xref: /libCEED/doc/sphinx/source/releasenotes.md (revision f1c40530bc6f02bc6ca847af718b8b41cc3fd3ca)
1# Changes/Release Notes
2
3On this page we provide a summary of the main API changes, new features and examples for each release of libCEED.
4
5(main)=
6
7## Current `main` branch
8
9### Interface changes
10
11- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
12- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
13
14### New features
15
16- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
17- Added `CeedInt_FMT` to support potential future use of larger interger sizes.
18- Added CEED_QFUNCTION_ATTR for setting compiler attributes/pragmas to CEED_QFUNCTION_HELPER and CEED_QFUNCTION
19- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
20Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
21This issue will be fixed in a future OCCA release.
22
23### Bugfix
24
25- Fix bug in setting device id for GPU backends.
26- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
27- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
28- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
29
30### Examples
31
32- Added various performance enhancements for {ref}`example-petsc-navier-stokes`.
33- Refactored {ref}`example-petsc-navier-stokes` to improve code reuse.
34- Added Shock Tube, Channel, and Flat Plate boundary layer problems to {ref}`example-petsc-navier-stokes`.
35- Added ability to use QFunctions for strong STG inflow in {ref}`example-petsc-navier-stokes`.
36
37### Maintainability
38
39- Refactored `/gpu/cuda/shared` and `/gpu/cuda/gen` as well as `/gpu/hip/shared` and `/gpu/hip/gen` backend to improve maintainablity and reduce duplicated code.
40- Enabled support for `p > 8` for `/gpu/*/shared` backends.
41
42(v0-10-1)=
43
44## v0.10.1 (Apr 11, 2022)
45
46### Interface changes
47
48- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
49
50### New features
51
52- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
53This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
54
55### Bugfix
56
57- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
58
59(v0-10)=
60
61## v0.10 (Mar 21, 2022)
62
63### Interface changes
64
65- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
66- 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`.
67- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
68- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
69- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
70- Warning added when compiling OCCA backend to alert users that this backend is experimental.
71- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
72- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
73- 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`.
74- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
75- 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.
76- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
77- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
78- 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.
79- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
80- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
81
82### New features
83
84- `CeedScalar` can now be set as `float` or `double` at compile time.
85- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
86- 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.
87- Remove need to guard library headers in QFunction source for code generation backends.
88- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
89- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
90- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
91- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
92- 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.
93- Added support for element matrix assembly in GPU backends.
94
95### Maintainability
96
97- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
98- `Include-what-you-use` makefile target added as `make iwyu`.
99- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
100- Put GPU JiTed kernel source code into separate files.
101- 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.
102
103(v0-9)=
104
105## v0.9 (Jul 6, 2021)
106
107### Interface changes
108
109- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
110
111### New features
112
113- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
114- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
115- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
116- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
117
118### Performance improvements
119
120### Examples
121
122- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
123- Fluid mechanics example adds GPU support and improves modularity.
124
125### Deprecated backends
126
127- 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.
128
129(v0-8)=
130
131## v0.8 (Mar 31, 2021)
132
133### Interface changes
134
135- Error handling improved to include enumerated error codes for C interface return values.
136- 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.
137
138### New features
139
140- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
141- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
142- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
143
144### Performance improvements
145
146- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
147- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
148
149### Examples
150
151- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
152- {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.
153- {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.
154- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
155
156(v0-7)=
157
158## v0.7 (Sep 29, 2020)
159
160### Interface changes
161
162- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
163  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
164  These changes improve support for mixed finite element methods.
165- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
166- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
167- 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.
168- 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.
169- 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.
170- 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.
171  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`.
172- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
173- 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.
174
175### New features
176
177- New HIP backend: `/gpu/hip/ref`.
178- CeedQFunction support for user `CUfunction`s in some backends
179
180### Performance improvements
181
182- OCCA backend rebuilt to facilitate future performance enhancements.
183- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
184
185### Examples
186
187- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
188
189### Deprecated backends
190
191- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
192
193(v0-6)=
194
195## v0.6 (Mar 29, 2020)
196
197libCEED v0.6 contains numerous new features and examples, as well as expanded
198documentation in [this new website](https://libceed.org).
199
200### New features
201
202- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
203  1-1 correspondence with the C interface, plus some convenience features.  For instance,
204  data stored in the {cpp:type}`CeedVector` structure are available without copy as
205  {py:class}`numpy.ndarray`.  Short tutorials are provided in
206  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
207- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
208  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
209  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
210  preconditioning ingredients and are used in the libCEED's multigrid examples.
211- The inverse of separable operators can be obtained using
212  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
213  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
214  especially for Laplacians and related operators.
215- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
216  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
217- Make public accessors for various attributes to facilitate writing composable code.
218- New backend: `/cpu/self/memcheck/serial`.
219- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
220  backends.  (Single source is coming soon for OCCA backends.)
221- Fix some missing edge cases in CUDA backend.
222
223### Performance Improvements
224
225- MAGMA backend performance optimization and non-tensor bases.
226- No-copy optimization in {c:func}`CeedOperatorApply`.
227
228### Interface changes
229
230- Replace {code}`CeedElemRestrictionCreateIdentity` and
231  {code}`CeedElemRestrictionCreateBlocked` with more flexible
232  {c:func}`CeedElemRestrictionCreateStrided` and
233  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
234- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
235- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
236  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
237  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
238  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
239  {c:func}`CeedElemRestrictionApply`.
240
241### Examples
242
243libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
244Notable additions include:
245
246- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
247  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
248
249- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
250  of domains (like the cube and sphere) by direct integration on a surface mesh;
251  demonstrates geometric dimension different from topological dimension.
252
253- PETSc {ref}`example-petsc-bps`:
254
255  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
256  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
257    and transparent CUDA support.
258  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
259    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
260    dimension different from topological dimension.
261
262- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
263  solver with algebraic multigrid coarse solve.
264
265- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
266  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
267  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
268  quasi-2D computational domain support.
269
270- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
271  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
272  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
273
274(v0-5)=
275
276## v0.5 (Sep 18, 2019)
277
278For this release, several improvements were made. Two new CUDA backends were added to
279the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
280performance using single-source {ref}`CeedQFunction`. From this release, users
281can define Q-Functions in a single source code independently of the targeted backend
282with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
283compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
284declaration, the {ref}`CeedQFunction` API has undergone a slight change:
285the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
286requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
287`QFunctionField` with eval mode `CEED EVAL GRAD`.
288
289Additionally, new CPU backends
290were included in this release, such as the `/cpu/self/opt/*` backends (which are
291written in pure C and use partial **E-vectors** to improve performance) and the
292`/cpu/self/ref/memcheck` backend (which relies upon the
293[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
294{ref}`CeedQFunction` have no undefined values).
295This release also included various performance improvements, bug fixes, new examples,
296and improved tests. Among these improvements, vectorized instructions for
297{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
298instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
299identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
300to include unstructured meshes handling were. For this expansion, the prior version of
301the PETSc BPs, which only included data associated with structured geometries, were
302renamed `bpsraw`, and the new version of the BPs, which can handle data associated
303with any unstructured geometry, were called `bps`. Additionally, other benchmark
304problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
305and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
306as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
307examples. Furthermoew, another standalone libCEED example, called `ex2`, which
308computes the surface area of a given mesh was added to this release.
309
310Backends available in this release:
311
312| CEED resource (`-ceed`)  | Backend                                             |
313|--------------------------|-----------------------------------------------------|
314| `/cpu/self/ref/serial`   | Serial reference implementation                     |
315| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
316| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
317| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
318| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
319| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
320| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
321| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
322| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
323| `/cpu/occa`              | Serial OCCA kernels                                 |
324| `/gpu/occa`              | CUDA OCCA kernels                                   |
325| `/omp/occa`              | OpenMP OCCA kernels                                 |
326| `/ocl/occa`              | OpenCL OCCA kernels                                 |
327| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
328| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
329| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
330| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
331| `/gpu/magma`             | CUDA MAGMA kernels                                  |
332
333Examples available in this release:
334
335:::{list-table}
336:header-rows: 1
337:widths: auto
338* - User code
339  - Example
340* - `ceed`
341  - * ex1 (volume)
342    * ex2 (surface)
343* - `mfem`
344  - * BP1 (scalar mass operator)
345    * BP3 (scalar Laplace operator)
346* - `petsc`
347  - * BP1 (scalar mass operator)
348    * BP2 (vector mass operator)
349    * BP3 (scalar Laplace operator)
350    * BP4 (vector Laplace operator)
351    * BP5 (collocated scalar Laplace operator)
352    * BP6 (collocated vector Laplace operator)
353    * Navier-Stokes
354* - `nek5000`
355  - * BP1 (scalar mass operator)
356    * BP3 (scalar Laplace operator)
357:::
358
359(v0-4)=
360
361## v0.4 (Apr 1, 2019)
362
363libCEED v0.4 was made again publicly available in the second full CEED software
364distribution, release CEED 2.0. This release contained notable features, such as
365four new CPU backends, two new GPU backends, CPU backend optimizations, initial
366support for operator composition, performance benchmarking, and a Navier-Stokes demo.
367The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
368backends process one element at a time and are intended for meshes with a smaller number
369of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
370eight interlaced elements and are intended for meshes with higher numbers of elements.
371The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
372performance. The `/cpu/self/xsmm/*` backends rely upon the
373[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
374performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
375The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
376performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
377parallelization approach, where each thread treats a finite element. Using just in time
378compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
379backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
380achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
381deteriorates very quickly when threads run out of registers.
382
383A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
384examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
385This example solves the time-dependent Navier-Stokes equations of compressible gas
386dynamics in a static Eulerian three-dimensional frame, using structured high-order
387finite/spectral element spatial discretizations and explicit high-order time-stepping
388(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
389so that the pointwise physics (defined at quadrature points) is separated from the
390parallelization and meshing concerns.
391
392Backends available in this release:
393
394| CEED resource (`-ceed`)  | Backend                                             |
395|--------------------------|-----------------------------------------------------|
396| `/cpu/self/ref/serial`   | Serial reference implementation                     |
397| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
398| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
399| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
400| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
401| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
402| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
403| `/cpu/occa`              | Serial OCCA kernels                                 |
404| `/gpu/occa`              | CUDA OCCA kernels                                   |
405| `/omp/occa`              | OpenMP OCCA kernels                                 |
406| `/ocl/occa`              | OpenCL OCCA kernels                                 |
407| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
408| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
409| `/gpu/magma`             | CUDA MAGMA kernels                                  |
410
411Examples available in this release:
412
413:::{list-table}
414:header-rows: 1
415:widths: auto
416* - User code
417  - Example
418* - `ceed`
419  - * ex1 (volume)
420* - `mfem`
421  - * BP1 (scalar mass operator)
422    * BP3 (scalar Laplace operator)
423* - `petsc`
424  - * BP1 (scalar mass operator)
425    * BP3 (scalar Laplace operator)
426    * Navier-Stokes
427* - `nek5000`
428  - * BP1 (scalar mass operator)
429    * BP3 (scalar Laplace operator)
430:::
431
432(v0-3)=
433
434## v0.3 (Sep 30, 2018)
435
436Notable features in this release include active/passive field interface, support for
437non-tensor bases, backend optimization, and improved Fortran interface. This release
438also focused on providing improved continuous integration, and many new tests with code
439coverage reports of about 90%. This release also provided a significant change to the
440public interface: a {ref}`CeedQFunction` can take any number of named input and output
441arguments while {ref}`CeedOperator` connects them to the actual data, which may be
442supplied explicitly to `CeedOperatorApply()` (active) or separately via
443`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
444of CeedQFunctions and composition of block solvers constructed using
445{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
446used in an optimized CPU backend. Although this is typically not visible to the user,
447it enables effective use of arbitrary-length SIMD while maintaining cache locality.
448This CPU backend also implements an algebraic factorization of tensor product gradients
449to perform fewer operations than standard application of interpolation and
450differentiation from nodes to quadrature points. This algebraic formulation
451automatically supports non-polynomial and non-interpolatory bases, thus is more general
452than the more common derivation in terms of Lagrange polynomials on the quadrature points.
453
454Backends available in this release:
455
456| CEED resource (`-ceed`) | Backend                                             |
457|-------------------------|-----------------------------------------------------|
458| `/cpu/self/blocked`     | Blocked reference implementation                    |
459| `/cpu/self/ref`         | Serial reference implementation                     |
460| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
461| `/cpu/occa`             | Serial OCCA kernels                                 |
462| `/gpu/occa`             | CUDA OCCA kernels                                   |
463| `/omp/occa`             | OpenMP OCCA kernels                                 |
464| `/ocl/occa`             | OpenCL OCCA kernels                                 |
465| `/gpu/magma`            | CUDA MAGMA kernels                                  |
466
467Examples available in this release:
468
469:::{list-table}
470:header-rows: 1
471:widths: auto
472* - User code
473  - Example
474* - `ceed`
475  - * ex1 (volume)
476* - `mfem`
477  - * BP1 (scalar mass operator)
478    * BP3 (scalar Laplace operator)
479* - `petsc`
480  - * BP1 (scalar mass operator)
481    * BP3 (scalar Laplace operator)
482* - `nek5000`
483  - * BP1 (scalar mass operator)
484    * BP3 (scalar Laplace operator)
485:::
486
487(v0-21)=
488
489## v0.21 (Sep 30, 2018)
490
491A MAGMA backend (which relies upon the
492[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
493release. This initial integration set up the framework of using MAGMA and provided the
494libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
495As any other backend, the MAGMA backend provides extended basic data structures for
496{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
497the fundamental CEED building blocks to work with the new data structures.
498In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
499but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
500internally, and thus seamlessly to the user, through the functions/methods that are
501provided to support them.
502
503Backends available in this release:
504
505| CEED resource (`-ceed`) | Backend                         |
506|-------------------------|---------------------------------|
507| `/cpu/self`             | Serial reference implementation |
508| `/cpu/occa`             | Serial OCCA kernels             |
509| `/gpu/occa`             | CUDA OCCA kernels               |
510| `/omp/occa`             | OpenMP OCCA kernels             |
511| `/ocl/occa`             | OpenCL OCCA kernels             |
512| `/gpu/magma`            | CUDA MAGMA kernels              |
513
514Examples available in this release:
515
516:::{list-table}
517:header-rows: 1
518:widths: auto
519* - User code
520  - Example
521* - `ceed`
522  - * ex1 (volume)
523* - `mfem`
524  - * BP1 (scalar mass operator)
525    * BP3 (scalar Laplace operator)
526* - `petsc`
527  - * BP1 (scalar mass operator)
528* - `nek5000`
529  - * BP1 (scalar mass operator)
530:::
531
532(v0-2)=
533
534## v0.2 (Mar 30, 2018)
535
536libCEED was made publicly available the first full CEED software distribution, release
537CEED 1.0. The distribution was made available using the Spack package manager to provide
538a common, easy-to-use build environment, where the user can build the CEED distribution
539with all dependencies. This release included a new Fortran interface for the library.
540This release also contained major improvements in the OCCA backend (including a new
541`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
542compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
543`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
544of the Laplace operator) was also added to this release.
545
546Backends available in this release:
547
548| CEED resource (`-ceed`) | Backend                         |
549|-------------------------|---------------------------------|
550| `/cpu/self`             | Serial reference implementation |
551| `/cpu/occa`             | Serial OCCA kernels             |
552| `/gpu/occa`             | CUDA OCCA kernels               |
553| `/omp/occa`             | OpenMP OCCA kernels             |
554| `/ocl/occa`             | OpenCL OCCA kernels             |
555
556Examples available in this release:
557
558:::{list-table}
559:header-rows: 1
560:widths: auto
561* - User code
562  - Example
563* - `ceed`
564  - * ex1 (volume)
565* - `mfem`
566  - * BP1 (scalar mass operator)
567    * BP3 (scalar Laplace operator)
568* - `petsc`
569  - * BP1 (scalar mass operator)
570* - `nek5000`
571  - * BP1 (scalar mass operator)
572:::
573
574(v0-1)=
575
576## v0.1 (Jan 3, 2018)
577
578Initial low-level API of the CEED project. The low-level API provides a set of Finite
579Elements kernels and components for writing new low-level kernels. Examples include:
580vector and sparse linear algebra, element matrix assembly over a batch of elements,
581partial assembly and action for efficient high-order operators like mass, diffusion,
582advection, etc. The main goal of the low-level API is to establish the basis for the
583high-level API. Also, identifying such low-level kernels and providing a reference
584implementation for them serves as the basis for specialized backend implementations.
585This release contained several backends: `/cpu/self`, and backends which rely upon the
586[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
587`/gpu/occa`, and `/omp/occa`.
588It also included several examples, in the `examples` folder:
589A standalone code that shows the usage of libCEED (with no external
590dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
591(with the application of the mass operator); and a `petsc` example to perform BP1
592(with the application of the mass operator).
593
594Backends available in this release:
595
596| CEED resource (`-ceed`) | Backend                         |
597|-------------------------|---------------------------------|
598| `/cpu/self`             | Serial reference implementation |
599| `/cpu/occa`             | Serial OCCA kernels             |
600| `/gpu/occa`             | CUDA OCCA kernels               |
601| `/omp/occa`             | OpenMP OCCA kernels             |
602
603Examples available in this release:
604
605| User code             | Example                           |
606|-----------------------|-----------------------------------|
607| `ceed`                | ex1 (scalar Laplace operator)     |
608| `mfem`                | BP1 (scalar mass operator)        |
609| `petsc`               | BP1 (scalar mass operator)        |
610```
611