xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 216bbcafa310ca9c8563aabbff60855a4edc792b)
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- Removed `CeedBasisSetNumQuadraturePoints` as redundant and bug-prone interface.
14
15### New features
16
17- Added {c:func}`CeedOperatorGetFieldByName` to access a specific `CeedOperatorField` by its name.
18- Update `/cpu/self/memcheck/*` backends to help verify `CeedVector` array access assumptions and `CeedQFunction` user output assumptions.
19- Update {c:func}`CeedOperatorLinearAssembleDiagonal` to provide default implementation that supports `CeedOperator` with multiple active bases.
20- Added Sycl backends `/gpu/sycl/ref` and `/gpu/sycl/shared`.
21- Added {c:func}`CeedBasisApplyAtPoints` for evaluation of values and derivatives at arbitrary points inside elements.
22- Added support for non-tensor $H(\text{curl})$ finite element spaces with {c:func}`CeedBasisCreateHcurl`.
23- 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)).
24- Added {c:func}`CeedOperatorLinearAssemblePointBlockDiagonalSymbolic` to create COO mapping for mapping out of {c:func}`CeedOperatorLinearAssemblePointBlockDiagonal`.
25- Added support for application codes which manage multiple {ref}`Ceed` objects, parallelized across OpenMP threads.
26
27### Examples
28
29#### {ref}`example-petsc-bps`
30
31- Requires PETSc version 3.19 or later.
32
33(v0-11)=
34
35## v0.11 (Dec 24, 2022)
36
37### Interface changes
38
39- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
40- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
41- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface.
42- Renamed `CEED_BASIS_COLLOCATED` to `CEED_BASIS_NONE` for clarity.
43Some 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.
44
45### New features
46
47- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
48- Improved support for $H(\text{div})$ bases.
49- Added `CeedInt_FMT` to support potential future use of larger integer sizes.
50- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`.
51- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
52Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
53This issue will be fixed in a future OCCA release.
54
55### Bugfix
56
57- Fix bug in setting device id for GPU backends.
58- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
59- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
60- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
61
62### Examples
63
64#### {ref}`example-petsc-navier-stokes`
65
66- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs.
67- Refactored to improve code reuse and modularity.
68- Support for primitive variables for more accurate boundary layers and all-speed flow.
69- Added $YZ\beta$ shock capturing scheme and Shock Tube example.
70- Added Channel example, with comparison to analytic solutions.
71- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations.
72- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions.
73- Added "freestream" boundary conditions based on HLLC Riemann solver.
74- Automated stabilization coefficients for different basis degree.
75
76#### {ref}`example-petsc-bps`
77
78- Support for convergence studies.
79
80### Maintainability
81
82- 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.
83- Enabled support for `p > 8` for `/gpu/*/shared` backends.
84- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
85- Improved test harness.
86
87(v0-10-1)=
88
89## v0.10.1 (Apr 11, 2022)
90
91### Interface changes
92
93- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
94
95### New features
96
97- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
98This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
99
100### Bugfix
101
102- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
103
104(v0-10)=
105
106## v0.10 (Mar 21, 2022)
107
108### Interface changes
109
110- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
111- 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`.
112- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
113- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
114- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
115- Warning added when compiling OCCA backend to alert users that this backend is experimental.
116- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
117- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
118- 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`.
119- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitialized arrays; require initialized data for {c:func}`CeedVectorGetArray`.
120- 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.
121- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retrieve user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
122- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
123- 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.
124- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
125- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
126
127### New features
128
129- `CeedScalar` can now be set as `float` or `double` at compile time.
130- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
131- 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.
132- Remove need to guard library headers in QFunction source for code generation backends.
133- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
134- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
135- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
136- Added support for non-tensor $H(\text{div})$ elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
137- 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.
138- Added support for element matrix assembly in GPU backends.
139
140### Maintainability
141
142- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
143- `Include-what-you-use` makefile target added as `make iwyu`.
144- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
145- Put GPU JiTed kernel source code into separate files.
146- 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.
147
148(v0-9)=
149
150## v0.9 (Jul 6, 2021)
151
152### Interface changes
153
154- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
155
156### New features
157
158- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
159- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
160- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
161- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
162
163### Performance improvements
164
165### Examples
166
167- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
168- Fluid mechanics example adds GPU support and improves modularity.
169
170### Deprecated backends
171
172- 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.
173
174(v0-8)=
175
176## v0.8 (Mar 31, 2021)
177
178### Interface changes
179
180- Error handling improved to include enumerated error codes for C interface return values.
181- 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.
182
183### New features
184
185- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
186- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
187- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
188
189### Performance improvements
190
191- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
192- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
193
194### Examples
195
196- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
197- {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.
198- {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.
199- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
200
201(v0-7)=
202
203## v0.7 (Sep 29, 2020)
204
205### Interface changes
206
207- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
208  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
209  These changes improve support for mixed finite element methods.
210- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
211- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
212- 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.
213- 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.
214- 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.
215- 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.
216  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`.
217- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
218- 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.
219
220### New features
221
222- New HIP backend: `/gpu/hip/ref`.
223- CeedQFunction support for user `CUfunction`s in some backends
224
225### Performance improvements
226
227- OCCA backend rebuilt to facilitate future performance enhancements.
228- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
229
230### Examples
231
232- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
233
234### Deprecated backends
235
236- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
237
238(v0-6)=
239
240## v0.6 (Mar 29, 2020)
241
242libCEED v0.6 contains numerous new features and examples, as well as expanded
243documentation in [this new website](https://libceed.org).
244
245### New features
246
247- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
248  1-1 correspondence with the C interface, plus some convenience features.  For instance,
249  data stored in the {cpp:type}`CeedVector` structure are available without copy as
250  {py:class}`numpy.ndarray`.  Short tutorials are provided in
251  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
252- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
253  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
254  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
255  preconditioning ingredients and are used in the libCEED's multigrid examples.
256- The inverse of separable operators can be obtained using
257  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
258  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
259  especially for Laplacians and related operators.
260- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
261  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
262- Make public accessors for various attributes to facilitate writing composable code.
263- New backend: `/cpu/self/memcheck/serial`.
264- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
265  backends.  (Single source is coming soon for OCCA backends.)
266- Fix some missing edge cases in CUDA backend.
267
268### Performance Improvements
269
270- MAGMA backend performance optimization and non-tensor bases.
271- No-copy optimization in {c:func}`CeedOperatorApply`.
272
273### Interface changes
274
275- Replace {code}`CeedElemRestrictionCreateIdentity` and
276  {code}`CeedElemRestrictionCreateBlocked` with more flexible
277  {c:func}`CeedElemRestrictionCreateStrided` and
278  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
279- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
280- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
281  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
282  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
283  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
284  {c:func}`CeedElemRestrictionApply`.
285
286### Examples
287
288libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
289Notable additions include:
290
291- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
292  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
293
294- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
295  of domains (like the cube and sphere) by direct integration on a surface mesh;
296  demonstrates geometric dimension different from topological dimension.
297
298- PETSc {ref}`example-petsc-bps`:
299
300  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
301  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
302    and transparent CUDA support.
303  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
304    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
305    dimension different from topological dimension.
306
307- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
308  solver with algebraic multigrid coarse solve.
309
310- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
311  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
312  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
313  quasi-2D computational domain support.
314
315- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
316  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
317  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
318
319(v0-5)=
320
321## v0.5 (Sep 18, 2019)
322
323For this release, several improvements were made. Two new CUDA backends were added to
324the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
325performance using single-source {ref}`CeedQFunction`. From this release, users
326can define Q-Functions in a single source code independently of the targeted backend
327with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
328compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
329declaration, the {ref}`CeedQFunction` API has undergone a slight change:
330the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
331requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
332`QFunctionField` with eval mode `CEED EVAL GRAD`.
333
334Additionally, new CPU backends
335were included in this release, such as the `/cpu/self/opt/*` backends (which are
336written in pure C and use partial **E-vectors** to improve performance) and the
337`/cpu/self/ref/memcheck` backend (which relies upon the
338[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
339{ref}`CeedQFunction` have no undefined values).
340This release also included various performance improvements, bug fixes, new examples,
341and improved tests. Among these improvements, vectorized instructions for
342{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
343instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
344identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
345to include unstructured meshes handling were. For this expansion, the prior version of
346the PETSc BPs, which only included data associated with structured geometries, were
347renamed `bpsraw`, and the new version of the BPs, which can handle data associated
348with any unstructured geometry, were called `bps`. Additionally, other benchmark
349problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
350and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
351as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
352examples. Furthermoew, another standalone libCEED example, called `ex2`, which
353computes the surface area of a given mesh was added to this release.
354
355Backends available in this release:
356
357| CEED resource (`-ceed`)  | Backend                                             |
358|--------------------------|-----------------------------------------------------|
359| `/cpu/self/ref/serial`   | Serial reference implementation                     |
360| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
361| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
362| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
363| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
364| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
365| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
366| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
367| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
368| `/cpu/occa`              | Serial OCCA kernels                                 |
369| `/gpu/occa`              | CUDA OCCA kernels                                   |
370| `/omp/occa`              | OpenMP OCCA kernels                                 |
371| `/ocl/occa`              | OpenCL OCCA kernels                                 |
372| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
373| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
374| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
375| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
376| `/gpu/magma`             | CUDA MAGMA kernels                                  |
377
378Examples available in this release:
379
380:::{list-table}
381:header-rows: 1
382:widths: auto
383* - User code
384  - Example
385* - `ceed`
386  - * ex1 (volume)
387    * ex2 (surface)
388* - `mfem`
389  - * BP1 (scalar mass operator)
390    * BP3 (scalar Laplace operator)
391* - `petsc`
392  - * BP1 (scalar mass operator)
393    * BP2 (vector mass operator)
394    * BP3 (scalar Laplace operator)
395    * BP4 (vector Laplace operator)
396    * BP5 (collocated scalar Laplace operator)
397    * BP6 (collocated vector Laplace operator)
398    * Navier-Stokes
399* - `nek5000`
400  - * BP1 (scalar mass operator)
401    * BP3 (scalar Laplace operator)
402:::
403
404(v0-4)=
405
406## v0.4 (Apr 1, 2019)
407
408libCEED v0.4 was made again publicly available in the second full CEED software
409distribution, release CEED 2.0. This release contained notable features, such as
410four new CPU backends, two new GPU backends, CPU backend optimizations, initial
411support for operator composition, performance benchmarking, and a Navier-Stokes demo.
412The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
413backends process one element at a time and are intended for meshes with a smaller number
414of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
415eight interlaced elements and are intended for meshes with higher numbers of elements.
416The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
417performance. The `/cpu/self/xsmm/*` backends rely upon the
418[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
419performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
420The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
421performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
422parallelization approach, where each thread treats a finite element. Using just in time
423compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
424backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
425achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
426deteriorates very quickly when threads run out of registers.
427
428A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
429examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
430This example solves the time-dependent Navier-Stokes equations of compressible gas
431dynamics in a static Eulerian three-dimensional frame, using structured high-order
432finite/spectral element spatial discretizations and explicit high-order time-stepping
433(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
434so that the pointwise physics (defined at quadrature points) is separated from the
435parallelization and meshing concerns.
436
437Backends available in this release:
438
439| CEED resource (`-ceed`)  | Backend                                             |
440|--------------------------|-----------------------------------------------------|
441| `/cpu/self/ref/serial`   | Serial reference implementation                     |
442| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
443| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
444| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
445| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
446| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
447| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
448| `/cpu/occa`              | Serial OCCA kernels                                 |
449| `/gpu/occa`              | CUDA OCCA kernels                                   |
450| `/omp/occa`              | OpenMP OCCA kernels                                 |
451| `/ocl/occa`              | OpenCL OCCA kernels                                 |
452| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
453| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
454| `/gpu/magma`             | CUDA MAGMA kernels                                  |
455
456Examples available in this release:
457
458:::{list-table}
459:header-rows: 1
460:widths: auto
461* - User code
462  - Example
463* - `ceed`
464  - * ex1 (volume)
465* - `mfem`
466  - * BP1 (scalar mass operator)
467    * BP3 (scalar Laplace operator)
468* - `petsc`
469  - * BP1 (scalar mass operator)
470    * BP3 (scalar Laplace operator)
471    * Navier-Stokes
472* - `nek5000`
473  - * BP1 (scalar mass operator)
474    * BP3 (scalar Laplace operator)
475:::
476
477(v0-3)=
478
479## v0.3 (Sep 30, 2018)
480
481Notable features in this release include active/passive field interface, support for
482non-tensor bases, backend optimization, and improved Fortran interface. This release
483also focused on providing improved continuous integration, and many new tests with code
484coverage reports of about 90%. This release also provided a significant change to the
485public interface: a {ref}`CeedQFunction` can take any number of named input and output
486arguments while {ref}`CeedOperator` connects them to the actual data, which may be
487supplied explicitly to `CeedOperatorApply()` (active) or separately via
488`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
489of CeedQFunctions and composition of block solvers constructed using
490{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
491used in an optimized CPU backend. Although this is typically not visible to the user,
492it enables effective use of arbitrary-length SIMD while maintaining cache locality.
493This CPU backend also implements an algebraic factorization of tensor product gradients
494to perform fewer operations than standard application of interpolation and
495differentiation from nodes to quadrature points. This algebraic formulation
496automatically supports non-polynomial and non-interpolatory bases, thus is more general
497than the more common derivation in terms of Lagrange polynomials on the quadrature points.
498
499Backends available in this release:
500
501| CEED resource (`-ceed`) | Backend                                             |
502|-------------------------|-----------------------------------------------------|
503| `/cpu/self/blocked`     | Blocked reference implementation                    |
504| `/cpu/self/ref`         | Serial reference implementation                     |
505| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
506| `/cpu/occa`             | Serial OCCA kernels                                 |
507| `/gpu/occa`             | CUDA OCCA kernels                                   |
508| `/omp/occa`             | OpenMP OCCA kernels                                 |
509| `/ocl/occa`             | OpenCL OCCA kernels                                 |
510| `/gpu/magma`            | CUDA MAGMA kernels                                  |
511
512Examples available in this release:
513
514:::{list-table}
515:header-rows: 1
516:widths: auto
517* - User code
518  - Example
519* - `ceed`
520  - * ex1 (volume)
521* - `mfem`
522  - * BP1 (scalar mass operator)
523    * BP3 (scalar Laplace operator)
524* - `petsc`
525  - * BP1 (scalar mass operator)
526    * BP3 (scalar Laplace operator)
527* - `nek5000`
528  - * BP1 (scalar mass operator)
529    * BP3 (scalar Laplace operator)
530:::
531
532(v0-21)=
533
534## v0.21 (Sep 30, 2018)
535
536A MAGMA backend (which relies upon the
537[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
538release. This initial integration set up the framework of using MAGMA and provided the
539libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
540As any other backend, the MAGMA backend provides extended basic data structures for
541{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
542the fundamental CEED building blocks to work with the new data structures.
543In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
544but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
545internally, and thus seamlessly to the user, through the functions/methods that are
546provided to support them.
547
548Backends available in this release:
549
550| CEED resource (`-ceed`) | Backend                         |
551|-------------------------|---------------------------------|
552| `/cpu/self`             | Serial reference implementation |
553| `/cpu/occa`             | Serial OCCA kernels             |
554| `/gpu/occa`             | CUDA OCCA kernels               |
555| `/omp/occa`             | OpenMP OCCA kernels             |
556| `/ocl/occa`             | OpenCL OCCA kernels             |
557| `/gpu/magma`            | CUDA MAGMA kernels              |
558
559Examples available in this release:
560
561:::{list-table}
562:header-rows: 1
563:widths: auto
564* - User code
565  - Example
566* - `ceed`
567  - * ex1 (volume)
568* - `mfem`
569  - * BP1 (scalar mass operator)
570    * BP3 (scalar Laplace operator)
571* - `petsc`
572  - * BP1 (scalar mass operator)
573* - `nek5000`
574  - * BP1 (scalar mass operator)
575:::
576
577(v0-2)=
578
579## v0.2 (Mar 30, 2018)
580
581libCEED was made publicly available the first full CEED software distribution, release
582CEED 1.0. The distribution was made available using the Spack package manager to provide
583a common, easy-to-use build environment, where the user can build the CEED distribution
584with all dependencies. This release included a new Fortran interface for the library.
585This release also contained major improvements in the OCCA backend (including a new
586`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
587compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
588`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
589of the Laplace operator) was also added to this release.
590
591Backends available in this release:
592
593| CEED resource (`-ceed`) | Backend                         |
594|-------------------------|---------------------------------|
595| `/cpu/self`             | Serial reference implementation |
596| `/cpu/occa`             | Serial OCCA kernels             |
597| `/gpu/occa`             | CUDA OCCA kernels               |
598| `/omp/occa`             | OpenMP OCCA kernels             |
599| `/ocl/occa`             | OpenCL OCCA kernels             |
600
601Examples available in this release:
602
603:::{list-table}
604:header-rows: 1
605:widths: auto
606* - User code
607  - Example
608* - `ceed`
609  - * ex1 (volume)
610* - `mfem`
611  - * BP1 (scalar mass operator)
612    * BP3 (scalar Laplace operator)
613* - `petsc`
614  - * BP1 (scalar mass operator)
615* - `nek5000`
616  - * BP1 (scalar mass operator)
617:::
618
619(v0-1)=
620
621## v0.1 (Jan 3, 2018)
622
623Initial low-level API of the CEED project. The low-level API provides a set of Finite
624Elements kernels and components for writing new low-level kernels. Examples include:
625vector and sparse linear algebra, element matrix assembly over a batch of elements,
626partial assembly and action for efficient high-order operators like mass, diffusion,
627advection, etc. The main goal of the low-level API is to establish the basis for the
628high-level API. Also, identifying such low-level kernels and providing a reference
629implementation for them serves as the basis for specialized backend implementations.
630This release contained several backends: `/cpu/self`, and backends which rely upon the
631[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
632`/gpu/occa`, and `/omp/occa`.
633It also included several examples, in the `examples` folder:
634A standalone code that shows the usage of libCEED (with no external
635dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
636(with the application of the mass operator); and a `petsc` example to perform BP1
637(with the application of the mass operator).
638
639Backends available in this release:
640
641| CEED resource (`-ceed`) | Backend                         |
642|-------------------------|---------------------------------|
643| `/cpu/self`             | Serial reference implementation |
644| `/cpu/occa`             | Serial OCCA kernels             |
645| `/gpu/occa`             | CUDA OCCA kernels               |
646| `/omp/occa`             | OpenMP OCCA kernels             |
647
648Examples available in this release:
649
650| User code             | Example                           |
651|-----------------------|-----------------------------------|
652| `ceed`                | ex1 (scalar Laplace operator)     |
653| `mfem`                | BP1 (scalar mass operator)        |
654| `petsc`               | BP1 (scalar mass operator)        |
655```
656