xref: /libCEED/doc/sphinx/source/releasenotes.md (revision c766f59897c34c3c78fee5b8ada37e6c961759ac)
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### Bugfix
10
11- Fix bug in setting device id for GPU backends.
12
13### Interface changes
14
15- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
16
17### New features
18
19- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
20
21### Bugfix
22
23- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
24- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
25- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
26
27### Examples
28
29- Added various performance enhancements for {ref}`example-petsc-navier-stokes`
30- Refactored {ref}`example-petsc-navier-stokes` to improve code reuse
31- Added Shock Tube, Channel, and Flat Plate boundary layer problems to {ref}`example-petsc-navier-stokes`
32
33(v0-10-1)=
34
35## v0.10.1 (Apr 11, 2022)
36
37### Interface changes
38
39- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
40
41### Bugfix
42
43- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
44
45(v0-10)=
46
47## v0.10 (Mar 21, 2022)
48
49### Interface changes
50
51- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
52- 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`.
53- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
54- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
55- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
56- Warning added when compiling OCCA backend to alert users that this backend is experimental.
57- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
58- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
59- 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`.
60- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
61- 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.
62- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
63- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
64- 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.
65- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
66- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
67
68### New features
69
70- `CeedScalar` can now be set as `float` or `double` at compile time.
71- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
72- 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.
73- Remove need to guard library headers in QFunction source for code generation backends.
74- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
75- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
76- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
77- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
78- 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.
79- Added support for element matrix assembly in GPU backends.
80
81### Maintainability
82
83- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
84- `Include-what-you-use` makefile target added as `make iwyu`.
85- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
86- Put GPU JiTed kernel source code into separate files.
87- 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.
88
89(v0-9)=
90
91## v0.9 (Jul 6, 2021)
92
93### Interface changes
94
95- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
96
97### New features
98
99- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
100- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
101- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
102- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
103
104### Performance improvements
105
106### Examples
107
108- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
109- Fluid mechanics example adds GPU support and improves modularity.
110
111### Deprecated backends
112
113- 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.
114
115(v0-8)=
116
117## v0.8 (Mar 31, 2021)
118
119### Interface changes
120
121- Error handling improved to include enumerated error codes for C interface return values.
122- 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.
123
124### New features
125
126- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
127- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
128- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
129
130### Performance improvements
131
132- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
133- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
134
135### Examples
136
137- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
138- {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.
139- {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.
140- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
141
142(v0-7)=
143
144## v0.7 (Sep 29, 2020)
145
146### Interface changes
147
148- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
149  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
150  These changes improve support for mixed finite element methods.
151- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
152- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
153- 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.
154- 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.
155- 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.
156- 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.
157  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`.
158- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
159- 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.
160
161### New features
162
163- New HIP backend: `/gpu/hip/ref`.
164- CeedQFunction support for user `CUfunction`s in some backends
165
166### Performance improvements
167
168- OCCA backend rebuilt to facilitate future performance enhancements.
169- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
170
171### Examples
172
173- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
174
175### Deprecated backends
176
177- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
178
179(v0-6)=
180
181## v0.6 (Mar 29, 2020)
182
183libCEED v0.6 contains numerous new features and examples, as well as expanded
184documentation in [this new website](https://libceed.org).
185
186### New features
187
188- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
189  1-1 correspondence with the C interface, plus some convenience features.  For instance,
190  data stored in the {cpp:type}`CeedVector` structure are available without copy as
191  {py:class}`numpy.ndarray`.  Short tutorials are provided in
192  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
193- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
194  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
195  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
196  preconditioning ingredients and are used in the libCEED's multigrid examples.
197- The inverse of separable operators can be obtained using
198  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
199  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
200  especially for Laplacians and related operators.
201- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
202  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
203- Make public accessors for various attributes to facilitate writing composable code.
204- New backend: `/cpu/self/memcheck/serial`.
205- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
206  backends.  (Single source is coming soon for OCCA backends.)
207- Fix some missing edge cases in CUDA backend.
208
209### Performance Improvements
210
211- MAGMA backend performance optimization and non-tensor bases.
212- No-copy optimization in {c:func}`CeedOperatorApply`.
213
214### Interface changes
215
216- Replace {code}`CeedElemRestrictionCreateIdentity` and
217  {code}`CeedElemRestrictionCreateBlocked` with more flexible
218  {c:func}`CeedElemRestrictionCreateStrided` and
219  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
220- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
221- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
222  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
223  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
224  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
225  {c:func}`CeedElemRestrictionApply`.
226
227### Examples
228
229libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
230Notable additions include:
231
232- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
233  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
234
235- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
236  of domains (like the cube and sphere) by direct integration on a surface mesh;
237  demonstrates geometric dimension different from topological dimension.
238
239- PETSc {ref}`example-petsc-bps`:
240
241  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
242  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
243    and transparent CUDA support.
244  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
245    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
246    dimension different from topological dimension.
247
248- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
249  solver with algebraic multigrid coarse solve.
250
251- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
252  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
253  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
254  quasi-2D computational domain support.
255
256- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
257  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
258  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
259
260(v0-5)=
261
262## v0.5 (Sep 18, 2019)
263
264For this release, several improvements were made. Two new CUDA backends were added to
265the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
266performance using single-source {ref}`CeedQFunction`. From this release, users
267can define Q-Functions in a single source code independently of the targeted backend
268with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
269compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
270declaration, the {ref}`CeedQFunction` API has undergone a slight change:
271the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
272requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
273`QFunctionField` with eval mode `CEED EVAL GRAD`.
274
275Additionally, new CPU backends
276were included in this release, such as the `/cpu/self/opt/*` backends (which are
277written in pure C and use partial **E-vectors** to improve performance) and the
278`/cpu/self/ref/memcheck` backend (which relies upon the
279[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
280{ref}`CeedQFunction` have no undefined values).
281This release also included various performance improvements, bug fixes, new examples,
282and improved tests. Among these improvements, vectorized instructions for
283{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
284instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
285identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
286to include unstructured meshes handling were. For this expansion, the prior version of
287the PETSc BPs, which only included data associated with structured geometries, were
288renamed `bpsraw`, and the new version of the BPs, which can handle data associated
289with any unstructured geometry, were called `bps`. Additionally, other benchmark
290problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
291and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
292as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
293examples. Furthermoew, another standalone libCEED example, called `ex2`, which
294computes the surface area of a given mesh was added to this release.
295
296Backends available in this release:
297
298| CEED resource (`-ceed`)  | Backend                                             |
299|--------------------------|-----------------------------------------------------|
300| `/cpu/self/ref/serial`   | Serial reference implementation                     |
301| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
302| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
303| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
304| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
305| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
306| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
307| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
308| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
309| `/cpu/occa`              | Serial OCCA kernels                                 |
310| `/gpu/occa`              | CUDA OCCA kernels                                   |
311| `/omp/occa`              | OpenMP OCCA kernels                                 |
312| `/ocl/occa`              | OpenCL OCCA kernels                                 |
313| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
314| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
315| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
316| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
317| `/gpu/magma`             | CUDA MAGMA kernels                                  |
318
319Examples available in this release:
320
321:::{list-table}
322:header-rows: 1
323:widths: auto
324* - User code
325  - Example
326* - `ceed`
327  - * ex1 (volume)
328    * ex2 (surface)
329* - `mfem`
330  - * BP1 (scalar mass operator)
331    * BP3 (scalar Laplace operator)
332* - `petsc`
333  - * BP1 (scalar mass operator)
334    * BP2 (vector mass operator)
335    * BP3 (scalar Laplace operator)
336    * BP4 (vector Laplace operator)
337    * BP5 (collocated scalar Laplace operator)
338    * BP6 (collocated vector Laplace operator)
339    * Navier-Stokes
340* - `nek5000`
341  - * BP1 (scalar mass operator)
342    * BP3 (scalar Laplace operator)
343:::
344
345(v0-4)=
346
347## v0.4 (Apr 1, 2019)
348
349libCEED v0.4 was made again publicly available in the second full CEED software
350distribution, release CEED 2.0. This release contained notable features, such as
351four new CPU backends, two new GPU backends, CPU backend optimizations, initial
352support for operator composition, performance benchmarking, and a Navier-Stokes demo.
353The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
354backends process one element at a time and are intended for meshes with a smaller number
355of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
356eight interlaced elements and are intended for meshes with higher numbers of elements.
357The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
358performance. The `/cpu/self/xsmm/*` backends rely upon the
359[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
360performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
361The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
362performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
363parallelization approach, where each thread treats a finite element. Using just in time
364compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
365backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
366achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
367deteriorates very quickly when threads run out of registers.
368
369A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
370examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
371This example solves the time-dependent Navier-Stokes equations of compressible gas
372dynamics in a static Eulerian three-dimensional frame, using structured high-order
373finite/spectral element spatial discretizations and explicit high-order time-stepping
374(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
375so that the pointwise physics (defined at quadrature points) is separated from the
376parallelization and meshing concerns.
377
378Backends available in this release:
379
380| CEED resource (`-ceed`)  | Backend                                             |
381|--------------------------|-----------------------------------------------------|
382| `/cpu/self/ref/serial`   | Serial reference implementation                     |
383| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
384| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
385| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
386| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
387| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
388| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
389| `/cpu/occa`              | Serial OCCA kernels                                 |
390| `/gpu/occa`              | CUDA OCCA kernels                                   |
391| `/omp/occa`              | OpenMP OCCA kernels                                 |
392| `/ocl/occa`              | OpenCL OCCA kernels                                 |
393| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
394| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
395| `/gpu/magma`             | CUDA MAGMA kernels                                  |
396
397Examples available in this release:
398
399:::{list-table}
400:header-rows: 1
401:widths: auto
402* - User code
403  - Example
404* - `ceed`
405  - * ex1 (volume)
406* - `mfem`
407  - * BP1 (scalar mass operator)
408    * BP3 (scalar Laplace operator)
409* - `petsc`
410  - * BP1 (scalar mass operator)
411    * BP3 (scalar Laplace operator)
412    * Navier-Stokes
413* - `nek5000`
414  - * BP1 (scalar mass operator)
415    * BP3 (scalar Laplace operator)
416:::
417
418(v0-3)=
419
420## v0.3 (Sep 30, 2018)
421
422Notable features in this release include active/passive field interface, support for
423non-tensor bases, backend optimization, and improved Fortran interface. This release
424also focused on providing improved continuous integration, and many new tests with code
425coverage reports of about 90%. This release also provided a significant change to the
426public interface: a {ref}`CeedQFunction` can take any number of named input and output
427arguments while {ref}`CeedOperator` connects them to the actual data, which may be
428supplied explicitly to `CeedOperatorApply()` (active) or separately via
429`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
430of CeedQFunctions and composition of block solvers constructed using
431{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
432used in an optimized CPU backend. Although this is typically not visible to the user,
433it enables effective use of arbitrary-length SIMD while maintaining cache locality.
434This CPU backend also implements an algebraic factorization of tensor product gradients
435to perform fewer operations than standard application of interpolation and
436differentiation from nodes to quadrature points. This algebraic formulation
437automatically supports non-polynomial and non-interpolatory bases, thus is more general
438than the more common derivation in terms of Lagrange polynomials on the quadrature points.
439
440Backends available in this release:
441
442| CEED resource (`-ceed`) | Backend                                             |
443|-------------------------|-----------------------------------------------------|
444| `/cpu/self/blocked`     | Blocked reference implementation                    |
445| `/cpu/self/ref`         | Serial reference implementation                     |
446| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
447| `/cpu/occa`             | Serial OCCA kernels                                 |
448| `/gpu/occa`             | CUDA OCCA kernels                                   |
449| `/omp/occa`             | OpenMP OCCA kernels                                 |
450| `/ocl/occa`             | OpenCL OCCA kernels                                 |
451| `/gpu/magma`            | CUDA MAGMA kernels                                  |
452
453Examples available in this release:
454
455:::{list-table}
456:header-rows: 1
457:widths: auto
458* - User code
459  - Example
460* - `ceed`
461  - * ex1 (volume)
462* - `mfem`
463  - * BP1 (scalar mass operator)
464    * BP3 (scalar Laplace operator)
465* - `petsc`
466  - * BP1 (scalar mass operator)
467    * BP3 (scalar Laplace operator)
468* - `nek5000`
469  - * BP1 (scalar mass operator)
470    * BP3 (scalar Laplace operator)
471:::
472
473(v0-21)=
474
475## v0.21 (Sep 30, 2018)
476
477A MAGMA backend (which relies upon the
478[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
479release. This initial integration set up the framework of using MAGMA and provided the
480libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
481As any other backend, the MAGMA backend provides extended basic data structures for
482{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
483the fundamental CEED building blocks to work with the new data structures.
484In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
485but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
486internally, and thus seamlessly to the user, through the functions/methods that are
487provided to support them.
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| `/gpu/magma`            | CUDA MAGMA kernels              |
499
500Examples available in this release:
501
502:::{list-table}
503:header-rows: 1
504:widths: auto
505* - User code
506  - Example
507* - `ceed`
508  - * ex1 (volume)
509* - `mfem`
510  - * BP1 (scalar mass operator)
511    * BP3 (scalar Laplace operator)
512* - `petsc`
513  - * BP1 (scalar mass operator)
514* - `nek5000`
515  - * BP1 (scalar mass operator)
516:::
517
518(v0-2)=
519
520## v0.2 (Mar 30, 2018)
521
522libCEED was made publicly available the first full CEED software distribution, release
523CEED 1.0. The distribution was made available using the Spack package manager to provide
524a common, easy-to-use build environment, where the user can build the CEED distribution
525with all dependencies. This release included a new Fortran interface for the library.
526This release also contained major improvements in the OCCA backend (including a new
527`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
528compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
529`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
530of the Laplace operator) was also added to this release.
531
532Backends available in this release:
533
534| CEED resource (`-ceed`) | Backend                         |
535|-------------------------|---------------------------------|
536| `/cpu/self`             | Serial reference implementation |
537| `/cpu/occa`             | Serial OCCA kernels             |
538| `/gpu/occa`             | CUDA OCCA kernels               |
539| `/omp/occa`             | OpenMP OCCA kernels             |
540| `/ocl/occa`             | OpenCL OCCA kernels             |
541
542Examples available in this release:
543
544:::{list-table}
545:header-rows: 1
546:widths: auto
547* - User code
548  - Example
549* - `ceed`
550  - * ex1 (volume)
551* - `mfem`
552  - * BP1 (scalar mass operator)
553    * BP3 (scalar Laplace operator)
554* - `petsc`
555  - * BP1 (scalar mass operator)
556* - `nek5000`
557  - * BP1 (scalar mass operator)
558:::
559
560(v0-1)=
561
562## v0.1 (Jan 3, 2018)
563
564Initial low-level API of the CEED project. The low-level API provides a set of Finite
565Elements kernels and components for writing new low-level kernels. Examples include:
566vector and sparse linear algebra, element matrix assembly over a batch of elements,
567partial assembly and action for efficient high-order operators like mass, diffusion,
568advection, etc. The main goal of the low-level API is to establish the basis for the
569high-level API. Also, identifying such low-level kernels and providing a reference
570implementation for them serves as the basis for specialized backend implementations.
571This release contained several backends: `/cpu/self`, and backends which rely upon the
572[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
573`/gpu/occa`, and `/omp/occa`.
574It also included several examples, in the `examples` folder:
575A standalone code that shows the usage of libCEED (with no external
576dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
577(with the application of the mass operator); and a `petsc` example to perform BP1
578(with the application of the mass operator).
579
580Backends available in this release:
581
582| CEED resource (`-ceed`) | Backend                         |
583|-------------------------|---------------------------------|
584| `/cpu/self`             | Serial reference implementation |
585| `/cpu/occa`             | Serial OCCA kernels             |
586| `/gpu/occa`             | CUDA OCCA kernels               |
587| `/omp/occa`             | OpenMP OCCA kernels             |
588
589Examples available in this release:
590
591| User code             | Example                           |
592|-----------------------|-----------------------------------|
593| `ceed`                | ex1 (scalar Laplace operator)     |
594| `mfem`                | BP1 (scalar mass operator)        |
595| `petsc`               | BP1 (scalar mass operator)        |
596```
597