xref: /libCEED/doc/sphinx/source/releasenotes.md (revision b0d170e7bc2e5c930ee481a47eb73044935a48a4)
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(v0-11)=
10
11## v0.11 (Dec 24, 2022)
12
13### Interface changes
14
15- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output.
16- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`.
17- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface.
18
19### New features
20
21- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user.
22- Improved support for $H(\text{div})$ bases.
23- Added `CeedInt_FMT` to support potential future use of larger interger sizes.
24- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`.
25- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled.
26Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend.
27This issue will be fixed in a future OCCA release.
28
29### Bugfix
30
31- Fix bug in setting device id for GPU backends.
32- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends.
33- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`.
34- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes.
35
36### Examples
37
38#### {ref}`example-petsc-navier-stokes`
39
40- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs.
41- Refactored to improve code reuse and modularity.
42- Support for primitive variables for more accurate boundary layers and all-speed flow.
43- Added $YZ\beta$ shock capturing scheme and Shock Tube example.
44- Added Channel example, with comparison to analytic solutions.
45- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations.
46- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions.
47- Added "freestream" boundary conditions based on HLLC Riemann solver.
48- Automated stabilization coefficients for different basis degree.
49
50#### {ref}`example-petsc-bps`
51
52- Support for convergence studies.
53
54### Maintainability
55
56- 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.
57- Enabled support for `p > 8` for `/gpu/*/shared` backends.
58- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`.
59- Improved test harness.
60
61(v0-10-1)=
62
63## v0.10.1 (Apr 11, 2022)
64
65### Interface changes
66
67- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application.
68
69### New features
70
71- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends).
72This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work.
73
74### Bugfix
75
76- Install JiT source files in install directory to fix GPU functionality for installed libCEED.
77
78(v0-10)=
79
80## v0.10 (Mar 21, 2022)
81
82### Interface changes
83
84- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields.
85- 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`.
86- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added.
87- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients.
88- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces.
89- Warning added when compiling OCCA backend to alert users that this backend is experimental.
90- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`.
91- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name.
92- 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`.
93- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`.
94- 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.
95- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`.
96- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums.
97- 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.
98- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy.
99- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator.
100
101### New features
102
103- `CeedScalar` can now be set as `float` or `double` at compile time.
104- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends.
105- 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.
106- Remove need to guard library headers in QFunction source for code generation backends.
107- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present.
108- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally.
109- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file.
110- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor.
111- 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.
112- Added support for element matrix assembly in GPU backends.
113
114### Maintainability
115
116- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
117- `Include-what-you-use` makefile target added as `make iwyu`.
118- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase.
119- Put GPU JiTed kernel source code into separate files.
120- 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.
121
122(v0-9)=
123
124## v0.9 (Jul 6, 2021)
125
126### Interface changes
127
128- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
129
130### New features
131
132- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
133- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
134- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
135- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
136
137### Performance improvements
138
139### Examples
140
141- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
142- Fluid mechanics example adds GPU support and improves modularity.
143
144### Deprecated backends
145
146- 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.
147
148(v0-8)=
149
150## v0.8 (Mar 31, 2021)
151
152### Interface changes
153
154- Error handling improved to include enumerated error codes for C interface return values.
155- 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.
156
157### New features
158
159- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
160- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
161- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
162
163### Performance improvements
164
165- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
166- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
167
168### Examples
169
170- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
171- {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.
172- {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.
173- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
174
175(v0-7)=
176
177## v0.7 (Sep 29, 2020)
178
179### Interface changes
180
181- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
182  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
183  These changes improve support for mixed finite element methods.
184- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
185- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
186- 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.
187- 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.
188- 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.
189- 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.
190  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`.
191- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
192- 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.
193
194### New features
195
196- New HIP backend: `/gpu/hip/ref`.
197- CeedQFunction support for user `CUfunction`s in some backends
198
199### Performance improvements
200
201- OCCA backend rebuilt to facilitate future performance enhancements.
202- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
203
204### Examples
205
206- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
207
208### Deprecated backends
209
210- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
211
212(v0-6)=
213
214## v0.6 (Mar 29, 2020)
215
216libCEED v0.6 contains numerous new features and examples, as well as expanded
217documentation in [this new website](https://libceed.org).
218
219### New features
220
221- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
222  1-1 correspondence with the C interface, plus some convenience features.  For instance,
223  data stored in the {cpp:type}`CeedVector` structure are available without copy as
224  {py:class}`numpy.ndarray`.  Short tutorials are provided in
225  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
226- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
227  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
228  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
229  preconditioning ingredients and are used in the libCEED's multigrid examples.
230- The inverse of separable operators can be obtained using
231  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
232  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
233  especially for Laplacians and related operators.
234- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
235  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
236- Make public accessors for various attributes to facilitate writing composable code.
237- New backend: `/cpu/self/memcheck/serial`.
238- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
239  backends.  (Single source is coming soon for OCCA backends.)
240- Fix some missing edge cases in CUDA backend.
241
242### Performance Improvements
243
244- MAGMA backend performance optimization and non-tensor bases.
245- No-copy optimization in {c:func}`CeedOperatorApply`.
246
247### Interface changes
248
249- Replace {code}`CeedElemRestrictionCreateIdentity` and
250  {code}`CeedElemRestrictionCreateBlocked` with more flexible
251  {c:func}`CeedElemRestrictionCreateStrided` and
252  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
253- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
254- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
255  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
256  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
257  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
258  {c:func}`CeedElemRestrictionApply`.
259
260### Examples
261
262libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
263Notable additions include:
264
265- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
266  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
267
268- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
269  of domains (like the cube and sphere) by direct integration on a surface mesh;
270  demonstrates geometric dimension different from topological dimension.
271
272- PETSc {ref}`example-petsc-bps`:
273
274  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
275  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
276    and transparent CUDA support.
277  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
278    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
279    dimension different from topological dimension.
280
281- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
282  solver with algebraic multigrid coarse solve.
283
284- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
285  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
286  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
287  quasi-2D computational domain support.
288
289- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
290  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
291  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
292
293(v0-5)=
294
295## v0.5 (Sep 18, 2019)
296
297For this release, several improvements were made. Two new CUDA backends were added to
298the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
299performance using single-source {ref}`CeedQFunction`. From this release, users
300can define Q-Functions in a single source code independently of the targeted backend
301with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
302compilation of the user provided {ref}`CeedQFunction` code. To allow a unified
303declaration, the {ref}`CeedQFunction` API has undergone a slight change:
304the `QFunctionField` parameter `ncomp` has been changed to `size`. This change
305requires setting the previous value of `ncomp` to `ncomp*dim` when adding a
306`QFunctionField` with eval mode `CEED EVAL GRAD`.
307
308Additionally, new CPU backends
309were included in this release, such as the `/cpu/self/opt/*` backends (which are
310written in pure C and use partial **E-vectors** to improve performance) and the
311`/cpu/self/ref/memcheck` backend (which relies upon the
312[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
313{ref}`CeedQFunction` have no undefined values).
314This release also included various performance improvements, bug fixes, new examples,
315and improved tests. Among these improvements, vectorized instructions for
316{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
317instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
318identity Q-Functions were introduced, and the PETSc benchmark problems were expanded
319to include unstructured meshes handling were. For this expansion, the prior version of
320the PETSc BPs, which only included data associated with structured geometries, were
321renamed `bpsraw`, and the new version of the BPs, which can handle data associated
322with any unstructured geometry, were called `bps`. Additionally, other benchmark
323problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
324and BP5 and BP6 (the collocated versions---for which the quadrature points are the same
325as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
326examples. Furthermoew, another standalone libCEED example, called `ex2`, which
327computes the surface area of a given mesh was added to this release.
328
329Backends available in this release:
330
331| CEED resource (`-ceed`)  | Backend                                             |
332|--------------------------|-----------------------------------------------------|
333| `/cpu/self/ref/serial`   | Serial reference implementation                     |
334| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
335| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
336| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
337| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
338| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
339| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
340| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
341| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
342| `/cpu/occa`              | Serial OCCA kernels                                 |
343| `/gpu/occa`              | CUDA OCCA kernels                                   |
344| `/omp/occa`              | OpenMP OCCA kernels                                 |
345| `/ocl/occa`              | OpenCL OCCA kernels                                 |
346| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
347| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
348| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
349| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
350| `/gpu/magma`             | CUDA MAGMA kernels                                  |
351
352Examples available in this release:
353
354:::{list-table}
355:header-rows: 1
356:widths: auto
357* - User code
358  - Example
359* - `ceed`
360  - * ex1 (volume)
361    * ex2 (surface)
362* - `mfem`
363  - * BP1 (scalar mass operator)
364    * BP3 (scalar Laplace operator)
365* - `petsc`
366  - * BP1 (scalar mass operator)
367    * BP2 (vector mass operator)
368    * BP3 (scalar Laplace operator)
369    * BP4 (vector Laplace operator)
370    * BP5 (collocated scalar Laplace operator)
371    * BP6 (collocated vector Laplace operator)
372    * Navier-Stokes
373* - `nek5000`
374  - * BP1 (scalar mass operator)
375    * BP3 (scalar Laplace operator)
376:::
377
378(v0-4)=
379
380## v0.4 (Apr 1, 2019)
381
382libCEED v0.4 was made again publicly available in the second full CEED software
383distribution, release CEED 2.0. This release contained notable features, such as
384four new CPU backends, two new GPU backends, CPU backend optimizations, initial
385support for operator composition, performance benchmarking, and a Navier-Stokes demo.
386The new CPU backends in this release came in two families. The `/cpu/self/*/serial`
387backends process one element at a time and are intended for meshes with a smaller number
388of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
389eight interlaced elements and are intended for meshes with higher numbers of elements.
390The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
391performance. The `/cpu/self/xsmm/*` backends rely upon the
392[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
393performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
394The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
395performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
396parallelization approach, where each thread treats a finite element. Using just in time
397compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
398backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
399achieve good peak performance for 1D, 2D, and low order 3D problems, but performance
400deteriorates very quickly when threads run out of registers.
401
402A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
403examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
404This example solves the time-dependent Navier-Stokes equations of compressible gas
405dynamics in a static Eulerian three-dimensional frame, using structured high-order
406finite/spectral element spatial discretizations and explicit high-order time-stepping
407(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
408so that the pointwise physics (defined at quadrature points) is separated from the
409parallelization and meshing concerns.
410
411Backends available in this release:
412
413| CEED resource (`-ceed`)  | Backend                                             |
414|--------------------------|-----------------------------------------------------|
415| `/cpu/self/ref/serial`   | Serial reference implementation                     |
416| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
417| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
418| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
419| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
420| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
421| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
422| `/cpu/occa`              | Serial OCCA kernels                                 |
423| `/gpu/occa`              | CUDA OCCA kernels                                   |
424| `/omp/occa`              | OpenMP OCCA kernels                                 |
425| `/ocl/occa`              | OpenCL OCCA kernels                                 |
426| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
427| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
428| `/gpu/magma`             | CUDA MAGMA kernels                                  |
429
430Examples available in this release:
431
432:::{list-table}
433:header-rows: 1
434:widths: auto
435* - User code
436  - Example
437* - `ceed`
438  - * ex1 (volume)
439* - `mfem`
440  - * BP1 (scalar mass operator)
441    * BP3 (scalar Laplace operator)
442* - `petsc`
443  - * BP1 (scalar mass operator)
444    * BP3 (scalar Laplace operator)
445    * Navier-Stokes
446* - `nek5000`
447  - * BP1 (scalar mass operator)
448    * BP3 (scalar Laplace operator)
449:::
450
451(v0-3)=
452
453## v0.3 (Sep 30, 2018)
454
455Notable features in this release include active/passive field interface, support for
456non-tensor bases, backend optimization, and improved Fortran interface. This release
457also focused on providing improved continuous integration, and many new tests with code
458coverage reports of about 90%. This release also provided a significant change to the
459public interface: a {ref}`CeedQFunction` can take any number of named input and output
460arguments while {ref}`CeedOperator` connects them to the actual data, which may be
461supplied explicitly to `CeedOperatorApply()` (active) or separately via
462`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
463of CeedQFunctions and composition of block solvers constructed using
464{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
465used in an optimized CPU backend. Although this is typically not visible to the user,
466it enables effective use of arbitrary-length SIMD while maintaining cache locality.
467This CPU backend also implements an algebraic factorization of tensor product gradients
468to perform fewer operations than standard application of interpolation and
469differentiation from nodes to quadrature points. This algebraic formulation
470automatically supports non-polynomial and non-interpolatory bases, thus is more general
471than the more common derivation in terms of Lagrange polynomials on the quadrature points.
472
473Backends available in this release:
474
475| CEED resource (`-ceed`) | Backend                                             |
476|-------------------------|-----------------------------------------------------|
477| `/cpu/self/blocked`     | Blocked reference implementation                    |
478| `/cpu/self/ref`         | Serial reference implementation                     |
479| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
480| `/cpu/occa`             | Serial OCCA kernels                                 |
481| `/gpu/occa`             | CUDA OCCA kernels                                   |
482| `/omp/occa`             | OpenMP OCCA kernels                                 |
483| `/ocl/occa`             | OpenCL OCCA kernels                                 |
484| `/gpu/magma`            | CUDA MAGMA kernels                                  |
485
486Examples available in this release:
487
488:::{list-table}
489:header-rows: 1
490:widths: auto
491* - User code
492  - Example
493* - `ceed`
494  - * ex1 (volume)
495* - `mfem`
496  - * BP1 (scalar mass operator)
497    * BP3 (scalar Laplace operator)
498* - `petsc`
499  - * BP1 (scalar mass operator)
500    * BP3 (scalar Laplace operator)
501* - `nek5000`
502  - * BP1 (scalar mass operator)
503    * BP3 (scalar Laplace operator)
504:::
505
506(v0-21)=
507
508## v0.21 (Sep 30, 2018)
509
510A MAGMA backend (which relies upon the
511[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
512release. This initial integration set up the framework of using MAGMA and provided the
513libCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
514As any other backend, the MAGMA backend provides extended basic data structures for
515{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
516the fundamental CEED building blocks to work with the new data structures.
517In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
518but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
519internally, and thus seamlessly to the user, through the functions/methods that are
520provided to support them.
521
522Backends available in this release:
523
524| CEED resource (`-ceed`) | Backend                         |
525|-------------------------|---------------------------------|
526| `/cpu/self`             | Serial reference implementation |
527| `/cpu/occa`             | Serial OCCA kernels             |
528| `/gpu/occa`             | CUDA OCCA kernels               |
529| `/omp/occa`             | OpenMP OCCA kernels             |
530| `/ocl/occa`             | OpenCL OCCA kernels             |
531| `/gpu/magma`            | CUDA MAGMA kernels              |
532
533Examples available in this release:
534
535:::{list-table}
536:header-rows: 1
537:widths: auto
538* - User code
539  - Example
540* - `ceed`
541  - * ex1 (volume)
542* - `mfem`
543  - * BP1 (scalar mass operator)
544    * BP3 (scalar Laplace operator)
545* - `petsc`
546  - * BP1 (scalar mass operator)
547* - `nek5000`
548  - * BP1 (scalar mass operator)
549:::
550
551(v0-2)=
552
553## v0.2 (Mar 30, 2018)
554
555libCEED was made publicly available the first full CEED software distribution, release
556CEED 1.0. The distribution was made available using the Spack package manager to provide
557a common, easy-to-use build environment, where the user can build the CEED distribution
558with all dependencies. This release included a new Fortran interface for the library.
559This release also contained major improvements in the OCCA backend (including a new
560`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
561compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
562`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
563of the Laplace operator) was also added to this release.
564
565Backends available in this release:
566
567| CEED resource (`-ceed`) | Backend                         |
568|-------------------------|---------------------------------|
569| `/cpu/self`             | Serial reference implementation |
570| `/cpu/occa`             | Serial OCCA kernels             |
571| `/gpu/occa`             | CUDA OCCA kernels               |
572| `/omp/occa`             | OpenMP OCCA kernels             |
573| `/ocl/occa`             | OpenCL OCCA kernels             |
574
575Examples available in this release:
576
577:::{list-table}
578:header-rows: 1
579:widths: auto
580* - User code
581  - Example
582* - `ceed`
583  - * ex1 (volume)
584* - `mfem`
585  - * BP1 (scalar mass operator)
586    * BP3 (scalar Laplace operator)
587* - `petsc`
588  - * BP1 (scalar mass operator)
589* - `nek5000`
590  - * BP1 (scalar mass operator)
591:::
592
593(v0-1)=
594
595## v0.1 (Jan 3, 2018)
596
597Initial low-level API of the CEED project. The low-level API provides a set of Finite
598Elements kernels and components for writing new low-level kernels. Examples include:
599vector and sparse linear algebra, element matrix assembly over a batch of elements,
600partial assembly and action for efficient high-order operators like mass, diffusion,
601advection, etc. The main goal of the low-level API is to establish the basis for the
602high-level API. Also, identifying such low-level kernels and providing a reference
603implementation for them serves as the basis for specialized backend implementations.
604This release contained several backends: `/cpu/self`, and backends which rely upon the
605[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
606`/gpu/occa`, and `/omp/occa`.
607It also included several examples, in the `examples` folder:
608A standalone code that shows the usage of libCEED (with no external
609dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
610(with the application of the mass operator); and a `petsc` example to perform BP1
611(with the application of the mass operator).
612
613Backends available in this release:
614
615| CEED resource (`-ceed`) | Backend                         |
616|-------------------------|---------------------------------|
617| `/cpu/self`             | Serial reference implementation |
618| `/cpu/occa`             | Serial OCCA kernels             |
619| `/gpu/occa`             | CUDA OCCA kernels               |
620| `/omp/occa`             | OpenMP OCCA kernels             |
621
622Examples available in this release:
623
624| User code             | Example                           |
625|-----------------------|-----------------------------------|
626| `ceed`                | ex1 (scalar Laplace operator)     |
627| `mfem`                | BP1 (scalar mass operator)        |
628| `petsc`               | BP1 (scalar mass operator)        |
629```
630