From 9214d37cb22ac28fc64c152dd6b222481dbcb351 Mon Sep 17 00:00:00 2001 From: Eugeniy Mikhailov Date: Tue, 23 Aug 2011 13:05:52 -0400 Subject: Added note about time execution estimate --- fortran/navy_four_levels/README | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) (limited to 'fortran') diff --git a/fortran/navy_four_levels/README b/fortran/navy_four_levels/README index 42c7e18..0f30a3a 100644 --- a/fortran/navy_four_levels/README +++ b/fortran/navy_four_levels/README @@ -1,4 +1,20 @@ -* Eugeniy is grumpy on 07/10/2007: +* Eugeniy is grumpy on 2011/07/10: Gosh, this is July 2011 and we are modifying Fortran 77 version to be Fortran 95. Is there a sense? + +* Eugeniy on 2011/08/23: +Navy chose very bad algorithm. +It seems to be a good idea to put normilezed dt=dz to avoid some unnesesary calculations +on each step. But results are devastating since code counter intuitively runs longer for +a shorter cell, and not surprisingly proportionally to time interval. + +Execution time proportional to the total number of grid nodes which can be calculated as + +Ngrid = tptz*zpts = (tmax-tmin)*Gamma_super / ((zmax*Gamma_super/(c*zpts)) *zpts + +Ngrid = (tmax-tmin)*c/zmax * zpts^2 + +where 'c' is speed of light, 'zmax' is length of the cell, and 'zpts' is number +of desired points along z axis. + -- cgit v1.2.3