usqcd-software/qopqdp

SciDAC QFT Operations library built on QDP
Latest commit b189c83 Jan 17, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
config Nov 18, 2007
examples Oct 5, 2018
include Feb 23, 2018
lib Jan 17, 2019
.gitignore Mar 21, 2015
AUTHORS Dec 15, 2006
COPYING Aug 27, 2005
ChangeLog Aug 29, 2011
INSTALL Nov 18, 2007
Makefile.am Oct 26, 2013
NEWS Aug 8, 2017
configure.ac Oct 5, 2018

```QOPQDP -- SciDAC Level 3 over QDP/C

===============================================================================
------------------------------------------------------
--- below are some notes on the asqtad conventions ---
--- which are also similar to the wilson case --------
------------------------------------------------------

The dslash is accessed by:

float mass,
QOP_F3_ColorVector *out,
QOP_F3_ColorVector *in,
QOP_evenodd_t eo_out,
QOP_evenodd_t eo_in);

where specifying different values of eo_out and eo_in will give the
corresponding parts of D.  If both are QOP_EVENODD then you get the
full matrix, both QOP_EVEN multiplies by just D_ee = m, and you can
also do things like eo_out=QOP_EVEN and eo_in=QOP_EVENODD.

The most important thing to note is that the inverter conventions have
now changed.  The result is now always of the form

out_subset = (D^-1)_subset in_subset

where subset is the one specified in the QOP_invert_arg_t and D is the
conventional Dslash.  In even-odd blocks this is

D = ( m     D_eo )
( D_oe  m    )

D^-1 = ( m A^-1      -D_eo B^-1 )
( -D_oe A^-1  m B^-1     )

with A = (m^2 - D_eo D_oe) and B = (m^2 - D_oe D_eo)

so specifying QOP_EVEN will now multiply by m/A instead of 1/(4A) as
it did before.  Thus the result will have to be scaled by 1/(4m) to
get the previous answer.  The main reason for this change is
consistency since now specifying QOP_EVENODD will give the full
inverse with the correct normalization which is very handy for writing
simple analysis codes.

Also starting in version 0.10.0 the residual returned is divided by the
norm of the input vector.
===============================================================================
*** Wilson Clover ***

Conventions for passing the Wilson clover term as raw data can be found in
lib/wilson_dslash_p.c
===============================================================================
```