This is the v1.0.0 release of libCEED. In accordance with SEMVAR, future releases that break the API or ABI will cause a major version increment. The changes from v0.12 are summarized below:

New features

• Add CeedOperatorCreateAtPoints which evaluates the CeedQFunction at arbitrary locations in each element, for use in Particle in Cell, Material Point Method, and similar methods.
• Add CeedElemRestrictionGetLLayout to provide L-vector layout for strided CeedElemRestriction created with CEED_BACKEND_STRIDES.
• Add CeedVectorReturnCeed and similar when parent Ceed context for a libCEED object is only needed once in a calling scope.
• Enable #pragma once for all JiT source; remove duplicate includes in JiT source string before compilation.
• Allow user to set additional compiler options for CUDA and HIP JiT. Specifically, directories set with CeedAddJitSourceRoot(ceed, "foo/bar") will be used to set -Ifoo/bar and defines set with CeedAddJitDefine(ceed, "foo=bar") will be used to set -Dfoo=bar.
• Added non-tensor basis support to code generation backends /gpu/cuda/gen and /gpu/hip/gen.
• Added support to code generation backends /gpu/cuda/gen and /gpu/hip/gen for operators with both tensor and non-tensor bases.
• Add CeedGetGitVersion() to access the Git commit and dirty state of the repository at build time.
• Add CeedGetBuildConfiguration() to access compilers, flags, and related information about the build environment.
• Add support for full CeedOperator assembly for operators with multiple active fields with different bases for CPU backends and /gpu/cuda/ref and /gpu/hip/gen backends.
• Add CeedVectorFilter to zero out components of a CeedVector that have absolute value below a specified threshold value.
• Add CeedOperatorSetEnableCudaGraph for CUDA Graph capture/replay on /gpu/cuda/gen composite operators. Enabled by default; use CEED_CUDA_ENABLE_GRAPH=0 to turn off.
• Add string names of ElemRestrictionTypes for human-readable output.
• Add /cpu/self/sve/serial and /cpu/self/sve/blocked vector-length-agnostic ARM SVE CPU backends.
• Add /cpu/self/gen/serial and /cpu/self/gen/blocked CPU just-in-time (JiT) compiled backends, inspired by the CUDA and ROCm JiT complied backends. JiT compiler and optimizer options can be set at compile or runtime with the environment variables CEED_CPU_JIT_CXX and CEED_CPU_JIT_OPT, respectively.
• Add unified environment variable interface. Supported variables are shown with CeedView() and can be get/set programmatically with CeedGet*() functions.
• Add even-odd (centro-symmetry) decomposition for tensor contractions, which halves the arithmetic when the 1D basis matrix is centro-symmetric. This reorders the summation, so results move by about an ulp; use CeedBasisSetUseEvenOdd per basis or the CEED_BASIS_USE_EVEN_ODD environment variable to turn it off.
• Add support for user-defined cache directory for /cpu/self/gen and CUDA Clang compilation pipelines. Defaults to $CEED_CACHE_BASE_DIR, $XDG_CACHE_HOME, or $HOME/.cache, in that order, and can be accessed and set programmatically via CeedGetCacheBaseDir and CeedSetCacheBaseDir, respectively.
• Add CeedElemRestrictionGetBlockedElemRestriction to prevent duplicate blocked CeedElemRestriction creation in CPU backends.
• Add /cpu/self/sme/serial and /cpu/self/sme/blocked ARM Scalable Matrix Extension (SME) CPU backends with SME outer-product instructions for tensor contractions.

Interface changes

• Add bool field type for CeedQFunctionContext and related interfaces to use bool fields.
• CEED_BASIS_COLLOCATED removed; users should only use CEED_BASIS_NONE.
• Remove unneeded pointer for CeedElemRestrictionGetELayout.
• Change QFunction source include file handling in JiT compilers
  • Add CEED_RUNNING_JIT_PASS compiler definition for wrapping header files that device JiT compilers cannot read
  • Users should now prefer #include <ceed/types.h> rather than #include <ceed.h> in QFunction source files
• Require use of Ceed*Destroy() on Ceed objects returned from Ceed*Get*().
• Rename CeedCompositeOperatorCreate() to CeedOperatorCreateComposite() for uniformity.
• Rename CeedCompositeOperator*() to CeedOperatorComposite*() for uniformity.
• Add build_objects parameter to CeedOperatorLinearAssembleQFunctionBuildOrUpdateFallback to allow for passing uninitialized vectors and restrictions
• Move JiT helper functions only required by SYCL backends to ceed/jit-tools-deprecated.h. These functions will be removed when the SYCL backends are updated to reflect the improvements in the CUDA and HIP backends.
• Add CeedOperatorLinearAssembleGetFlopsEstimate, CeedOperatorLinearAssembleDiagonalGetFlopsEstimate, and CeedOperatorLinearAssemblePointBlockDiagonalGetFlopsEstimate to estimate FLOPs for linear operator assembly.
• Add CeedBasisGetChebyshevData to access interpolation data data from DoFs to Chebyshev polynomials and reduce duplicated code.
• Add CeedBasisHasCollocatedGrad to check whether collocated gradient is suppported for a CeedBasis.
• Rename CeedBasisGetCollocatedGrad to CeedBasisGetCollocatedGrad1D and change signature to take const CeedScalar ** like other getters; the output array is set to an internally cached array.

Examples

• Add deal.II example with CEED BP suite.

Maintainability

• OCCA backends were retired.
• Use clang-tidy to automatically fix if-else blocks which are missing braces.
• Fixed bugs with CeedRegistryGetList() and clarified internals.
• All environment variables are now defined in a central header: include/ceed/ceed-env-list.h. This header is used to generate all code needed to set and get the environment variables.

Documentation PDF: libCEED.pdf (github.com)