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