xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 9e201c85545dd39529c090846df629a32c15659b)
1# Changes/Release Notes
2
3On this page we provide a summary of the main API changes, new features and examples for each release of libCEED.
4
5(main)=
6
7## Current `main` branch
8
9### Interface changes
10
11- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
12- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
13
14### New features
15
16- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
17- Added `CeedInt_FMT` to support potential future use of larger interger sizes.
18- Added CEED_QFUNCTION_ATTR for setting compiler attributes/pragmas to CEED_QFUNCTION_HELPER and CEED_QFUNCTION
19
20### Bugfix
21
22- Fix bug in setting device id for GPU backends.
23- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
24- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
25- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
26
27### Examples
28
29- Added various performance enhancements for {ref}`example-petsc-navier-stokes`.
30- Refactored {ref}`example-petsc-navier-stokes` to improve code reuse.
31- Added Shock Tube, Channel, and Flat Plate boundary layer problems to {ref}`example-petsc-navier-stokes`.
32- Added ability to use QFunctions for strong STG inflow in {ref}`example-petsc-navier-stokes`.
33
34### Maintainability
35
36- Refactored `/gpu/cuda/shared` and `/gpu/cuda/gen` as well as `/gpu/hip/shared` and `/gpu/hip/gen` backend to improve maintainablity and reduce duplicated code.
37- Enabled support for `p > 8` for `/gpu/*/shared` backends.
38
39(v0-10-1)=
40
41## v0.10.1 (Apr 11, 2022)
42
43### Interface changes
44
45- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
46
47### New features
48
49- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
50This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
51
52### Bugfix
53
54- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
55
56(v0-10)=
57
58## v0.10 (Mar 21, 2022)
59
60### Interface changes
61
62- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
63- 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`.
64- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
65- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
66- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
67- Warning added when compiling OCCA backend to alert users that this backend is experimental.
68- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
69- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
70- 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`.
71- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
72- 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.
73- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
74- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
75- 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.
76- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
77- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
78
79### New features
80
81- `CeedScalar` can now be set as `float` or `double` at compile time.
82- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
83- 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.
84- Remove need to guard library headers in QFunction source for code generation backends.
85- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
86- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
87- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
88- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
89- 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.
90- Added support for element matrix assembly in GPU backends.
91
92### Maintainability
93
94- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
95- `Include-what-you-use` makefile target added as `make iwyu`.
96- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
97- Put GPU JiTed kernel source code into separate files.
98- 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.
99
100(v0-9)=
101
102## v0.9 (Jul 6, 2021)
103
104### Interface changes
105
106- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
107
108### New features
109
110- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
111- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
112- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
113- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
114
115### Performance improvements
116
117### Examples
118
119- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
120- Fluid mechanics example adds GPU support and improves modularity.
121
122### Deprecated backends
123
124- 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.
125
126(v0-8)=
127
128## v0.8 (Mar 31, 2021)
129
130### Interface changes
131
132- Error handling improved to include enumerated error codes for C interface return values.
133- 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.
134
135### New features
136
137- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
138- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
139- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
140
141### Performance improvements
142
143- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
144- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
145
146### Examples
147
148- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
149- {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.
150- {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.
151- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
152
153(v0-7)=
154
155## v0.7 (Sep 29, 2020)
156
157### Interface changes
158
159- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
160  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
161  These changes improve support for mixed finite element methods.
162- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
163- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
164- 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.
165- 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.
166- 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.
167- 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.
168  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`.
169- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
170- 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.
171
172### New features
173
174- New HIP backend: `/gpu/hip/ref`.
175- CeedQFunction support for user `CUfunction`s in some backends
176
177### Performance improvements
178
179- OCCA backend rebuilt to facilitate future performance enhancements.
180- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
181
182### Examples
183
184- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
185
186### Deprecated backends
187
188- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
189
190(v0-6)=
191
192## v0.6 (Mar 29, 2020)
193
194libCEED v0.6 contains numerous new features and examples, as well as expanded
195documentation in [this new website](https://libceed.org).
196
197### New features
198
199- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
200  1-1 correspondence with the C interface, plus some convenience features.  For instance,
201  data stored in the {cpp:type}`CeedVector` structure are available without copy as
202  {py:class}`numpy.ndarray`.  Short tutorials are provided in
203  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
204- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
205  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
206  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
207  preconditioning ingredients and are used in the libCEED's multigrid examples.
208- The inverse of separable operators can be obtained using
209  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
210  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
211  especially for Laplacians and related operators.
212- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
213  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
214- Make public accessors for various attributes to facilitate writing composable code.
215- New backend: `/cpu/self/memcheck/serial`.
216- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
217  backends.  (Single source is coming soon for OCCA backends.)
218- Fix some missing edge cases in CUDA backend.
219
220### Performance Improvements
221
222- MAGMA backend performance optimization and non-tensor bases.
223- No-copy optimization in {c:func}`CeedOperatorApply`.
224
225### Interface changes
226
227- Replace {code}`CeedElemRestrictionCreateIdentity` and
228  {code}`CeedElemRestrictionCreateBlocked` with more flexible
229  {c:func}`CeedElemRestrictionCreateStrided` and
230  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
231- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
232- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
233  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
234  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
235  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
236  {c:func}`CeedElemRestrictionApply`.
237
238### Examples
239
240libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
241Notable additions include:
242
243- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
244  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
245
246- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
247  of domains (like the cube and sphere) by direct integration on a surface mesh;
248  demonstrates geometric dimension different from topological dimension.
249
250- PETSc {ref}`example-petsc-bps`:
251
252  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
253  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
254    and transparent CUDA support.
255  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
256    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
257    dimension different from topological dimension.
258
259- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
260  solver with algebraic multigrid coarse solve.
261
262- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
263  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
264  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
265  quasi-2D computational domain support.
266
267- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
268  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
269  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
270
271(v0-5)=
272
273## v0.5 (Sep 18, 2019)
274
275For this release, several improvements were made. Two new CUDA backends were added to
276the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
277performance using single-source {ref}`CeedQFunction`. From this release, users
278can define Q-Functions in a single source code independently of the targeted backend
279with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
280compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
281declaration, the {ref}`CeedQFunction` API has undergone a slight change:
282the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
283requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
284`QFunctionField` with eval mode `CEED EVAL GRAD`.
285
286Additionally, new CPU backends
287were included in this release, such as the `/cpu/self/opt/*` backends (which are
288written in pure C and use partial **E-vectors** to improve performance) and the
289`/cpu/self/ref/memcheck` backend (which relies upon the
290[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
291{ref}`CeedQFunction` have no undefined values).
292This release also included various performance improvements, bug fixes, new examples,
293and improved tests. Among these improvements, vectorized instructions for
294{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
295instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
296identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
297to include unstructured meshes handling were. For this expansion, the prior version of
298the PETSc BPs, which only included data associated with structured geometries, were
299renamed `bpsraw`, and the new version of the BPs, which can handle data associated
300with any unstructured geometry, were called `bps`. Additionally, other benchmark
301problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
302and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
303as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
304examples. Furthermoew, another standalone libCEED example, called `ex2`, which
305computes the surface area of a given mesh was added to this release.
306
307Backends available in this release:
308
309| CEED resource (`-ceed`)  | Backend                                             |
310|--------------------------|-----------------------------------------------------|
311| `/cpu/self/ref/serial`   | Serial reference implementation                     |
312| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
313| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
314| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
315| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
316| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
317| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
318| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
319| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
320| `/cpu/occa`              | Serial OCCA kernels                                 |
321| `/gpu/occa`              | CUDA OCCA kernels                                   |
322| `/omp/occa`              | OpenMP OCCA kernels                                 |
323| `/ocl/occa`              | OpenCL OCCA kernels                                 |
324| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
325| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
326| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
327| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
328| `/gpu/magma`             | CUDA MAGMA kernels                                  |
329
330Examples available in this release:
331
332:::{list-table}
333:header-rows: 1
334:widths: auto
335* - User code
336  - Example
337* - `ceed`
338  - * ex1 (volume)
339    * ex2 (surface)
340* - `mfem`
341  - * BP1 (scalar mass operator)
342    * BP3 (scalar Laplace operator)
343* - `petsc`
344  - * BP1 (scalar mass operator)
345    * BP2 (vector mass operator)
346    * BP3 (scalar Laplace operator)
347    * BP4 (vector Laplace operator)
348    * BP5 (collocated scalar Laplace operator)
349    * BP6 (collocated vector Laplace operator)
350    * Navier-Stokes
351* - `nek5000`
352  - * BP1 (scalar mass operator)
353    * BP3 (scalar Laplace operator)
354:::
355
356(v0-4)=
357
358## v0.4 (Apr 1, 2019)
359
360libCEED v0.4 was made again publicly available in the second full CEED software
361distribution, release CEED 2.0. This release contained notable features, such as
362four new CPU backends, two new GPU backends, CPU backend optimizations, initial
363support for operator composition, performance benchmarking, and a Navier-Stokes demo.
364The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
365backends process one element at a time and are intended for meshes with a smaller number
366of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
367eight interlaced elements and are intended for meshes with higher numbers of elements.
368The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
369performance. The `/cpu/self/xsmm/*` backends rely upon the
370[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
371performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
372The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
373performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
374parallelization approach, where each thread treats a finite element. Using just in time
375compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
376backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
377achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
378deteriorates very quickly when threads run out of registers.
379
380A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
381examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
382This example solves the time-dependent Navier-Stokes equations of compressible gas
383dynamics in a static Eulerian three-dimensional frame, using structured high-order
384finite/spectral element spatial discretizations and explicit high-order time-stepping
385(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
386so that the pointwise physics (defined at quadrature points) is separated from the
387parallelization and meshing concerns.
388
389Backends available in this release:
390
391| CEED resource (`-ceed`)  | Backend                                             |
392|--------------------------|-----------------------------------------------------|
393| `/cpu/self/ref/serial`   | Serial reference implementation                     |
394| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
395| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
396| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
397| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
398| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
399| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
400| `/cpu/occa`              | Serial OCCA kernels                                 |
401| `/gpu/occa`              | CUDA OCCA kernels                                   |
402| `/omp/occa`              | OpenMP OCCA kernels                                 |
403| `/ocl/occa`              | OpenCL OCCA kernels                                 |
404| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
405| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
406| `/gpu/magma`             | CUDA MAGMA kernels                                  |
407
408Examples available in this release:
409
410:::{list-table}
411:header-rows: 1
412:widths: auto
413* - User code
414  - Example
415* - `ceed`
416  - * ex1 (volume)
417* - `mfem`
418  - * BP1 (scalar mass operator)
419    * BP3 (scalar Laplace operator)
420* - `petsc`
421  - * BP1 (scalar mass operator)
422    * BP3 (scalar Laplace operator)
423    * Navier-Stokes
424* - `nek5000`
425  - * BP1 (scalar mass operator)
426    * BP3 (scalar Laplace operator)
427:::
428
429(v0-3)=
430
431## v0.3 (Sep 30, 2018)
432
433Notable features in this release include active/passive field interface, support for
434non-tensor bases, backend optimization, and improved Fortran interface. This release
435also focused on providing improved continuous integration, and many new tests with code
436coverage reports of about 90%. This release also provided a significant change to the
437public interface: a {ref}`CeedQFunction` can take any number of named input and output
438arguments while {ref}`CeedOperator` connects them to the actual data, which may be
439supplied explicitly to `CeedOperatorApply()` (active) or separately via
440`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
441of CeedQFunctions and composition of block solvers constructed using
442{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
443used in an optimized CPU backend. Although this is typically not visible to the user,
444it enables effective use of arbitrary-length SIMD while maintaining cache locality.
445This CPU backend also implements an algebraic factorization of tensor product gradients
446to perform fewer operations than standard application of interpolation and
447differentiation from nodes to quadrature points. This algebraic formulation
448automatically supports non-polynomial and non-interpolatory bases, thus is more general
449than the more common derivation in terms of Lagrange polynomials on the quadrature points.
450
451Backends available in this release:
452
453| CEED resource (`-ceed`) | Backend                                             |
454|-------------------------|-----------------------------------------------------|
455| `/cpu/self/blocked`     | Blocked reference implementation                    |
456| `/cpu/self/ref`         | Serial reference implementation                     |
457| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
458| `/cpu/occa`             | Serial OCCA kernels                                 |
459| `/gpu/occa`             | CUDA OCCA kernels                                   |
460| `/omp/occa`             | OpenMP OCCA kernels                                 |
461| `/ocl/occa`             | OpenCL OCCA kernels                                 |
462| `/gpu/magma`            | CUDA MAGMA kernels                                  |
463
464Examples available in this release:
465
466:::{list-table}
467:header-rows: 1
468:widths: auto
469* - User code
470  - Example
471* - `ceed`
472  - * ex1 (volume)
473* - `mfem`
474  - * BP1 (scalar mass operator)
475    * BP3 (scalar Laplace operator)
476* - `petsc`
477  - * BP1 (scalar mass operator)
478    * BP3 (scalar Laplace operator)
479* - `nek5000`
480  - * BP1 (scalar mass operator)
481    * BP3 (scalar Laplace operator)
482:::
483
484(v0-21)=
485
486## v0.21 (Sep 30, 2018)
487
488A MAGMA backend (which relies upon the
489[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
490release. This initial integration set up the framework of using MAGMA and provided the
491libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
492As any other backend, the MAGMA backend provides extended basic data structures for
493{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
494the fundamental CEED building blocks to work with the new data structures.
495In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
496but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
497internally, and thus seamlessly to the user, through the functions/methods that are
498provided to support them.
499
500Backends available in this release:
501
502| CEED resource (`-ceed`) | Backend                         |
503|-------------------------|---------------------------------|
504| `/cpu/self`             | Serial reference implementation |
505| `/cpu/occa`             | Serial OCCA kernels             |
506| `/gpu/occa`             | CUDA OCCA kernels               |
507| `/omp/occa`             | OpenMP OCCA kernels             |
508| `/ocl/occa`             | OpenCL OCCA kernels             |
509| `/gpu/magma`            | CUDA MAGMA 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(v0-2)=
530
531## v0.2 (Mar 30, 2018)
532
533libCEED was made publicly available the first full CEED software distribution, release
534CEED 1.0. The distribution was made available using the Spack package manager to provide
535a common, easy-to-use build environment, where the user can build the CEED distribution
536with all dependencies. This release included a new Fortran interface for the library.
537This release also contained major improvements in the OCCA backend (including a new
538`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
539compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
540`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
541of the Laplace operator) was also added to this release.
542
543Backends available in this release:
544
545| CEED resource (`-ceed`) | Backend                         |
546|-------------------------|---------------------------------|
547| `/cpu/self`             | Serial reference implementation |
548| `/cpu/occa`             | Serial OCCA kernels             |
549| `/gpu/occa`             | CUDA OCCA kernels               |
550| `/omp/occa`             | OpenMP OCCA kernels             |
551| `/ocl/occa`             | OpenCL OCCA kernels             |
552
553Examples available in this release:
554
555:::{list-table}
556:header-rows: 1
557:widths: auto
558* - User code
559  - Example
560* - `ceed`
561  - * ex1 (volume)
562* - `mfem`
563  - * BP1 (scalar mass operator)
564    * BP3 (scalar Laplace operator)
565* - `petsc`
566  - * BP1 (scalar mass operator)
567* - `nek5000`
568  - * BP1 (scalar mass operator)
569:::
570
571(v0-1)=
572
573## v0.1 (Jan 3, 2018)
574
575Initial low-level API of the CEED project. The low-level API provides a set of Finite
576Elements kernels and components for writing new low-level kernels. Examples include:
577vector and sparse linear algebra, element matrix assembly over a batch of elements,
578partial assembly and action for efficient high-order operators like mass, diffusion,
579advection, etc. The main goal of the low-level API is to establish the basis for the
580high-level API. Also, identifying such low-level kernels and providing a reference
581implementation for them serves as the basis for specialized backend implementations.
582This release contained several backends: `/cpu/self`, and backends which rely upon the
583[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
584`/gpu/occa`, and `/omp/occa`.
585It also included several examples, in the `examples` folder:
586A standalone code that shows the usage of libCEED (with no external
587dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
588(with the application of the mass operator); and a `petsc` example to perform BP1
589(with the application of the mass operator).
590
591Backends available in this release:
592
593| CEED resource (`-ceed`) | Backend                         |
594|-------------------------|---------------------------------|
595| `/cpu/self`             | Serial reference implementation |
596| `/cpu/occa`             | Serial OCCA kernels             |
597| `/gpu/occa`             | CUDA OCCA kernels               |
598| `/omp/occa`             | OpenMP OCCA kernels             |
599
600Examples available in this release:
601
602| User code             | Example                           |
603|-----------------------|-----------------------------------|
604| `ceed`                | ex1 (scalar Laplace operator)     |
605| `mfem`                | BP1 (scalar mass operator)        |
606| `petsc`               | BP1 (scalar mass operator)        |
607```
608