1bcb2dfaeSJed Brown# Changes/Release Notes 2bcb2dfaeSJed Brown 3bcb2dfaeSJed BrownOn this page we provide a summary of the main API changes, new features and examples 4bcb2dfaeSJed Brownfor each release of libCEED. 5bcb2dfaeSJed Brown 6bcb2dfaeSJed Brown(main)= 7bcb2dfaeSJed Brown 8bcb2dfaeSJed Brown## Current `main` branch 9bcb2dfaeSJed Brown 107e7773b5SJeremy L Thompson### Interface changes 117e7773b5SJeremy L Thompson 127e7773b5SJeremy L Thompson- Update {c:func} `CeedQFunctionGetFields` and {c:func} `CeedOperatorGetFields` to include number of fields. 1343bbe138SJeremy L Thompson- 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`, promoted to the public API. 14*f04ea552SJeremy L Thompson- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added. 157e7773b5SJeremy L Thompson 16bcb2dfaeSJed Brown### Maintainability 17bcb2dfaeSJed Brown 18bcb2dfaeSJed Brown- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage. 19bcb2dfaeSJed Brown 20bcb2dfaeSJed Brown(v0-9)= 21bcb2dfaeSJed Brown 22bcb2dfaeSJed Brown## v0.9 (Jul 6, 2021) 23bcb2dfaeSJed Brown 24bcb2dfaeSJed Brown### Interface changes 25bcb2dfaeSJed Brown 26bcb2dfaeSJed Brown- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact. 27bcb2dfaeSJed Brown 28bcb2dfaeSJed Brown### New features 29bcb2dfaeSJed Brown 30bcb2dfaeSJed Brown- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use. 31bcb2dfaeSJed Brown- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends. 32bcb2dfaeSJed Brown- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations. 33bcb2dfaeSJed Brown- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed). 34bcb2dfaeSJed Brown 35bcb2dfaeSJed Brown### Performance improvements 36bcb2dfaeSJed Brown 37bcb2dfaeSJed Brown### Examples 38bcb2dfaeSJed Brown 39bcb2dfaeSJed Brown- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations. 40bcb2dfaeSJed Brown- Fluid mechanics example adds GPU support and improves modularity. 41bcb2dfaeSJed Brown 42bcb2dfaeSJed Brown### Deprecated backends 43bcb2dfaeSJed Brown 44bcb2dfaeSJed Brown- 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. 45bcb2dfaeSJed Brown 46bcb2dfaeSJed Brown(v0-8)= 47bcb2dfaeSJed Brown 48bcb2dfaeSJed Brown## v0.8 (Mar 31, 2021) 49bcb2dfaeSJed Brown 50bcb2dfaeSJed Brown### Interface changes 51bcb2dfaeSJed Brown 52bcb2dfaeSJed Brown- Error handling improved to include enumerated error codes for C interface return values. 53bcb2dfaeSJed Brown- 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. 54bcb2dfaeSJed Brown 55bcb2dfaeSJed Brown### New features 56bcb2dfaeSJed Brown 57bcb2dfaeSJed Brown- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features. 58bcb2dfaeSJed Brown- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly. 59bcb2dfaeSJed Brown- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators. 60bcb2dfaeSJed Brown 61bcb2dfaeSJed Brown### Performance improvements 62bcb2dfaeSJed Brown 63bcb2dfaeSJed Brown- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`. 64bcb2dfaeSJed Brown- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`. 65bcb2dfaeSJed Brown 66bcb2dfaeSJed Brown### Examples 67bcb2dfaeSJed Brown 68bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions. 69bcb2dfaeSJed Brown- {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. 70bcb2dfaeSJed Brown- {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. 71bcb2dfaeSJed Brown- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree. 72bcb2dfaeSJed Brown 73bcb2dfaeSJed Brown(v0-7)= 74bcb2dfaeSJed Brown 75bcb2dfaeSJed Brown## v0.7 (Sep 29, 2020) 76bcb2dfaeSJed Brown 77bcb2dfaeSJed Brown### Interface changes 78bcb2dfaeSJed Brown 79bcb2dfaeSJed Brown- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors. 80bcb2dfaeSJed Brown As a result, the {code}`indices` parameter has been replaced with {code}`offsets` and the {code}`nnodes` parameter has been replaced with {code}`lsize`. 81bcb2dfaeSJed Brown These changes improve support for mixed finite element methods. 82bcb2dfaeSJed Brown- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature. 83bcb2dfaeSJed Brown- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity. 84bcb2dfaeSJed Brown- 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. 85bcb2dfaeSJed Brown- 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. 86bcb2dfaeSJed Brown- 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. 87bcb2dfaeSJed Brown- 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. 88bcb2dfaeSJed Brown 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`. 89bcb2dfaeSJed Brown- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory. 90bcb2dfaeSJed Brown- 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. 91bcb2dfaeSJed Brown 92bcb2dfaeSJed Brown### New features 93bcb2dfaeSJed Brown 94bcb2dfaeSJed Brown- New HIP backend: `/gpu/hip/ref`. 95bcb2dfaeSJed Brown- CeedQFunction support for user `CUfunction`s in some backends 96bcb2dfaeSJed Brown 97bcb2dfaeSJed Brown### Performance improvements 98bcb2dfaeSJed Brown 99bcb2dfaeSJed Brown- OCCA backend rebuilt to facilitate future performance enhancements. 100bcb2dfaeSJed Brown- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`. 101bcb2dfaeSJed Brown 102bcb2dfaeSJed Brown### Examples 103bcb2dfaeSJed Brown 104bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions. 105bcb2dfaeSJed Brown 106bcb2dfaeSJed Brown### Deprecated backends 107bcb2dfaeSJed Brown 108bcb2dfaeSJed Brown- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`. 109bcb2dfaeSJed Brown 110bcb2dfaeSJed Brown(v0-6)= 111bcb2dfaeSJed Brown 112bcb2dfaeSJed Brown## v0.6 (Mar 29, 2020) 113bcb2dfaeSJed Brown 114bcb2dfaeSJed BrownlibCEED v0.6 contains numerous new features and examples, as well as expanded 115bcb2dfaeSJed Browndocumentation in [this new website](https://libceed.readthedocs.io). 116bcb2dfaeSJed Brown 117bcb2dfaeSJed Brown### New features 118bcb2dfaeSJed Brown 119bcb2dfaeSJed Brown- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly 120bcb2dfaeSJed Brown 1-1 correspondence with the C interface, plus some convenience features. For instance, 121bcb2dfaeSJed Brown data stored in the {cpp:type}`CeedVector` structure are available without copy as 122bcb2dfaeSJed Brown {py:class}`numpy.ndarray`. Short tutorials are provided in 123bcb2dfaeSJed Brown [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/). 124bcb2dfaeSJed Brown- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point, 125bcb2dfaeSJed Brown {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal 126bcb2dfaeSJed Brown ({c:func}`CeedOperatorAssembleLinearDiagonal`). These operations are useful for 127bcb2dfaeSJed Brown preconditioning ingredients and are used in the libCEED's multigrid examples. 128bcb2dfaeSJed Brown- The inverse of separable operators can be obtained using 129bcb2dfaeSJed Brown {c:func}`CeedOperatorCreateFDMElementInverse` and applied with 130bcb2dfaeSJed Brown {c:func}`CeedOperatorApply`. This is a useful preconditioning ingredient, 131bcb2dfaeSJed Brown especially for Laplacians and related operators. 132bcb2dfaeSJed Brown- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`, 133bcb2dfaeSJed Brown {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`. 134bcb2dfaeSJed Brown- Make public accessors for various attributes to facilitate writing composable code. 135bcb2dfaeSJed Brown- New backend: `/cpu/self/memcheck/serial`. 136bcb2dfaeSJed Brown- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA 137bcb2dfaeSJed Brown backends. (Single source is coming soon for OCCA backends.) 138bcb2dfaeSJed Brown- Fix some missing edge cases in CUDA backend. 139bcb2dfaeSJed Brown 140bcb2dfaeSJed Brown### Performance Improvements 141bcb2dfaeSJed Brown 142bcb2dfaeSJed Brown- MAGMA backend performance optimization and non-tensor bases. 143bcb2dfaeSJed Brown- No-copy optimization in {c:func}`CeedOperatorApply`. 144bcb2dfaeSJed Brown 145bcb2dfaeSJed Brown### Interface changes 146bcb2dfaeSJed Brown 147bcb2dfaeSJed Brown- Replace {code}`CeedElemRestrictionCreateIdentity` and 148bcb2dfaeSJed Brown {code}`CeedElemRestrictionCreateBlocked` with more flexible 149bcb2dfaeSJed Brown {c:func}`CeedElemRestrictionCreateStrided` and 150bcb2dfaeSJed Brown {c:func}`CeedElemRestrictionCreateBlockedStrided`. 151bcb2dfaeSJed Brown- Add arguments to {c:func}`CeedQFunctionCreateIdentity`. 152bcb2dfaeSJed Brown- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification 153bcb2dfaeSJed Brown with {cpp:enum}`CeedInterlaceMode`. This is now an attribute of the 154bcb2dfaeSJed Brown {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no 155bcb2dfaeSJed Brown longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and 156bcb2dfaeSJed Brown {c:func}`CeedElemRestrictionApply`. 157bcb2dfaeSJed Brown 158bcb2dfaeSJed Brown### Examples 159bcb2dfaeSJed Brown 160bcb2dfaeSJed BrownlibCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`. 161bcb2dfaeSJed BrownNotable additions include: 162bcb2dfaeSJed Brown 163bcb2dfaeSJed Brown- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of 164bcb2dfaeSJed Brown a domain in 1, 2, and 3 dimensions by applying a Laplacian. 165bcb2dfaeSJed Brown 166bcb2dfaeSJed Brown- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area 167bcb2dfaeSJed Brown of domains (like the cube and sphere) by direct integration on a surface mesh; 168bcb2dfaeSJed Brown demonstrates geometric dimension different from topological dimension. 169bcb2dfaeSJed Brown 170bcb2dfaeSJed Brown- PETSc {ref}`example-petsc-bps`: 171bcb2dfaeSJed Brown 172bcb2dfaeSJed Brown - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support. 173bcb2dfaeSJed Brown - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements 174bcb2dfaeSJed Brown and transparent CUDA support. 175bcb2dfaeSJed Brown - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`): 176bcb2dfaeSJed Brown generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric 177bcb2dfaeSJed Brown dimension different from topological dimension. 178bcb2dfaeSJed Brown 179bcb2dfaeSJed Brown- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid 180bcb2dfaeSJed Brown solver with algebraic multigrid coarse solve. 181bcb2dfaeSJed Brown 182bcb2dfaeSJed Brown- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly 183bcb2dfaeSJed Brown `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`), 184bcb2dfaeSJed Brown implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and 185bcb2dfaeSJed Brown quasi-2D computational domain support. 186bcb2dfaeSJed Brown 187bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for 188bcb2dfaeSJed Brown linear elasticity, small-strain hyperelasticity, and globalized finite-strain 189bcb2dfaeSJed Brown hyperelasticity using p-multigrid with algebraic multigrid coarse solve. 190bcb2dfaeSJed Brown 191bcb2dfaeSJed Brown(v0-5)= 192bcb2dfaeSJed Brown 193bcb2dfaeSJed Brown## v0.5 (Sep 18, 2019) 194bcb2dfaeSJed Brown 195bcb2dfaeSJed BrownFor this release, several improvements were made. Two new CUDA backends were added to 196bcb2dfaeSJed Brownthe family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art 197bcb2dfaeSJed Brownperformance using single-source {ref}`CeedQFunction`. From this release, users 198bcb2dfaeSJed Browncan define Q-Functions in a single source code independently of the targeted backend 199bcb2dfaeSJed Brownwith the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU 200bcb2dfaeSJed Browncompilation of the user provided {ref}`CeedQFunction` code. To allow a unified 201bcb2dfaeSJed Browndeclaration, the {ref}`CeedQFunction` API has undergone a slight change: 202bcb2dfaeSJed Brownthe `QFunctionField` parameter `ncomp` has been changed to `size`. This change 203bcb2dfaeSJed Brownrequires setting the previous value of `ncomp` to `ncomp*dim` when adding a 204bcb2dfaeSJed Brown`QFunctionField` with eval mode `CEED EVAL GRAD`. 205bcb2dfaeSJed Brown 206bcb2dfaeSJed BrownAdditionally, new CPU backends 207bcb2dfaeSJed Brownwere included in this release, such as the `/cpu/self/opt/*` backends (which are 208bcb2dfaeSJed Brownwritten in pure C and use partial **E-vectors** to improve performance) and the 209bcb2dfaeSJed Brown`/cpu/self/ref/memcheck` backend (which relies upon the 210bcb2dfaeSJed Brown[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user 211bcb2dfaeSJed Brown{ref}`CeedQFunction` have no undefined values). 212bcb2dfaeSJed BrownThis release also included various performance improvements, bug fixes, new examples, 213bcb2dfaeSJed Brownand improved tests. Among these improvements, vectorized instructions for 214bcb2dfaeSJed Brown{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD` 215bcb2dfaeSJed Browninstead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and 216bcb2dfaeSJed Brownidentity Q-Functions were introduced, and the PETSc benchmark problems were expanded 217bcb2dfaeSJed Brownto include unstructured meshes handling were. For this expansion, the prior version of 218bcb2dfaeSJed Brownthe PETSc BPs, which only included data associated with structured geometries, were 219bcb2dfaeSJed Brownrenamed `bpsraw`, and the new version of the BPs, which can handle data associated 220bcb2dfaeSJed Brownwith any unstructured geometry, were called `bps`. Additionally, other benchmark 221bcb2dfaeSJed Brownproblems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively), 222bcb2dfaeSJed Brownand BP5 and BP6 (the collocated versions---for which the quadrature points are the same 223bcb2dfaeSJed Brownas the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc 224bcb2dfaeSJed Brownexamples. Furthermoew, another standalone libCEED example, called `ex2`, which 225bcb2dfaeSJed Browncomputes the surface area of a given mesh was added to this release. 226bcb2dfaeSJed Brown 227bcb2dfaeSJed BrownBackends available in this release: 228bcb2dfaeSJed Brown 22968e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 23068e843eeSJed Brown|--------------------------|-----------------------------------------------------| 23168e843eeSJed Brown| `/cpu/self/ref/serial` | Serial reference implementation | 23268e843eeSJed Brown| `/cpu/self/ref/blocked` | Blocked reference implementation | 23368e843eeSJed Brown| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks | 23468e843eeSJed Brown| `/cpu/self/opt/serial` | Serial optimized C implementation | 23568e843eeSJed Brown| `/cpu/self/opt/blocked` | Blocked optimized C implementation | 23668e843eeSJed Brown| `/cpu/self/avx/serial` | Serial AVX implementation | 23768e843eeSJed Brown| `/cpu/self/avx/blocked` | Blocked AVX implementation | 23868e843eeSJed Brown| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 23968e843eeSJed Brown| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 24068e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 24168e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 24268e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 24368e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 24468e843eeSJed Brown| `/gpu/cuda/ref` | Reference pure CUDA kernels | 24568e843eeSJed Brown| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 24668e843eeSJed Brown| `/gpu/cuda/shared` | Optimized pure CUDA kernels using shared memory | 24768e843eeSJed Brown| `/gpu/cuda/gen` | Optimized pure CUDA kernels using code generation | 24868e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 249bcb2dfaeSJed Brown 250bcb2dfaeSJed BrownExamples available in this release: 251bcb2dfaeSJed Brown 25268e843eeSJed Brown:::{list-table} 25368e843eeSJed Brown:header-rows: 1 25468e843eeSJed Brown:widths: auto 25568e843eeSJed Brown* - User code 25668e843eeSJed Brown - Example 25768e843eeSJed Brown* - `ceed` 25868e843eeSJed Brown - * ex1 (volume) 25968e843eeSJed Brown * ex2 (surface) 26068e843eeSJed Brown* - `mfem` 26168e843eeSJed Brown - * BP1 (scalar mass operator) 26268e843eeSJed Brown * BP3 (scalar Laplace operator) 26368e843eeSJed Brown* - `petsc` 26468e843eeSJed Brown - * BP1 (scalar mass operator) 26568e843eeSJed Brown * BP2 (vector mass operator) 26668e843eeSJed Brown * BP3 (scalar Laplace operator) 26768e843eeSJed Brown * BP4 (vector Laplace operator) 26868e843eeSJed Brown * BP5 (collocated scalar Laplace operator) 26968e843eeSJed Brown * BP6 (collocated vector Laplace operator) 27068e843eeSJed Brown * Navier-Stokes 27168e843eeSJed Brown* - `nek5000` 27268e843eeSJed Brown - * BP1 (scalar mass operator) 27368e843eeSJed Brown * BP3 (scalar Laplace operator) 27468e843eeSJed Brown::: 275bcb2dfaeSJed Brown 276bcb2dfaeSJed Brown(v0-4)= 277bcb2dfaeSJed Brown 278bcb2dfaeSJed Brown## v0.4 (Apr 1, 2019) 279bcb2dfaeSJed Brown 280bcb2dfaeSJed BrownlibCEED v0.4 was made again publicly available in the second full CEED software 281bcb2dfaeSJed Browndistribution, release CEED 2.0. This release contained notable features, such as 282bcb2dfaeSJed Brownfour new CPU backends, two new GPU backends, CPU backend optimizations, initial 283bcb2dfaeSJed Brownsupport for operator composition, performance benchmarking, and a Navier-Stokes demo. 284bcb2dfaeSJed BrownThe new CPU backends in this release came in two families. The `/cpu/self/*/serial` 285bcb2dfaeSJed Brownbackends process one element at a time and are intended for meshes with a smaller number 286bcb2dfaeSJed Brownof high order elements. The `/cpu/self/*/blocked` backends process blocked batches of 287bcb2dfaeSJed Browneight interlaced elements and are intended for meshes with higher numbers of elements. 288bcb2dfaeSJed BrownThe `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU 289bcb2dfaeSJed Brownperformance. The `/cpu/self/xsmm/*` backends rely upon the 290bcb2dfaeSJed Brown[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU 291bcb2dfaeSJed Brownperformance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA. 292bcb2dfaeSJed BrownThe `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable 293bcb2dfaeSJed Brownperformance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple 294bcb2dfaeSJed Brownparallelization approach, where each thread treats a finite element. Using just in time 295bcb2dfaeSJed Browncompilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this 296bcb2dfaeSJed Brownbackend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend 297bcb2dfaeSJed Brownachieve good peak performance for 1D, 2D, and low order 3D problems, but performance 298bcb2dfaeSJed Browndeteriorates very quickly when threads run out of registers. 299bcb2dfaeSJed Brown 300bcb2dfaeSJed BrownA new explicit time-stepping Navier-Stokes solver was added to the family of libCEED 301bcb2dfaeSJed Brownexamples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`). 302bcb2dfaeSJed BrownThis example solves the time-dependent Navier-Stokes equations of compressible gas 303bcb2dfaeSJed Browndynamics in a static Eulerian three-dimensional frame, using structured high-order 304bcb2dfaeSJed Brownfinite/spectral element spatial discretizations and explicit high-order time-stepping 305bcb2dfaeSJed Brown(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc, 306bcb2dfaeSJed Brownso that the pointwise physics (defined at quadrature points) is separated from the 307bcb2dfaeSJed Brownparallelization and meshing concerns. 308bcb2dfaeSJed Brown 309bcb2dfaeSJed BrownBackends available in this release: 310bcb2dfaeSJed Brown 31168e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 31268e843eeSJed Brown|--------------------------|-----------------------------------------------------| 31368e843eeSJed Brown| `/cpu/self/ref/serial` | Serial reference implementation | 31468e843eeSJed Brown| `/cpu/self/ref/blocked` | Blocked reference implementation | 31568e843eeSJed Brown| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 31668e843eeSJed Brown| `/cpu/self/avx/serial` | Serial AVX implementation | 31768e843eeSJed Brown| `/cpu/self/avx/blocked` | Blocked AVX implementation | 31868e843eeSJed Brown| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 31968e843eeSJed Brown| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 32068e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 32168e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 32268e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 32368e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 32468e843eeSJed Brown| `/gpu/cuda/ref` | Reference pure CUDA kernels | 32568e843eeSJed Brown| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 32668e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 327bcb2dfaeSJed Brown 328bcb2dfaeSJed BrownExamples available in this release: 329bcb2dfaeSJed Brown 33068e843eeSJed Brown:::{list-table} 33168e843eeSJed Brown:header-rows: 1 33268e843eeSJed Brown:widths: auto 33368e843eeSJed Brown* - User code 33468e843eeSJed Brown - Example 33568e843eeSJed Brown* - `ceed` 33668e843eeSJed Brown - * ex1 (volume) 33768e843eeSJed Brown* - `mfem` 33868e843eeSJed Brown - * BP1 (scalar mass operator) 33968e843eeSJed Brown * BP3 (scalar Laplace operator) 34068e843eeSJed Brown* - `petsc` 34168e843eeSJed Brown - * BP1 (scalar mass operator) 34268e843eeSJed Brown * BP3 (scalar Laplace operator) 34368e843eeSJed Brown * Navier-Stokes 34468e843eeSJed Brown* - `nek5000` 34568e843eeSJed Brown - * BP1 (scalar mass operator) 34668e843eeSJed Brown * BP3 (scalar Laplace operator) 34768e843eeSJed Brown::: 348bcb2dfaeSJed Brown 349bcb2dfaeSJed Brown(v0-3)= 350bcb2dfaeSJed Brown 351bcb2dfaeSJed Brown## v0.3 (Sep 30, 2018) 352bcb2dfaeSJed Brown 353bcb2dfaeSJed BrownNotable features in this release include active/passive field interface, support for 354bcb2dfaeSJed Brownnon-tensor bases, backend optimization, and improved Fortran interface. This release 355bcb2dfaeSJed Brownalso focused on providing improved continuous integration, and many new tests with code 356bcb2dfaeSJed Browncoverage reports of about 90%. This release also provided a significant change to the 357bcb2dfaeSJed Brownpublic interface: a {ref}`CeedQFunction` can take any number of named input and output 358bcb2dfaeSJed Brownarguments while {ref}`CeedOperator` connects them to the actual data, which may be 359bcb2dfaeSJed Brownsupplied explicitly to `CeedOperatorApply()` (active) or separately via 360bcb2dfaeSJed Brown`CeedOperatorSetField()` (passive). This interface change enables reusable libraries 361bcb2dfaeSJed Brownof CeedQFunctions and composition of block solvers constructed using 362bcb2dfaeSJed Brown{ref}`CeedOperator`. A concept of blocked restriction was added to this release and 363bcb2dfaeSJed Brownused in an optimized CPU backend. Although this is typically not visible to the user, 364bcb2dfaeSJed Brownit enables effective use of arbitrary-length SIMD while maintaining cache locality. 365bcb2dfaeSJed BrownThis CPU backend also implements an algebraic factorization of tensor product gradients 366bcb2dfaeSJed Brownto perform fewer operations than standard application of interpolation and 367bcb2dfaeSJed Browndifferentiation from nodes to quadrature points. This algebraic formulation 368bcb2dfaeSJed Brownautomatically supports non-polynomial and non-interpolatory bases, thus is more general 369bcb2dfaeSJed Brownthan the more common derivation in terms of Lagrange polynomials on the quadrature points. 370bcb2dfaeSJed Brown 371bcb2dfaeSJed BrownBackends available in this release: 372bcb2dfaeSJed Brown 37368e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 37468e843eeSJed Brown|-------------------------|-----------------------------------------------------| 37568e843eeSJed Brown| `/cpu/self/blocked` | Blocked reference implementation | 37668e843eeSJed Brown| `/cpu/self/ref` | Serial reference implementation | 37768e843eeSJed Brown| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 37868e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 37968e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 38068e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 38168e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 38268e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 383bcb2dfaeSJed Brown 384bcb2dfaeSJed BrownExamples available in this release: 385bcb2dfaeSJed Brown 38668e843eeSJed Brown:::{list-table} 38768e843eeSJed Brown:header-rows: 1 38868e843eeSJed Brown:widths: auto 38968e843eeSJed Brown* - User code 39068e843eeSJed Brown - Example 39168e843eeSJed Brown* - `ceed` 39268e843eeSJed Brown - * ex1 (volume) 39368e843eeSJed Brown* - `mfem` 39468e843eeSJed Brown - * BP1 (scalar mass operator) 39568e843eeSJed Brown * BP3 (scalar Laplace operator) 39668e843eeSJed Brown* - `petsc` 39768e843eeSJed Brown - * BP1 (scalar mass operator) 39868e843eeSJed Brown * BP3 (scalar Laplace operator) 39968e843eeSJed Brown* - `nek5000` 40068e843eeSJed Brown - * BP1 (scalar mass operator) 40168e843eeSJed Brown * BP3 (scalar Laplace operator) 40268e843eeSJed Brown::: 403bcb2dfaeSJed Brown 404bcb2dfaeSJed Brown(v0-21)= 405bcb2dfaeSJed Brown 406bcb2dfaeSJed Brown## v0.21 (Sep 30, 2018) 407bcb2dfaeSJed Brown 408bcb2dfaeSJed BrownA MAGMA backend (which relies upon the 409bcb2dfaeSJed Brown[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this 410bcb2dfaeSJed Brownrelease. This initial integration set up the framework of using MAGMA and provided the 411bcb2dfaeSJed BrownlibCEED functionality through MAGMA kernels as one of libCEED’s computational backends. 412bcb2dfaeSJed BrownAs any other backend, the MAGMA backend provides extended basic data structures for 413bcb2dfaeSJed Brown{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements 414bcb2dfaeSJed Brownthe fundamental CEED building blocks to work with the new data structures. 415bcb2dfaeSJed BrownIn general, the MAGMA-specific data structures keep the libCEED pointers to CPU data 416bcb2dfaeSJed Brownbut also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled 417bcb2dfaeSJed Browninternally, and thus seamlessly to the user, through the functions/methods that are 418bcb2dfaeSJed Brownprovided to support them. 419bcb2dfaeSJed Brown 420bcb2dfaeSJed BrownBackends available in this release: 421bcb2dfaeSJed Brown 42268e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 42368e843eeSJed Brown|-------------------------|---------------------------------| 42468e843eeSJed Brown| `/cpu/self` | Serial reference implementation | 42568e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 42668e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 42768e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 42868e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 42968e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 430bcb2dfaeSJed Brown 431bcb2dfaeSJed BrownExamples available in this release: 432bcb2dfaeSJed Brown 43368e843eeSJed Brown:::{list-table} 43468e843eeSJed Brown:header-rows: 1 43568e843eeSJed Brown:widths: auto 43668e843eeSJed Brown* - User code 43768e843eeSJed Brown - Example 43868e843eeSJed Brown* - `ceed` 43968e843eeSJed Brown - * ex1 (volume) 44068e843eeSJed Brown* - `mfem` 44168e843eeSJed Brown - * BP1 (scalar mass operator) 44268e843eeSJed Brown * BP3 (scalar Laplace operator) 44368e843eeSJed Brown* - `petsc` 44468e843eeSJed Brown - * BP1 (scalar mass operator) 44568e843eeSJed Brown* - `nek5000` 44668e843eeSJed Brown - * BP1 (scalar mass operator) 44768e843eeSJed Brown::: 448bcb2dfaeSJed Brown 449bcb2dfaeSJed Brown(v0-2)= 450bcb2dfaeSJed Brown 451bcb2dfaeSJed Brown## v0.2 (Mar 30, 2018) 452bcb2dfaeSJed Brown 453bcb2dfaeSJed BrownlibCEED was made publicly available the first full CEED software distribution, release 454bcb2dfaeSJed BrownCEED 1.0. The distribution was made available using the Spack package manager to provide 455bcb2dfaeSJed Browna common, easy-to-use build environment, where the user can build the CEED distribution 456bcb2dfaeSJed Brownwith all dependencies. This release included a new Fortran interface for the library. 457bcb2dfaeSJed BrownThis release also contained major improvements in the OCCA backend (including a new 458bcb2dfaeSJed Brown`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to 459bcb2dfaeSJed Browncompute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an 460bcb2dfaeSJed Brown`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application 461bcb2dfaeSJed Brownof the Laplace operator) was also added to this release. 462bcb2dfaeSJed Brown 463bcb2dfaeSJed BrownBackends available in this release: 464bcb2dfaeSJed Brown 46568e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 46668e843eeSJed Brown|-------------------------|---------------------------------| 46768e843eeSJed Brown| `/cpu/self` | Serial reference implementation | 46868e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 46968e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 47068e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 47168e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 472bcb2dfaeSJed Brown 473bcb2dfaeSJed BrownExamples available in this release: 474bcb2dfaeSJed Brown 47568e843eeSJed Brown:::{list-table} 47668e843eeSJed Brown:header-rows: 1 47768e843eeSJed Brown:widths: auto 47868e843eeSJed Brown* - User code 47968e843eeSJed Brown - Example 48068e843eeSJed Brown* - `ceed` 48168e843eeSJed Brown - * ex1 (volume) 48268e843eeSJed Brown* - `mfem` 48368e843eeSJed Brown - * BP1 (scalar mass operator) 48468e843eeSJed Brown * BP3 (scalar Laplace operator) 48568e843eeSJed Brown* - `petsc` 48668e843eeSJed Brown - * BP1 (scalar mass operator) 48768e843eeSJed Brown* - `nek5000` 48868e843eeSJed Brown - * BP1 (scalar mass operator) 48968e843eeSJed Brown::: 490bcb2dfaeSJed Brown 491bcb2dfaeSJed Brown(v0-1)= 492bcb2dfaeSJed Brown 493bcb2dfaeSJed Brown## v0.1 (Jan 3, 2018) 494bcb2dfaeSJed Brown 495bcb2dfaeSJed BrownInitial low-level API of the CEED project. The low-level API provides a set of Finite 496bcb2dfaeSJed BrownElements kernels and components for writing new low-level kernels. Examples include: 497bcb2dfaeSJed Brownvector and sparse linear algebra, element matrix assembly over a batch of elements, 498bcb2dfaeSJed Brownpartial assembly and action for efficient high-order operators like mass, diffusion, 499bcb2dfaeSJed Brownadvection, etc. The main goal of the low-level API is to establish the basis for the 500bcb2dfaeSJed Brownhigh-level API. Also, identifying such low-level kernels and providing a reference 501bcb2dfaeSJed Brownimplementation for them serves as the basis for specialized backend implementations. 502bcb2dfaeSJed BrownThis release contained several backends: `/cpu/self`, and backends which rely upon the 503bcb2dfaeSJed Brown[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`, 504bcb2dfaeSJed Brown`/gpu/occa`, and `/omp/occa`. 505bcb2dfaeSJed BrownIt also included several examples, in the `examples` folder: 506bcb2dfaeSJed BrownA standalone code that shows the usage of libCEED (with no external 507bcb2dfaeSJed Browndependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1 508bcb2dfaeSJed Brown(with the application of the mass operator); and a `petsc` example to perform BP1 509bcb2dfaeSJed Brown(with the application of the mass operator). 510bcb2dfaeSJed Brown 511bcb2dfaeSJed BrownBackends available in this release: 512bcb2dfaeSJed Brown 51368e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 51468e843eeSJed Brown|-------------------------|---------------------------------| 51568e843eeSJed Brown| `/cpu/self` | Serial reference implementation | 51668e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 51768e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 51868e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 519bcb2dfaeSJed Brown 520bcb2dfaeSJed BrownExamples available in this release: 521bcb2dfaeSJed Brown 522bcb2dfaeSJed Brown| User code | Example | 52368e843eeSJed Brown|-----------------------|-----------------------------------| 52468e843eeSJed Brown| `ceed` | ex1 (scalar Laplace operator) | 52568e843eeSJed Brown| `mfem` | BP1 (scalar mass operator) | 52668e843eeSJed Brown| `petsc` | BP1 (scalar mass operator) | 527bcb2dfaeSJed Brown``` 528