On 8/8/2012 9:14 AM, David Mertens wrote:
Interesting. I had no idea that memory mapping and pthreads were so
closely linked.
Sorry, the pthreads work was mainly me and Rob
with some fixes to the existing code where invalid
assumptions were being made (which broke for win32
pthreads---of course).
--Chris
David
On Wed, Aug 8, 2012 at 5:41 AM, chm <[email protected]> wrote:
On 8/7/2012 11:35 PM, chm wrote:
On 8/7/2012 11:04 AM, David Mertens wrote:
On Tue, Aug 7, 2012 at 9:08 AM, Chris Marshall
<[email protected]> wrote:
From the PDL-2.4.10 Release_Notes:
* PDL now supports pthreads on all platforms as well
as a new, auto-parallelization capability for PDL
threadloops across multiple processors using the
PDL::ParallelCPU module.
--Chris
Ah, sorry Chris. Some day I will stop saying "this might not work on
Windows," and with my luck, that day I will be wrong. :-)
Did you implement this on Windows? That seems like it would have been
pretty darn hard!
I have to give credit to Leon Timmermans' File::Map
which made this possible. His cross-platform
implementation of mmap semantics and the bug
fixes needed for the PDL implementation were key
to making the new functionality possible.
Also, I believe you were the one who originally
researched the possibility for some sort of
win32 mmap implementation to allow the PDL
memory mapping to work. :-)
Of course, it won't be fully buildable until we
get some sort of Alien::Pthreads to make sure
that the windows installs have the win32 pthread
library available....
--Chris
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
