ppopp-topic-selection

PPoPP Topic Selection

Decide the venue before drafting. PPoPP — the ACM SIGPLAN Symposium on Principles and Practice of

Parallel Programming — is the premier forum for work whose **center of gravity is parallelism and

concurrency**: parallel languages, compilers, and runtimes; lock-free/wait-free data structures;

GPU/accelerator and NUMA performance; parallel algorithms; and scalability studies. A technically

strong paper whose real contribution is a compiler pass, a program logic, or a cache design is

respected here and then rejected as out of scope. The question PPoPP reviewers ask first is: **is

the parallelism the point, or is it the setting?**

The routing question that matters most

PPoPP sits inside the co-located HPCA/CGO/PPoPP/CC week, so its nearest neighbors are one

hallway away and the boundary is real. Use the "is the parallelism the point?" test: if you removed

the concurrency/parallel-performance story, is there still a paper? If yes, the contribution lives

elsewhere.

Sibling-venue routing table

| Signal in your project | Better home | Why |

|---|---|---|

| The contribution is a parallel algorithm, runtime, synchronization mechanism, or a concurrency-correctness + scalability result | PPoPP | This is PPoPP's center; both correctness and speedup are judged |

| The contribution is a compiler optimization or code-generation technique | CGO (or PLDI) | Backend/optimization is CGO's named center; PLDI for broader PL implementation |

| The contribution is a new program logic, type system, or semantics for concurrency | POPL | Foundational concurrency theory over a measured parallel system |

| The contribution is hardware/microarchitecture (coherence, memory system, accelerator design) | ASPLOS / HPCA | The result is about the machine, not the parallel program |

| The core is a large-scale HPC application or systems-at-scale deployment | SC / ICS / IPDPS | Application/HPC-systems framing over a general parallel-programming lesson |

| The core is a theoretical parallel-algorithm bound with no implementation | SPAA | Theory of parallelism; PPoPP wants practice alongside principle |

| A concurrency feature of a managed language / object system | OOPSLA / PLDI | Language-design framing dominates |

Contribution shapes PPoPP rewards

• Concurrent data structures — a new lock-free/wait-free structure or synchronization primitive

with a correctness argument (linearizability, progress) and throughput/scalability under

contention.

• Parallel runtimes and schedulers — work-stealing, task graphs, futures, load balancing, with

measured overhead and scaling on real workloads.

• GPU / accelerator programming — techniques, languages, or libraries that raise occupancy,

reduce divergence, or exploit heterogeneity, with speedups over strong GPU baselines.

• Parallel algorithms in practice — graph, numerical, or sparse kernels whose parallel design

and NUMA/locality engineering deliver a measured win.

• Memory-model and concurrency-correctness work — race/linearizability checking, weak-memory

reasoning, or verified concurrency, positioned as a parallel-programming enabler.

• **Compilers/languages for parallelism** — DSLs, parallel IRs, and runtime-coupled compilation

where the parallel-execution lesson (not a generic pass) is the contribution.

The remove-the-parallelism and swap-the-machine tests

Two quick tests sharpen a borderline verdict:

• Remove-the-parallelism test: delete the concurrency and the scaling story. If a contribution

remains (a better pass, a cleaner semantics, a faster core), route to CGO/PLDI/POPL/ASPLOS.

• Swap-the-machine test: if your result is a single-configuration speedup that would vanish or

reverse on another core count, socket topology, or GPU generation, it is a microbenchmark, not a

parallel-programming contribution — PPoPP wants a lesson that survives the sweep.

Evidence maturity, without the cliché

Fit is necessary but not sufficient. A synchronization idea shown only at one thread count is a

workshop poster; a runtime evaluated only on microbenchmarks needs real workloads before the

research track; a scalability claim with no baseline is not yet a paper. PPoPP's twin bar —

correct under concurrency and measurably scalable — is what promotes an idea from "interesting"

to "PPoPP-shaped."

Cheap reconnaissance before committing

[Scope]   scan the last two PPoPP programs (dblp, conf.researchr.org) for your subarea
          -> 3+ recent papers = a reviewer pool exists; 0 = mismatch or a CGO/ASPLOS home
[Citations] is your bibliography majority PPoPP/PLDI/ASPLOS/SC/SPAA parallel venues?
          -> majority elsewhere => reviewers read you as a visitor; reframe the intro
[Calendar] PPoPP's summer deadline vs CGO/PLDI/ASPLOS/SC dates (same co-located family)
          -> route to the nearest honest fit; a marginal preference rarely justifies a year

Decision procedure

[Point test] remove the parallelism -> is there still a contribution? yes => reroute
[Claim type] concurrent structure / runtime / GPU / parallel algorithm / memory-model / parallel-language
[PPoPP vs CGO/PLDI] is a compiler pass the real result? -> CGO/PLDI
[PPoPP vs POPL] is it concurrency theory with no measured system? -> POPL
[PPoPP vs ASPLOS/HPCA] is it about the machine? -> ASPLOS/HPCA
[Verdict] PPoPP research track / sibling venue, with a one-line reason

Run this before the writing skills; a wrong venue decision wastes every later step. When the

verdict is PPoPP, continue with ppopp-workflow for the calendar and ppopp-writing-style for the

paper shape.