xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 709403c1d8a3197f71934fa05519eedc9835a633)
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
22### Examples
23
24#### {ref}`example-petsc-bps`
25
26- Requires PETSc version 3.19 or later
27
28(v0-11)=
29
30## v0.11 (Dec 24, 2022)
31
32### Interface changes
33
34- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
35- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
36- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface.
37
38### New features
39
40- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
41- Improved support for $H(\text{div})$ bases.
42- Added `CeedInt_FMT` to support potential future use of larger integer sizes.
43- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`.
44- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
45Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
46This issue will be fixed in a future OCCA release.
47
48### Bugfix
49
50- Fix bug in setting device id for GPU backends.
51- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
52- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
53- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
54
55### Examples
56
57#### {ref}`example-petsc-navier-stokes`
58
59- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs.
60- Refactored to improve code reuse and modularity.
61- Support for primitive variables for more accurate boundary layers and all-speed flow.
62- Added $YZ\beta$ shock capturing scheme and Shock Tube example.
63- Added Channel example, with comparison to analytic solutions.
64- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations.
65- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions.
66- Added "freestream" boundary conditions based on HLLC Riemann solver.
67- Automated stabilization coefficients for different basis degree.
68
69#### {ref}`example-petsc-bps`
70
71- Support for convergence studies.
72
73### Maintainability
74
75- 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.
76- Enabled support for `p > 8` for `/gpu/*/shared` backends.
77- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
78- Improved test harness.
79
80(v0-10-1)=
81
82## v0.10.1 (Apr 11, 2022)
83
84### Interface changes
85
86- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
87
88### New features
89
90- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
91This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
92
93### Bugfix
94
95- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
96
97(v0-10)=
98
99## v0.10 (Mar 21, 2022)
100
101### Interface changes
102
103- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
104- 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`.
105- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
106- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
107- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
108- Warning added when compiling OCCA backend to alert users that this backend is experimental.
109- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
110- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
111- 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`.
112- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
113- 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.
114- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
115- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
116- 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.
117- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
118- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
119
120### New features
121
122- `CeedScalar` can now be set as `float` or `double` at compile time.
123- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
124- 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.
125- Remove need to guard library headers in QFunction source for code generation backends.
126- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
127- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
128- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
129- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
130- 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.
131- Added support for element matrix assembly in GPU backends.
132
133### Maintainability
134
135- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
136- `Include-what-you-use` makefile target added as `make iwyu`.
137- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
138- Put GPU JiTed kernel source code into separate files.
139- 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.
140
141(v0-9)=
142
143## v0.9 (Jul 6, 2021)
144
145### Interface changes
146
147- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
148
149### New features
150
151- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
152- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
153- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
154- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
155
156### Performance improvements
157
158### Examples
159
160- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
161- Fluid mechanics example adds GPU support and improves modularity.
162
163### Deprecated backends
164
165- 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.
166
167(v0-8)=
168
169## v0.8 (Mar 31, 2021)
170
171### Interface changes
172
173- Error handling improved to include enumerated error codes for C interface return values.
174- 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.
175
176### New features
177
178- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
179- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
180- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
181
182### Performance improvements
183
184- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
185- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
186
187### Examples
188
189- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
190- {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.
191- {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.
192- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
193
194(v0-7)=
195
196## v0.7 (Sep 29, 2020)
197
198### Interface changes
199
200- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
201  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
202  These changes improve support for mixed finite element methods.
203- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
204- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
205- 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.
206- 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.
207- 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.
208- 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.
209  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`.
210- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
211- 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.
212
213### New features
214
215- New HIP backend: `/gpu/hip/ref`.
216- CeedQFunction support for user `CUfunction`s in some backends
217
218### Performance improvements
219
220- OCCA backend rebuilt to facilitate future performance enhancements.
221- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
222
223### Examples
224
225- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
226
227### Deprecated backends
228
229- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
230
231(v0-6)=
232
233## v0.6 (Mar 29, 2020)
234
235libCEED v0.6 contains numerous new features and examples, as well as expanded
236documentation in [this new website](https://libceed.org).
237
238### New features
239
240- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
241  1-1 correspondence with the C interface, plus some convenience features.  For instance,
242  data stored in the {cpp:type}`CeedVector` structure are available without copy as
243  {py:class}`numpy.ndarray`.  Short tutorials are provided in
244  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
245- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
246  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
247  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
248  preconditioning ingredients and are used in the libCEED's multigrid examples.
249- The inverse of separable operators can be obtained using
250  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
251  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
252  especially for Laplacians and related operators.
253- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
254  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
255- Make public accessors for various attributes to facilitate writing composable code.
256- New backend: `/cpu/self/memcheck/serial`.
257- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
258  backends.  (Single source is coming soon for OCCA backends.)
259- Fix some missing edge cases in CUDA backend.
260
261### Performance Improvements
262
263- MAGMA backend performance optimization and non-tensor bases.
264- No-copy optimization in {c:func}`CeedOperatorApply`.
265
266### Interface changes
267
268- Replace {code}`CeedElemRestrictionCreateIdentity` and
269  {code}`CeedElemRestrictionCreateBlocked` with more flexible
270  {c:func}`CeedElemRestrictionCreateStrided` and
271  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
272- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
273- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
274  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
275  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
276  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
277  {c:func}`CeedElemRestrictionApply`.
278
279### Examples
280
281libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
282Notable additions include:
283
284- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
285  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
286
287- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
288  of domains (like the cube and sphere) by direct integration on a surface mesh;
289  demonstrates geometric dimension different from topological dimension.
290
291- PETSc {ref}`example-petsc-bps`:
292
293  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
294  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
295    and transparent CUDA support.
296  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
297    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
298    dimension different from topological dimension.
299
300- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
301  solver with algebraic multigrid coarse solve.
302
303- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
304  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
305  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
306  quasi-2D computational domain support.
307
308- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
309  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
310  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
311
312(v0-5)=
313
314## v0.5 (Sep 18, 2019)
315
316For this release, several improvements were made. Two new CUDA backends were added to
317the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
318performance using single-source {ref}`CeedQFunction`. From this release, users
319can define Q-Functions in a single source code independently of the targeted backend
320with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
321compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
322declaration, the {ref}`CeedQFunction` API has undergone a slight change:
323the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
324requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
325`QFunctionField` with eval mode `CEED EVAL GRAD`.
326
327Additionally, new CPU backends
328were included in this release, such as the `/cpu/self/opt/*` backends (which are
329written in pure C and use partial **E-vectors** to improve performance) and the
330`/cpu/self/ref/memcheck` backend (which relies upon the
331[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
332{ref}`CeedQFunction` have no undefined values).
333This release also included various performance improvements, bug fixes, new examples,
334and improved tests. Among these improvements, vectorized instructions for
335{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
336instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
337identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
338to include unstructured meshes handling were. For this expansion, the prior version of
339the PETSc BPs, which only included data associated with structured geometries, were
340renamed `bpsraw`, and the new version of the BPs, which can handle data associated
341with any unstructured geometry, were called `bps`. Additionally, other benchmark
342problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
343and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
344as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
345examples. Furthermoew, another standalone libCEED example, called `ex2`, which
346computes the surface area of a given mesh was added to this release.
347
348Backends available in this release:
349
350| CEED resource (`-ceed`)  | Backend                                             |
351|--------------------------|-----------------------------------------------------|
352| `/cpu/self/ref/serial`   | Serial reference implementation                     |
353| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
354| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
355| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
356| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
357| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
358| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
359| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
360| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
361| `/cpu/occa`              | Serial OCCA kernels                                 |
362| `/gpu/occa`              | CUDA OCCA kernels                                   |
363| `/omp/occa`              | OpenMP OCCA kernels                                 |
364| `/ocl/occa`              | OpenCL OCCA kernels                                 |
365| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
366| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
367| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
368| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
369| `/gpu/magma`             | CUDA MAGMA kernels                                  |
370
371Examples available in this release:
372
373:::{list-table}
374:header-rows: 1
375:widths: auto
376* - User code
377  - Example
378* - `ceed`
379  - * ex1 (volume)
380    * ex2 (surface)
381* - `mfem`
382  - * BP1 (scalar mass operator)
383    * BP3 (scalar Laplace operator)
384* - `petsc`
385  - * BP1 (scalar mass operator)
386    * BP2 (vector mass operator)
387    * BP3 (scalar Laplace operator)
388    * BP4 (vector Laplace operator)
389    * BP5 (collocated scalar Laplace operator)
390    * BP6 (collocated vector Laplace operator)
391    * Navier-Stokes
392* - `nek5000`
393  - * BP1 (scalar mass operator)
394    * BP3 (scalar Laplace operator)
395:::
396
397(v0-4)=
398
399## v0.4 (Apr 1, 2019)
400
401libCEED v0.4 was made again publicly available in the second full CEED software
402distribution, release CEED 2.0. This release contained notable features, such as
403four new CPU backends, two new GPU backends, CPU backend optimizations, initial
404support for operator composition, performance benchmarking, and a Navier-Stokes demo.
405The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
406backends process one element at a time and are intended for meshes with a smaller number
407of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
408eight interlaced elements and are intended for meshes with higher numbers of elements.
409The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
410performance. The `/cpu/self/xsmm/*` backends rely upon the
411[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
412performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
413The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
414performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
415parallelization approach, where each thread treats a finite element. Using just in time
416compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
417backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
418achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
419deteriorates very quickly when threads run out of registers.
420
421A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
422examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
423This example solves the time-dependent Navier-Stokes equations of compressible gas
424dynamics in a static Eulerian three-dimensional frame, using structured high-order
425finite/spectral element spatial discretizations and explicit high-order time-stepping
426(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
427so that the pointwise physics (defined at quadrature points) is separated from the
428parallelization and meshing concerns.
429
430Backends available in this release:
431
432| CEED resource (`-ceed`)  | Backend                                             |
433|--------------------------|-----------------------------------------------------|
434| `/cpu/self/ref/serial`   | Serial reference implementation                     |
435| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
436| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
437| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
438| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
439| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
440| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
441| `/cpu/occa`              | Serial OCCA kernels                                 |
442| `/gpu/occa`              | CUDA OCCA kernels                                   |
443| `/omp/occa`              | OpenMP OCCA kernels                                 |
444| `/ocl/occa`              | OpenCL OCCA kernels                                 |
445| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
446| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
447| `/gpu/magma`             | CUDA MAGMA kernels                                  |
448
449Examples available in this release:
450
451:::{list-table}
452:header-rows: 1
453:widths: auto
454* - User code
455  - Example
456* - `ceed`
457  - * ex1 (volume)
458* - `mfem`
459  - * BP1 (scalar mass operator)
460    * BP3 (scalar Laplace operator)
461* - `petsc`
462  - * BP1 (scalar mass operator)
463    * BP3 (scalar Laplace operator)
464    * Navier-Stokes
465* - `nek5000`
466  - * BP1 (scalar mass operator)
467    * BP3 (scalar Laplace operator)
468:::
469
470(v0-3)=
471
472## v0.3 (Sep 30, 2018)
473
474Notable features in this release include active/passive field interface, support for
475non-tensor bases, backend optimization, and improved Fortran interface. This release
476also focused on providing improved continuous integration, and many new tests with code
477coverage reports of about 90%. This release also provided a significant change to the
478public interface: a {ref}`CeedQFunction` can take any number of named input and output
479arguments while {ref}`CeedOperator` connects them to the actual data, which may be
480supplied explicitly to `CeedOperatorApply()` (active) or separately via
481`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
482of CeedQFunctions and composition of block solvers constructed using
483{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
484used in an optimized CPU backend. Although this is typically not visible to the user,
485it enables effective use of arbitrary-length SIMD while maintaining cache locality.
486This CPU backend also implements an algebraic factorization of tensor product gradients
487to perform fewer operations than standard application of interpolation and
488differentiation from nodes to quadrature points. This algebraic formulation
489automatically supports non-polynomial and non-interpolatory bases, thus is more general
490than the more common derivation in terms of Lagrange polynomials on the quadrature points.
491
492Backends available in this release:
493
494| CEED resource (`-ceed`) | Backend                                             |
495|-------------------------|-----------------------------------------------------|
496| `/cpu/self/blocked`     | Blocked reference implementation                    |
497| `/cpu/self/ref`         | Serial reference implementation                     |
498| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
499| `/cpu/occa`             | Serial OCCA kernels                                 |
500| `/gpu/occa`             | CUDA OCCA kernels                                   |
501| `/omp/occa`             | OpenMP OCCA kernels                                 |
502| `/ocl/occa`             | OpenCL OCCA kernels                                 |
503| `/gpu/magma`            | CUDA MAGMA kernels                                  |
504
505Examples available in this release:
506
507:::{list-table}
508:header-rows: 1
509:widths: auto
510* - User code
511  - Example
512* - `ceed`
513  - * ex1 (volume)
514* - `mfem`
515  - * BP1 (scalar mass operator)
516    * BP3 (scalar Laplace operator)
517* - `petsc`
518  - * BP1 (scalar mass operator)
519    * BP3 (scalar Laplace operator)
520* - `nek5000`
521  - * BP1 (scalar mass operator)
522    * BP3 (scalar Laplace operator)
523:::
524
525(v0-21)=
526
527## v0.21 (Sep 30, 2018)
528
529A MAGMA backend (which relies upon the
530[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
531release. This initial integration set up the framework of using MAGMA and provided the
532libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
533As any other backend, the MAGMA backend provides extended basic data structures for
534{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
535the fundamental CEED building blocks to work with the new data structures.
536In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
537but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
538internally, and thus seamlessly to the user, through the functions/methods that are
539provided to support them.
540
541Backends available in this release:
542
543| CEED resource (`-ceed`) | Backend                         |
544|-------------------------|---------------------------------|
545| `/cpu/self`             | Serial reference implementation |
546| `/cpu/occa`             | Serial OCCA kernels             |
547| `/gpu/occa`             | CUDA OCCA kernels               |
548| `/omp/occa`             | OpenMP OCCA kernels             |
549| `/ocl/occa`             | OpenCL OCCA kernels             |
550| `/gpu/magma`            | CUDA MAGMA kernels              |
551
552Examples available in this release:
553
554:::{list-table}
555:header-rows: 1
556:widths: auto
557* - User code
558  - Example
559* - `ceed`
560  - * ex1 (volume)
561* - `mfem`
562  - * BP1 (scalar mass operator)
563    * BP3 (scalar Laplace operator)
564* - `petsc`
565  - * BP1 (scalar mass operator)
566* - `nek5000`
567  - * BP1 (scalar mass operator)
568:::
569
570(v0-2)=
571
572## v0.2 (Mar 30, 2018)
573
574libCEED was made publicly available the first full CEED software distribution, release
575CEED 1.0. The distribution was made available using the Spack package manager to provide
576a common, easy-to-use build environment, where the user can build the CEED distribution
577with all dependencies. This release included a new Fortran interface for the library.
578This release also contained major improvements in the OCCA backend (including a new
579`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
580compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
581`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
582of the Laplace operator) was also added to this release.
583
584Backends available in this release:
585
586| CEED resource (`-ceed`) | Backend                         |
587|-------------------------|---------------------------------|
588| `/cpu/self`             | Serial reference implementation |
589| `/cpu/occa`             | Serial OCCA kernels             |
590| `/gpu/occa`             | CUDA OCCA kernels               |
591| `/omp/occa`             | OpenMP OCCA kernels             |
592| `/ocl/occa`             | OpenCL OCCA kernels             |
593
594Examples available in this release:
595
596:::{list-table}
597:header-rows: 1
598:widths: auto
599* - User code
600  - Example
601* - `ceed`
602  - * ex1 (volume)
603* - `mfem`
604  - * BP1 (scalar mass operator)
605    * BP3 (scalar Laplace operator)
606* - `petsc`
607  - * BP1 (scalar mass operator)
608* - `nek5000`
609  - * BP1 (scalar mass operator)
610:::
611
612(v0-1)=
613
614## v0.1 (Jan 3, 2018)
615
616Initial low-level API of the CEED project. The low-level API provides a set of Finite
617Elements kernels and components for writing new low-level kernels. Examples include:
618vector and sparse linear algebra, element matrix assembly over a batch of elements,
619partial assembly and action for efficient high-order operators like mass, diffusion,
620advection, etc. The main goal of the low-level API is to establish the basis for the
621high-level API. Also, identifying such low-level kernels and providing a reference
622implementation for them serves as the basis for specialized backend implementations.
623This release contained several backends: `/cpu/self`, and backends which rely upon the
624[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
625`/gpu/occa`, and `/omp/occa`.
626It also included several examples, in the `examples` folder:
627A standalone code that shows the usage of libCEED (with no external
628dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
629(with the application of the mass operator); and a `petsc` example to perform BP1
630(with the application of the mass operator).
631
632Backends available in this release:
633
634| CEED resource (`-ceed`) | Backend                         |
635|-------------------------|---------------------------------|
636| `/cpu/self`             | Serial reference implementation |
637| `/cpu/occa`             | Serial OCCA kernels             |
638| `/gpu/occa`             | CUDA OCCA kernels               |
639| `/omp/occa`             | OpenMP OCCA kernels             |
640
641Examples available in this release:
642
643| User code             | Example                           |
644|-----------------------|-----------------------------------|
645| `ceed`                | ex1 (scalar Laplace operator)     |
646| `mfem`                | BP1 (scalar mass operator)        |
647| `petsc`               | BP1 (scalar mass operator)        |
648```
649