|
Test
Your EQ #147— Answer
|
Answer
6
Each process
has to keep track of its own threads. It has to allocate
stack space for each one. It has to have a program counter
and stack pointer for each one. It also has to manage
the registers associated with these values, filling in
the appropriate values for the thread it wants to execute.
Finally, it needs to have a timer (provided by the OS)
to wake up the scheduler thread, and that thread does
context switching among the other threads as appropriate.
Possible problems include:
- Scheduling
is not a trivial thing to program. Having user programs
do this means that they could easily mess it up and
fail to work at all. However, user-level thread software
libraries are available to do this for you.
- Because
the OS doesn't know that the threads are running,
one thread blocking on an I/O operation will appear
to the OS that the whole process is blocked, and will
remain blocked until the I/O completes. Even if other
threads could run, they will not.
- The
kernel will schedule the processes based on their
priorities. High-priority threads within a low-priority
process will be scheduled with low priority relative
to threads in other processes.
- Overhead:
all of the thread scheduling takes time. But the context
switches are cheap.
Contributor:
Naveen PN
Published: October-2002
