• 0 Posts
  • 115 Comments
Joined 2 years ago
Cake day: June 22nd, 2023
  • wewbull@feddit.uktoProgrammer Humor@programming.dev•Coding chessEnglish
    751·
    4 days ago

    At university I had an introductory C course where one assignment was to write a program that searched a 4x4 array of booleans for groups of cells set to true. Groups had to be rectangles, powers of 2 in width and height, and could wrap (i.e. they could go off the right edge and back on the left edge). We had to submit our programs by e-mail and printed form one week later. The prof. marked the paper versions and the TA ran and tested the digital. One slight problem, if you used the university owned printers, they charged for print outs. A few pence per page to cover costs and stop people abusing the rather nice high quality printers the computer faculty had.

    I’d always enjoyed programming and whilst C was new to me, using another language wasn’t a big problem. As I worked on it I realised the problem wasn’t as straightforward as I first thought, but I spent a few hours on it that evening and had a solution I was happy with.

    Penny was a student on the course whose approach to academia was memorization. She didn’t consume, process, and apply concepts. She just remembered them. Her favourite subject was maths. While the rest of us were struggling to derive some formula, she’d have just committed the process to memory.

    Penny was complaining a lot on this programming assignment. She didn’t understand why the assignment was so hard for an introductory class. I didn’t judge. I know some people find programming hard, but I didn’t feel I could help her much without jeopardising my own mark. There’s only so much uniqueness in a small program and if she just copied my solution we’d both get penalised for plagiarism. I did mention to her the cases I’d found tricky to get right was when two groups overlapped. If one group completely covered a smaller one you’d only report the bigger one, but if not you’d report both groups.

    I heard, through her boyfriend, that that week had involved many long evenings working on this assignment, but she turned up at the next class solution in hand. Obviously stressed, she carried a pile of paper of several hundred pages. She had written a program that consisted of an if-statement for every possible group size and location. About a hundred different possible groups. Each condition written with constant value indices into the array. To cope with the overlapping groups problem, checks for smaller groups also checked that no larger group also covered this area. No loops. No search algorithm. Just a linear program of if-statements.

    Apparently debugging this has been a nightmare. Cut and paste errors everywhere, but when I’d told her about overlapping groups aspect it had blown her mind. There always seemed to be a combination she hadn’t accounted for. Multiple times she thought she was done, only to find a corner case she’d missed. And just to kick her when she was down, she’d paid for multiple printouts, each one costing about £10 only to find a problem afterwards.

    This consistent A grade student who sailed through everything by relying on her memory had been broken by being asked to create an algorithm rather than remember one. She got credit for submitting a solution that compiled and solved some cases, but I doubt the professor got past the first page of that huge printout.

    Penny had worked really hard for that D.

  • wewbull@feddit.uktoLinux@lemmy.ml•Is there a way to monitor memory bandwidth utilisation?English
    3·
    1 month ago

    Depends on two things

    1. What performance counters your CPU has.
    2. What you consider to be a memory load.

    Tools like perf on Linux can get you access to your processors performance counters and you’ll be able to see how many “events” occur while a process is running.

    What’s an event? Well they can be configured to monitor all sorts of things in the CPU. Instructions executed, Interrupts, page table misses, and on some loads / stores.

    Memory systems on a CPU aren’t straightforward though. They contain multiple levels of cache, each of which reduces the number of accesses which go to the next layer. So depending on which level you measure, you’ll get different numbers.