Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[oneMKL][DFT] Full revision of DFT parts of specs + addition of {F,B}…
…WD_STRIDES configuration parameters (#506) * [DFT] removing REAL_COMPLEX and CCE_FORMAT from the list of config_value as 1) this is the only appearance of REAL_COMPLEX found in the sources (the specs lack much clarification about it if we keep it there), 2) the changes targeted herein - addition of {F,B}WD_STRIDES - would result in even more nightmare-ish additions of other formats than CCE_FORMAT (and clarifications thereof) that would be supported only for 1D real transfomrs on CPU devices by closed-source oneMKL anyways... Also removing config_param::PACKED_FORMAT as its very existence is questionable (and confusing imo) if CCE is the only format registered in the specs for conjugate-even complex sequences anyways * [DFT] Revisions to the landing page, introducing and generalizing notations to be used thereafter * [DFT] Revisions to descriptor page + introduction of jargon therein to refer to it thereafter. * [DFT] Revision to the page for scoped enumeration types + removing the page about {FOR|BACK}WARD_SCALE as it doesn't add any relevant information that is not already specified therein * [DFT] Revisions to the page for data layout configuration and integrating all specs pertaining to batched DFTs therein to make it self-contained (removing number_of_transforms.rst) * [DFT] specializing compute_forward call in REAL_REAL example from storage_formats to avoid ambiguity. * [DFT] removing WORKSPACE configuration parameter and corresponding values --------- Co-authored-by: Parks, Helen <helen.parks@intel.com> Co-authored-by: Finlay <finlay.marno@codeplay.com>
- Loading branch information