xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 609df0c8aa4664800c4e2fa3c255c816aed70bd1)
1# Changes/Release Notes
2
3On this page we provide a summary of the main API changes, new features and examples
4for each release of libCEED.
5
6(main)=
7
8## Current `main` branch
9
10### Interface changes
11
12- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
13- Promote to the public API: QFunction and Operator field objects, `CeedQFunctionField` and `CeedOperatorField`, and associated getters, {c:func}`CeedQFunctionGetFields`; {c:func}`CeedQFunctionFieldGetName`; {c:func}`CeedQFunctionFieldGetSize`; {c:func}`CeedQFunctionFieldGetEvalMode`; {c:func}`CeedOperatorGetFields`; {c:func}`CeedOperatorFieldGetElemRestriction`; {c:func}`CeedOperatorFieldGetBasis`; and {c:func}`CeedOperatorFieldGetVector`.
14- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
15- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
16- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
17- Warning added when compiling OCCA backend to alert users that this backend is experimental.
18- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
19- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
20- Clarify documentation for {c:func}`CeedVectorTakeArray`; this function will error if {c:func}`CeedVectorSetArray` with `copy_mode == CEED_USE_POINTER` was not previously called for the corresponding `CeedMemType`.
21- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
22- Added {c:func}`CeedQFunctionContextRegisterDouble` and {c:func}`CeedQFunctionContextRegisterInt32` with {c:func}`CeedQFunctionContextSetDouble` and {c:func}`CeedQFunctionContextSetInt32` to facilitate easy updating of {c:struct}`CeedQFunctionContext` data by user defined field names.
23- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
24- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
25- 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.
26- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
27- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
28
29### New features
30
31- `CeedScalar` can now be set as `float` or `double` at compile time.
32- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
33- 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.
34- Remove need to guard library headers in QFunction source for code generation backends.
35- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
36- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
37- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
38- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
39- 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.
40
41### Maintainability
42
43- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
44- `Include-what-you-use` makefile target added as `make iwyu`.
45- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
46- Put GPU JiTed kernel source code into separate files.
47- 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.
48
49(v0-9)=
50
51## v0.9 (Jul 6, 2021)
52
53### Interface changes
54
55- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
56
57### New features
58
59- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
60- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
61- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
62- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
63
64### Performance improvements
65
66### Examples
67
68- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
69- Fluid mechanics example adds GPU support and improves modularity.
70
71### Deprecated backends
72
73- 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.
74
75(v0-8)=
76
77## v0.8 (Mar 31, 2021)
78
79### Interface changes
80
81- Error handling improved to include enumerated error codes for C interface return values.
82- 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.
83
84### New features
85
86- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
87- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
88- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
89
90### Performance improvements
91
92- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
93- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
94
95### Examples
96
97- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
98- {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.
99- {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.
100- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
101
102(v0-7)=
103
104## v0.7 (Sep 29, 2020)
105
106### Interface changes
107
108- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
109  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
110  These changes improve support for mixed finite element methods.
111- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
112- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
113- 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.
114- 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.
115- 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.
116- 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.
117  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`.
118- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
119- 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.
120
121### New features
122
123- New HIP backend: `/gpu/hip/ref`.
124- CeedQFunction support for user `CUfunction`s in some backends
125
126### Performance improvements
127
128- OCCA backend rebuilt to facilitate future performance enhancements.
129- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
130
131### Examples
132
133- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
134
135### Deprecated backends
136
137- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
138
139(v0-6)=
140
141## v0.6 (Mar 29, 2020)
142
143libCEED v0.6 contains numerous new features and examples, as well as expanded
144documentation in [this new website](https://libceed.org).
145
146### New features
147
148- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
149  1-1 correspondence with the C interface, plus some convenience features.  For instance,
150  data stored in the {cpp:type}`CeedVector` structure are available without copy as
151  {py:class}`numpy.ndarray`.  Short tutorials are provided in
152  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
153- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
154  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
155  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
156  preconditioning ingredients and are used in the libCEED's multigrid examples.
157- The inverse of separable operators can be obtained using
158  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
159  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
160  especially for Laplacians and related operators.
161- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
162  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
163- Make public accessors for various attributes to facilitate writing composable code.
164- New backend: `/cpu/self/memcheck/serial`.
165- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
166  backends.  (Single source is coming soon for OCCA backends.)
167- Fix some missing edge cases in CUDA backend.
168
169### Performance Improvements
170
171- MAGMA backend performance optimization and non-tensor bases.
172- No-copy optimization in {c:func}`CeedOperatorApply`.
173
174### Interface changes
175
176- Replace {code}`CeedElemRestrictionCreateIdentity` and
177  {code}`CeedElemRestrictionCreateBlocked` with more flexible
178  {c:func}`CeedElemRestrictionCreateStrided` and
179  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
180- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
181- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
182  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
183  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
184  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
185  {c:func}`CeedElemRestrictionApply`.
186
187### Examples
188
189libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
190Notable additions include:
191
192- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
193  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
194
195- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
196  of domains (like the cube and sphere) by direct integration on a surface mesh;
197  demonstrates geometric dimension different from topological dimension.
198
199- PETSc {ref}`example-petsc-bps`:
200
201  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
202  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
203    and transparent CUDA support.
204  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
205    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
206    dimension different from topological dimension.
207
208- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
209  solver with algebraic multigrid coarse solve.
210
211- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
212  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
213  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
214  quasi-2D computational domain support.
215
216- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
217  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
218  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
219
220(v0-5)=
221
222## v0.5 (Sep 18, 2019)
223
224For this release, several improvements were made. Two new CUDA backends were added to
225the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
226performance using single-source {ref}`CeedQFunction`. From this release, users
227can define Q-Functions in a single source code independently of the targeted backend
228with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
229compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
230declaration, the {ref}`CeedQFunction` API has undergone a slight change:
231the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
232requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
233`QFunctionField` with eval mode `CEED EVAL GRAD`.
234
235Additionally, new CPU backends
236were included in this release, such as the `/cpu/self/opt/*` backends (which are
237written in pure C and use partial **E-vectors** to improve performance) and the
238`/cpu/self/ref/memcheck` backend (which relies upon the
239[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
240{ref}`CeedQFunction` have no undefined values).
241This release also included various performance improvements, bug fixes, new examples,
242and improved tests. Among these improvements, vectorized instructions for
243{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
244instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
245identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
246to include unstructured meshes handling were. For this expansion, the prior version of
247the PETSc BPs, which only included data associated with structured geometries, were
248renamed `bpsraw`, and the new version of the BPs, which can handle data associated
249with any unstructured geometry, were called `bps`. Additionally, other benchmark
250problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
251and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
252as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
253examples. Furthermoew, another standalone libCEED example, called `ex2`, which
254computes the surface area of a given mesh was added to this release.
255
256Backends available in this release:
257
258| CEED resource (`-ceed`)  | Backend                                             |
259|--------------------------|-----------------------------------------------------|
260| `/cpu/self/ref/serial`   | Serial reference implementation                     |
261| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
262| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
263| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
264| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
265| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
266| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
267| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
268| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
269| `/cpu/occa`              | Serial OCCA kernels                                 |
270| `/gpu/occa`              | CUDA OCCA kernels                                   |
271| `/omp/occa`              | OpenMP OCCA kernels                                 |
272| `/ocl/occa`              | OpenCL OCCA kernels                                 |
273| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
274| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
275| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
276| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
277| `/gpu/magma`             | CUDA MAGMA kernels                                  |
278
279Examples available in this release:
280
281:::{list-table}
282:header-rows: 1
283:widths: auto
284* - User code
285  - Example
286* - `ceed`
287  - * ex1 (volume)
288    * ex2 (surface)
289* - `mfem`
290  - * BP1 (scalar mass operator)
291    * BP3 (scalar Laplace operator)
292* - `petsc`
293  - * BP1 (scalar mass operator)
294    * BP2 (vector mass operator)
295    * BP3 (scalar Laplace operator)
296    * BP4 (vector Laplace operator)
297    * BP5 (collocated scalar Laplace operator)
298    * BP6 (collocated vector Laplace operator)
299    * Navier-Stokes
300* - `nek5000`
301  - * BP1 (scalar mass operator)
302    * BP3 (scalar Laplace operator)
303:::
304
305(v0-4)=
306
307## v0.4 (Apr 1, 2019)
308
309libCEED v0.4 was made again publicly available in the second full CEED software
310distribution, release CEED 2.0. This release contained notable features, such as
311four new CPU backends, two new GPU backends, CPU backend optimizations, initial
312support for operator composition, performance benchmarking, and a Navier-Stokes demo.
313The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
314backends process one element at a time and are intended for meshes with a smaller number
315of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
316eight interlaced elements and are intended for meshes with higher numbers of elements.
317The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
318performance. The `/cpu/self/xsmm/*` backends rely upon the
319[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
320performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
321The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
322performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
323parallelization approach, where each thread treats a finite element. Using just in time
324compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
325backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
326achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
327deteriorates very quickly when threads run out of registers.
328
329A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
330examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
331This example solves the time-dependent Navier-Stokes equations of compressible gas
332dynamics in a static Eulerian three-dimensional frame, using structured high-order
333finite/spectral element spatial discretizations and explicit high-order time-stepping
334(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
335so that the pointwise physics (defined at quadrature points) is separated from the
336parallelization and meshing concerns.
337
338Backends available in this release:
339
340| CEED resource (`-ceed`)  | Backend                                             |
341|--------------------------|-----------------------------------------------------|
342| `/cpu/self/ref/serial`   | Serial reference implementation                     |
343| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
344| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
345| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
346| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
347| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
348| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
349| `/cpu/occa`              | Serial OCCA kernels                                 |
350| `/gpu/occa`              | CUDA OCCA kernels                                   |
351| `/omp/occa`              | OpenMP OCCA kernels                                 |
352| `/ocl/occa`              | OpenCL OCCA kernels                                 |
353| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
354| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
355| `/gpu/magma`             | CUDA MAGMA kernels                                  |
356
357Examples available in this release:
358
359:::{list-table}
360:header-rows: 1
361:widths: auto
362* - User code
363  - Example
364* - `ceed`
365  - * ex1 (volume)
366* - `mfem`
367  - * BP1 (scalar mass operator)
368    * BP3 (scalar Laplace operator)
369* - `petsc`
370  - * BP1 (scalar mass operator)
371    * BP3 (scalar Laplace operator)
372    * Navier-Stokes
373* - `nek5000`
374  - * BP1 (scalar mass operator)
375    * BP3 (scalar Laplace operator)
376:::
377
378(v0-3)=
379
380## v0.3 (Sep 30, 2018)
381
382Notable features in this release include active/passive field interface, support for
383non-tensor bases, backend optimization, and improved Fortran interface. This release
384also focused on providing improved continuous integration, and many new tests with code
385coverage reports of about 90%. This release also provided a significant change to the
386public interface: a {ref}`CeedQFunction` can take any number of named input and output
387arguments while {ref}`CeedOperator` connects them to the actual data, which may be
388supplied explicitly to `CeedOperatorApply()` (active) or separately via
389`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
390of CeedQFunctions and composition of block solvers constructed using
391{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
392used in an optimized CPU backend. Although this is typically not visible to the user,
393it enables effective use of arbitrary-length SIMD while maintaining cache locality.
394This CPU backend also implements an algebraic factorization of tensor product gradients
395to perform fewer operations than standard application of interpolation and
396differentiation from nodes to quadrature points. This algebraic formulation
397automatically supports non-polynomial and non-interpolatory bases, thus is more general
398than the more common derivation in terms of Lagrange polynomials on the quadrature points.
399
400Backends available in this release:
401
402| CEED resource (`-ceed`) | Backend                                             |
403|-------------------------|-----------------------------------------------------|
404| `/cpu/self/blocked`     | Blocked reference implementation                    |
405| `/cpu/self/ref`         | Serial reference implementation                     |
406| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
407| `/cpu/occa`             | Serial OCCA kernels                                 |
408| `/gpu/occa`             | CUDA OCCA kernels                                   |
409| `/omp/occa`             | OpenMP OCCA kernels                                 |
410| `/ocl/occa`             | OpenCL OCCA kernels                                 |
411| `/gpu/magma`            | CUDA MAGMA kernels                                  |
412
413Examples available in this release:
414
415:::{list-table}
416:header-rows: 1
417:widths: auto
418* - User code
419  - Example
420* - `ceed`
421  - * ex1 (volume)
422* - `mfem`
423  - * BP1 (scalar mass operator)
424    * BP3 (scalar Laplace operator)
425* - `petsc`
426  - * BP1 (scalar mass operator)
427    * BP3 (scalar Laplace operator)
428* - `nek5000`
429  - * BP1 (scalar mass operator)
430    * BP3 (scalar Laplace operator)
431:::
432
433(v0-21)=
434
435## v0.21 (Sep 30, 2018)
436
437A MAGMA backend (which relies upon the
438[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
439release. This initial integration set up the framework of using MAGMA and provided the
440libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
441As any other backend, the MAGMA backend provides extended basic data structures for
442{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
443the fundamental CEED building blocks to work with the new data structures.
444In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
445but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
446internally, and thus seamlessly to the user, through the functions/methods that are
447provided to support them.
448
449Backends available in this release:
450
451| CEED resource (`-ceed`) | Backend                         |
452|-------------------------|---------------------------------|
453| `/cpu/self`             | Serial reference implementation |
454| `/cpu/occa`             | Serial OCCA kernels             |
455| `/gpu/occa`             | CUDA OCCA kernels               |
456| `/omp/occa`             | OpenMP OCCA kernels             |
457| `/ocl/occa`             | OpenCL OCCA kernels             |
458| `/gpu/magma`            | CUDA MAGMA kernels              |
459
460Examples available in this release:
461
462:::{list-table}
463:header-rows: 1
464:widths: auto
465* - User code
466  - Example
467* - `ceed`
468  - * ex1 (volume)
469* - `mfem`
470  - * BP1 (scalar mass operator)
471    * BP3 (scalar Laplace operator)
472* - `petsc`
473  - * BP1 (scalar mass operator)
474* - `nek5000`
475  - * BP1 (scalar mass operator)
476:::
477
478(v0-2)=
479
480## v0.2 (Mar 30, 2018)
481
482libCEED was made publicly available the first full CEED software distribution, release
483CEED 1.0. The distribution was made available using the Spack package manager to provide
484a common, easy-to-use build environment, where the user can build the CEED distribution
485with all dependencies. This release included a new Fortran interface for the library.
486This release also contained major improvements in the OCCA backend (including a new
487`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
488compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
489`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
490of the Laplace operator) was also added to this release.
491
492Backends available in this release:
493
494| CEED resource (`-ceed`) | Backend                         |
495|-------------------------|---------------------------------|
496| `/cpu/self`             | Serial reference implementation |
497| `/cpu/occa`             | Serial OCCA kernels             |
498| `/gpu/occa`             | CUDA OCCA kernels               |
499| `/omp/occa`             | OpenMP OCCA kernels             |
500| `/ocl/occa`             | OpenCL OCCA kernels             |
501
502Examples available in this release:
503
504:::{list-table}
505:header-rows: 1
506:widths: auto
507* - User code
508  - Example
509* - `ceed`
510  - * ex1 (volume)
511* - `mfem`
512  - * BP1 (scalar mass operator)
513    * BP3 (scalar Laplace operator)
514* - `petsc`
515  - * BP1 (scalar mass operator)
516* - `nek5000`
517  - * BP1 (scalar mass operator)
518:::
519
520(v0-1)=
521
522## v0.1 (Jan 3, 2018)
523
524Initial low-level API of the CEED project. The low-level API provides a set of Finite
525Elements kernels and components for writing new low-level kernels. Examples include:
526vector and sparse linear algebra, element matrix assembly over a batch of elements,
527partial assembly and action for efficient high-order operators like mass, diffusion,
528advection, etc. The main goal of the low-level API is to establish the basis for the
529high-level API. Also, identifying such low-level kernels and providing a reference
530implementation for them serves as the basis for specialized backend implementations.
531This release contained several backends: `/cpu/self`, and backends which rely upon the
532[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
533`/gpu/occa`, and `/omp/occa`.
534It also included several examples, in the `examples` folder:
535A standalone code that shows the usage of libCEED (with no external
536dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
537(with the application of the mass operator); and a `petsc` example to perform BP1
538(with the application of the mass operator).
539
540Backends available in this release:
541
542| CEED resource (`-ceed`) | Backend                         |
543|-------------------------|---------------------------------|
544| `/cpu/self`             | Serial reference implementation |
545| `/cpu/occa`             | Serial OCCA kernels             |
546| `/gpu/occa`             | CUDA OCCA kernels               |
547| `/omp/occa`             | OpenMP OCCA kernels             |
548
549Examples available in this release:
550
551| User code             | Example                           |
552|-----------------------|-----------------------------------|
553| `ceed`                | ex1 (scalar Laplace operator)     |
554| `mfem`                | BP1 (scalar mass operator)        |
555| `petsc`               | BP1 (scalar mass operator)        |
556```
557