Paper reading homework

Locks avoid race conditions but limit parallelism. In this lecture, we look at RCU, which can allow for more parallelism for certain kernel operations.

In preparation for lecture, read the paper RCU Usage in the Linux kernel: one decade later.

Figure 8 shows the performance of acquiring a read-write lock for reading. The read-write lock's implementation is similar to the one you implemented in the lab. Looking at your read_acuire implementation and the overhead shown in Figure 8 in the paper, what operation in your implementation becomes more expensive as more cores acquire the lock in parallel in read mode?

Submit your answer in an ASCII text file named homework.txt to the corresponding "Lecture N" assignment on Gradescope.

---

Questions or comments regarding 6.1810? Send e-mail to the course staff at 61810-staff@lists.csail.mit.edu.