xref: /libCEED/doc/sphinx/source/releasenotes.md (revision a0154adecfab8547cdc0febbbf40ac009dbe9d1d)
1# Changes/Release Notes
2
3On this page we provide a summary of the main API changes, new features and examples
4for each release of libCEED.
5
6(main)=
7
8## Current `main` branch
9
10### Interface changes
11
12- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
13
14### Bugfix
15
16- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
17
18(v0-10)=
19
20## v0.10 (Mar 21, 2022)
21
22### Interface changes
23
24- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
25- 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`.
26- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
27- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
28- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
29- Warning added when compiling OCCA backend to alert users that this backend is experimental.
30- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
31- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
32- 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`.
33- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
34- 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.
35- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
36- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
37- 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.
38- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
39- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
40
41### New features
42
43- `CeedScalar` can now be set as `float` or `double` at compile time.
44- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
45- 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.
46- Remove need to guard library headers in QFunction source for code generation backends.
47- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
48- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
49- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
50- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
51- 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.
52- Added support for element matrix assembly in GPU backends.
53
54### Maintainability
55
56- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
57- `Include-what-you-use` makefile target added as `make iwyu`.
58- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
59- Put GPU JiTed kernel source code into separate files.
60- 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.
61
62(v0-9)=
63
64## v0.9 (Jul 6, 2021)
65
66### Interface changes
67
68- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
69
70### New features
71
72- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
73- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
74- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
75- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
76
77### Performance improvements
78
79### Examples
80
81- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
82- Fluid mechanics example adds GPU support and improves modularity.
83
84### Deprecated backends
85
86- 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.
87
88(v0-8)=
89
90## v0.8 (Mar 31, 2021)
91
92### Interface changes
93
94- Error handling improved to include enumerated error codes for C interface return values.
95- 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.
96
97### New features
98
99- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
100- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
101- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
102
103### Performance improvements
104
105- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
106- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
107
108### Examples
109
110- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
111- {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.
112- {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.
113- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
114
115(v0-7)=
116
117## v0.7 (Sep 29, 2020)
118
119### Interface changes
120
121- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
122  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
123  These changes improve support for mixed finite element methods.
124- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
125- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
126- 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.
127- 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.
128- 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.
129- 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.
130  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`.
131- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
132- 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.
133
134### New features
135
136- New HIP backend: `/gpu/hip/ref`.
137- CeedQFunction support for user `CUfunction`s in some backends
138
139### Performance improvements
140
141- OCCA backend rebuilt to facilitate future performance enhancements.
142- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
143
144### Examples
145
146- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
147
148### Deprecated backends
149
150- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
151
152(v0-6)=
153
154## v0.6 (Mar 29, 2020)
155
156libCEED v0.6 contains numerous new features and examples, as well as expanded
157documentation in [this new website](https://libceed.org).
158
159### New features
160
161- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
162  1-1 correspondence with the C interface, plus some convenience features.  For instance,
163  data stored in the {cpp:type}`CeedVector` structure are available without copy as
164  {py:class}`numpy.ndarray`.  Short tutorials are provided in
165  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
166- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
167  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
168  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
169  preconditioning ingredients and are used in the libCEED's multigrid examples.
170- The inverse of separable operators can be obtained using
171  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
172  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
173  especially for Laplacians and related operators.
174- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
175  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
176- Make public accessors for various attributes to facilitate writing composable code.
177- New backend: `/cpu/self/memcheck/serial`.
178- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
179  backends.  (Single source is coming soon for OCCA backends.)
180- Fix some missing edge cases in CUDA backend.
181
182### Performance Improvements
183
184- MAGMA backend performance optimization and non-tensor bases.
185- No-copy optimization in {c:func}`CeedOperatorApply`.
186
187### Interface changes
188
189- Replace {code}`CeedElemRestrictionCreateIdentity` and
190  {code}`CeedElemRestrictionCreateBlocked` with more flexible
191  {c:func}`CeedElemRestrictionCreateStrided` and
192  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
193- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
194- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
195  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
196  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
197  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
198  {c:func}`CeedElemRestrictionApply`.
199
200### Examples
201
202libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
203Notable additions include:
204
205- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
206  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
207
208- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
209  of domains (like the cube and sphere) by direct integration on a surface mesh;
210  demonstrates geometric dimension different from topological dimension.
211
212- PETSc {ref}`example-petsc-bps`:
213
214  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
215  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
216    and transparent CUDA support.
217  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
218    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
219    dimension different from topological dimension.
220
221- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
222  solver with algebraic multigrid coarse solve.
223
224- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
225  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
226  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
227  quasi-2D computational domain support.
228
229- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
230  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
231  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
232
233(v0-5)=
234
235## v0.5 (Sep 18, 2019)
236
237For this release, several improvements were made. Two new CUDA backends were added to
238the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
239performance using single-source {ref}`CeedQFunction`. From this release, users
240can define Q-Functions in a single source code independently of the targeted backend
241with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
242compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
243declaration, the {ref}`CeedQFunction` API has undergone a slight change:
244the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
245requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
246`QFunctionField` with eval mode `CEED EVAL GRAD`.
247
248Additionally, new CPU backends
249were included in this release, such as the `/cpu/self/opt/*` backends (which are
250written in pure C and use partial **E-vectors** to improve performance) and the
251`/cpu/self/ref/memcheck` backend (which relies upon the
252[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
253{ref}`CeedQFunction` have no undefined values).
254This release also included various performance improvements, bug fixes, new examples,
255and improved tests. Among these improvements, vectorized instructions for
256{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
257instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
258identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
259to include unstructured meshes handling were. For this expansion, the prior version of
260the PETSc BPs, which only included data associated with structured geometries, were
261renamed `bpsraw`, and the new version of the BPs, which can handle data associated
262with any unstructured geometry, were called `bps`. Additionally, other benchmark
263problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
264and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
265as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
266examples. Furthermoew, another standalone libCEED example, called `ex2`, which
267computes the surface area of a given mesh was added to this release.
268
269Backends available in this release:
270
271| CEED resource (`-ceed`)  | Backend                                             |
272|--------------------------|-----------------------------------------------------|
273| `/cpu/self/ref/serial`   | Serial reference implementation                     |
274| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
275| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
276| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
277| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
278| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
279| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
280| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
281| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
282| `/cpu/occa`              | Serial OCCA kernels                                 |
283| `/gpu/occa`              | CUDA OCCA kernels                                   |
284| `/omp/occa`              | OpenMP OCCA kernels                                 |
285| `/ocl/occa`              | OpenCL OCCA kernels                                 |
286| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
287| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
288| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
289| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
290| `/gpu/magma`             | CUDA MAGMA kernels                                  |
291
292Examples available in this release:
293
294:::{list-table}
295:header-rows: 1
296:widths: auto
297* - User code
298  - Example
299* - `ceed`
300  - * ex1 (volume)
301    * ex2 (surface)
302* - `mfem`
303  - * BP1 (scalar mass operator)
304    * BP3 (scalar Laplace operator)
305* - `petsc`
306  - * BP1 (scalar mass operator)
307    * BP2 (vector mass operator)
308    * BP3 (scalar Laplace operator)
309    * BP4 (vector Laplace operator)
310    * BP5 (collocated scalar Laplace operator)
311    * BP6 (collocated vector Laplace operator)
312    * Navier-Stokes
313* - `nek5000`
314  - * BP1 (scalar mass operator)
315    * BP3 (scalar Laplace operator)
316:::
317
318(v0-4)=
319
320## v0.4 (Apr 1, 2019)
321
322libCEED v0.4 was made again publicly available in the second full CEED software
323distribution, release CEED 2.0. This release contained notable features, such as
324four new CPU backends, two new GPU backends, CPU backend optimizations, initial
325support for operator composition, performance benchmarking, and a Navier-Stokes demo.
326The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
327backends process one element at a time and are intended for meshes with a smaller number
328of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
329eight interlaced elements and are intended for meshes with higher numbers of elements.
330The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
331performance. The `/cpu/self/xsmm/*` backends rely upon the
332[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
333performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
334The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
335performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
336parallelization approach, where each thread treats a finite element. Using just in time
337compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
338backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
339achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
340deteriorates very quickly when threads run out of registers.
341
342A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
343examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
344This example solves the time-dependent Navier-Stokes equations of compressible gas
345dynamics in a static Eulerian three-dimensional frame, using structured high-order
346finite/spectral element spatial discretizations and explicit high-order time-stepping
347(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
348so that the pointwise physics (defined at quadrature points) is separated from the
349parallelization and meshing concerns.
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/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
358| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
359| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
360| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
361| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
362| `/cpu/occa`              | Serial OCCA kernels                                 |
363| `/gpu/occa`              | CUDA OCCA kernels                                   |
364| `/omp/occa`              | OpenMP OCCA kernels                                 |
365| `/ocl/occa`              | OpenCL OCCA kernels                                 |
366| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
367| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
368| `/gpu/magma`             | CUDA MAGMA kernels                                  |
369
370Examples available in this release:
371
372:::{list-table}
373:header-rows: 1
374:widths: auto
375* - User code
376  - Example
377* - `ceed`
378  - * ex1 (volume)
379* - `mfem`
380  - * BP1 (scalar mass operator)
381    * BP3 (scalar Laplace operator)
382* - `petsc`
383  - * BP1 (scalar mass operator)
384    * BP3 (scalar Laplace operator)
385    * Navier-Stokes
386* - `nek5000`
387  - * BP1 (scalar mass operator)
388    * BP3 (scalar Laplace operator)
389:::
390
391(v0-3)=
392
393## v0.3 (Sep 30, 2018)
394
395Notable features in this release include active/passive field interface, support for
396non-tensor bases, backend optimization, and improved Fortran interface. This release
397also focused on providing improved continuous integration, and many new tests with code
398coverage reports of about 90%. This release also provided a significant change to the
399public interface: a {ref}`CeedQFunction` can take any number of named input and output
400arguments while {ref}`CeedOperator` connects them to the actual data, which may be
401supplied explicitly to `CeedOperatorApply()` (active) or separately via
402`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
403of CeedQFunctions and composition of block solvers constructed using
404{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
405used in an optimized CPU backend. Although this is typically not visible to the user,
406it enables effective use of arbitrary-length SIMD while maintaining cache locality.
407This CPU backend also implements an algebraic factorization of tensor product gradients
408to perform fewer operations than standard application of interpolation and
409differentiation from nodes to quadrature points. This algebraic formulation
410automatically supports non-polynomial and non-interpolatory bases, thus is more general
411than the more common derivation in terms of Lagrange polynomials on the quadrature points.
412
413Backends available in this release:
414
415| CEED resource (`-ceed`) | Backend                                             |
416|-------------------------|-----------------------------------------------------|
417| `/cpu/self/blocked`     | Blocked reference implementation                    |
418| `/cpu/self/ref`         | Serial reference implementation                     |
419| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
420| `/cpu/occa`             | Serial OCCA kernels                                 |
421| `/gpu/occa`             | CUDA OCCA kernels                                   |
422| `/omp/occa`             | OpenMP OCCA kernels                                 |
423| `/ocl/occa`             | OpenCL OCCA kernels                                 |
424| `/gpu/magma`            | CUDA MAGMA kernels                                  |
425
426Examples available in this release:
427
428:::{list-table}
429:header-rows: 1
430:widths: auto
431* - User code
432  - Example
433* - `ceed`
434  - * ex1 (volume)
435* - `mfem`
436  - * BP1 (scalar mass operator)
437    * BP3 (scalar Laplace operator)
438* - `petsc`
439  - * BP1 (scalar mass operator)
440    * BP3 (scalar Laplace operator)
441* - `nek5000`
442  - * BP1 (scalar mass operator)
443    * BP3 (scalar Laplace operator)
444:::
445
446(v0-21)=
447
448## v0.21 (Sep 30, 2018)
449
450A MAGMA backend (which relies upon the
451[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
452release. This initial integration set up the framework of using MAGMA and provided the
453libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
454As any other backend, the MAGMA backend provides extended basic data structures for
455{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
456the fundamental CEED building blocks to work with the new data structures.
457In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
458but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
459internally, and thus seamlessly to the user, through the functions/methods that are
460provided to support them.
461
462Backends available in this release:
463
464| CEED resource (`-ceed`) | Backend                         |
465|-------------------------|---------------------------------|
466| `/cpu/self`             | Serial reference implementation |
467| `/cpu/occa`             | Serial OCCA kernels             |
468| `/gpu/occa`             | CUDA OCCA kernels               |
469| `/omp/occa`             | OpenMP OCCA kernels             |
470| `/ocl/occa`             | OpenCL OCCA kernels             |
471| `/gpu/magma`            | CUDA MAGMA kernels              |
472
473Examples available in this release:
474
475:::{list-table}
476:header-rows: 1
477:widths: auto
478* - User code
479  - Example
480* - `ceed`
481  - * ex1 (volume)
482* - `mfem`
483  - * BP1 (scalar mass operator)
484    * BP3 (scalar Laplace operator)
485* - `petsc`
486  - * BP1 (scalar mass operator)
487* - `nek5000`
488  - * BP1 (scalar mass operator)
489:::
490
491(v0-2)=
492
493## v0.2 (Mar 30, 2018)
494
495libCEED was made publicly available the first full CEED software distribution, release
496CEED 1.0. The distribution was made available using the Spack package manager to provide
497a common, easy-to-use build environment, where the user can build the CEED distribution
498with all dependencies. This release included a new Fortran interface for the library.
499This release also contained major improvements in the OCCA backend (including a new
500`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
501compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
502`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
503of the Laplace operator) was also added to this release.
504
505Backends available in this release:
506
507| CEED resource (`-ceed`) | Backend                         |
508|-------------------------|---------------------------------|
509| `/cpu/self`             | Serial reference implementation |
510| `/cpu/occa`             | Serial OCCA kernels             |
511| `/gpu/occa`             | CUDA OCCA kernels               |
512| `/omp/occa`             | OpenMP OCCA kernels             |
513| `/ocl/occa`             | OpenCL OCCA kernels             |
514
515Examples available in this release:
516
517:::{list-table}
518:header-rows: 1
519:widths: auto
520* - User code
521  - Example
522* - `ceed`
523  - * ex1 (volume)
524* - `mfem`
525  - * BP1 (scalar mass operator)
526    * BP3 (scalar Laplace operator)
527* - `petsc`
528  - * BP1 (scalar mass operator)
529* - `nek5000`
530  - * BP1 (scalar mass operator)
531:::
532
533(v0-1)=
534
535## v0.1 (Jan 3, 2018)
536
537Initial low-level API of the CEED project. The low-level API provides a set of Finite
538Elements kernels and components for writing new low-level kernels. Examples include:
539vector and sparse linear algebra, element matrix assembly over a batch of elements,
540partial assembly and action for efficient high-order operators like mass, diffusion,
541advection, etc. The main goal of the low-level API is to establish the basis for the
542high-level API. Also, identifying such low-level kernels and providing a reference
543implementation for them serves as the basis for specialized backend implementations.
544This release contained several backends: `/cpu/self`, and backends which rely upon the
545[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
546`/gpu/occa`, and `/omp/occa`.
547It also included several examples, in the `examples` folder:
548A standalone code that shows the usage of libCEED (with no external
549dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
550(with the application of the mass operator); and a `petsc` example to perform BP1
551(with the application of the mass operator).
552
553Backends available in this release:
554
555| CEED resource (`-ceed`) | Backend                         |
556|-------------------------|---------------------------------|
557| `/cpu/self`             | Serial reference implementation |
558| `/cpu/occa`             | Serial OCCA kernels             |
559| `/gpu/occa`             | CUDA OCCA kernels               |
560| `/omp/occa`             | OpenMP OCCA kernels             |
561
562Examples available in this release:
563
564| User code             | Example                           |
565|-----------------------|-----------------------------------|
566| `ceed`                | ex1 (scalar Laplace operator)     |
567| `mfem`                | BP1 (scalar mass operator)        |
568| `petsc`               | BP1 (scalar mass operator)        |
569```
570