Feature/multi rhs #1318
Merged
Feature/multi rhs #1318
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…0 seconds. This allows us to clean up various calls to saveTuneCache()
…ew function max and max_deviation
…plate - make it more useful
…rence_wrapper> conversion
…wrappers with a generic multi-rhs interface and a cleanup of the internals
…iring compiler flags
…ty for both gauge and color-spinor fields
…and writing to distinct fields. blas::ax is now just a wrapper around this, and introduce blas::axy which exposes this functionality
…Field_cref instead of ColorSpinorField_ref where the vectors are read only
…oexcept - this allows for move assignment instead of copy assignment when std::vector::resize is called
… of crefs, we would endup self recursing if passing a vector of refs: added explicit make_cset helper, to create explicit cref sets to avoid this
…es that allow us to avoid the copying that would occur if we used std::vector::resize
…cantly boosts the performance of the blas_test, removing a lot of copies
… of pointers. As part of this change, the multi-blas function interfaces have been overhauled to utilized a new specialized type vector_ref, to provide a universal interface
…nto feature/multi-rhs
…z vectors in ARPACK, not Schur vectors
…nto feature/multi-rhs
I left those intentionally, but happy to discuss their removal. Those were left to ensure that when the solver finishes, the tunecache is dumped which I naively thought was desirable. Given it would be dumped anyway when Thoughts? |
…basis used in preconditioners to Chebyshev
hummingtree
reviewed
Sep 28, 2022
mathiaswagner
approved these changes
Sep 30, 2022
hummingtree
reviewed
Oct 5, 2022
hummingtree
reviewed
Oct 5, 2022
hummingtree
reviewed
Oct 5, 2022
hummingtree
reviewed
Oct 5, 2022
hummingtree
reviewed
Oct 5, 2022
hummingtree
approved these changes
Oct 5, 2022
weinbe2
approved these changes
Oct 17, 2022
|
Merging this now. I've verified manually:
I've updated the description with a few late additions to this PR that were added during review. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR is focussed on framework evolution to get us ready for multi-RHS workflows
vector_refclass which replaces the prior use ofstd::vector<ColorSpinorField_ref>DiracandDiracMatrixclassesstd::vector<ColorSpinorField>instead ofstd::vector<ColorSpinorField*>FieldTmpwhich provides a cache for temporary objects. Deployed in theDiracoperators to negate the need to allocate and maintain explicit temporary vectorsmaxandmax_deviationreduce kernels, useful for correctness testingio_test, for testing both gauge and color-spinor storage when QIO is enabledconstcorrect, and doxygen is now added for all functions.Update
Some additional fixes added during review:
max_res_increasedefault parameter to ensure more reliable convergence in ctestQUDA_INVERFACE_NVTX=ON