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