Update fucntion declarations and cuda kernel

control/compute_z.h

@@ -107,7 +107,7 @@ void chameleon_pzsyrk(cham_uplo_t uplo, cham_trans_t trans, CHAMELEON_Complex64_
 void chameleon_pzsyr2k(cham_uplo_t uplo, cham_trans_t trans, CHAMELEON_Complex64_t alpha, CHAM_desc_t *A, CHAM_desc_t *B, CHAMELEON_Complex64_t beta, CHAM_desc_t *C, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
 void chameleon_pzsytrf(cham_uplo_t uplo, CHAM_desc_t *A, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
 void chameleon_pztile2band(cham_uplo_t uplo, CHAM_desc_t *A, CHAM_desc_t *descAB, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
-void chameleon_pztpgqrt( int genD, int L, CHAM_desc_t *V1, CHAM_desc_t *T1, CHAM_desc_t *V2, CHAM_desc_t *T2, CHAM_desc_t *Q1, CHAM_desc_t *Q2, CHAM_desc_t *D, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request );
+void chameleon_pztpgqrt( int kt, int L, CHAM_desc_t *V2, CHAM_desc_t *T2, CHAM_desc_t *Q1, CHAM_desc_t *Q2, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request );
 void chameleon_pztpqrt( int L, CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *T, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request );
 void chameleon_pztradd(cham_uplo_t uplo, cham_trans_t trans, CHAMELEON_Complex64_t alpha, CHAM_desc_t *A, CHAMELEON_Complex64_t beta, CHAM_desc_t *B, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
 void chameleon_pztrmm(cham_side_t side, cham_uplo_t uplo, cham_trans_t transA, cham_diag_t diag, CHAMELEON_Complex64_t alpha, CHAM_desc_t *A, CHAM_desc_t *B, RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
@@ -130,7 +130,8 @@ void chameleon_pzbuild( cham_uplo_t uplo, CHAM_desc_t *A, void *user_data, void*
 
 void chameleon_pzgelqf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
                               RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
-void chameleon_pzgeqrf_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
+void chameleon_pzgeqrf_param( int genD, int K, const libhqr_tree_t *qrtree,
+                              CHAM_desc_t *A, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
                               RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
 void chameleon_pzunmlq_param( int genD, const libhqr_tree_t *qrtree, cham_side_t side, cham_trans_t trans,
                               CHAM_desc_t *A, CHAM_desc_t *B, CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
@@ -141,7 +142,8 @@ void chameleon_pzunmqr_param( int genD, const libhqr_tree_t *qrtree, cham_side_t
 void chameleon_pzunglq_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t *A, CHAM_desc_t *Q,
                               CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
                               RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
-void chameleon_pzungqr_param( int genD, const libhqr_tree_t *qrtree, CHAM_desc_t *A, CHAM_desc_t *Q,
+void chameleon_pzungqr_param( int genD, int K, const libhqr_tree_t *qrtree,
+                              CHAM_desc_t *A, CHAM_desc_t *Q,
                               CHAM_desc_t *TS, CHAM_desc_t *TT, CHAM_desc_t *D,
                               RUNTIME_sequence_t *sequence, RUNTIME_request_t *request);
 

cudablas/compute/cuda_ztpmqrt.c

@@ -47,7 +47,7 @@ CUDA_ztpmqrt( cham_side_t side, cham_trans_t trans,
     else {
         m1 = M;
         n1 = K;
-        ldwork  = m1;
+        ldwork = chameleon_max( K, chameleon_max( M, N ) );
         ldworkc = IB;
         ws = ldwork * IB;
     }