1# Changes/Release Notes 2 3On this page we provide a summary of the main API changes, new features and examples for each release of libCEED. 4 5(main)= 6 7## Current `main` branch 8 9### Interface changes 10 11- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output. 12 13### Bugfix 14 15- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends. 16- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`. 17 18### Examples 19 20- Added various performance enhancements for {ref}`example-petsc-navier-stokes` 21 22(v0-10-1)= 23 24## v0.10.1 (Apr 11, 2022) 25 26### Interface changes 27 28- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application. 29 30### Bugfix 31 32- Install JiT source files in install directory to fix GPU functionality for installed libCEED. 33 34(v0-10)= 35 36## v0.10 (Mar 21, 2022) 37 38### Interface changes 39 40- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields. 41- Promote to the public API: QFunction and Operator field objects, `CeedQFunctionField` and `CeedOperatorField`, and associated getters, {c:func}`CeedQFunctionGetFields`; {c:func}`CeedQFunctionFieldGetName`; {c:func}`CeedQFunctionFieldGetSize`; {c:func}`CeedQFunctionFieldGetEvalMode`; {c:func}`CeedOperatorGetFields`; {c:func}`CeedOperatorFieldGetElemRestriction`; {c:func}`CeedOperatorFieldGetBasis`; and {c:func}`CeedOperatorFieldGetVector`. 42- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added. 43- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients. 44- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces. 45- Warning added when compiling OCCA backend to alert users that this backend is experimental. 46- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`. 47- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name. 48- Clarify documentation for {c:func}`CeedVectorTakeArray`; this function will error if {c:func}`CeedVectorSetArray` with `copy_mode == CEED_USE_POINTER` was not previously called for the corresponding `CeedMemType`. 49- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`. 50- Added {c:func}`CeedQFunctionContextRegisterDouble` and {c:func}`CeedQFunctionContextRegisterInt32` with {c:func}`CeedQFunctionContextSetDouble` and {c:func}`CeedQFunctionContextSetInt32` to facilitate easy updating of {c:struct}`CeedQFunctionContext` data by user defined field names. 51- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`. 52- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums. 53- Added type `CeedSize` equivalent to `ptrdiff_t` for array sizes in {c:func}`CeedVectorCreate`, {c:func}`CeedVectorGetLength`, `CeedElemRestrictionCreate*`, {c:func}`CeedElemRestrictionGetLVectorSize`, and {c:func}`CeedOperatorLinearAssembleSymbolic`. This is a breaking change. 54- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy. 55- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator. 56 57### New features 58 59- `CeedScalar` can now be set as `float` or `double` at compile time. 60- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends. 61- Added support for JiT of QFunctions with `#include "relative/path/local-file.h"` statements for additional local files. Note that files included with `""` are searched relative to the current file first, then by compiler paths (as with `<>` includes). To use this feature, one should adhere to relative paths only, not compiler flags like `-I`, which the JiT will not be aware of. 62- Remove need to guard library headers in QFunction source for code generation backends. 63- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present. 64- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally. 65- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file. 66- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor. 67- Added {c:func}`CeedQFunctionSetContextWritable` and read-only access to `CeedQFunctionContext` data as an optional feature to improve GPU performance. By default, calling the `CeedQFunctionUser` during {c:func}`CeedQFunctionApply` is assumed to write into the `CeedQFunctionContext` data, consistent with the previous behavior. Note that if a user asserts that their `CeedQFunctionUser` does not write into the `CeedQFunctionContext` data, they are responsible for the validity of this assertion. 68- Added support for element matrix assembly in GPU backends. 69 70### Maintainability 71 72- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage. 73- `Include-what-you-use` makefile target added as `make iwyu`. 74- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase. 75- Put GPU JiTed kernel source code into separate files. 76- Dropped legacy version support in PETSc based examples to better utilize PETSc DMPlex and Mat updates to support libCEED; current minimum PETSc version for the examples is v3.17. 77 78(v0-9)= 79 80## v0.9 (Jul 6, 2021) 81 82### Interface changes 83 84- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact. 85 86### New features 87 88- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use. 89- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends. 90- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations. 91- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed). 92 93### Performance improvements 94 95### Examples 96 97- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations. 98- Fluid mechanics example adds GPU support and improves modularity. 99 100### Deprecated backends 101 102- The `/cpu/self/tmpl` and `/cpu/self/tmpl/sub` backends have been removed. These backends were intially added to test the backend inheritance mechanism, but this mechanism is now widely used and tested in multiple backends. 103 104(v0-8)= 105 106## v0.8 (Mar 31, 2021) 107 108### Interface changes 109 110- Error handling improved to include enumerated error codes for C interface return values. 111- Installed headers that will follow semantic versioning were moved to {code}`include/ceed` directory. These headers have been renamed from {code}`ceed-*.h` to {code}`ceed/*.h`. Placeholder headers with the old naming schema are currently provided, but these headers will be removed in the libCEED v0.9 release. 112 113### New features 114 115- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features. 116- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly. 117- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators. 118 119### Performance improvements 120 121- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`. 122- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`. 123 124### Examples 125 126- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions. 127- {ref}`example-petsc-elasticity` example updated with Neo-Hookean hyperelasticity in current configuration as well as improved Neo-Hookean hyperelasticity exploring storage vs computation tradeoffs. 128- {ref}`example-petsc-navier-stokes` example updated with isentropic traveling vortex test case, an analytical solution to the Euler equations that is useful for testing boundary conditions, discretization stability, and order of accuracy. 129- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree. 130 131(v0-7)= 132 133## v0.7 (Sep 29, 2020) 134 135### Interface changes 136 137- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors. 138 As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`. 139 These changes improve support for mixed finite element methods. 140- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature. 141- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity. 142- Linear Operators can be assembled as point-block diagonal matrices with {c:func}`CeedOperatorLinearAssemblePointBlockDiagonal`, provided in row-major form in a {code}`ncomp` by {code}`ncomp` block per node. 143- Diagonal assemble interface changed to accept a {ref}`CeedVector` instead of a pointer to a {ref}`CeedVector` to reduce memory movement when interfacing with calling code. 144- Added {c:func}`CeedOperatorLinearAssembleAddDiagonal` and {c:func}`CeedOperatorLinearAssembleAddPointBlockDiagonal` for improved future integration with codes such as MFEM that compose the action of {ref}`CeedOperator`s external to libCEED. 145- Added {c:func}`CeedVectorTakeAray` to sync and remove libCEED read/write access to an allocated array and pass ownership of the array to the caller. 146 This function is recommended over {c:func}`CeedVectorSyncArray` when the {code}`CeedVector` has an array owned by the caller that was set by {c:func}`CeedVectorSetArray`. 147- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory. 148- Added {c:func}`CeedOperatorMultigridLevelCreate`, {c:func}`CeedOperatorMultigridLevelCreateTensorH1`, and {c:func}`CeedOperatorMultigridLevelCreateH1` to facilitate creation of multigrid prolongation, restriction, and coarse grid operators using a common quadrature space. 149 150### New features 151 152- New HIP backend: `/gpu/hip/ref`. 153- CeedQFunction support for user `CUfunction`s in some backends 154 155### Performance improvements 156 157- OCCA backend rebuilt to facilitate future performance enhancements. 158- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`. 159 160### Examples 161 162- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions. 163 164### Deprecated backends 165 166- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`. 167 168(v0-6)= 169 170## v0.6 (Mar 29, 2020) 171 172libCEED v0.6 contains numerous new features and examples, as well as expanded 173documentation in [this new website](https://libceed.org). 174 175### New features 176 177- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly 178 1-1 correspondence with the C interface, plus some convenience features. For instance, 179 data stored in the {cpp:type}`CeedVector` structure are available without copy as 180 {py:class}`numpy.ndarray`. Short tutorials are provided in 181 [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/). 182- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point, 183 {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal 184 ({c:func}`CeedOperatorAssembleLinearDiagonal`). These operations are useful for 185 preconditioning ingredients and are used in the libCEED's multigrid examples. 186- The inverse of separable operators can be obtained using 187 {c:func}`CeedOperatorCreateFDMElementInverse` and applied with 188 {c:func}`CeedOperatorApply`. This is a useful preconditioning ingredient, 189 especially for Laplacians and related operators. 190- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`, 191 {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`. 192- Make public accessors for various attributes to facilitate writing composable code. 193- New backend: `/cpu/self/memcheck/serial`. 194- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA 195 backends. (Single source is coming soon for OCCA backends.) 196- Fix some missing edge cases in CUDA backend. 197 198### Performance Improvements 199 200- MAGMA backend performance optimization and non-tensor bases. 201- No-copy optimization in {c:func}`CeedOperatorApply`. 202 203### Interface changes 204 205- Replace {code}`CeedElemRestrictionCreateIdentity` and 206 {code}`CeedElemRestrictionCreateBlocked` with more flexible 207 {c:func}`CeedElemRestrictionCreateStrided` and 208 {c:func}`CeedElemRestrictionCreateBlockedStrided`. 209- Add arguments to {c:func}`CeedQFunctionCreateIdentity`. 210- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification 211 with {cpp:enum}`CeedInterlaceMode`. This is now an attribute of the 212 {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no 213 longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and 214 {c:func}`CeedElemRestrictionApply`. 215 216### Examples 217 218libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`. 219Notable additions include: 220 221- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of 222 a domain in 1, 2, and 3 dimensions by applying a Laplacian. 223 224- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area 225 of domains (like the cube and sphere) by direct integration on a surface mesh; 226 demonstrates geometric dimension different from topological dimension. 227 228- PETSc {ref}`example-petsc-bps`: 229 230 - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support. 231 - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements 232 and transparent CUDA support. 233 - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`): 234 generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric 235 dimension different from topological dimension. 236 237- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid 238 solver with algebraic multigrid coarse solve. 239 240- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly 241 `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`), 242 implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and 243 quasi-2D computational domain support. 244 245- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for 246 linear elasticity, small-strain hyperelasticity, and globalized finite-strain 247 hyperelasticity using p-multigrid with algebraic multigrid coarse solve. 248 249(v0-5)= 250 251## v0.5 (Sep 18, 2019) 252 253For this release, several improvements were made. Two new CUDA backends were added to 254the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art 255performance using single-source {ref}`CeedQFunction`. From this release, users 256can define Q-Functions in a single source code independently of the targeted backend 257with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU 258compilation of the user provided {ref}`CeedQFunction` code. To allow a unified 259declaration, the {ref}`CeedQFunction` API has undergone a slight change: 260the `QFunctionField` parameter `ncomp` has been changed to `size`. This change 261requires setting the previous value of `ncomp` to `ncomp*dim` when adding a 262`QFunctionField` with eval mode `CEED EVAL GRAD`. 263 264Additionally, new CPU backends 265were included in this release, such as the `/cpu/self/opt/*` backends (which are 266written in pure C and use partial **E-vectors** to improve performance) and the 267`/cpu/self/ref/memcheck` backend (which relies upon the 268[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user 269{ref}`CeedQFunction` have no undefined values). 270This release also included various performance improvements, bug fixes, new examples, 271and improved tests. Among these improvements, vectorized instructions for 272{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD` 273instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and 274identity Q-Functions were introduced, and the PETSc benchmark problems were expanded 275to include unstructured meshes handling were. For this expansion, the prior version of 276the PETSc BPs, which only included data associated with structured geometries, were 277renamed `bpsraw`, and the new version of the BPs, which can handle data associated 278with any unstructured geometry, were called `bps`. Additionally, other benchmark 279problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively), 280and BP5 and BP6 (the collocated versions---for which the quadrature points are the same 281as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc 282examples. Furthermoew, another standalone libCEED example, called `ex2`, which 283computes the surface area of a given mesh was added to this release. 284 285Backends available in this release: 286 287| CEED resource (`-ceed`) | Backend | 288|--------------------------|-----------------------------------------------------| 289| `/cpu/self/ref/serial` | Serial reference implementation | 290| `/cpu/self/ref/blocked` | Blocked reference implementation | 291| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks | 292| `/cpu/self/opt/serial` | Serial optimized C implementation | 293| `/cpu/self/opt/blocked` | Blocked optimized C implementation | 294| `/cpu/self/avx/serial` | Serial AVX implementation | 295| `/cpu/self/avx/blocked` | Blocked AVX implementation | 296| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 297| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 298| `/cpu/occa` | Serial OCCA kernels | 299| `/gpu/occa` | CUDA OCCA kernels | 300| `/omp/occa` | OpenMP OCCA kernels | 301| `/ocl/occa` | OpenCL OCCA kernels | 302| `/gpu/cuda/ref` | Reference pure CUDA kernels | 303| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 304| `/gpu/cuda/shared` | Optimized pure CUDA kernels using shared memory | 305| `/gpu/cuda/gen` | Optimized pure CUDA kernels using code generation | 306| `/gpu/magma` | CUDA MAGMA kernels | 307 308Examples available in this release: 309 310:::{list-table} 311:header-rows: 1 312:widths: auto 313* - User code 314 - Example 315* - `ceed` 316 - * ex1 (volume) 317 * ex2 (surface) 318* - `mfem` 319 - * BP1 (scalar mass operator) 320 * BP3 (scalar Laplace operator) 321* - `petsc` 322 - * BP1 (scalar mass operator) 323 * BP2 (vector mass operator) 324 * BP3 (scalar Laplace operator) 325 * BP4 (vector Laplace operator) 326 * BP5 (collocated scalar Laplace operator) 327 * BP6 (collocated vector Laplace operator) 328 * Navier-Stokes 329* - `nek5000` 330 - * BP1 (scalar mass operator) 331 * BP3 (scalar Laplace operator) 332::: 333 334(v0-4)= 335 336## v0.4 (Apr 1, 2019) 337 338libCEED v0.4 was made again publicly available in the second full CEED software 339distribution, release CEED 2.0. This release contained notable features, such as 340four new CPU backends, two new GPU backends, CPU backend optimizations, initial 341support for operator composition, performance benchmarking, and a Navier-Stokes demo. 342The new CPU backends in this release came in two families. The `/cpu/self/*/serial` 343backends process one element at a time and are intended for meshes with a smaller number 344of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of 345eight interlaced elements and are intended for meshes with higher numbers of elements. 346The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU 347performance. The `/cpu/self/xsmm/*` backends rely upon the 348[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU 349performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA. 350The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable 351performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple 352parallelization approach, where each thread treats a finite element. Using just in time 353compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this 354backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend 355achieve good peak performance for 1D, 2D, and low order 3D problems, but performance 356deteriorates very quickly when threads run out of registers. 357 358A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED 359examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`). 360This example solves the time-dependent Navier-Stokes equations of compressible gas 361dynamics in a static Eulerian three-dimensional frame, using structured high-order 362finite/spectral element spatial discretizations and explicit high-order time-stepping 363(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc, 364so that the pointwise physics (defined at quadrature points) is separated from the 365parallelization and meshing concerns. 366 367Backends available in this release: 368 369| CEED resource (`-ceed`) | Backend | 370|--------------------------|-----------------------------------------------------| 371| `/cpu/self/ref/serial` | Serial reference implementation | 372| `/cpu/self/ref/blocked` | Blocked reference implementation | 373| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 374| `/cpu/self/avx/serial` | Serial AVX implementation | 375| `/cpu/self/avx/blocked` | Blocked AVX implementation | 376| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 377| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 378| `/cpu/occa` | Serial OCCA kernels | 379| `/gpu/occa` | CUDA OCCA kernels | 380| `/omp/occa` | OpenMP OCCA kernels | 381| `/ocl/occa` | OpenCL OCCA kernels | 382| `/gpu/cuda/ref` | Reference pure CUDA kernels | 383| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 384| `/gpu/magma` | CUDA MAGMA kernels | 385 386Examples available in this release: 387 388:::{list-table} 389:header-rows: 1 390:widths: auto 391* - User code 392 - Example 393* - `ceed` 394 - * ex1 (volume) 395* - `mfem` 396 - * BP1 (scalar mass operator) 397 * BP3 (scalar Laplace operator) 398* - `petsc` 399 - * BP1 (scalar mass operator) 400 * BP3 (scalar Laplace operator) 401 * Navier-Stokes 402* - `nek5000` 403 - * BP1 (scalar mass operator) 404 * BP3 (scalar Laplace operator) 405::: 406 407(v0-3)= 408 409## v0.3 (Sep 30, 2018) 410 411Notable features in this release include active/passive field interface, support for 412non-tensor bases, backend optimization, and improved Fortran interface. This release 413also focused on providing improved continuous integration, and many new tests with code 414coverage reports of about 90%. This release also provided a significant change to the 415public interface: a {ref}`CeedQFunction` can take any number of named input and output 416arguments while {ref}`CeedOperator` connects them to the actual data, which may be 417supplied explicitly to `CeedOperatorApply()` (active) or separately via 418`CeedOperatorSetField()` (passive). This interface change enables reusable libraries 419of CeedQFunctions and composition of block solvers constructed using 420{ref}`CeedOperator`. A concept of blocked restriction was added to this release and 421used in an optimized CPU backend. Although this is typically not visible to the user, 422it enables effective use of arbitrary-length SIMD while maintaining cache locality. 423This CPU backend also implements an algebraic factorization of tensor product gradients 424to perform fewer operations than standard application of interpolation and 425differentiation from nodes to quadrature points. This algebraic formulation 426automatically supports non-polynomial and non-interpolatory bases, thus is more general 427than the more common derivation in terms of Lagrange polynomials on the quadrature points. 428 429Backends available in this release: 430 431| CEED resource (`-ceed`) | Backend | 432|-------------------------|-----------------------------------------------------| 433| `/cpu/self/blocked` | Blocked reference implementation | 434| `/cpu/self/ref` | Serial reference implementation | 435| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 436| `/cpu/occa` | Serial OCCA kernels | 437| `/gpu/occa` | CUDA OCCA kernels | 438| `/omp/occa` | OpenMP OCCA kernels | 439| `/ocl/occa` | OpenCL OCCA kernels | 440| `/gpu/magma` | CUDA MAGMA kernels | 441 442Examples available in this release: 443 444:::{list-table} 445:header-rows: 1 446:widths: auto 447* - User code 448 - Example 449* - `ceed` 450 - * ex1 (volume) 451* - `mfem` 452 - * BP1 (scalar mass operator) 453 * BP3 (scalar Laplace operator) 454* - `petsc` 455 - * BP1 (scalar mass operator) 456 * BP3 (scalar Laplace operator) 457* - `nek5000` 458 - * BP1 (scalar mass operator) 459 * BP3 (scalar Laplace operator) 460::: 461 462(v0-21)= 463 464## v0.21 (Sep 30, 2018) 465 466A MAGMA backend (which relies upon the 467[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this 468release. This initial integration set up the framework of using MAGMA and provided the 469libCEED functionality through MAGMA kernels as one of libCEED’s computational backends. 470As any other backend, the MAGMA backend provides extended basic data structures for 471{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements 472the fundamental CEED building blocks to work with the new data structures. 473In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data 474but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled 475internally, and thus seamlessly to the user, through the functions/methods that are 476provided to support them. 477 478Backends available in this release: 479 480| CEED resource (`-ceed`) | Backend | 481|-------------------------|---------------------------------| 482| `/cpu/self` | Serial reference implementation | 483| `/cpu/occa` | Serial OCCA kernels | 484| `/gpu/occa` | CUDA OCCA kernels | 485| `/omp/occa` | OpenMP OCCA kernels | 486| `/ocl/occa` | OpenCL OCCA kernels | 487| `/gpu/magma` | CUDA MAGMA kernels | 488 489Examples available in this release: 490 491:::{list-table} 492:header-rows: 1 493:widths: auto 494* - User code 495 - Example 496* - `ceed` 497 - * ex1 (volume) 498* - `mfem` 499 - * BP1 (scalar mass operator) 500 * BP3 (scalar Laplace operator) 501* - `petsc` 502 - * BP1 (scalar mass operator) 503* - `nek5000` 504 - * BP1 (scalar mass operator) 505::: 506 507(v0-2)= 508 509## v0.2 (Mar 30, 2018) 510 511libCEED was made publicly available the first full CEED software distribution, release 512CEED 1.0. The distribution was made available using the Spack package manager to provide 513a common, easy-to-use build environment, where the user can build the CEED distribution 514with all dependencies. This release included a new Fortran interface for the library. 515This release also contained major improvements in the OCCA backend (including a new 516`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to 517compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an 518`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application 519of the Laplace operator) was also added to this release. 520 521Backends available in this release: 522 523| CEED resource (`-ceed`) | Backend | 524|-------------------------|---------------------------------| 525| `/cpu/self` | Serial reference implementation | 526| `/cpu/occa` | Serial OCCA kernels | 527| `/gpu/occa` | CUDA OCCA kernels | 528| `/omp/occa` | OpenMP OCCA kernels | 529| `/ocl/occa` | OpenCL OCCA kernels | 530 531Examples available in this release: 532 533:::{list-table} 534:header-rows: 1 535:widths: auto 536* - User code 537 - Example 538* - `ceed` 539 - * ex1 (volume) 540* - `mfem` 541 - * BP1 (scalar mass operator) 542 * BP3 (scalar Laplace operator) 543* - `petsc` 544 - * BP1 (scalar mass operator) 545* - `nek5000` 546 - * BP1 (scalar mass operator) 547::: 548 549(v0-1)= 550 551## v0.1 (Jan 3, 2018) 552 553Initial low-level API of the CEED project. The low-level API provides a set of Finite 554Elements kernels and components for writing new low-level kernels. Examples include: 555vector and sparse linear algebra, element matrix assembly over a batch of elements, 556partial assembly and action for efficient high-order operators like mass, diffusion, 557advection, etc. The main goal of the low-level API is to establish the basis for the 558high-level API. Also, identifying such low-level kernels and providing a reference 559implementation for them serves as the basis for specialized backend implementations. 560This release contained several backends: `/cpu/self`, and backends which rely upon the 561[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`, 562`/gpu/occa`, and `/omp/occa`. 563It also included several examples, in the `examples` folder: 564A standalone code that shows the usage of libCEED (with no external 565dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1 566(with the application of the mass operator); and a `petsc` example to perform BP1 567(with the application of the mass operator). 568 569Backends available in this release: 570 571| CEED resource (`-ceed`) | Backend | 572|-------------------------|---------------------------------| 573| `/cpu/self` | Serial reference implementation | 574| `/cpu/occa` | Serial OCCA kernels | 575| `/gpu/occa` | CUDA OCCA kernels | 576| `/omp/occa` | OpenMP OCCA kernels | 577 578Examples available in this release: 579 580| User code | Example | 581|-----------------------|-----------------------------------| 582| `ceed` | ex1 (scalar Laplace operator) | 583| `mfem` | BP1 (scalar mass operator) | 584| `petsc` | BP1 (scalar mass operator) | 585``` 586