xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 7b39487de4ab72ae5651458a5c845e3d06436385)
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
22### Examples
23
24- Added various performance enhancements for {ref}`example-petsc-navier-stokes`
25
26(v0-10-1)=
27
28## v0.10.1 (Apr 11, 2022)
29
30### Interface changes
31
32- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
33
34### Bugfix
35
36- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
37
38(v0-10)=
39
40## v0.10 (Mar 21, 2022)
41
42### Interface changes
43
44- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
45- 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`.
46- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
47- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
48- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
49- Warning added when compiling OCCA backend to alert users that this backend is experimental.
50- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
51- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
52- 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`.
53- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
54- 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.
55- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
56- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
57- 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.
58- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
59- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
60
61### New features
62
63- `CeedScalar` can now be set as `float` or `double` at compile time.
64- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
65- 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.
66- Remove need to guard library headers in QFunction source for code generation backends.
67- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
68- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
69- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
70- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
71- 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.
72- Added support for element matrix assembly in GPU backends.
73
74### Maintainability
75
76- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
77- `Include-what-you-use` makefile target added as `make iwyu`.
78- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
79- Put GPU JiTed kernel source code into separate files.
80- 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.
81
82(v0-9)=
83
84## v0.9 (Jul 6, 2021)
85
86### Interface changes
87
88- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
89
90### New features
91
92- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
93- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
94- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
95- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
96
97### Performance improvements
98
99### Examples
100
101- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
102- Fluid mechanics example adds GPU support and improves modularity.
103
104### Deprecated backends
105
106- 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.
107
108(v0-8)=
109
110## v0.8 (Mar 31, 2021)
111
112### Interface changes
113
114- Error handling improved to include enumerated error codes for C interface return values.
115- 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.
116
117### New features
118
119- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
120- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
121- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
122
123### Performance improvements
124
125- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
126- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
127
128### Examples
129
130- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
131- {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.
132- {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.
133- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
134
135(v0-7)=
136
137## v0.7 (Sep 29, 2020)
138
139### Interface changes
140
141- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
142  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
143  These changes improve support for mixed finite element methods.
144- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
145- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
146- 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.
147- 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.
148- 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.
149- 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.
150  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`.
151- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
152- 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.
153
154### New features
155
156- New HIP backend: `/gpu/hip/ref`.
157- CeedQFunction support for user `CUfunction`s in some backends
158
159### Performance improvements
160
161- OCCA backend rebuilt to facilitate future performance enhancements.
162- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
163
164### Examples
165
166- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
167
168### Deprecated backends
169
170- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
171
172(v0-6)=
173
174## v0.6 (Mar 29, 2020)
175
176libCEED v0.6 contains numerous new features and examples, as well as expanded
177documentation in [this new website](https://libceed.org).
178
179### New features
180
181- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
182  1-1 correspondence with the C interface, plus some convenience features.  For instance,
183  data stored in the {cpp:type}`CeedVector` structure are available without copy as
184  {py:class}`numpy.ndarray`.  Short tutorials are provided in
185  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
186- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
187  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
188  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
189  preconditioning ingredients and are used in the libCEED's multigrid examples.
190- The inverse of separable operators can be obtained using
191  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
192  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
193  especially for Laplacians and related operators.
194- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
195  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
196- Make public accessors for various attributes to facilitate writing composable code.
197- New backend: `/cpu/self/memcheck/serial`.
198- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
199  backends.  (Single source is coming soon for OCCA backends.)
200- Fix some missing edge cases in CUDA backend.
201
202### Performance Improvements
203
204- MAGMA backend performance optimization and non-tensor bases.
205- No-copy optimization in {c:func}`CeedOperatorApply`.
206
207### Interface changes
208
209- Replace {code}`CeedElemRestrictionCreateIdentity` and
210  {code}`CeedElemRestrictionCreateBlocked` with more flexible
211  {c:func}`CeedElemRestrictionCreateStrided` and
212  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
213- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
214- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
215  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
216  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
217  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
218  {c:func}`CeedElemRestrictionApply`.
219
220### Examples
221
222libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
223Notable additions include:
224
225- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
226  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
227
228- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
229  of domains (like the cube and sphere) by direct integration on a surface mesh;
230  demonstrates geometric dimension different from topological dimension.
231
232- PETSc {ref}`example-petsc-bps`:
233
234  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
235  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
236    and transparent CUDA support.
237  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
238    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
239    dimension different from topological dimension.
240
241- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
242  solver with algebraic multigrid coarse solve.
243
244- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
245  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
246  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
247  quasi-2D computational domain support.
248
249- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
250  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
251  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
252
253(v0-5)=
254
255## v0.5 (Sep 18, 2019)
256
257For this release, several improvements were made. Two new CUDA backends were added to
258the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
259performance using single-source {ref}`CeedQFunction`. From this release, users
260can define Q-Functions in a single source code independently of the targeted backend
261with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
262compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
263declaration, the {ref}`CeedQFunction` API has undergone a slight change:
264the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
265requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
266`QFunctionField` with eval mode `CEED EVAL GRAD`.
267
268Additionally, new CPU backends
269were included in this release, such as the `/cpu/self/opt/*` backends (which are
270written in pure C and use partial **E-vectors** to improve performance) and the
271`/cpu/self/ref/memcheck` backend (which relies upon the
272[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
273{ref}`CeedQFunction` have no undefined values).
274This release also included various performance improvements, bug fixes, new examples,
275and improved tests. Among these improvements, vectorized instructions for
276{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
277instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
278identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
279to include unstructured meshes handling were. For this expansion, the prior version of
280the PETSc BPs, which only included data associated with structured geometries, were
281renamed `bpsraw`, and the new version of the BPs, which can handle data associated
282with any unstructured geometry, were called `bps`. Additionally, other benchmark
283problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
284and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
285as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
286examples. Furthermoew, another standalone libCEED example, called `ex2`, which
287computes the surface area of a given mesh was added to this release.
288
289Backends available in this release:
290
291| CEED resource (`-ceed`)  | Backend                                             |
292|--------------------------|-----------------------------------------------------|
293| `/cpu/self/ref/serial`   | Serial reference implementation                     |
294| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
295| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
296| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
297| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
298| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
299| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
300| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
301| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
302| `/cpu/occa`              | Serial OCCA kernels                                 |
303| `/gpu/occa`              | CUDA OCCA kernels                                   |
304| `/omp/occa`              | OpenMP OCCA kernels                                 |
305| `/ocl/occa`              | OpenCL OCCA kernels                                 |
306| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
307| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
308| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
309| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
310| `/gpu/magma`             | CUDA MAGMA kernels                                  |
311
312Examples available in this release:
313
314:::{list-table}
315:header-rows: 1
316:widths: auto
317* - User code
318  - Example
319* - `ceed`
320  - * ex1 (volume)
321    * ex2 (surface)
322* - `mfem`
323  - * BP1 (scalar mass operator)
324    * BP3 (scalar Laplace operator)
325* - `petsc`
326  - * BP1 (scalar mass operator)
327    * BP2 (vector mass operator)
328    * BP3 (scalar Laplace operator)
329    * BP4 (vector Laplace operator)
330    * BP5 (collocated scalar Laplace operator)
331    * BP6 (collocated vector Laplace operator)
332    * Navier-Stokes
333* - `nek5000`
334  - * BP1 (scalar mass operator)
335    * BP3 (scalar Laplace operator)
336:::
337
338(v0-4)=
339
340## v0.4 (Apr 1, 2019)
341
342libCEED v0.4 was made again publicly available in the second full CEED software
343distribution, release CEED 2.0. This release contained notable features, such as
344four new CPU backends, two new GPU backends, CPU backend optimizations, initial
345support for operator composition, performance benchmarking, and a Navier-Stokes demo.
346The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
347backends process one element at a time and are intended for meshes with a smaller number
348of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
349eight interlaced elements and are intended for meshes with higher numbers of elements.
350The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
351performance. The `/cpu/self/xsmm/*` backends rely upon the
352[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
353performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
354The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
355performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
356parallelization approach, where each thread treats a finite element. Using just in time
357compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
358backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
359achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
360deteriorates very quickly when threads run out of registers.
361
362A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
363examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
364This example solves the time-dependent Navier-Stokes equations of compressible gas
365dynamics in a static Eulerian three-dimensional frame, using structured high-order
366finite/spectral element spatial discretizations and explicit high-order time-stepping
367(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
368so that the pointwise physics (defined at quadrature points) is separated from the
369parallelization and meshing concerns.
370
371Backends available in this release:
372
373| CEED resource (`-ceed`)  | Backend                                             |
374|--------------------------|-----------------------------------------------------|
375| `/cpu/self/ref/serial`   | Serial reference implementation                     |
376| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
377| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
378| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
379| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
380| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
381| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
382| `/cpu/occa`              | Serial OCCA kernels                                 |
383| `/gpu/occa`              | CUDA OCCA kernels                                   |
384| `/omp/occa`              | OpenMP OCCA kernels                                 |
385| `/ocl/occa`              | OpenCL OCCA kernels                                 |
386| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
387| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
388| `/gpu/magma`             | CUDA MAGMA kernels                                  |
389
390Examples available in this release:
391
392:::{list-table}
393:header-rows: 1
394:widths: auto
395* - User code
396  - Example
397* - `ceed`
398  - * ex1 (volume)
399* - `mfem`
400  - * BP1 (scalar mass operator)
401    * BP3 (scalar Laplace operator)
402* - `petsc`
403  - * BP1 (scalar mass operator)
404    * BP3 (scalar Laplace operator)
405    * Navier-Stokes
406* - `nek5000`
407  - * BP1 (scalar mass operator)
408    * BP3 (scalar Laplace operator)
409:::
410
411(v0-3)=
412
413## v0.3 (Sep 30, 2018)
414
415Notable features in this release include active/passive field interface, support for
416non-tensor bases, backend optimization, and improved Fortran interface. This release
417also focused on providing improved continuous integration, and many new tests with code
418coverage reports of about 90%. This release also provided a significant change to the
419public interface: a {ref}`CeedQFunction` can take any number of named input and output
420arguments while {ref}`CeedOperator` connects them to the actual data, which may be
421supplied explicitly to `CeedOperatorApply()` (active) or separately via
422`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
423of CeedQFunctions and composition of block solvers constructed using
424{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
425used in an optimized CPU backend. Although this is typically not visible to the user,
426it enables effective use of arbitrary-length SIMD while maintaining cache locality.
427This CPU backend also implements an algebraic factorization of tensor product gradients
428to perform fewer operations than standard application of interpolation and
429differentiation from nodes to quadrature points. This algebraic formulation
430automatically supports non-polynomial and non-interpolatory bases, thus is more general
431than the more common derivation in terms of Lagrange polynomials on the quadrature points.
432
433Backends available in this release:
434
435| CEED resource (`-ceed`) | Backend                                             |
436|-------------------------|-----------------------------------------------------|
437| `/cpu/self/blocked`     | Blocked reference implementation                    |
438| `/cpu/self/ref`         | Serial reference implementation                     |
439| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
440| `/cpu/occa`             | Serial OCCA kernels                                 |
441| `/gpu/occa`             | CUDA OCCA kernels                                   |
442| `/omp/occa`             | OpenMP OCCA kernels                                 |
443| `/ocl/occa`             | OpenCL OCCA kernels                                 |
444| `/gpu/magma`            | CUDA MAGMA kernels                                  |
445
446Examples available in this release:
447
448:::{list-table}
449:header-rows: 1
450:widths: auto
451* - User code
452  - Example
453* - `ceed`
454  - * ex1 (volume)
455* - `mfem`
456  - * BP1 (scalar mass operator)
457    * BP3 (scalar Laplace operator)
458* - `petsc`
459  - * BP1 (scalar mass operator)
460    * BP3 (scalar Laplace operator)
461* - `nek5000`
462  - * BP1 (scalar mass operator)
463    * BP3 (scalar Laplace operator)
464:::
465
466(v0-21)=
467
468## v0.21 (Sep 30, 2018)
469
470A MAGMA backend (which relies upon the
471[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
472release. This initial integration set up the framework of using MAGMA and provided the
473libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
474As any other backend, the MAGMA backend provides extended basic data structures for
475{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
476the fundamental CEED building blocks to work with the new data structures.
477In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
478but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
479internally, and thus seamlessly to the user, through the functions/methods that are
480provided to support them.
481
482Backends available in this release:
483
484| CEED resource (`-ceed`) | Backend                         |
485|-------------------------|---------------------------------|
486| `/cpu/self`             | Serial reference implementation |
487| `/cpu/occa`             | Serial OCCA kernels             |
488| `/gpu/occa`             | CUDA OCCA kernels               |
489| `/omp/occa`             | OpenMP OCCA kernels             |
490| `/ocl/occa`             | OpenCL OCCA kernels             |
491| `/gpu/magma`            | CUDA MAGMA kernels              |
492
493Examples available in this release:
494
495:::{list-table}
496:header-rows: 1
497:widths: auto
498* - User code
499  - Example
500* - `ceed`
501  - * ex1 (volume)
502* - `mfem`
503  - * BP1 (scalar mass operator)
504    * BP3 (scalar Laplace operator)
505* - `petsc`
506  - * BP1 (scalar mass operator)
507* - `nek5000`
508  - * BP1 (scalar mass operator)
509:::
510
511(v0-2)=
512
513## v0.2 (Mar 30, 2018)
514
515libCEED was made publicly available the first full CEED software distribution, release
516CEED 1.0. The distribution was made available using the Spack package manager to provide
517a common, easy-to-use build environment, where the user can build the CEED distribution
518with all dependencies. This release included a new Fortran interface for the library.
519This release also contained major improvements in the OCCA backend (including a new
520`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
521compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
522`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
523of the Laplace operator) was also added to this release.
524
525Backends available in this release:
526
527| CEED resource (`-ceed`) | Backend                         |
528|-------------------------|---------------------------------|
529| `/cpu/self`             | Serial reference implementation |
530| `/cpu/occa`             | Serial OCCA kernels             |
531| `/gpu/occa`             | CUDA OCCA kernels               |
532| `/omp/occa`             | OpenMP OCCA kernels             |
533| `/ocl/occa`             | OpenCL OCCA kernels             |
534
535Examples available in this release:
536
537:::{list-table}
538:header-rows: 1
539:widths: auto
540* - User code
541  - Example
542* - `ceed`
543  - * ex1 (volume)
544* - `mfem`
545  - * BP1 (scalar mass operator)
546    * BP3 (scalar Laplace operator)
547* - `petsc`
548  - * BP1 (scalar mass operator)
549* - `nek5000`
550  - * BP1 (scalar mass operator)
551:::
552
553(v0-1)=
554
555## v0.1 (Jan 3, 2018)
556
557Initial low-level API of the CEED project. The low-level API provides a set of Finite
558Elements kernels and components for writing new low-level kernels. Examples include:
559vector and sparse linear algebra, element matrix assembly over a batch of elements,
560partial assembly and action for efficient high-order operators like mass, diffusion,
561advection, etc. The main goal of the low-level API is to establish the basis for the
562high-level API. Also, identifying such low-level kernels and providing a reference
563implementation for them serves as the basis for specialized backend implementations.
564This release contained several backends: `/cpu/self`, and backends which rely upon the
565[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
566`/gpu/occa`, and `/omp/occa`.
567It also included several examples, in the `examples` folder:
568A standalone code that shows the usage of libCEED (with no external
569dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
570(with the application of the mass operator); and a `petsc` example to perform BP1
571(with the application of the mass operator).
572
573Backends available in this release:
574
575| CEED resource (`-ceed`) | Backend                         |
576|-------------------------|---------------------------------|
577| `/cpu/self`             | Serial reference implementation |
578| `/cpu/occa`             | Serial OCCA kernels             |
579| `/gpu/occa`             | CUDA OCCA kernels               |
580| `/omp/occa`             | OpenMP OCCA kernels             |
581
582Examples available in this release:
583
584| User code             | Example                           |
585|-----------------------|-----------------------------------|
586| `ceed`                | ex1 (scalar Laplace operator)     |
587| `mfem`                | BP1 (scalar mass operator)        |
588| `petsc`               | BP1 (scalar mass operator)        |
589```
590