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. 13ce4822f6SJeremy L Thompson- 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`. 14f04ea552SJeremy L Thompson- Clarify and document conditions where `CeedQFunction` and `CeedOperator` become immutable and no further fields or suboperators can be added. 1570a7ffb3SJeremy L Thompson- Add {c:func}`CeedOperatorLinearAssembleQFunctionBuildOrUpdate` to reduce object creation overhead in assembly of CeedOperator preconditioning ingredients. 164db537f9SJeremy L Thompson- Promote {c:func}`CeedOperatorCheckReady`to the public API to facilitate interactive interfaces. 17dcc1e3ecSJeremy L Thompson- Warning added when compiling OCCA backend to alert users that this backend is experimental. 189a1d3511SJeremy L Thompson- `ceed-backend.h`, `ceed-hash.h`, and `ceed-khash.h` removed. Users should use `ceed/backend.h`, `ceed/hash.h`, and `ceed/khash.h`. 1943e1b16fSJeremy L Thompson- Added {c:func}`CeedQFunctionGetKernelName`; refactored {c:func}`CeedQFunctionGetSourcePath` to exclude function kernel name. 207e7773b5SJeremy L Thompson 21f479eb23SJeremy L Thompson### New features 22f479eb23SJeremy L Thompson 23f479eb23SJeremy L Thompson- `CeedScalar` can now be set as `float` or `double` at compile time. 2430601ac0SJeremy L Thompson- Added JiT utilities in `ceed/jit-tools.h` to reduce duplicated code in GPU backends. 25fb3c7d02SJeremy L Thompson- 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. 2623dfbf5bSJeremy L Thompson- Remove need to guard library headers in QFunction source for code generation backends. 27*3f21f6b1SJeremy L Thompson- `CeedDebugEnv()` macro created to provide debugging outputs when Ceed context is not present. 28f479eb23SJeremy L Thompson 29bcb2dfaeSJed Brown### Maintainability 30bcb2dfaeSJed Brown 31bcb2dfaeSJed Brown- Refactored preconditioner support internally to facilitate future development and improve GPU completeness/test coverage. 32db52d626SJeremy L Thompson- `Include-what-you-use` makefile target added as `make iwyu`. 33bcb2dfaeSJed Brown 34bcb2dfaeSJed Brown(v0-9)= 35bcb2dfaeSJed Brown 36bcb2dfaeSJed Brown## v0.9 (Jul 6, 2021) 37bcb2dfaeSJed Brown 38bcb2dfaeSJed Brown### Interface changes 39bcb2dfaeSJed Brown 40bcb2dfaeSJed Brown- Minor modification in error handling macro to silence pedantic warnings when compiling with Clang, but no functional impact. 41bcb2dfaeSJed Brown 42bcb2dfaeSJed Brown### New features 43bcb2dfaeSJed Brown 44bcb2dfaeSJed Brown- Add {c:func}`CeedVectorAXPY` and {c:func}`CeedVectorPointwiseMult` as a convenience for stand-alone testing and internal use. 45bcb2dfaeSJed Brown- Add `CEED_QFUNCTION_HELPER` macro to properly annotate QFunction helper functions for code generation backends. 46bcb2dfaeSJed Brown- Add `CeedPragmaOptimizeOff` macro for code that is sensitive to floating point errors from fast math optimizations. 47bcb2dfaeSJed Brown- Rust support: split `libceed-sys` crate out of `libceed` and [publish both on crates.io](https://crates.io/crates/libceed). 48bcb2dfaeSJed Brown 49bcb2dfaeSJed Brown### Performance improvements 50bcb2dfaeSJed Brown 51bcb2dfaeSJed Brown### Examples 52bcb2dfaeSJed Brown 53bcb2dfaeSJed Brown- Solid mechanics mini-app updated to explore the performance impacts of various formulations in the initial and current configurations. 54bcb2dfaeSJed Brown- Fluid mechanics example adds GPU support and improves modularity. 55bcb2dfaeSJed Brown 56bcb2dfaeSJed Brown### Deprecated backends 57bcb2dfaeSJed Brown 58bcb2dfaeSJed 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. 59bcb2dfaeSJed Brown 60bcb2dfaeSJed Brown(v0-8)= 61bcb2dfaeSJed Brown 62bcb2dfaeSJed Brown## v0.8 (Mar 31, 2021) 63bcb2dfaeSJed Brown 64bcb2dfaeSJed Brown### Interface changes 65bcb2dfaeSJed Brown 66bcb2dfaeSJed Brown- Error handling improved to include enumerated error codes for C interface return values. 67bcb2dfaeSJed 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. 68bcb2dfaeSJed Brown 69bcb2dfaeSJed Brown### New features 70bcb2dfaeSJed Brown 71bcb2dfaeSJed Brown- Julia and Rust interfaces added, providing a nearly 1-1 correspondence with the C interface, plus some convenience features. 72bcb2dfaeSJed Brown- Static libraries can be built with `make STATIC=1` and the pkg-config file is installed accordingly. 73bcb2dfaeSJed Brown- Add {c:func}`CeedOperatorLinearAssembleSymbolic` and {c:func}`CeedOperatorLinearAssemble` to support full assembly of libCEED operators. 74bcb2dfaeSJed Brown 75bcb2dfaeSJed Brown### Performance improvements 76bcb2dfaeSJed Brown 77bcb2dfaeSJed Brown- New HIP MAGMA backends for hipMAGMA library users: `/gpu/hip/magma` and `/gpu/hip/magma/det`. 78bcb2dfaeSJed Brown- New HIP backends for improved tensor basis performance: `/gpu/hip/shared` and `/gpu/hip/gen`. 79bcb2dfaeSJed Brown 80bcb2dfaeSJed Brown### Examples 81bcb2dfaeSJed Brown 82bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` example updated with traction boundary conditions and improved Dirichlet boundary conditions. 83bcb2dfaeSJed 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. 84bcb2dfaeSJed 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. 85bcb2dfaeSJed Brown- {ref}`example-petsc-navier-stokes` example updated with support for performing convergence study and plotting order of convergence by polynomial degree. 86bcb2dfaeSJed Brown 87bcb2dfaeSJed Brown(v0-7)= 88bcb2dfaeSJed Brown 89bcb2dfaeSJed Brown## v0.7 (Sep 29, 2020) 90bcb2dfaeSJed Brown 91bcb2dfaeSJed Brown### Interface changes 92bcb2dfaeSJed Brown 93bcb2dfaeSJed Brown- Replace limited {code}`CeedInterlaceMode` with more flexible component stride {code}`compstride` in {code}`CeedElemRestriction` constructors. 94bcb2dfaeSJed 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`. 95bcb2dfaeSJed Brown These changes improve support for mixed finite element methods. 96bcb2dfaeSJed Brown- Replace various uses of {code}`Ceed*Get*Status` with {code}`Ceed*Is*` in the backend API to match common nomenclature. 97bcb2dfaeSJed Brown- Replace {code}`CeedOperatorAssembleLinearDiagonal` with {c:func}`CeedOperatorLinearAssembleDiagonal` for clarity. 98bcb2dfaeSJed 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. 99bcb2dfaeSJed 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. 100bcb2dfaeSJed 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. 101bcb2dfaeSJed 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. 102bcb2dfaeSJed 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`. 103bcb2dfaeSJed Brown- Added {code}`CeedQFunctionContext` object to manage user QFunction context data and reduce copies between device and host memory. 104bcb2dfaeSJed 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. 105bcb2dfaeSJed Brown 106bcb2dfaeSJed Brown### New features 107bcb2dfaeSJed Brown 108bcb2dfaeSJed Brown- New HIP backend: `/gpu/hip/ref`. 109bcb2dfaeSJed Brown- CeedQFunction support for user `CUfunction`s in some backends 110bcb2dfaeSJed Brown 111bcb2dfaeSJed Brown### Performance improvements 112bcb2dfaeSJed Brown 113bcb2dfaeSJed Brown- OCCA backend rebuilt to facilitate future performance enhancements. 114bcb2dfaeSJed Brown- Petsc BPs suite improved to reduce noise due to multiple calls to {code}`mpiexec`. 115bcb2dfaeSJed Brown 116bcb2dfaeSJed Brown### Examples 117bcb2dfaeSJed Brown 118bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` example updated with strain energy computation and more flexible boundary conditions. 119bcb2dfaeSJed Brown 120bcb2dfaeSJed Brown### Deprecated backends 121bcb2dfaeSJed Brown 122bcb2dfaeSJed Brown- The `/gpu/cuda/reg` backend has been removed, with its core features moved into `/gpu/cuda/ref` and `/gpu/cuda/shared`. 123bcb2dfaeSJed Brown 124bcb2dfaeSJed Brown(v0-6)= 125bcb2dfaeSJed Brown 126bcb2dfaeSJed Brown## v0.6 (Mar 29, 2020) 127bcb2dfaeSJed Brown 128bcb2dfaeSJed BrownlibCEED v0.6 contains numerous new features and examples, as well as expanded 129bcb2dfaeSJed Browndocumentation in [this new website](https://libceed.readthedocs.io). 130bcb2dfaeSJed Brown 131bcb2dfaeSJed Brown### New features 132bcb2dfaeSJed Brown 133bcb2dfaeSJed Brown- New Python interface using [CFFI](https://cffi.readthedocs.io/) provides a nearly 134bcb2dfaeSJed Brown 1-1 correspondence with the C interface, plus some convenience features. For instance, 135bcb2dfaeSJed Brown data stored in the {cpp:type}`CeedVector` structure are available without copy as 136bcb2dfaeSJed Brown {py:class}`numpy.ndarray`. Short tutorials are provided in 137bcb2dfaeSJed Brown [Binder](https://mybinder.org/v2/gh/CEED/libCEED/main?urlpath=lab/tree/examples/tutorials/). 138bcb2dfaeSJed Brown- Linear QFunctions can be assembled as block-diagonal matrices (per quadrature point, 139bcb2dfaeSJed Brown {c:func}`CeedOperatorAssembleLinearQFunction`) or to evaluate the diagonal 140bcb2dfaeSJed Brown ({c:func}`CeedOperatorAssembleLinearDiagonal`). These operations are useful for 141bcb2dfaeSJed Brown preconditioning ingredients and are used in the libCEED's multigrid examples. 142bcb2dfaeSJed Brown- The inverse of separable operators can be obtained using 143bcb2dfaeSJed Brown {c:func}`CeedOperatorCreateFDMElementInverse` and applied with 144bcb2dfaeSJed Brown {c:func}`CeedOperatorApply`. This is a useful preconditioning ingredient, 145bcb2dfaeSJed Brown especially for Laplacians and related operators. 146bcb2dfaeSJed Brown- New functions: {c:func}`CeedVectorNorm`, {c:func}`CeedOperatorApplyAdd`, 147bcb2dfaeSJed Brown {c:func}`CeedQFunctionView`, {c:func}`CeedOperatorView`. 148bcb2dfaeSJed Brown- Make public accessors for various attributes to facilitate writing composable code. 149bcb2dfaeSJed Brown- New backend: `/cpu/self/memcheck/serial`. 150bcb2dfaeSJed Brown- QFunctions using variable-length array (VLA) pointer constructs can be used with CUDA 151bcb2dfaeSJed Brown backends. (Single source is coming soon for OCCA backends.) 152bcb2dfaeSJed Brown- Fix some missing edge cases in CUDA backend. 153bcb2dfaeSJed Brown 154bcb2dfaeSJed Brown### Performance Improvements 155bcb2dfaeSJed Brown 156bcb2dfaeSJed Brown- MAGMA backend performance optimization and non-tensor bases. 157bcb2dfaeSJed Brown- No-copy optimization in {c:func}`CeedOperatorApply`. 158bcb2dfaeSJed Brown 159bcb2dfaeSJed Brown### Interface changes 160bcb2dfaeSJed Brown 161bcb2dfaeSJed Brown- Replace {code}`CeedElemRestrictionCreateIdentity` and 162bcb2dfaeSJed Brown {code}`CeedElemRestrictionCreateBlocked` with more flexible 163bcb2dfaeSJed Brown {c:func}`CeedElemRestrictionCreateStrided` and 164bcb2dfaeSJed Brown {c:func}`CeedElemRestrictionCreateBlockedStrided`. 165bcb2dfaeSJed Brown- Add arguments to {c:func}`CeedQFunctionCreateIdentity`. 166bcb2dfaeSJed Brown- Replace ambiguous uses of {cpp:enum}`CeedTransposeMode` for L-vector identification 167bcb2dfaeSJed Brown with {cpp:enum}`CeedInterlaceMode`. This is now an attribute of the 168bcb2dfaeSJed Brown {cpp:type}`CeedElemRestriction` (see {c:func}`CeedElemRestrictionCreate`) and no 169bcb2dfaeSJed Brown longer passed as `lmode` arguments to {c:func}`CeedOperatorSetField` and 170bcb2dfaeSJed Brown {c:func}`CeedElemRestrictionApply`. 171bcb2dfaeSJed Brown 172bcb2dfaeSJed Brown### Examples 173bcb2dfaeSJed Brown 174bcb2dfaeSJed BrownlibCEED-0.6 contains greatly expanded examples with {ref}`new documentation <Examples>`. 175bcb2dfaeSJed BrownNotable additions include: 176bcb2dfaeSJed Brown 177bcb2dfaeSJed Brown- Standalone {ref}`ex2-surface` ({file}`examples/ceed/ex2-surface`): compute the area of 178bcb2dfaeSJed Brown a domain in 1, 2, and 3 dimensions by applying a Laplacian. 179bcb2dfaeSJed Brown 180bcb2dfaeSJed Brown- PETSc {ref}`example-petsc-area` ({file}`examples/petsc/area.c`): computes surface area 181bcb2dfaeSJed Brown of domains (like the cube and sphere) by direct integration on a surface mesh; 182bcb2dfaeSJed Brown demonstrates geometric dimension different from topological dimension. 183bcb2dfaeSJed Brown 184bcb2dfaeSJed Brown- PETSc {ref}`example-petsc-bps`: 185bcb2dfaeSJed Brown 186bcb2dfaeSJed Brown - {file}`examples/petsc/bpsraw.c` (formerly `bps.c`): transparent CUDA support. 187bcb2dfaeSJed Brown - {file}`examples/petsc/bps.c` (formerly `bpsdmplex.c`): performance improvements 188bcb2dfaeSJed Brown and transparent CUDA support. 189bcb2dfaeSJed Brown - {ref}`example-petsc-bps-sphere` ({file}`examples/petsc/bpssphere.c`): 190bcb2dfaeSJed Brown generalizations of all CEED BPs to the surface of the sphere; demonstrates geometric 191bcb2dfaeSJed Brown dimension different from topological dimension. 192bcb2dfaeSJed Brown 193bcb2dfaeSJed Brown- {ref}`example-petsc-multigrid` ({file}`examples/petsc/multigrid.c`): new p-multigrid 194bcb2dfaeSJed Brown solver with algebraic multigrid coarse solve. 195bcb2dfaeSJed Brown 196bcb2dfaeSJed Brown- {ref}`example-petsc-navier-stokes` ({file}`examples/fluids/navierstokes.c`; formerly 197bcb2dfaeSJed Brown `examples/navier-stokes`): unstructured grid support (using PETSc's `DMPlex`), 198bcb2dfaeSJed Brown implicit time integration, SU/SUPG stabilization, free-slip boundary conditions, and 199bcb2dfaeSJed Brown quasi-2D computational domain support. 200bcb2dfaeSJed Brown 201bcb2dfaeSJed Brown- {ref}`example-petsc-elasticity` ({file}`examples/solids/elasticity.c`): new solver for 202bcb2dfaeSJed Brown linear elasticity, small-strain hyperelasticity, and globalized finite-strain 203bcb2dfaeSJed Brown hyperelasticity using p-multigrid with algebraic multigrid coarse solve. 204bcb2dfaeSJed Brown 205bcb2dfaeSJed Brown(v0-5)= 206bcb2dfaeSJed Brown 207bcb2dfaeSJed Brown## v0.5 (Sep 18, 2019) 208bcb2dfaeSJed Brown 209bcb2dfaeSJed BrownFor this release, several improvements were made. Two new CUDA backends were added to 210bcb2dfaeSJed Brownthe family of backends, of which, the new `cuda-gen` backend achieves state-of-the-art 211bcb2dfaeSJed Brownperformance using single-source {ref}`CeedQFunction`. From this release, users 212bcb2dfaeSJed Browncan define Q-Functions in a single source code independently of the targeted backend 213bcb2dfaeSJed Brownwith the aid of a new macro `CEED QFUNCTION` to support JIT (Just-In-Time) and CPU 214bcb2dfaeSJed Browncompilation of the user provided {ref}`CeedQFunction` code. To allow a unified 215bcb2dfaeSJed Browndeclaration, the {ref}`CeedQFunction` API has undergone a slight change: 216bcb2dfaeSJed Brownthe `QFunctionField` parameter `ncomp` has been changed to `size`. This change 217bcb2dfaeSJed Brownrequires setting the previous value of `ncomp` to `ncomp*dim` when adding a 218bcb2dfaeSJed Brown`QFunctionField` with eval mode `CEED EVAL GRAD`. 219bcb2dfaeSJed Brown 220bcb2dfaeSJed BrownAdditionally, new CPU backends 221bcb2dfaeSJed Brownwere included in this release, such as the `/cpu/self/opt/*` backends (which are 222bcb2dfaeSJed Brownwritten in pure C and use partial **E-vectors** to improve performance) and the 223bcb2dfaeSJed Brown`/cpu/self/ref/memcheck` backend (which relies upon the 224bcb2dfaeSJed Brown[Valgrind](http://valgrind.org/) Memcheck tool to help verify that user 225bcb2dfaeSJed Brown{ref}`CeedQFunction` have no undefined values). 226bcb2dfaeSJed BrownThis release also included various performance improvements, bug fixes, new examples, 227bcb2dfaeSJed Brownand improved tests. Among these improvements, vectorized instructions for 228bcb2dfaeSJed Brown{ref}`CeedQFunction` code compiled for CPU were enhanced by using `CeedPragmaSIMD` 229bcb2dfaeSJed Browninstead of `CeedPragmaOMP`, implementation of a {ref}`CeedQFunction` gallery and 230bcb2dfaeSJed Brownidentity Q-Functions were introduced, and the PETSc benchmark problems were expanded 231bcb2dfaeSJed Brownto include unstructured meshes handling were. For this expansion, the prior version of 232bcb2dfaeSJed Brownthe PETSc BPs, which only included data associated with structured geometries, were 233bcb2dfaeSJed Brownrenamed `bpsraw`, and the new version of the BPs, which can handle data associated 234bcb2dfaeSJed Brownwith any unstructured geometry, were called `bps`. Additionally, other benchmark 235bcb2dfaeSJed Brownproblems, namely BP2 and BP4 (the vector-valued versions of BP1 and BP3, respectively), 236bcb2dfaeSJed Brownand BP5 and BP6 (the collocated versions---for which the quadrature points are the same 237bcb2dfaeSJed Brownas the Gauss Lobatto nodes---of BP3 and BP4 respectively) were added to the PETSc 238bcb2dfaeSJed Brownexamples. Furthermoew, another standalone libCEED example, called `ex2`, which 239bcb2dfaeSJed Browncomputes the surface area of a given mesh was added to this release. 240bcb2dfaeSJed Brown 241bcb2dfaeSJed BrownBackends available in this release: 242bcb2dfaeSJed Brown 24368e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 24468e843eeSJed Brown|--------------------------|-----------------------------------------------------| 24568e843eeSJed Brown| `/cpu/self/ref/serial` | Serial reference implementation | 24668e843eeSJed Brown| `/cpu/self/ref/blocked` | Blocked reference implementation | 24768e843eeSJed Brown| `/cpu/self/ref/memcheck` | Memcheck backend, undefined value checks | 24868e843eeSJed Brown| `/cpu/self/opt/serial` | Serial optimized C implementation | 24968e843eeSJed Brown| `/cpu/self/opt/blocked` | Blocked optimized C implementation | 25068e843eeSJed Brown| `/cpu/self/avx/serial` | Serial AVX implementation | 25168e843eeSJed Brown| `/cpu/self/avx/blocked` | Blocked AVX implementation | 25268e843eeSJed Brown| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 25368e843eeSJed Brown| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 25468e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 25568e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 25668e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 25768e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 25868e843eeSJed Brown| `/gpu/cuda/ref` | Reference pure CUDA kernels | 25968e843eeSJed Brown| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 26068e843eeSJed Brown| `/gpu/cuda/shared` | Optimized pure CUDA kernels using shared memory | 26168e843eeSJed Brown| `/gpu/cuda/gen` | Optimized pure CUDA kernels using code generation | 26268e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 263bcb2dfaeSJed Brown 264bcb2dfaeSJed BrownExamples available in this release: 265bcb2dfaeSJed Brown 26668e843eeSJed Brown:::{list-table} 26768e843eeSJed Brown:header-rows: 1 26868e843eeSJed Brown:widths: auto 26968e843eeSJed Brown* - User code 27068e843eeSJed Brown - Example 27168e843eeSJed Brown* - `ceed` 27268e843eeSJed Brown - * ex1 (volume) 27368e843eeSJed Brown * ex2 (surface) 27468e843eeSJed Brown* - `mfem` 27568e843eeSJed Brown - * BP1 (scalar mass operator) 27668e843eeSJed Brown * BP3 (scalar Laplace operator) 27768e843eeSJed Brown* - `petsc` 27868e843eeSJed Brown - * BP1 (scalar mass operator) 27968e843eeSJed Brown * BP2 (vector mass operator) 28068e843eeSJed Brown * BP3 (scalar Laplace operator) 28168e843eeSJed Brown * BP4 (vector Laplace operator) 28268e843eeSJed Brown * BP5 (collocated scalar Laplace operator) 28368e843eeSJed Brown * BP6 (collocated vector Laplace operator) 28468e843eeSJed Brown * Navier-Stokes 28568e843eeSJed Brown* - `nek5000` 28668e843eeSJed Brown - * BP1 (scalar mass operator) 28768e843eeSJed Brown * BP3 (scalar Laplace operator) 28868e843eeSJed Brown::: 289bcb2dfaeSJed Brown 290bcb2dfaeSJed Brown(v0-4)= 291bcb2dfaeSJed Brown 292bcb2dfaeSJed Brown## v0.4 (Apr 1, 2019) 293bcb2dfaeSJed Brown 294bcb2dfaeSJed BrownlibCEED v0.4 was made again publicly available in the second full CEED software 295bcb2dfaeSJed Browndistribution, release CEED 2.0. This release contained notable features, such as 296bcb2dfaeSJed Brownfour new CPU backends, two new GPU backends, CPU backend optimizations, initial 297bcb2dfaeSJed Brownsupport for operator composition, performance benchmarking, and a Navier-Stokes demo. 298bcb2dfaeSJed BrownThe new CPU backends in this release came in two families. The `/cpu/self/*/serial` 299bcb2dfaeSJed Brownbackends process one element at a time and are intended for meshes with a smaller number 300bcb2dfaeSJed Brownof high order elements. The `/cpu/self/*/blocked` backends process blocked batches of 301bcb2dfaeSJed Browneight interlaced elements and are intended for meshes with higher numbers of elements. 302bcb2dfaeSJed BrownThe `/cpu/self/avx/*` backends rely upon AVX instructions to provide vectorized CPU 303bcb2dfaeSJed Brownperformance. The `/cpu/self/xsmm/*` backends rely upon the 304bcb2dfaeSJed Brown[LIBXSMM](http://github.com/hfp/libxsmm) package to provide vectorized CPU 305bcb2dfaeSJed Brownperformance. The `/gpu/cuda/*` backends provide GPU performance strictly using CUDA. 306bcb2dfaeSJed BrownThe `/gpu/cuda/ref` backend is a reference CUDA backend, providing reasonable 307bcb2dfaeSJed Brownperformance for most problem configurations. The `/gpu/cuda/reg` backend uses a simple 308bcb2dfaeSJed Brownparallelization approach, where each thread treats a finite element. Using just in time 309bcb2dfaeSJed Browncompilation, provided by nvrtc (NVidia Runtime Compiler), and runtime parameters, this 310bcb2dfaeSJed Brownbackend unroll loops and map memory address to registers. The `/gpu/cuda/reg` backend 311bcb2dfaeSJed Brownachieve good peak performance for 1D, 2D, and low order 3D problems, but performance 312bcb2dfaeSJed Browndeteriorates very quickly when threads run out of registers. 313bcb2dfaeSJed Brown 314bcb2dfaeSJed BrownA new explicit time-stepping Navier-Stokes solver was added to the family of libCEED 315bcb2dfaeSJed Brownexamples in the `examples/petsc` directory (see {ref}`example-petsc-navier-stokes`). 316bcb2dfaeSJed BrownThis example solves the time-dependent Navier-Stokes equations of compressible gas 317bcb2dfaeSJed Browndynamics in a static Eulerian three-dimensional frame, using structured high-order 318bcb2dfaeSJed Brownfinite/spectral element spatial discretizations and explicit high-order time-stepping 319bcb2dfaeSJed Brown(available in PETSc). Moreover, the Navier-Stokes example was developed using PETSc, 320bcb2dfaeSJed Brownso that the pointwise physics (defined at quadrature points) is separated from the 321bcb2dfaeSJed Brownparallelization and meshing concerns. 322bcb2dfaeSJed Brown 323bcb2dfaeSJed BrownBackends available in this release: 324bcb2dfaeSJed Brown 32568e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 32668e843eeSJed Brown|--------------------------|-----------------------------------------------------| 32768e843eeSJed Brown| `/cpu/self/ref/serial` | Serial reference implementation | 32868e843eeSJed Brown| `/cpu/self/ref/blocked` | Blocked reference implementation | 32968e843eeSJed Brown| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 33068e843eeSJed Brown| `/cpu/self/avx/serial` | Serial AVX implementation | 33168e843eeSJed Brown| `/cpu/self/avx/blocked` | Blocked AVX implementation | 33268e843eeSJed Brown| `/cpu/self/xsmm/serial` | Serial LIBXSMM implementation | 33368e843eeSJed Brown| `/cpu/self/xsmm/blocked` | Blocked LIBXSMM implementation | 33468e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 33568e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 33668e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 33768e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 33868e843eeSJed Brown| `/gpu/cuda/ref` | Reference pure CUDA kernels | 33968e843eeSJed Brown| `/gpu/cuda/reg` | Pure CUDA kernels using one thread per element | 34068e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 341bcb2dfaeSJed Brown 342bcb2dfaeSJed BrownExamples available in this release: 343bcb2dfaeSJed Brown 34468e843eeSJed Brown:::{list-table} 34568e843eeSJed Brown:header-rows: 1 34668e843eeSJed Brown:widths: auto 34768e843eeSJed Brown* - User code 34868e843eeSJed Brown - Example 34968e843eeSJed Brown* - `ceed` 35068e843eeSJed Brown - * ex1 (volume) 35168e843eeSJed Brown* - `mfem` 35268e843eeSJed Brown - * BP1 (scalar mass operator) 35368e843eeSJed Brown * BP3 (scalar Laplace operator) 35468e843eeSJed Brown* - `petsc` 35568e843eeSJed Brown - * BP1 (scalar mass operator) 35668e843eeSJed Brown * BP3 (scalar Laplace operator) 35768e843eeSJed Brown * Navier-Stokes 35868e843eeSJed Brown* - `nek5000` 35968e843eeSJed Brown - * BP1 (scalar mass operator) 36068e843eeSJed Brown * BP3 (scalar Laplace operator) 36168e843eeSJed Brown::: 362bcb2dfaeSJed Brown 363bcb2dfaeSJed Brown(v0-3)= 364bcb2dfaeSJed Brown 365bcb2dfaeSJed Brown## v0.3 (Sep 30, 2018) 366bcb2dfaeSJed Brown 367bcb2dfaeSJed BrownNotable features in this release include active/passive field interface, support for 368bcb2dfaeSJed Brownnon-tensor bases, backend optimization, and improved Fortran interface. This release 369bcb2dfaeSJed Brownalso focused on providing improved continuous integration, and many new tests with code 370bcb2dfaeSJed Browncoverage reports of about 90%. This release also provided a significant change to the 371bcb2dfaeSJed Brownpublic interface: a {ref}`CeedQFunction` can take any number of named input and output 372bcb2dfaeSJed Brownarguments while {ref}`CeedOperator` connects them to the actual data, which may be 373bcb2dfaeSJed Brownsupplied explicitly to `CeedOperatorApply()` (active) or separately via 374bcb2dfaeSJed Brown`CeedOperatorSetField()` (passive). This interface change enables reusable libraries 375bcb2dfaeSJed Brownof CeedQFunctions and composition of block solvers constructed using 376bcb2dfaeSJed Brown{ref}`CeedOperator`. A concept of blocked restriction was added to this release and 377bcb2dfaeSJed Brownused in an optimized CPU backend. Although this is typically not visible to the user, 378bcb2dfaeSJed Brownit enables effective use of arbitrary-length SIMD while maintaining cache locality. 379bcb2dfaeSJed BrownThis CPU backend also implements an algebraic factorization of tensor product gradients 380bcb2dfaeSJed Brownto perform fewer operations than standard application of interpolation and 381bcb2dfaeSJed Browndifferentiation from nodes to quadrature points. This algebraic formulation 382bcb2dfaeSJed Brownautomatically supports non-polynomial and non-interpolatory bases, thus is more general 383bcb2dfaeSJed Brownthan the more common derivation in terms of Lagrange polynomials on the quadrature points. 384bcb2dfaeSJed Brown 385bcb2dfaeSJed BrownBackends available in this release: 386bcb2dfaeSJed Brown 38768e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 38868e843eeSJed Brown|-------------------------|-----------------------------------------------------| 38968e843eeSJed Brown| `/cpu/self/blocked` | Blocked reference implementation | 39068e843eeSJed Brown| `/cpu/self/ref` | Serial reference implementation | 39168e843eeSJed Brown| `/cpu/self/tmpl` | Backend template, defaults to `/cpu/self/blocked` | 39268e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 39368e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 39468e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 39568e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 39668e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 397bcb2dfaeSJed Brown 398bcb2dfaeSJed BrownExamples available in this release: 399bcb2dfaeSJed Brown 40068e843eeSJed Brown:::{list-table} 40168e843eeSJed Brown:header-rows: 1 40268e843eeSJed Brown:widths: auto 40368e843eeSJed Brown* - User code 40468e843eeSJed Brown - Example 40568e843eeSJed Brown* - `ceed` 40668e843eeSJed Brown - * ex1 (volume) 40768e843eeSJed Brown* - `mfem` 40868e843eeSJed Brown - * BP1 (scalar mass operator) 40968e843eeSJed Brown * BP3 (scalar Laplace operator) 41068e843eeSJed Brown* - `petsc` 41168e843eeSJed Brown - * BP1 (scalar mass operator) 41268e843eeSJed Brown * BP3 (scalar Laplace operator) 41368e843eeSJed Brown* - `nek5000` 41468e843eeSJed Brown - * BP1 (scalar mass operator) 41568e843eeSJed Brown * BP3 (scalar Laplace operator) 41668e843eeSJed Brown::: 417bcb2dfaeSJed Brown 418bcb2dfaeSJed Brown(v0-21)= 419bcb2dfaeSJed Brown 420bcb2dfaeSJed Brown## v0.21 (Sep 30, 2018) 421bcb2dfaeSJed Brown 422bcb2dfaeSJed BrownA MAGMA backend (which relies upon the 423bcb2dfaeSJed Brown[MAGMA](https://bitbucket.org/icl/magma) package) was integrated in libCEED for this 424bcb2dfaeSJed Brownrelease. This initial integration set up the framework of using MAGMA and provided the 425bcb2dfaeSJed BrownlibCEED functionality through MAGMA kernels as one of libCEED’s computational backends. 426bcb2dfaeSJed BrownAs any other backend, the MAGMA backend provides extended basic data structures for 427bcb2dfaeSJed Brown{ref}`CeedVector`, {ref}`CeedElemRestriction`, and {ref}`CeedOperator`, and implements 428bcb2dfaeSJed Brownthe fundamental CEED building blocks to work with the new data structures. 429bcb2dfaeSJed BrownIn general, the MAGMA-specific data structures keep the libCEED pointers to CPU data 430bcb2dfaeSJed Brownbut also add corresponding device (e.g., GPU) pointers to the data. Coherency is handled 431bcb2dfaeSJed Browninternally, and thus seamlessly to the user, through the functions/methods that are 432bcb2dfaeSJed Brownprovided to support them. 433bcb2dfaeSJed Brown 434bcb2dfaeSJed BrownBackends available in this release: 435bcb2dfaeSJed Brown 43668e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 43768e843eeSJed Brown|-------------------------|---------------------------------| 43868e843eeSJed Brown| `/cpu/self` | Serial reference implementation | 43968e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 44068e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 44168e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 44268e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 44368e843eeSJed Brown| `/gpu/magma` | CUDA MAGMA kernels | 444bcb2dfaeSJed Brown 445bcb2dfaeSJed BrownExamples available in this release: 446bcb2dfaeSJed Brown 44768e843eeSJed Brown:::{list-table} 44868e843eeSJed Brown:header-rows: 1 44968e843eeSJed Brown:widths: auto 45068e843eeSJed Brown* - User code 45168e843eeSJed Brown - Example 45268e843eeSJed Brown* - `ceed` 45368e843eeSJed Brown - * ex1 (volume) 45468e843eeSJed Brown* - `mfem` 45568e843eeSJed Brown - * BP1 (scalar mass operator) 45668e843eeSJed Brown * BP3 (scalar Laplace operator) 45768e843eeSJed Brown* - `petsc` 45868e843eeSJed Brown - * BP1 (scalar mass operator) 45968e843eeSJed Brown* - `nek5000` 46068e843eeSJed Brown - * BP1 (scalar mass operator) 46168e843eeSJed Brown::: 462bcb2dfaeSJed Brown 463bcb2dfaeSJed Brown(v0-2)= 464bcb2dfaeSJed Brown 465bcb2dfaeSJed Brown## v0.2 (Mar 30, 2018) 466bcb2dfaeSJed Brown 467bcb2dfaeSJed BrownlibCEED was made publicly available the first full CEED software distribution, release 468bcb2dfaeSJed BrownCEED 1.0. The distribution was made available using the Spack package manager to provide 469bcb2dfaeSJed Browna common, easy-to-use build environment, where the user can build the CEED distribution 470bcb2dfaeSJed Brownwith all dependencies. This release included a new Fortran interface for the library. 471bcb2dfaeSJed BrownThis release also contained major improvements in the OCCA backend (including a new 472bcb2dfaeSJed Brown`/ocl/occa` backend) and new examples. The standalone libCEED example was modified to 473bcb2dfaeSJed Browncompute the volume volume of a given mesh (in 1D, 2D, or 3D) and placed in an 474bcb2dfaeSJed Brown`examples/ceed` subfolder. A new `mfem` example to perform BP3 (with the application 475bcb2dfaeSJed Brownof the Laplace operator) was also added to this release. 476bcb2dfaeSJed Brown 477bcb2dfaeSJed BrownBackends available in this release: 478bcb2dfaeSJed Brown 47968e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 48068e843eeSJed Brown|-------------------------|---------------------------------| 48168e843eeSJed Brown| `/cpu/self` | Serial reference implementation | 48268e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 48368e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 48468e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 48568e843eeSJed Brown| `/ocl/occa` | OpenCL OCCA kernels | 486bcb2dfaeSJed Brown 487bcb2dfaeSJed BrownExamples available in this release: 488bcb2dfaeSJed Brown 48968e843eeSJed Brown:::{list-table} 49068e843eeSJed Brown:header-rows: 1 49168e843eeSJed Brown:widths: auto 49268e843eeSJed Brown* - User code 49368e843eeSJed Brown - Example 49468e843eeSJed Brown* - `ceed` 49568e843eeSJed Brown - * ex1 (volume) 49668e843eeSJed Brown* - `mfem` 49768e843eeSJed Brown - * BP1 (scalar mass operator) 49868e843eeSJed Brown * BP3 (scalar Laplace operator) 49968e843eeSJed Brown* - `petsc` 50068e843eeSJed Brown - * BP1 (scalar mass operator) 50168e843eeSJed Brown* - `nek5000` 50268e843eeSJed Brown - * BP1 (scalar mass operator) 50368e843eeSJed Brown::: 504bcb2dfaeSJed Brown 505bcb2dfaeSJed Brown(v0-1)= 506bcb2dfaeSJed Brown 507bcb2dfaeSJed Brown## v0.1 (Jan 3, 2018) 508bcb2dfaeSJed Brown 509bcb2dfaeSJed BrownInitial low-level API of the CEED project. The low-level API provides a set of Finite 510bcb2dfaeSJed BrownElements kernels and components for writing new low-level kernels. Examples include: 511bcb2dfaeSJed Brownvector and sparse linear algebra, element matrix assembly over a batch of elements, 512bcb2dfaeSJed Brownpartial assembly and action for efficient high-order operators like mass, diffusion, 513bcb2dfaeSJed Brownadvection, etc. The main goal of the low-level API is to establish the basis for the 514bcb2dfaeSJed Brownhigh-level API. Also, identifying such low-level kernels and providing a reference 515bcb2dfaeSJed Brownimplementation for them serves as the basis for specialized backend implementations. 516bcb2dfaeSJed BrownThis release contained several backends: `/cpu/self`, and backends which rely upon the 517bcb2dfaeSJed Brown[OCCA](http://github.com/libocca/occa) package, such as `/cpu/occa`, 518bcb2dfaeSJed Brown`/gpu/occa`, and `/omp/occa`. 519bcb2dfaeSJed BrownIt also included several examples, in the `examples` folder: 520bcb2dfaeSJed BrownA standalone code that shows the usage of libCEED (with no external 521bcb2dfaeSJed Browndependencies) to apply the Laplace operator, `ex1`; an `mfem` example to perform BP1 522bcb2dfaeSJed Brown(with the application of the mass operator); and a `petsc` example to perform BP1 523bcb2dfaeSJed Brown(with the application of the mass operator). 524bcb2dfaeSJed Brown 525bcb2dfaeSJed BrownBackends available in this release: 526bcb2dfaeSJed Brown 52768e843eeSJed Brown| CEED resource (`-ceed`) | Backend | 52868e843eeSJed Brown|-------------------------|---------------------------------| 52968e843eeSJed Brown| `/cpu/self` | Serial reference implementation | 53068e843eeSJed Brown| `/cpu/occa` | Serial OCCA kernels | 53168e843eeSJed Brown| `/gpu/occa` | CUDA OCCA kernels | 53268e843eeSJed Brown| `/omp/occa` | OpenMP OCCA kernels | 533bcb2dfaeSJed Brown 534bcb2dfaeSJed BrownExamples available in this release: 535bcb2dfaeSJed Brown 536bcb2dfaeSJed Brown| User code | Example | 53768e843eeSJed Brown|-----------------------|-----------------------------------| 53868e843eeSJed Brown| `ceed` | ex1 (scalar Laplace operator) | 53968e843eeSJed Brown| `mfem` | BP1 (scalar mass operator) | 54068e843eeSJed Brown| `petsc` | BP1 (scalar mass operator) | 541bcb2dfaeSJed Brown``` 542