xref: /libCEED/doc/sphinx/source/releasenotes.md (revision a2cba4be6838e1a18742bb6cef1b14b2cf87e1c8)
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
13### New features
14
15- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
16
17### Bugfix
18
19- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
20- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
21- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
22
23### Examples
24
25- Added various performance enhancements for {ref}`example-petsc-navier-stokes`
26
27(v0-10-1)=
28
29## v0.10.1 (Apr 11, 2022)
30
31### Interface changes
32
33- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
34
35### Bugfix
36
37- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
38
39(v0-10)=
40
41## v0.10 (Mar 21, 2022)
42
43### Interface changes
44
45- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
46- 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`.
47- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
48- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
49- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
50- Warning added when compiling OCCA backend to alert users that this backend is experimental.
51- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
52- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
53- 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`.
54- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
55- 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.
56- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
57- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
58- 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.
59- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
60- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
61
62### New features
63
64- `CeedScalar` can now be set as `float` or `double` at compile time.
65- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
66- 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.
67- Remove need to guard library headers in QFunction source for code generation backends.
68- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
69- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
70- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
71- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
72- 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.
73- Added support for element matrix assembly in GPU backends.
74
75### Maintainability
76
77- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
78- `Include-what-you-use` makefile target added as `make iwyu`.
79- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
80- Put GPU JiTed kernel source code into separate files.
81- 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.
82
83(v0-9)=
84
85## v0.9 (Jul 6, 2021)
86
87### Interface changes
88
89- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
90
91### New features
92
93- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
94- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
95- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
96- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
97
98### Performance improvements
99
100### Examples
101
102- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
103- Fluid mechanics example adds GPU support and improves modularity.
104
105### Deprecated backends
106
107- 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.
108
109(v0-8)=
110
111## v0.8 (Mar 31, 2021)
112
113### Interface changes
114
115- Error handling improved to include enumerated error codes for C interface return values.
116- 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.
117
118### New features
119
120- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
121- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
122- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
123
124### Performance improvements
125
126- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
127- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
128
129### Examples
130
131- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
132- {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.
133- {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.
134- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
135
136(v0-7)=
137
138## v0.7 (Sep 29, 2020)
139
140### Interface changes
141
142- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
143  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
144  These changes improve support for mixed finite element methods.
145- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
146- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
147- 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.
148- 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.
149- 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.
150- 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.
151  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`.
152- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
153- 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.
154
155### New features
156
157- New HIP backend: `/gpu/hip/ref`.
158- CeedQFunction support for user `CUfunction`s in some backends
159
160### Performance improvements
161
162- OCCA backend rebuilt to facilitate future performance enhancements.
163- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
164
165### Examples
166
167- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
168
169### Deprecated backends
170
171- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
172
173(v0-6)=
174
175## v0.6 (Mar 29, 2020)
176
177libCEED v0.6 contains numerous new features and examples, as well as expanded
178documentation in [this new website](https://libceed.org).
179
180### New features
181
182- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
183  1-1 correspondence with the C interface, plus some convenience features.  For instance,
184  data stored in the {cpp:type}`CeedVector` structure are available without copy as
185  {py:class}`numpy.ndarray`.  Short tutorials are provided in
186  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
187- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
188  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
189  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
190  preconditioning ingredients and are used in the libCEED's multigrid examples.
191- The inverse of separable operators can be obtained using
192  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
193  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
194  especially for Laplacians and related operators.
195- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
196  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
197- Make public accessors for various attributes to facilitate writing composable code.
198- New backend: `/cpu/self/memcheck/serial`.
199- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
200  backends.  (Single source is coming soon for OCCA backends.)
201- Fix some missing edge cases in CUDA backend.
202
203### Performance Improvements
204
205- MAGMA backend performance optimization and non-tensor bases.
206- No-copy optimization in {c:func}`CeedOperatorApply`.
207
208### Interface changes
209
210- Replace {code}`CeedElemRestrictionCreateIdentity` and
211  {code}`CeedElemRestrictionCreateBlocked` with more flexible
212  {c:func}`CeedElemRestrictionCreateStrided` and
213  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
214- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
215- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
216  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
217  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
218  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
219  {c:func}`CeedElemRestrictionApply`.
220
221### Examples
222
223libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
224Notable additions include:
225
226- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
227  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
228
229- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
230  of domains (like the cube and sphere) by direct integration on a surface mesh;
231  demonstrates geometric dimension different from topological dimension.
232
233- PETSc {ref}`example-petsc-bps`:
234
235  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
236  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
237    and transparent CUDA support.
238  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
239    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
240    dimension different from topological dimension.
241
242- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
243  solver with algebraic multigrid coarse solve.
244
245- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
246  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
247  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
248  quasi-2D computational domain support.
249
250- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
251  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
252  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
253
254(v0-5)=
255
256## v0.5 (Sep 18, 2019)
257
258For this release, several improvements were made. Two new CUDA backends were added to
259the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
260performance using single-source {ref}`CeedQFunction`. From this release, users
261can define Q-Functions in a single source code independently of the targeted backend
262with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
263compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
264declaration, the {ref}`CeedQFunction` API has undergone a slight change:
265the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
266requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
267`QFunctionField` with eval mode `CEED EVAL GRAD`.
268
269Additionally, new CPU backends
270were included in this release, such as the `/cpu/self/opt/*` backends (which are
271written in pure C and use partial **E-vectors** to improve performance) and the
272`/cpu/self/ref/memcheck` backend (which relies upon the
273[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
274{ref}`CeedQFunction` have no undefined values).
275This release also included various performance improvements, bug fixes, new examples,
276and improved tests. Among these improvements, vectorized instructions for
277{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
278instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
279identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
280to include unstructured meshes handling were. For this expansion, the prior version of
281the PETSc BPs, which only included data associated with structured geometries, were
282renamed `bpsraw`, and the new version of the BPs, which can handle data associated
283with any unstructured geometry, were called `bps`. Additionally, other benchmark
284problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
285and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
286as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
287examples. Furthermoew, another standalone libCEED example, called `ex2`, which
288computes the surface area of a given mesh was added to this release.
289
290Backends available in this release:
291
292| CEED resource (`-ceed`)  | Backend                                             |
293|--------------------------|-----------------------------------------------------|
294| `/cpu/self/ref/serial`   | Serial reference implementation                     |
295| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
296| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
297| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
298| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
299| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
300| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
301| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
302| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
303| `/cpu/occa`              | Serial OCCA kernels                                 |
304| `/gpu/occa`              | CUDA OCCA kernels                                   |
305| `/omp/occa`              | OpenMP OCCA kernels                                 |
306| `/ocl/occa`              | OpenCL OCCA kernels                                 |
307| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
308| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
309| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
310| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
311| `/gpu/magma`             | CUDA MAGMA kernels                                  |
312
313Examples available in this release:
314
315:::{list-table}
316:header-rows: 1
317:widths: auto
318* - User code
319  - Example
320* - `ceed`
321  - * ex1 (volume)
322    * ex2 (surface)
323* - `mfem`
324  - * BP1 (scalar mass operator)
325    * BP3 (scalar Laplace operator)
326* - `petsc`
327  - * BP1 (scalar mass operator)
328    * BP2 (vector mass operator)
329    * BP3 (scalar Laplace operator)
330    * BP4 (vector Laplace operator)
331    * BP5 (collocated scalar Laplace operator)
332    * BP6 (collocated vector Laplace operator)
333    * Navier-Stokes
334* - `nek5000`
335  - * BP1 (scalar mass operator)
336    * BP3 (scalar Laplace operator)
337:::
338
339(v0-4)=
340
341## v0.4 (Apr 1, 2019)
342
343libCEED v0.4 was made again publicly available in the second full CEED software
344distribution, release CEED 2.0. This release contained notable features, such as
345four new CPU backends, two new GPU backends, CPU backend optimizations, initial
346support for operator composition, performance benchmarking, and a Navier-Stokes demo.
347The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
348backends process one element at a time and are intended for meshes with a smaller number
349of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
350eight interlaced elements and are intended for meshes with higher numbers of elements.
351The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
352performance. The `/cpu/self/xsmm/*` backends rely upon the
353[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
354performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
355The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
356performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
357parallelization approach, where each thread treats a finite element. Using just in time
358compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
359backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
360achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
361deteriorates very quickly when threads run out of registers.
362
363A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
364examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
365This example solves the time-dependent Navier-Stokes equations of compressible gas
366dynamics in a static Eulerian three-dimensional frame, using structured high-order
367finite/spectral element spatial discretizations and explicit high-order time-stepping
368(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
369so that the pointwise physics (defined at quadrature points) is separated from the
370parallelization and meshing concerns.
371
372Backends available in this release:
373
374| CEED resource (`-ceed`)  | Backend                                             |
375|--------------------------|-----------------------------------------------------|
376| `/cpu/self/ref/serial`   | Serial reference implementation                     |
377| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
378| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
379| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
380| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
381| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
382| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
383| `/cpu/occa`              | Serial OCCA kernels                                 |
384| `/gpu/occa`              | CUDA OCCA kernels                                   |
385| `/omp/occa`              | OpenMP OCCA kernels                                 |
386| `/ocl/occa`              | OpenCL OCCA kernels                                 |
387| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
388| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
389| `/gpu/magma`             | CUDA MAGMA kernels                                  |
390
391Examples available in this release:
392
393:::{list-table}
394:header-rows: 1
395:widths: auto
396* - User code
397  - Example
398* - `ceed`
399  - * ex1 (volume)
400* - `mfem`
401  - * BP1 (scalar mass operator)
402    * BP3 (scalar Laplace operator)
403* - `petsc`
404  - * BP1 (scalar mass operator)
405    * BP3 (scalar Laplace operator)
406    * Navier-Stokes
407* - `nek5000`
408  - * BP1 (scalar mass operator)
409    * BP3 (scalar Laplace operator)
410:::
411
412(v0-3)=
413
414## v0.3 (Sep 30, 2018)
415
416Notable features in this release include active/passive field interface, support for
417non-tensor bases, backend optimization, and improved Fortran interface. This release
418also focused on providing improved continuous integration, and many new tests with code
419coverage reports of about 90%. This release also provided a significant change to the
420public interface: a {ref}`CeedQFunction` can take any number of named input and output
421arguments while {ref}`CeedOperator` connects them to the actual data, which may be
422supplied explicitly to `CeedOperatorApply()` (active) or separately via
423`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
424of CeedQFunctions and composition of block solvers constructed using
425{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
426used in an optimized CPU backend. Although this is typically not visible to the user,
427it enables effective use of arbitrary-length SIMD while maintaining cache locality.
428This CPU backend also implements an algebraic factorization of tensor product gradients
429to perform fewer operations than standard application of interpolation and
430differentiation from nodes to quadrature points. This algebraic formulation
431automatically supports non-polynomial and non-interpolatory bases, thus is more general
432than the more common derivation in terms of Lagrange polynomials on the quadrature points.
433
434Backends available in this release:
435
436| CEED resource (`-ceed`) | Backend                                             |
437|-------------------------|-----------------------------------------------------|
438| `/cpu/self/blocked`     | Blocked reference implementation                    |
439| `/cpu/self/ref`         | Serial reference implementation                     |
440| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
441| `/cpu/occa`             | Serial OCCA kernels                                 |
442| `/gpu/occa`             | CUDA OCCA kernels                                   |
443| `/omp/occa`             | OpenMP OCCA kernels                                 |
444| `/ocl/occa`             | OpenCL OCCA kernels                                 |
445| `/gpu/magma`            | CUDA MAGMA kernels                                  |
446
447Examples available in this release:
448
449:::{list-table}
450:header-rows: 1
451:widths: auto
452* - User code
453  - Example
454* - `ceed`
455  - * ex1 (volume)
456* - `mfem`
457  - * BP1 (scalar mass operator)
458    * BP3 (scalar Laplace operator)
459* - `petsc`
460  - * BP1 (scalar mass operator)
461    * BP3 (scalar Laplace operator)
462* - `nek5000`
463  - * BP1 (scalar mass operator)
464    * BP3 (scalar Laplace operator)
465:::
466
467(v0-21)=
468
469## v0.21 (Sep 30, 2018)
470
471A MAGMA backend (which relies upon the
472[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
473release. This initial integration set up the framework of using MAGMA and provided the
474libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
475As any other backend, the MAGMA backend provides extended basic data structures for
476{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
477the fundamental CEED building blocks to work with the new data structures.
478In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
479but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
480internally, and thus seamlessly to the user, through the functions/methods that are
481provided to support them.
482
483Backends available in this release:
484
485| CEED resource (`-ceed`) | Backend                         |
486|-------------------------|---------------------------------|
487| `/cpu/self`             | Serial reference implementation |
488| `/cpu/occa`             | Serial OCCA kernels             |
489| `/gpu/occa`             | CUDA OCCA kernels               |
490| `/omp/occa`             | OpenMP OCCA kernels             |
491| `/ocl/occa`             | OpenCL OCCA kernels             |
492| `/gpu/magma`            | CUDA MAGMA kernels              |
493
494Examples available in this release:
495
496:::{list-table}
497:header-rows: 1
498:widths: auto
499* - User code
500  - Example
501* - `ceed`
502  - * ex1 (volume)
503* - `mfem`
504  - * BP1 (scalar mass operator)
505    * BP3 (scalar Laplace operator)
506* - `petsc`
507  - * BP1 (scalar mass operator)
508* - `nek5000`
509  - * BP1 (scalar mass operator)
510:::
511
512(v0-2)=
513
514## v0.2 (Mar 30, 2018)
515
516libCEED was made publicly available the first full CEED software distribution, release
517CEED 1.0. The distribution was made available using the Spack package manager to provide
518a common, easy-to-use build environment, where the user can build the CEED distribution
519with all dependencies. This release included a new Fortran interface for the library.
520This release also contained major improvements in the OCCA backend (including a new
521`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
522compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
523`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
524of the Laplace operator) was also added to this release.
525
526Backends available in this release:
527
528| CEED resource (`-ceed`) | Backend                         |
529|-------------------------|---------------------------------|
530| `/cpu/self`             | Serial reference implementation |
531| `/cpu/occa`             | Serial OCCA kernels             |
532| `/gpu/occa`             | CUDA OCCA kernels               |
533| `/omp/occa`             | OpenMP OCCA kernels             |
534| `/ocl/occa`             | OpenCL OCCA kernels             |
535
536Examples available in this release:
537
538:::{list-table}
539:header-rows: 1
540:widths: auto
541* - User code
542  - Example
543* - `ceed`
544  - * ex1 (volume)
545* - `mfem`
546  - * BP1 (scalar mass operator)
547    * BP3 (scalar Laplace operator)
548* - `petsc`
549  - * BP1 (scalar mass operator)
550* - `nek5000`
551  - * BP1 (scalar mass operator)
552:::
553
554(v0-1)=
555
556## v0.1 (Jan 3, 2018)
557
558Initial low-level API of the CEED project. The low-level API provides a set of Finite
559Elements kernels and components for writing new low-level kernels. Examples include:
560vector and sparse linear algebra, element matrix assembly over a batch of elements,
561partial assembly and action for efficient high-order operators like mass, diffusion,
562advection, etc. The main goal of the low-level API is to establish the basis for the
563high-level API. Also, identifying such low-level kernels and providing a reference
564implementation for them serves as the basis for specialized backend implementations.
565This release contained several backends: `/cpu/self`, and backends which rely upon the
566[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
567`/gpu/occa`, and `/omp/occa`.
568It also included several examples, in the `examples` folder:
569A standalone code that shows the usage of libCEED (with no external
570dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
571(with the application of the mass operator); and a `petsc` example to perform BP1
572(with the application of the mass operator).
573
574Backends available in this release:
575
576| CEED resource (`-ceed`) | Backend                         |
577|-------------------------|---------------------------------|
578| `/cpu/self`             | Serial reference implementation |
579| `/cpu/occa`             | Serial OCCA kernels             |
580| `/gpu/occa`             | CUDA OCCA kernels               |
581| `/omp/occa`             | OpenMP OCCA kernels             |
582
583Examples available in this release:
584
585| User code             | Example                           |
586|-----------------------|-----------------------------------|
587| `ceed`                | ex1 (scalar Laplace operator)     |
588| `mfem`                | BP1 (scalar mass operator)        |
589| `petsc`               | BP1 (scalar mass operator)        |
590```
591