xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 667e613fe678313d77f7966d97cc228a73b32933)
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(v0-10)=
13
14## v0.10 (Mar 16, 2022)
15
16### Interface changes
17
18- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
19- 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`.
20- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
21- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
22- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
23- Warning added when compiling OCCA backend to alert users that this backend is experimental.
24- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
25- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
26- 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`.
27- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
28- 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.
29- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
30- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
31- 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.
32- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
33- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
34
35### New features
36
37- `CeedScalar` can now be set as `float` or `double` at compile time.
38- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
39- 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.
40- Remove need to guard library headers in QFunction source for code generation backends.
41- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
42- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
43- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
44- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
45- 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.
46- Added support for element matrix assembly in GPU backends.
47
48### Maintainability
49
50- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
51- `Include-what-you-use` makefile target added as `make iwyu`.
52- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
53- Put GPU JiTed kernel source code into separate files.
54- 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.
55
56(v0-9)=
57
58## v0.9 (Jul 6, 2021)
59
60### Interface changes
61
62- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
63
64### New features
65
66- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
67- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
68- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
69- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
70
71### Performance improvements
72
73### Examples
74
75- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
76- Fluid mechanics example adds GPU support and improves modularity.
77
78### Deprecated backends
79
80- 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.
81
82(v0-8)=
83
84## v0.8 (Mar 31, 2021)
85
86### Interface changes
87
88- Error handling improved to include enumerated error codes for C interface return values.
89- 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.
90
91### New features
92
93- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
94- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
95- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
96
97### Performance improvements
98
99- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
100- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
101
102### Examples
103
104- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
105- {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.
106- {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.
107- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
108
109(v0-7)=
110
111## v0.7 (Sep 29, 2020)
112
113### Interface changes
114
115- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
116  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
117  These changes improve support for mixed finite element methods.
118- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
119- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
120- 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.
121- 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.
122- 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.
123- 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.
124  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`.
125- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
126- 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.
127
128### New features
129
130- New HIP backend: `/gpu/hip/ref`.
131- CeedQFunction support for user `CUfunction`s in some backends
132
133### Performance improvements
134
135- OCCA backend rebuilt to facilitate future performance enhancements.
136- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
137
138### Examples
139
140- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
141
142### Deprecated backends
143
144- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
145
146(v0-6)=
147
148## v0.6 (Mar 29, 2020)
149
150libCEED v0.6 contains numerous new features and examples, as well as expanded
151documentation in [this new website](https://libceed.org).
152
153### New features
154
155- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
156  1-1 correspondence with the C interface, plus some convenience features.  For instance,
157  data stored in the {cpp:type}`CeedVector` structure are available without copy as
158  {py:class}`numpy.ndarray`.  Short tutorials are provided in
159  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
160- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
161  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
162  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
163  preconditioning ingredients and are used in the libCEED's multigrid examples.
164- The inverse of separable operators can be obtained using
165  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
166  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
167  especially for Laplacians and related operators.
168- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
169  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
170- Make public accessors for various attributes to facilitate writing composable code.
171- New backend: `/cpu/self/memcheck/serial`.
172- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
173  backends.  (Single source is coming soon for OCCA backends.)
174- Fix some missing edge cases in CUDA backend.
175
176### Performance Improvements
177
178- MAGMA backend performance optimization and non-tensor bases.
179- No-copy optimization in {c:func}`CeedOperatorApply`.
180
181### Interface changes
182
183- Replace {code}`CeedElemRestrictionCreateIdentity` and
184  {code}`CeedElemRestrictionCreateBlocked` with more flexible
185  {c:func}`CeedElemRestrictionCreateStrided` and
186  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
187- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
188- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
189  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
190  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
191  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
192  {c:func}`CeedElemRestrictionApply`.
193
194### Examples
195
196libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
197Notable additions include:
198
199- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
200  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
201
202- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
203  of domains (like the cube and sphere) by direct integration on a surface mesh;
204  demonstrates geometric dimension different from topological dimension.
205
206- PETSc {ref}`example-petsc-bps`:
207
208  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
209  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
210    and transparent CUDA support.
211  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
212    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
213    dimension different from topological dimension.
214
215- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
216  solver with algebraic multigrid coarse solve.
217
218- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
219  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
220  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
221  quasi-2D computational domain support.
222
223- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
224  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
225  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
226
227(v0-5)=
228
229## v0.5 (Sep 18, 2019)
230
231For this release, several improvements were made. Two new CUDA backends were added to
232the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
233performance using single-source {ref}`CeedQFunction`. From this release, users
234can define Q-Functions in a single source code independently of the targeted backend
235with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
236compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
237declaration, the {ref}`CeedQFunction` API has undergone a slight change:
238the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
239requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
240`QFunctionField` with eval mode `CEED EVAL GRAD`.
241
242Additionally, new CPU backends
243were included in this release, such as the `/cpu/self/opt/*` backends (which are
244written in pure C and use partial **E-vectors** to improve performance) and the
245`/cpu/self/ref/memcheck` backend (which relies upon the
246[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
247{ref}`CeedQFunction` have no undefined values).
248This release also included various performance improvements, bug fixes, new examples,
249and improved tests. Among these improvements, vectorized instructions for
250{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
251instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
252identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
253to include unstructured meshes handling were. For this expansion, the prior version of
254the PETSc BPs, which only included data associated with structured geometries, were
255renamed `bpsraw`, and the new version of the BPs, which can handle data associated
256with any unstructured geometry, were called `bps`. Additionally, other benchmark
257problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
258and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
259as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
260examples. Furthermoew, another standalone libCEED example, called `ex2`, which
261computes the surface area of a given mesh was added to this release.
262
263Backends available in this release:
264
265| CEED resource (`-ceed`)  | Backend                                             |
266|--------------------------|-----------------------------------------------------|
267| `/cpu/self/ref/serial`   | Serial reference implementation                     |
268| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
269| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
270| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
271| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
272| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
273| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
274| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
275| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
276| `/cpu/occa`              | Serial OCCA kernels                                 |
277| `/gpu/occa`              | CUDA OCCA kernels                                   |
278| `/omp/occa`              | OpenMP OCCA kernels                                 |
279| `/ocl/occa`              | OpenCL OCCA kernels                                 |
280| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
281| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
282| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
283| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
284| `/gpu/magma`             | CUDA MAGMA kernels                                  |
285
286Examples available in this release:
287
288:::{list-table}
289:header-rows: 1
290:widths: auto
291* - User code
292  - Example
293* - `ceed`
294  - * ex1 (volume)
295    * ex2 (surface)
296* - `mfem`
297  - * BP1 (scalar mass operator)
298    * BP3 (scalar Laplace operator)
299* - `petsc`
300  - * BP1 (scalar mass operator)
301    * BP2 (vector mass operator)
302    * BP3 (scalar Laplace operator)
303    * BP4 (vector Laplace operator)
304    * BP5 (collocated scalar Laplace operator)
305    * BP6 (collocated vector Laplace operator)
306    * Navier-Stokes
307* - `nek5000`
308  - * BP1 (scalar mass operator)
309    * BP3 (scalar Laplace operator)
310:::
311
312(v0-4)=
313
314## v0.4 (Apr 1, 2019)
315
316libCEED v0.4 was made again publicly available in the second full CEED software
317distribution, release CEED 2.0. This release contained notable features, such as
318four new CPU backends, two new GPU backends, CPU backend optimizations, initial
319support for operator composition, performance benchmarking, and a Navier-Stokes demo.
320The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
321backends process one element at a time and are intended for meshes with a smaller number
322of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
323eight interlaced elements and are intended for meshes with higher numbers of elements.
324The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
325performance. The `/cpu/self/xsmm/*` backends rely upon the
326[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
327performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
328The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
329performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
330parallelization approach, where each thread treats a finite element. Using just in time
331compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
332backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
333achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
334deteriorates very quickly when threads run out of registers.
335
336A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
337examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
338This example solves the time-dependent Navier-Stokes equations of compressible gas
339dynamics in a static Eulerian three-dimensional frame, using structured high-order
340finite/spectral element spatial discretizations and explicit high-order time-stepping
341(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
342so that the pointwise physics (defined at quadrature points) is separated from the
343parallelization and meshing concerns.
344
345Backends available in this release:
346
347| CEED resource (`-ceed`)  | Backend                                             |
348|--------------------------|-----------------------------------------------------|
349| `/cpu/self/ref/serial`   | Serial reference implementation                     |
350| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
351| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
352| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
353| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
354| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
355| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
356| `/cpu/occa`              | Serial OCCA kernels                                 |
357| `/gpu/occa`              | CUDA OCCA kernels                                   |
358| `/omp/occa`              | OpenMP OCCA kernels                                 |
359| `/ocl/occa`              | OpenCL OCCA kernels                                 |
360| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
361| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
362| `/gpu/magma`             | CUDA MAGMA kernels                                  |
363
364Examples available in this release:
365
366:::{list-table}
367:header-rows: 1
368:widths: auto
369* - User code
370  - Example
371* - `ceed`
372  - * ex1 (volume)
373* - `mfem`
374  - * BP1 (scalar mass operator)
375    * BP3 (scalar Laplace operator)
376* - `petsc`
377  - * BP1 (scalar mass operator)
378    * BP3 (scalar Laplace operator)
379    * Navier-Stokes
380* - `nek5000`
381  - * BP1 (scalar mass operator)
382    * BP3 (scalar Laplace operator)
383:::
384
385(v0-3)=
386
387## v0.3 (Sep 30, 2018)
388
389Notable features in this release include active/passive field interface, support for
390non-tensor bases, backend optimization, and improved Fortran interface. This release
391also focused on providing improved continuous integration, and many new tests with code
392coverage reports of about 90%. This release also provided a significant change to the
393public interface: a {ref}`CeedQFunction` can take any number of named input and output
394arguments while {ref}`CeedOperator` connects them to the actual data, which may be
395supplied explicitly to `CeedOperatorApply()` (active) or separately via
396`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
397of CeedQFunctions and composition of block solvers constructed using
398{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
399used in an optimized CPU backend. Although this is typically not visible to the user,
400it enables effective use of arbitrary-length SIMD while maintaining cache locality.
401This CPU backend also implements an algebraic factorization of tensor product gradients
402to perform fewer operations than standard application of interpolation and
403differentiation from nodes to quadrature points. This algebraic formulation
404automatically supports non-polynomial and non-interpolatory bases, thus is more general
405than the more common derivation in terms of Lagrange polynomials on the quadrature points.
406
407Backends available in this release:
408
409| CEED resource (`-ceed`) | Backend                                             |
410|-------------------------|-----------------------------------------------------|
411| `/cpu/self/blocked`     | Blocked reference implementation                    |
412| `/cpu/self/ref`         | Serial reference implementation                     |
413| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
414| `/cpu/occa`             | Serial OCCA kernels                                 |
415| `/gpu/occa`             | CUDA OCCA kernels                                   |
416| `/omp/occa`             | OpenMP OCCA kernels                                 |
417| `/ocl/occa`             | OpenCL OCCA kernels                                 |
418| `/gpu/magma`            | CUDA MAGMA kernels                                  |
419
420Examples available in this release:
421
422:::{list-table}
423:header-rows: 1
424:widths: auto
425* - User code
426  - Example
427* - `ceed`
428  - * ex1 (volume)
429* - `mfem`
430  - * BP1 (scalar mass operator)
431    * BP3 (scalar Laplace operator)
432* - `petsc`
433  - * BP1 (scalar mass operator)
434    * BP3 (scalar Laplace operator)
435* - `nek5000`
436  - * BP1 (scalar mass operator)
437    * BP3 (scalar Laplace operator)
438:::
439
440(v0-21)=
441
442## v0.21 (Sep 30, 2018)
443
444A MAGMA backend (which relies upon the
445[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
446release. This initial integration set up the framework of using MAGMA and provided the
447libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
448As any other backend, the MAGMA backend provides extended basic data structures for
449{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
450the fundamental CEED building blocks to work with the new data structures.
451In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
452but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
453internally, and thus seamlessly to the user, through the functions/methods that are
454provided to support them.
455
456Backends available in this release:
457
458| CEED resource (`-ceed`) | Backend                         |
459|-------------------------|---------------------------------|
460| `/cpu/self`             | Serial reference implementation |
461| `/cpu/occa`             | Serial OCCA kernels             |
462| `/gpu/occa`             | CUDA OCCA kernels               |
463| `/omp/occa`             | OpenMP OCCA kernels             |
464| `/ocl/occa`             | OpenCL OCCA kernels             |
465| `/gpu/magma`            | CUDA MAGMA kernels              |
466
467Examples available in this release:
468
469:::{list-table}
470:header-rows: 1
471:widths: auto
472* - User code
473  - Example
474* - `ceed`
475  - * ex1 (volume)
476* - `mfem`
477  - * BP1 (scalar mass operator)
478    * BP3 (scalar Laplace operator)
479* - `petsc`
480  - * BP1 (scalar mass operator)
481* - `nek5000`
482  - * BP1 (scalar mass operator)
483:::
484
485(v0-2)=
486
487## v0.2 (Mar 30, 2018)
488
489libCEED was made publicly available the first full CEED software distribution, release
490CEED 1.0. The distribution was made available using the Spack package manager to provide
491a common, easy-to-use build environment, where the user can build the CEED distribution
492with all dependencies. This release included a new Fortran interface for the library.
493This release also contained major improvements in the OCCA backend (including a new
494`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
495compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
496`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
497of the Laplace operator) was also added to this release.
498
499Backends available in this release:
500
501| CEED resource (`-ceed`) | Backend                         |
502|-------------------------|---------------------------------|
503| `/cpu/self`             | Serial reference implementation |
504| `/cpu/occa`             | Serial OCCA kernels             |
505| `/gpu/occa`             | CUDA OCCA kernels               |
506| `/omp/occa`             | OpenMP OCCA kernels             |
507| `/ocl/occa`             | OpenCL OCCA kernels             |
508
509Examples available in this release:
510
511:::{list-table}
512:header-rows: 1
513:widths: auto
514* - User code
515  - Example
516* - `ceed`
517  - * ex1 (volume)
518* - `mfem`
519  - * BP1 (scalar mass operator)
520    * BP3 (scalar Laplace operator)
521* - `petsc`
522  - * BP1 (scalar mass operator)
523* - `nek5000`
524  - * BP1 (scalar mass operator)
525:::
526
527(v0-1)=
528
529## v0.1 (Jan 3, 2018)
530
531Initial low-level API of the CEED project. The low-level API provides a set of Finite
532Elements kernels and components for writing new low-level kernels. Examples include:
533vector and sparse linear algebra, element matrix assembly over a batch of elements,
534partial assembly and action for efficient high-order operators like mass, diffusion,
535advection, etc. The main goal of the low-level API is to establish the basis for the
536high-level API. Also, identifying such low-level kernels and providing a reference
537implementation for them serves as the basis for specialized backend implementations.
538This release contained several backends: `/cpu/self`, and backends which rely upon the
539[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
540`/gpu/occa`, and `/omp/occa`.
541It also included several examples, in the `examples` folder:
542A standalone code that shows the usage of libCEED (with no external
543dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
544(with the application of the mass operator); and a `petsc` example to perform BP1
545(with the application of the mass operator).
546
547Backends available in this release:
548
549| CEED resource (`-ceed`) | Backend                         |
550|-------------------------|---------------------------------|
551| `/cpu/self`             | Serial reference implementation |
552| `/cpu/occa`             | Serial OCCA kernels             |
553| `/gpu/occa`             | CUDA OCCA kernels               |
554| `/omp/occa`             | OpenMP OCCA kernels             |
555
556Examples available in this release:
557
558| User code             | Example                           |
559|-----------------------|-----------------------------------|
560| `ceed`                | ex1 (scalar Laplace operator)     |
561| `mfem`                | BP1 (scalar mass operator)        |
562| `petsc`               | BP1 (scalar mass operator)        |
563```
564