Skip to content

Latest commit

 

History

History
151 lines (114 loc) · 9.65 KB

CHANGELOG.md

File metadata and controls

151 lines (114 loc) · 9.65 KB
Raw

Change Log

All notable changes to HiOp are documented in this file.

Version 1.1.0: Checkpointing, advanced restarts, and misc fixes

  • Checkpointing for quasi-Newton solver: new API for load and saving checkpoints, user options for rollover restarts, and scalable I/O via axom::sidre.
  • New function for advanced warmstarts.
  • Updated spack build with build cache.

Version 1.0.3: Moving limits and misc fixes

  • Update modules for CI tests on LLNL LC by @nychiang in #679
  • Update cmake build system to require RAJA when GPU compute mode is used by @nychiang in #676
  • Moving limits options for NLP IPM solvers by @cnpetra in #681

Version 1.0.2: Misc fixes

  • Removed deprecated ALG2 for cusparseCsr2cscEx2 by @cnpetra in #671
  • Addressed fixed buffer size vulnerability for vsnprintf by @nychiang in #673
  • Removed stringent -Wall and -Werror from release builds to avoid downstream compilation errors

Version 1.0.1: C++17 compatible and misc fixes

Default C++ standard remains C++14

  • Fix LLNL CI by @nychiang in #663
  • C++17 support: compile -Wall/-Werror proof by @tepperly in #653
  • Fix a bug in copying an empty matrix into a bigger matrix by @nychiang in #666
  • Fix the approach used to update mu by @nychiang in #664

Version 1.0.0: Mature solvers interfaces and execution backends

Notable new features

Interfaces of various solvers reached an equilibrium point after HiOp was interfaced with multiple optimization front-ends (e.g., power grid ACOPF and SC-ACOPF problems and topology optimization) both on CPUs and GPUs. The PriDec solver reached exascale on Frontier after minor communication optimizations. The quasi-Newton interior-point solver received a couple of updates that increase robustness. The Newton interior-point solver can fully operate on GPUs with select GPU linear solvers (CUSOLVER-LU and Gingko).

  • Instrumentation of RAJA sparse matrix class with execution spaces by @cnpetra in #589
  • Fix Assignment Typo in hiopMatrixSparseCsrCuda.cpp by @pate7 in #612
  • Use failure not failed in PNNL commit status posting by @cameronrutherford in #609
  • rebuild modules on quartz by @nychiang in #619
  • Use constraint violation in checkTermination by @nychiang in #617
  • MPI communication optimization by @rothpc in #613
  • fix memory leaks in inertia-free alg and condensed linsys by @nychiang in #622
  • Update IPM algorithm for the dense solver by @nychiang in #616
  • Use integer preprocessor macros for version information by @tepperly in #627
  • use compound vec in bicg IR by @nychiang in #621
  • Use bicg ir in the quasi-Newton solver by @nychiang in #620
  • Add support to MPI in C/Fortran examples by @nychiang in #633
  • Refactor CUSOLVER-LU module and interface by @pelesh in #634
  • Add MPI unit test for DenseEx4 by @nychiang in #644
  • Add more options to control NLP scaling by @nychiang in #649
  • Development of the feasibility restoration in the quasi-Newton solver by @nychiang in #647
  • GPU linear solver interface by @pelesh in #650

New Contributors

  • @pate7 made their first contribution in #612
  • @rothpc made their first contribution in #613

Version 0.7.2: Execution spaces abstractions and misc fixes

This release hosts a series of comprehensive internal developments and software re-engineering to improve the portability and performance on accelerators/GPU platforms. No changes to the user interface permeated under this release.

Notable new features

A new execution space abstraction is introduced to allow multiple hardware backends to run concurrently. The proposed design differentiates between "memory backend" and "execution policies" to allow using RAJA with Umpire-managed memory, RAJA with Cuda- or Hip-managed memory, RAJA with std memory, Cuda/Hip kernels with Cuda-/Hip- or Umpire-managed memory, etc.

  • Execution spaces: support for memory backends and execution policies by @cnpetra in #543
  • Build: Cuda without raja by @cnpetra in #579
  • Update of RAJA-based dense matrix to support runtime execution spaces by @cnpetra in #580
  • Reorganization of device namespace by @cnpetra in #582
  • RAJA Vector int with ExecSpace by @cnpetra in #583
  • Instrumentation of host vectors with execution spaces by @cnpetra in #584
  • Remove copy from/to device methods in vector classes by @cnpetra in #587
  • Add support for Raja with OpenMP into LLNL CI by @nychiang in #566

New vector classes using vendor-provided API were introduced and documentation was updated/improved

  • Development of hiopVectorCuda by @nychiang in #572
  • Implementation of hiopVectorHip by @nychiang in #590
  • Update user manual by @nychiang in #591
  • Update the code comments in hiopVector classes by @nychiang in #592

Refinement of triangular solver implementation for Ginkgo by @fritzgoebel in #585

Bug fixes

  • Refine the computation in normal equation system by @nychiang in #530
  • Fix static culibos issue #567 by @nychiang in #568
  • Fix segfault, remove nonsymmetric ginkgo solver by @fritzgoebel in #548
  • Calculate the inverse objective scale correctly. by @tepperly in #570
  • Fix hiopVectorRajaPar::copyToStartingAt_w_pattern by @nychiang in #569
  • Gitlab pipeline refactor by @CameronRutherford in #597

New Contributors

  • @tepperly made their first contribution in #570

Full Changelog: https://github.com/LLNL/hiop/compare/v0.7.1...v0.7.2

Version 0.7.1: Miscellaneous fixes to build system

This minor release fixes a couple of issues found in the build system after the major release 0.7 of HiOp.

Version 0.7.0: Fortran interface and misc fixes and improvements:

  • Fortran interface and examples
  • Bug fixing for sparse device linear solvers
  • Implementation of CUDA CSR matrices
  • Iterative refinement within CUSOLVER linear solver class
  • Improved robustness and performance of mixed dense-sparse solver for AMD/HIP

Version 0.6.2: Initial ginkgo solver integration and misc fixes

This tag provides an initial integration with ginko, fixes a couple of issues, and add options for (outer) iterative refinement.

Version 0.6.1: HIP linear algebra workaround and update for RAJA > v0.14 (March 31, 2022)

This version/tag provides a workaround for an issue in the HIP BLAS and updates the RAJA code to better operate with the newer versions of RAJA.

Version 0.6.0: Release of the PriDec optimization and improved GPU computations (March 31, 2022)

The salient features of v0.6.0 are

  • the release of the primal decomposition (PriDec) solver for structured two-stage problems
  • improved support for (NVIDIA) GPUs for solving sparse optimization problems via NVIDIA's cuSOLVER API and newly developed condensed optimization kernels.

Other notable capabilities include

  • improved accuracy in the computations of the search directions via Krylov-based iterative refinement
  • design of a matrix interface for sparse matrices in compressed sparse row format and (capable) CPU reference implementation

Version 0.5.4: Elastic mode, Krylov solvers, and misc bug fixes (March 2, 2022)

New algorithmic features related to the NLP solver(s) and associated linear algebra KKT systems

  • soft feasibility restoration
  • Relaxer of equality constraints at the NLP formulation level
  • Krylov interfaces and implementation for CG and BiCGStab (ready for device computations)
  • protype of the condensed linear system and initial Krylov-based iterative refinement
  • update of the Magma solver class for the latest Magma API
  • elastic mode

This release also includes several bug fixes.

Version 0.5.3: xSDK compliance (Dec 3, 2021)

xSDK compliance

Version 0.5.2: xSDK compliance and misc bug fixes (Dec 2, 2021)

  • fixed bugs in the IPM solver: gradient scaling on CUDA, unscaled objective in the user callbacks, lambda capture fix in axpy for ROCm
  • exported sparse config in cmake
  • added user options for the algorithm parameters in PriDec solver

Version 0.5.1: Objective scaling factor fix (Oct 21, 2021)

Modified the computation of the scaling factor to use the user-specified initial point

Version 0.5.0: MDS device computations, and porting of sparse kernels (Sep 30, 2021)

The salient features of this major release are

  • update of the interface to MAGMA and capability for running mixed dense-sparse (MDS) problems solely in the device memory space
  • added interface PARDISO linear solver
  • porting of the sparse linear algebra kernels to device via RAJA performance portability layer
  • various optimizations and bug fixes for the RAJA-based dense linear algebra kernels
  • Primal decomposition solver HiOp-PriDec available as a release candidate