Imports

26.2–26.3. Multithreaded Algorithms (Historical 2_4 Compatibility)

The historical 2_4 module name is retained for compatibility. Its main text now covers §§26.2–26.3: parallel-recurrence definitions and all-input analysis for P-MATMUL, P-MERGE, and P-MERGE-SORT. The historical parallel Strassen names are imported separately from an explicitly labeled Chapter 26 extension.

Main results

The imported modules provide an executable value/work/span layer for attaching costs directly to algorithm results, executable and proved-correct P-ADD and P-MATMUL algorithms, exact equalities between their recorded costs and their execution recurrences, monotonicity and adjacent-power sandwiches for all four matrix costs, power-of-two solutions, and all-input Theta results for P-ADD and executable P-MATMUL. The earlier idealized P-MATMUL recurrence retains its all-input upper bounds. The imports also provide monotonicity, adjacent-power sandwiches, and all-input Theta results for P-MERGE and P-MERGE-SORT. The compatibility extension retains the corresponding Strassen recurrence results without presenting them as Chapter 26 main text. The executable binary lower bound used by P-MERGE additionally has a complete duplicate-sensitive partition theorem and logarithmic work/span bounds. CLRS.­Chapter27.­MergeSplit and CLRS.­Chapter27.­pMerge implement the actual midpoint/binary-search P-MERGE control structure, and CLRS.­Chapter27.­pMerge_correct proves its sortedness, permutation, and exact-length specification. The pointwise CLRS.­Chapter27.­pMerge_work_lower and CLRS.­Chapter27.­pMerge_work_upper theorems prove its actual execution has linear work. The theorem CLRS.­Chapter27.­pMerge_span_upper proves the universal quadratic-logarithmic span upper bound, and sorted interleaved power-of-two inputs attain the matching bound through CLRS.­Chapter27.­pMerge_interleaved_span_lower. CLRS.­Chapter27.­pMergeSort executes the textbook midpoint split, parallel recursive sorting, and sequential P-MERGE phase; CLRS.­Chapter27.­pMergeSort_correct proves sortedness, permutation, and exact output length for every input. The exact step equations CLRS.­Chapter27.­pMergeSort_work_step_eq and CLRS.­Chapter27.­pMergeSort_span_step_eq connect its carried costs to the recursion. The pointwise pair CLRS.­Chapter27.­pMergeSort_work_lower and CLRS.­Chapter27.­pMergeSort_work_upper proves executable Θ(n log n) work; CLRS.­Chapter27.­pMergeSort_span_upper gives the universal cubic-log span bound, while CLRS.­Chapter27.­pMergeSort_worstFamily_span_lower gives the matching recursive witness family.

Completion boundary

The executable P-MERGE-SORT cost gap is closed, and the parallel-Strassen compatibility extension has been isolated from the main-text recurrence modules. No represented Chapter 26 main-text core proof obligation remains.

Implementation details