mpi cholesky: Add flops information

mpi/examples/matrix_decomposition/mpi_cholesky_codelets.c

@@ -20,6 +20,64 @@
 #include <limits.h>
 #include <math.h>
 
+/* This is from magma
+
+  -- Innovative Computing Laboratory
+  -- Electrical Engineering and Computer Science Department
+  -- University of Tennessee
+  -- (C) Copyright 2009
+
+  Redistribution  and  use  in  source and binary forms, with or without
+  modification,  are  permitted  provided  that the following conditions
+  are met:
+
+  * Redistributions  of  source  code  must  retain  the above copyright
+    notice,  this  list  of  conditions  and  the  following  disclaimer.
+  * Redistributions  in  binary  form must reproduce the above copyright
+    notice,  this list of conditions and the following disclaimer in the
+    documentation  and/or other materials provided with the distribution.
+  * Neither  the  name of the University of Tennessee, Knoxville nor the
+    names of its contributors may be used to endorse or promote products
+    derived from this software without specific prior written permission.
+
+  THIS  SOFTWARE  IS  PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+  ``AS IS''  AND  ANY  EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+  LIMITED  TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+  A  PARTICULAR  PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+  HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+  SPECIAL,  EXEMPLARY,  OR  CONSEQUENTIAL  DAMAGES  (INCLUDING,  BUT NOT
+  LIMITED  TO,  PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+  DATA,  OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+  THEORY  OF  LIABILITY,  WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+  (INCLUDING  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+  OF  THIS  SOFTWARE,  EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+  */
+
+#define FMULS_POTRF(__n) ((double)(__n) * (((1. / 6.) * (double)(__n) + 0.5) * (double)(__n) + (1. / 3.)))
+#define FADDS_POTRF(__n) ((double)(__n) * (((1. / 6.) * (double)(__n)      ) * (double)(__n) - (1. / 6.)))
+
+#define FLOPS_SPOTRF(__n) (     FMULS_POTRF((__n)) +       FADDS_POTRF((__n)) )
+
+#define FMULS_TRMM_2(__m, __n) (0.5 * (double)(__n) * (double)(__m) * ((double)(__m)+1.))
+#define FADDS_TRMM_2(__m, __n) (0.5 * (double)(__n) * (double)(__m) * ((double)(__m)-1.))
+
+#define FMULS_TRMM(__m, __n) ( /*( (__side) == PlasmaLeft ) ? FMULS_TRMM_2((__m), (__n)) :*/ FMULS_TRMM_2((__n), (__m)) )
+#define FADDS_TRMM(__m, __n) ( /*( (__side) == PlasmaLeft ) ? FADDS_TRMM_2((__m), (__n)) :*/ FADDS_TRMM_2((__n), (__m)) )
+
+#define FMULS_TRSM FMULS_TRMM
+#define FADDS_TRSM FMULS_TRMM
+
+#define FLOPS_STRSM(__m, __n) (     FMULS_TRSM((__m), (__n)) +       FADDS_TRSM((__m), (__n)) )
+
+
+#define FMULS_GEMM(__m, __n, __k) ((double)(__m) * (double)(__n) * (double)(__k))
+#define FADDS_GEMM(__m, __n, __k) ((double)(__m) * (double)(__n) * (double)(__k))
+
+#define FLOPS_SGEMM(__m, __n, __k) (     FMULS_GEMM((__m), (__n), (__k)) +       FADDS_GEMM((__m), (__n), (__k)) )
+
+/* End of magma code */
+
 /*
  *	Create the codelets
  */
@@ -72,6 +130,7 @@ static void run_cholesky(starpu_data_handle_t **data_handles, int rank, int node
 {
 	unsigned k, m, n;
 	unsigned unbound_prio = STARPU_MAX_PRIO == INT_MAX && STARPU_MIN_PRIO == INT_MIN;
+	unsigned nn = size/nblocks;
 
 	for (k = 0; k < nblocks; k++)
 	{
@@ -80,6 +139,7 @@ static void run_cholesky(starpu_data_handle_t **data_handles, int rank, int node
 		starpu_mpi_task_insert(MPI_COMM_WORLD, &cl11,
 				       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k) : STARPU_MAX_PRIO,
 				       STARPU_RW, data_handles[k][k],
+				       STARPU_FLOPS, (double) FLOPS_SPOTRF(nn),
 				       0);
 
 		for (m = k+1; m<nblocks; m++)
@@ -88,6 +148,7 @@ static void run_cholesky(starpu_data_handle_t **data_handles, int rank, int node
 					       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k - m) : (m == k+1)?STARPU_MAX_PRIO:STARPU_DEFAULT_PRIO,
 					       STARPU_R, data_handles[k][k],
 					       STARPU_RW, data_handles[m][k],
+					       STARPU_FLOPS, (double) FLOPS_STRSM(nn, nn),
 					       0);
 
 			starpu_mpi_cache_flush(MPI_COMM_WORLD, data_handles[k][k]);
@@ -104,6 +165,7 @@ static void run_cholesky(starpu_data_handle_t **data_handles, int rank, int node
 						       STARPU_R, data_handles[n][k],
 						       STARPU_R, data_handles[m][k],
 						       STARPU_RW | STARPU_COMMUTE, data_handles[m][n],
+						       STARPU_FLOPS, (double) FLOPS_SGEMM(nn, nn, nn),
 						       0);
 			}
 
@@ -120,6 +182,7 @@ static void run_cholesky_column(starpu_data_handle_t **data_handles, int rank, i
 {
 	unsigned k, m, n;
 	unsigned unbound_prio = STARPU_MAX_PRIO == INT_MAX && STARPU_MIN_PRIO == INT_MIN;
+	unsigned nn = size/nblocks;
 
 	/* Column */
 	for (n = 0; n<nblocks; n++)
@@ -137,6 +200,7 @@ static void run_cholesky_column(starpu_data_handle_t **data_handles, int rank, i
 						       STARPU_R, data_handles[n][k],
 						       STARPU_R, data_handles[m][k],
 						       STARPU_RW | STARPU_COMMUTE, data_handles[m][n],
+						       STARPU_FLOPS, (double) FLOPS_SGEMM(nn, nn, nn),
 						       0);
 			}
 			k = n;
@@ -147,6 +211,7 @@ static void run_cholesky_column(starpu_data_handle_t **data_handles, int rank, i
 						       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k - m) : (m == k+1)?STARPU_MAX_PRIO:STARPU_DEFAULT_PRIO,
 						       STARPU_R, data_handles[k][k],
 						       STARPU_RW, data_handles[m][k],
+						       STARPU_FLOPS, (double) FLOPS_STRSM(nn, nn),
 						       0);
 			}
 			else
@@ -155,6 +220,7 @@ static void run_cholesky_column(starpu_data_handle_t **data_handles, int rank, i
 				starpu_mpi_task_insert(MPI_COMM_WORLD, &cl11,
 						       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k) : STARPU_MAX_PRIO,
 						       STARPU_RW, data_handles[k][k],
+						       STARPU_FLOPS, (double) FLOPS_SPOTRF(nn),
 						       0);
 			}
 		}
@@ -175,6 +241,7 @@ static void run_cholesky_antidiagonal(starpu_data_handle_t **data_handles, int r
 	unsigned a, c;
 	unsigned k, m, n;
 	unsigned unbound_prio = STARPU_MAX_PRIO == INT_MAX && STARPU_MIN_PRIO == INT_MIN;
+	unsigned nn = size/nblocks;
 
 	/* double-antidiagonal number:
 	 * - a=0 contains (0,0) plus (1,0)
@@ -205,6 +272,7 @@ static void run_cholesky_antidiagonal(starpu_data_handle_t **data_handles, int r
 						       STARPU_R, data_handles[n][k],
 						       STARPU_R, data_handles[m][k],
 						       STARPU_RW | STARPU_COMMUTE, data_handles[m][n],
+						       STARPU_FLOPS, (double) FLOPS_SGEMM(nn, nn, nn),
 						       0);
 			}
 
@@ -216,6 +284,7 @@ static void run_cholesky_antidiagonal(starpu_data_handle_t **data_handles, int r
 						       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k - m) : (m == k+1)?STARPU_MAX_PRIO:STARPU_DEFAULT_PRIO,
 						       STARPU_R, data_handles[k][k],
 						       STARPU_RW, data_handles[m][k],
+						       STARPU_FLOPS, (double) FLOPS_STRSM(nn, nn),
 						       0);
 			}
 			else
@@ -224,6 +293,7 @@ static void run_cholesky_antidiagonal(starpu_data_handle_t **data_handles, int r
 				starpu_mpi_task_insert(MPI_COMM_WORLD, &cl11,
 						       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k) : STARPU_MAX_PRIO,
 						       STARPU_RW, data_handles[k][k],
+						       STARPU_FLOPS, (double) FLOPS_SPOTRF(nn),
 						       0);
 			}
 		}
@@ -246,6 +316,7 @@ static void run_cholesky_antidiagonal(starpu_data_handle_t **data_handles, int r
 						       STARPU_R, data_handles[n][k],
 						       STARPU_R, data_handles[m][k],
 						       STARPU_RW | STARPU_COMMUTE, data_handles[m][n],
+						       STARPU_FLOPS, (double) FLOPS_SGEMM(nn, nn, nn),
 						       0);
 			}
 			/* non-diagonal block, solve */
@@ -254,6 +325,7 @@ static void run_cholesky_antidiagonal(starpu_data_handle_t **data_handles, int r
 					       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k - m) : (m == k+1)?STARPU_MAX_PRIO:STARPU_DEFAULT_PRIO,
 					       STARPU_R, data_handles[k][k],
 					       STARPU_RW, data_handles[m][k],
+					       STARPU_FLOPS, (double) FLOPS_STRSM(nn, nn),
 					       0);
 		}
 
@@ -273,6 +345,7 @@ static void run_cholesky_prio(starpu_data_handle_t **data_handles, int rank, int
 	unsigned a;
 	int k, m, n;
 	unsigned unbound_prio = STARPU_MAX_PRIO == INT_MAX && STARPU_MIN_PRIO == INT_MIN;
+	unsigned nn = size/nblocks;
 
 	/*
 	 * This is basically similar to above, except that we shift k according to the priorities set in the algorithm, so that prio ~ 2*a or 2*a+1
@@ -301,6 +374,7 @@ static void run_cholesky_prio(starpu_data_handle_t **data_handles, int rank, int
 				starpu_mpi_task_insert(MPI_COMM_WORLD, &cl11,
 						       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k) : STARPU_MAX_PRIO,
 						       STARPU_RW, data_handles[k][k],
+						       STARPU_FLOPS, (double) FLOPS_SPOTRF(nn),
 						       0);
 			}
 			else
@@ -310,6 +384,7 @@ static void run_cholesky_prio(starpu_data_handle_t **data_handles, int rank, int
 						       STARPU_PRIORITY, noprio ? STARPU_DEFAULT_PRIO : unbound_prio ? (int)(2*nblocks - 2*k - m) : (m == k+1)?STARPU_MAX_PRIO:STARPU_DEFAULT_PRIO,
 						       STARPU_R, data_handles[k][k],
 						       STARPU_RW, data_handles[m][k],
+						       STARPU_FLOPS, (double) FLOPS_STRSM(nn, nn),
 						       0);
 			}
 
@@ -327,6 +402,7 @@ static void run_cholesky_prio(starpu_data_handle_t **data_handles, int rank, int
 							       STARPU_R, data_handles[n][k],
 							       STARPU_R, data_handles[m][k],
 							       STARPU_RW | STARPU_COMMUTE, data_handles[m][n],
+							       STARPU_FLOPS, (double) FLOPS_SGEMM(nn, nn, nn),
 							       0);
 				}
 			}