Upstream tracking issues

[ronawho]

Upstream issues this work would benefit from:

Performance:

• Optimize comm patterns to avoid needing usePrimitiveComm:
  • Parallelize DefaultRectangular array assignment chapel-lang/chapel#14478
  • Optimize DefaultRectangular <=> Block transfers chapel-lang/chapel#14479
• Enable -schpl_serializeSlices by default
  • Enable slice serialization by default chapel-lang/chapel#14530
• Avoid need for localAccess when using direct indexing:
  • Optimize indexing into distributed arrays with domain values chapel-lang/chapel#14480
• Improve performance on InfiniBand networks:
  • Improve gasnet-ibv performance chapel-lang/chapel#14438

Avoid non-user facing features:

• Avoid using non user-facing features for planner lock:
  • Create a mutex/lock library chapel-lang/chapel#9881
  • Request: Attributes for locale-private global variables chapel-lang/chapel#12300
• Avoid using default-intent-is-ref pragma:
  • Support user-specified default intents for record types chapel-lang/chapel#13319
• Add official lower-level comm operations
  • [Multiresolution Design Philosophy] Expose user-friendly chpl_comm primitives chapel-lang/chapel#13052

Elegance:

• Use promoted/zippered expressions when possible instead of direct indexing
  • Multidimensional zippered iteration (or promotion) kills performance chapel-lang/chapel#13147

Older or not as relevant:

• Reduce interference between Chapel and separate OpenMP/Pthread runtimes:
  • Reduce interference between Qthreads and OpenMP chapel-lang/chapel#9882