[Gnu-arch-users] Re: tla library-add - high cpu

[John Goerzen]

Tom Lord <address@hidden> writes:

> Rob is running on a 64-bit machine.  Patching is a fairly
> regexp-intensive operation (due to computing inventories -- those
> =tagging-method regexps).  The default DFA and NFA cache sizes of Rx
> is measured in bytes -- but allocated in multiples of the machine's
> word-size.  So, in effect, the cache size was cut in half on a 64 bit
> machine.  Word has it that just bumping the cache size fixed the
> problem.

I can confirm this.  tla, and hackerlab's build process (especially
regardining unidata) were *painful* on Alpha -- seriously.  My Alpha
is only 600MHz, but still, it's not THAT slow.

I bumped up both defaults from 1MB to 2MB and bumped up the default in
the unidata generator as well.  Doing that made a tremendous
difference.  Processing that unidata.txt file literally went from
about 5 minutes to 10 seconds or so.  An improvement of several orders
of magnitude also appeared for several tla commands.

Perhaps the cache size should just be increased, or specified in terms
of the size of data stored there on each platform?  (ie, x * sizeof(foo))

-- John