Final Capstone: Submission/Report
The problem, from first principles
02-apply-optimization-measure-improvement produced raw benchmark numbers for ONE optimization attempt. A real capstone submission needs to turn that into an honest, structured verdict, and, crucially, needs a rule for what actually counts as a PASSING submission when comparing several candidate optimizations: a big speedup on broken code, or a correct-but-slower result, must never be treated as a win.
From theory to code
Implement build_capstone_report (turning one optimization's raw stats into a structured pass/fail verdict) and compare_capstone_reports (picking the best PASSING submission out of several).
Constraints
build_capstone_report(project_name, baseline_stats, optimized_stats, correctness_verified)computesspeedup_factor = baseline_stats["median"] / optimized_stats["median"], andpassed = correctness_verified and speedup_factor > 1.0.- Returns a dict with
"project_name","speedup_factor","correctness_verified","passed", and a one-line"summary"string mentioning the project name. compare_capstone_reports(reports)returns the highest-speedup_factorreport AMONG ONLY the ones withpassed=True; if none pass, returns{"best_project": None, "best_speedup_factor": None, "num_passing": 0}.
Hints
Open one at a time. Each gives away a little more than the last.
Hint 1
passed requires BOTH conditions at once (correctness_verified AND speedup_factor > 1.0): a correct-but-not-actually-faster submission fails just as much as an incorrect-but-fast one does.
Hint 2
compare_capstone_reports should FILTER to passed=True reports first, THEN find the max by speedup_factor among only those survivors: never compare a failing report's speedup against a passing one's.
Keywords
No results yet
Run your code to check it against the examples, or submit to run every test.