Consider a RAM organized in blocks. There are multiple processes running on the system. Every application gets below information.
(Thread T, Memory Block M, time t, R/W) which essentially tells that the thread T was using memory block M at time t and operation could be read or write.
Memory conflict is defined as –
– Multiple read operations at the same location are not cause of conflict.
– One write operation between x+5 to x-5 to location M, will be cause of conflict for a thread accessing location M at time x where x is some time in standard unit of time measurement.
– Example – If thread T1 accessed memory location M at time x+1 and if a thread T2 accesses location M before time x+6, then T1 and T2 are candidate of conflict given one of them does write operation.
You are given with the list of threads accessing memory locations, you have to find all conflicts.
Input: (1, 512, 1, R) (2, 432, 2, W) (3, 512, 3, R) (4, 932, 4, R) (5, 512, 5, W) (6, 932, 6, R) (7, 835, 7, R) (8, 432, 8, R) Output: Thread 1 & 3 conflict with thread 5 All other operations are safe.
We strongly recommend you to minimize your browser and try this yourself first.
The idea is to sort all threads by memory block and if memory block is same, then by time. Once we have all threads sorted, we can traverse all threads one by one. For every thread being traversed, we simply need to check previous adjacent threads of same block as threads are sorted by time.
Below is C++ implementation of this idea.
threads 3 and 5 conflict. threads 1 and 5 conflict. All other operations are same
Time complexity: The above solution uses sorting to sort threads. Sorting can be done in O(nLogn) time. Then it prints all conflicts. Printing all conflicts takes O(n + m) time where m is number of conflicts. So overall time complexity is O(nLogn + m).