LDWG Meeting minutes: April 16, 2018

Attendees

• Mario Baldi (Cisco)
• Antonin Bas (Barefoot)
• Gordon Brebner (Xilinx)
• Mihai Budiu (VMware)
• Calin Cascaval (Barefoot)
• Andy Fingerhut (Cisco)
• Nate Foster (Cornell)
• Vladimir Gurevich (Barefoot)
• Andy Keep (Cisco)
• Milad Sharif (Barefoot)
• Chris Sommers (Keysight)
• Han Wang (Barefoot)
• Bo Liu (?)
• Sksodhi (Juniper)

5 minutes: Announcements (Nate and Gordon)

• Workshop on in-kernel networking; interested in P4; ping Nate for an invite.
• Need to finalize a minor version update of P4_16 before June Workshop.

15 minutes: Updates from Sub-Working Groups

• Serializability (Andy K)

• Working notes: https://github.com/akeep/p4-proposals
• What to do with serializable enums?
• Example: enum { A; B; C;}. How does 4 convert?
• Mihai proposed a type-safe enum.
• Compiler checking of range of values.
• Summary of email-discussion of efficiency of serializability
• Action Item: finalize safety issues.

• Control-Plane Types (Andy F)

• Mostly bringing compiler experts up to speed.
• So far, no language changes needed.
• Outcomes would affect P4Runtime / PSA
• Action item: sub-group meeting this week.

• Modularity (Nate)

• Discussed use cases
• Discussed type families for data-like types
• Started thinking about broader language changes that might facilitate modular programming
• Short-term: type families proposal
• Longer-term: start of P4_18?
• Action Item: sub-working group meeting on non-LDWG weeks

60 minutes: Technical Issues

• Table Sizes (Andy F)

  • Issue: https://github.com/p4lang/p4-spec/issues/514
  • PR: https://github.com/p4lang/p4-spec/pull/603
  • Status: PR has been approved for 3 weeks
  • Action Item: Andy F to add issue to type check integer Saturating

• Saturating Arithmetic (Calin)

  • Issue: https://github.com/p4lang/p4-spec/wiki/saturating
  • Spec PR: https://github.com/p4lang/p4-spec/pull/609
  • Status: saturating data types considered too heavy; we decided to add saturating operations.
  • Implementation: easy for compiler, just need to file issues with the targets like Bmv2.
  • Discussed whether we prefer |+| or [+]? Go with |+|
  • Action Item: merge PRs

• Parser Values Sets (Han)

  • Issue: https://github.com/p4lang/p4-spec/issues/565
  • Spec PR: https://github.com/p4lang/p4-spec/pull/602
  • p4c PR: https://github.com/p4lang/p4c/pull/1213
  • Status: main issue is naming of keys of parser values sets and match_types.
  • Seems like we need something very much like a table without actions.
  • Action Item: Han and Vladimir to convene a sub-working group.

• Named Parameters (Calin)

  • Issue: https://github.com/p4lang/p4-spec/issues/598
  • Spec PR: https://github.com/p4lang/p4-spec/pull/599
  • Based on "arg = val" syntax
  • Can omit optional parameters by writing "_"
  • For simplicity, don't want optional parameters to affect control-plane APIs
  • Open issues:
    1. Restrictions on mixing named/unnamed/optional parameters 1b. Overloading
    2. Default values: None, _, etc.
  • See Mihai's arithmetic examples with v1model.p4 p4c/testdata/p4_16_samples
  • Action Item: Mihai to generate some examples that don't require None; Calin to iterate on design.

• Grammar Syntax Tweaks (Chris and Mihai)

  • One Issue: https://github.com/p4lang/p4-spec/issues/605
  • A different Spec PR: https://github.com/p4lang/p4-spec/pull/608

Next meeting

May 7th @ Cisco.