Remove unused variables, duplicate code

src/C-interface/bml_parallel.c

@@ -322,6 +322,7 @@ bml_mpi_recv_matrix(
             break;
         default:
             LOG_ERROR("unknown matrix type\n");
+            return NULL;
             break;
     }
 }

src/C-interface/csr/bml_allocate_csr.c

@@ -21,8 +21,6 @@ csr_noinit_table(
     int i, lwr;
     const int alloc_size =
         INIT_SLOT_STORAGE_SIZE >= tsize ? INIT_SLOT_STORAGE_SIZE : tsize;
-    static int powers[] =
-        { 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 };
     static int powersof2[] =
         { 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768,
         65536, 131072, 262144

src/C-interface/csr/bml_normalize_csr.c

@@ -30,6 +30,7 @@ bml_accumulate_offdiag_csr(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/csr/bml_parallel_csr.c

@@ -144,6 +144,7 @@ bml_mpi_recv_matrix_csr(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/csr/bml_threshold_csr_typed.c

@@ -89,7 +89,6 @@ void TYPED_FUNC(
         const int annz = A->data_[i]->NNZ_;
         for (int pos = 0; pos < annz; pos++)
         {
-            const int j = cols[pos];
             if (is_above_threshold(vals[pos], threshold))
             {
                 if (rlen < pos)

src/C-interface/dense/bml_allocate_dense.c

@@ -338,7 +338,6 @@ bml_update_domain_dense(
 {
     bml_domain_t *A_domain = A->domain;
 
-    int myRank = bml_getMyRank();
     int nprocs = bml_getNRanks();
 
     for (int i = 0; i < nprocs; i++)

src/C-interface/dense/bml_normalize_dense.c

@@ -33,6 +33,7 @@ bml_accumulate_offdiag_dense(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/dense/bml_parallel_dense.c

@@ -208,6 +208,7 @@ bml_mpi_recv_matrix_dense(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/dense/bml_parallel_dense_typed.c

@@ -33,8 +33,6 @@ void TYPED_FUNC(
     bml_matrix_dense_t * A)
 {
 #ifdef BML_USE_MPI
-    int myRank = bml_getMyRank();
-
     int N = A->N;
 
     REAL_T *A_matrix = A->matrix;

src/C-interface/dense/bml_utilities_dense_typed.c

@@ -33,7 +33,10 @@ void TYPED_FUNC(
     char header1[20], header2[20], header3[20], header4[20], header5[20];
     int hdimx, nnz, irow, icol;
     int values_read;
-    double real_part, imaginary_part;
+    double real_part;
+#if (defined(SINGLE_COMPLEX) || defined(DOUBLE_COMPLEX))
+    double imaginary_part;
+#endif
 
     int N = A->N;
 

src/C-interface/distributed2d/bml_allocate_distributed2d.c

@@ -12,8 +12,6 @@
 
 /* MPI communicator for all the distributed2d matrices */
 static MPI_Comm s_comm = MPI_COMM_NULL;
-static MPI_Comm r_comm = MPI_COMM_NULL;
-static MPI_Comm c_comm = MPI_COMM_NULL;
 
 void
 bml_setcomm_distributed2d(

src/C-interface/distributed2d/bml_diagonalize_distributed2d_typed.c

@@ -220,8 +220,8 @@ void TYPED_FUNC(
     int ione = 1;
     int izero = 0;
     int np0 = NUMROC(&m, &mb, &my_prow, &izero, &np_rows);
-    int nq0 = NUMROC(&m, &mb, &my_pcol, &izero, &np_cols);
 #if defined(SINGLE_REAL) || defined(DOUBLE_REAL)
+    int nq0 = NUMROC(&m, &mb, &my_pcol, &izero, &np_cols);
     int lwork = MAX(1 + 6 * m + 2 * np0 * nq0,
                     3 * m + MAX(mb * (np0 + 1), 3 * mb)) + 2 * m;
 #else

src/C-interface/ellblock/bml_normalize_ellblock.c

@@ -30,6 +30,7 @@ bml_accumulate_offdiag_ellblock(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/ellblock/bml_parallel_ellblock.c

@@ -146,6 +146,7 @@ bml_mpi_recv_matrix_ellblock(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/ellblock/bml_threshold_ellblock_typed.c

@@ -105,8 +105,6 @@ void TYPED_FUNC(
     int *A_nnzb = A->nnzb;
     int *bsize = A->bsize;
 
-    int rlen;
-
     for (int ib = 0; ib < NB; ib++)
     {
         for (int jp = 0; jp < A_nnzb[ib]; jp++)

src/C-interface/ellpack/bml_add_ellpack_typed.c

@@ -239,8 +239,6 @@ double TYPED_FUNC(
     int *B_nnz = B->nnz;
     int *B_index = B->index;
 
-    int ind, ind2;
-
     REAL_T *A_value = (REAL_T *) A->value;
     REAL_T *B_value = (REAL_T *) B->value;
 

src/C-interface/ellpack/bml_adjungate_triangle_ellpack_typed.c

@@ -34,7 +34,6 @@ void TYPED_FUNC(
     int A_M = A->M;
     int l;
     int ll;
-    int j;
     REAL_T *A_value = (REAL_T *) A->value;
     int *A_index = A->index;
     int *A_nnz = A->nnz;

src/C-interface/ellpack/bml_allocate_ellpack.c

@@ -328,7 +328,6 @@ bml_update_domain_ellpack(
 {
     bml_domain_t *A_domain = A->domain;
 
-    int myRank = bml_getMyRank();
     int nprocs = bml_getNRanks();
 
     for (int i = 0; i < nprocs; i++)

src/C-interface/ellpack/bml_allocate_ellpack_typed.c

@@ -123,7 +123,7 @@ void TYPED_FUNC(
 #else
     memset(A->nnz, 0, A->N * sizeof(int));
     memset(A->index, 0, A->N * A->M * sizeof(int));
-    memset(A->value, 0.0, A->N * A->M * sizeof(REAL_T));
+    memset(A_value, 0.0, A->N * A->M * sizeof(REAL_T));
 #endif
 
 #endif // conditional for offload
@@ -226,12 +226,12 @@ bml_matrix_ellpack_t *TYPED_FUNC(
     A->csrRowPtr = bml_allocate_memory(sizeof(int) * (N + 1));
     A->csrVal = bml_allocate_memory(sizeof(REAL_T) * N * M);
 #endif
-    REAL_T *A_value = A->value;
 
     A->domain = bml_default_domain(N, M, distrib_mode);
     A->domain2 = bml_default_domain(N, M, distrib_mode);
 
 #if defined(USE_OMP_OFFLOAD)
+    REAL_T *A_value = A->value;
     int *A_nnz = A->nnz;
     int *A_index = A->index;
     int NM = N * M;

src/C-interface/ellpack/bml_getters_ellpack_typed.c

@@ -46,7 +46,7 @@ void *TYPED_FUNC(
     }
     for (int j_index = 0; j_index < A->nnz[i]; j_index++)
     {
-        if (A->index[ROWMAJOR(i, j_index, A->N, A->M)] == j)
+        if (A_index[ROWMAJOR(i, j_index, A->N, A->M)] == j)
         {
             return &A_value[ROWMAJOR(i, j_index, A->N, A->M)];
         }

src/C-interface/ellpack/bml_introspection_ellpack_typed.c

@@ -28,8 +28,6 @@ double TYPED_FUNC(
     double threshold)
 {
     int nnzs = 0;
-    int i;
-    int j;
     double sparsity;
     REAL_T *A_value = (REAL_T *) A->value;
     int A_N = A->N;

src/C-interface/ellpack/bml_parallel_ellpack.c

@@ -207,6 +207,7 @@ bml_mpi_recv_matrix_ellpack(
 #endif
         default:
             LOG_ERROR("unknown precision\n");
+            return NULL;
             break;
     }
 }

src/C-interface/ellpack/bml_parallel_ellpack_typed.c

@@ -31,8 +31,9 @@ void TYPED_FUNC(
     bml_allGatherVParallel_ellpack) (
     bml_matrix_ellpack_t * A)
 {
-    int myRank = bml_getMyRank();
-    int nRanks = bml_getNRanks();
+#ifdef BML_USE_MPI
+    //int myRank = bml_getMyRank();
+    //int nRanks = bml_getNRanks();
 
     int N = A->N;
     int M = A->M;
@@ -43,7 +44,6 @@ void TYPED_FUNC(
 
     REAL_T *A_value = (REAL_T *) A->value;
 
-#ifdef BML_USE_MPI
 /*
     for (int i = 0; i < nRanks; i++)
     {

src/C-interface/ellpack/bml_scale_ellpack_typed.c

@@ -27,47 +27,13 @@
  */
 bml_matrix_ellpack_t *TYPED_FUNC(
     bml_scale_ellpack_new) (
-    void *_scale_factor,
+    void *scale_factor,
     bml_matrix_ellpack_t * A)
 {
-    REAL_T *scale_factor = _scale_factor;
     bml_matrix_ellpack_t *B = TYPED_FUNC(bml_copy_ellpack_new) (A);
 
-    REAL_T *B_value = B->value;
-    int myRank = bml_getMyRank();
-    int nElems = B->domain->localRowExtent[myRank] * B->M;
-    int startIndex = B->domain->localDispl[myRank];
-    int inc = 1;
-
-    int N = A->N;
-    int M = A->M;
-
-    int *A_nnz = A->nnz;
-    int *A_index = A->index;
-    REAL_T *A_value = A->value;
-
-    int *B_nnz = B->nnz;
-    int *B_index = B->index;
-    REAL_T scale = *scale_factor;
-#ifdef USE_OMP_OFFLOAD
-#pragma omp target teams distribute parallel for collapse(2)
-    for (int i = 0; i < N; i++)
-    {
-        for (int j = 0; j < M; j++)
-        {
-            B_value[ROWMAJOR(i, j, M, N)] =
-                scale * A_value[ROWMAJOR(i, j, M, N)];
-        }
-    }
-#else // offload conditional
+    TYPED_FUNC(bml_scale_ellpack)(scale_factor, A, B);
 
-#ifdef NOBLAS
-    LOG_ERROR("No BLAS library");
-#else
-    C_BLAS(SCAL) (&nElems, scale_factor, &(B_value[startIndex]), &inc);
-#endif
-
-#endif // offload conditional
     return B;
 }
 
@@ -84,43 +50,38 @@ void TYPED_FUNC(
     bml_matrix_ellpack_t * A,
     bml_matrix_ellpack_t * B)
 {
+    // copy necessary so that B has the same structure as A
     if (A != B)
     {
         TYPED_FUNC(bml_copy_ellpack) (A, B);
     }
 
     REAL_T *scale_factor = _scale_factor;
     REAL_T *B_value = B->value;
-    int myRank = bml_getMyRank();
-    int nElems = B->domain->localRowExtent[myRank] * B->M;
-    int startIndex = B->domain->localDispl[myRank];
-    int inc = 1;
 
     int N = A->N;
     int M = A->M;
 
-    int *A_nnz = A->nnz;
-    int *A_index = A->index;
-    REAL_T *A_value = A->value;
-
-    int *B_nnz = B->nnz;
-    int *B_index = B->index;
-    REAL_T scale = *scale_factor;
 #ifdef USE_OMP_OFFLOAD
+    REAL_T scale = *scale_factor;
 #pragma omp target teams distribute parallel for collapse(2)
     for (int i = 0; i < N; i++)
     {
         for (int j = 0; j < M; j++)
         {
             B_value[ROWMAJOR(i, j, M, N)] =
-                scale * A_value[ROWMAJOR(i, j, M, N)];
+                scale * B_value[ROWMAJOR(i, j, M, N)];
         }
     }
 #else // offload conditional
 
 #ifdef NOBLAS
     LOG_ERROR("No BLAS library");
 #else
+    int myRank = bml_getMyRank();
+    int nElems = B->domain->localRowExtent[myRank] * B->M;
+    int startIndex = B->domain->localDispl[myRank];
+    int inc = 1;
     C_BLAS(SCAL) (&nElems, scale_factor, &(B_value[startIndex]), &inc);
 #endif
 
@@ -134,17 +95,11 @@ void TYPED_FUNC(
 {
     REAL_T *scale_factor = _scale_factor;
     REAL_T *A_value = A->value;
-    int myRank = bml_getMyRank();
-    int number_elements = A->domain->localRowExtent[myRank] * A->M;
-    int startIndex = A->domain->localDispl[myRank];
-    int inc = 1;
 
 #ifdef USE_OMP_OFFLOAD
     int N = A->N;
     int M = A->M;
 
-    int *A_nnz = A->nnz;
-    int *A_index = A->index;
     REAL_T scale = *scale_factor;
     size_t MbyN = N * M;
 #pragma omp target teams distribute parallel for map(to:MbyN,scale)
@@ -157,6 +112,10 @@ void TYPED_FUNC(
 #ifdef NOBLAS
     LOG_ERROR("No BLAS library");
 #else
+    int myRank = bml_getMyRank();
+    int number_elements = A->domain->localRowExtent[myRank] * A->M;
+    int startIndex = A->domain->localDispl[myRank];
+    int inc = 1;
     C_BLAS(SCAL) (&number_elements, scale_factor, &(A_value[startIndex]),
                   &inc);
 #endif

src/C-interface/ellpack/bml_setters_ellpack_typed.c

@@ -32,15 +32,12 @@ void TYPED_FUNC(
 {
     int A_N = A->N;
     int A_M = A->M;
-    int l;
-    int ll;
 
     REAL_T *A_value = (REAL_T *) A->value;
     int *A_index = A->index;
     int *A_nnz = A->nnz;
 
 #ifdef USE_OMP_OFFLOAD
-//#pragma omp target
 #pragma omp target update from(A_nnz[:A_N], A_index[:A_N*A_M], A_value[:A_N*A_M])
 #endif
     A_value[ROWMAJOR(i, A_nnz[i], A_N, A_M)] = *((REAL_T *) element);

src/C-interface/ellpack/bml_trace_ellpack_typed.c

@@ -44,9 +44,9 @@ double TYPED_FUNC(
     int myRank = bml_getMyRank();
     int rowMin = A_localRowMin[myRank];
     int rowMax = A_localRowMax[myRank];
-    int numrows = rowMax - rowMin;
 
 #ifdef USE_OMP_OFFLOAD
+    int numrows = rowMax - rowMin;
     REAL_T *diag;
     diag = (REAL_T *) calloc(numrows, sizeof(REAL_T));
 #pragma omp target enter data map(to:diag[:numrows])
@@ -130,10 +130,10 @@ double TYPED_FUNC(
     int B_N = B->N;
     int B_M = B->M;
 
+#ifdef USE_OMP_OFFLOAD
     REAL_T *B_value = (REAL_T *) B->value;
     int *B_index = (int *) B->index;
     int *B_nnz = (int *) B->nnz;
-#ifdef USE_OMP_OFFLOAD
 #pragma omp target update from(A_nnz[:A_N], A_index[:A_N*A_M], A_value[:A_N*A_M])
 #pragma omp target update from(B_nnz[:B_N], B_index[:B_N*B_M], B_value[:B_N*B_M])
 #endif