xref: /libCEED/doc/sphinx/source/releasenotes.md (revision a71fcd9fac4e7a8dfa69a197fd7b41b8f31fd6a3)
1# Changes/Release Notes
2
3On this page we provide a summary of the main API changes, new features and examples for each release of libCEED.
4
5(main)=
6
7## Current `main` branch
8
9### Interface changes
10
11(v0-10-1)=
12
13## v0.10.1 (Apr 11, 2022)
14
15### Interface changes
16
17- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
18
19### Bugfix
20
21- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
22
23(v0-10)=
24
25## v0.10 (Mar 21, 2022)
26
27### Interface changes
28
29- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
30- 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`.
31- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
32- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
33- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
34- Warning added when compiling OCCA backend to alert users that this backend is experimental.
35- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
36- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
37- 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`.
38- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
39- 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.
40- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
41- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
42- 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.
43- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
44- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
45
46### New features
47
48- `CeedScalar` can now be set as `float` or `double` at compile time.
49- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
50- 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.
51- Remove need to guard library headers in QFunction source for code generation backends.
52- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
53- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
54- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
55- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
56- 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.
57- Added support for element matrix assembly in GPU backends.
58
59### Maintainability
60
61- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
62- `Include-what-you-use` makefile target added as `make iwyu`.
63- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
64- Put GPU JiTed kernel source code into separate files.
65- 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.
66
67(v0-9)=
68
69## v0.9 (Jul 6, 2021)
70
71### Interface changes
72
73- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
74
75### New features
76
77- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
78- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
79- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
80- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
81
82### Performance improvements
83
84### Examples
85
86- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
87- Fluid mechanics example adds GPU support and improves modularity.
88
89### Deprecated backends
90
91- 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.
92
93(v0-8)=
94
95## v0.8 (Mar 31, 2021)
96
97### Interface changes
98
99- Error handling improved to include enumerated error codes for C interface return values.
100- 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.
101
102### New features
103
104- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
105- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
106- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
107
108### Performance improvements
109
110- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
111- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
112
113### Examples
114
115- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
116- {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.
117- {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.
118- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
119
120(v0-7)=
121
122## v0.7 (Sep 29, 2020)
123
124### Interface changes
125
126- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
127  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
128  These changes improve support for mixed finite element methods.
129- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
130- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
131- 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.
132- 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.
133- 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.
134- 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.
135  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`.
136- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
137- 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.
138
139### New features
140
141- New HIP backend: `/gpu/hip/ref`.
142- CeedQFunction support for user `CUfunction`s in some backends
143
144### Performance improvements
145
146- OCCA backend rebuilt to facilitate future performance enhancements.
147- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
148
149### Examples
150
151- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
152
153### Deprecated backends
154
155- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
156
157(v0-6)=
158
159## v0.6 (Mar 29, 2020)
160
161libCEED v0.6 contains numerous new features and examples, as well as expanded
162documentation in [this new website](https://libceed.org).
163
164### New features
165
166- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
167  1-1 correspondence with the C interface, plus some convenience features.  For instance,
168  data stored in the {cpp:type}`CeedVector` structure are available without copy as
169  {py:class}`numpy.ndarray`.  Short tutorials are provided in
170  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
171- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
172  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
173  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
174  preconditioning ingredients and are used in the libCEED's multigrid examples.
175- The inverse of separable operators can be obtained using
176  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
177  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
178  especially for Laplacians and related operators.
179- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
180  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
181- Make public accessors for various attributes to facilitate writing composable code.
182- New backend: `/cpu/self/memcheck/serial`.
183- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
184  backends.  (Single source is coming soon for OCCA backends.)
185- Fix some missing edge cases in CUDA backend.
186
187### Performance Improvements
188
189- MAGMA backend performance optimization and non-tensor bases.
190- No-copy optimization in {c:func}`CeedOperatorApply`.
191
192### Interface changes
193
194- Replace {code}`CeedElemRestrictionCreateIdentity` and
195  {code}`CeedElemRestrictionCreateBlocked` with more flexible
196  {c:func}`CeedElemRestrictionCreateStrided` and
197  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
198- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
199- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
200  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
201  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
202  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
203  {c:func}`CeedElemRestrictionApply`.
204
205### Examples
206
207libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
208Notable additions include:
209
210- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
211  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
212
213- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
214  of domains (like the cube and sphere) by direct integration on a surface mesh;
215  demonstrates geometric dimension different from topological dimension.
216
217- PETSc {ref}`example-petsc-bps`:
218
219  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
220  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
221    and transparent CUDA support.
222  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
223    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
224    dimension different from topological dimension.
225
226- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
227  solver with algebraic multigrid coarse solve.
228
229- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
230  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
231  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
232  quasi-2D computational domain support.
233
234- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
235  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
236  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
237
238(v0-5)=
239
240## v0.5 (Sep 18, 2019)
241
242For this release, several improvements were made. Two new CUDA backends were added to
243the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
244performance using single-source {ref}`CeedQFunction`. From this release, users
245can define Q-Functions in a single source code independently of the targeted backend
246with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
247compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
248declaration, the {ref}`CeedQFunction` API has undergone a slight change:
249the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
250requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
251`QFunctionField` with eval mode `CEED EVAL GRAD`.
252
253Additionally, new CPU backends
254were included in this release, such as the `/cpu/self/opt/*` backends (which are
255written in pure C and use partial **E-vectors** to improve performance) and the
256`/cpu/self/ref/memcheck` backend (which relies upon the
257[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
258{ref}`CeedQFunction` have no undefined values).
259This release also included various performance improvements, bug fixes, new examples,
260and improved tests. Among these improvements, vectorized instructions for
261{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
262instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
263identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
264to include unstructured meshes handling were. For this expansion, the prior version of
265the PETSc BPs, which only included data associated with structured geometries, were
266renamed `bpsraw`, and the new version of the BPs, which can handle data associated
267with any unstructured geometry, were called `bps`. Additionally, other benchmark
268problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
269and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
270as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
271examples. Furthermoew, another standalone libCEED example, called `ex2`, which
272computes the surface area of a given mesh was added to this release.
273
274Backends available in this release:
275
276| CEED resource (`-ceed`)  | Backend                                             |
277|--------------------------|-----------------------------------------------------|
278| `/cpu/self/ref/serial`   | Serial reference implementation                     |
279| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
280| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
281| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
282| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
283| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
284| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
285| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
286| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
287| `/cpu/occa`              | Serial OCCA kernels                                 |
288| `/gpu/occa`              | CUDA OCCA kernels                                   |
289| `/omp/occa`              | OpenMP OCCA kernels                                 |
290| `/ocl/occa`              | OpenCL OCCA kernels                                 |
291| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
292| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
293| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
294| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
295| `/gpu/magma`             | CUDA MAGMA kernels                                  |
296
297Examples available in this release:
298
299:::{list-table}
300:header-rows: 1
301:widths: auto
302* - User code
303  - Example
304* - `ceed`
305  - * ex1 (volume)
306    * ex2 (surface)
307* - `mfem`
308  - * BP1 (scalar mass operator)
309    * BP3 (scalar Laplace operator)
310* - `petsc`
311  - * BP1 (scalar mass operator)
312    * BP2 (vector mass operator)
313    * BP3 (scalar Laplace operator)
314    * BP4 (vector Laplace operator)
315    * BP5 (collocated scalar Laplace operator)
316    * BP6 (collocated vector Laplace operator)
317    * Navier-Stokes
318* - `nek5000`
319  - * BP1 (scalar mass operator)
320    * BP3 (scalar Laplace operator)
321:::
322
323(v0-4)=
324
325## v0.4 (Apr 1, 2019)
326
327libCEED v0.4 was made again publicly available in the second full CEED software
328distribution, release CEED 2.0. This release contained notable features, such as
329four new CPU backends, two new GPU backends, CPU backend optimizations, initial
330support for operator composition, performance benchmarking, and a Navier-Stokes demo.
331The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
332backends process one element at a time and are intended for meshes with a smaller number
333of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
334eight interlaced elements and are intended for meshes with higher numbers of elements.
335The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
336performance. The `/cpu/self/xsmm/*` backends rely upon the
337[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
338performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
339The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
340performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
341parallelization approach, where each thread treats a finite element. Using just in time
342compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
343backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
344achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
345deteriorates very quickly when threads run out of registers.
346
347A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
348examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
349This example solves the time-dependent Navier-Stokes equations of compressible gas
350dynamics in a static Eulerian three-dimensional frame, using structured high-order
351finite/spectral element spatial discretizations and explicit high-order time-stepping
352(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
353so that the pointwise physics (defined at quadrature points) is separated from the
354parallelization and meshing concerns.
355
356Backends available in this release:
357
358| CEED resource (`-ceed`)  | Backend                                             |
359|--------------------------|-----------------------------------------------------|
360| `/cpu/self/ref/serial`   | Serial reference implementation                     |
361| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
362| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
363| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
364| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
365| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
366| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
367| `/cpu/occa`              | Serial OCCA kernels                                 |
368| `/gpu/occa`              | CUDA OCCA kernels                                   |
369| `/omp/occa`              | OpenMP OCCA kernels                                 |
370| `/ocl/occa`              | OpenCL OCCA kernels                                 |
371| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
372| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
373| `/gpu/magma`             | CUDA MAGMA kernels                                  |
374
375Examples available in this release:
376
377:::{list-table}
378:header-rows: 1
379:widths: auto
380* - User code
381  - Example
382* - `ceed`
383  - * ex1 (volume)
384* - `mfem`
385  - * BP1 (scalar mass operator)
386    * BP3 (scalar Laplace operator)
387* - `petsc`
388  - * BP1 (scalar mass operator)
389    * BP3 (scalar Laplace operator)
390    * Navier-Stokes
391* - `nek5000`
392  - * BP1 (scalar mass operator)
393    * BP3 (scalar Laplace operator)
394:::
395
396(v0-3)=
397
398## v0.3 (Sep 30, 2018)
399
400Notable features in this release include active/passive field interface, support for
401non-tensor bases, backend optimization, and improved Fortran interface. This release
402also focused on providing improved continuous integration, and many new tests with code
403coverage reports of about 90%. This release also provided a significant change to the
404public interface: a {ref}`CeedQFunction` can take any number of named input and output
405arguments while {ref}`CeedOperator` connects them to the actual data, which may be
406supplied explicitly to `CeedOperatorApply()` (active) or separately via
407`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
408of CeedQFunctions and composition of block solvers constructed using
409{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
410used in an optimized CPU backend. Although this is typically not visible to the user,
411it enables effective use of arbitrary-length SIMD while maintaining cache locality.
412This CPU backend also implements an algebraic factorization of tensor product gradients
413to perform fewer operations than standard application of interpolation and
414differentiation from nodes to quadrature points. This algebraic formulation
415automatically supports non-polynomial and non-interpolatory bases, thus is more general
416than the more common derivation in terms of Lagrange polynomials on the quadrature points.
417
418Backends available in this release:
419
420| CEED resource (`-ceed`) | Backend                                             |
421|-------------------------|-----------------------------------------------------|
422| `/cpu/self/blocked`     | Blocked reference implementation                    |
423| `/cpu/self/ref`         | Serial reference implementation                     |
424| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
425| `/cpu/occa`             | Serial OCCA kernels                                 |
426| `/gpu/occa`             | CUDA OCCA kernels                                   |
427| `/omp/occa`             | OpenMP OCCA kernels                                 |
428| `/ocl/occa`             | OpenCL OCCA kernels                                 |
429| `/gpu/magma`            | CUDA MAGMA kernels                                  |
430
431Examples available in this release:
432
433:::{list-table}
434:header-rows: 1
435:widths: auto
436* - User code
437  - Example
438* - `ceed`
439  - * ex1 (volume)
440* - `mfem`
441  - * BP1 (scalar mass operator)
442    * BP3 (scalar Laplace operator)
443* - `petsc`
444  - * BP1 (scalar mass operator)
445    * BP3 (scalar Laplace operator)
446* - `nek5000`
447  - * BP1 (scalar mass operator)
448    * BP3 (scalar Laplace operator)
449:::
450
451(v0-21)=
452
453## v0.21 (Sep 30, 2018)
454
455A MAGMA backend (which relies upon the
456[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
457release. This initial integration set up the framework of using MAGMA and provided the
458libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
459As any other backend, the MAGMA backend provides extended basic data structures for
460{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
461the fundamental CEED building blocks to work with the new data structures.
462In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
463but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
464internally, and thus seamlessly to the user, through the functions/methods that are
465provided to support them.
466
467Backends available in this release:
468
469| CEED resource (`-ceed`) | Backend                         |
470|-------------------------|---------------------------------|
471| `/cpu/self`             | Serial reference implementation |
472| `/cpu/occa`             | Serial OCCA kernels             |
473| `/gpu/occa`             | CUDA OCCA kernels               |
474| `/omp/occa`             | OpenMP OCCA kernels             |
475| `/ocl/occa`             | OpenCL OCCA kernels             |
476| `/gpu/magma`            | CUDA MAGMA kernels              |
477
478Examples available in this release:
479
480:::{list-table}
481:header-rows: 1
482:widths: auto
483* - User code
484  - Example
485* - `ceed`
486  - * ex1 (volume)
487* - `mfem`
488  - * BP1 (scalar mass operator)
489    * BP3 (scalar Laplace operator)
490* - `petsc`
491  - * BP1 (scalar mass operator)
492* - `nek5000`
493  - * BP1 (scalar mass operator)
494:::
495
496(v0-2)=
497
498## v0.2 (Mar 30, 2018)
499
500libCEED was made publicly available the first full CEED software distribution, release
501CEED 1.0. The distribution was made available using the Spack package manager to provide
502a common, easy-to-use build environment, where the user can build the CEED distribution
503with all dependencies. This release included a new Fortran interface for the library.
504This release also contained major improvements in the OCCA backend (including a new
505`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
506compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
507`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
508of the Laplace operator) was also added to this release.
509
510Backends available in this release:
511
512| CEED resource (`-ceed`) | Backend                         |
513|-------------------------|---------------------------------|
514| `/cpu/self`             | Serial reference implementation |
515| `/cpu/occa`             | Serial OCCA kernels             |
516| `/gpu/occa`             | CUDA OCCA kernels               |
517| `/omp/occa`             | OpenMP OCCA kernels             |
518| `/ocl/occa`             | OpenCL OCCA kernels             |
519
520Examples available in this release:
521
522:::{list-table}
523:header-rows: 1
524:widths: auto
525* - User code
526  - Example
527* - `ceed`
528  - * ex1 (volume)
529* - `mfem`
530  - * BP1 (scalar mass operator)
531    * BP3 (scalar Laplace operator)
532* - `petsc`
533  - * BP1 (scalar mass operator)
534* - `nek5000`
535  - * BP1 (scalar mass operator)
536:::
537
538(v0-1)=
539
540## v0.1 (Jan 3, 2018)
541
542Initial low-level API of the CEED project. The low-level API provides a set of Finite
543Elements kernels and components for writing new low-level kernels. Examples include:
544vector and sparse linear algebra, element matrix assembly over a batch of elements,
545partial assembly and action for efficient high-order operators like mass, diffusion,
546advection, etc. The main goal of the low-level API is to establish the basis for the
547high-level API. Also, identifying such low-level kernels and providing a reference
548implementation for them serves as the basis for specialized backend implementations.
549This release contained several backends: `/cpu/self`, and backends which rely upon the
550[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
551`/gpu/occa`, and `/omp/occa`.
552It also included several examples, in the `examples` folder:
553A standalone code that shows the usage of libCEED (with no external
554dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
555(with the application of the mass operator); and a `petsc` example to perform BP1
556(with the application of the mass operator).
557
558Backends available in this release:
559
560| CEED resource (`-ceed`) | Backend                         |
561|-------------------------|---------------------------------|
562| `/cpu/self`             | Serial reference implementation |
563| `/cpu/occa`             | Serial OCCA kernels             |
564| `/gpu/occa`             | CUDA OCCA kernels               |
565| `/omp/occa`             | OpenMP OCCA kernels             |
566
567Examples available in this release:
568
569| User code             | Example                           |
570|-----------------------|-----------------------------------|
571| `ceed`                | ex1 (scalar Laplace operator)     |
572| `mfem`                | BP1 (scalar mass operator)        |
573| `petsc`               | BP1 (scalar mass operator)        |
574```
575