xref: /libCEED/doc/sphinx/source/releasenotes.md (revision a49029fe2b683c1941ca0e371db6091ee3c727b1)
1# Changes/Release Notes
2
3On this page we provide a summary of the main API changes, new features and examples
4for each release of libCEED.
5
6(main)=
7
8## Current `main` branch
9
10### Interface changes
11
12- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
13
14(v0-10)=
15
16## v0.10 (Mar 21, 2022)
17
18### Interface changes
19
20- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
21- 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`.
22- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
23- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
24- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
25- Warning added when compiling OCCA backend to alert users that this backend is experimental.
26- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
27- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
28- 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`.
29- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
30- 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.
31- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
32- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
33- 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.
34- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
35- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
36
37### New features
38
39- `CeedScalar` can now be set as `float` or `double` at compile time.
40- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
41- 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.
42- Remove need to guard library headers in QFunction source for code generation backends.
43- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
44- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
45- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
46- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
47- 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.
48- Added support for element matrix assembly in GPU backends.
49
50### Maintainability
51
52- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
53- `Include-what-you-use` makefile target added as `make iwyu`.
54- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
55- Put GPU JiTed kernel source code into separate files.
56- 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.
57
58(v0-9)=
59
60## v0.9 (Jul 6, 2021)
61
62### Interface changes
63
64- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
65
66### New features
67
68- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
69- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
70- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
71- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
72
73### Performance improvements
74
75### Examples
76
77- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
78- Fluid mechanics example adds GPU support and improves modularity.
79
80### Deprecated backends
81
82- 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.
83
84(v0-8)=
85
86## v0.8 (Mar 31, 2021)
87
88### Interface changes
89
90- Error handling improved to include enumerated error codes for C interface return values.
91- 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.
92
93### New features
94
95- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
96- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
97- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
98
99### Performance improvements
100
101- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
102- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
103
104### Examples
105
106- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
107- {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.
108- {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.
109- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
110
111(v0-7)=
112
113## v0.7 (Sep 29, 2020)
114
115### Interface changes
116
117- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
118  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
119  These changes improve support for mixed finite element methods.
120- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
121- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
122- 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.
123- 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.
124- 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.
125- 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.
126  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`.
127- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
128- 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.
129
130### New features
131
132- New HIP backend: `/gpu/hip/ref`.
133- CeedQFunction support for user `CUfunction`s in some backends
134
135### Performance improvements
136
137- OCCA backend rebuilt to facilitate future performance enhancements.
138- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
139
140### Examples
141
142- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
143
144### Deprecated backends
145
146- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
147
148(v0-6)=
149
150## v0.6 (Mar 29, 2020)
151
152libCEED v0.6 contains numerous new features and examples, as well as expanded
153documentation in [this new website](https://libceed.org).
154
155### New features
156
157- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
158  1-1 correspondence with the C interface, plus some convenience features.  For instance,
159  data stored in the {cpp:type}`CeedVector` structure are available without copy as
160  {py:class}`numpy.ndarray`.  Short tutorials are provided in
161  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
162- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
163  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
164  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
165  preconditioning ingredients and are used in the libCEED's multigrid examples.
166- The inverse of separable operators can be obtained using
167  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
168  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
169  especially for Laplacians and related operators.
170- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
171  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
172- Make public accessors for various attributes to facilitate writing composable code.
173- New backend: `/cpu/self/memcheck/serial`.
174- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
175  backends.  (Single source is coming soon for OCCA backends.)
176- Fix some missing edge cases in CUDA backend.
177
178### Performance Improvements
179
180- MAGMA backend performance optimization and non-tensor bases.
181- No-copy optimization in {c:func}`CeedOperatorApply`.
182
183### Interface changes
184
185- Replace {code}`CeedElemRestrictionCreateIdentity` and
186  {code}`CeedElemRestrictionCreateBlocked` with more flexible
187  {c:func}`CeedElemRestrictionCreateStrided` and
188  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
189- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
190- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
191  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
192  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
193  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
194  {c:func}`CeedElemRestrictionApply`.
195
196### Examples
197
198libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
199Notable additions include:
200
201- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
202  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
203
204- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
205  of domains (like the cube and sphere) by direct integration on a surface mesh;
206  demonstrates geometric dimension different from topological dimension.
207
208- PETSc {ref}`example-petsc-bps`:
209
210  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
211  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
212    and transparent CUDA support.
213  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
214    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
215    dimension different from topological dimension.
216
217- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
218  solver with algebraic multigrid coarse solve.
219
220- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
221  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
222  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
223  quasi-2D computational domain support.
224
225- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
226  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
227  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
228
229(v0-5)=
230
231## v0.5 (Sep 18, 2019)
232
233For this release, several improvements were made. Two new CUDA backends were added to
234the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
235performance using single-source {ref}`CeedQFunction`. From this release, users
236can define Q-Functions in a single source code independently of the targeted backend
237with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
238compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
239declaration, the {ref}`CeedQFunction` API has undergone a slight change:
240the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
241requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
242`QFunctionField` with eval mode `CEED EVAL GRAD`.
243
244Additionally, new CPU backends
245were included in this release, such as the `/cpu/self/opt/*` backends (which are
246written in pure C and use partial **E-vectors** to improve performance) and the
247`/cpu/self/ref/memcheck` backend (which relies upon the
248[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
249{ref}`CeedQFunction` have no undefined values).
250This release also included various performance improvements, bug fixes, new examples,
251and improved tests. Among these improvements, vectorized instructions for
252{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
253instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
254identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
255to include unstructured meshes handling were. For this expansion, the prior version of
256the PETSc BPs, which only included data associated with structured geometries, were
257renamed `bpsraw`, and the new version of the BPs, which can handle data associated
258with any unstructured geometry, were called `bps`. Additionally, other benchmark
259problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
260and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
261as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
262examples. Furthermoew, another standalone libCEED example, called `ex2`, which
263computes the surface area of a given mesh was added to this release.
264
265Backends available in this release:
266
267| CEED resource (`-ceed`)  | Backend                                             |
268|--------------------------|-----------------------------------------------------|
269| `/cpu/self/ref/serial`   | Serial reference implementation                     |
270| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
271| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
272| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
273| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
274| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
275| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
276| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
277| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
278| `/cpu/occa`              | Serial OCCA kernels                                 |
279| `/gpu/occa`              | CUDA OCCA kernels                                   |
280| `/omp/occa`              | OpenMP OCCA kernels                                 |
281| `/ocl/occa`              | OpenCL OCCA kernels                                 |
282| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
283| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
284| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
285| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
286| `/gpu/magma`             | CUDA MAGMA kernels                                  |
287
288Examples available in this release:
289
290:::{list-table}
291:header-rows: 1
292:widths: auto
293* - User code
294  - Example
295* - `ceed`
296  - * ex1 (volume)
297    * ex2 (surface)
298* - `mfem`
299  - * BP1 (scalar mass operator)
300    * BP3 (scalar Laplace operator)
301* - `petsc`
302  - * BP1 (scalar mass operator)
303    * BP2 (vector mass operator)
304    * BP3 (scalar Laplace operator)
305    * BP4 (vector Laplace operator)
306    * BP5 (collocated scalar Laplace operator)
307    * BP6 (collocated vector Laplace operator)
308    * Navier-Stokes
309* - `nek5000`
310  - * BP1 (scalar mass operator)
311    * BP3 (scalar Laplace operator)
312:::
313
314(v0-4)=
315
316## v0.4 (Apr 1, 2019)
317
318libCEED v0.4 was made again publicly available in the second full CEED software
319distribution, release CEED 2.0. This release contained notable features, such as
320four new CPU backends, two new GPU backends, CPU backend optimizations, initial
321support for operator composition, performance benchmarking, and a Navier-Stokes demo.
322The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
323backends process one element at a time and are intended for meshes with a smaller number
324of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
325eight interlaced elements and are intended for meshes with higher numbers of elements.
326The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
327performance. The `/cpu/self/xsmm/*` backends rely upon the
328[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
329performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
330The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
331performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
332parallelization approach, where each thread treats a finite element. Using just in time
333compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
334backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
335achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
336deteriorates very quickly when threads run out of registers.
337
338A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
339examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
340This example solves the time-dependent Navier-Stokes equations of compressible gas
341dynamics in a static Eulerian three-dimensional frame, using structured high-order
342finite/spectral element spatial discretizations and explicit high-order time-stepping
343(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
344so that the pointwise physics (defined at quadrature points) is separated from the
345parallelization and meshing concerns.
346
347Backends available in this release:
348
349| CEED resource (`-ceed`)  | Backend                                             |
350|--------------------------|-----------------------------------------------------|
351| `/cpu/self/ref/serial`   | Serial reference implementation                     |
352| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
353| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
354| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
355| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
356| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
357| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
358| `/cpu/occa`              | Serial OCCA kernels                                 |
359| `/gpu/occa`              | CUDA OCCA kernels                                   |
360| `/omp/occa`              | OpenMP OCCA kernels                                 |
361| `/ocl/occa`              | OpenCL OCCA kernels                                 |
362| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
363| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
364| `/gpu/magma`             | CUDA MAGMA kernels                                  |
365
366Examples available in this release:
367
368:::{list-table}
369:header-rows: 1
370:widths: auto
371* - User code
372  - Example
373* - `ceed`
374  - * ex1 (volume)
375* - `mfem`
376  - * BP1 (scalar mass operator)
377    * BP3 (scalar Laplace operator)
378* - `petsc`
379  - * BP1 (scalar mass operator)
380    * BP3 (scalar Laplace operator)
381    * Navier-Stokes
382* - `nek5000`
383  - * BP1 (scalar mass operator)
384    * BP3 (scalar Laplace operator)
385:::
386
387(v0-3)=
388
389## v0.3 (Sep 30, 2018)
390
391Notable features in this release include active/passive field interface, support for
392non-tensor bases, backend optimization, and improved Fortran interface. This release
393also focused on providing improved continuous integration, and many new tests with code
394coverage reports of about 90%. This release also provided a significant change to the
395public interface: a {ref}`CeedQFunction` can take any number of named input and output
396arguments while {ref}`CeedOperator` connects them to the actual data, which may be
397supplied explicitly to `CeedOperatorApply()` (active) or separately via
398`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
399of CeedQFunctions and composition of block solvers constructed using
400{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
401used in an optimized CPU backend. Although this is typically not visible to the user,
402it enables effective use of arbitrary-length SIMD while maintaining cache locality.
403This CPU backend also implements an algebraic factorization of tensor product gradients
404to perform fewer operations than standard application of interpolation and
405differentiation from nodes to quadrature points. This algebraic formulation
406automatically supports non-polynomial and non-interpolatory bases, thus is more general
407than the more common derivation in terms of Lagrange polynomials on the quadrature points.
408
409Backends available in this release:
410
411| CEED resource (`-ceed`) | Backend                                             |
412|-------------------------|-----------------------------------------------------|
413| `/cpu/self/blocked`     | Blocked reference implementation                    |
414| `/cpu/self/ref`         | Serial reference implementation                     |
415| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
416| `/cpu/occa`             | Serial OCCA kernels                                 |
417| `/gpu/occa`             | CUDA OCCA kernels                                   |
418| `/omp/occa`             | OpenMP OCCA kernels                                 |
419| `/ocl/occa`             | OpenCL OCCA kernels                                 |
420| `/gpu/magma`            | CUDA MAGMA kernels                                  |
421
422Examples available in this release:
423
424:::{list-table}
425:header-rows: 1
426:widths: auto
427* - User code
428  - Example
429* - `ceed`
430  - * ex1 (volume)
431* - `mfem`
432  - * BP1 (scalar mass operator)
433    * BP3 (scalar Laplace operator)
434* - `petsc`
435  - * BP1 (scalar mass operator)
436    * BP3 (scalar Laplace operator)
437* - `nek5000`
438  - * BP1 (scalar mass operator)
439    * BP3 (scalar Laplace operator)
440:::
441
442(v0-21)=
443
444## v0.21 (Sep 30, 2018)
445
446A MAGMA backend (which relies upon the
447[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
448release. This initial integration set up the framework of using MAGMA and provided the
449libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
450As any other backend, the MAGMA backend provides extended basic data structures for
451{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
452the fundamental CEED building blocks to work with the new data structures.
453In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
454but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
455internally, and thus seamlessly to the user, through the functions/methods that are
456provided to support them.
457
458Backends available in this release:
459
460| CEED resource (`-ceed`) | Backend                         |
461|-------------------------|---------------------------------|
462| `/cpu/self`             | Serial reference implementation |
463| `/cpu/occa`             | Serial OCCA kernels             |
464| `/gpu/occa`             | CUDA OCCA kernels               |
465| `/omp/occa`             | OpenMP OCCA kernels             |
466| `/ocl/occa`             | OpenCL OCCA kernels             |
467| `/gpu/magma`            | CUDA MAGMA kernels              |
468
469Examples available in this release:
470
471:::{list-table}
472:header-rows: 1
473:widths: auto
474* - User code
475  - Example
476* - `ceed`
477  - * ex1 (volume)
478* - `mfem`
479  - * BP1 (scalar mass operator)
480    * BP3 (scalar Laplace operator)
481* - `petsc`
482  - * BP1 (scalar mass operator)
483* - `nek5000`
484  - * BP1 (scalar mass operator)
485:::
486
487(v0-2)=
488
489## v0.2 (Mar 30, 2018)
490
491libCEED was made publicly available the first full CEED software distribution, release
492CEED 1.0. The distribution was made available using the Spack package manager to provide
493a common, easy-to-use build environment, where the user can build the CEED distribution
494with all dependencies. This release included a new Fortran interface for the library.
495This release also contained major improvements in the OCCA backend (including a new
496`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
497compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
498`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
499of the Laplace operator) was also added to this release.
500
501Backends available in this release:
502
503| CEED resource (`-ceed`) | Backend                         |
504|-------------------------|---------------------------------|
505| `/cpu/self`             | Serial reference implementation |
506| `/cpu/occa`             | Serial OCCA kernels             |
507| `/gpu/occa`             | CUDA OCCA kernels               |
508| `/omp/occa`             | OpenMP OCCA kernels             |
509| `/ocl/occa`             | OpenCL OCCA kernels             |
510
511Examples available in this release:
512
513:::{list-table}
514:header-rows: 1
515:widths: auto
516* - User code
517  - Example
518* - `ceed`
519  - * ex1 (volume)
520* - `mfem`
521  - * BP1 (scalar mass operator)
522    * BP3 (scalar Laplace operator)
523* - `petsc`
524  - * BP1 (scalar mass operator)
525* - `nek5000`
526  - * BP1 (scalar mass operator)
527:::
528
529(v0-1)=
530
531## v0.1 (Jan 3, 2018)
532
533Initial low-level API of the CEED project. The low-level API provides a set of Finite
534Elements kernels and components for writing new low-level kernels. Examples include:
535vector and sparse linear algebra, element matrix assembly over a batch of elements,
536partial assembly and action for efficient high-order operators like mass, diffusion,
537advection, etc. The main goal of the low-level API is to establish the basis for the
538high-level API. Also, identifying such low-level kernels and providing a reference
539implementation for them serves as the basis for specialized backend implementations.
540This release contained several backends: `/cpu/self`, and backends which rely upon the
541[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
542`/gpu/occa`, and `/omp/occa`.
543It also included several examples, in the `examples` folder:
544A standalone code that shows the usage of libCEED (with no external
545dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
546(with the application of the mass operator); and a `petsc` example to perform BP1
547(with the application of the mass operator).
548
549Backends available in this release:
550
551| CEED resource (`-ceed`) | Backend                         |
552|-------------------------|---------------------------------|
553| `/cpu/self`             | Serial reference implementation |
554| `/cpu/occa`             | Serial OCCA kernels             |
555| `/gpu/occa`             | CUDA OCCA kernels               |
556| `/omp/occa`             | OpenMP OCCA kernels             |
557
558Examples available in this release:
559
560| User code             | Example                           |
561|-----------------------|-----------------------------------|
562| `ceed`                | ex1 (scalar Laplace operator)     |
563| `mfem`                | BP1 (scalar mass operator)        |
564| `petsc`               | BP1 (scalar mass operator)        |
565```
566