xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 07328cdeb7cc6fe1ac9596b69aea7fe9a5beddd3)
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- Update `CeedOperatorContext*` functions to `CeedOperator*Context*` functions for consistency.
12For example, `CeedOperatorContextGetFieldLabel` was renamed to `CeedOperatorGetContextFieldLabel`.
13
14### New features
15
16- Added {c:func}`CeedOperatorGetFieldByName` to access a specific `CeedOperatorField` by its name.
17- Update `/cpu/self/memcheck/*` backends to help verify `CeedVector` array access assumptions and `CeedQFunction` user output assumptions.
18- Update {c:func}`CeedOperatorLinearAssembleDiagonal` to provide default implementation that supports `CeedOperator` with multiple active bases.
19- Added Sycl backends `/gpu/sycl/ref` and `/gpu/sycl/shared`.
20- Added {c:func}`CeedBasisApplyAtPoints` for evalution of values and derivaties at arbitrary points inside elements.
21- Added support for non-tensor $H(\text{curl})$ finite element spaces with {c:func}`CeedBasisCreateHcurl`.
22- Added {c:func}`CeedElemRestrictionCreateCurlOriented`, similar to {c:func}`CeedElemRestrictionCreateOriented`, for element restrictions requiring more general element transformations such as those for high-order $H(\text{curl})$ spaces on tetrahedra (see [https://dl.acm.org/doi/pdf/10.1145/3524456](https://dl.acm.org/doi/pdf/10.1145/3524456)).
23- Added {c:func}`CeedOperatorLinearAssemblePointBlockDiagonalSymbolic` to create COO mapping for mapping out of {c:func}`CeedOperatorLinearAssemblePointBlockDiagonal`
24
25### Examples
26
27#### {ref}`example-petsc-bps`
28
29- Requires PETSc version 3.19 or later.
30
31(v0-11)=
32
33## v0.11 (Dec 24, 2022)
34
35### Interface changes
36
37- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
38- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
39- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface.
40- Renamed `CEED_BASIS_COLLOCATED` to `CEED_BASIS_NONE` for clarity.
41Some users previously misinterpreted a `CeedOperator` field using `CEED_BASIS_COLLOCATED` as meaning that the entire `CeedOperator` used a quadrature space that is collocated with the nodal space of the active bases.
42
43### New features
44
45- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
46- Improved support for $H(\text{div})$ bases.
47- Added `CeedInt_FMT` to support potential future use of larger integer sizes.
48- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`.
49- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
50Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
51This issue will be fixed in a future OCCA release.
52
53### Bugfix
54
55- Fix bug in setting device id for GPU backends.
56- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
57- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
58- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
59
60### Examples
61
62#### {ref}`example-petsc-navier-stokes`
63
64- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs.
65- Refactored to improve code reuse and modularity.
66- Support for primitive variables for more accurate boundary layers and all-speed flow.
67- Added $YZ\beta$ shock capturing scheme and Shock Tube example.
68- Added Channel example, with comparison to analytic solutions.
69- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations.
70- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions.
71- Added "freestream" boundary conditions based on HLLC Riemann solver.
72- Automated stabilization coefficients for different basis degree.
73
74#### {ref}`example-petsc-bps`
75
76- Support for convergence studies.
77
78### Maintainability
79
80- 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.
81- Enabled support for `p > 8` for `/gpu/*/shared` backends.
82- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
83- Improved test harness.
84
85(v0-10-1)=
86
87## v0.10.1 (Apr 11, 2022)
88
89### Interface changes
90
91- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
92
93### New features
94
95- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
96This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
97
98### Bugfix
99
100- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
101
102(v0-10)=
103
104## v0.10 (Mar 21, 2022)
105
106### Interface changes
107
108- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
109- 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`.
110- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
111- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
112- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
113- Warning added when compiling OCCA backend to alert users that this backend is experimental.
114- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
115- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
116- 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`.
117- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
118- 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.
119- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
120- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
121- 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.
122- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
123- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
124
125### New features
126
127- `CeedScalar` can now be set as `float` or `double` at compile time.
128- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
129- 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.
130- Remove need to guard library headers in QFunction source for code generation backends.
131- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
132- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
133- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
134- Added support for non-tensor $H(\text{div})$ elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
135- 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.
136- Added support for element matrix assembly in GPU backends.
137
138### Maintainability
139
140- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
141- `Include-what-you-use` makefile target added as `make iwyu`.
142- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
143- Put GPU JiTed kernel source code into separate files.
144- 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.
145
146(v0-9)=
147
148## v0.9 (Jul 6, 2021)
149
150### Interface changes
151
152- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
153
154### New features
155
156- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
157- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
158- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
159- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
160
161### Performance improvements
162
163### Examples
164
165- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
166- Fluid mechanics example adds GPU support and improves modularity.
167
168### Deprecated backends
169
170- 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.
171
172(v0-8)=
173
174## v0.8 (Mar 31, 2021)
175
176### Interface changes
177
178- Error handling improved to include enumerated error codes for C interface return values.
179- 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.
180
181### New features
182
183- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
184- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
185- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
186
187### Performance improvements
188
189- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
190- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
191
192### Examples
193
194- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
195- {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.
196- {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.
197- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
198
199(v0-7)=
200
201## v0.7 (Sep 29, 2020)
202
203### Interface changes
204
205- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
206  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
207  These changes improve support for mixed finite element methods.
208- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
209- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
210- 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.
211- 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.
212- 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.
213- 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.
214  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`.
215- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
216- 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.
217
218### New features
219
220- New HIP backend: `/gpu/hip/ref`.
221- CeedQFunction support for user `CUfunction`s in some backends
222
223### Performance improvements
224
225- OCCA backend rebuilt to facilitate future performance enhancements.
226- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
227
228### Examples
229
230- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
231
232### Deprecated backends
233
234- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
235
236(v0-6)=
237
238## v0.6 (Mar 29, 2020)
239
240libCEED v0.6 contains numerous new features and examples, as well as expanded
241documentation in [this new website](https://libceed.org).
242
243### New features
244
245- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
246  1-1 correspondence with the C interface, plus some convenience features.  For instance,
247  data stored in the {cpp:type}`CeedVector` structure are available without copy as
248  {py:class}`numpy.ndarray`.  Short tutorials are provided in
249  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
250- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
251  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
252  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
253  preconditioning ingredients and are used in the libCEED's multigrid examples.
254- The inverse of separable operators can be obtained using
255  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
256  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
257  especially for Laplacians and related operators.
258- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
259  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
260- Make public accessors for various attributes to facilitate writing composable code.
261- New backend: `/cpu/self/memcheck/serial`.
262- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
263  backends.  (Single source is coming soon for OCCA backends.)
264- Fix some missing edge cases in CUDA backend.
265
266### Performance Improvements
267
268- MAGMA backend performance optimization and non-tensor bases.
269- No-copy optimization in {c:func}`CeedOperatorApply`.
270
271### Interface changes
272
273- Replace {code}`CeedElemRestrictionCreateIdentity` and
274  {code}`CeedElemRestrictionCreateBlocked` with more flexible
275  {c:func}`CeedElemRestrictionCreateStrided` and
276  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
277- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
278- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
279  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
280  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
281  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
282  {c:func}`CeedElemRestrictionApply`.
283
284### Examples
285
286libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
287Notable additions include:
288
289- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
290  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
291
292- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
293  of domains (like the cube and sphere) by direct integration on a surface mesh;
294  demonstrates geometric dimension different from topological dimension.
295
296- PETSc {ref}`example-petsc-bps`:
297
298  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
299  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
300    and transparent CUDA support.
301  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
302    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
303    dimension different from topological dimension.
304
305- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
306  solver with algebraic multigrid coarse solve.
307
308- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
309  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
310  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
311  quasi-2D computational domain support.
312
313- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
314  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
315  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
316
317(v0-5)=
318
319## v0.5 (Sep 18, 2019)
320
321For this release, several improvements were made. Two new CUDA backends were added to
322the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
323performance using single-source {ref}`CeedQFunction`. From this release, users
324can define Q-Functions in a single source code independently of the targeted backend
325with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
326compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
327declaration, the {ref}`CeedQFunction` API has undergone a slight change:
328the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
329requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
330`QFunctionField` with eval mode `CEED EVAL GRAD`.
331
332Additionally, new CPU backends
333were included in this release, such as the `/cpu/self/opt/*` backends (which are
334written in pure C and use partial **E-vectors** to improve performance) and the
335`/cpu/self/ref/memcheck` backend (which relies upon the
336[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
337{ref}`CeedQFunction` have no undefined values).
338This release also included various performance improvements, bug fixes, new examples,
339and improved tests. Among these improvements, vectorized instructions for
340{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
341instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
342identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
343to include unstructured meshes handling were. For this expansion, the prior version of
344the PETSc BPs, which only included data associated with structured geometries, were
345renamed `bpsraw`, and the new version of the BPs, which can handle data associated
346with any unstructured geometry, were called `bps`. Additionally, other benchmark
347problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
348and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
349as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
350examples. Furthermoew, another standalone libCEED example, called `ex2`, which
351computes the surface area of a given mesh was added to this release.
352
353Backends available in this release:
354
355| CEED resource (`-ceed`)  | Backend                                             |
356|--------------------------|-----------------------------------------------------|
357| `/cpu/self/ref/serial`   | Serial reference implementation                     |
358| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
359| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
360| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
361| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
362| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
363| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
364| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
365| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
366| `/cpu/occa`              | Serial OCCA kernels                                 |
367| `/gpu/occa`              | CUDA OCCA kernels                                   |
368| `/omp/occa`              | OpenMP OCCA kernels                                 |
369| `/ocl/occa`              | OpenCL OCCA kernels                                 |
370| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
371| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
372| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
373| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
374| `/gpu/magma`             | CUDA MAGMA kernels                                  |
375
376Examples available in this release:
377
378:::{list-table}
379:header-rows: 1
380:widths: auto
381* - User code
382  - Example
383* - `ceed`
384  - * ex1 (volume)
385    * ex2 (surface)
386* - `mfem`
387  - * BP1 (scalar mass operator)
388    * BP3 (scalar Laplace operator)
389* - `petsc`
390  - * BP1 (scalar mass operator)
391    * BP2 (vector mass operator)
392    * BP3 (scalar Laplace operator)
393    * BP4 (vector Laplace operator)
394    * BP5 (collocated scalar Laplace operator)
395    * BP6 (collocated vector Laplace operator)
396    * Navier-Stokes
397* - `nek5000`
398  - * BP1 (scalar mass operator)
399    * BP3 (scalar Laplace operator)
400:::
401
402(v0-4)=
403
404## v0.4 (Apr 1, 2019)
405
406libCEED v0.4 was made again publicly available in the second full CEED software
407distribution, release CEED 2.0. This release contained notable features, such as
408four new CPU backends, two new GPU backends, CPU backend optimizations, initial
409support for operator composition, performance benchmarking, and a Navier-Stokes demo.
410The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
411backends process one element at a time and are intended for meshes with a smaller number
412of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
413eight interlaced elements and are intended for meshes with higher numbers of elements.
414The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
415performance. The `/cpu/self/xsmm/*` backends rely upon the
416[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
417performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
418The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
419performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
420parallelization approach, where each thread treats a finite element. Using just in time
421compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
422backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
423achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
424deteriorates very quickly when threads run out of registers.
425
426A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
427examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
428This example solves the time-dependent Navier-Stokes equations of compressible gas
429dynamics in a static Eulerian three-dimensional frame, using structured high-order
430finite/spectral element spatial discretizations and explicit high-order time-stepping
431(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
432so that the pointwise physics (defined at quadrature points) is separated from the
433parallelization and meshing concerns.
434
435Backends available in this release:
436
437| CEED resource (`-ceed`)  | Backend                                             |
438|--------------------------|-----------------------------------------------------|
439| `/cpu/self/ref/serial`   | Serial reference implementation                     |
440| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
441| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
442| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
443| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
444| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
445| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
446| `/cpu/occa`              | Serial OCCA kernels                                 |
447| `/gpu/occa`              | CUDA OCCA kernels                                   |
448| `/omp/occa`              | OpenMP OCCA kernels                                 |
449| `/ocl/occa`              | OpenCL OCCA kernels                                 |
450| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
451| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
452| `/gpu/magma`             | CUDA MAGMA kernels                                  |
453
454Examples available in this release:
455
456:::{list-table}
457:header-rows: 1
458:widths: auto
459* - User code
460  - Example
461* - `ceed`
462  - * ex1 (volume)
463* - `mfem`
464  - * BP1 (scalar mass operator)
465    * BP3 (scalar Laplace operator)
466* - `petsc`
467  - * BP1 (scalar mass operator)
468    * BP3 (scalar Laplace operator)
469    * Navier-Stokes
470* - `nek5000`
471  - * BP1 (scalar mass operator)
472    * BP3 (scalar Laplace operator)
473:::
474
475(v0-3)=
476
477## v0.3 (Sep 30, 2018)
478
479Notable features in this release include active/passive field interface, support for
480non-tensor bases, backend optimization, and improved Fortran interface. This release
481also focused on providing improved continuous integration, and many new tests with code
482coverage reports of about 90%. This release also provided a significant change to the
483public interface: a {ref}`CeedQFunction` can take any number of named input and output
484arguments while {ref}`CeedOperator` connects them to the actual data, which may be
485supplied explicitly to `CeedOperatorApply()` (active) or separately via
486`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
487of CeedQFunctions and composition of block solvers constructed using
488{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
489used in an optimized CPU backend. Although this is typically not visible to the user,
490it enables effective use of arbitrary-length SIMD while maintaining cache locality.
491This CPU backend also implements an algebraic factorization of tensor product gradients
492to perform fewer operations than standard application of interpolation and
493differentiation from nodes to quadrature points. This algebraic formulation
494automatically supports non-polynomial and non-interpolatory bases, thus is more general
495than the more common derivation in terms of Lagrange polynomials on the quadrature points.
496
497Backends available in this release:
498
499| CEED resource (`-ceed`) | Backend                                             |
500|-------------------------|-----------------------------------------------------|
501| `/cpu/self/blocked`     | Blocked reference implementation                    |
502| `/cpu/self/ref`         | Serial reference implementation                     |
503| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
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| `/gpu/magma`            | CUDA MAGMA kernels                                  |
509
510Examples available in this release:
511
512:::{list-table}
513:header-rows: 1
514:widths: auto
515* - User code
516  - Example
517* - `ceed`
518  - * ex1 (volume)
519* - `mfem`
520  - * BP1 (scalar mass operator)
521    * BP3 (scalar Laplace operator)
522* - `petsc`
523  - * BP1 (scalar mass operator)
524    * BP3 (scalar Laplace operator)
525* - `nek5000`
526  - * BP1 (scalar mass operator)
527    * BP3 (scalar Laplace operator)
528:::
529
530(v0-21)=
531
532## v0.21 (Sep 30, 2018)
533
534A MAGMA backend (which relies upon the
535[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
536release. This initial integration set up the framework of using MAGMA and provided the
537libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
538As any other backend, the MAGMA backend provides extended basic data structures for
539{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
540the fundamental CEED building blocks to work with the new data structures.
541In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
542but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
543internally, and thus seamlessly to the user, through the functions/methods that are
544provided to support them.
545
546Backends available in this release:
547
548| CEED resource (`-ceed`) | Backend                         |
549|-------------------------|---------------------------------|
550| `/cpu/self`             | Serial reference implementation |
551| `/cpu/occa`             | Serial OCCA kernels             |
552| `/gpu/occa`             | CUDA OCCA kernels               |
553| `/omp/occa`             | OpenMP OCCA kernels             |
554| `/ocl/occa`             | OpenCL OCCA kernels             |
555| `/gpu/magma`            | CUDA MAGMA kernels              |
556
557Examples available in this release:
558
559:::{list-table}
560:header-rows: 1
561:widths: auto
562* - User code
563  - Example
564* - `ceed`
565  - * ex1 (volume)
566* - `mfem`
567  - * BP1 (scalar mass operator)
568    * BP3 (scalar Laplace operator)
569* - `petsc`
570  - * BP1 (scalar mass operator)
571* - `nek5000`
572  - * BP1 (scalar mass operator)
573:::
574
575(v0-2)=
576
577## v0.2 (Mar 30, 2018)
578
579libCEED was made publicly available the first full CEED software distribution, release
580CEED 1.0. The distribution was made available using the Spack package manager to provide
581a common, easy-to-use build environment, where the user can build the CEED distribution
582with all dependencies. This release included a new Fortran interface for the library.
583This release also contained major improvements in the OCCA backend (including a new
584`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
585compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
586`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
587of the Laplace operator) was also added to this release.
588
589Backends available in this release:
590
591| CEED resource (`-ceed`) | Backend                         |
592|-------------------------|---------------------------------|
593| `/cpu/self`             | Serial reference implementation |
594| `/cpu/occa`             | Serial OCCA kernels             |
595| `/gpu/occa`             | CUDA OCCA kernels               |
596| `/omp/occa`             | OpenMP OCCA kernels             |
597| `/ocl/occa`             | OpenCL OCCA kernels             |
598
599Examples available in this release:
600
601:::{list-table}
602:header-rows: 1
603:widths: auto
604* - User code
605  - Example
606* - `ceed`
607  - * ex1 (volume)
608* - `mfem`
609  - * BP1 (scalar mass operator)
610    * BP3 (scalar Laplace operator)
611* - `petsc`
612  - * BP1 (scalar mass operator)
613* - `nek5000`
614  - * BP1 (scalar mass operator)
615:::
616
617(v0-1)=
618
619## v0.1 (Jan 3, 2018)
620
621Initial low-level API of the CEED project. The low-level API provides a set of Finite
622Elements kernels and components for writing new low-level kernels. Examples include:
623vector and sparse linear algebra, element matrix assembly over a batch of elements,
624partial assembly and action for efficient high-order operators like mass, diffusion,
625advection, etc. The main goal of the low-level API is to establish the basis for the
626high-level API. Also, identifying such low-level kernels and providing a reference
627implementation for them serves as the basis for specialized backend implementations.
628This release contained several backends: `/cpu/self`, and backends which rely upon the
629[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
630`/gpu/occa`, and `/omp/occa`.
631It also included several examples, in the `examples` folder:
632A standalone code that shows the usage of libCEED (with no external
633dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
634(with the application of the mass operator); and a `petsc` example to perform BP1
635(with the application of the mass operator).
636
637Backends available in this release:
638
639| CEED resource (`-ceed`) | Backend                         |
640|-------------------------|---------------------------------|
641| `/cpu/self`             | Serial reference implementation |
642| `/cpu/occa`             | Serial OCCA kernels             |
643| `/gpu/occa`             | CUDA OCCA kernels               |
644| `/omp/occa`             | OpenMP OCCA kernels             |
645
646Examples available in this release:
647
648| User code             | Example                           |
649|-----------------------|-----------------------------------|
650| `ceed`                | ex1 (scalar Laplace operator)     |
651| `mfem`                | BP1 (scalar mass operator)        |
652| `petsc`               | BP1 (scalar mass operator)        |
653```
654