7-Day Catch-Up Schedule for the Concurrency Retake

Exam date: Monday, April 13, 2026 at 13:30
Minimum practical target: about 18/39
Safe target: 20/39

Day intensity: Mon/Thu/Sun = full days | Tue/Wed/Fri = work days (light) | Sat = birthday (light)

Use the site as your base. Use the real slides and past exams as the check that you actually understand the material. Use the generated practice exams only as extra calibration.

Keep the same study loop all week: 1. Read one study-guide section 2. Do one related source question or source-derived mini-question 3. Close notes and write the rule from memory 4. Make one small variation yourself so you are not just recognizing wording

Day 1 - Monday, April 6 (full day)

Read: patterns, scan, permute
Memorize the core decision rules for map, zipWith, stencil, fold, scanl, scanr, backpermute, permute
Review slides 12, 13, and 14 for map/stencil, gather/scatter, fold/scan, and segmented scan
Do source-backed questions on data-parallel patterns and scan from 2024-exam2.pdf and 2024-exam3.pdf
Make at least 6 mini-questions yourself by changing a pattern question into a nearby variant
Do the flashcards for Data-Parallel Patterns and Parallel Scan

Checkpoint: you should be able to identify the dependency shape of a question before thinking about the exact answer.

Day 2 - Tuesday, April 7 (work day - light)

Read: gpu and re-read the scan sections on inclusive vs exclusive scan, tiled scan, and chained scan
Review slides 10 and 14 for warp lockstep, occupancy, memory coalescing, gather vs scatter, and scan communication
Write from memory: warp, branch divergence, occupancy, coalescing, gather/backpermute vs scatter/permute, reduce-then-scan, scan-then-propagate, chained scan
Do the GPU and scan source MC/material from 2024-exam2.pdf and 2024-exam3.pdf
End with one short closed-notes flashcard pass

Checkpoint: you should be able to explain why gather/backpermute is usually preferred over scatter/permute on a GPU.

Day 3 - Wednesday, April 8 (work day - light)

Read: primitives and locking
Review slides 06 and 08 for STM fairness/progress limits, retry, orElse, and starvation
Do the fetch-or source question from 2024-exam3.pdf
Do the IORef/MVar/TVar properties source question from 2024-exam3.pdf
Write the IORef/MVar/TVar property table from memory twice
Write the 4 deadlock conditions from memory
Do the flashcards for Concurrency Primitives and Locking & Synchronization

Checkpoint: you should be able to judge atomic, wait-free, deadlock-free, and starvation-free separately without mixing them up.

Day 4 - Thursday, April 9 (full day)

Read: work-span and definitions
Review slide 15 for RAM/PRAM, work, span, Brent's theorem, overhead, efficient vs optimal
Practice setting up recurrences before solving them
Solve the source-style algorithms: eagle from 2024-exam2.pdf, erumpent from 2024-exam3.pdf, undertow from 2020-exam2.pdf, tierra from 2020-exam3.pdf
Do 2 processor-count / worthwhile-parallelism questions
Make 3 fresh recurrence variants by changing the local work term or number of recursive calls

Checkpoint: you should be able to write the work and span recurrences immediately, before doing algebra.

Day 5 - Friday, April 10 (work day - light)

Make a mixed set of 10 to 15 short questions from source material and your own variants
Include 3 pattern-selection questions
Include 2 scan questions
Include 2 GPU questions
Include 3 concurrency-property/locking questions
Include 2 work/span questions
Solve them closed-notes in one sitting
Review only misses, and fix the rule that caused the miss

Checkpoint: you should be solving variants correctly, not just recognizing old questions.

Day 6 - Saturday, April 11 (birthday - light)

Do one short flashcard pass
Re-do one missed question from patterns/scan/GPU
Re-do one missed question from primitives/locking
Re-do one missed question from work/span
Update your one-page summary sheet with only the rules you still hesitate on
Stop early rather than dragging through low-quality study

Checkpoint: the topics that used to feel shaky should now feel predictable.

Day 7 - Sunday, April 12 (full day)

Do one full timed 2-hour session
Use a mix of source questions, your own variants, and only a few generated practice questions as extras
Start in this order: fast conceptual points first, work/span second, stretch questions last
Grade yourself honestly
Sort every miss into one of 3 causes: fact gap, pattern-selection gap, recurrence setup gap
Re-read only the matching study-guide sections
Finish with one final flashcard pass
Stop in time to sleep properly

Checkpoint: aim to reach at least 18/39, and preferably 20/39, in a timed setting.

Exam Day - Monday, April 13 at 13:30

First pass: secure the fast conceptual points quickly
Second pass: do work/span and processor-count questions
Third pass: spend remaining time on stretch questions
For pattern questions: identify the dependency shape first
For concurrency-property questions: judge each property independently
For work/span questions: write the recurrence first, then solve it
Do not chase perfection; chase the cleanest route to 20 points

Source-Backed Topic Map

Topic	Best local sources	Why it matters
GPU architecture	`slides/10-GPGPU.pdf`, `slides/16-Conclusion.pdf`, GPU MC from `exams/2024-exam2.pdf`	Fast bankable points
Patterns	`slides/12-Data-Parallelism-1.pdf`, `slides/13-Data-Parallelism-2.pdf`, pattern questions in `exams/2024-exam2.pdf` and `exams/2024-exam3.pdf`	Very high yield
Scan	`slides/14-Data-Parallelism-3.pdf`, scan questions in `exams/2024-exam2.pdf` and `exams/2024-exam3.pdf`	Very high yield
IORef / MVar / TVar	`slides/06-STM-1.pdf`, `slides/08-STM-2.pdf`, property question in `exams/2024-exam3.pdf`	Fast conceptual points
Locking / atomics	`slides/06-STM-1.pdf`, `slides/10-GPGPU.pdf`, `fetch-or` in `exams/2024-exam3.pdf`	Reliable short-answer points
Work/span	`slides/15-Work-Span.pdf`, `exams/2024-exam2.pdf`, `exams/2024-exam3.pdf`, `exams/2020-exam2.pdf`, `exams/2020-exam3.pdf`	Main medium-difficulty scoring bucket