Notes on Project 2

• Formatting, variable/function/method naming, and comments were greatly improved.
• Most groups had working virtual->physical mappings, page swappings, and forking. These kernels responded well to multiple, large processes running in a small main memory.
• Most kernels broke under stress, but usually only under reasonably severe stress (many processes, some forking, forcing some page swapping, and a memory with fewer page frames than the TLB had mappings.) That's not bad.
• There were frequent allocation/deallocation mistakes. These mistakes, unfortunately, tended not to cause crashes on the CS Suns. Rather, in spite of the errant pointers and prematurely deleted objects, the programs continued to run correct for most inputs. These errors turned up when run on different systems. When you see the comments regarding your errors, be sure you understand what they really mean.
• There was an overuse of hash tables in many places. Nobody lost credit for it, but hash tables are not an ``inexpensive'' data structure, in that they tend to require more space than other data structures. People used them as backing stores, and used complicated structures like a hash table of hash tables. Such a structure will require large amounts of memory, and aren't really appropriate for kernel programming.
• There was a common confusion between (or misuse of) page numbers and starting page addresses. Many groups would use a mask to set the offset bits of an address to zero, and then refer to it as a ``page number''. It is not a page number. It is the address of the first byte on that page. The page number would be obtained by shifting the virtual address down by nine bits. There are only 2²³ different page numbers. Be sure that you can see the difference between these two!