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- Update `CeedOperatorContext*` functions to `CeedOperator*Context*` functions for consistency. 12For example, `CeedOperatorContextGetFieldLabel` was renamed to `CeedOperatorGetContextFieldLabel`. 13 14### New features 15 16- Added {c:func}`CeedOperatorGetFieldByName` to access a specific `CeedOperatorField` by its name 17- Update `/cpu/self/memcheck/*` backends to help verify `CeedVector` array access assumptions and `CeedQFunction` user output assumptions. 18- Update {c:func}`CeedOperatorLinearAssembleDiagonal` to provide default implementation that supports `CeedOperator` with multiple active bases. 19 20### Examples 21 22#### {ref}`example-petsc-bps` 23 24- Requires PETSc version 3.19 or later 25 26(v0-11)= 27 28## v0.11 (Dec 24, 2022) 29 30### Interface changes 31 32- Added {c:func}`CeedOperatorSetName` for more readable {c:func}`CeedOperatorView` output. 33- Added {c:func}`CeedBasisCreateProjection` to facilitate interpolation between nodes for separate `CeedBases`. 34- Rename and move {c:func}`CeedCompositeOperatorGetNumSub` and {c:func}`CeedCompositeOperatorGetSubList` to public interface. 35 36### New features 37 38- Update `/cpu/self/memcheck/*` backends to help verify `CeedQFunctionContext` data sizes provided by user. 39- Improved support for $H(\text{div})$ bases. 40- Added `CeedInt_FMT` to support potential future use of larger integer sizes. 41- Added `CEED_QFUNCTION_ATTR` for setting compiler attributes/pragmas to `CEED_QFUNCTION_HELPER` and `CEED_QFUNCTION`. 42- OCCA backend updated to latest OCCA release; DPC++ and OMP OCCA modes enabled. 43Due to a limitation of the OCCA parser, typedefs are required to use pointers to arrays in QFunctions with the OCCA backend. 44This issue will be fixed in a future OCCA release. 45 46### Bugfix 47 48- Fix bug in setting device id for GPU backends. 49- Fix storing of indices for `CeedElemRestriction` on the host with GPU backends. 50- Fix `CeedElemRestriction` sizing for {c:func}`CeedOperatorAssemblePointBlockDiagonal`. 51- Fix bugs in CPU implementation of {c:func}`CeedOperatorLinearAssemble` when there are different number of active input modes and active output modes. 52 53### Examples 54 55#### {ref}`example-petsc-navier-stokes` 56 57- Various performance enhancements, analytic matrix-free and assembled Jacobian, and PETSc solver configurations for GPUs. 58- Refactored to improve code reuse and modularity. 59- Support for primitive variables for more accurate boundary layers and all-speed flow. 60- Added $YZ\beta$ shock capturing scheme and Shock Tube example. 61- Added Channel example, with comparison to analytic solutions. 62- Added Flat Plate with boundary layer mesh and compressible Blasius inflow condition based on Chebyshev collocation solution of the Blasius equations. 63- Added strong and weak synthetic turbulence generation (STG) inflow boundary conditions. 64- Added "freestream" boundary conditions based on HLLC Riemann solver. 65- Automated stabilization coefficients for different basis degree. 66 67#### {ref}`example-petsc-bps` 68 69- Support for convergence studies. 70 71### Maintainability 72 73- 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. 74- Enabled support for `p > 8` for `/gpu/*/shared` backends. 75- Switch to `clang-format` over `astyle` for automatic formatting; Makefile command changed to `make format` from `make style`. 76- Improved test harness. 77 78(v0-10-1)= 79 80## v0.10.1 (Apr 11, 2022) 81 82### Interface changes 83 84- Added {c:func}`CeedQFunctionSetUserFlopsEstimate` and {c:func}`CeedOperatorGetFlopsEstimate` to facilitate estimating FLOPs in operator application. 85 86### New features 87 88- Switched MAGMA backends to use runtime compilation for tensor basis kernels (and element restriction kernels, in non-deterministic `/gpu/*/magma` backends). 89This reduces time to compile the library and increases the range of parameters for which the MAGMA tensor basis kernels will work. 90 91### Bugfix 92 93- Install JiT source files in install directory to fix GPU functionality for installed libCEED. 94 95(v0-10)= 96 97## v0.10 (Mar 21, 2022) 98 99### Interface changes 100 101- Update {c:func}`CeedQFunctionGetFields` and {c:func}`CeedOperatorGetFields` to include number of fields. 102- 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`. 103- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added. 104- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients. 105- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces. 106- Warning added when compiling OCCA backend to alert users that this backend is experimental. 107- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`. 108- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name. 109- 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`. 110- Added {c:func}`CeedVectorGetArrayWrite` that allows access to uninitalized arrays; require initalized data for {c:func}`CeedVectorGetArray`. 111- 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. 112- Added {c:func}`CeedQFunctionContextGetFieldDescriptions` to retreive user defined descriptions of fields that are registered with `CeedQFunctionContextRegister*`. 113- Renamed `CeedElemTopology` entries for clearer namespacing between libCEED enums. 114- 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. 115- Added {c:func}`CeedOperatorSetQFunctionUpdated` to facilitate QFunction data re-use between operators sharing the same quadrature space, such as in a multigrid hierarchy. 116- Added {c:func}`CeedOperatorGetActiveVectorLengths` to get shape of CeedOperator. 117 118### New features 119 120- `CeedScalar` can now be set as `float` or `double` at compile time. 121- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends. 122- 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. 123- Remove need to guard library headers in QFunction source for code generation backends. 124- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present. 125- Added {c:func}`CeedStringAllocCopy` to reduce repeated code for copying strings internally. 126- Added {c:func}`CeedPathConcatenate` to facilitate loading kernel source files with a path relative to the current file. 127- Added support for non-tensor H(div) elements, to include CPU backend implementations and {c:func}`CeedBasisCreateHdiv` convenience constructor. 128- 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. 129- Added support for element matrix assembly in GPU backends. 130 131### Maintainability 132 133- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage. 134- `Include-what-you-use` makefile target added as `make iwyu`. 135- Create backend constant `CEED_FIELD_MAX` to reduce magic numbers in codebase. 136- Put GPU JiTed kernel source code into separate files. 137- 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. 138 139(v0-9)= 140 141## v0.9 (Jul 6, 2021) 142 143### Interface changes 144 145- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact. 146 147### New features 148 149- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use. 150- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends. 151- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations. 152- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed). 153 154### Performance improvements 155 156### Examples 157 158- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations. 159- Fluid mechanics example adds GPU support and improves modularity. 160 161### Deprecated backends 162 163- 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. 164 165(v0-8)= 166 167## v0.8 (Mar 31, 2021) 168 169### Interface changes 170 171- Error handling improved to include enumerated error codes for C interface return values. 172- 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. 173 174### New features 175 176- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features. 177- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly. 178- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators. 179 180### Performance improvements 181 182- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`. 183- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`. 184 185### Examples 186 187- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions. 188- {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. 189- {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. 190- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree. 191 192(v0-7)= 193 194## v0.7 (Sep 29, 2020) 195 196### Interface changes 197 198- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors. 199 As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`. 200 These changes improve support for mixed finite element methods. 201- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature. 202- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity. 203- 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. 204- 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. 205- 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. 206- 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. 207 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`. 208- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory. 209- 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. 210 211### New features 212 213- New HIP backend: `/gpu/hip/ref`. 214- CeedQFunction support for user `CUfunction`s in some backends 215 216### Performance improvements 217 218- OCCA backend rebuilt to facilitate future performance enhancements. 219- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`. 220 221### Examples 222 223- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions. 224 225### Deprecated backends 226 227- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`. 228 229(v0-6)= 230 231## v0.6 (Mar 29, 2020) 232 233libCEED v0.6 contains numerous new features and examples, as well as expanded 234documentation in [this new website](https://libceed.org). 235 236### New features 237 238- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly 239 1-1 correspondence with the C interface, plus some convenience features. For instance, 240 data stored in the {cpp:type}`CeedVector` structure are available without copy as 241 {py:class}`numpy.ndarray`. Short tutorials are provided in 242 [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/). 243- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point, 244 {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal 245 ({c:func}`CeedOperatorAssembleLinearDiagonal`). These operations are useful for 246 preconditioning ingredients and are used in the libCEED's multigrid examples. 247- The inverse of separable operators can be obtained using 248 {c:func}`CeedOperatorCreateFDMElementInverse` and applied with 249 {c:func}`CeedOperatorApply`. This is a useful preconditioning ingredient, 250 especially for Laplacians and related operators. 251- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`, 252 {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`. 253- Make public accessors for various attributes to facilitate writing composable code. 254- New backend: `/cpu/self/memcheck/serial`. 255- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA 256 backends. (Single source is coming soon for OCCA backends.) 257- Fix some missing edge cases in CUDA backend. 258 259### Performance Improvements 260 261- MAGMA backend performance optimization and non-tensor bases. 262- No-copy optimization in {c:func}`CeedOperatorApply`. 263 264### Interface changes 265 266- Replace {code}`CeedElemRestrictionCreateIdentity` and 267 {code}`CeedElemRestrictionCreateBlocked` with more flexible 268 {c:func}`CeedElemRestrictionCreateStrided` and 269 {c:func}`CeedElemRestrictionCreateBlockedStrided`. 270- Add arguments to {c:func}`CeedQFunctionCreateIdentity`. 271- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification 272 with {cpp:enum}`CeedInterlaceMode`. This is now an attribute of the 273 {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no 274 longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and 275 {c:func}`CeedElemRestrictionApply`. 276 277### Examples 278 279libCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`. 280Notable additions include: 281 282- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of 283 a domain in 1, 2, and 3 dimensions by applying a Laplacian. 284 285- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area 286 of domains (like the cube and sphere) by direct integration on a surface mesh; 287 demonstrates geometric dimension different from topological dimension. 288 289- PETSc {ref}`example-petsc-bps`: 290 291 - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support. 292 - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements 293 and transparent CUDA support. 294 - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`): 295 generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric 296 dimension different from topological dimension. 297 298- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid 299 solver with algebraic multigrid coarse solve. 300 301- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly 302 `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`), 303 implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and 304 quasi-2D computational domain support. 305 306- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for 307 linear elasticity, small-strain hyperelasticity, and globalized finite-strain 308 hyperelasticity using p-multigrid with algebraic multigrid coarse solve. 309 310(v0-5)= 311 312## v0.5 (Sep 18, 2019) 313 314For this release, several improvements were made. Two new CUDA backends were added to 315the family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art 316performance using single-source {ref}`CeedQFunction`. From this release, users 317can define Q-Functions in a single source code independently of the targeted backend 318with the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU 319compilation of the user provided {ref}`CeedQFunction` code. To allow a unified 320declaration, the {ref}`CeedQFunction` API has undergone a slight change: 321the `QFunctionField` parameter `ncomp` has been changed to `size`. This change 322requires setting the previous value of `ncomp` to `ncomp*dim` when adding a 323`QFunctionField` with eval mode `CEED EVAL GRAD`. 324 325Additionally, new CPU backends 326were included in this release, such as the `/cpu/self/opt/*` backends (which are 327written in pure C and use partial **E-vectors** to improve performance) and the 328`/cpu/self/ref/memcheck` backend (which relies upon the 329[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user 330{ref}`CeedQFunction` have no undefined values). 331This release also included various performance improvements, bug fixes, new examples, 332and improved tests. Among these improvements, vectorized instructions for 333{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD` 334instead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and 335identity Q-Functions were introduced, and the PETSc benchmark problems were expanded 336to include unstructured meshes handling were. For this expansion, the prior version of 337the PETSc BPs, which only included data associated with structured geometries, were 338renamed `bpsraw`, and the new version of the BPs, which can handle data associated 339with any unstructured geometry, were called `bps`. Additionally, other benchmark 340problems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively), 341and BP5 and BP6 (the collocated versions---for which the quadrature points are the same 342as the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc 343examples. Furthermoew, another standalone libCEED example, called `ex2`, which 344computes the surface area of a given mesh was added to this release. 345 346Backends available in this release: 347 348| CEED resource (`-ceed`) | Backend | 349|--------------------------|-----------------------------------------------------| 350| `/cpu/self/ref/serial` | Serial reference implementation | 351| `/cpu/self/ref/blocked` | Blocked reference implementation | 352| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks | 353| `/cpu/self/opt/serial` | Serial optimized C implementation | 354| `/cpu/self/opt/blocked` | Blocked optimized C implementation | 355| `/cpu/self/avx/serial` | Serial AVX implementation | 356| `/cpu/self/avx/blocked` | Blocked AVX implementation | 357| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 358| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 359| `/cpu/occa` | Serial OCCA kernels | 360| `/gpu/occa` | CUDA OCCA kernels | 361| `/omp/occa` | OpenMP OCCA kernels | 362| `/ocl/occa` | OpenCL OCCA kernels | 363| `/gpu/cuda/ref` | Reference pure CUDA kernels | 364| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 365| `/gpu/cuda/shared` | Optimized pure CUDA kernels using shared memory | 366| `/gpu/cuda/gen` | Optimized pure CUDA kernels using code generation | 367| `/gpu/magma` | CUDA MAGMA kernels | 368 369Examples available in this release: 370 371:::{list-table} 372:header-rows: 1 373:widths: auto 374* - User code 375 - Example 376* - `ceed` 377 - * ex1 (volume) 378 * ex2 (surface) 379* - `mfem` 380 - * BP1 (scalar mass operator) 381 * BP3 (scalar Laplace operator) 382* - `petsc` 383 - * BP1 (scalar mass operator) 384 * BP2 (vector mass operator) 385 * BP3 (scalar Laplace operator) 386 * BP4 (vector Laplace operator) 387 * BP5 (collocated scalar Laplace operator) 388 * BP6 (collocated vector Laplace operator) 389 * Navier-Stokes 390* - `nek5000` 391 - * BP1 (scalar mass operator) 392 * BP3 (scalar Laplace operator) 393::: 394 395(v0-4)= 396 397## v0.4 (Apr 1, 2019) 398 399libCEED v0.4 was made again publicly available in the second full CEED software 400distribution, release CEED 2.0. This release contained notable features, such as 401four new CPU backends, two new GPU backends, CPU backend optimizations, initial 402support for operator composition, performance benchmarking, and a Navier-Stokes demo. 403The new CPU backends in this release came in two families. The `/cpu/self/*/serial` 404backends process one element at a time and are intended for meshes with a smaller number 405of high order elements. The `/cpu/self/*/blocked` backends process blocked batches of 406eight interlaced elements and are intended for meshes with higher numbers of elements. 407The `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU 408performance. The `/cpu/self/xsmm/*` backends rely upon the 409[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU 410performance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA. 411The `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable 412performance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple 413parallelization approach, where each thread treats a finite element. Using just in time 414compilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this 415backend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend 416achieve good peak performance for 1D, 2D, and low order 3D problems, but performance 417deteriorates very quickly when threads run out of registers. 418 419A new explicit time-stepping Navier-Stokes solver was added to the family of libCEED 420examples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`). 421This example solves the time-dependent Navier-Stokes equations of compressible gas 422dynamics in a static Eulerian three-dimensional frame, using structured high-order 423finite/spectral element spatial discretizations and explicit high-order time-stepping 424(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc, 425so that the pointwise physics (defined at quadrature points) is separated from the 426parallelization and meshing concerns. 427 428Backends available in this release: 429 430| CEED resource (`-ceed`) | Backend | 431|--------------------------|-----------------------------------------------------| 432| `/cpu/self/ref/serial` | Serial reference implementation | 433| `/cpu/self/ref/blocked` | Blocked reference implementation | 434| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 435| `/cpu/self/avx/serial` | Serial AVX implementation | 436| `/cpu/self/avx/blocked` | Blocked AVX implementation | 437| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 438| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 439| `/cpu/occa` | Serial OCCA kernels | 440| `/gpu/occa` | CUDA OCCA kernels | 441| `/omp/occa` | OpenMP OCCA kernels | 442| `/ocl/occa` | OpenCL OCCA kernels | 443| `/gpu/cuda/ref` | Reference pure CUDA kernels | 444| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 445| `/gpu/magma` | CUDA MAGMA kernels | 446 447Examples available in this release: 448 449:::{list-table} 450:header-rows: 1 451:widths: auto 452* - User code 453 - Example 454* - `ceed` 455 - * ex1 (volume) 456* - `mfem` 457 - * BP1 (scalar mass operator) 458 * BP3 (scalar Laplace operator) 459* - `petsc` 460 - * BP1 (scalar mass operator) 461 * BP3 (scalar Laplace operator) 462 * Navier-Stokes 463* - `nek5000` 464 - * BP1 (scalar mass operator) 465 * BP3 (scalar Laplace operator) 466::: 467 468(v0-3)= 469 470## v0.3 (Sep 30, 2018) 471 472Notable features in this release include active/passive field interface, support for 473non-tensor bases, backend optimization, and improved Fortran interface. This release 474also focused on providing improved continuous integration, and many new tests with code 475coverage reports of about 90%. This release also provided a significant change to the 476public interface: a {ref}`CeedQFunction` can take any number of named input and output 477arguments while {ref}`CeedOperator` connects them to the actual data, which may be 478supplied explicitly to `CeedOperatorApply()` (active) or separately via 479`CeedOperatorSetField()` (passive). This interface change enables reusable libraries 480of CeedQFunctions and composition of block solvers constructed using 481{ref}`CeedOperator`. A concept of blocked restriction was added to this release and 482used in an optimized CPU backend. Although this is typically not visible to the user, 483it enables effective use of arbitrary-length SIMD while maintaining cache locality. 484This CPU backend also implements an algebraic factorization of tensor product gradients 485to perform fewer operations than standard application of interpolation and 486differentiation from nodes to quadrature points. This algebraic formulation 487automatically supports non-polynomial and non-interpolatory bases, thus is more general 488than the more common derivation in terms of Lagrange polynomials on the quadrature points. 489 490Backends available in this release: 491 492| CEED resource (`-ceed`) | Backend | 493|-------------------------|-----------------------------------------------------| 494| `/cpu/self/blocked` | Blocked reference implementation | 495| `/cpu/self/ref` | Serial reference implementation | 496| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 497| `/cpu/occa` | Serial OCCA kernels | 498| `/gpu/occa` | CUDA OCCA kernels | 499| `/omp/occa` | OpenMP OCCA kernels | 500| `/ocl/occa` | OpenCL OCCA kernels | 501| `/gpu/magma` | CUDA MAGMA kernels | 502 503Examples available in this release: 504 505:::{list-table} 506:header-rows: 1 507:widths: auto 508* - User code 509 - Example 510* - `ceed` 511 - * ex1 (volume) 512* - `mfem` 513 - * BP1 (scalar mass operator) 514 * BP3 (scalar Laplace operator) 515* - `petsc` 516 - * BP1 (scalar mass operator) 517 * BP3 (scalar Laplace operator) 518* - `nek5000` 519 - * BP1 (scalar mass operator) 520 * BP3 (scalar Laplace operator) 521::: 522 523(v0-21)= 524 525## v0.21 (Sep 30, 2018) 526 527A MAGMA backend (which relies upon the 528[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this 529release. This initial integration set up the framework of using MAGMA and provided the 530libCEED functionality through MAGMA kernels as one of libCEED’s computational backends. 531As any other backend, the MAGMA backend provides extended basic data structures for 532{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements 533the fundamental CEED building blocks to work with the new data structures. 534In general, the MAGMA-specific data structures keep the libCEED pointers to CPU data 535but also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled 536internally, and thus seamlessly to the user, through the functions/methods that are 537provided to support them. 538 539Backends available in this release: 540 541| CEED resource (`-ceed`) | Backend | 542|-------------------------|---------------------------------| 543| `/cpu/self` | Serial reference implementation | 544| `/cpu/occa` | Serial OCCA kernels | 545| `/gpu/occa` | CUDA OCCA kernels | 546| `/omp/occa` | OpenMP OCCA kernels | 547| `/ocl/occa` | OpenCL OCCA kernels | 548| `/gpu/magma` | CUDA MAGMA kernels | 549 550Examples available in this release: 551 552:::{list-table} 553:header-rows: 1 554:widths: auto 555* - User code 556 - Example 557* - `ceed` 558 - * ex1 (volume) 559* - `mfem` 560 - * BP1 (scalar mass operator) 561 * BP3 (scalar Laplace operator) 562* - `petsc` 563 - * BP1 (scalar mass operator) 564* - `nek5000` 565 - * BP1 (scalar mass operator) 566::: 567 568(v0-2)= 569 570## v0.2 (Mar 30, 2018) 571 572libCEED was made publicly available the first full CEED software distribution, release 573CEED 1.0. The distribution was made available using the Spack package manager to provide 574a common, easy-to-use build environment, where the user can build the CEED distribution 575with all dependencies. This release included a new Fortran interface for the library. 576This release also contained major improvements in the OCCA backend (including a new 577`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to 578compute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an 579`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application 580of the Laplace operator) was also added to this release. 581 582Backends available in this release: 583 584| CEED resource (`-ceed`) | Backend | 585|-------------------------|---------------------------------| 586| `/cpu/self` | Serial reference implementation | 587| `/cpu/occa` | Serial OCCA kernels | 588| `/gpu/occa` | CUDA OCCA kernels | 589| `/omp/occa` | OpenMP OCCA kernels | 590| `/ocl/occa` | OpenCL OCCA kernels | 591 592Examples available in this release: 593 594:::{list-table} 595:header-rows: 1 596:widths: auto 597* - User code 598 - Example 599* - `ceed` 600 - * ex1 (volume) 601* - `mfem` 602 - * BP1 (scalar mass operator) 603 * BP3 (scalar Laplace operator) 604* - `petsc` 605 - * BP1 (scalar mass operator) 606* - `nek5000` 607 - * BP1 (scalar mass operator) 608::: 609 610(v0-1)= 611 612## v0.1 (Jan 3, 2018) 613 614Initial low-level API of the CEED project. The low-level API provides a set of Finite 615Elements kernels and components for writing new low-level kernels. Examples include: 616vector and sparse linear algebra, element matrix assembly over a batch of elements, 617partial assembly and action for efficient high-order operators like mass, diffusion, 618advection, etc. The main goal of the low-level API is to establish the basis for the 619high-level API. Also, identifying such low-level kernels and providing a reference 620implementation for them serves as the basis for specialized backend implementations. 621This release contained several backends: `/cpu/self`, and backends which rely upon the 622[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`, 623`/gpu/occa`, and `/omp/occa`. 624It also included several examples, in the `examples` folder: 625A standalone code that shows the usage of libCEED (with no external 626dependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1 627(with the application of the mass operator); and a `petsc` example to perform BP1 628(with the application of the mass operator). 629 630Backends available in this release: 631 632| CEED resource (`-ceed`) | Backend | 633|-------------------------|---------------------------------| 634| `/cpu/self` | Serial reference implementation | 635| `/cpu/occa` | Serial OCCA kernels | 636| `/gpu/occa` | CUDA OCCA kernels | 637| `/omp/occa` | OpenMP OCCA kernels | 638 639Examples available in this release: 640 641| User code | Example | 642|-----------------------|-----------------------------------| 643| `ceed` | ex1 (scalar Laplace operator) | 644| `mfem` | BP1 (scalar mass operator) | 645| `petsc` | BP1 (scalar mass operator) | 646``` 647