On 1 Sep 2015, at 16:01, Comer Duncan comer.duncan@gmail.com wrote:
I have a thorn I am constructing from a Kranc script. it compiles ok. However, upon looking at its schedule.inc file there is a section in which apparently doubling up occurs. Here is the section:
schedule formgk AT CCTK_INITIAL before CT_MultiLevel { LANG: C SYNC: CT_g READS: grid::x(Everywhere) READS: grid::y(Everywhere) READS: grid::z(Everywhere) READS: grid::x(Everywhere) READS: grid::y(Everywhere) READS: grid::z(Everywhere) WRITES: CT_BrillAnalytic::g11(Interior) WRITES: CT_BrillAnalytic::g12(Interior) WRITES: CT_BrillAnalytic::g13(Interior) WRITES: CT_BrillAnalytic::g22(Interior) WRITES: CT_BrillAnalytic::g23(Interior) WRITES: CT_BrillAnalytic::g33(Interior) } "formgk"
My question is how come there is a doubling of a READS grid::x(Everywhere) along with the same for y and z? I don't see any problem in the script which obviously induces such behavior. So, does this indicate that there is a problem somewhere or should I ignore the doubling of the READS for x, y, and z? This script prepares initial data so this would not occur repeatedly.
Hi Comer,
This looks like a bug in Kranc. Different parts of Kranc are probably contributing accesses to the coordinate variables, and Kranc is not de-duplicating them. Since Cactus/Carpet currently don't look at these entries, it should be harmless for the moment. I have created an issue on Kranc's issue tracker: https://github.com/ianhinder/Kranc/issues/131. Would you be OK with sending me the script which has this problem, so I can attach it to the issue? Note that this would make the script publicly-available. If this is not OK, you can just send it to me privately.