xref: /libCEED/doc/sphinx/source/releasenotes.md (revision 7e7773b5b3fd61233915b5fe058a2730d1639517)
1bcb2dfaeSJed Brown# Changes/Release Notes
2bcb2dfaeSJed Brown
3bcb2dfaeSJed BrownOn this page we provide a summary of the main API changes, new features and examples
4bcb2dfaeSJed Brownfor each release of libCEED.
5bcb2dfaeSJed Brown
6bcb2dfaeSJed Brown(main)=
7bcb2dfaeSJed Brown
8bcb2dfaeSJed Brown## Current `main` branch
9bcb2dfaeSJed Brown
10*7e7773b5SJeremy L Thompson### Interface changes
11*7e7773b5SJeremy L Thompson
12*7e7773b5SJeremy L Thompson- Update {c:func} `CeedQFunctionGetFields` and {c:func} `CeedOperatorGetFields` to include number of fields.
13*7e7773b5SJeremy L Thompson
14bcb2dfaeSJed Brown### Maintainability
15bcb2dfaeSJed Brown
16bcb2dfaeSJed Brown- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage.
17bcb2dfaeSJed Brown
18bcb2dfaeSJed Brown(v0-9)=
19bcb2dfaeSJed Brown
20bcb2dfaeSJed Brown## v0.9 (Jul 6, 2021)
21bcb2dfaeSJed Brown
22bcb2dfaeSJed Brown### Interface changes
23bcb2dfaeSJed Brown
24bcb2dfaeSJed Brown- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact.
25bcb2dfaeSJed Brown
26bcb2dfaeSJed Brown### New features
27bcb2dfaeSJed Brown
28bcb2dfaeSJed Brown- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use.
29bcb2dfaeSJed Brown- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends.
30bcb2dfaeSJed Brown- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations.
31bcb2dfaeSJed Brown- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed).
32bcb2dfaeSJed Brown
33bcb2dfaeSJed Brown### Performance improvements
34bcb2dfaeSJed Brown
35bcb2dfaeSJed Brown### Examples
36bcb2dfaeSJed Brown
37bcb2dfaeSJed Brown- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations.
38bcb2dfaeSJed Brown- Fluid mechanics example adds GPU support and improves modularity.
39bcb2dfaeSJed Brown
40bcb2dfaeSJed Brown### Deprecated backends
41bcb2dfaeSJed Brown
42bcb2dfaeSJed Brown- 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.
43bcb2dfaeSJed Brown
44bcb2dfaeSJed Brown(v0-8)=
45bcb2dfaeSJed Brown
46bcb2dfaeSJed Brown## v0.8 (Mar 31, 2021)
47bcb2dfaeSJed Brown
48bcb2dfaeSJed Brown### Interface changes
49bcb2dfaeSJed Brown
50bcb2dfaeSJed Brown- Error handling improved to include enumerated error codes for C interface return values.
51bcb2dfaeSJed Brown- 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.
52bcb2dfaeSJed Brown
53bcb2dfaeSJed Brown### New features
54bcb2dfaeSJed Brown
55bcb2dfaeSJed Brown- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features.
56bcb2dfaeSJed Brown- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly.
57bcb2dfaeSJed Brown- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators.
58bcb2dfaeSJed Brown
59bcb2dfaeSJed Brown### Performance improvements
60bcb2dfaeSJed Brown
61bcb2dfaeSJed Brown- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`.
62bcb2dfaeSJed Brown- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`.
63bcb2dfaeSJed Brown
64bcb2dfaeSJed Brown### Examples
65bcb2dfaeSJed Brown
66bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions.
67bcb2dfaeSJed Brown- {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.
68bcb2dfaeSJed Brown- {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.
69bcb2dfaeSJed Brown- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree.
70bcb2dfaeSJed Brown
71bcb2dfaeSJed Brown(v0-7)=
72bcb2dfaeSJed Brown
73bcb2dfaeSJed Brown## v0.7 (Sep 29, 2020)
74bcb2dfaeSJed Brown
75bcb2dfaeSJed Brown### Interface changes
76bcb2dfaeSJed Brown
77bcb2dfaeSJed Brown- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors.
78bcb2dfaeSJed Brown  As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`.
79bcb2dfaeSJed Brown  These changes improve support for mixed finite element methods.
80bcb2dfaeSJed Brown- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature.
81bcb2dfaeSJed Brown- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity.
82bcb2dfaeSJed Brown- 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.
83bcb2dfaeSJed Brown- 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.
84bcb2dfaeSJed Brown- 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.
85bcb2dfaeSJed Brown- 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.
86bcb2dfaeSJed Brown  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`.
87bcb2dfaeSJed Brown- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory.
88bcb2dfaeSJed Brown- 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.
89bcb2dfaeSJed Brown
90bcb2dfaeSJed Brown### New features
91bcb2dfaeSJed Brown
92bcb2dfaeSJed Brown- New HIP backend: `/gpu/hip/ref`.
93bcb2dfaeSJed Brown- CeedQFunction support for user `CUfunction`s in some backends
94bcb2dfaeSJed Brown
95bcb2dfaeSJed Brown### Performance improvements
96bcb2dfaeSJed Brown
97bcb2dfaeSJed Brown- OCCA backend rebuilt to facilitate future performance enhancements.
98bcb2dfaeSJed Brown- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`.
99bcb2dfaeSJed Brown
100bcb2dfaeSJed Brown### Examples
101bcb2dfaeSJed Brown
102bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions.
103bcb2dfaeSJed Brown
104bcb2dfaeSJed Brown### Deprecated backends
105bcb2dfaeSJed Brown
106bcb2dfaeSJed Brown- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`.
107bcb2dfaeSJed Brown
108bcb2dfaeSJed Brown(v0-6)=
109bcb2dfaeSJed Brown
110bcb2dfaeSJed Brown## v0.6 (Mar 29, 2020)
111bcb2dfaeSJed Brown
112bcb2dfaeSJed BrownlibCEED v0.6 contains numerous new features and examples, as well as expanded
113bcb2dfaeSJed Browndocumentation in [this new website](https://libceed.readthedocs.io).
114bcb2dfaeSJed Brown
115bcb2dfaeSJed Brown### New features
116bcb2dfaeSJed Brown
117bcb2dfaeSJed Brown- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly
118bcb2dfaeSJed Brown  1-1 correspondence with the C interface, plus some convenience features.  For instance,
119bcb2dfaeSJed Brown  data stored in the {cpp:type}`CeedVector` structure are available without copy as
120bcb2dfaeSJed Brown  {py:class}`numpy.ndarray`.  Short tutorials are provided in
121bcb2dfaeSJed Brown  [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/).
122bcb2dfaeSJed Brown- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point,
123bcb2dfaeSJed Brown  {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal
124bcb2dfaeSJed Brown  ({c:func}`CeedOperatorAssembleLinearDiagonal`).  These operations are useful for
125bcb2dfaeSJed Brown  preconditioning ingredients and are used in the libCEED's multigrid examples.
126bcb2dfaeSJed Brown- The inverse of separable operators can be obtained using
127bcb2dfaeSJed Brown  {c:func}`CeedOperatorCreateFDMElementInverse` and applied with
128bcb2dfaeSJed Brown  {c:func}`CeedOperatorApply`.  This is a useful preconditioning ingredient,
129bcb2dfaeSJed Brown  especially for Laplacians and related operators.
130bcb2dfaeSJed Brown- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`,
131bcb2dfaeSJed Brown  {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`.
132bcb2dfaeSJed Brown- Make public accessors for various attributes to facilitate writing composable code.
133bcb2dfaeSJed Brown- New backend: `/cpu/self/memcheck/serial`.
134bcb2dfaeSJed Brown- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA
135bcb2dfaeSJed Brown  backends.  (Single source is coming soon for OCCA backends.)
136bcb2dfaeSJed Brown- Fix some missing edge cases in CUDA backend.
137bcb2dfaeSJed Brown
138bcb2dfaeSJed Brown### Performance Improvements
139bcb2dfaeSJed Brown
140bcb2dfaeSJed Brown- MAGMA backend performance optimization and non-tensor bases.
141bcb2dfaeSJed Brown- No-copy optimization in {c:func}`CeedOperatorApply`.
142bcb2dfaeSJed Brown
143bcb2dfaeSJed Brown### Interface changes
144bcb2dfaeSJed Brown
145bcb2dfaeSJed Brown- Replace {code}`CeedElemRestrictionCreateIdentity` and
146bcb2dfaeSJed Brown  {code}`CeedElemRestrictionCreateBlocked` with more flexible
147bcb2dfaeSJed Brown  {c:func}`CeedElemRestrictionCreateStrided` and
148bcb2dfaeSJed Brown  {c:func}`CeedElemRestrictionCreateBlockedStrided`.
149bcb2dfaeSJed Brown- Add arguments to {c:func}`CeedQFunctionCreateIdentity`.
150bcb2dfaeSJed Brown- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification
151bcb2dfaeSJed Brown  with {cpp:enum}`CeedInterlaceMode`.  This is now an attribute of the
152bcb2dfaeSJed Brown  {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no
153bcb2dfaeSJed Brown  longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and
154bcb2dfaeSJed Brown  {c:func}`CeedElemRestrictionApply`.
155bcb2dfaeSJed Brown
156bcb2dfaeSJed Brown### Examples
157bcb2dfaeSJed Brown
158bcb2dfaeSJed BrownlibCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`.
159bcb2dfaeSJed BrownNotable additions include:
160bcb2dfaeSJed Brown
161bcb2dfaeSJed Brown- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of
162bcb2dfaeSJed Brown  a domain in 1, 2, and 3 dimensions by applying a Laplacian.
163bcb2dfaeSJed Brown
164bcb2dfaeSJed Brown- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area
165bcb2dfaeSJed Brown  of domains (like the cube and sphere) by direct integration on a surface mesh;
166bcb2dfaeSJed Brown  demonstrates geometric dimension different from topological dimension.
167bcb2dfaeSJed Brown
168bcb2dfaeSJed Brown- PETSc {ref}`example-petsc-bps`:
169bcb2dfaeSJed Brown
170bcb2dfaeSJed Brown  - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support.
171bcb2dfaeSJed Brown  - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements
172bcb2dfaeSJed Brown    and transparent CUDA support.
173bcb2dfaeSJed Brown  - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`):
174bcb2dfaeSJed Brown    generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric
175bcb2dfaeSJed Brown    dimension different from topological dimension.
176bcb2dfaeSJed Brown
177bcb2dfaeSJed Brown- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid
178bcb2dfaeSJed Brown  solver with algebraic multigrid coarse solve.
179bcb2dfaeSJed Brown
180bcb2dfaeSJed Brown- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly
181bcb2dfaeSJed Brown  `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`),
182bcb2dfaeSJed Brown  implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and
183bcb2dfaeSJed Brown  quasi-2D computational domain support.
184bcb2dfaeSJed Brown
185bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for
186bcb2dfaeSJed Brown  linear elasticity, small-strain hyperelasticity, and globalized finite-strain
187bcb2dfaeSJed Brown  hyperelasticity using p-multigrid with algebraic multigrid coarse solve.
188bcb2dfaeSJed Brown
189bcb2dfaeSJed Brown(v0-5)=
190bcb2dfaeSJed Brown
191bcb2dfaeSJed Brown## v0.5 (Sep 18, 2019)
192bcb2dfaeSJed Brown
193bcb2dfaeSJed BrownFor this release, several improvements were made. Two new CUDA backends were added to
194bcb2dfaeSJed Brownthe family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art
195bcb2dfaeSJed Brownperformance using single-source {ref}`CeedQFunction`. From this release, users
196bcb2dfaeSJed Browncan define Q-Functions in a single source code independently of the targeted backend
197bcb2dfaeSJed Brownwith the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU
198bcb2dfaeSJed Browncompilation of the user provided {ref}`CeedQFunction` code. To allow a unified
199bcb2dfaeSJed Browndeclaration, the {ref}`CeedQFunction` API has undergone a slight change:
200bcb2dfaeSJed Brownthe `QFunctionField` parameter `ncomp` has been changed to `size`. This change
201bcb2dfaeSJed Brownrequires setting the previous value of `ncomp` to `ncomp*dim` when adding a
202bcb2dfaeSJed Brown`QFunctionField` with eval mode `CEED EVAL GRAD`.
203bcb2dfaeSJed Brown
204bcb2dfaeSJed BrownAdditionally, new CPU backends
205bcb2dfaeSJed Brownwere included in this release, such as the `/cpu/self/opt/*` backends (which are
206bcb2dfaeSJed Brownwritten in pure C and use partial **E-vectors** to improve performance) and the
207bcb2dfaeSJed Brown`/cpu/self/ref/memcheck` backend (which relies upon the
208bcb2dfaeSJed Brown[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user
209bcb2dfaeSJed Brown{ref}`CeedQFunction` have no undefined values).
210bcb2dfaeSJed BrownThis release also included various performance improvements, bug fixes, new examples,
211bcb2dfaeSJed Brownand improved tests. Among these improvements, vectorized instructions for
212bcb2dfaeSJed Brown{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD`
213bcb2dfaeSJed Browninstead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and
214bcb2dfaeSJed Brownidentity Q-Functions were introduced, and the PETSc benchmark problems were expanded
215bcb2dfaeSJed Brownto include unstructured meshes handling were. For this expansion, the prior version of
216bcb2dfaeSJed Brownthe PETSc BPs, which only included data associated with structured geometries, were
217bcb2dfaeSJed Brownrenamed `bpsraw`, and the new version of the BPs, which can handle data associated
218bcb2dfaeSJed Brownwith any unstructured geometry, were called `bps`. Additionally, other benchmark
219bcb2dfaeSJed Brownproblems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively),
220bcb2dfaeSJed Brownand BP5 and BP6 (the collocated versions---for which the quadrature points are the same
221bcb2dfaeSJed Brownas the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc
222bcb2dfaeSJed Brownexamples. Furthermoew, another standalone libCEED example, called `ex2`, which
223bcb2dfaeSJed Browncomputes the surface area of a given mesh was added to this release.
224bcb2dfaeSJed Brown
225bcb2dfaeSJed BrownBackends available in this release:
226bcb2dfaeSJed Brown
22768e843eeSJed Brown| CEED resource (`-ceed`)  | Backend                                             |
22868e843eeSJed Brown|--------------------------|-----------------------------------------------------|
22968e843eeSJed Brown| `/cpu/self/ref/serial`   | Serial reference implementation                     |
23068e843eeSJed Brown| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
23168e843eeSJed Brown| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks            |
23268e843eeSJed Brown| `/cpu/self/opt/serial`   | Serial optimized C implementation                   |
23368e843eeSJed Brown| `/cpu/self/opt/blocked`  | Blocked optimized C implementation                  |
23468e843eeSJed Brown| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
23568e843eeSJed Brown| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
23668e843eeSJed Brown| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
23768e843eeSJed Brown| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
23868e843eeSJed Brown| `/cpu/occa`              | Serial OCCA kernels                                 |
23968e843eeSJed Brown| `/gpu/occa`              | CUDA OCCA kernels                                   |
24068e843eeSJed Brown| `/omp/occa`              | OpenMP OCCA kernels                                 |
24168e843eeSJed Brown| `/ocl/occa`              | OpenCL OCCA kernels                                 |
24268e843eeSJed Brown| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
24368e843eeSJed Brown| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
24468e843eeSJed Brown| `/gpu/cuda/shared`       | Optimized pure CUDA kernels using shared memory     |
24568e843eeSJed Brown| `/gpu/cuda/gen`          | Optimized pure CUDA kernels using code generation   |
24668e843eeSJed Brown| `/gpu/magma`             | CUDA MAGMA kernels                                  |
247bcb2dfaeSJed Brown
248bcb2dfaeSJed BrownExamples available in this release:
249bcb2dfaeSJed Brown
25068e843eeSJed Brown:::{list-table}
25168e843eeSJed Brown:header-rows: 1
25268e843eeSJed Brown:widths: auto
25368e843eeSJed Brown* - User code
25468e843eeSJed Brown  - Example
25568e843eeSJed Brown* - `ceed`
25668e843eeSJed Brown  - * ex1 (volume)
25768e843eeSJed Brown    * ex2 (surface)
25868e843eeSJed Brown* - `mfem`
25968e843eeSJed Brown  - * BP1 (scalar mass operator)
26068e843eeSJed Brown    * BP3 (scalar Laplace operator)
26168e843eeSJed Brown* - `petsc`
26268e843eeSJed Brown  - * BP1 (scalar mass operator)
26368e843eeSJed Brown    * BP2 (vector mass operator)
26468e843eeSJed Brown    * BP3 (scalar Laplace operator)
26568e843eeSJed Brown    * BP4 (vector Laplace operator)
26668e843eeSJed Brown    * BP5 (collocated scalar Laplace operator)
26768e843eeSJed Brown    * BP6 (collocated vector Laplace operator)
26868e843eeSJed Brown    * Navier-Stokes
26968e843eeSJed Brown* - `nek5000`
27068e843eeSJed Brown  - * BP1 (scalar mass operator)
27168e843eeSJed Brown    * BP3 (scalar Laplace operator)
27268e843eeSJed Brown:::
273bcb2dfaeSJed Brown
274bcb2dfaeSJed Brown(v0-4)=
275bcb2dfaeSJed Brown
276bcb2dfaeSJed Brown## v0.4 (Apr 1, 2019)
277bcb2dfaeSJed Brown
278bcb2dfaeSJed BrownlibCEED v0.4 was made again publicly available in the second full CEED software
279bcb2dfaeSJed Browndistribution, release CEED 2.0. This release contained notable features, such as
280bcb2dfaeSJed Brownfour new CPU backends, two new GPU backends, CPU backend optimizations, initial
281bcb2dfaeSJed Brownsupport for operator composition, performance benchmarking, and a Navier-Stokes demo.
282bcb2dfaeSJed BrownThe new CPU backends in this release came in two families. The `/cpu/self/*/serial`
283bcb2dfaeSJed Brownbackends process one element at a time and are intended for meshes with a smaller number
284bcb2dfaeSJed Brownof high order elements. The `/cpu/self/*/blocked` backends process blocked batches of
285bcb2dfaeSJed Browneight interlaced elements and are intended for meshes with higher numbers of elements.
286bcb2dfaeSJed BrownThe `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU
287bcb2dfaeSJed Brownperformance. The `/cpu/self/xsmm/*` backends rely upon the
288bcb2dfaeSJed Brown[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU
289bcb2dfaeSJed Brownperformance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA.
290bcb2dfaeSJed BrownThe `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable
291bcb2dfaeSJed Brownperformance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple
292bcb2dfaeSJed Brownparallelization approach, where each thread treats a finite element. Using just in time
293bcb2dfaeSJed Browncompilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this
294bcb2dfaeSJed Brownbackend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend
295bcb2dfaeSJed Brownachieve good peak performance for 1D, 2D, and low order 3D problems, but performance
296bcb2dfaeSJed Browndeteriorates very quickly when threads run out of registers.
297bcb2dfaeSJed Brown
298bcb2dfaeSJed BrownA new explicit time-stepping Navier-Stokes solver was added to the family of libCEED
299bcb2dfaeSJed Brownexamples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`).
300bcb2dfaeSJed BrownThis example solves the time-dependent Navier-Stokes equations of compressible gas
301bcb2dfaeSJed Browndynamics in a static Eulerian three-dimensional frame, using structured high-order
302bcb2dfaeSJed Brownfinite/spectral element spatial discretizations and explicit high-order time-stepping
303bcb2dfaeSJed Brown(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc,
304bcb2dfaeSJed Brownso that the pointwise physics (defined at quadrature points) is separated from the
305bcb2dfaeSJed Brownparallelization and meshing concerns.
306bcb2dfaeSJed Brown
307bcb2dfaeSJed BrownBackends available in this release:
308bcb2dfaeSJed Brown
30968e843eeSJed Brown| CEED resource (`-ceed`)  | Backend                                             |
31068e843eeSJed Brown|--------------------------|-----------------------------------------------------|
31168e843eeSJed Brown| `/cpu/self/ref/serial`   | Serial reference implementation                     |
31268e843eeSJed Brown| `/cpu/self/ref/blocked`  | Blocked reference implementation                    |
31368e843eeSJed Brown| `/cpu/self/tmpl`         | Backend template, defaults to `/cpu/self/blocked`   |
31468e843eeSJed Brown| `/cpu/self/avx/serial`   | Serial AVX implementation                           |
31568e843eeSJed Brown| `/cpu/self/avx/blocked`  | Blocked AVX implementation                          |
31668e843eeSJed Brown| `/cpu/self/xsmm/serial`  | Serial LIBXSMM implementation                       |
31768e843eeSJed Brown| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation                      |
31868e843eeSJed Brown| `/cpu/occa`              | Serial OCCA kernels                                 |
31968e843eeSJed Brown| `/gpu/occa`              | CUDA OCCA kernels                                   |
32068e843eeSJed Brown| `/omp/occa`              | OpenMP OCCA kernels                                 |
32168e843eeSJed Brown| `/ocl/occa`              | OpenCL OCCA kernels                                 |
32268e843eeSJed Brown| `/gpu/cuda/ref`          | Reference pure CUDA kernels                         |
32368e843eeSJed Brown| `/gpu/cuda/reg`          | Pure CUDA kernels using one thread per element      |
32468e843eeSJed Brown| `/gpu/magma`             | CUDA MAGMA kernels                                  |
325bcb2dfaeSJed Brown
326bcb2dfaeSJed BrownExamples available in this release:
327bcb2dfaeSJed Brown
32868e843eeSJed Brown:::{list-table}
32968e843eeSJed Brown:header-rows: 1
33068e843eeSJed Brown:widths: auto
33168e843eeSJed Brown* - User code
33268e843eeSJed Brown  - Example
33368e843eeSJed Brown* - `ceed`
33468e843eeSJed Brown  - * ex1 (volume)
33568e843eeSJed Brown* - `mfem`
33668e843eeSJed Brown  - * BP1 (scalar mass operator)
33768e843eeSJed Brown    * BP3 (scalar Laplace operator)
33868e843eeSJed Brown* - `petsc`
33968e843eeSJed Brown  - * BP1 (scalar mass operator)
34068e843eeSJed Brown    * BP3 (scalar Laplace operator)
34168e843eeSJed Brown    * Navier-Stokes
34268e843eeSJed Brown* - `nek5000`
34368e843eeSJed Brown  - * BP1 (scalar mass operator)
34468e843eeSJed Brown    * BP3 (scalar Laplace operator)
34568e843eeSJed Brown:::
346bcb2dfaeSJed Brown
347bcb2dfaeSJed Brown(v0-3)=
348bcb2dfaeSJed Brown
349bcb2dfaeSJed Brown## v0.3 (Sep 30, 2018)
350bcb2dfaeSJed Brown
351bcb2dfaeSJed BrownNotable features in this release include active/passive field interface, support for
352bcb2dfaeSJed Brownnon-tensor bases, backend optimization, and improved Fortran interface. This release
353bcb2dfaeSJed Brownalso focused on providing improved continuous integration, and many new tests with code
354bcb2dfaeSJed Browncoverage reports of about 90%. This release also provided a significant change to the
355bcb2dfaeSJed Brownpublic interface: a {ref}`CeedQFunction` can take any number of named input and output
356bcb2dfaeSJed Brownarguments while {ref}`CeedOperator` connects them to the actual data, which may be
357bcb2dfaeSJed Brownsupplied explicitly to `CeedOperatorApply()` (active) or separately via
358bcb2dfaeSJed Brown`CeedOperatorSetField()` (passive). This interface change enables reusable libraries
359bcb2dfaeSJed Brownof CeedQFunctions and composition of block solvers constructed using
360bcb2dfaeSJed Brown{ref}`CeedOperator`. A concept of blocked restriction was added to this release and
361bcb2dfaeSJed Brownused in an optimized CPU backend. Although this is typically not visible to the user,
362bcb2dfaeSJed Brownit enables effective use of arbitrary-length SIMD while maintaining cache locality.
363bcb2dfaeSJed BrownThis CPU backend also implements an algebraic factorization of tensor product gradients
364bcb2dfaeSJed Brownto perform fewer operations than standard application of interpolation and
365bcb2dfaeSJed Browndifferentiation from nodes to quadrature points. This algebraic formulation
366bcb2dfaeSJed Brownautomatically supports non-polynomial and non-interpolatory bases, thus is more general
367bcb2dfaeSJed Brownthan the more common derivation in terms of Lagrange polynomials on the quadrature points.
368bcb2dfaeSJed Brown
369bcb2dfaeSJed BrownBackends available in this release:
370bcb2dfaeSJed Brown
37168e843eeSJed Brown| CEED resource (`-ceed`) | Backend                                             |
37268e843eeSJed Brown|-------------------------|-----------------------------------------------------|
37368e843eeSJed Brown| `/cpu/self/blocked`     | Blocked reference implementation                    |
37468e843eeSJed Brown| `/cpu/self/ref`         | Serial reference implementation                     |
37568e843eeSJed Brown| `/cpu/self/tmpl`        | Backend template, defaults to `/cpu/self/blocked`   |
37668e843eeSJed Brown| `/cpu/occa`             | Serial OCCA kernels                                 |
37768e843eeSJed Brown| `/gpu/occa`             | CUDA OCCA kernels                                   |
37868e843eeSJed Brown| `/omp/occa`             | OpenMP OCCA kernels                                 |
37968e843eeSJed Brown| `/ocl/occa`             | OpenCL OCCA kernels                                 |
38068e843eeSJed Brown| `/gpu/magma`            | CUDA MAGMA kernels                                  |
381bcb2dfaeSJed Brown
382bcb2dfaeSJed BrownExamples available in this release:
383bcb2dfaeSJed Brown
38468e843eeSJed Brown:::{list-table}
38568e843eeSJed Brown:header-rows: 1
38668e843eeSJed Brown:widths: auto
38768e843eeSJed Brown* - User code
38868e843eeSJed Brown  - Example
38968e843eeSJed Brown* - `ceed`
39068e843eeSJed Brown  - * ex1 (volume)
39168e843eeSJed Brown* - `mfem`
39268e843eeSJed Brown  - * BP1 (scalar mass operator)
39368e843eeSJed Brown    * BP3 (scalar Laplace operator)
39468e843eeSJed Brown* - `petsc`
39568e843eeSJed Brown  - * BP1 (scalar mass operator)
39668e843eeSJed Brown    * BP3 (scalar Laplace operator)
39768e843eeSJed Brown* - `nek5000`
39868e843eeSJed Brown  - * BP1 (scalar mass operator)
39968e843eeSJed Brown    * BP3 (scalar Laplace operator)
40068e843eeSJed Brown:::
401bcb2dfaeSJed Brown
402bcb2dfaeSJed Brown(v0-21)=
403bcb2dfaeSJed Brown
404bcb2dfaeSJed Brown## v0.21 (Sep 30, 2018)
405bcb2dfaeSJed Brown
406bcb2dfaeSJed BrownA MAGMA backend (which relies upon the
407bcb2dfaeSJed Brown[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this
408bcb2dfaeSJed Brownrelease. This initial integration set up the framework of using MAGMA and provided the
409bcb2dfaeSJed BrownlibCEED functionality through MAGMA kernels as one of libCEED’s computational backends.
410bcb2dfaeSJed BrownAs any other backend, the MAGMA backend provides extended basic data structures for
411bcb2dfaeSJed Brown{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements
412bcb2dfaeSJed Brownthe fundamental CEED building blocks to work with the new data structures.
413bcb2dfaeSJed BrownIn general, the MAGMA-specific data structures keep the libCEED pointers to CPU data
414bcb2dfaeSJed Brownbut also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled
415bcb2dfaeSJed Browninternally, and thus seamlessly to the user, through the functions/methods that are
416bcb2dfaeSJed Brownprovided to support them.
417bcb2dfaeSJed Brown
418bcb2dfaeSJed BrownBackends available in this release:
419bcb2dfaeSJed Brown
42068e843eeSJed Brown| CEED resource (`-ceed`) | Backend                         |
42168e843eeSJed Brown|-------------------------|---------------------------------|
42268e843eeSJed Brown| `/cpu/self`             | Serial reference implementation |
42368e843eeSJed Brown| `/cpu/occa`             | Serial OCCA kernels             |
42468e843eeSJed Brown| `/gpu/occa`             | CUDA OCCA kernels               |
42568e843eeSJed Brown| `/omp/occa`             | OpenMP OCCA kernels             |
42668e843eeSJed Brown| `/ocl/occa`             | OpenCL OCCA kernels             |
42768e843eeSJed Brown| `/gpu/magma`            | CUDA MAGMA kernels              |
428bcb2dfaeSJed Brown
429bcb2dfaeSJed BrownExamples available in this release:
430bcb2dfaeSJed Brown
43168e843eeSJed Brown:::{list-table}
43268e843eeSJed Brown:header-rows: 1
43368e843eeSJed Brown:widths: auto
43468e843eeSJed Brown* - User code
43568e843eeSJed Brown  - Example
43668e843eeSJed Brown* - `ceed`
43768e843eeSJed Brown  - * ex1 (volume)
43868e843eeSJed Brown* - `mfem`
43968e843eeSJed Brown  - * BP1 (scalar mass operator)
44068e843eeSJed Brown    * BP3 (scalar Laplace operator)
44168e843eeSJed Brown* - `petsc`
44268e843eeSJed Brown  - * BP1 (scalar mass operator)
44368e843eeSJed Brown* - `nek5000`
44468e843eeSJed Brown  - * BP1 (scalar mass operator)
44568e843eeSJed Brown:::
446bcb2dfaeSJed Brown
447bcb2dfaeSJed Brown(v0-2)=
448bcb2dfaeSJed Brown
449bcb2dfaeSJed Brown## v0.2 (Mar 30, 2018)
450bcb2dfaeSJed Brown
451bcb2dfaeSJed BrownlibCEED was made publicly available the first full CEED software distribution, release
452bcb2dfaeSJed BrownCEED 1.0. The distribution was made available using the Spack package manager to provide
453bcb2dfaeSJed Browna common, easy-to-use build environment, where the user can build the CEED distribution
454bcb2dfaeSJed Brownwith all dependencies. This release included a new Fortran interface for the library.
455bcb2dfaeSJed BrownThis release also contained major improvements in the OCCA backend (including a new
456bcb2dfaeSJed Brown`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to
457bcb2dfaeSJed Browncompute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an
458bcb2dfaeSJed Brown`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application
459bcb2dfaeSJed Brownof the Laplace operator) was also added to this release.
460bcb2dfaeSJed Brown
461bcb2dfaeSJed BrownBackends available in this release:
462bcb2dfaeSJed Brown
46368e843eeSJed Brown| CEED resource (`-ceed`) | Backend                         |
46468e843eeSJed Brown|-------------------------|---------------------------------|
46568e843eeSJed Brown| `/cpu/self`             | Serial reference implementation |
46668e843eeSJed Brown| `/cpu/occa`             | Serial OCCA kernels             |
46768e843eeSJed Brown| `/gpu/occa`             | CUDA OCCA kernels               |
46868e843eeSJed Brown| `/omp/occa`             | OpenMP OCCA kernels             |
46968e843eeSJed Brown| `/ocl/occa`             | OpenCL OCCA kernels             |
470bcb2dfaeSJed Brown
471bcb2dfaeSJed BrownExamples available in this release:
472bcb2dfaeSJed Brown
47368e843eeSJed Brown:::{list-table}
47468e843eeSJed Brown:header-rows: 1
47568e843eeSJed Brown:widths: auto
47668e843eeSJed Brown* - User code
47768e843eeSJed Brown  - Example
47868e843eeSJed Brown* - `ceed`
47968e843eeSJed Brown  - * ex1 (volume)
48068e843eeSJed Brown* - `mfem`
48168e843eeSJed Brown  - * BP1 (scalar mass operator)
48268e843eeSJed Brown    * BP3 (scalar Laplace operator)
48368e843eeSJed Brown* - `petsc`
48468e843eeSJed Brown  - * BP1 (scalar mass operator)
48568e843eeSJed Brown* - `nek5000`
48668e843eeSJed Brown  - * BP1 (scalar mass operator)
48768e843eeSJed Brown:::
488bcb2dfaeSJed Brown
489bcb2dfaeSJed Brown(v0-1)=
490bcb2dfaeSJed Brown
491bcb2dfaeSJed Brown## v0.1 (Jan 3, 2018)
492bcb2dfaeSJed Brown
493bcb2dfaeSJed BrownInitial low-level API of the CEED project. The low-level API provides a set of Finite
494bcb2dfaeSJed BrownElements kernels and components for writing new low-level kernels. Examples include:
495bcb2dfaeSJed Brownvector and sparse linear algebra, element matrix assembly over a batch of elements,
496bcb2dfaeSJed Brownpartial assembly and action for efficient high-order operators like mass, diffusion,
497bcb2dfaeSJed Brownadvection, etc. The main goal of the low-level API is to establish the basis for the
498bcb2dfaeSJed Brownhigh-level API. Also, identifying such low-level kernels and providing a reference
499bcb2dfaeSJed Brownimplementation for them serves as the basis for specialized backend implementations.
500bcb2dfaeSJed BrownThis release contained several backends: `/cpu/self`, and backends which rely upon the
501bcb2dfaeSJed Brown[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`,
502bcb2dfaeSJed Brown`/gpu/occa`, and `/omp/occa`.
503bcb2dfaeSJed BrownIt also included several examples, in the `examples` folder:
504bcb2dfaeSJed BrownA standalone code that shows the usage of libCEED (with no external
505bcb2dfaeSJed Browndependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1
506bcb2dfaeSJed Brown(with the application of the mass operator); and a `petsc` example to perform BP1
507bcb2dfaeSJed Brown(with the application of the mass operator).
508bcb2dfaeSJed Brown
509bcb2dfaeSJed BrownBackends available in this release:
510bcb2dfaeSJed Brown
51168e843eeSJed Brown| CEED resource (`-ceed`) | Backend                         |
51268e843eeSJed Brown|-------------------------|---------------------------------|
51368e843eeSJed Brown| `/cpu/self`             | Serial reference implementation |
51468e843eeSJed Brown| `/cpu/occa`             | Serial OCCA kernels             |
51568e843eeSJed Brown| `/gpu/occa`             | CUDA OCCA kernels               |
51668e843eeSJed Brown| `/omp/occa`             | OpenMP OCCA kernels             |
517bcb2dfaeSJed Brown
518bcb2dfaeSJed BrownExamples available in this release:
519bcb2dfaeSJed Brown
520bcb2dfaeSJed Brown| User code             | Example                           |
52168e843eeSJed Brown|-----------------------|-----------------------------------|
52268e843eeSJed Brown| `ceed`                | ex1 (scalar Laplace operator)     |
52368e843eeSJed Brown| `mfem`                | BP1 (scalar mass operator)        |
52468e843eeSJed Brown| `petsc`               | BP1 (scalar mass operator)        |
525bcb2dfaeSJed Brown```
526