Most Repeated Topics in GATE CSE: The Top 25 Concepts

The most repeated topics in GATE CSE are C pointer and array tracing, CPU scheduling metrics, eigenvalue identities, Bayes theorem and synchronization code checks. The ranking below scores concepts from GATE CS papers of 2000–2026 and explains the top 25.

Scores are analytical confidence values from an evidence model, not probabilities that a question will appear.

Key takeaways

• All ten technical subjects appear in the top 25. No subject is safe to skip.
• Operating Systems has the most top-25 entries (four), followed by Programming and DS, Engineering Mathematics and DBMS with three each.
• NAT is a listed format for most of the top 25. These are calculation templates, and NAT carries no negative marking.
• Top concepts recur in many variants, so practise the variants rather than memorising past answers.
• This is a priority list, not a study sequence. Learn prerequisites first.

How the ranking was built

Past questions from 2000–2026 were grouped into concepts as GATE actually tested them, not by textbook chapter. That produced 216 concepts across ten technical subjects. Each concept was scored from 0 to 10 on eight factors, and the weighted total gives a score out of 100:

Concepts scoring 80 or more are Very High (Tier 1). Exactly 100 concepts score 70 or more, and those form the Top 100. The year lists behind the scores are reconstructed from public question archives and are approximate. The method is explained step by step in how to use GATE CSE previous year papers.

The top 25 at a glance

This article stops at rank 25. The book ranks all 100 concepts with the historical and recent evidence for each, and pairs every Tier 1 and Tier 2 concept with practice questions covering its recurring variants.

What each concept looks like in the paper

Grouped by subject. Years mentioned are approximate reconstructions.

Programming and Data Structures (Ranks 1, 8, 20)

• Pointer tracing. A 2-D array accessed through plain and row pointers; report the printed sum. Present in every shift since 2018.
• Recursion tracing. Return value, number of calls counted with a static or global variable, or print order.
• BSTs. Build a tree from an insertion sequence, spot an impossible search path, or delete a node using its successor.

Operating Systems (Ranks 2, 5, 9, 21)

• CPU scheduling. Average waiting or turnaround time under SRTF, round robin or priority with staggered arrivals. A multilevel queue variant appeared in 2026.
• Synchronization. Given turn-flag or semaphore code, decide mutual exclusion, progress, bounded waiting and deadlock. As an MSQ it gives no partial credit, so check every property.
• Paging. Levels, bits per level and page-table size, combined with a TLB in a single NAT in the 2024–2026 papers.
• Page replacement. FIFO, LRU and Optimal fault counts, sometimes paired with a Belady's anomaly check.

Engineering Mathematics (Ranks 3, 4, 19)

• Eigenvalues. Determinant or trace of A^k, A^-1 or A + kI from given eigenvalues. Shortcuts make it a 60-second NAT.
• Bayes. Posterior probability of one of two or three sources with prior shares and defect rates. The story changes; the arithmetic does not.
• Graph counting. Vertices or edges from degree data, or planar graphs using Euler's formula and e ≤ 3v − 6.

COA (Ranks 6, 11)

• Cache mapping. Tag, index and offset widths, or the reverse: associativity or cache size from a tag width. The 2027 syllabus names cache memory mapping explicitly.
• Pipelining. Clock cycles for a few instructions in a 5-stage pipeline with data hazards, with forwarding on or off. Pipeline hazards are named in the syllabus too.

Databases (Ranks 7, 17, 18)

• SQL. Rows returned by nested queries using NOT EXISTS, NOT IN with NULLs, or ALL over an empty subquery. Analysts flagged tricky SQL in 2023–2026.
• Serializability. Draw the precedence graph, then find or count equivalent serial orders. Present in every paper 2019–2026.
• Normal forms. Highest normal form from functional dependencies, and MSQs on whether a decomposition is lossless, dependency-preserving or in BCNF.

Theory of Computation (Ranks 10, 16)

• Language classes. Which of four languages are regular, DCFL, CFL or not CFL. Asked essentially every year since 2004, now as an MSQ.
• Regular expressions. Write or identify the expression for a described language, or pick the ones equivalent to a given expression.

Computer Networks (Ranks 12, 13)

• TCP congestion control. Track cwnd and ssthresh through timeouts and triple duplicate ACKs, or compare Tahoe and Reno behaviour.
• CIDR and subnetting. Hosts per subnet, the smallest block for a requirement, or a broadcast address. Both CN topics are named in the 2027 syllabus, and the cut in other CN topics is likely to raise their share.

Compiler Design (Ranks 14, 22)

• SDT. Evaluate a synthesised attribute for a short input with unusual operators, or tell S-attributed from L-attributed definitions.
• LR parsing. Find shift-reduce or reduce-reduce conflicts, or classify a grammar as LR(0), SLR, LALR or CLR.

Digital Logic (Ranks 15, 23)

• Boolean algebra. Pick the one invalid identity, or the forms equivalent to a function. Analysts reported higher than expected Boolean weight in 2026 shift 2.
• K-maps. Minimal SOP or POS with don't-cares, and counts of prime implicants, essential prime implicants or literals.

Algorithms (Ranks 24, 25)

• Loop complexity. Doubling loops, j += i loops and i*i ≤ n conditions. Present in every paper since 2010.
• Dijkstra. Extraction order, final distances, relaxation counts and behaviour with a negative edge. The 2026 paper asked it with modified conditions.

What the top 25 have in common

• They are procedures, not definitions. Tracing, counting and computing dominate, so almost all of them can be drilled.
• Several are one idea under two subject names. Paging access time in OS and average memory access time in COA use the same hit-ratio average. Link-state routing in Networks is Dijkstra on the topology. Serializability in DBMS and synchronization in OS both ask which interleavings are safe. Prepare each idea once, in both vocabularies.
• The top of some subjects is narrow. In Compiler Design, nothing outside SDT, LR parsing and LL(1) scores above 79.

How to use this list

1. Treat Ranks 1–50 as non-negotiable. All fifty are Tier 1. On 2018–2026 evidence, full command of every variant of these concepts covers roughly half of the 85 technical marks.
2. Do not study in rank order. Rank reflects evidence strength, not prerequisites. Learn recursion before recursive tree functions, attribute closure before normal forms, and asymptotic notation before recurrences.
3. Use the ranks in the exam too. Attempt questions on drilled concepts early and quickly. Lower-ranked ones can wait until certain marks are secured, but attempt them when they are NAT or MSQ, which carry no negative marking.
4. Cover the syllabus gaps. Some topics named in the 2027 syllabus score below 70 because they have little history: tabular (Quine–McCluskey) minimisation, hardwired versus microprogrammed control, the socket API, and DNS and HTTP. They still need preparation. Confirm the current syllabus at gate2027.iitm.ac.in.
5. Balance with weightage. Pair this list with GATE CSE subject-wise weightage so time per subject matches marks, then plan the weeks with the GATE CSE 2027 preparation strategy.

For subject deep dives, start with Operating Systems important topics and Engineering Mathematics important topics.