java.io fix and speedup

[Mark Wielaard]

Hi,

On Sat, 2003-01-04 at 06:43, Jeff Sturm wrote:
> On 4 Jan 2003, Mark Wielaard wrote:
> > If the mmap patch from Jeff is useable in
> > general then that would give us much bigger improvements.
> 
> Below is the patch I used (against 3.2 branch).  It only attempts to
> mmap() for read access.  I don't know whether this is a good idea, versus
> waiting for a possible java.nio implementation.
> 
> (I got the idea from Per Bothner, in a message I can't find now.)

Nice idea, thanks. Attached is the patch against current CVS for people
that are interested in playing with it. In my tests there was a small
speedup but nothing spectacular. The system time dropped, but user time
was up a bit. I think the recent addition of the position field made
most of the difference for me. But maybe for some other workloads this
does make a big difference. If someone has I/O intensive benchmarks this
might be an interesting patch to try out.

Cheers,

Mark
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mmap.patch
Type: text/x-patch
Size: 6274 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/java-patches/attachments/20030118/44007816/attachment.bin>