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