AFAIK, Dekker's algorithm provides a strict turn based solution. It assumes that the two processes are always competing for the resource.
After one process has completed using the resource it will not be able to make use of the resource again until the other process has used it atleast once. This is done so as to satisfy the progress requirement.
So if there are two processes and if one process requires use of the resource
aggresively and if the other process uses the resource rarely then Dekker's
algorithm may not be the best solution to ensure performance.
I guess all processors should have some kind of 'Test and Set' instructions which is required to implement locks in SMP.
john
On 05 Jan 2004 10:35:55 +0200, Momchil Velikov <velco@fadata.bg> wrote:
"pankaj" == pankaj chauhan <chauhan_ait@yahoo.co.in> writes:
pankaj> hi all , pankaj> i have a small doubt : Is it possible to have pankaj> synchronization between processes running on different pankaj> processors in a SMP without any harware support for pankaj> that. I mean can we implement mutual exclusion in pankaj> access of shared data without any 'lock ' prefix , or pankaj> atomic_swap , or atomic test_and_set instructions in pankaj> the instruction set of the machine. If yes then How ??
Search for Dekker's algorithm.
~velco
-- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
-- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
