xref: /libCEED/doc/sphinx/source/releasenotes.md (revision caee03026e6576cbf7a399c2fc51bb918c77f451)
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### Other
24
25- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
26
27### Bugfix
28
29- Fix bug in setting device id for GPU backends.
30- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
31- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
32- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
33
34### Examples
35
36- Added various performance enhancements for {ref}`example-petsc-navier-stokes`.
37- Refactored {ref}`example-petsc-navier-stokes` to improve code reuse.
38- Added Shock Tube, Channel, and Flat Plate boundary layer problems to {ref}`example-petsc-navier-stokes`.
39- Added ability to use QFunctions for strong STG inflow in {ref}`example-petsc-navier-stokes`.
40
41### Maintainability
42
43- 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.
44- Enabled support for `p > 8` for `/gpu/*/shared` backends.
45
46(v0-10-1)=
47
48## v0.10.1 (Apr 11, 2022)
49
50### Interface changes
51
52- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
53
54### New features
55
56- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
57This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
58
59### Bugfix
60
61- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
62
63(v0-10)=
64
65## v0.10 (Mar 21, 2022)
66
67### Interface changes
68
69- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
70- 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`.
71- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
72- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
73- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
74- Warning added when compiling OCCA backend to alert users that this backend is experimental.
75- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
76- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
77- 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`.
78- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
79- 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.
80- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
81- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
82- 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.
83- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
84- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
85
86### New features
87
88- `CeedScalar` can now be set as `float` or `double` at compile time.
89- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
90- 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.
91- Remove need to guard library headers in QFunction source for code generation backends.
92- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
93- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
94- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
95- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
96- 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.
97- Added support for element matrix assembly in GPU backends.
98
99### Maintainability
100
101- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
102- `Include-what-you-use` makefile target added as `make iwyu`.
103- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
104- Put GPU JiTed kernel source code into separate files.
105- 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.
106
107(v0-9)=
108
109## v0.9 (Jul 6, 2021)
110
111### Interface changes
112
113- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
114
115### New features
116
117- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
118- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
119- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
120- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
121
122### Performance improvements
123
124### Examples
125
126- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
127- Fluid mechanics example adds GPU support and improves modularity.
128
129### Deprecated backends
130
131- 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.
132
133(v0-8)=
134
135## v0.8 (Mar 31, 2021)
136
137### Interface changes
138
139- Error handling improved to include enumerated error codes for C interface return values.
140- 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.
141
142### New features
143
144- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
145- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
146- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
147
148### Performance improvements
149
150- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
151- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
152
153### Examples
154
155- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
156- {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.
157- {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.
158- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
159
160(v0-7)=
161
162## v0.7 (Sep 29, 2020)
163
164### Interface changes
165
166- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
167  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
168  These changes improve support for mixed finite element methods.
169- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
170- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
171- 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.
172- 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.
173- 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.
174- 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.
175  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`.
176- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
177- 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.
178
179### New features
180
181- New HIP backend: `/gpu/hip/ref`.
182- CeedQFunction support for user `CUfunction`s in some backends
183
184### Performance improvements
185
186- OCCA backend rebuilt to facilitate future performance enhancements.
187- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
188
189### Examples
190
191- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
192
193### Deprecated backends
194
195- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
196
197(v0-6)=
198
199## v0.6 (Mar 29, 2020)
200
201libCEED v0.6 contains numerous new features and examples, as well as expanded
202documentation in [this new website](https://libceed.org).
203
204### New features
205
206- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
207  1-1 correspondence with the C interface, plus some convenience features.  For instance,
208  data stored in the {cpp:type}`CeedVector` structure are available without copy as
209  {py:class}`numpy.ndarray`.  Short tutorials are provided in
210  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
211- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
212  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
213  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
214  preconditioning ingredients and are used in the libCEED's multigrid examples.
215- The inverse of separable operators can be obtained using
216  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
217  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
218  especially for Laplacians and related operators.
219- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
220  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
221- Make public accessors for various attributes to facilitate writing composable code.
222- New backend: `/cpu/self/memcheck/serial`.
223- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
224  backends.  (Single source is coming soon for OCCA backends.)
225- Fix some missing edge cases in CUDA backend.
226
227### Performance Improvements
228
229- MAGMA backend performance optimization and non-tensor bases.
230- No-copy optimization in {c:func}`CeedOperatorApply`.
231
232### Interface changes
233
234- Replace {code}`CeedElemRestrictionCreateIdentity` and
235  {code}`CeedElemRestrictionCreateBlocked` with more flexible
236  {c:func}`CeedElemRestrictionCreateStrided` and
237  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
238- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
239- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
240  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
241  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
242  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
243  {c:func}`CeedElemRestrictionApply`.
244
245### Examples
246
247libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
248Notable additions include:
249
250- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
251  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
252
253- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
254  of domains (like the cube and sphere) by direct integration on a surface mesh;
255  demonstrates geometric dimension different from topological dimension.
256
257- PETSc {ref}`example-petsc-bps`:
258
259  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
260  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
261    and transparent CUDA support.
262  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
263    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
264    dimension different from topological dimension.
265
266- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
267  solver with algebraic multigrid coarse solve.
268
269- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
270  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
271  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
272  quasi-2D computational domain support.
273
274- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
275  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
276  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
277
278(v0-5)=
279
280## v0.5 (Sep 18, 2019)
281
282For this release, several improvements were made. Two new CUDA backends were added to
283the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
284performance using single-source {ref}`CeedQFunction`. From this release, users
285can define Q-Functions in a single source code independently of the targeted backend
286with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
287compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
288declaration, the {ref}`CeedQFunction` API has undergone a slight change:
289the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
290requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
291`QFunctionField` with eval mode `CEED EVAL GRAD`.
292
293Additionally, new CPU backends
294were included in this release, such as the `/cpu/self/opt/*` backends (which are
295written in pure C and use partial **E-vectors** to improve performance) and the
296`/cpu/self/ref/memcheck` backend (which relies upon the
297[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
298{ref}`CeedQFunction` have no undefined values).
299This release also included various performance improvements, bug fixes, new examples,
300and improved tests. Among these improvements, vectorized instructions for
301{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
302instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
303identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
304to include unstructured meshes handling were. For this expansion, the prior version of
305the PETSc BPs, which only included data associated with structured geometries, were
306renamed `bpsraw`, and the new version of the BPs, which can handle data associated
307with any unstructured geometry, were called `bps`. Additionally, other benchmark
308problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
309and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
310as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
311examples. Furthermoew, another standalone libCEED example, called `ex2`, which
312computes the surface area of a given mesh was added to this release.
313
314Backends available in this release:
315
316| CEED resource (`-ceed`)  | Backend                                             |
317|--------------------------|-----------------------------------------------------|
318| `/cpu/self/ref/serial`   | Serial reference implementation                     |
319| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
320| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
321| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
322| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
323| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
324| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
325| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
326| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
327| `/cpu/occa`              | Serial OCCA kernels                                 |
328| `/gpu/occa`              | CUDA OCCA kernels                                   |
329| `/omp/occa`              | OpenMP OCCA kernels                                 |
330| `/ocl/occa`              | OpenCL OCCA kernels                                 |
331| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
332| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
333| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
334| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
335| `/gpu/magma`             | CUDA MAGMA kernels                                  |
336
337Examples available in this release:
338
339:::{list-table}
340:header-rows: 1
341:widths: auto
342* - User code
343  - Example
344* - `ceed`
345  - * ex1 (volume)
346    * ex2 (surface)
347* - `mfem`
348  - * BP1 (scalar mass operator)
349    * BP3 (scalar Laplace operator)
350* - `petsc`
351  - * BP1 (scalar mass operator)
352    * BP2 (vector mass operator)
353    * BP3 (scalar Laplace operator)
354    * BP4 (vector Laplace operator)
355    * BP5 (collocated scalar Laplace operator)
356    * BP6 (collocated vector Laplace operator)
357    * Navier-Stokes
358* - `nek5000`
359  - * BP1 (scalar mass operator)
360    * BP3 (scalar Laplace operator)
361:::
362
363(v0-4)=
364
365## v0.4 (Apr 1, 2019)
366
367libCEED v0.4 was made again publicly available in the second full CEED software
368distribution, release CEED 2.0. This release contained notable features, such as
369four new CPU backends, two new GPU backends, CPU backend optimizations, initial
370support for operator composition, performance benchmarking, and a Navier-Stokes demo.
371The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
372backends process one element at a time and are intended for meshes with a smaller number
373of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
374eight interlaced elements and are intended for meshes with higher numbers of elements.
375The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
376performance. The `/cpu/self/xsmm/*` backends rely upon the
377[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
378performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
379The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
380performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
381parallelization approach, where each thread treats a finite element. Using just in time
382compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
383backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
384achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
385deteriorates very quickly when threads run out of registers.
386
387A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
388examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
389This example solves the time-dependent Navier-Stokes equations of compressible gas
390dynamics in a static Eulerian three-dimensional frame, using structured high-order
391finite/spectral element spatial discretizations and explicit high-order time-stepping
392(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
393so that the pointwise physics (defined at quadrature points) is separated from the
394parallelization and meshing concerns.
395
396Backends available in this release:
397
398| CEED resource (`-ceed`)  | Backend                                             |
399|--------------------------|-----------------------------------------------------|
400| `/cpu/self/ref/serial`   | Serial reference implementation                     |
401| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
402| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
403| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
404| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
405| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
406| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
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/cuda/ref`          | Reference pure CUDA kernels                         |
412| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
413| `/gpu/magma`             | CUDA MAGMA kernels                                  |
414
415Examples available in this release:
416
417:::{list-table}
418:header-rows: 1
419:widths: auto
420* - User code
421  - Example
422* - `ceed`
423  - * ex1 (volume)
424* - `mfem`
425  - * BP1 (scalar mass operator)
426    * BP3 (scalar Laplace operator)
427* - `petsc`
428  - * BP1 (scalar mass operator)
429    * BP3 (scalar Laplace operator)
430    * Navier-Stokes
431* - `nek5000`
432  - * BP1 (scalar mass operator)
433    * BP3 (scalar Laplace operator)
434:::
435
436(v0-3)=
437
438## v0.3 (Sep 30, 2018)
439
440Notable features in this release include active/passive field interface, support for
441non-tensor bases, backend optimization, and improved Fortran interface. This release
442also focused on providing improved continuous integration, and many new tests with code
443coverage reports of about 90%. This release also provided a significant change to the
444public interface: a {ref}`CeedQFunction` can take any number of named input and output
445arguments while {ref}`CeedOperator` connects them to the actual data, which may be
446supplied explicitly to `CeedOperatorApply()` (active) or separately via
447`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
448of CeedQFunctions and composition of block solvers constructed using
449{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
450used in an optimized CPU backend. Although this is typically not visible to the user,
451it enables effective use of arbitrary-length SIMD while maintaining cache locality.
452This CPU backend also implements an algebraic factorization of tensor product gradients
453to perform fewer operations than standard application of interpolation and
454differentiation from nodes to quadrature points. This algebraic formulation
455automatically supports non-polynomial and non-interpolatory bases, thus is more general
456than the more common derivation in terms of Lagrange polynomials on the quadrature points.
457
458Backends available in this release:
459
460| CEED resource (`-ceed`) | Backend                                             |
461|-------------------------|-----------------------------------------------------|
462| `/cpu/self/blocked`     | Blocked reference implementation                    |
463| `/cpu/self/ref`         | Serial reference implementation                     |
464| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
465| `/cpu/occa`             | Serial OCCA kernels                                 |
466| `/gpu/occa`             | CUDA OCCA kernels                                   |
467| `/omp/occa`             | OpenMP OCCA kernels                                 |
468| `/ocl/occa`             | OpenCL OCCA kernels                                 |
469| `/gpu/magma`            | CUDA MAGMA kernels                                  |
470
471Examples available in this release:
472
473:::{list-table}
474:header-rows: 1
475:widths: auto
476* - User code
477  - Example
478* - `ceed`
479  - * ex1 (volume)
480* - `mfem`
481  - * BP1 (scalar mass operator)
482    * BP3 (scalar Laplace operator)
483* - `petsc`
484  - * BP1 (scalar mass operator)
485    * BP3 (scalar Laplace operator)
486* - `nek5000`
487  - * BP1 (scalar mass operator)
488    * BP3 (scalar Laplace operator)
489:::
490
491(v0-21)=
492
493## v0.21 (Sep 30, 2018)
494
495A MAGMA backend (which relies upon the
496[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
497release. This initial integration set up the framework of using MAGMA and provided the
498libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
499As any other backend, the MAGMA backend provides extended basic data structures for
500{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
501the fundamental CEED building blocks to work with the new data structures.
502In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
503but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
504internally, and thus seamlessly to the user, through the functions/methods that are
505provided to support them.
506
507Backends available in this release:
508
509| CEED resource (`-ceed`) | Backend                         |
510|-------------------------|---------------------------------|
511| `/cpu/self`             | Serial reference implementation |
512| `/cpu/occa`             | Serial OCCA kernels             |
513| `/gpu/occa`             | CUDA OCCA kernels               |
514| `/omp/occa`             | OpenMP OCCA kernels             |
515| `/ocl/occa`             | OpenCL OCCA kernels             |
516| `/gpu/magma`            | CUDA MAGMA kernels              |
517
518Examples available in this release:
519
520:::{list-table}
521:header-rows: 1
522:widths: auto
523* - User code
524  - Example
525* - `ceed`
526  - * ex1 (volume)
527* - `mfem`
528  - * BP1 (scalar mass operator)
529    * BP3 (scalar Laplace operator)
530* - `petsc`
531  - * BP1 (scalar mass operator)
532* - `nek5000`
533  - * BP1 (scalar mass operator)
534:::
535
536(v0-2)=
537
538## v0.2 (Mar 30, 2018)
539
540libCEED was made publicly available the first full CEED software distribution, release
541CEED 1.0. The distribution was made available using the Spack package manager to provide
542a common, easy-to-use build environment, where the user can build the CEED distribution
543with all dependencies. This release included a new Fortran interface for the library.
544This release also contained major improvements in the OCCA backend (including a new
545`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
546compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
547`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
548of the Laplace operator) was also added to this release.
549
550Backends available in this release:
551
552| CEED resource (`-ceed`) | Backend                         |
553|-------------------------|---------------------------------|
554| `/cpu/self`             | Serial reference implementation |
555| `/cpu/occa`             | Serial OCCA kernels             |
556| `/gpu/occa`             | CUDA OCCA kernels               |
557| `/omp/occa`             | OpenMP OCCA kernels             |
558| `/ocl/occa`             | OpenCL OCCA kernels             |
559
560Examples available in this release:
561
562:::{list-table}
563:header-rows: 1
564:widths: auto
565* - User code
566  - Example
567* - `ceed`
568  - * ex1 (volume)
569* - `mfem`
570  - * BP1 (scalar mass operator)
571    * BP3 (scalar Laplace operator)
572* - `petsc`
573  - * BP1 (scalar mass operator)
574* - `nek5000`
575  - * BP1 (scalar mass operator)
576:::
577
578(v0-1)=
579
580## v0.1 (Jan 3, 2018)
581
582Initial low-level API of the CEED project. The low-level API provides a set of Finite
583Elements kernels and components for writing new low-level kernels. Examples include:
584vector and sparse linear algebra, element matrix assembly over a batch of elements,
585partial assembly and action for efficient high-order operators like mass, diffusion,
586advection, etc. The main goal of the low-level API is to establish the basis for the
587high-level API. Also, identifying such low-level kernels and providing a reference
588implementation for them serves as the basis for specialized backend implementations.
589This release contained several backends: `/cpu/self`, and backends which rely upon the
590[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
591`/gpu/occa`, and `/omp/occa`.
592It also included several examples, in the `examples` folder:
593A standalone code that shows the usage of libCEED (with no external
594dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
595(with the application of the mass operator); and a `petsc` example to perform BP1
596(with the application of the mass operator).
597
598Backends available in this release:
599
600| CEED resource (`-ceed`) | Backend                         |
601|-------------------------|---------------------------------|
602| `/cpu/self`             | Serial reference implementation |
603| `/cpu/occa`             | Serial OCCA kernels             |
604| `/gpu/occa`             | CUDA OCCA kernels               |
605| `/omp/occa`             | OpenMP OCCA kernels             |
606
607Examples available in this release:
608
609| User code             | Example                           |
610|-----------------------|-----------------------------------|
611| `ceed`                | ex1 (scalar Laplace operator)     |
612| `mfem`                | BP1 (scalar mass operator)        |
613| `petsc`               | BP1 (scalar mass operator)        |
614```
615