ut-parla/Parla.py

Issues

• Relatively aggressive logging may affect performance

  #36 opened 4 years ago
  1

• Integration Testing

  #34 opened 4 years ago
  0

• clone_here could not maintain major order

  #31 opened 4 years ago
  1

• Not setting memory argument can cause crashes

  #30 opened 4 years ago
  1

• Scheduler gives unnecessary warnings

  #29 opened 4 years ago
  2

• VCUs have no defined meaning and need guidelines in general and a definition for GPUs specifically

  #28 opened 4 years ago
  1

• Type aliases

  #26 opened 4 years ago
  3

• Vague module semantics

  #25 opened 4 years ago
  2

• Uploading docs built from master

  #24 opened 4 years ago
  1

• "Virtual Compute Unit" should be documentated and potentially renamed

  #23 opened 4 years ago
  4

• Scheduler Awareness of Persistent Data

  #19 opened 4 years ago
  1

• clone_here May Swap Current Device

  #18 opened 4 years ago
  7

• Buggy Cross-GPU Synchronization?

  #17 opened 4 years ago
  10

• Tasks should run on their own streams

  #16 opened 4 years ago
  0

• Poor Scaling In Matmul and Exp Demos

  #15 opened 4 years ago
  40

• Tasks can't modify nonlocal variables

  #14 opened 4 years ago
  3

• Parla components import order errors

  #13 opened 4 years ago
  1

• Segmentation Fault with VECs when multithreaded

  #12 opened 4 years ago
  2

• VECs should be optional

  #11 opened 4 years ago
  1

• Some import mechanisms bypass a customized builtins.__import__

  #10 opened 4 years ago
  7

• Can't load MKL into a VEC

  #9 opened 4 years ago
  3

• VECs don't work in interactive Python (REPL)

  #8 opened 4 years ago
  14

• Parla task scheduler may limit scaling due to lock usage and polling

  #3 opened 5 years ago
  1

• Separate examples, benchmarks, and integration tests

  #2 opened 5 years ago
  2