B-Theory
← All courses

53 videos · about 235 minutes

Correspondence Matrices — Advanced

Explore representations, shared computation, exact comparisons and recognition through 53 videos connecting correspondence matrices to implementation and benchmark evidence.

Follow how a Boolean rule becomes a matrix, shared program or packed output, then examine the costs, comparison boundaries and exact checks behind performance claims.

Start the course →Start with From Logic to Code

Code and benchmark results

Explore the implementation and consult the published benchmark results alongside the videos. Check the workload, required output and measurement boundary for each comparison.

Correspondence Matrices repository on GitHub →Benchmark results and comparison reports →

1. Foundations and representations

  1. Boolean rules and exact decisions4:33
  2. Expression, table, and function4:25
  3. A truth function in a matrix layout5:34
  4. The sixteen two-input operator CMs5:17
  5. Logical matrices and tensor construction6:57
  6. A filled grid and a shared program4:40
  7. Live support and display dimensions4:33
  8. Choose the required answer4:19

2. Execution, sharing and reuse

  1. Repeated work in a raw tree4:44
  2. Reusing identical subexpressions4:30
  3. Canonicalization Interning4:36
  4. Shared nodes and actual reuse4:50
  5. Flattening without losing sharing9:45
  6. Identifying a distinct compiler mechanism4:21
  7. CM-IR Persistence9:36
  8. Eager and deferred computation3:22
  9. Packing truth outputs into bits5:36
  10. Small operator tokens and construction routes5:25
  11. Choosing materialization boundaries4:26
  12. Selector Width Limit3:36

3. Applications, benchmarking and evidence

  1. Output cones and semantic support4:45
  2. Exact policy rewrites and their cost4:15
  3. Safe reuse when configurations change4:36
  4. Parallel work and complete request cost3:25
  5. Instructions, operations, and memory4:24
  6. When preparation is repaid3:25
  7. The boundaries of a timing result3:33
  8. Animation and experimental evidence4:10
  9. Guards for a fair exact comparison3:24
  10. The corrected B2/B4 comparison4:36
  11. What replication adds4:32
  12. Epfl Parity3:20
  13. Reading timing ratios and intervals3:12
  14. Keeping comparison scope explicit3:13
  15. Why there is no universal winner chart3:08
  16. Choosing an eligible representation3:15
  17. Match the tool to the artifact4:15
  18. Binding a video to its evidence3:13
  19. Preserving the correction trail3:06

4. Recognition and exact verification

  1. Advice behind an exact check4:37
  2. Exact decomposition and learned transfer5:59
  3. Supervision, ranking, and transfer4:31
  4. The exact source-ANF core4:30
  5. Routing, restart, and conversion3:19
  6. Exact screening and artifact admission5:00
  7. Policy gates and fixed exact arms3:29
  8. Matching the best-artifact task3:27
  9. Fresh sources and host transfer3:20
  10. Rewrites across complete tasks4:15
  11. Exact transfer without a runtime win3:04
  12. The cost of abstaining3:05
  13. Recognition D104:47
  14. Task-aware BDD and SAT advice4:38