xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 356036fa84f714fa73ef64c9a80ce2028dde816f)
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
24### Examples
25
26#### {ref}`example-petsc-bps`
27
28- Requires PETSc version 3.19 or later.
29
30(v0-11)=
31
32## v0.11 (Dec 24, 2022)
33
34### Interface changes
35
36- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
37- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
38- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface.
39- Renamed `CEED_BASIS_COLLOCATED` to `CEED_BASIS_NONE` for clarity.
40
41### New features
42
43- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
44- Improved support for $H(\text{div})$ bases.
45- Added `CeedInt_FMT` to support potential future use of larger integer sizes.
46- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`.
47- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
48Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
49This issue will be fixed in a future OCCA release.
50
51### Bugfix
52
53- Fix bug in setting device id for GPU backends.
54- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
55- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
56- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
57
58### Examples
59
60#### {ref}`example-petsc-navier-stokes`
61
62- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs.
63- Refactored to improve code reuse and modularity.
64- Support for primitive variables for more accurate boundary layers and all-speed flow.
65- Added $YZ\beta$ shock capturing scheme and Shock Tube example.
66- Added Channel example, with comparison to analytic solutions.
67- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations.
68- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions.
69- Added "freestream" boundary conditions based on HLLC Riemann solver.
70- Automated stabilization coefficients for different basis degree.
71
72#### {ref}`example-petsc-bps`
73
74- Support for convergence studies.
75
76### Maintainability
77
78- 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.
79- Enabled support for `p > 8` for `/gpu/*/shared` backends.
80- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
81- Improved test harness.
82
83(v0-10-1)=
84
85## v0.10.1 (Apr 11, 2022)
86
87### Interface changes
88
89- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
90
91### New features
92
93- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
94This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
95
96### Bugfix
97
98- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
99
100(v0-10)=
101
102## v0.10 (Mar 21, 2022)
103
104### Interface changes
105
106- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
107- 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`.
108- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
109- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
110- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
111- Warning added when compiling OCCA backend to alert users that this backend is experimental.
112- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
113- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
114- 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`.
115- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
116- 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.
117- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
118- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
119- 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.
120- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
121- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
122
123### New features
124
125- `CeedScalar` can now be set as `float` or `double` at compile time.
126- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
127- 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.
128- Remove need to guard library headers in QFunction source for code generation backends.
129- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
130- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
131- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
132- Added support for non-tensor $H(\text{div})$ elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
133- 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.
134- Added support for element matrix assembly in GPU backends.
135
136### Maintainability
137
138- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
139- `Include-what-you-use` makefile target added as `make iwyu`.
140- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
141- Put GPU JiTed kernel source code into separate files.
142- 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.
143
144(v0-9)=
145
146## v0.9 (Jul 6, 2021)
147
148### Interface changes
149
150- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
151
152### New features
153
154- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
155- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
156- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
157- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
158
159### Performance improvements
160
161### Examples
162
163- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
164- Fluid mechanics example adds GPU support and improves modularity.
165
166### Deprecated backends
167
168- 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.
169
170(v0-8)=
171
172## v0.8 (Mar 31, 2021)
173
174### Interface changes
175
176- Error handling improved to include enumerated error codes for C interface return values.
177- 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.
178
179### New features
180
181- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
182- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
183- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
184
185### Performance improvements
186
187- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
188- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
189
190### Examples
191
192- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
193- {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.
194- {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.
195- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
196
197(v0-7)=
198
199## v0.7 (Sep 29, 2020)
200
201### Interface changes
202
203- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
204  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
205  These changes improve support for mixed finite element methods.
206- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
207- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
208- 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.
209- 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.
210- 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.
211- 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.
212  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`.
213- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
214- 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.
215
216### New features
217
218- New HIP backend: `/gpu/hip/ref`.
219- CeedQFunction support for user `CUfunction`s in some backends
220
221### Performance improvements
222
223- OCCA backend rebuilt to facilitate future performance enhancements.
224- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
225
226### Examples
227
228- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
229
230### Deprecated backends
231
232- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
233
234(v0-6)=
235
236## v0.6 (Mar 29, 2020)
237
238libCEED v0.6 contains numerous new features and examples, as well as expanded
239documentation in [this new website](https://libceed.org).
240
241### New features
242
243- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
244  1-1 correspondence with the C interface, plus some convenience features.  For instance,
245  data stored in the {cpp:type}`CeedVector` structure are available without copy as
246  {py:class}`numpy.ndarray`.  Short tutorials are provided in
247  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
248- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
249  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
250  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
251  preconditioning ingredients and are used in the libCEED's multigrid examples.
252- The inverse of separable operators can be obtained using
253  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
254  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
255  especially for Laplacians and related operators.
256- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
257  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
258- Make public accessors for various attributes to facilitate writing composable code.
259- New backend: `/cpu/self/memcheck/serial`.
260- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
261  backends.  (Single source is coming soon for OCCA backends.)
262- Fix some missing edge cases in CUDA backend.
263
264### Performance Improvements
265
266- MAGMA backend performance optimization and non-tensor bases.
267- No-copy optimization in {c:func}`CeedOperatorApply`.
268
269### Interface changes
270
271- Replace {code}`CeedElemRestrictionCreateIdentity` and
272  {code}`CeedElemRestrictionCreateBlocked` with more flexible
273  {c:func}`CeedElemRestrictionCreateStrided` and
274  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
275- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
276- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
277  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
278  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
279  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
280  {c:func}`CeedElemRestrictionApply`.
281
282### Examples
283
284libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
285Notable additions include:
286
287- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
288  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
289
290- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
291  of domains (like the cube and sphere) by direct integration on a surface mesh;
292  demonstrates geometric dimension different from topological dimension.
293
294- PETSc {ref}`example-petsc-bps`:
295
296  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
297  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
298    and transparent CUDA support.
299  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
300    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
301    dimension different from topological dimension.
302
303- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
304  solver with algebraic multigrid coarse solve.
305
306- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
307  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
308  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
309  quasi-2D computational domain support.
310
311- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
312  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
313  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
314
315(v0-5)=
316
317## v0.5 (Sep 18, 2019)
318
319For this release, several improvements were made. Two new CUDA backends were added to
320the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
321performance using single-source {ref}`CeedQFunction`. From this release, users
322can define Q-Functions in a single source code independently of the targeted backend
323with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
324compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
325declaration, the {ref}`CeedQFunction` API has undergone a slight change:
326the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
327requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
328`QFunctionField` with eval mode `CEED EVAL GRAD`.
329
330Additionally, new CPU backends
331were included in this release, such as the `/cpu/self/opt/*` backends (which are
332written in pure C and use partial **E-vectors** to improve performance) and the
333`/cpu/self/ref/memcheck` backend (which relies upon the
334[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
335{ref}`CeedQFunction` have no undefined values).
336This release also included various performance improvements, bug fixes, new examples,
337and improved tests. Among these improvements, vectorized instructions for
338{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
339instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
340identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
341to include unstructured meshes handling were. For this expansion, the prior version of
342the PETSc BPs, which only included data associated with structured geometries, were
343renamed `bpsraw`, and the new version of the BPs, which can handle data associated
344with any unstructured geometry, were called `bps`. Additionally, other benchmark
345problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
346and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
347as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
348examples. Furthermoew, another standalone libCEED example, called `ex2`, which
349computes the surface area of a given mesh was added to this release.
350
351Backends available in this release:
352
353| CEED resource (`-ceed`)  | Backend                                             |
354|--------------------------|-----------------------------------------------------|
355| `/cpu/self/ref/serial`   | Serial reference implementation                     |
356| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
357| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
358| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
359| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
360| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
361| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
362| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
363| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
364| `/cpu/occa`              | Serial OCCA kernels                                 |
365| `/gpu/occa`              | CUDA OCCA kernels                                   |
366| `/omp/occa`              | OpenMP OCCA kernels                                 |
367| `/ocl/occa`              | OpenCL OCCA kernels                                 |
368| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
369| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
370| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
371| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
372| `/gpu/magma`             | CUDA MAGMA kernels                                  |
373
374Examples available in this release:
375
376:::{list-table}
377:header-rows: 1
378:widths: auto
379* - User code
380  - Example
381* - `ceed`
382  - * ex1 (volume)
383    * ex2 (surface)
384* - `mfem`
385  - * BP1 (scalar mass operator)
386    * BP3 (scalar Laplace operator)
387* - `petsc`
388  - * BP1 (scalar mass operator)
389    * BP2 (vector mass operator)
390    * BP3 (scalar Laplace operator)
391    * BP4 (vector Laplace operator)
392    * BP5 (collocated scalar Laplace operator)
393    * BP6 (collocated vector Laplace operator)
394    * Navier-Stokes
395* - `nek5000`
396  - * BP1 (scalar mass operator)
397    * BP3 (scalar Laplace operator)
398:::
399
400(v0-4)=
401
402## v0.4 (Apr 1, 2019)
403
404libCEED v0.4 was made again publicly available in the second full CEED software
405distribution, release CEED 2.0. This release contained notable features, such as
406four new CPU backends, two new GPU backends, CPU backend optimizations, initial
407support for operator composition, performance benchmarking, and a Navier-Stokes demo.
408The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
409backends process one element at a time and are intended for meshes with a smaller number
410of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
411eight interlaced elements and are intended for meshes with higher numbers of elements.
412The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
413performance. The `/cpu/self/xsmm/*` backends rely upon the
414[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
415performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
416The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
417performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
418parallelization approach, where each thread treats a finite element. Using just in time
419compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
420backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
421achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
422deteriorates very quickly when threads run out of registers.
423
424A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
425examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
426This example solves the time-dependent Navier-Stokes equations of compressible gas
427dynamics in a static Eulerian three-dimensional frame, using structured high-order
428finite/spectral element spatial discretizations and explicit high-order time-stepping
429(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
430so that the pointwise physics (defined at quadrature points) is separated from the
431parallelization and meshing concerns.
432
433Backends available in this release:
434
435| CEED resource (`-ceed`)  | Backend                                             |
436|--------------------------|-----------------------------------------------------|
437| `/cpu/self/ref/serial`   | Serial reference implementation                     |
438| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
439| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
440| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
441| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
442| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
443| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
444| `/cpu/occa`              | Serial OCCA kernels                                 |
445| `/gpu/occa`              | CUDA OCCA kernels                                   |
446| `/omp/occa`              | OpenMP OCCA kernels                                 |
447| `/ocl/occa`              | OpenCL OCCA kernels                                 |
448| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
449| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
450| `/gpu/magma`             | CUDA MAGMA kernels                                  |
451
452Examples available in this release:
453
454:::{list-table}
455:header-rows: 1
456:widths: auto
457* - User code
458  - Example
459* - `ceed`
460  - * ex1 (volume)
461* - `mfem`
462  - * BP1 (scalar mass operator)
463    * BP3 (scalar Laplace operator)
464* - `petsc`
465  - * BP1 (scalar mass operator)
466    * BP3 (scalar Laplace operator)
467    * Navier-Stokes
468* - `nek5000`
469  - * BP1 (scalar mass operator)
470    * BP3 (scalar Laplace operator)
471:::
472
473(v0-3)=
474
475## v0.3 (Sep 30, 2018)
476
477Notable features in this release include active/passive field interface, support for
478non-tensor bases, backend optimization, and improved Fortran interface. This release
479also focused on providing improved continuous integration, and many new tests with code
480coverage reports of about 90%. This release also provided a significant change to the
481public interface: a {ref}`CeedQFunction` can take any number of named input and output
482arguments while {ref}`CeedOperator` connects them to the actual data, which may be
483supplied explicitly to `CeedOperatorApply()` (active) or separately via
484`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
485of CeedQFunctions and composition of block solvers constructed using
486{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
487used in an optimized CPU backend. Although this is typically not visible to the user,
488it enables effective use of arbitrary-length SIMD while maintaining cache locality.
489This CPU backend also implements an algebraic factorization of tensor product gradients
490to perform fewer operations than standard application of interpolation and
491differentiation from nodes to quadrature points. This algebraic formulation
492automatically supports non-polynomial and non-interpolatory bases, thus is more general
493than the more common derivation in terms of Lagrange polynomials on the quadrature points.
494
495Backends available in this release:
496
497| CEED resource (`-ceed`) | Backend                                             |
498|-------------------------|-----------------------------------------------------|
499| `/cpu/self/blocked`     | Blocked reference implementation                    |
500| `/cpu/self/ref`         | Serial reference implementation                     |
501| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
502| `/cpu/occa`             | Serial OCCA kernels                                 |
503| `/gpu/occa`             | CUDA OCCA kernels                                   |
504| `/omp/occa`             | OpenMP OCCA kernels                                 |
505| `/ocl/occa`             | OpenCL OCCA kernels                                 |
506| `/gpu/magma`            | CUDA MAGMA kernels                                  |
507
508Examples available in this release:
509
510:::{list-table}
511:header-rows: 1
512:widths: auto
513* - User code
514  - Example
515* - `ceed`
516  - * ex1 (volume)
517* - `mfem`
518  - * BP1 (scalar mass operator)
519    * BP3 (scalar Laplace operator)
520* - `petsc`
521  - * BP1 (scalar mass operator)
522    * BP3 (scalar Laplace operator)
523* - `nek5000`
524  - * BP1 (scalar mass operator)
525    * BP3 (scalar Laplace operator)
526:::
527
528(v0-21)=
529
530## v0.21 (Sep 30, 2018)
531
532A MAGMA backend (which relies upon the
533[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
534release. This initial integration set up the framework of using MAGMA and provided the
535libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
536As any other backend, the MAGMA backend provides extended basic data structures for
537{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
538the fundamental CEED building blocks to work with the new data structures.
539In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
540but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
541internally, and thus seamlessly to the user, through the functions/methods that are
542provided to support them.
543
544Backends available in this release:
545
546| CEED resource (`-ceed`) | Backend                         |
547|-------------------------|---------------------------------|
548| `/cpu/self`             | Serial reference implementation |
549| `/cpu/occa`             | Serial OCCA kernels             |
550| `/gpu/occa`             | CUDA OCCA kernels               |
551| `/omp/occa`             | OpenMP OCCA kernels             |
552| `/ocl/occa`             | OpenCL OCCA kernels             |
553| `/gpu/magma`            | CUDA MAGMA kernels              |
554
555Examples available in this release:
556
557:::{list-table}
558:header-rows: 1
559:widths: auto
560* - User code
561  - Example
562* - `ceed`
563  - * ex1 (volume)
564* - `mfem`
565  - * BP1 (scalar mass operator)
566    * BP3 (scalar Laplace operator)
567* - `petsc`
568  - * BP1 (scalar mass operator)
569* - `nek5000`
570  - * BP1 (scalar mass operator)
571:::
572
573(v0-2)=
574
575## v0.2 (Mar 30, 2018)
576
577libCEED was made publicly available the first full CEED software distribution, release
578CEED 1.0. The distribution was made available using the Spack package manager to provide
579a common, easy-to-use build environment, where the user can build the CEED distribution
580with all dependencies. This release included a new Fortran interface for the library.
581This release also contained major improvements in the OCCA backend (including a new
582`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
583compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
584`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
585of the Laplace operator) was also added to this release.
586
587Backends available in this release:
588
589| CEED resource (`-ceed`) | Backend                         |
590|-------------------------|---------------------------------|
591| `/cpu/self`             | Serial reference implementation |
592| `/cpu/occa`             | Serial OCCA kernels             |
593| `/gpu/occa`             | CUDA OCCA kernels               |
594| `/omp/occa`             | OpenMP OCCA kernels             |
595| `/ocl/occa`             | OpenCL OCCA kernels             |
596
597Examples available in this release:
598
599:::{list-table}
600:header-rows: 1
601:widths: auto
602* - User code
603  - Example
604* - `ceed`
605  - * ex1 (volume)
606* - `mfem`
607  - * BP1 (scalar mass operator)
608    * BP3 (scalar Laplace operator)
609* - `petsc`
610  - * BP1 (scalar mass operator)
611* - `nek5000`
612  - * BP1 (scalar mass operator)
613:::
614
615(v0-1)=
616
617## v0.1 (Jan 3, 2018)
618
619Initial low-level API of the CEED project. The low-level API provides a set of Finite
620Elements kernels and components for writing new low-level kernels. Examples include:
621vector and sparse linear algebra, element matrix assembly over a batch of elements,
622partial assembly and action for efficient high-order operators like mass, diffusion,
623advection, etc. The main goal of the low-level API is to establish the basis for the
624high-level API. Also, identifying such low-level kernels and providing a reference
625implementation for them serves as the basis for specialized backend implementations.
626This release contained several backends: `/cpu/self`, and backends which rely upon the
627[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
628`/gpu/occa`, and `/omp/occa`.
629It also included several examples, in the `examples` folder:
630A standalone code that shows the usage of libCEED (with no external
631dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
632(with the application of the mass operator); and a `petsc` example to perform BP1
633(with the application of the mass operator).
634
635Backends available in this release:
636
637| CEED resource (`-ceed`) | Backend                         |
638|-------------------------|---------------------------------|
639| `/cpu/self`             | Serial reference implementation |
640| `/cpu/occa`             | Serial OCCA kernels             |
641| `/gpu/occa`             | CUDA OCCA kernels               |
642| `/omp/occa`             | OpenMP OCCA kernels             |
643
644Examples available in this release:
645
646| User code             | Example                           |
647|-----------------------|-----------------------------------|
648| `ceed`                | ex1 (scalar Laplace operator)     |
649| `mfem`                | BP1 (scalar mass operator)        |
650| `petsc`               | BP1 (scalar mass operator)        |
651```
652