Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Support for Homeogeneous Self-Dual embeddings #93

Open
poulson opened this issue Aug 19, 2015 · 0 comments
Open

Support for Homeogeneous Self-Dual embeddings #93

poulson opened this issue Aug 19, 2015 · 0 comments

Comments

@poulson
Copy link
Member

poulson commented Aug 19, 2015

Elemental is still lacking support for detecting primal and/or dual infeasibility in its IPMs, and the appropriate response should be to make use of a Homogeneous Self-Dual embeddings, at least for LPs and SOCPs.

Issue #92 is somewhat of a dependency before transitioning to HSD solvers as the default.

Also, it should be verified that the number of iterations and overall solution time typically decreases with the HSD formulation.
benson31 added a commit to benson31/Elemental that referenced this issue Jan 30, 2020
@benson31
Fix lbann half issues (elemental#93)
7d581e9 
* tick up version to 1.3.3

* fix issues with CPU half

* rework the copy interface

* work on Copy

* add a warmup run to the gemm test

* Various updates to copy dispatch; no more ETI for Copy

* fix an issue with the cuda half type's assignment operators

* make gpu_half_type assignment operators into templates

* Update include/El/blas_like/level1/CopyLocal.hpp

Co-Authored-By: Tim Moon <[email protected]>

* remove some debugging output

* add decls for BaseDistMatrix copy and copyasync

* make gpu_half_type streamable

* be a little more clever about casting to __half

* fix some things

* add overloads of sqrt and pow for half types

* add unary minus for gpu half type

* fix an issue where NVCC decided that rvalue references have value semantics

* add an overload of Log for gpu_half_type

* add exception-throwing bitwise operators for half. This is to appease Aluminum at compile-time and should never be encounted IRL.

* add a bunch of missing library symbols

* add a bunch of transendental functions, etc, for half types

* fix some missing symbols when compiling without half support

* Fixes to instantiate Read/Write with gpu half type

* Add overload for instantiate with half types

* add a few missing symbols

* add a write impl for gpu half matrices

* temp: dispatch gemv through Gemm for __half

* Fix the GEMV as GEMM call

* temporary error-throw for unhandled case

* remove the incy != 1 gemv case

* patch around an issue in the Half library

* expose AbstractMatrix interface to Print()

Co-authored-by: Tim Moon <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@poulson