The following five files are useful for running regressions with
ACL2(p):

  parallel-full.lisp
  parallel-resource-based.lisp
  parallel-top-level.lisp
  pseudo-parallel.lisp
  serial.lisp

For example:

(time nice make -j 8 regression-fresh ACL2=/projects/acl2/devel/ccl-saved_acl2p ACL2_CUSTOMIZATION=/projects/acl2/devel/acl2-customization-files/parallel-full.lisp) >& logs/make-regression-par-ccl-j-8-feb27.log&

Note: Sometimes parallelism is defeated or modified for individual
books.  To find out where:

  cd books
  grep -l acl2-par `find . -name "*.acl2"`

For example, hints/basic-tests.acl2 has this:

  #+acl2-par
  ; computed hints that modify state
  (set-waterfall-parallelism nil)

Here is a note relevant to ACL2(p) from David Rager.

  Some of the books will perform better under :resource-based waterfall
  parallelism than :full waterfall parallelism.  This is because these
  books have so many subgoals and such a strange dependency tree between
  those subgoals that the waterfall parallelism resources become
  exhausted and the proofs must execute serially at times to maintain a
  stable system.  As such, it may be good to override any :full setting
  that a customization file provides and use :resource-based waterfall
  parallelism instead -- just for these few books.  Note that even
  without overriding the waterfall parallelism setting, the system will
  remain stable.  Changing the waterfall parallelism mode is just a
  performance optimization.  As of April 2012, the books to which this
  statement may apply are the following:

  models/jvm/m5/apprentice.lisp
  coi/termination/assuming/complex.lisp
  concurrent-programs/bakery/stutter2.lisp
