4 messages, 2 authors, 2001-08-27 · open the first message on its own page
From: Ralph Blach <hidden>
Date: 2001-08-27 11:38:45
Is there any support for having a special case for the 405 that would limit the number of processes to 255. This means that the pid and context could be the same. This might be a performance gain. Chip ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
From: Dan Malek <hidden>
Date: 2001-08-27 18:14:46
Ralph Blach wrote:
Is there any support for having a special case for the 405 that would limit the number of processes to 255. This means that the pid and context could be the same. This might be a performance gain.
The PID and context are already reused until there are more than 255 (or whatever the MMU can support) active processes. There isn't a special case needed, this is generic to all PowerPC with small context space resources. -- Dan ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
From: Ralph Blach <hidden>
Date: 2001-08-27 18:22:36
Dan Perhaps i dont understand. It looks like from the code that process id and pids dont match. Could you explain how this code works. Chip Dan Malek wrote:
Ralph Blach wrote:quoted
Is there any support for having a special case for the 405 that would limit the number of processes to 255. This means that the pid and context could be the same. This might be a performance gain.The PID and context are already reused until there are more than 255 (or whatever the MMU can support) active processes. There isn't a special case needed, this is generic to all PowerPC with small context space resources. -- Dan
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
From: Dan Malek <hidden>
Date: 2001-08-27 18:52:45
Ralph Blach wrote:
Perhaps i dont understand. It looks like from the code that process id and pids dont match. Could you explain how this code works.
The "contexts" are associated with an MM context, not with a process ID. Typically a process has one MM context, but there are cases (like multithreaded address spaces) that will share an MM context and all get the same TLB context. A "process" is somewhat blurry in Linux, but you can still think of it as an address space plus a processor thread. We are really concerned what happens to the address space and how that affects the MMU, not what the thread actually does, or how many share the address space. -- Dan ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/