xref: /libCEED/doc/sphinx/source/releasenotes.md (revision d34e270f91eeccdaafc024e5afc9304ed3811350)
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
26### New features
27
28- `CeedScalar` can now be set as `float` or `double` at compile time.
29- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
30- 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.
31- Remove need to guard library headers in QFunction source for code generation backends.
32- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
33- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
34- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
35- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
36- 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.
37
38### Maintainability
39
40- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
41- `Include-what-you-use` makefile target added as `make iwyu`.
42- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
43- Put GPU JiTed kernel source code into separate files.
44- 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.
45
46(v0-9)=
47
48## v0.9 (Jul 6, 2021)
49
50### Interface changes
51
52- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
53
54### New features
55
56- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
57- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
58- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
59- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
60
61### Performance improvements
62
63### Examples
64
65- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
66- Fluid mechanics example adds GPU support and improves modularity.
67
68### Deprecated backends
69
70- 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.
71
72(v0-8)=
73
74## v0.8 (Mar 31, 2021)
75
76### Interface changes
77
78- Error handling improved to include enumerated error codes for C interface return values.
79- 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.
80
81### New features
82
83- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
84- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
85- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
86
87### Performance improvements
88
89- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
90- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
91
92### Examples
93
94- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
95- {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.
96- {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.
97- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
98
99(v0-7)=
100
101## v0.7 (Sep 29, 2020)
102
103### Interface changes
104
105- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
106  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
107  These changes improve support for mixed finite element methods.
108- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
109- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
110- 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.
111- 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.
112- 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.
113- 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.
114  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`.
115- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
116- 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.
117
118### New features
119
120- New HIP backend: `/gpu/hip/ref`.
121- CeedQFunction support for user `CUfunction`s in some backends
122
123### Performance improvements
124
125- OCCA backend rebuilt to facilitate future performance enhancements.
126- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
127
128### Examples
129
130- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
131
132### Deprecated backends
133
134- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
135
136(v0-6)=
137
138## v0.6 (Mar 29, 2020)
139
140libCEED v0.6 contains numerous new features and examples, as well as expanded
141documentation in [this new website](https://libceed.readthedocs.io).
142
143### New features
144
145- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
146  1-1 correspondence with the C interface, plus some convenience features.  For instance,
147  data stored in the {cpp:type}`CeedVector` structure are available without copy as
148  {py:class}`numpy.ndarray`.  Short tutorials are provided in
149  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
150- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
151  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
152  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
153  preconditioning ingredients and are used in the libCEED's multigrid examples.
154- The inverse of separable operators can be obtained using
155  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
156  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
157  especially for Laplacians and related operators.
158- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
159  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
160- Make public accessors for various attributes to facilitate writing composable code.
161- New backend: `/cpu/self/memcheck/serial`.
162- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
163  backends.  (Single source is coming soon for OCCA backends.)
164- Fix some missing edge cases in CUDA backend.
165
166### Performance Improvements
167
168- MAGMA backend performance optimization and non-tensor bases.
169- No-copy optimization in {c:func}`CeedOperatorApply`.
170
171### Interface changes
172
173- Replace {code}`CeedElemRestrictionCreateIdentity` and
174  {code}`CeedElemRestrictionCreateBlocked` with more flexible
175  {c:func}`CeedElemRestrictionCreateStrided` and
176  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
177- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
178- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
179  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
180  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
181  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
182  {c:func}`CeedElemRestrictionApply`.
183
184### Examples
185
186libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
187Notable additions include:
188
189- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
190  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
191
192- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
193  of domains (like the cube and sphere) by direct integration on a surface mesh;
194  demonstrates geometric dimension different from topological dimension.
195
196- PETSc {ref}`example-petsc-bps`:
197
198  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
199  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
200    and transparent CUDA support.
201  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
202    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
203    dimension different from topological dimension.
204
205- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
206  solver with algebraic multigrid coarse solve.
207
208- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
209  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
210  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
211  quasi-2D computational domain support.
212
213- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
214  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
215  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
216
217(v0-5)=
218
219## v0.5 (Sep 18, 2019)
220
221For this release, several improvements were made. Two new CUDA backends were added to
222the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
223performance using single-source {ref}`CeedQFunction`. From this release, users
224can define Q-Functions in a single source code independently of the targeted backend
225with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
226compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
227declaration, the {ref}`CeedQFunction` API has undergone a slight change:
228the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
229requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
230`QFunctionField` with eval mode `CEED EVAL GRAD`.
231
232Additionally, new CPU backends
233were included in this release, such as the `/cpu/self/opt/*` backends (which are
234written in pure C and use partial **E-vectors** to improve performance) and the
235`/cpu/self/ref/memcheck` backend (which relies upon the
236[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
237{ref}`CeedQFunction` have no undefined values).
238This release also included various performance improvements, bug fixes, new examples,
239and improved tests. Among these improvements, vectorized instructions for
240{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
241instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
242identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
243to include unstructured meshes handling were. For this expansion, the prior version of
244the PETSc BPs, which only included data associated with structured geometries, were
245renamed `bpsraw`, and the new version of the BPs, which can handle data associated
246with any unstructured geometry, were called `bps`. Additionally, other benchmark
247problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
248and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
249as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
250examples. Furthermoew, another standalone libCEED example, called `ex2`, which
251computes the surface area of a given mesh was added to this release.
252
253Backends available in this release:
254
255| CEED resource (`-ceed`)  | Backend                                             |
256|--------------------------|-----------------------------------------------------|
257| `/cpu/self/ref/serial`   | Serial reference implementation                     |
258| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
259| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
260| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
261| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
262| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
263| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
264| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
265| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
266| `/cpu/occa`              | Serial OCCA kernels                                 |
267| `/gpu/occa`              | CUDA OCCA kernels                                   |
268| `/omp/occa`              | OpenMP OCCA kernels                                 |
269| `/ocl/occa`              | OpenCL OCCA kernels                                 |
270| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
271| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
272| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
273| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
274| `/gpu/magma`             | CUDA MAGMA kernels                                  |
275
276Examples available in this release:
277
278:::{list-table}
279:header-rows: 1
280:widths: auto
281* - User code
282  - Example
283* - `ceed`
284  - * ex1 (volume)
285    * ex2 (surface)
286* - `mfem`
287  - * BP1 (scalar mass operator)
288    * BP3 (scalar Laplace operator)
289* - `petsc`
290  - * BP1 (scalar mass operator)
291    * BP2 (vector mass operator)
292    * BP3 (scalar Laplace operator)
293    * BP4 (vector Laplace operator)
294    * BP5 (collocated scalar Laplace operator)
295    * BP6 (collocated vector Laplace operator)
296    * Navier-Stokes
297* - `nek5000`
298  - * BP1 (scalar mass operator)
299    * BP3 (scalar Laplace operator)
300:::
301
302(v0-4)=
303
304## v0.4 (Apr 1, 2019)
305
306libCEED v0.4 was made again publicly available in the second full CEED software
307distribution, release CEED 2.0. This release contained notable features, such as
308four new CPU backends, two new GPU backends, CPU backend optimizations, initial
309support for operator composition, performance benchmarking, and a Navier-Stokes demo.
310The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
311backends process one element at a time and are intended for meshes with a smaller number
312of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
313eight interlaced elements and are intended for meshes with higher numbers of elements.
314The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
315performance. The `/cpu/self/xsmm/*` backends rely upon the
316[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
317performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
318The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
319performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
320parallelization approach, where each thread treats a finite element. Using just in time
321compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
322backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
323achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
324deteriorates very quickly when threads run out of registers.
325
326A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
327examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
328This example solves the time-dependent Navier-Stokes equations of compressible gas
329dynamics in a static Eulerian three-dimensional frame, using structured high-order
330finite/spectral element spatial discretizations and explicit high-order time-stepping
331(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
332so that the pointwise physics (defined at quadrature points) is separated from the
333parallelization and meshing concerns.
334
335Backends available in this release:
336
337| CEED resource (`-ceed`)  | Backend                                             |
338|--------------------------|-----------------------------------------------------|
339| `/cpu/self/ref/serial`   | Serial reference implementation                     |
340| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
341| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
342| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
343| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
344| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
345| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
346| `/cpu/occa`              | Serial OCCA kernels                                 |
347| `/gpu/occa`              | CUDA OCCA kernels                                   |
348| `/omp/occa`              | OpenMP OCCA kernels                                 |
349| `/ocl/occa`              | OpenCL OCCA kernels                                 |
350| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
351| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
352| `/gpu/magma`             | CUDA MAGMA kernels                                  |
353
354Examples available in this release:
355
356:::{list-table}
357:header-rows: 1
358:widths: auto
359* - User code
360  - Example
361* - `ceed`
362  - * ex1 (volume)
363* - `mfem`
364  - * BP1 (scalar mass operator)
365    * BP3 (scalar Laplace operator)
366* - `petsc`
367  - * BP1 (scalar mass operator)
368    * BP3 (scalar Laplace operator)
369    * Navier-Stokes
370* - `nek5000`
371  - * BP1 (scalar mass operator)
372    * BP3 (scalar Laplace operator)
373:::
374
375(v0-3)=
376
377## v0.3 (Sep 30, 2018)
378
379Notable features in this release include active/passive field interface, support for
380non-tensor bases, backend optimization, and improved Fortran interface. This release
381also focused on providing improved continuous integration, and many new tests with code
382coverage reports of about 90%. This release also provided a significant change to the
383public interface: a {ref}`CeedQFunction` can take any number of named input and output
384arguments while {ref}`CeedOperator` connects them to the actual data, which may be
385supplied explicitly to `CeedOperatorApply()` (active) or separately via
386`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
387of CeedQFunctions and composition of block solvers constructed using
388{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
389used in an optimized CPU backend. Although this is typically not visible to the user,
390it enables effective use of arbitrary-length SIMD while maintaining cache locality.
391This CPU backend also implements an algebraic factorization of tensor product gradients
392to perform fewer operations than standard application of interpolation and
393differentiation from nodes to quadrature points. This algebraic formulation
394automatically supports non-polynomial and non-interpolatory bases, thus is more general
395than the more common derivation in terms of Lagrange polynomials on the quadrature points.
396
397Backends available in this release:
398
399| CEED resource (`-ceed`) | Backend                                             |
400|-------------------------|-----------------------------------------------------|
401| `/cpu/self/blocked`     | Blocked reference implementation                    |
402| `/cpu/self/ref`         | Serial reference implementation                     |
403| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
404| `/cpu/occa`             | Serial OCCA kernels                                 |
405| `/gpu/occa`             | CUDA OCCA kernels                                   |
406| `/omp/occa`             | OpenMP OCCA kernels                                 |
407| `/ocl/occa`             | OpenCL OCCA kernels                                 |
408| `/gpu/magma`            | CUDA MAGMA kernels                                  |
409
410Examples available in this release:
411
412:::{list-table}
413:header-rows: 1
414:widths: auto
415* - User code
416  - Example
417* - `ceed`
418  - * ex1 (volume)
419* - `mfem`
420  - * BP1 (scalar mass operator)
421    * BP3 (scalar Laplace operator)
422* - `petsc`
423  - * BP1 (scalar mass operator)
424    * BP3 (scalar Laplace operator)
425* - `nek5000`
426  - * BP1 (scalar mass operator)
427    * BP3 (scalar Laplace operator)
428:::
429
430(v0-21)=
431
432## v0.21 (Sep 30, 2018)
433
434A MAGMA backend (which relies upon the
435[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
436release. This initial integration set up the framework of using MAGMA and provided the
437libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
438As any other backend, the MAGMA backend provides extended basic data structures for
439{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
440the fundamental CEED building blocks to work with the new data structures.
441In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
442but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
443internally, and thus seamlessly to the user, through the functions/methods that are
444provided to support them.
445
446Backends available in this release:
447
448| CEED resource (`-ceed`) | Backend                         |
449|-------------------------|---------------------------------|
450| `/cpu/self`             | Serial reference implementation |
451| `/cpu/occa`             | Serial OCCA kernels             |
452| `/gpu/occa`             | CUDA OCCA kernels               |
453| `/omp/occa`             | OpenMP OCCA kernels             |
454| `/ocl/occa`             | OpenCL OCCA kernels             |
455| `/gpu/magma`            | CUDA MAGMA kernels              |
456
457Examples available in this release:
458
459:::{list-table}
460:header-rows: 1
461:widths: auto
462* - User code
463  - Example
464* - `ceed`
465  - * ex1 (volume)
466* - `mfem`
467  - * BP1 (scalar mass operator)
468    * BP3 (scalar Laplace operator)
469* - `petsc`
470  - * BP1 (scalar mass operator)
471* - `nek5000`
472  - * BP1 (scalar mass operator)
473:::
474
475(v0-2)=
476
477## v0.2 (Mar 30, 2018)
478
479libCEED was made publicly available the first full CEED software distribution, release
480CEED 1.0. The distribution was made available using the Spack package manager to provide
481a common, easy-to-use build environment, where the user can build the CEED distribution
482with all dependencies. This release included a new Fortran interface for the library.
483This release also contained major improvements in the OCCA backend (including a new
484`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
485compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
486`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
487of the Laplace operator) was also added to this release.
488
489Backends available in this release:
490
491| CEED resource (`-ceed`) | Backend                         |
492|-------------------------|---------------------------------|
493| `/cpu/self`             | Serial reference implementation |
494| `/cpu/occa`             | Serial OCCA kernels             |
495| `/gpu/occa`             | CUDA OCCA kernels               |
496| `/omp/occa`             | OpenMP OCCA kernels             |
497| `/ocl/occa`             | OpenCL OCCA kernels             |
498
499Examples available in this release:
500
501:::{list-table}
502:header-rows: 1
503:widths: auto
504* - User code
505  - Example
506* - `ceed`
507  - * ex1 (volume)
508* - `mfem`
509  - * BP1 (scalar mass operator)
510    * BP3 (scalar Laplace operator)
511* - `petsc`
512  - * BP1 (scalar mass operator)
513* - `nek5000`
514  - * BP1 (scalar mass operator)
515:::
516
517(v0-1)=
518
519## v0.1 (Jan 3, 2018)
520
521Initial low-level API of the CEED project. The low-level API provides a set of Finite
522Elements kernels and components for writing new low-level kernels. Examples include:
523vector and sparse linear algebra, element matrix assembly over a batch of elements,
524partial assembly and action for efficient high-order operators like mass, diffusion,
525advection, etc. The main goal of the low-level API is to establish the basis for the
526high-level API. Also, identifying such low-level kernels and providing a reference
527implementation for them serves as the basis for specialized backend implementations.
528This release contained several backends: `/cpu/self`, and backends which rely upon the
529[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
530`/gpu/occa`, and `/omp/occa`.
531It also included several examples, in the `examples` folder:
532A standalone code that shows the usage of libCEED (with no external
533dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
534(with the application of the mass operator); and a `petsc` example to perform BP1
535(with the application of the mass operator).
536
537Backends available in this release:
538
539| CEED resource (`-ceed`) | Backend                         |
540|-------------------------|---------------------------------|
541| `/cpu/self`             | Serial reference implementation |
542| `/cpu/occa`             | Serial OCCA kernels             |
543| `/gpu/occa`             | CUDA OCCA kernels               |
544| `/omp/occa`             | OpenMP OCCA kernels             |
545
546Examples available in this release:
547
548| User code             | Example                           |
549|-----------------------|-----------------------------------|
550| `ceed`                | ex1 (scalar Laplace operator)     |
551| `mfem`                | BP1 (scalar mass operator)        |
552| `petsc`               | BP1 (scalar mass operator)        |
553```
554