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