search

sa2c / HiRep

HiRep repository
GNU General Public License v2.0
0 stars 4 forks source link

Spn smearing #2

Open theHoHsiao opened 4 months ago

theHoHsiao commented 4 months ago

We implemented the Wuppertal smearing and APE smearing for both connected and disconnected measurements.

fzierler commented 3 months ago

Hi, as discussed with Paul, I found that the code here is actually slower than the one developed on top of the old HiRep code (using the old complex numbers). Specifically, I tested this code here and compared it to the one at https://github.com/fzierler/hirep-smear on DiaL3 for the connected measurements on a 64x16^3 lattice using the same input file and same flags in the Makefile. I find that the code here is about 50% slower. I don't know what could have caused this behaviour.

fzierler commented 2 months ago

Hi, I also tried to test this code for a source position different from (0,0,0,0). However, this leads to a failure of the Inverter on some of the configurations. Typically, this depends on the MPI partitioning and the specific machine (or maybe compiler). Further testing shows that this appears even without smearing in the conventional code for point sources.

I suggest we remove the option to place the source at an arbitrary point for the smearing code.

fzierler commented 1 month ago

Hi, currently this code does not measure the disconnected pieces because a function call to this measurement was lost, when I ported the code to the recent version of HiRep. I have fixed this by now and created PR for Paul's branch here: https://github.com/theHoHsiao/HiRep_smear/pull/1