xref: /libCEED/doc/sphinx/source/releasenotes.md (revision dfdfb19e57f3fc18ccfd612ad88e5987a5e7a393)
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### Bugfix
11
12- Fix bug in setting device id for GPU backends.
13
14### Interface changes
15
16(v0-10)=
17
18## v0.10 (Mar 16, 2022)
19
20### Interface changes
21
22- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
23- 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`.
24- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
25- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
26- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
27- Warning added when compiling OCCA backend to alert users that this backend is experimental.
28- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
29- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
30- 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`.
31- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
32- 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.
33- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
34- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
35- 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.
36- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
37- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
38
39### New features
40
41- `CeedScalar` can now be set as `float` or `double` at compile time.
42- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
43- 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.
44- Remove need to guard library headers in QFunction source for code generation backends.
45- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
46- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
47- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
48- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
49- 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.
50- Added support for element matrix assembly in GPU backends.
51
52### Maintainability
53
54- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
55- `Include-what-you-use` makefile target added as `make iwyu`.
56- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
57- Put GPU JiTed kernel source code into separate files.
58- 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.
59
60(v0-9)=
61
62## v0.9 (Jul 6, 2021)
63
64### Interface changes
65
66- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
67
68### New features
69
70- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
71- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
72- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
73- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
74
75### Performance improvements
76
77### Examples
78
79- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
80- Fluid mechanics example adds GPU support and improves modularity.
81
82### Deprecated backends
83
84- 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.
85
86(v0-8)=
87
88## v0.8 (Mar 31, 2021)
89
90### Interface changes
91
92- Error handling improved to include enumerated error codes for C interface return values.
93- 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.
94
95### New features
96
97- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
98- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
99- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
100
101### Performance improvements
102
103- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
104- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
105
106### Examples
107
108- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
109- {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.
110- {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.
111- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
112
113(v0-7)=
114
115## v0.7 (Sep 29, 2020)
116
117### Interface changes
118
119- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
120  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
121  These changes improve support for mixed finite element methods.
122- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
123- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
124- 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.
125- 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.
126- 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.
127- 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.
128  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`.
129- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
130- 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.
131
132### New features
133
134- New HIP backend: `/gpu/hip/ref`.
135- CeedQFunction support for user `CUfunction`s in some backends
136
137### Performance improvements
138
139- OCCA backend rebuilt to facilitate future performance enhancements.
140- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
141
142### Examples
143
144- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
145
146### Deprecated backends
147
148- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
149
150(v0-6)=
151
152## v0.6 (Mar 29, 2020)
153
154libCEED v0.6 contains numerous new features and examples, as well as expanded
155documentation in [this new website](https://libceed.org).
156
157### New features
158
159- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
160  1-1 correspondence with the C interface, plus some convenience features.  For instance,
161  data stored in the {cpp:type}`CeedVector` structure are available without copy as
162  {py:class}`numpy.ndarray`.  Short tutorials are provided in
163  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
164- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
165  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
166  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
167  preconditioning ingredients and are used in the libCEED's multigrid examples.
168- The inverse of separable operators can be obtained using
169  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
170  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
171  especially for Laplacians and related operators.
172- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
173  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
174- Make public accessors for various attributes to facilitate writing composable code.
175- New backend: `/cpu/self/memcheck/serial`.
176- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
177  backends.  (Single source is coming soon for OCCA backends.)
178- Fix some missing edge cases in CUDA backend.
179
180### Performance Improvements
181
182- MAGMA backend performance optimization and non-tensor bases.
183- No-copy optimization in {c:func}`CeedOperatorApply`.
184
185### Interface changes
186
187- Replace {code}`CeedElemRestrictionCreateIdentity` and
188  {code}`CeedElemRestrictionCreateBlocked` with more flexible
189  {c:func}`CeedElemRestrictionCreateStrided` and
190  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
191- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
192- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
193  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
194  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
195  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
196  {c:func}`CeedElemRestrictionApply`.
197
198### Examples
199
200libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
201Notable additions include:
202
203- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
204  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
205
206- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
207  of domains (like the cube and sphere) by direct integration on a surface mesh;
208  demonstrates geometric dimension different from topological dimension.
209
210- PETSc {ref}`example-petsc-bps`:
211
212  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
213  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
214    and transparent CUDA support.
215  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
216    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
217    dimension different from topological dimension.
218
219- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
220  solver with algebraic multigrid coarse solve.
221
222- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
223  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
224  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
225  quasi-2D computational domain support.
226
227- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
228  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
229  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
230
231(v0-5)=
232
233## v0.5 (Sep 18, 2019)
234
235For this release, several improvements were made. Two new CUDA backends were added to
236the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
237performance using single-source {ref}`CeedQFunction`. From this release, users
238can define Q-Functions in a single source code independently of the targeted backend
239with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
240compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
241declaration, the {ref}`CeedQFunction` API has undergone a slight change:
242the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
243requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
244`QFunctionField` with eval mode `CEED EVAL GRAD`.
245
246Additionally, new CPU backends
247were included in this release, such as the `/cpu/self/opt/*` backends (which are
248written in pure C and use partial **E-vectors** to improve performance) and the
249`/cpu/self/ref/memcheck` backend (which relies upon the
250[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
251{ref}`CeedQFunction` have no undefined values).
252This release also included various performance improvements, bug fixes, new examples,
253and improved tests. Among these improvements, vectorized instructions for
254{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
255instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
256identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
257to include unstructured meshes handling were. For this expansion, the prior version of
258the PETSc BPs, which only included data associated with structured geometries, were
259renamed `bpsraw`, and the new version of the BPs, which can handle data associated
260with any unstructured geometry, were called `bps`. Additionally, other benchmark
261problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
262and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
263as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
264examples. Furthermoew, another standalone libCEED example, called `ex2`, which
265computes the surface area of a given mesh was added to this release.
266
267Backends available in this release:
268
269| CEED resource (`-ceed`)  | Backend                                             |
270|--------------------------|-----------------------------------------------------|
271| `/cpu/self/ref/serial`   | Serial reference implementation                     |
272| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
273| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
274| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
275| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
276| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
277| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
278| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
279| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
280| `/cpu/occa`              | Serial OCCA kernels                                 |
281| `/gpu/occa`              | CUDA OCCA kernels                                   |
282| `/omp/occa`              | OpenMP OCCA kernels                                 |
283| `/ocl/occa`              | OpenCL OCCA kernels                                 |
284| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
285| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
286| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
287| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
288| `/gpu/magma`             | CUDA MAGMA kernels                                  |
289
290Examples available in this release:
291
292:::{list-table}
293:header-rows: 1
294:widths: auto
295* - User code
296  - Example
297* - `ceed`
298  - * ex1 (volume)
299    * ex2 (surface)
300* - `mfem`
301  - * BP1 (scalar mass operator)
302    * BP3 (scalar Laplace operator)
303* - `petsc`
304  - * BP1 (scalar mass operator)
305    * BP2 (vector mass operator)
306    * BP3 (scalar Laplace operator)
307    * BP4 (vector Laplace operator)
308    * BP5 (collocated scalar Laplace operator)
309    * BP6 (collocated vector Laplace operator)
310    * Navier-Stokes
311* - `nek5000`
312  - * BP1 (scalar mass operator)
313    * BP3 (scalar Laplace operator)
314:::
315
316(v0-4)=
317
318## v0.4 (Apr 1, 2019)
319
320libCEED v0.4 was made again publicly available in the second full CEED software
321distribution, release CEED 2.0. This release contained notable features, such as
322four new CPU backends, two new GPU backends, CPU backend optimizations, initial
323support for operator composition, performance benchmarking, and a Navier-Stokes demo.
324The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
325backends process one element at a time and are intended for meshes with a smaller number
326of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
327eight interlaced elements and are intended for meshes with higher numbers of elements.
328The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
329performance. The `/cpu/self/xsmm/*` backends rely upon the
330[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
331performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
332The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
333performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
334parallelization approach, where each thread treats a finite element. Using just in time
335compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
336backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
337achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
338deteriorates very quickly when threads run out of registers.
339
340A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
341examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
342This example solves the time-dependent Navier-Stokes equations of compressible gas
343dynamics in a static Eulerian three-dimensional frame, using structured high-order
344finite/spectral element spatial discretizations and explicit high-order time-stepping
345(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
346so that the pointwise physics (defined at quadrature points) is separated from the
347parallelization and meshing concerns.
348
349Backends available in this release:
350
351| CEED resource (`-ceed`)  | Backend                                             |
352|--------------------------|-----------------------------------------------------|
353| `/cpu/self/ref/serial`   | Serial reference implementation                     |
354| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
355| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
356| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
357| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
358| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
359| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
360| `/cpu/occa`              | Serial OCCA kernels                                 |
361| `/gpu/occa`              | CUDA OCCA kernels                                   |
362| `/omp/occa`              | OpenMP OCCA kernels                                 |
363| `/ocl/occa`              | OpenCL OCCA kernels                                 |
364| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
365| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
366| `/gpu/magma`             | CUDA MAGMA kernels                                  |
367
368Examples available in this release:
369
370:::{list-table}
371:header-rows: 1
372:widths: auto
373* - User code
374  - Example
375* - `ceed`
376  - * ex1 (volume)
377* - `mfem`
378  - * BP1 (scalar mass operator)
379    * BP3 (scalar Laplace operator)
380* - `petsc`
381  - * BP1 (scalar mass operator)
382    * BP3 (scalar Laplace operator)
383    * Navier-Stokes
384* - `nek5000`
385  - * BP1 (scalar mass operator)
386    * BP3 (scalar Laplace operator)
387:::
388
389(v0-3)=
390
391## v0.3 (Sep 30, 2018)
392
393Notable features in this release include active/passive field interface, support for
394non-tensor bases, backend optimization, and improved Fortran interface. This release
395also focused on providing improved continuous integration, and many new tests with code
396coverage reports of about 90%. This release also provided a significant change to the
397public interface: a {ref}`CeedQFunction` can take any number of named input and output
398arguments while {ref}`CeedOperator` connects them to the actual data, which may be
399supplied explicitly to `CeedOperatorApply()` (active) or separately via
400`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
401of CeedQFunctions and composition of block solvers constructed using
402{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
403used in an optimized CPU backend. Although this is typically not visible to the user,
404it enables effective use of arbitrary-length SIMD while maintaining cache locality.
405This CPU backend also implements an algebraic factorization of tensor product gradients
406to perform fewer operations than standard application of interpolation and
407differentiation from nodes to quadrature points. This algebraic formulation
408automatically supports non-polynomial and non-interpolatory bases, thus is more general
409than the more common derivation in terms of Lagrange polynomials on the quadrature points.
410
411Backends available in this release:
412
413| CEED resource (`-ceed`) | Backend                                             |
414|-------------------------|-----------------------------------------------------|
415| `/cpu/self/blocked`     | Blocked reference implementation                    |
416| `/cpu/self/ref`         | Serial reference implementation                     |
417| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
418| `/cpu/occa`             | Serial OCCA kernels                                 |
419| `/gpu/occa`             | CUDA OCCA kernels                                   |
420| `/omp/occa`             | OpenMP OCCA kernels                                 |
421| `/ocl/occa`             | OpenCL OCCA kernels                                 |
422| `/gpu/magma`            | CUDA MAGMA kernels                                  |
423
424Examples available in this release:
425
426:::{list-table}
427:header-rows: 1
428:widths: auto
429* - User code
430  - Example
431* - `ceed`
432  - * ex1 (volume)
433* - `mfem`
434  - * BP1 (scalar mass operator)
435    * BP3 (scalar Laplace operator)
436* - `petsc`
437  - * BP1 (scalar mass operator)
438    * BP3 (scalar Laplace operator)
439* - `nek5000`
440  - * BP1 (scalar mass operator)
441    * BP3 (scalar Laplace operator)
442:::
443
444(v0-21)=
445
446## v0.21 (Sep 30, 2018)
447
448A MAGMA backend (which relies upon the
449[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
450release. This initial integration set up the framework of using MAGMA and provided the
451libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
452As any other backend, the MAGMA backend provides extended basic data structures for
453{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
454the fundamental CEED building blocks to work with the new data structures.
455In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
456but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
457internally, and thus seamlessly to the user, through the functions/methods that are
458provided to support them.
459
460Backends available in this release:
461
462| CEED resource (`-ceed`) | Backend                         |
463|-------------------------|---------------------------------|
464| `/cpu/self`             | Serial reference implementation |
465| `/cpu/occa`             | Serial OCCA kernels             |
466| `/gpu/occa`             | CUDA OCCA kernels               |
467| `/omp/occa`             | OpenMP OCCA kernels             |
468| `/ocl/occa`             | OpenCL OCCA kernels             |
469| `/gpu/magma`            | CUDA MAGMA kernels              |
470
471Examples available in this release:
472
473:::{list-table}
474:header-rows: 1
475:widths: auto
476* - User code
477  - Example
478* - `ceed`
479  - * ex1 (volume)
480* - `mfem`
481  - * BP1 (scalar mass operator)
482    * BP3 (scalar Laplace operator)
483* - `petsc`
484  - * BP1 (scalar mass operator)
485* - `nek5000`
486  - * BP1 (scalar mass operator)
487:::
488
489(v0-2)=
490
491## v0.2 (Mar 30, 2018)
492
493libCEED was made publicly available the first full CEED software distribution, release
494CEED 1.0. The distribution was made available using the Spack package manager to provide
495a common, easy-to-use build environment, where the user can build the CEED distribution
496with all dependencies. This release included a new Fortran interface for the library.
497This release also contained major improvements in the OCCA backend (including a new
498`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
499compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
500`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
501of the Laplace operator) was also added to this release.
502
503Backends available in this release:
504
505| CEED resource (`-ceed`) | Backend                         |
506|-------------------------|---------------------------------|
507| `/cpu/self`             | Serial reference implementation |
508| `/cpu/occa`             | Serial OCCA kernels             |
509| `/gpu/occa`             | CUDA OCCA kernels               |
510| `/omp/occa`             | OpenMP OCCA kernels             |
511| `/ocl/occa`             | OpenCL OCCA kernels             |
512
513Examples available in this release:
514
515:::{list-table}
516:header-rows: 1
517:widths: auto
518* - User code
519  - Example
520* - `ceed`
521  - * ex1 (volume)
522* - `mfem`
523  - * BP1 (scalar mass operator)
524    * BP3 (scalar Laplace operator)
525* - `petsc`
526  - * BP1 (scalar mass operator)
527* - `nek5000`
528  - * BP1 (scalar mass operator)
529:::
530
531(v0-1)=
532
533## v0.1 (Jan 3, 2018)
534
535Initial low-level API of the CEED project. The low-level API provides a set of Finite
536Elements kernels and components for writing new low-level kernels. Examples include:
537vector and sparse linear algebra, element matrix assembly over a batch of elements,
538partial assembly and action for efficient high-order operators like mass, diffusion,
539advection, etc. The main goal of the low-level API is to establish the basis for the
540high-level API. Also, identifying such low-level kernels and providing a reference
541implementation for them serves as the basis for specialized backend implementations.
542This release contained several backends: `/cpu/self`, and backends which rely upon the
543[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
544`/gpu/occa`, and `/omp/occa`.
545It also included several examples, in the `examples` folder:
546A standalone code that shows the usage of libCEED (with no external
547dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
548(with the application of the mass operator); and a `petsc` example to perform BP1
549(with the application of the mass operator).
550
551Backends available in this release:
552
553| CEED resource (`-ceed`) | Backend                         |
554|-------------------------|---------------------------------|
555| `/cpu/self`             | Serial reference implementation |
556| `/cpu/occa`             | Serial OCCA kernels             |
557| `/gpu/occa`             | CUDA OCCA kernels               |
558| `/omp/occa`             | OpenMP OCCA kernels             |
559
560Examples available in this release:
561
562| User code             | Example                           |
563|-----------------------|-----------------------------------|
564| `ceed`                | ex1 (scalar Laplace operator)     |
565| `mfem`                | BP1 (scalar mass operator)        |
566| `petsc`               | BP1 (scalar mass operator)        |
567```
568