Operating Systems 2014F Lecture 10

By using base 16 vs. base 10 - yes the alphabet is larger. when you talk about hex vs. decimal, you are changing how we read it. There's something more correct about using hex vs. base 10. A Hexadecimal digit represents 4 bits. How many bits does base 10 represent? (somewhere between 8 and 16 - a fractional bit) it's messy. Decimal is not the native base of a computer. Base 16 maps perfectly to base 2. That's why you also see octal - 3 bits vs. 4 bits. (but a reduced range), when you use hexadecimal it represents 4 bits, and is a bit cleaner.

When you say offset is a power of two, the page size is also a power of two. the storage used to encode the page size determines the range of the offset. It is a specific byte in a page. What is the range of bytes that can go in a page. Classic x86

what is the range of offsets of a 4 k page? The offsets are 0 - 4095. Another thing that you will need for the assignment, it's a question that refers to Peterson's algorithm. It doesn't work on modern machines. The big idea is the memory hierarchy. Key assumption: when one processor writes to memory the other processors will see that write, that value. Is that true in modern systems? No.

Concurrency Mechanisms:

locks condition variables Semaphores

you build all of these using the same mechanisms. Let's you check a value and change it in one atomic operation. The basic functionality of the hardware. The semantics are different in these higher level constructs. When do you use what? What are the purpose of these things? there's huge amounts of overlap. what you are talking about with these things, is that you have some sort of storage that has atomic operations so you can check it's value adn change it, all at one time (so you don't have race conditions when you manipulate the value) then you have a queue of threads.

The difference between these, is related to how you deal with the interaction between the storage and the queue of threads. IF you have a lock, grab lock, and release the lock, does releasing the lock always succeed? Yes, if you have the lock, and you release it, it always succeeds. grab lock can block. when we talk about blocking, it's going to sit there and wait for the condition to happen, nothing is going to happen to that thread until that condition is met. You often have a try lock, try lock means check to see if the lock is grabbable or not. it will grab the lock if it's available, but if it's not available it will return, it's non-blocking, because maybe you don't want to sit there and block. Most of the time you will try to grab the lock and if the lock fails, it will wait until grabbing the lock succeeds.

what if 5 processes try to grab the lock all at once, what do the rest do? they are all blocked, there are two ways for them to wait. One is to just go into an infinite loop, it just keeps trying trying trying until it succeeds. It is sometimes a good idea, this is called a spin lock. If the CPU with running the spin lock, doing nothing else, it is useless. Not very good for making use of efficient resources. You expect the time for you to wait is very small. no one is going to be using hte lock structure for very long. it's better to sit there and wait. It's like if you go to a gov't office, and you want something to happen, you will probably have to wait in line. When you are spin locking, you aresitting in the queue doing nothing. What you can also do instead is you can have a wait queue: in here we say, they queue up, they wait, you are freeing up the resources to do something else. the particular processor running the thread, it encounters things behind it, it puts that thread to sleep. what does sleep mean? Sleep means, the thread goes to sleep that implies something about the scheduling of the thread. IT is kicked off the cpu it's running on. I am not going to consume any resources right now, when you wake up, in biological terms, it's still consuming resources. It's still consuming space resources (still consuming ram, still living in cache but it is not being actively run by a core. When you say a thread yields, I don't need this CPU anymore. when you call sleep you are making a system call that says, I don't want to do anything for at least 10 seconds. You are taken out of the list of running processes, you are put on the list of block processes, and you wait for the timer to go off. Then your process is brought back in and is allowed to run. It depends on the time of the CPU. In general you should sleep at least for that amount of time.