[m-rev.] diff: disable GC during startup

[Julien Fischer]

On Fri, 6 May 2011, Peter Wang wrote:

> On 2011-05-05, Peter Wang <novalazy at gmail.com> wrote:
>> [On my machine, this reduced the run time of an empty Mercury program in
>> the hlc.gc grade from 2.7 ms to 1.1 ms.  The equivalent C program takes
>> 0.26 ms.  The Mercury libraries were statically linked.  Measured with a
>> simple C program which does fork/exec/wait in a loop to avoid shell
>> overhead.  CPU frequency scaling was disabled.]
>
> For an unknown reason, this change doesn't only affect the startup time.
> I timed the compilation of most compiler/*.m modules, with and without
> this change.  Most modules took less time to compile, but some took
> longer.

Presumably one of the effects of your change is to shift when
collections occur.  Possibly they are just occurring at more fortuitous
spots in the case where you get a speedup?

Julien.
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------