Repository navigation

Sparse GPU implementation - #35

Closed

pelesh wants to merge 35 commits into

developfrom

perk/opflow/hiop-sparse-gpu

Closed

Sparse GPU implementation#35
pelesh wants to merge 35 commits into
developfrom
perk/opflow/hiop-sparse-gpu

Commits on Dec 6, 2023
- OPFLOW: initial implementation of RAJA/HiOp sparse GPU-based solver
  
  Abhyankar, Shrirang G
  authored and
  wperkins
  committed
  d3e0ac0
- Spit out sparse Jacobian indexes for debugging
  wperkins
  committed
  80e8599
- Count equality Jacobian nonzeros and assign bus/gen/load locations
  wperkins
  committed
  8670724
- Equality Jacobian nonzero count matches
  wperkins
  committed
  0287bd7
- Inequality Jacobian nonzero count matches
  wperkins
  committed
  02d6619
- Log event begin/end unbalanced in equality Jacobian for some reason
  wperkins
  committed
  33902a5
- Checkpoint - Equality Jacobian assembly runs
  wperkins
  committed
  cd0bd76
- Checkpoint: equality Jacobian indexes correct (w/o lines)
  wperkins
  committed
  8acb1d9
- Add line to/from locations for sparse equality Jacobian
  wperkins
  committed
  3bf98af
- GPU computed sparse equality Jacobian indexes correct
  wperkins
  committed
  335ddd1
- Add some additional line information for sparse Jacobian assembly
  wperkins
  committed
  94ddd33
- Sparse GPU equality Jacobian assembly is coded and running (results untested)
  wperkins
  committed
  46efbde
- Streamline debug printing of equality Jacobian
  wperkins
  committed
  84a45be
- Equality Jacobian is correct, but apparently needs to be reordered
  wperkins
  committed
  fae9cd9
- Inequality Jacobian indexes are correct (for lines)
  wperkins
  committed
  6890b32
- Inequality Jacobian values for line limits added
  wperkins
  committed
  d60a5ff
- Minor changes to inequality Jacobian line limit values
  wperkins
  committed
  523932d
- Sort Jacobian indexes, but not values.
  wperkins
  committed
  fd65166
- Correctly reorder Jacobian values, but using the wrong method
  wperkins
  committed
  067469f
- Actually use GPU-computed Jacobian; reduce steps to reorder Jacobian values
  wperkins
  committed
  8fb2e2c
- Print (CPU) Hessian non-zero count, indexes, and values
  wperkins
  committed
  fe0b6ea
- Add Hessian indexes to line params; make sure Hessian index is allocated
  wperkins
  committed
  fcf9074
- Attempt to count Hessian non-zeros (wrong, of course)
  wperkins
  committed
  8a2c4d3
- Temporarily capture and report Hessian row/col indexes used
  wperkins
  committed
  eaa1cf0
- GPU Hessian indexes match model but not resulting matrix?
  wperkins
  committed
  d241781
- First shot at GPU computed Hessian values (untested)
  wperkins
  committed
  d15445b
- Sort Hessian indexes and permute values; clean up allocations
  wperkins
  committed
  ca85561
- Add some minor debug messages about Hessian size
  wperkins
  committed
  8e11b60
- Allocate Hessian differently to avoid unnecessary structure
  wperkins
  committed
  86f787d
- Add an entry for generator reactive power even if it's unused
  wperkins
  committed
  e3905c7
- GPU assembled Hessian non-zero count and indexes are correct
  wperkins
  committed
  1e78787
- Mark Hessian contributions with a code
  wperkins
  committed
  86ff70d
- Checkpoint: GPU Hessian values incorrect -- new approach maybe?
  wperkins
  committed
  718359f
- Fix Hessian indexing problems in line limits
  wperkins
  committed
  baa1c9d
- Apply pre-commmit fixes
  wperkins
  committed
  f05b8d7

Commits

Commits on Dec 6, 2023