On Monday 12 July 2004 10:58, Anand wrote: > Hi Folks, > I have few doubts on how memory is managed in Multi Processor > Systems(SMP)? > Right now Linux uses 3G:1G split for a uniprocessor (Though patches > are there for 32G) > > 1) I was jus wondering how the 1G memory is managed on a SMP? > Suppose i have a 1Gig RAM and 3 procesors (32 bit),Is the 1 Gig > memory split among the 3 processors ?? No the memory is not split between the different CPUs. I don't believe ther is any significant difference in MM for SMP vs non-SMP unless you are on a NUMA architecture. > Since each processor can address 4G memory ,Do they have large > Virtual Address space? > > 2) How is the process switching taken care on SMP? > (ie) on a UP when when a process time slice is over ,its context is > saved in the PSW or the task_struct! > and next time it gets its timeslice , it loads the info form the > task_struct, > but on a SMP, Assume i have 3 procesors ,In the first time slice the > Process runs in Processor P1, > Is it possible for the process to run in Processor P2 when it gets it > time slice back ? > If possibe how is it taken care? It is possible for a process to switch to a different CPU. However, the scheduler tries to avoid that. > 3)Does the Linux scheduler diffrenciates between a Process and a > thread ? All i can see in the scheduler code is a the task > represented by task_struct, which is getting scheduled, > From which i assume ,that the scheduler schedules only the Process > (coz task_struct represents a Process :-)!!) > But how is the user space threads (which is part of a Process),Kernel > treads scheduled ? A thread is just a process that shares a mm with other tasks. -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
