Introducing other tests for rectangular matrices

tests/dbcsr_performance_multiply.F

@@ -559,10 +559,10 @@ SUBROUTINE perf_multiply(mp_group, mp_env, io_unit, &
       ! Row & column distributions
       row_dist_c => dbcsr_distribution_row_dist(dbcsr_distribution(matrix_c))
       col_dist_c => dbcsr_distribution_col_dist(dbcsr_distribution(matrix_c))
-      row_dist_a => row_dist_c
+      row_dist_a => dbcsr_distribution_row_dist(dbcsr_distribution(matrix_a))
       col_dist_a => dbcsr_distribution_col_dist(dbcsr_distribution(matrix_a))
       row_dist_b => dbcsr_distribution_row_dist(dbcsr_distribution(matrix_b))
-      col_dist_b => col_dist_c
+      col_dist_b => dbcsr_distribution_col_dist(dbcsr_distribution(matrix_b))
 
       CALL dbcsr_distribution_new(dist_a, mp_env, row_dist_a, col_dist_a)
       CALL dbcsr_distribution_new(dist_b, mp_env, row_dist_b, col_dist_b)
@@ -711,7 +711,7 @@ SUBROUTINE perf_multiply(mp_group, mp_env, io_unit, &
                   "  threshold = ", chksum_threshold
                chksum_err = .TRUE.
             ENDIF
-            IF (chksum_err) DBCSR_ABORT("Test failed!")
+            IF (chksum_err) DBCSR_ABORT("Wrong Checksums. Test failed!")
          ENDIF
          WRITE (io_unit, *)
          WRITE (io_unit, *)

tests/inputs/test_rect1_dense.perf

@@ -56,5 +56,5 @@ F
 # checksum (check, threshold, references)
 T
 0.1E-10
-0.627999017637509E+10
-0.239280442954758E+09
+0.675410189104774E+08
+0.247302643276505E+08

tests/inputs/test_rect1_sparse.perf

@@ -56,6 +56,5 @@ F
 # checksum (check, threshold, references)
 T
 0.1E-10
-0.885789976224201E+09
-0.847877708261093E+09
-
+0.515505547242877E+08
+0.173182002166628E+09

tests/inputs/test_rect2_dense.perf

@@ -0,0 +1,60 @@
+# npcols MPI grid, 0 leaves MPI to find the best grid.
+# Note that the total number of processors must be divisible per npcols
+1
+# operation
+dbcsr_multiply
+# matrix sizes (M, N, K)
+100
+100
+1000
+# sparsity (A, B, C)
+0.0d0
+0.0d0
+0.0d0
+# transposes
+T
+N
+# symmetries
+N
+N
+N
+# data type
+3
+# alpha (real, imag)
+1.0d0
+0.0d0
+# beta (real, imag)
+1.0d0
+0.0d0          
+# limits (0 means full size)
+# row
+0
+0
+# col
+0
+0
+# k
+0
+0
+# retain sparsity (T/F)
+F
+# number of repetitions
+1
+# number of different blocks to read (m, n, k)
+1
+1
+1
+# the m blocks (multiplicity, block size, ...)
+1
+5
+# the n blocks (multiplicity, block size, ...)
+1
+5
+# the k blocks (multiplicity, block size, ...)
+1
+5
+# checksum (check, threshold, references)
+T
+0.1E-10
+0.628267155865818E+09
+0.182181529654306E+08

tests/inputs/test_rect2_sparse.perf

@@ -0,0 +1,60 @@
+# npcols MPI grid, 0 leaves MPI to find the best grid.
+# Note that the total number of processors must be divisible per npcols
+1
+# operation
+dbcsr_multiply
+# matrix sizes (M, N, K)
+1000
+1000
+5000
+# sparsity (A, B, C)
+0.9d0
+0.9d0
+0.9d0
+# transposes
+T
+N
+# symmetries
+N
+N
+N
+# data type
+3
+# alpha (real, imag)
+1.0d0
+0.0d0
+# beta (real, imag)
+1.0d0
+0.0d0          
+# limits (0 means full size)
+# row
+0
+0
+# col
+0
+0
+# k
+0
+0
+# retain sparsity (T/F)
+F
+# number of repetitions
+1
+# number of different blocks to read (m, n, k)
+1
+1
+1
+# the m blocks (multiplicity, block size, ...)
+1
+5
+# the n blocks (multiplicity, block size, ...)
+1
+5
+# the k blocks (multiplicity, block size, ...)
+1
+5
+# checksum (check, threshold, references)
+T
+0.1E-10
+0.176967186312957E+09
+0.149098274728932E+09