Evan Weinberg

Results 46 issues of Evan Weinberg

Reminder for self: add documentation/clean up the routines constructHostCloverField and loadCloverQuda. I'm making this as a note to myself because I lost a good deal of time trying to understand...

clean-up
documentation

We should add a ctest which covers at least the most commonly used solvers, sweeping over (as appropriate for the compile): * Wilson- and staggered-type fermions. * All mixed precision...

feature
clean-up

As a step towards developing MG-accelerated staggered RHMC we should create a `staggered_multigrid_evolve_test` in the spirit of `multigrid_evolve_test`. This should be straightforward on first pass (HMC only). Adding support for...

feature

The MILC HISQ MG interface currently implements its own functions for parsing MG param input files in the (hideous) spirit of the old command line interface before CLI11 was added....

feature
clean-up

The MILC HISQ MG interface currently hard codes a handful of default values for the various MG-related structures. This is largely redundant of the `setStaggeredMGInvertParam` and `setStaggeredMultigridParam` functions in `utils/set_params.cpp`....

feature
clean-up

In `lib/interface_quda.cpp`, the routines `loadGaugeQuda` and `loadSloppyGaugeQuda`, while distinct, do a lot of redundant work. They both contain logic to create the sloppy, preconditioned, and refinement gauge fields. One easy...

clean-up

This draft PR demonstrates an interface to allocate and utilize static MPI comms buffers for the `gather`/`scatter` routines in `Cabana_Halo.hpp`, as opposed to reallocating a new buffer with each call....

performance

Right now, there are a few issues (in my opinion) with the Coulomb gauge fixing convergence criteria, both in QUDA and analogously in MILC. This manifested itself in testing offloading...

bug
question
clean-up

Currently, the `t_boundary` member of the `GaugeField` class is serving double-duty. * In the case of applying/removing staggered phases, it keeps track of what boundary conditions the field had _when...

clean-up

As the title notes, the function `polarSu3` in `su3_project.cuh` uses a `quda::sincos` call here: https://github.com/lattice/quda/blob/develop/include/su3_project.cuh#L106 This could be replaced with a `sincospi`, improving performance and reducing/eliminating the stack frame in...

clean-up
optimization