xref: /libCEED/doc/sphinx/source/releasenotes.md (revision cdb4dd67cd8edd07e3207a1c021923acf216e178)
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
31(v0-10-1)=
32
33## v0.10.1 (Apr 11, 2022)
34
35### Interface changes
36
37- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
38
39### Bugfix
40
41- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
42
43(v0-10)=
44
45## v0.10 (Mar 21, 2022)
46
47### Interface changes
48
49- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
50- 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`.
51- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
52- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
53- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
54- Warning added when compiling OCCA backend to alert users that this backend is experimental.
55- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
56- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
57- 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`.
58- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
59- 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.
60- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
61- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
62- 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.
63- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
64- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
65
66### New features
67
68- `CeedScalar` can now be set as `float` or `double` at compile time.
69- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
70- 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.
71- Remove need to guard library headers in QFunction source for code generation backends.
72- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
73- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
74- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
75- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
76- 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.
77- Added support for element matrix assembly in GPU backends.
78
79### Maintainability
80
81- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
82- `Include-what-you-use` makefile target added as `make iwyu`.
83- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
84- Put GPU JiTed kernel source code into separate files.
85- 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.
86
87(v0-9)=
88
89## v0.9 (Jul 6, 2021)
90
91### Interface changes
92
93- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
94
95### New features
96
97- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
98- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
99- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
100- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
101
102### Performance improvements
103
104### Examples
105
106- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
107- Fluid mechanics example adds GPU support and improves modularity.
108
109### Deprecated backends
110
111- 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.
112
113(v0-8)=
114
115## v0.8 (Mar 31, 2021)
116
117### Interface changes
118
119- Error handling improved to include enumerated error codes for C interface return values.
120- 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.
121
122### New features
123
124- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
125- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
126- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
127
128### Performance improvements
129
130- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
131- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
132
133### Examples
134
135- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
136- {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.
137- {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.
138- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
139
140(v0-7)=
141
142## v0.7 (Sep 29, 2020)
143
144### Interface changes
145
146- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
147  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
148  These changes improve support for mixed finite element methods.
149- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
150- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
151- 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.
152- 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.
153- 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.
154- 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.
155  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`.
156- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
157- 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.
158
159### New features
160
161- New HIP backend: `/gpu/hip/ref`.
162- CeedQFunction support for user `CUfunction`s in some backends
163
164### Performance improvements
165
166- OCCA backend rebuilt to facilitate future performance enhancements.
167- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
168
169### Examples
170
171- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
172
173### Deprecated backends
174
175- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
176
177(v0-6)=
178
179## v0.6 (Mar 29, 2020)
180
181libCEED v0.6 contains numerous new features and examples, as well as expanded
182documentation in [this new website](https://libceed.org).
183
184### New features
185
186- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
187  1-1 correspondence with the C interface, plus some convenience features.  For instance,
188  data stored in the {cpp:type}`CeedVector` structure are available without copy as
189  {py:class}`numpy.ndarray`.  Short tutorials are provided in
190  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
191- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
192  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
193  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
194  preconditioning ingredients and are used in the libCEED's multigrid examples.
195- The inverse of separable operators can be obtained using
196  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
197  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
198  especially for Laplacians and related operators.
199- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
200  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
201- Make public accessors for various attributes to facilitate writing composable code.
202- New backend: `/cpu/self/memcheck/serial`.
203- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
204  backends.  (Single source is coming soon for OCCA backends.)
205- Fix some missing edge cases in CUDA backend.
206
207### Performance Improvements
208
209- MAGMA backend performance optimization and non-tensor bases.
210- No-copy optimization in {c:func}`CeedOperatorApply`.
211
212### Interface changes
213
214- Replace {code}`CeedElemRestrictionCreateIdentity` and
215  {code}`CeedElemRestrictionCreateBlocked` with more flexible
216  {c:func}`CeedElemRestrictionCreateStrided` and
217  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
218- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
219- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
220  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
221  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
222  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
223  {c:func}`CeedElemRestrictionApply`.
224
225### Examples
226
227libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
228Notable additions include:
229
230- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
231  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
232
233- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
234  of domains (like the cube and sphere) by direct integration on a surface mesh;
235  demonstrates geometric dimension different from topological dimension.
236
237- PETSc {ref}`example-petsc-bps`:
238
239  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
240  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
241    and transparent CUDA support.
242  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
243    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
244    dimension different from topological dimension.
245
246- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
247  solver with algebraic multigrid coarse solve.
248
249- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
250  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
251  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
252  quasi-2D computational domain support.
253
254- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
255  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
256  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
257
258(v0-5)=
259
260## v0.5 (Sep 18, 2019)
261
262For this release, several improvements were made. Two new CUDA backends were added to
263the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
264performance using single-source {ref}`CeedQFunction`. From this release, users
265can define Q-Functions in a single source code independently of the targeted backend
266with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
267compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
268declaration, the {ref}`CeedQFunction` API has undergone a slight change:
269the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
270requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
271`QFunctionField` with eval mode `CEED EVAL GRAD`.
272
273Additionally, new CPU backends
274were included in this release, such as the `/cpu/self/opt/*` backends (which are
275written in pure C and use partial **E-vectors** to improve performance) and the
276`/cpu/self/ref/memcheck` backend (which relies upon the
277[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
278{ref}`CeedQFunction` have no undefined values).
279This release also included various performance improvements, bug fixes, new examples,
280and improved tests. Among these improvements, vectorized instructions for
281{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
282instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
283identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
284to include unstructured meshes handling were. For this expansion, the prior version of
285the PETSc BPs, which only included data associated with structured geometries, were
286renamed `bpsraw`, and the new version of the BPs, which can handle data associated
287with any unstructured geometry, were called `bps`. Additionally, other benchmark
288problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
289and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
290as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
291examples. Furthermoew, another standalone libCEED example, called `ex2`, which
292computes the surface area of a given mesh was added to this release.
293
294Backends available in this release:
295
296| CEED resource (`-ceed`)  | Backend                                             |
297|--------------------------|-----------------------------------------------------|
298| `/cpu/self/ref/serial`   | Serial reference implementation                     |
299| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
300| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
301| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
302| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
303| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
304| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
305| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
306| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
307| `/cpu/occa`              | Serial OCCA kernels                                 |
308| `/gpu/occa`              | CUDA OCCA kernels                                   |
309| `/omp/occa`              | OpenMP OCCA kernels                                 |
310| `/ocl/occa`              | OpenCL OCCA kernels                                 |
311| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
312| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
313| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
314| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
315| `/gpu/magma`             | CUDA MAGMA kernels                                  |
316
317Examples available in this release:
318
319:::{list-table}
320:header-rows: 1
321:widths: auto
322* - User code
323  - Example
324* - `ceed`
325  - * ex1 (volume)
326    * ex2 (surface)
327* - `mfem`
328  - * BP1 (scalar mass operator)
329    * BP3 (scalar Laplace operator)
330* - `petsc`
331  - * BP1 (scalar mass operator)
332    * BP2 (vector mass operator)
333    * BP3 (scalar Laplace operator)
334    * BP4 (vector Laplace operator)
335    * BP5 (collocated scalar Laplace operator)
336    * BP6 (collocated vector Laplace operator)
337    * Navier-Stokes
338* - `nek5000`
339  - * BP1 (scalar mass operator)
340    * BP3 (scalar Laplace operator)
341:::
342
343(v0-4)=
344
345## v0.4 (Apr 1, 2019)
346
347libCEED v0.4 was made again publicly available in the second full CEED software
348distribution, release CEED 2.0. This release contained notable features, such as
349four new CPU backends, two new GPU backends, CPU backend optimizations, initial
350support for operator composition, performance benchmarking, and a Navier-Stokes demo.
351The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
352backends process one element at a time and are intended for meshes with a smaller number
353of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
354eight interlaced elements and are intended for meshes with higher numbers of elements.
355The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
356performance. The `/cpu/self/xsmm/*` backends rely upon the
357[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
358performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
359The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
360performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
361parallelization approach, where each thread treats a finite element. Using just in time
362compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
363backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
364achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
365deteriorates very quickly when threads run out of registers.
366
367A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
368examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
369This example solves the time-dependent Navier-Stokes equations of compressible gas
370dynamics in a static Eulerian three-dimensional frame, using structured high-order
371finite/spectral element spatial discretizations and explicit high-order time-stepping
372(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
373so that the pointwise physics (defined at quadrature points) is separated from the
374parallelization and meshing concerns.
375
376Backends available in this release:
377
378| CEED resource (`-ceed`)  | Backend                                             |
379|--------------------------|-----------------------------------------------------|
380| `/cpu/self/ref/serial`   | Serial reference implementation                     |
381| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
382| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
383| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
384| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
385| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
386| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
387| `/cpu/occa`              | Serial OCCA kernels                                 |
388| `/gpu/occa`              | CUDA OCCA kernels                                   |
389| `/omp/occa`              | OpenMP OCCA kernels                                 |
390| `/ocl/occa`              | OpenCL OCCA kernels                                 |
391| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
392| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
393| `/gpu/magma`             | CUDA MAGMA kernels                                  |
394
395Examples available in this release:
396
397:::{list-table}
398:header-rows: 1
399:widths: auto
400* - User code
401  - Example
402* - `ceed`
403  - * ex1 (volume)
404* - `mfem`
405  - * BP1 (scalar mass operator)
406    * BP3 (scalar Laplace operator)
407* - `petsc`
408  - * BP1 (scalar mass operator)
409    * BP3 (scalar Laplace operator)
410    * Navier-Stokes
411* - `nek5000`
412  - * BP1 (scalar mass operator)
413    * BP3 (scalar Laplace operator)
414:::
415
416(v0-3)=
417
418## v0.3 (Sep 30, 2018)
419
420Notable features in this release include active/passive field interface, support for
421non-tensor bases, backend optimization, and improved Fortran interface. This release
422also focused on providing improved continuous integration, and many new tests with code
423coverage reports of about 90%. This release also provided a significant change to the
424public interface: a {ref}`CeedQFunction` can take any number of named input and output
425arguments while {ref}`CeedOperator` connects them to the actual data, which may be
426supplied explicitly to `CeedOperatorApply()` (active) or separately via
427`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
428of CeedQFunctions and composition of block solvers constructed using
429{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
430used in an optimized CPU backend. Although this is typically not visible to the user,
431it enables effective use of arbitrary-length SIMD while maintaining cache locality.
432This CPU backend also implements an algebraic factorization of tensor product gradients
433to perform fewer operations than standard application of interpolation and
434differentiation from nodes to quadrature points. This algebraic formulation
435automatically supports non-polynomial and non-interpolatory bases, thus is more general
436than the more common derivation in terms of Lagrange polynomials on the quadrature points.
437
438Backends available in this release:
439
440| CEED resource (`-ceed`) | Backend                                             |
441|-------------------------|-----------------------------------------------------|
442| `/cpu/self/blocked`     | Blocked reference implementation                    |
443| `/cpu/self/ref`         | Serial reference implementation                     |
444| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
445| `/cpu/occa`             | Serial OCCA kernels                                 |
446| `/gpu/occa`             | CUDA OCCA kernels                                   |
447| `/omp/occa`             | OpenMP OCCA kernels                                 |
448| `/ocl/occa`             | OpenCL OCCA kernels                                 |
449| `/gpu/magma`            | CUDA MAGMA kernels                                  |
450
451Examples available in this release:
452
453:::{list-table}
454:header-rows: 1
455:widths: auto
456* - User code
457  - Example
458* - `ceed`
459  - * ex1 (volume)
460* - `mfem`
461  - * BP1 (scalar mass operator)
462    * BP3 (scalar Laplace operator)
463* - `petsc`
464  - * BP1 (scalar mass operator)
465    * BP3 (scalar Laplace operator)
466* - `nek5000`
467  - * BP1 (scalar mass operator)
468    * BP3 (scalar Laplace operator)
469:::
470
471(v0-21)=
472
473## v0.21 (Sep 30, 2018)
474
475A MAGMA backend (which relies upon the
476[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
477release. This initial integration set up the framework of using MAGMA and provided the
478libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
479As any other backend, the MAGMA backend provides extended basic data structures for
480{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
481the fundamental CEED building blocks to work with the new data structures.
482In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
483but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
484internally, and thus seamlessly to the user, through the functions/methods that are
485provided to support them.
486
487Backends available in this release:
488
489| CEED resource (`-ceed`) | Backend                         |
490|-------------------------|---------------------------------|
491| `/cpu/self`             | Serial reference implementation |
492| `/cpu/occa`             | Serial OCCA kernels             |
493| `/gpu/occa`             | CUDA OCCA kernels               |
494| `/omp/occa`             | OpenMP OCCA kernels             |
495| `/ocl/occa`             | OpenCL OCCA kernels             |
496| `/gpu/magma`            | CUDA MAGMA kernels              |
497
498Examples available in this release:
499
500:::{list-table}
501:header-rows: 1
502:widths: auto
503* - User code
504  - Example
505* - `ceed`
506  - * ex1 (volume)
507* - `mfem`
508  - * BP1 (scalar mass operator)
509    * BP3 (scalar Laplace operator)
510* - `petsc`
511  - * BP1 (scalar mass operator)
512* - `nek5000`
513  - * BP1 (scalar mass operator)
514:::
515
516(v0-2)=
517
518## v0.2 (Mar 30, 2018)
519
520libCEED was made publicly available the first full CEED software distribution, release
521CEED 1.0. The distribution was made available using the Spack package manager to provide
522a common, easy-to-use build environment, where the user can build the CEED distribution
523with all dependencies. This release included a new Fortran interface for the library.
524This release also contained major improvements in the OCCA backend (including a new
525`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
526compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
527`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
528of the Laplace operator) was also added to this release.
529
530Backends available in this release:
531
532| CEED resource (`-ceed`) | Backend                         |
533|-------------------------|---------------------------------|
534| `/cpu/self`             | Serial reference implementation |
535| `/cpu/occa`             | Serial OCCA kernels             |
536| `/gpu/occa`             | CUDA OCCA kernels               |
537| `/omp/occa`             | OpenMP OCCA kernels             |
538| `/ocl/occa`             | OpenCL OCCA kernels             |
539
540Examples available in this release:
541
542:::{list-table}
543:header-rows: 1
544:widths: auto
545* - User code
546  - Example
547* - `ceed`
548  - * ex1 (volume)
549* - `mfem`
550  - * BP1 (scalar mass operator)
551    * BP3 (scalar Laplace operator)
552* - `petsc`
553  - * BP1 (scalar mass operator)
554* - `nek5000`
555  - * BP1 (scalar mass operator)
556:::
557
558(v0-1)=
559
560## v0.1 (Jan 3, 2018)
561
562Initial low-level API of the CEED project. The low-level API provides a set of Finite
563Elements kernels and components for writing new low-level kernels. Examples include:
564vector and sparse linear algebra, element matrix assembly over a batch of elements,
565partial assembly and action for efficient high-order operators like mass, diffusion,
566advection, etc. The main goal of the low-level API is to establish the basis for the
567high-level API. Also, identifying such low-level kernels and providing a reference
568implementation for them serves as the basis for specialized backend implementations.
569This release contained several backends: `/cpu/self`, and backends which rely upon the
570[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
571`/gpu/occa`, and `/omp/occa`.
572It also included several examples, in the `examples` folder:
573A standalone code that shows the usage of libCEED (with no external
574dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
575(with the application of the mass operator); and a `petsc` example to perform BP1
576(with the application of the mass operator).
577
578Backends available in this release:
579
580| CEED resource (`-ceed`) | Backend                         |
581|-------------------------|---------------------------------|
582| `/cpu/self`             | Serial reference implementation |
583| `/cpu/occa`             | Serial OCCA kernels             |
584| `/gpu/occa`             | CUDA OCCA kernels               |
585| `/omp/occa`             | OpenMP OCCA kernels             |
586
587Examples available in this release:
588
589| User code             | Example                           |
590|-----------------------|-----------------------------------|
591| `ceed`                | ex1 (scalar Laplace operator)     |
592| `mfem`                | BP1 (scalar mass operator)        |
593| `petsc`               | BP1 (scalar mass operator)        |
594```
595