Private Repo for collaboration and work in EC527 labs and project
| Taks | Subtask | status | notes |
|---|---|---|---|
| B&O 5.3 | |||
| B&O 6.5 | |||
| B&O 6.6 | |||
| B&O Blocking | |||
| Part 1 | run code, compute data, interpret on spreadsheet | ||
| Note | current loops in main need to free vecs | ||
| need to make a vec_free function | |||
| Part 2 | Optimize MMM using loop interchange | ||
| Part 3 | MMM with blocking | ||
| Part 4 | Matrix transpose with loop interchange and blocking | ||
| This whole assigment seems far more messy than the first unfortunately |
proposed for vec_free(), will test
void vec_free(vec_ptr vec) //return 1 for good, 0 for fail { free(vec->data); free(vec); }
Check list for where a given portion may be
| Task | Subtasks | Status | Notes |
|---|---|---|---|
| B&0 Reading | J:Started | ||
| Part 1 | use test_combine1-7.c | ||
| a | compile, run, graph CPE, tinker, compare with book | ||
| b | doubles & multiply, unroll loop by up to 10, graph CPE, explain | ||
| c | same as b, but multiple accumulators and reassoc. transform | ||
| submit | modified code, tables, graphs, explanations | ||
| Part 2 | create test_dot.c, seems to require part 1b | ||
| submit | code with original dot & best versions, describe optimizations, CPE and graphs for both | ||
| Part 3 | test_branch.c | ||
| finish coding each init_vectors to show data dependence | |||
| generate CPE graphs | |||
| submit | modified code & description, CPE plots wrt different data, explanation of CPE; possibly assembly level | ||
| Part 4 | make test_eval.c, problem 5.20 from B&O | ||
| submit | decription of methods, code, plots and CPEs of original and best, justify performance maybe assembly | ||
| Part 5 | complain somehow |