f_fermion_eo broken? (Defect #327)


Added by Christopher Pinke almost 5 years ago. Updated almost 5 years ago.


Status:Done Start date:22 Oct 2012
Priority:High Due date:
Assignee:Christopher Pinke % Done:

100%

Category:-
Target version:-

Description

At cdf48d884ad7aa9d3bd068382ee0ec00af45b296 the f_fermion_eo kernel tests all pass. However, test 7 and 8 give the same result, what seems really strange since they should access different gauge links. This has to be investigated!


Related issues

blocks CL2QCD - Unit Test #325: test opencl_module_hmc Done 22 Oct 2012

History

Updated by Christopher Pinke almost 5 years ago

  • Status changed from New to In Progress

Updated by Christopher Pinke almost 5 years ago

Tests 7 and 8 give the same result just because independently of Even or Odd as an argument, all gauge links are visited during the kernel's evaluation. Because the tests have cold input vectors, this then gives the same result.

With c8023b7fdbdb67bbdb213a2cde01a2a7ff646e0c I restructured the test so that each test uses the same rnd numbers. This then results in different results from tests 9 and 10, which is the same situation as above but with the same rnd input vectors.

To conclude, the kernel is not broken.

However, tests for nontrivial BC and chemical potential should be added.

  • % Done changed from 0 to 80

Updated by Christopher Pinke almost 5 years ago

Commit:76541d49b87822aa95b41d63044f7d8ec0a8662 gives tests for chemical potential and nontrivial BCs.

  • Status changed from In Progress to Done
  • % Done changed from 80 to 100

Also available in: Atom PDF