Organize element restriction variants in ref backend

Operator full assembly with oriented or curl-conforming element restrictions (RT or ND elements)

Address PR comment on hidden variable and clarify by renaming

Improve element restriction for H(curl) spaces by reorganizing loops, including use of int8_t for tridiagonal matrix

Update backends for unified ElemRestrictionCreate variants for all restriction types (default, oriented, strided)

Fix bugs for blocked ElemRestriction apply

enum CeedRestrictionType for CeedElemRestriction type

Element restriction for high-order (> 1) H(curl) elements requiring more general orientation transformations

Adds CeedElemRestrictionCreateCurlOriented which takes a tridiagonal element-wise transfo

Element restriction for high-order (> 1) H(curl) elements requiring more general orientation transformations

Adds CeedElemRestrictionCreateCurlOriented which takes a tridiagonal element-wise transformation matrix, typically with entries {-1, 0, 1}.

show more ...

Add sycl/gen backend and other sycl changes (#1258)

---------

Co-authored-by: Kris Rowe <kris.rowe@anl.gov>
Co-authored-by: Kris Rowe <krowe@anl.gov>
Co-authored-by: Jed Brown <jed@jedbrown.org

Add sycl/gen backend and other sycl changes (#1258)

---------

Co-authored-by: Kris Rowe <kris.rowe@anl.gov>
Co-authored-by: Kris Rowe <krowe@anl.gov>
Co-authored-by: Jed Brown <jed@jedbrown.org>
Co-authored-by: Varsha Madananth <vmadananth@uan-0002.head.cm.americas.sgi.com>
Co-authored-by: James Wright <jrwrigh.iii@gmail.com>

show more ...

Revert undesired change to ceed-ref-tensor.c

Combine ceed-avx-tensor-f64 and -f32 into a single file for all precisions

Update LIBXSMM backend (#1248)

* Fix LIBXSMM kernel generation calls after 9c0e481 in https://github.com/libxsmm/libxsmm

* Update LIBXSMM interface to work with main branch after commit 1f4cdad (

Update LIBXSMM backend (#1248)

* Fix LIBXSMM kernel generation calls after 9c0e481 in https://github.com/libxsmm/libxsmm

* Update LIBXSMM interface to work with main branch after commit 1f4cdad (in preparation for v2)

* Allow user specified BLAS_LIB for LIBXSMM dependency in Makefile

* LIBXSMM does not require kernels to be released

See https://github.com/libxsmm/libxsmm/issues/783\#issuecomment-1596655284.

* Improvements for non-tensor CPU-based CeedBasisApply for q_comp > 1

* Revert previous commit since it's faster to apply in P*Q panels, remove an unncessary LIBXSMM kernel compilation

* Remove an unused macro

* make format

* Rely on LIBXSMM to cache JIT'd kernels

* LIBXSMM dispatched kernels for xsmm/serial backend

* Combine ceed-xsmm-tensor-fp64 and -fp32 into single file for all precisions

* Address PR comments

* Update GitLab CI LIBXSMM version

show more ...

doc - fixup errors

debug - add debug color enum

memcheck - error if QFunction application gives NaNs

memcheck - use CeedSize in for loops as needed

memcheck - warn in CEED_DEBUG if NaN remains after restoring write-only access

Add headers missed by IWYU to hip-ref

Ensure initialization for new L1 norm handling in GPU backends

CUDA: improvements for handling large CeedVectors requiring CeedSize for length

Force use of CeedSize in calculating global idx

simplify logic of hipblas use for norm

HIP: Improve support for CeedVectors that are longer than the max size of 32 bit integers

Very minor backend style and whitespace fixes

Fix -fsanitize=address bugs

note: ‘snprintf’ output between 21 and 31 bytes into a destination of size 30

The assembly test had inconsistent ordering of arguments between the
qfunction itself and t

Fix -fsanitize=address bugs

note: ‘snprintf’ output between 21 and 31 bytes into a destination of size 30

The assembly test had inconsistent ordering of arguments between the
qfunction itself and the call to CeedQFunctionAddInput.

show more ...