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