xref: /libCEED/doc/sphinx/source/releasenotes.md (revision edf0491998c1d524f2f70fdd683669b8904cb3b6)
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
21### Examples
22
23#### {ref}`example-petsc-bps`
24
25- Requires PETSc version 3.19 or later
26
27(v0-11)=
28
29## v0.11 (Dec 24, 2022)
30
31### Interface changes
32
33- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
34- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
35- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface.
36
37### New features
38
39- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
40- Improved support for $H(\text{div})$ bases.
41- Added `CeedInt_FMT` to support potential future use of larger integer sizes.
42- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`.
43- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
44Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
45This issue will be fixed in a future OCCA release.
46
47### Bugfix
48
49- Fix bug in setting device id for GPU backends.
50- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
51- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
52- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
53
54### Examples
55
56#### {ref}`example-petsc-navier-stokes`
57
58- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs.
59- Refactored to improve code reuse and modularity.
60- Support for primitive variables for more accurate boundary layers and all-speed flow.
61- Added $YZ\beta$ shock capturing scheme and Shock Tube example.
62- Added Channel example, with comparison to analytic solutions.
63- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations.
64- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions.
65- Added "freestream" boundary conditions based on HLLC Riemann solver.
66- Automated stabilization coefficients for different basis degree.
67
68#### {ref}`example-petsc-bps`
69
70- Support for convergence studies.
71
72### Maintainability
73
74- 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.
75- Enabled support for `p > 8` for `/gpu/*/shared` backends.
76- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
77- Improved test harness.
78
79(v0-10-1)=
80
81## v0.10.1 (Apr 11, 2022)
82
83### Interface changes
84
85- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
86
87### New features
88
89- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
90This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
91
92### Bugfix
93
94- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
95
96(v0-10)=
97
98## v0.10 (Mar 21, 2022)
99
100### Interface changes
101
102- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
103- 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`.
104- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
105- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
106- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
107- Warning added when compiling OCCA backend to alert users that this backend is experimental.
108- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
109- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
110- 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`.
111- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
112- 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.
113- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
114- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
115- 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.
116- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
117- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
118
119### New features
120
121- `CeedScalar` can now be set as `float` or `double` at compile time.
122- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
123- 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.
124- Remove need to guard library headers in QFunction source for code generation backends.
125- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
126- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
127- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
128- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
129- 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.
130- Added support for element matrix assembly in GPU backends.
131
132### Maintainability
133
134- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
135- `Include-what-you-use` makefile target added as `make iwyu`.
136- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
137- Put GPU JiTed kernel source code into separate files.
138- 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.
139
140(v0-9)=
141
142## v0.9 (Jul 6, 2021)
143
144### Interface changes
145
146- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
147
148### New features
149
150- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
151- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
152- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
153- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
154
155### Performance improvements
156
157### Examples
158
159- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
160- Fluid mechanics example adds GPU support and improves modularity.
161
162### Deprecated backends
163
164- 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.
165
166(v0-8)=
167
168## v0.8 (Mar 31, 2021)
169
170### Interface changes
171
172- Error handling improved to include enumerated error codes for C interface return values.
173- 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.
174
175### New features
176
177- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
178- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
179- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
180
181### Performance improvements
182
183- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
184- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
185
186### Examples
187
188- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
189- {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.
190- {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.
191- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
192
193(v0-7)=
194
195## v0.7 (Sep 29, 2020)
196
197### Interface changes
198
199- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
200  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
201  These changes improve support for mixed finite element methods.
202- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
203- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
204- 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.
205- 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.
206- 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.
207- 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.
208  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`.
209- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
210- 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.
211
212### New features
213
214- New HIP backend: `/gpu/hip/ref`.
215- CeedQFunction support for user `CUfunction`s in some backends
216
217### Performance improvements
218
219- OCCA backend rebuilt to facilitate future performance enhancements.
220- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
221
222### Examples
223
224- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
225
226### Deprecated backends
227
228- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
229
230(v0-6)=
231
232## v0.6 (Mar 29, 2020)
233
234libCEED v0.6 contains numerous new features and examples, as well as expanded
235documentation in [this new website](https://libceed.org).
236
237### New features
238
239- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
240  1-1 correspondence with the C interface, plus some convenience features.  For instance,
241  data stored in the {cpp:type}`CeedVector` structure are available without copy as
242  {py:class}`numpy.ndarray`.  Short tutorials are provided in
243  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
244- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
245  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
246  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
247  preconditioning ingredients and are used in the libCEED's multigrid examples.
248- The inverse of separable operators can be obtained using
249  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
250  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
251  especially for Laplacians and related operators.
252- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
253  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
254- Make public accessors for various attributes to facilitate writing composable code.
255- New backend: `/cpu/self/memcheck/serial`.
256- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
257  backends.  (Single source is coming soon for OCCA backends.)
258- Fix some missing edge cases in CUDA backend.
259
260### Performance Improvements
261
262- MAGMA backend performance optimization and non-tensor bases.
263- No-copy optimization in {c:func}`CeedOperatorApply`.
264
265### Interface changes
266
267- Replace {code}`CeedElemRestrictionCreateIdentity` and
268  {code}`CeedElemRestrictionCreateBlocked` with more flexible
269  {c:func}`CeedElemRestrictionCreateStrided` and
270  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
271- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
272- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
273  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
274  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
275  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
276  {c:func}`CeedElemRestrictionApply`.
277
278### Examples
279
280libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
281Notable additions include:
282
283- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
284  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
285
286- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
287  of domains (like the cube and sphere) by direct integration on a surface mesh;
288  demonstrates geometric dimension different from topological dimension.
289
290- PETSc {ref}`example-petsc-bps`:
291
292  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
293  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
294    and transparent CUDA support.
295  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
296    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
297    dimension different from topological dimension.
298
299- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
300  solver with algebraic multigrid coarse solve.
301
302- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
303  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
304  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
305  quasi-2D computational domain support.
306
307- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
308  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
309  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
310
311(v0-5)=
312
313## v0.5 (Sep 18, 2019)
314
315For this release, several improvements were made. Two new CUDA backends were added to
316the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
317performance using single-source {ref}`CeedQFunction`. From this release, users
318can define Q-Functions in a single source code independently of the targeted backend
319with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
320compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
321declaration, the {ref}`CeedQFunction` API has undergone a slight change:
322the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
323requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
324`QFunctionField` with eval mode `CEED EVAL GRAD`.
325
326Additionally, new CPU backends
327were included in this release, such as the `/cpu/self/opt/*` backends (which are
328written in pure C and use partial **E-vectors** to improve performance) and the
329`/cpu/self/ref/memcheck` backend (which relies upon the
330[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
331{ref}`CeedQFunction` have no undefined values).
332This release also included various performance improvements, bug fixes, new examples,
333and improved tests. Among these improvements, vectorized instructions for
334{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
335instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
336identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
337to include unstructured meshes handling were. For this expansion, the prior version of
338the PETSc BPs, which only included data associated with structured geometries, were
339renamed `bpsraw`, and the new version of the BPs, which can handle data associated
340with any unstructured geometry, were called `bps`. Additionally, other benchmark
341problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
342and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
343as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
344examples. Furthermoew, another standalone libCEED example, called `ex2`, which
345computes the surface area of a given mesh was added to this release.
346
347Backends available in this release:
348
349| CEED resource (`-ceed`)  | Backend                                             |
350|--------------------------|-----------------------------------------------------|
351| `/cpu/self/ref/serial`   | Serial reference implementation                     |
352| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
353| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
354| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
355| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
356| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
357| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
358| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
359| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
360| `/cpu/occa`              | Serial OCCA kernels                                 |
361| `/gpu/occa`              | CUDA OCCA kernels                                   |
362| `/omp/occa`              | OpenMP OCCA kernels                                 |
363| `/ocl/occa`              | OpenCL OCCA kernels                                 |
364| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
365| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
366| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
367| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
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    * ex2 (surface)
380* - `mfem`
381  - * BP1 (scalar mass operator)
382    * BP3 (scalar Laplace operator)
383* - `petsc`
384  - * BP1 (scalar mass operator)
385    * BP2 (vector mass operator)
386    * BP3 (scalar Laplace operator)
387    * BP4 (vector Laplace operator)
388    * BP5 (collocated scalar Laplace operator)
389    * BP6 (collocated vector Laplace operator)
390    * Navier-Stokes
391* - `nek5000`
392  - * BP1 (scalar mass operator)
393    * BP3 (scalar Laplace operator)
394:::
395
396(v0-4)=
397
398## v0.4 (Apr 1, 2019)
399
400libCEED v0.4 was made again publicly available in the second full CEED software
401distribution, release CEED 2.0. This release contained notable features, such as
402four new CPU backends, two new GPU backends, CPU backend optimizations, initial
403support for operator composition, performance benchmarking, and a Navier-Stokes demo.
404The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
405backends process one element at a time and are intended for meshes with a smaller number
406of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
407eight interlaced elements and are intended for meshes with higher numbers of elements.
408The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
409performance. The `/cpu/self/xsmm/*` backends rely upon the
410[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
411performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
412The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
413performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
414parallelization approach, where each thread treats a finite element. Using just in time
415compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
416backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
417achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
418deteriorates very quickly when threads run out of registers.
419
420A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
421examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
422This example solves the time-dependent Navier-Stokes equations of compressible gas
423dynamics in a static Eulerian three-dimensional frame, using structured high-order
424finite/spectral element spatial discretizations and explicit high-order time-stepping
425(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
426so that the pointwise physics (defined at quadrature points) is separated from the
427parallelization and meshing concerns.
428
429Backends available in this release:
430
431| CEED resource (`-ceed`)  | Backend                                             |
432|--------------------------|-----------------------------------------------------|
433| `/cpu/self/ref/serial`   | Serial reference implementation                     |
434| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
435| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
436| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
437| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
438| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
439| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
440| `/cpu/occa`              | Serial OCCA kernels                                 |
441| `/gpu/occa`              | CUDA OCCA kernels                                   |
442| `/omp/occa`              | OpenMP OCCA kernels                                 |
443| `/ocl/occa`              | OpenCL OCCA kernels                                 |
444| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
445| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
446| `/gpu/magma`             | CUDA MAGMA kernels                                  |
447
448Examples available in this release:
449
450:::{list-table}
451:header-rows: 1
452:widths: auto
453* - User code
454  - Example
455* - `ceed`
456  - * ex1 (volume)
457* - `mfem`
458  - * BP1 (scalar mass operator)
459    * BP3 (scalar Laplace operator)
460* - `petsc`
461  - * BP1 (scalar mass operator)
462    * BP3 (scalar Laplace operator)
463    * Navier-Stokes
464* - `nek5000`
465  - * BP1 (scalar mass operator)
466    * BP3 (scalar Laplace operator)
467:::
468
469(v0-3)=
470
471## v0.3 (Sep 30, 2018)
472
473Notable features in this release include active/passive field interface, support for
474non-tensor bases, backend optimization, and improved Fortran interface. This release
475also focused on providing improved continuous integration, and many new tests with code
476coverage reports of about 90%. This release also provided a significant change to the
477public interface: a {ref}`CeedQFunction` can take any number of named input and output
478arguments while {ref}`CeedOperator` connects them to the actual data, which may be
479supplied explicitly to `CeedOperatorApply()` (active) or separately via
480`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
481of CeedQFunctions and composition of block solvers constructed using
482{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
483used in an optimized CPU backend. Although this is typically not visible to the user,
484it enables effective use of arbitrary-length SIMD while maintaining cache locality.
485This CPU backend also implements an algebraic factorization of tensor product gradients
486to perform fewer operations than standard application of interpolation and
487differentiation from nodes to quadrature points. This algebraic formulation
488automatically supports non-polynomial and non-interpolatory bases, thus is more general
489than the more common derivation in terms of Lagrange polynomials on the quadrature points.
490
491Backends available in this release:
492
493| CEED resource (`-ceed`) | Backend                                             |
494|-------------------------|-----------------------------------------------------|
495| `/cpu/self/blocked`     | Blocked reference implementation                    |
496| `/cpu/self/ref`         | Serial reference implementation                     |
497| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
498| `/cpu/occa`             | Serial OCCA kernels                                 |
499| `/gpu/occa`             | CUDA OCCA kernels                                   |
500| `/omp/occa`             | OpenMP OCCA kernels                                 |
501| `/ocl/occa`             | OpenCL OCCA kernels                                 |
502| `/gpu/magma`            | CUDA MAGMA kernels                                  |
503
504Examples available in this release:
505
506:::{list-table}
507:header-rows: 1
508:widths: auto
509* - User code
510  - Example
511* - `ceed`
512  - * ex1 (volume)
513* - `mfem`
514  - * BP1 (scalar mass operator)
515    * BP3 (scalar Laplace operator)
516* - `petsc`
517  - * BP1 (scalar mass operator)
518    * BP3 (scalar Laplace operator)
519* - `nek5000`
520  - * BP1 (scalar mass operator)
521    * BP3 (scalar Laplace operator)
522:::
523
524(v0-21)=
525
526## v0.21 (Sep 30, 2018)
527
528A MAGMA backend (which relies upon the
529[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
530release. This initial integration set up the framework of using MAGMA and provided the
531libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
532As any other backend, the MAGMA backend provides extended basic data structures for
533{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
534the fundamental CEED building blocks to work with the new data structures.
535In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
536but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
537internally, and thus seamlessly to the user, through the functions/methods that are
538provided to support them.
539
540Backends available in this release:
541
542| CEED resource (`-ceed`) | Backend                         |
543|-------------------------|---------------------------------|
544| `/cpu/self`             | Serial reference implementation |
545| `/cpu/occa`             | Serial OCCA kernels             |
546| `/gpu/occa`             | CUDA OCCA kernels               |
547| `/omp/occa`             | OpenMP OCCA kernels             |
548| `/ocl/occa`             | OpenCL OCCA kernels             |
549| `/gpu/magma`            | CUDA MAGMA kernels              |
550
551Examples available in this release:
552
553:::{list-table}
554:header-rows: 1
555:widths: auto
556* - User code
557  - Example
558* - `ceed`
559  - * ex1 (volume)
560* - `mfem`
561  - * BP1 (scalar mass operator)
562    * BP3 (scalar Laplace operator)
563* - `petsc`
564  - * BP1 (scalar mass operator)
565* - `nek5000`
566  - * BP1 (scalar mass operator)
567:::
568
569(v0-2)=
570
571## v0.2 (Mar 30, 2018)
572
573libCEED was made publicly available the first full CEED software distribution, release
574CEED 1.0. The distribution was made available using the Spack package manager to provide
575a common, easy-to-use build environment, where the user can build the CEED distribution
576with all dependencies. This release included a new Fortran interface for the library.
577This release also contained major improvements in the OCCA backend (including a new
578`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
579compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
580`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
581of the Laplace operator) was also added to this release.
582
583Backends available in this release:
584
585| CEED resource (`-ceed`) | Backend                         |
586|-------------------------|---------------------------------|
587| `/cpu/self`             | Serial reference implementation |
588| `/cpu/occa`             | Serial OCCA kernels             |
589| `/gpu/occa`             | CUDA OCCA kernels               |
590| `/omp/occa`             | OpenMP OCCA kernels             |
591| `/ocl/occa`             | OpenCL OCCA kernels             |
592
593Examples available in this release:
594
595:::{list-table}
596:header-rows: 1
597:widths: auto
598* - User code
599  - Example
600* - `ceed`
601  - * ex1 (volume)
602* - `mfem`
603  - * BP1 (scalar mass operator)
604    * BP3 (scalar Laplace operator)
605* - `petsc`
606  - * BP1 (scalar mass operator)
607* - `nek5000`
608  - * BP1 (scalar mass operator)
609:::
610
611(v0-1)=
612
613## v0.1 (Jan 3, 2018)
614
615Initial low-level API of the CEED project. The low-level API provides a set of Finite
616Elements kernels and components for writing new low-level kernels. Examples include:
617vector and sparse linear algebra, element matrix assembly over a batch of elements,
618partial assembly and action for efficient high-order operators like mass, diffusion,
619advection, etc. The main goal of the low-level API is to establish the basis for the
620high-level API. Also, identifying such low-level kernels and providing a reference
621implementation for them serves as the basis for specialized backend implementations.
622This release contained several backends: `/cpu/self`, and backends which rely upon the
623[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
624`/gpu/occa`, and `/omp/occa`.
625It also included several examples, in the `examples` folder:
626A standalone code that shows the usage of libCEED (with no external
627dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
628(with the application of the mass operator); and a `petsc` example to perform BP1
629(with the application of the mass operator).
630
631Backends available in this release:
632
633| CEED resource (`-ceed`) | Backend                         |
634|-------------------------|---------------------------------|
635| `/cpu/self`             | Serial reference implementation |
636| `/cpu/occa`             | Serial OCCA kernels             |
637| `/gpu/occa`             | CUDA OCCA kernels               |
638| `/omp/occa`             | OpenMP OCCA kernels             |
639
640Examples available in this release:
641
642| User code             | Example                           |
643|-----------------------|-----------------------------------|
644| `ceed`                | ex1 (scalar Laplace operator)     |
645| `mfem`                | BP1 (scalar mass operator)        |
646| `petsc`               | BP1 (scalar mass operator)        |
647```
648