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