Download COMPUT~1 and more Slides Computer Science in PDF only on Docsity!
Module 20: Multi-core Computing Multi-processor Scheduling
Lecture 40: Multi-core Computing Synchronization
The Lecture Contains:
Process Dispatching
Process Scheduling
Thread Scheduling
Anatomy of Downloading Window
The Producer-Consumer Problem
What is Wrong?
A Possible Scenario
Race Conditions
Critical Section Problem
Problem Abstraction
Module 20: Multi-core Computing Multi-processor Scheduling
Lecture 40: Multi-core Computing Synchronization
Process Dispatching
After assignment, deciding who is selected from among the pool of waiting processes Process dispatching. Single processor multiprogramming strategies may be counter-productive here. Priorities and process history may not be sufficient.
Process Scheduling
Single queue of processes or if multiple priority is used, multiple priority queues, all feeding into a common pool of processors. Multi-server queuing model: multiple-queue/single queue, multiple server system. Inference: Specific scheduling policy does not have much effect as the number of processors increase. Conclusion: Use FCFS with priority levels.
Module 20: Multi-core Computing Multi-processor Scheduling
Lecture 40: Multi-core Computing Synchronization
Anatomy of Downloading Window
The Producer-Consumer Problem
Produce(item_t item) { while (count==bufsz); buffer[in]=item; in=(in+1)%bufsz; count=count+1; }
Consume(item_t *item) { while(count==0); *item=buffer[out]; out=(out+1)%bufsz; count=count-1; } Shared Variables buffer, count
Module 20: Multi-core Computing Multi-processor Scheduling
Lecture 40: Multi-core Computing Synchronization
What is Wrong?
Variable count is shared. Both process read and modify this variable The assembly code for these two statement may be something like following:
//count=count+1 //count=count- P1: load R1,count C1: load R1,count P2: add R1,1 C2: sub R1, P3: store count,R1 C3: store count,R
Two processes run concurrently (Single or multiple CPUs)
A Possible Scenario
The producer and consumer processes may be scheduled in any order and may be preempted. Consider the following sequence of statements P1, CSwitch , C1, C2, C3, CSwitch , P2, P3.
Module 20: Multi-core Computing Multi-processor Scheduling
Lecture 40: Multi-core Computing Synchronization
Race Conditions
Situation when several concurrent processes operate on the same variable and the result of the computation depends upon the order in which they execut. In the preceding example, the value of count could be 4, 5 or 6. C1,P1,P2,P3,C2,C3 final value 4 C1,C2,C3,P1,P2,P3 final value 5 P1,C1,C2,C3,P2,P3 final value 6
Critical Section Problem
We model the problem using the notion of Critical Sections. Critical sections are with respect to the shared data. There are n processes, each sharing some common resource. Each wants to modify the common resource. For each process, define the region of code where it accesses a shared piece of data as a critical section. In a correct system of multiple cooperating processes, Only one process must be inside its critical section.
Problem Abstraction
Processes execute code similar to the following.
while (1) { Non critical code Entry code Critical Section Code Exit Code Non Critical Code }
