Asplos Workflow

ASPLOS Workflow

An ASPLOS campaign is a pipeline with two entry gates per edition and a revision side-channel, not a single deadline sprint. All dates below are the verified 2027 cycle (CFP and conference site, checked 2026-07-08); reopen the official pages before committing a team to any of them.

The 2027 cycle as a calendar

There is no separate abstract deadline in 2027, and no mid-year third round — the edition moved from three deadlines to two. Missing September means waiting for the next edition's first gate.

Working backward from September 9, 2026

A credible countdown for a systems paper with hardware-dependent evidence:

T-10 weeks (by ~Jul 1)   Freeze the claim list; run asplos-topic-selection;
                         choose evidence instruments (silicon / FPGA / simulator).
T-8 weeks                Evaluation infrastructure done: baselines building,
                         workloads chosen, simulator configs pinned.
T-6 weeks                First full results sweep; kill or descope claims the
                         data will not carry. Start the two-page draft NOW —
                         it is the part rapid review actually reads.
T-4 weeks                Full draft; internal red-team against the rapid-review
                         question set; ablations running.
T-2 weeks                Freeze experiments. Citation-format pass (full names,
                         no "et al.", DOIs). Anonymity sweep.
T-1 week                 HotCRP entry created; PDF uploaded; co-author sign-off.
T-0  Sep 9, 11:59 PM AoE HotCRP status shows *complete*, not draft.

Role assignments that prevent known failures

• Two-page owner — one author owns pages 1-2 as a standalone artifact and rehearses them against asplos-writing-style; rapid review reads nothing else.
• Evidence-integrity owner — keeps the experiment matrix (claims × platforms × baselines) current so the rebuttal window never requires new runs under pressure.
• Compliance owner — template, 11-page geometry, citation full-name rule, GenAI disclosure, HotCRP fields; format violations risk rejection without review.
• Response owner — reserves the response window dates (Dec 1-4 for the September cycle) on every author's calendar at submission time, not after reviews arrive.

Branch planning at notification (December 21, 2026)

1. Accept → switch to asplos-camera-ready immediately; camera-ready lands ~6 weeks out, and artifact evaluation (asplos-artifact-evaluation) runs in parallel with its own appendix and deadlines (2027 AE calendar 待核实).
2. Major Revision → treat as a funded contract, not a consolation: the revision is due at the camera-ready deadline and counts as a submission; scope revision work to exactly what the decision letter requires, tracked in a change log that will become the resubmission note.
3. Reject at rapid review → the diagnosis is almost always positioning: the first two pages did not establish the intersection or the evidence. Fix framing before considering a new venue.
4. Reject after full review → mine the reviews for the evidence matrix, then re-run asplos-topic-selection; some full-review rejects are strong ISCA, SOSP, or PLDI papers mis-aimed at the intersection.

Cross-edition cautions

• The two-deadline structure is itself new for 2027; do not assume it for 2028.
• ASPLOS logistics genuinely change year to year — the 2025 edition (30th, Rotterdam) ran jointly with EuroSys; 2026 (31st) was standalone in Pittsburgh; 2027 is in Crete. Travel, visa, and registration planning must restart per edition (fees and attendance rules for 2027: 待核实).
• Program leadership rotates every edition; there is no standing editorial memory to appeal to. A policy answer from a previous year's chairs is expired.

Hardware lead times are schedule items

Systems evidence has procurement physics that a paper calendar must absorb:

• Cluster or testbed reservations for the T-8 to T-2 sweep window book out weeks ahead; reserve at claim-freeze time, not first-results time.
• FPGA synthesis iterations and simulator calibration runs are days-long serial steps — one redesign late in the countdown costs a week, not an evening.
• Access to pre-production parts (new memory devices, engineering samples) often comes with NDA review of the paper text; start that review before T-4 or it becomes the critical path.

Between notification and Crete

An accepted September-cycle paper has a compressed post-decision runway: camera-ready and any Major Revision work land ~6 weeks after December 21 — squarely across the holiday break, which the six-week plan must acknowledge — followed by artifact evaluation, then travel logistics for an April conference in Greece (visa timelines for some team members exceed the gap; start at notification). Presentation format, registration requirements, and speaker obligations for 2027 were 待核实 at pack-check time: read the acceptance email before assigning the talk.

Pipelining across gates

Teams running multiple projects should treat the two gates as a pipeline: a paper that misses the September evidence bar can target the next edition's first gate with its evaluation matrix already built, while the response and revision windows of a September submission land in December-February — plan author bandwidth so a Major Revision does not collide with the next project's deadline sprint. One person cannot own the revision of paper A and pages 1-2 of paper B in the same six weeks.

Output format

[Cycle chosen] September 2026 / next edition (reason)
[Countdown health] on-track items · at-risk items with dates
[Owners] two-page / evidence / compliance / response — names attached?
[Branch plans] accept · major-revision · rapid-reject · full-reject, one line each
[Calendar holds] Dec 1-4 response window + 6-week revision block reserved? Y/N
[Facts to re-verify] <official pages to reopen before relying on this plan>

Source: brycewang-stanford/Awesome-Journal-Skills → ASPLOS-Skills/skills/asplos-workflow/SKILL.md