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