On Sun, Dec 21, 1997 at 01:00:51PM -0600, Miguel de Icaza wrote:

quoted

minor side effects of polishing the cache routines ...

  Ralf

What exactly did the polishing do?  I saw that we are down in a couple
of tests.

Most of these values where we went down aren't stable across several
runs.  The improoved values for the exec & sh test however were pretty
stable.  The optimization was getting rid of flushing the second level
cache in r4k_flush_page_to_ram_d32i32().

  Ralf