAMELEON_CL_CB(ztsmlq_hetra1, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (4.0*M+starpu_matrix_get_nx(task->handles[3]))*M*M)
CHAMELEON_CL_CB(ztsmqr_hetra1, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (4.0*M+starpu_matrix_get_nx(task->handles[3]))*M*M)
-CHAMELEON_CL_CB(ztsqrt, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), 2. *M* M*M)
CHAMELEON_CL_CB(ztstrf, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), M* M*M)
-CHAMELEON_CL_CB(zttlqt, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), 1. *M* M*M)
-CHAMELEON_CL_CB(zttmlq, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (2.0*M+starpu_matrix_get_nx(task->handles[3]))*M*M)
-CHAMELEON_CL_CB(zttmqr, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), (2.0*M+starpu_matrix_get_nx(task->handles[3]))*M*M)
-CHAMELEON_CL_CB(zttqrt, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), 1. *M* M*M)
CHAMELEON_CL_CB(zunmlq, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), 2. *M* M*M)
CHAMELEON_CL_CB(zunmqr, starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), starpu_matrix_get_nx(task->handles[0]), 2. *M* M*M)
diff --git a/runtime/starpu/codelets/codelet_ztpmlqt.c b/runtime/starpu/codelets/codelet_ztpmlqt.c
index 714b45762488101d571b7ed4a7f1132a256d5301..8dffa4ff2292c448c710713a7556bec3920f3d73 100644
--- a/runtime/starpu/codelets/codelet_ztpmlqt.c
+++ b/runtime/starpu/codelets/codelet_ztpmlqt.c
@@ -11,7 +11,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-07
* @precisions normal z -> s d c
*
*/
@@ -37,6 +37,7 @@ static void cl_ztpmlqt_cpu_func(void *descr[], void *cl_arg)
CHAMELEON_Complex64_t *B;
int ldb;
CHAMELEON_Complex64_t *WORK;
+ size_t lwork;
V = (const CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
T = (const CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
@@ -45,13 +46,15 @@ static void cl_ztpmlqt_cpu_func(void *descr[], void *cl_arg)
WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* ib * nb */
starpu_codelet_unpack_args( cl_arg, &side, &trans, &M, &N, &K, &L, &ib,
- &ldv, &ldt, &lda, &ldb );
+ &ldv, &ldt, &lda, &ldb, &lwork );
CORE_ztpmlqt( side, trans, M, N, K, L, ib,
V, ldv, T, ldt, A, lda, B, ldb, WORK );
+
+ (void)lwork;
}
-#if defined(CHAMELEON_USE_CUDA) && 0
+#if defined(CHAMELEON_USE_CUDA)
static void cl_ztpmlqt_cuda_func(void *descr[], void *cl_arg)
{
cham_side_t side;
@@ -70,6 +73,7 @@ static void cl_ztpmlqt_cuda_func(void *descr[], void *cl_arg)
cuDoubleComplex *B;
int ldb;
cuDoubleComplex *W;
+ size_t lwork;
V = (const cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[0]);
T = (const cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[1]);
@@ -78,14 +82,14 @@ static void cl_ztpmlqt_cuda_func(void *descr[], void *cl_arg)
W = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[4]); /* 2*ib*nb */
starpu_codelet_unpack_args( cl_arg, &side, &trans, &M, &N, &K, &L, &ib,
- &ldv, &ldt, &lda, &ldb );
+ &ldv, &ldt, &lda, &ldb, &lwork );
RUNTIME_getStream(stream);
CUDA_ztpmlqt(
side, trans, M, N, K, L, ib,
V, ldv, T, ldt, A, lda, B, ldb,
- W, stream );
+ W, lwork, stream );
#ifndef STARPU_CUDA_ASYNC
cudaStreamSynchronize( stream );
@@ -97,8 +101,7 @@ static void cl_ztpmlqt_cuda_func(void *descr[], void *cl_arg)
/*
* Codelet definition
*/
-CODELETS_CPU(ztpmlqt, 5, cl_ztpmlqt_cpu_func)
-//CODELETS(ztpmlqt, 5, cl_ztpmlqt_cpu_func, cl_ztpmlqt_cuda_func, STARPU_CUDA_ASYNC)
+CODELETS(ztpmlqt, 5, cl_ztpmlqt_cpu_func, cl_ztpmlqt_cuda_func, STARPU_CUDA_ASYNC)
void
INSERT_TASK_ztpmlqt( const RUNTIME_option_t *options,
@@ -136,6 +139,7 @@ INSERT_TASK_ztpmlqt( const RUNTIME_option_t *options,
STARPU_VALUE, &lda, sizeof(int),
STARPU_RW, RTBLKADDR(B, CHAMELEON_Complex64_t, Bm, Bn),
STARPU_VALUE, &ldb, sizeof(int),
+ STARPU_VALUE, &(options->ws_wsize), sizeof(size_t),
/* Other options */
STARPU_SCRATCH, options->ws_worker,
STARPU_PRIORITY, options->priority,
diff --git a/runtime/starpu/codelets/codelet_ztpmqrt.c b/runtime/starpu/codelets/codelet_ztpmqrt.c
index 40f83ab2d137bfe6c9de11d58a4ad05307edb78e..6684e59f82d579cd129a1ff40c25dd377bc6166e 100644
--- a/runtime/starpu/codelets/codelet_ztpmqrt.c
+++ b/runtime/starpu/codelets/codelet_ztpmqrt.c
@@ -11,7 +11,7 @@
*
* @version 1.0.0
* @author Mathieu Faverge
- * @date 2016-12-15
+ * @date 2018-11-07
* @precisions normal z -> s d c
*
*/
@@ -37,6 +37,7 @@ static void cl_ztpmqrt_cpu_func(void *descr[], void *cl_arg)
CHAMELEON_Complex64_t *B;
int ldb;
CHAMELEON_Complex64_t *WORK;
+ size_t lwork;
V = (const CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
T = (const CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
@@ -45,10 +46,12 @@ static void cl_ztpmqrt_cpu_func(void *descr[], void *cl_arg)
WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* ib * nb */
starpu_codelet_unpack_args( cl_arg, &side, &trans, &M, &N, &K, &L, &ib,
- &ldv, &ldt, &lda, &ldb );
+ &ldv, &ldt, &lda, &ldb, &lwork );
CORE_ztpmqrt( side, trans, M, N, K, L, ib,
V, ldv, T, ldt, A, lda, B, ldb, WORK );
+
+ (void)lwork;
}
@@ -71,22 +74,23 @@ static void cl_ztpmqrt_cuda_func(void *descr[], void *cl_arg)
cuDoubleComplex *B;
int ldb;
cuDoubleComplex *W;
+ size_t lwork;
V = (const cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[0]);
T = (const cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[1]);
A = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[2]);
B = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[3]);
- W = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[4]); /* 2*ib*nb */
+ W = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[4]); /* 3*ib*nb */
starpu_codelet_unpack_args( cl_arg, &side, &trans, &M, &N, &K, &L, &ib,
- &ldv, &ldt, &lda, &ldb );
+ &ldv, &ldt, &lda, &ldb, &lwork );
RUNTIME_getStream(stream);
CUDA_ztpmqrt(
side, trans, M, N, K, L, ib,
V, ldv, T, ldt, A, lda, B, ldb,
- W, stream );
+ W, lwork, stream );
#ifndef STARPU_CUDA_ASYNC
cudaStreamSynchronize( stream );
@@ -102,12 +106,12 @@ CODELETS(ztpmqrt, 5, cl_ztpmqrt_cpu_func, cl_ztpmqrt_cuda_func, STARPU_CUDA_ASYN
void
INSERT_TASK_ztpmqrt( const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int M, int N, int K, int L, int ib, int nb,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt,
- const CHAM_desc_t *A, int Am, int An, int lda,
- const CHAM_desc_t *B, int Bm, int Bn, int ldb )
+ cham_side_t side, cham_trans_t trans,
+ int M, int N, int K, int L, int ib, int nb,
+ const CHAM_desc_t *V, int Vm, int Vn, int ldv,
+ const CHAM_desc_t *T, int Tm, int Tn, int ldt,
+ const CHAM_desc_t *A, int Am, int An, int lda,
+ const CHAM_desc_t *B, int Bm, int Bn, int ldb )
{
struct starpu_codelet *codelet = &cl_ztpmqrt;
void (*callback)(void*) = options->profiling ? cl_ztpmqrt_callback : NULL;
@@ -136,6 +140,7 @@ INSERT_TASK_ztpmqrt( const RUNTIME_option_t *options,
STARPU_VALUE, &lda, sizeof(int),
STARPU_RW, RTBLKADDR(B, CHAMELEON_Complex64_t, Bm, Bn),
STARPU_VALUE, &ldb, sizeof(int),
+ STARPU_VALUE, &(options->ws_wsize), sizeof(size_t),
/* Other options */
STARPU_SCRATCH, options->ws_worker,
STARPU_PRIORITY, options->priority,
diff --git a/runtime/starpu/codelets/codelet_ztslqt.c b/runtime/starpu/codelets/codelet_ztslqt.c
deleted file mode 100644
index 870e1349af880473ae1e92787e84e0a6ad3df035..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_ztslqt.c
+++ /dev/null
@@ -1,174 +0,0 @@
-/**
- *
- * @file starpu/codelet_ztslqt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztslqt StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztslqt computes a LQ factorization of a rectangular matrix
- * formed by coupling side-by-side a complex M-by-M
- * lower triangular tile A1 and a complex M-by-N tile A2:
- *
- * | A1 A2 | = L * Q
- *
- * The tile Q is represented as a product of elementary reflectors
- *
- * Q = H(k)' . . . H(2)' H(1)', where k = min(M,N).
- *
- * Each H(i) has the form
- *
- * H(i) = I - tau * v * v'
- *
- * where tau is a complex scalar, and v is a complex vector with
- * v(1:i-1) = 0 and v(i) = 1; conjg(v(i+1:n)) is stored on exit in
- * A2(i,1:n), and tau in TAU(i).
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of rows of the tile A1 and A2. M >= 0.
- * The number of columns of the tile A1.
- *
- * @param[in] N
- * The number of columns of the tile A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M-by-M tile A1.
- * On exit, the elements on and below the diagonal of the array
- * contain the M-by-M lower trapezoidal tile L;
- * the elements above the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M).
- *
- * @param[in,out] A2
- * On entry, the M-by-N tile A2.
- * On exit, all the elements with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_ztslqt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_ztslqt;
- void (*callback)(void*) = options->profiling ? cl_ztslqt_callback : NULL;
- CHAMELEON_starpu_ws_t *h_work = (CHAMELEON_starpu_ws_t*)(options->ws_host);
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_W(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &m, sizeof(int),
- STARPU_VALUE, &n, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_W, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* max( nb * (ib+1), ib * (ib+nb) ) */
- STARPU_SCRATCH, options->ws_worker,
- /* /\* 2 * ib * (nb+ib) + nb *\/ */
- STARPU_VALUE, &h_work, sizeof(CHAMELEON_starpu_ws_t *),
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "ztslqt",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_ztslqt_cpu_func(void *descr[], void *cl_arg)
-{
- CHAMELEON_starpu_ws_t *h_work;
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU, *WORK;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- TAU= (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]); /* nb + ib*nb */
-
- starpu_codelet_unpack_args(cl_arg, &m, &n, &ib, &lda1, &lda2, &ldt, &h_work);
-
- WORK = TAU + chameleon_max( m, n );
- CORE_ztslqt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS_CPU(ztslqt, 4, cl_ztslqt_cpu_func)
diff --git a/runtime/starpu/codelets/codelet_ztsmlq.c b/runtime/starpu/codelets/codelet_ztsmlq.c
deleted file mode 100644
index b0a2e38ec8ead1eb249c0e29ba5f98f8a8e91a90..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_ztsmlq.c
+++ /dev/null
@@ -1,266 +0,0 @@
-/**
- *
- * @file starpu/codelet_ztsmlq.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsmlq StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Azzam Haidar
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztsmlq overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 A2 | * Q
- * | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 A2 | * Q**H
- * | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(k)' . . . H(2)' H(1)'
- *
- * as returned by CORE_ZTSLQT.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- * M2 = M1 if side == ChamRight.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- * N2 = N1 if side == ChamLeft.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTSLQT in the first k rows of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size
- * LDWORK-by-M1 if side == ChamLeft
- * LDWORK-by-IB if side == ChamRight
- *
- * @param[in] LDWORK
- * The leading dimension of the array WORK.
- * LDWORK >= max(1,IB) if side == ChamLeft
- * LDWORK >= max(1,N1) if side == ChamRight
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_ztsmlq(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- struct starpu_codelet *codelet = &cl_ztsmlq;
- void (*callback)(void*) = options->profiling ? cl_ztsmlq_callback : NULL;
- int ldwork = side == ChamLeft ? ib : nb;
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_R(V, Vm, Vn);
- CHAMELEON_ACCESS_R(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &side, sizeof(int),
- STARPU_VALUE, &trans, sizeof(int),
- STARPU_VALUE, &m1, sizeof(int),
- STARPU_VALUE, &n1, sizeof(int),
- STARPU_VALUE, &m2, sizeof(int),
- STARPU_VALUE, &n2, sizeof(int),
- STARPU_VALUE, &k, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_R, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn),
- STARPU_VALUE, &ldv, sizeof(int),
- STARPU_R, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* max( ib*nb, 2*ib*nb ) */
- STARPU_SCRATCH, options->ws_worker,
- STARPU_VALUE, &ldwork, sizeof(int),
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "ztsmlq",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_ztsmlq_cpu_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]);
- WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* ib * nb */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- CORE_ztsmlq(side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-#if defined(CHAMELEON_USE_CUDA)
-static void cl_ztsmlq_cuda_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- cuDoubleComplex *A1;
- int lda1;
- cuDoubleComplex *A2;
- int lda2;
- cuDoubleComplex *V;
- int ldv;
- cuDoubleComplex *T;
- int ldt;
- cuDoubleComplex *W, *WC;
- int ldwork;
- int ldworkc;
-
- A1 = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[3]);
- W = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[4]); /* 2*ib*nb */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- WC = W + ib * ldwork;
- ldworkc = (side == ChamLeft) ? m1 : ib;
-
- RUNTIME_getStream(stream);
-
- CUDA_ztsmlq( side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt,
- W, ldwork, WC, ldworkc, stream );
-
-#ifndef STARPU_CUDA_ASYNC
- cudaStreamSynchronize( stream );
-#endif
-}
-#endif /* defined(CHAMELEON_USE_CUDA) */
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS(ztsmlq, 5, cl_ztsmlq_cpu_func, cl_ztsmlq_cuda_func, STARPU_CUDA_ASYNC)
diff --git a/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c b/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c
index 8996121be9fbce09fb360a4cbe0c7ec262cff5fc..d68e2bebf803ac1c449c30435dec5ea1883ba86e 100644
--- a/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c
+++ b/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c
@@ -15,7 +15,7 @@
* @author Hatem Ltaief
* @author Mathieu Faverge
* @author Azzam Haidar
- * @date 2010-11-15
+ * @date 2018-11-07
* @precisions normal z -> c d s
*
*/
@@ -106,8 +106,8 @@ static void cl_ztsmlq_hetra1_cpu_func(void *descr[], void *cl_arg)
T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]);
WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* ib * nb */
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k,
- &ib, &nb, &lda1, &lda2, &ldv, &ldt, &ldwork);
+ starpu_codelet_unpack_args( cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k,
+ &ib, &nb, &lda1, &lda2, &ldv, &ldt, &ldwork);
CORE_ztsmlq_hetra1(side, trans, m1, n1, m2, n2, k,
ib, A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
}
diff --git a/runtime/starpu/codelets/codelet_ztsmqr.c b/runtime/starpu/codelets/codelet_ztsmqr.c
deleted file mode 100644
index c38a8fbd129656fc5681a25a00bd7a57cefdb6e9..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_ztsmqr.c
+++ /dev/null
@@ -1,271 +0,0 @@
-/**
- *
- * @file starpu/codelet_ztsmqr.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsmqr StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Azzam Haidar
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztsmqr overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 A2 | * Q
- * | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 A2 | * Q**H
- * | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k)
- *
- * as returned by CORE_ZTSQRT.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- * M2 = M1 if side == ChamRight.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- * N2 = N1 if side == ChamLeft.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTSQRT in the first k columns of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size
- * LDWORK-by-N1 if side == ChamLeft
- * LDWORK-by-IB if side == ChamRight
- *
- * @param[in] LDWORK
- * The leading dimension of the array WORK.
- * LDWORK >= max(1,IB) if side == ChamLeft
- * LDWORK >= max(1,M1) if side == ChamRight
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_ztsmqr(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_ztsmqr;
- void (*callback)(void*) = options->profiling ? cl_ztsmqr_callback : NULL;
- int ldwork = side == ChamLeft ? ib : nb;
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_R(V, Vm, Vn);
- CHAMELEON_ACCESS_R(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &side, sizeof(int),
- STARPU_VALUE, &trans, sizeof(int),
- STARPU_VALUE, &m1, sizeof(int),
- STARPU_VALUE, &n1, sizeof(int),
- STARPU_VALUE, &m2, sizeof(int),
- STARPU_VALUE, &n2, sizeof(int),
- STARPU_VALUE, &k, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_R, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn),
- STARPU_VALUE, &ldv, sizeof(int),
- STARPU_R, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* max( ib*nb, 2*ib*nb ) */
- STARPU_SCRATCH, options->ws_worker,
- STARPU_VALUE, &ldwork, sizeof(int),
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_USE_MPI)
- STARPU_EXECUTE_ON_NODE, A2->get_rankof(A2, A2m, A2n),
-#endif
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "ztsmqr",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_ztsmqr_cpu_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]);
- WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* ib * nb */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- CORE_ztsmqr(side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-#if defined(CHAMELEON_USE_CUDA)
-static void cl_ztsmqr_cuda_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- cuDoubleComplex *A1;
- int lda1;
- cuDoubleComplex *A2;
- int lda2;
- cuDoubleComplex *V;
- int ldv;
- cuDoubleComplex *T;
- int ldt;
- cuDoubleComplex *W, *WC;
- int ldwork;
- int ldworkc;
-
- A1 = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[3]);
- W = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[4]); /* 2*ib*nb */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- WC = W + ib * (side == ChamLeft ? m1 : n1);
- ldworkc = (side == ChamLeft) ? m2 : ib;
-
- RUNTIME_getStream(stream);
-
- CUDA_ztsmqr(
- side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt,
- W, ldwork, WC, ldworkc, stream );
-
-#ifndef STARPU_CUDA_ASYNC
- cudaStreamSynchronize( stream );
-#endif
-}
-#endif /* defined(CHAMELEON_USE_CUDA) */
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS(ztsmqr, 5, cl_ztsmqr_cpu_func, cl_ztsmqr_cuda_func, STARPU_CUDA_ASYNC)
diff --git a/runtime/starpu/codelets/codelet_ztsqrt.c b/runtime/starpu/codelets/codelet_ztsqrt.c
deleted file mode 100644
index cb93ba7c22c3f832f365f293545366f03863bd6c..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_ztsqrt.c
+++ /dev/null
@@ -1,166 +0,0 @@
-/**
- *
- * @file starpu/codelet_ztsqrt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon ztsqrt StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Jakub Kurzak
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_ztsqrt computes a QR factorization of a rectangular matrix
- * formed by coupling a complex N-by-N upper triangular tile A1
- * on top of a complex M-by-N tile A2:
- *
- * | A1 | = Q * R
- * | A2 |
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of columns of the tile A2. M >= 0.
- *
- * @param[in] N
- * The number of rows of the tile A1.
- * The number of columns of the tiles A1 and A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the N-by-N tile A1.
- * On exit, the elements on and above the diagonal of the array
- * contain the N-by-N upper trapezoidal tile R;
- * the elements below the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,N).
- *
- * @param[in,out] A2
- * On entry, the M-by-N tile A2.
- * On exit, all the elements with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_ztsqrt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_ztsqrt;
- void (*callback)(void*) = options->profiling ? cl_ztsqrt_callback : NULL;
- CHAMELEON_starpu_ws_t *h_work = (CHAMELEON_starpu_ws_t*)(options->ws_host);
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_W(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &m, sizeof(int),
- STARPU_VALUE, &n, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_W, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn ),
- STARPU_VALUE, &ldt, sizeof(int),
- /* max( nb * (ib+1), ib * (ib+nb) ) */
- STARPU_SCRATCH, options->ws_worker,
- /* 2 * ib * (nb+ib) + nb */
- STARPU_VALUE, &h_work, sizeof(CHAMELEON_starpu_ws_t *),
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_USE_MPI)
- STARPU_EXECUTE_ON_NODE, A2->get_rankof(A2, A2m, A2n),
-#endif
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "ztsqrt",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_ztsqrt_cpu_func(void *descr[], void *cl_arg)
-{
- CHAMELEON_starpu_ws_t *h_work;
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU, *WORK;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- TAU= (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]); /* nb + ib*nb */
-
- starpu_codelet_unpack_args(cl_arg, &m, &n, &ib, &lda1, &lda2, &ldt, &h_work);
-
- WORK = TAU + chameleon_max( m, n );
- CORE_ztsqrt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS_CPU(ztsqrt, 4, cl_ztsqrt_cpu_func)
diff --git a/runtime/starpu/codelets/codelet_zttlqt.c b/runtime/starpu/codelets/codelet_zttlqt.c
deleted file mode 100644
index a673832082e5aaa466b31b53230f0f6421c77333..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_zttlqt.c
+++ /dev/null
@@ -1,173 +0,0 @@
-/**
- *
- * @file starpu/codelet_zttlqt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttlqt StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttlqt computes a LQ factorization of a rectangular matrix
- * formed by coupling side-by-side a complex M-by-M lower triangular tile A1
- * and a complex M-by-N lower triangular tile A2:
- *
- * | A1 A2 | = L * Q
- *
- * The tile Q is represented as a product of elementary reflectors
- *
- * Q = H(k)' . . . H(2)' H(1)', where k = min(M,N).
- *
- * Each H(i) has the form
- *
- * H(i) = I - tau * v * v'
- *
- * where tau is a complex scalar, and v is a complex vector with
- * v(1:i-1) = 0 and v(i) = 1; conjg(v(i+1:n)) is stored on exit in
- * A2(i,1:n), and tau in TAU(i).
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of rows of the tile A1 and A2. M >= 0.
- * The number of columns of the tile A1.
- *
- * @param[in] N
- * The number of columns of the tile A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M-by-M tile A1.
- * On exit, the elements on and below the diagonal of the array
- * contain the M-by-M lower trapezoidal tile L;
- * the elements above the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,N).
- *
- * @param[in,out] A2
- * On entry, the M-by-N lower triangular tile A2.
- * On exit, the elements on and below the diagonal of the array
- * with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the array A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[in,out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_zttlqt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_zttlqt;
- void (*callback)(void*) = options->profiling ? cl_zttlqt_callback : NULL;
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_W(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &m, sizeof(int),
- STARPU_VALUE, &n, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_W, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* nb * (ib+1) */
- STARPU_SCRATCH, options->ws_worker,
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "zttlqt",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_zttlqt_cpu_func(void *descr[], void *cl_arg)
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- TAU = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]); /* nb * (ib+1) */
-
- starpu_codelet_unpack_args(cl_arg, &m, &n, &ib, &lda1, &lda2, &ldt);
-
- WORK = TAU + chameleon_max( m, n );
-
- CORE_zttlqt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS_CPU(zttlqt, 4, cl_zttlqt_cpu_func)
diff --git a/runtime/starpu/codelets/codelet_zttmlq.c b/runtime/starpu/codelets/codelet_zttmlq.c
deleted file mode 100644
index c2924eafcef47724732584b52ab235208c578a78..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_zttmlq.c
+++ /dev/null
@@ -1,212 +0,0 @@
-/**
- *
- * @file starpu/codelet_zttmlq.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttmlq StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttmlq overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 (N1 == N2) with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 | * Q
- * | A2 | | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 | * Q**H
- * | A2 | | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k)
- *
- * as returned by CORE_zttqrt.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTTQRT in the first k rows of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[out] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size LDWORK-by-N1.
- *
- * @param[in] LDWORK
- * The dimension of the array WORK. LDWORK >= max(1,IB).
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_zttmlq(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_zttmlq;
- void (*callback)(void*) = options->profiling ? cl_zttmlq_callback : NULL;
- int ldwork = side == ChamLeft ? ib : nb;
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_R(V, Vm, Vn);
- CHAMELEON_ACCESS_R(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &side, sizeof(int),
- STARPU_VALUE, &trans, sizeof(int),
- STARPU_VALUE, &m1, sizeof(int),
- STARPU_VALUE, &n1, sizeof(int),
- STARPU_VALUE, &m2, sizeof(int),
- STARPU_VALUE, &n2, sizeof(int),
- STARPU_VALUE, &k, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_R, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn),
- STARPU_VALUE, &ldv, sizeof(int),
- STARPU_R, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* nb * ib */
- STARPU_SCRATCH, options->ws_worker,
- STARPU_VALUE, &ldwork, sizeof(int),
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "zttmlq",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_zttmlq_cpu_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]);
- WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* nb * ib */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- CORE_zttmlq(side, trans, m1, n1, m2, n2, k, ib, A1, lda1,
- A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS_CPU(zttmlq, 5, cl_zttmlq_cpu_func)
diff --git a/runtime/starpu/codelets/codelet_zttmqr.c b/runtime/starpu/codelets/codelet_zttmqr.c
deleted file mode 100644
index d485d16b9bcf582e1a86489612ad61b7724eb004..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_zttmqr.c
+++ /dev/null
@@ -1,269 +0,0 @@
-/**
- *
- * @file starpu/codelet_zttmqr.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttmqr StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttmqr overwrites the general complex M1-by-N1 tile A1 and
- * M2-by-N2 tile A2 (N1 == N2) with
- *
- * SIDE = 'L' SIDE = 'R'
- * TRANS = 'N': Q * | A1 | | A1 | * Q
- * | A2 | | A2 |
- *
- * TRANS = 'C': Q**H * | A1 | | A1 | * Q**H
- * | A2 | | A2 |
- *
- * where Q is a complex unitary matrix defined as the product of k
- * elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k)
- *
- * as returned by CORE_zttqrt.
- *
- *******************************************************************************
- *
- * @param[in] side
- * @arg ChamLeft : apply Q or Q**H from the Left;
- * @arg ChamRight : apply Q or Q**H from the Right.
- *
- * @param[in] trans
- * @arg ChamNoTrans : No transpose, apply Q;
- * @arg ChamConjTrans : ConjTranspose, apply Q**H.
- *
- * @param[in] M1
- * The number of rows of the tile A1. M1 >= 0.
- *
- * @param[in] N1
- * The number of columns of the tile A1. N1 >= 0.
- *
- * @param[in] M2
- * The number of rows of the tile A2. M2 >= 0.
- * M2 = M1 if side == ChamRight.
- *
- * @param[in] N2
- * The number of columns of the tile A2. N2 >= 0.
- * N2 = N1 if side == ChamLeft.
- *
- * @param[in] K
- * The number of elementary reflectors whose product defines
- * the matrix Q.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the M1-by-N1 tile A1.
- * On exit, A1 is overwritten by the application of Q.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,M1).
- *
- * @param[in,out] A2
- * On entry, the M2-by-N2 tile A2.
- * On exit, A2 is overwritten by the application of Q.
- *
- * @param[in] LDA2
- * The leading dimension of the tile A2. LDA2 >= max(1,M2).
- *
- * @param[in] V
- * The i-th row must contain the vector which defines the
- * elementary reflector H(i), for i = 1,2,...,k, as returned by
- * CORE_ZTTQRT in the first k columns of its array argument V.
- *
- * @param[in] LDV
- * The leading dimension of the array V. LDV >= max(1,K).
- *
- * @param[in] T
- * The IB-by-N1 triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] WORK
- * Workspace array of size
- * LDWORK-by-N1 if side == ChamLeft
- * LDWORK-by-IB if side == ChamRight
- *
- * @param[in] LDWORK
- * The leading dimension of the array WORK.
- * LDWORK >= max(1,IB) if side == ChamLeft
- * LDWORK >= max(1,M1) if side == ChamRight
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_zttmqr(const RUNTIME_option_t *options,
- cham_side_t side, cham_trans_t trans,
- int m1, int n1, int m2, int n2, int k, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *V, int Vm, int Vn, int ldv,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_zttmqr;
- void (*callback)(void*) = options->profiling ? cl_zttmqr_callback : NULL;
- int ldwork = side == ChamLeft ? ib : nb;
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_R(V, Vm, Vn);
- CHAMELEON_ACCESS_R(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &side, sizeof(int),
- STARPU_VALUE, &trans, sizeof(int),
- STARPU_VALUE, &m1, sizeof(int),
- STARPU_VALUE, &n1, sizeof(int),
- STARPU_VALUE, &m2, sizeof(int),
- STARPU_VALUE, &n2, sizeof(int),
- STARPU_VALUE, &k, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_R, RTBLKADDR(V, CHAMELEON_Complex64_t, Vm, Vn),
- STARPU_VALUE, &ldv, sizeof(int),
- STARPU_R, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* max( ib*nb, 2*ib*nb ) */
- STARPU_SCRATCH, options->ws_worker,
- STARPU_VALUE, &ldwork, sizeof(int),
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_USE_MPI)
- STARPU_EXECUTE_ON_NODE, A2->get_rankof(A2, A2m, A2n),
-#endif
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "zttmqr",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_zttmqr_cpu_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *V;
- int ldv;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *WORK;
- int ldwork;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]);
- WORK = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[4]); /* ib * nb */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- CORE_zttmqr(side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
-}
-
-#if defined(CHAMELEON_USE_CUDA)
-static void cl_zttmqr_cuda_func(void *descr[], void *cl_arg)
-{
- cham_side_t side;
- cham_trans_t trans;
- int m1;
- int n1;
- int m2;
- int n2;
- int k;
- int ib;
- cuDoubleComplex *A1;
- int lda1;
- cuDoubleComplex *A2;
- int lda2;
- cuDoubleComplex *V;
- int ldv;
- cuDoubleComplex *T;
- int ldt;
- cuDoubleComplex *W, *WC;
- int ldwork;
- int ldworkc;
-
- A1 = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[1]);
- V = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[2]);
- T = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[3]);
- W = (cuDoubleComplex *)STARPU_MATRIX_GET_PTR(descr[4]); /* 2*ib*nb */
-
- starpu_codelet_unpack_args(cl_arg, &side, &trans, &m1, &n1, &m2, &n2, &k, &ib,
- &lda1, &lda2, &ldv, &ldt, &ldwork);
-
- WC = W + ib * (side == ChamLeft ? m1 : n1);
- ldworkc = (side == ChamLeft) ? m2 : ib;
-
- RUNTIME_getStream(stream);
-
- CUDA_zttmqr(
- side, trans, m1, n1, m2, n2, k, ib,
- A1, lda1, A2, lda2, V, ldv, T, ldt,
- W, ldwork, WC, ldworkc, stream );
-
-#ifndef STARPU_CUDA_ASYNC
- cudaStreamSynchronize( stream );
-#endif
-}
-#endif /* defined(CHAMELEON_USE_CUDA) */
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS(zttmqr, 5, cl_zttmqr_cpu_func, cl_zttmqr_cuda_func, STARPU_CUDA_ASYNC)
diff --git a/runtime/starpu/codelets/codelet_zttqrt.c b/runtime/starpu/codelets/codelet_zttqrt.c
deleted file mode 100644
index 39d52185fdc68654e966b34db700b31733cde5c6..0000000000000000000000000000000000000000
--- a/runtime/starpu/codelets/codelet_zttqrt.c
+++ /dev/null
@@ -1,176 +0,0 @@
-/**
- *
- * @file starpu/codelet_zttqrt.c
- *
- * @copyright 2009-2014 The University of Tennessee and The University of
- * Tennessee Research Foundation. All rights reserved.
- * @copyright 2012-2018 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
- * Univ. Bordeaux. All rights reserved.
- *
- ***
- *
- * @brief Chameleon zttqrt StarPU codelet
- *
- * @version 1.0.0
- * @comment This file has been automatically generated
- * from Plasma 2.5.0 for CHAMELEON 1.0.0
- * @author Hatem Ltaief
- * @author Dulceneia Becker
- * @author Mathieu Faverge
- * @author Emmanuel Agullo
- * @author Cedric Castagnede
- * @date 2010-11-15
- * @precisions normal z -> c d s
- *
- */
-#include "chameleon_starpu.h"
-#include "runtime_codelet_z.h"
-
-/**
- *
- * @ingroup INSERT_TASK_Complex64_t
- *
- * CORE_zttqrt computes a QR factorization of a rectangular matrix
- * formed by coupling a complex N-by-N upper triangular tile A1
- * on top of a complex M-by-N upper trapezoidal tile A2:
- *
- * | A1 | = Q * R
- * | A2 |
- *
- * The tile Q is represented as a product of elementary reflectors
- *
- * Q = H(1) H(2) . . . H(k), where k = min(M,N).
- *
- * Each H(i) has the form
- *
- * H(i) = I - tau * v * v'
- *
- * where tau is a complex scalar, and v is a complex vector with
- * v(1:i-1) = 0 and v(i) = 1; v(i+1:m) is stored on exit in A2(1:m,i),
- * and tau in TAU(i).
- *
- *******************************************************************************
- *
- * @param[in] M
- * The number of rows of the tile A2. M >= 0.
- *
- * @param[in] N
- * The number of columns of the tile A1 and A2. N >= 0.
- *
- * @param[in] IB
- * The inner-blocking size. IB >= 0.
- *
- * @param[in,out] A1
- * On entry, the N-by-N tile A1.
- * On exit, the elements on and above the diagonal of the array
- * contain the N-by-N upper trapezoidal tile R;
- * the elements below the diagonal are not referenced.
- *
- * @param[in] LDA1
- * The leading dimension of the array A1. LDA1 >= max(1,N).
- *
- * @param[in,out] A2
- * On entry, the M-by-N upper triangular tile A2.
- * On exit, the elements on and above the diagonal of the array
- * with the array TAU, represent
- * the unitary tile Q as a product of elementary reflectors
- * (see Further Details).
- *
- * @param[in] LDA2
- * The leading dimension of the array A2. LDA2 >= max(1,M).
- *
- * @param[out] T
- * The IB-by-N triangular factor T of the block reflector.
- * T is upper triangular by block (economic storage);
- * The rest of the array is not referenced.
- *
- * @param[in] LDT
- * The leading dimension of the array T. LDT >= IB.
- *
- * @param[out] TAU
- * The scalar factors of the elementary reflectors (see Further
- * Details).
- *
- * @param[in,out] WORK
- *
- *******************************************************************************
- *
- * @return
- * \retval CHAMELEON_SUCCESS successful exit
- * \retval <0 if -i, the i-th argument had an illegal value
- *
- */
-
-void INSERT_TASK_zttqrt(const RUNTIME_option_t *options,
- int m, int n, int ib, int nb,
- const CHAM_desc_t *A1, int A1m, int A1n, int lda1,
- const CHAM_desc_t *A2, int A2m, int A2n, int lda2,
- const CHAM_desc_t *T, int Tm, int Tn, int ldt)
-{
- (void)nb;
- struct starpu_codelet *codelet = &cl_zttqrt;
- void (*callback)(void*) = options->profiling ? cl_zttqrt_callback : NULL;
-
- CHAMELEON_BEGIN_ACCESS_DECLARATION;
- CHAMELEON_ACCESS_RW(A1, A1m, A1n);
- CHAMELEON_ACCESS_RW(A2, A2m, A2n);
- CHAMELEON_ACCESS_W(T, Tm, Tn);
- CHAMELEON_END_ACCESS_DECLARATION;
-
- starpu_insert_task(
- starpu_mpi_codelet(codelet),
- STARPU_VALUE, &m, sizeof(int),
- STARPU_VALUE, &n, sizeof(int),
- STARPU_VALUE, &ib, sizeof(int),
- STARPU_RW, RTBLKADDR(A1, CHAMELEON_Complex64_t, A1m, A1n),
- STARPU_VALUE, &lda1, sizeof(int),
- STARPU_RW, RTBLKADDR(A2, CHAMELEON_Complex64_t, A2m, A2n),
- STARPU_VALUE, &lda2, sizeof(int),
- STARPU_W, RTBLKADDR(T, CHAMELEON_Complex64_t, Tm, Tn),
- STARPU_VALUE, &ldt, sizeof(int),
- /* nb * (ib+1) */
- STARPU_SCRATCH, options->ws_worker,
- STARPU_PRIORITY, options->priority,
- STARPU_CALLBACK, callback,
-#if defined(CHAMELEON_USE_MPI)
- STARPU_EXECUTE_ON_NODE, A2->get_rankof(A2, A2m, A2n),
-#endif
-#if defined(CHAMELEON_CODELETS_HAVE_NAME)
- STARPU_NAME, "zttqrt",
-#endif
- 0);
-}
-
-
-#if !defined(CHAMELEON_SIMULATION)
-static void cl_zttqrt_cpu_func(void *descr[], void *cl_arg)
-{
- int m;
- int n;
- int ib;
- CHAMELEON_Complex64_t *A1;
- int lda1;
- CHAMELEON_Complex64_t *A2;
- int lda2;
- CHAMELEON_Complex64_t *T;
- int ldt;
- CHAMELEON_Complex64_t *TAU;
- CHAMELEON_Complex64_t *WORK;
-
- A1 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[0]);
- A2 = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[1]);
- T = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[2]);
- TAU = (CHAMELEON_Complex64_t *)STARPU_MATRIX_GET_PTR(descr[3]); /* nb * (ib+1) */
-
- starpu_codelet_unpack_args(cl_arg, &m, &n, &ib, &lda1, &lda2, &ldt);
-
- WORK = TAU + chameleon_max( m, n );
-
- CORE_zttqrt(m, n, ib, A1, lda1, A2, lda2, T, ldt, TAU, WORK);
-}
-#endif /* !defined(CHAMELEON_SIMULATION) */
-
-/*
- * Codelet definition
- */
-CODELETS_CPU(zttqrt, 4, cl_zttqrt_cpu_func)
diff --git a/runtime/starpu/control/runtime_zlocality.c b/runtime/starpu/control/runtime_zlocality.c
index 0b2a7bc282dd6d39e852cb6d2679367d8b213f31..025e300c62775d79287593f316fb5671dea40345 100644
--- a/runtime/starpu/control/runtime_zlocality.c
+++ b/runtime/starpu/control/runtime_zlocality.c
@@ -15,7 +15,7 @@
* @author Cedric Augonnet
* @author Mathieu Faverge
* @author Cedric Castagnede
- * @date 2011-06-01
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -58,23 +58,15 @@ void RUNTIME_zlocality_allrestrict( uint32_t where )
/* QR */
cl_zgeqrt_restrict_where( where );
- cl_ztsqrt_restrict_where( where );
+ cl_ztpqrt_restrict_where( where );
cl_zunmqr_restrict_where( where );
- cl_ztsmqr_restrict_where( where );
-
- /* QR-RH */
-/* cl_zttqrt_restrict_where( where ); */
-/* cl_zttmqr_restrict_where( where ); */
+ cl_ztpmqrt_restrict_where( where );
/* LQ */
cl_zgelqt_restrict_where( where );
- cl_ztslqt_restrict_where( where );
+ cl_ztplqt_restrict_where( where );
cl_zunmlq_restrict_where( where );
- cl_ztsmlq_restrict_where( where );
-
- /* LQ-RH */
-/* cl_zttlqt_restrict_where( where ); */
-/* cl_zttmlq_restrict_where( where ); */
+ cl_ztpmlqt_restrict_where( where );
}
@@ -112,23 +104,15 @@ void RUNTIME_zlocality_onerestrict( cham_tasktype_t kernel, uint32_t where )
/* QR */
case TASK_GEQRT: cl_zgeqrt_restrict_where( where ); break;
+ case TASK_TPQRT: cl_ztpqrt_restrict_where( where ); break;
case TASK_UNMQR: cl_zunmqr_restrict_where( where ); break;
- case TASK_TSMQR: cl_ztsmqr_restrict_where( where ); break;
- case TASK_TSQRT: cl_ztsqrt_restrict_where( where ); break;
-
- /* QR-RH */
-/* case TASK_TTMQR: cl_zttmqr_restrict_where( where ); break; */
-/* case TASK_TTQRT: cl_zttqrt_restrict_where( where ); break; */
+ case TASK_TPMQRT: cl_ztpmqrt_restrict_where( where ); break;
/* LQ */
case TASK_GELQT: cl_zgelqt_restrict_where( where ); break;
+ case TASK_TPLQT: cl_ztplqt_restrict_where( where ); break;
case TASK_UNMLQ: cl_zunmlq_restrict_where( where ); break;
- case TASK_TSMLQ: cl_ztsmlq_restrict_where( where ); break;
- case TASK_TSLQT: cl_ztslqt_restrict_where( where ); break;
-
- /* LQ-RH */
-/* case TASK_TTMLQ: cl_zttmlq_restrict_where( where ); break; */
-/* case TASK_TTLQT: cl_zttlqt_restrict_where( where ); break; */
+ case TASK_TPMLQT: cl_ztpmlqt_restrict_where( where ); break;
default:
return;
@@ -167,23 +151,15 @@ void RUNTIME_zlocality_allrestore( )
/* QR */
cl_zgeqrt_restore_where();
- cl_ztsqrt_restore_where();
+ cl_ztpqrt_restore_where();
cl_zunmqr_restore_where();
- cl_ztsmqr_restore_where();
-
- /* QR-RH */
-/* cl_zttqrt_restore_where(); */
-/* cl_zttmqr_restore_where(); */
+ cl_ztpmqrt_restore_where();
/* LQ */
cl_zgelqt_restore_where();
- cl_ztslqt_restore_where();
+ cl_ztplqt_restore_where();
cl_zunmlq_restore_where();
- cl_ztsmlq_restore_where();
-
- /* LQ-RH */
-/* cl_zttlqt_restore_where(); */
-/* cl_zttmlq_restore_where(); */
+ cl_ztpmlqt_restore_where();
}
@@ -221,23 +197,15 @@ void RUNTIME_zlocality_onerestore( cham_tasktype_t kernel )
/* QR */
case TASK_GEQRT: cl_zgeqrt_restore_where(); break;
+ case TASK_TPQRT: cl_ztpqrt_restore_where(); break;
case TASK_UNMQR: cl_zunmqr_restore_where(); break;
- case TASK_TSMQR: cl_ztsmqr_restore_where(); break;
- case TASK_TSQRT: cl_ztsqrt_restore_where(); break;
-
- /* QR-RH */
-/* case TASK_TTMQR: cl_zttmqr_restore_where(); break; */
-/* case TASK_TTQRT: cl_zttqrt_restore_where(); break; */
+ case TASK_TPMQRT: cl_ztpmqrt_restore_where(); break;
/* LQ */
case TASK_GELQT: cl_zgelqt_restore_where(); break;
+ case TASK_TPLQT: cl_ztplqt_restore_where(); break;
case TASK_UNMLQ: cl_zunmlq_restore_where(); break;
- case TASK_TSMLQ: cl_ztsmlq_restore_where(); break;
- case TASK_TSLQT: cl_ztslqt_restore_where(); break;
-
- /* LQ-RH */
-/* case TASK_TTMLQ: cl_zttmlq_restore_where(); break; */
-/* case TASK_TTLQT: cl_zttlqt_restore_where(); break; */
+ case TASK_TPMLQT: cl_ztpmlqt_restore_where(); break;
default:
return;
diff --git a/runtime/starpu/control/runtime_zprofiling.c b/runtime/starpu/control/runtime_zprofiling.c
index 82af2b32836b8e3d6400f17cadd86b8f4f02de00..848746b7f5e4a9c64f060d06393983382907e0f1 100644
--- a/runtime/starpu/control/runtime_zprofiling.c
+++ b/runtime/starpu/control/runtime_zprofiling.c
@@ -15,7 +15,7 @@
* @author Cedric Augonnet
* @author Mathieu Faverge
* @author Cedric Castagnede
- * @date 2011-06-01
+ * @date 2018-11-08
* @precisions normal z -> s d c
*
*/
@@ -43,21 +43,18 @@ void RUNTIME_zdisplay_allprofile()
profiling_display_zgelqt_info();
profiling_display_zgeqrt_info();
profiling_display_zgessm_info();
- profiling_display_zgetrf_info();
profiling_display_zgetrf_incpiv_info();
+ profiling_display_zgetrf_info();
profiling_display_zgetrf_nopiv_info();
profiling_display_zlauum_info();
profiling_display_zpotrf_info();
profiling_display_zssssm_info();
+ profiling_display_ztplqt_info();
+ profiling_display_ztpmlqt_info();
+ profiling_display_ztpmqrt_info();
+ profiling_display_ztpqrt_info();
profiling_display_ztrtri_info();
- profiling_display_ztslqt_info();
- profiling_display_ztsmqr_info();
- profiling_display_ztsqrt_info();
profiling_display_ztstrf_info();
- profiling_display_zttlqt_info();
- profiling_display_zttmlq_info();
- profiling_display_zttmqr_info();
- profiling_display_zttqrt_info();
profiling_display_zunmlq_info();
profiling_display_zunmqr_info();
@@ -78,7 +75,7 @@ void RUNTIME_zdisplay_oneprofile( cham_tasktype_t kernel )
case TASK_SYMM: profiling_display_zsymm_info(); break;
case TASK_SYR2K: profiling_display_zsyr2k_info(); break;
case TASK_SYRK: profiling_display_zsyrk_info(); break;
- case TASK_TRMM: profiling_display_ztrmm_info(); break;
+ case TASK_TRMM: profiling_display_ztrmm_info(); break;
case TASK_TRSM: profiling_display_ztrsm_info(); break;
/* Lapack */
@@ -92,14 +89,13 @@ void RUNTIME_zdisplay_oneprofile( cham_tasktype_t kernel )
case TASK_POTRF: profiling_display_zpotrf_info(); break;
case TASK_SSSSM: profiling_display_zssssm_info(); break;
case TASK_TRTRI: profiling_display_ztrtri_info(); break;
- case TASK_TSLQT: profiling_display_ztslqt_info(); break;
- case TASK_TSMQR: profiling_display_ztsmqr_info(); break;
- case TASK_TSQRT: profiling_display_ztsqrt_info(); break;
case TASK_TSTRF: profiling_display_ztstrf_info(); break;
- case TASK_TTLQT: profiling_display_zttlqt_info(); break;
- case TASK_TTMLQ: profiling_display_zttmlq_info(); break;
- case TASK_TTMQR: profiling_display_zttmqr_info(); break;
- case TASK_TTQRT: profiling_display_zttqrt_info(); break;
+
+ case TASK_TPLQT: profiling_display_ztplqt_info(); break;
+ case TASK_TPMLQT: profiling_display_ztpmlqt_info(); break;
+ case TASK_TPMQRT: profiling_display_ztpmqrt_info(); break;
+ case TASK_TPQRT: profiling_display_ztpqrt_info(); break;
+
case TASK_UNMLQ: profiling_display_zunmlq_info(); break;
case TASK_UNMQR: profiling_display_zunmqr_info(); break;
diff --git a/runtime/starpu/include/runtime_codelet_z.h b/runtime/starpu/include/runtime_codelet_z.h
index 30b5f282db608afbb653daadf4b2c5159d2142be..509abacfc9a3dd9c2fd09729f8a7e7a351778476 100644
--- a/runtime/starpu/include/runtime_codelet_z.h
+++ b/runtime/starpu/include/runtime_codelet_z.h
@@ -15,7 +15,7 @@
* @author Cedric Augonnet
* @author Mathieu Faverge
* @author Cedric Castagnede
- * @date 2011-06-01
+ * @date 2018-11-08
* @precisions normal z -> c d s
*
*/
@@ -78,17 +78,9 @@ ZCODELETS_HEADER(tplqt)
ZCODELETS_HEADER(tpqrt)
ZCODELETS_HEADER(tpmlqt)
ZCODELETS_HEADER(tpmqrt)
-ZCODELETS_HEADER(tslqt)
-ZCODELETS_HEADER(tsmlq)
-ZCODELETS_HEADER(tsmqr)
ZCODELETS_HEADER(tsmlq_hetra1)
ZCODELETS_HEADER(tsmqr_hetra1)
-ZCODELETS_HEADER(tsqrt)
ZCODELETS_HEADER(tstrf)
-ZCODELETS_HEADER(ttlqt)
-ZCODELETS_HEADER(ttmlq)
-ZCODELETS_HEADER(ttmqr)
-ZCODELETS_HEADER(ttqrt)
ZCODELETS_HEADER(unmlq)
ZCODELETS_HEADER(unmqr)