Make new Accessors independent of Ginkgo #708

thoasm · 2021-02-21T18:15:17Z

Move the accessor scaled_reduced_row_major and reduced_row_major to the global folder accessor/ in order for others to use it without compiling Ginkgo. row_major was not copied (it must not be moved since it is part of the public interface).
A separate range was also implemented because it is needed to use the accessors.

Additionally, the following improvements / updates were performed:

Move each accessor to its own file
Add a subdirectory core/test/accessor, where the accessors are tested
Create separate test files for each accessor
Improve naming of helper functions / structures to me more descriptive

Changes done to block_col_major and row_major when migrating:

Remove gko::dim constructor parameter (creating is only possible with std::array now)
Remove copy_from functionality (since the gko::acc::range does not support it)
Change the value_type from length_type and stride_type from const to non-const

TODO:

Wait for Multidimensional row-major and column-major accessors #707 to be merged
Move accessors from Multidimensional row-major and column-major accessors #707 to accessor/ directory

thoasm · 2021-02-21T18:15:29Z

label!

Slaedr · 2021-02-23T08:12:27Z

One preliminary question - should the CB-GMRES-specific accessor functionality be in the new folder, or should it be part of Ginkgo proper? Would all multi-precision and layout helper code needed for Ginkgo end up in the new accessor folder? One option would be to keep fundamental accessor stuff in the new folder (those which would be needed for many types of things including matrices, preconditioners, solvers etc.) going forward, but keep solver-specific accessor code in the respective core (non-public) subfolders.

thoasm · 2021-02-23T19:18:21Z

Sorry, I missed this comment.
I think it should be local to each solver, unless it is truly a generic solution that can apply to all existing accessors. Otherwise, it is harder to maintain.

thoasm · 2021-02-24T06:08:14Z

format!

codecov · 2021-02-24T13:50:42Z

Codecov Report

Merging #708 (b11ca27) into develop (b293b9f) will decrease coverage by 0.01%.
The diff coverage is 97.37%.

@@             Coverage Diff             @@
##           develop     #708      +/-   ##
===========================================
- Coverage    92.42%   92.41%   -0.02%     
===========================================
  Files          369      380      +11     
  Lines        27520    27486      -34     
===========================================
- Hits         25436    25400      -36     
- Misses        2084     2086       +2

Impacted Files	Coverage Δ
core/solver/cb_gmres_accessor.hpp	`100.00% <ø> (ø)`
core/test/base/dim.cpp	`100.00% <ø> (ø)`
core/test/base/range_accessors.cpp	`100.00% <ø> (ø)`
include/ginkgo/core/base/dim.hpp	`88.88% <ø> (-2.61%)`	⬇️
include/ginkgo/core/base/range_accessors.hpp	`73.68% <ø> (-21.56%)`	⬇️
accessor/range.hpp	`81.81% <81.81%> (ø)`
accessor/reduced_row_major.hpp	`89.65% <89.65%> (ø)`
accessor/accessor_references.hpp	`94.00% <92.59%> (ø)`
accessor/accessor_helper.hpp	`94.49% <94.49%> (ø)`
accessor/block_col_major.hpp	`100.00% <100.00%> (ø)`
... and 26 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b293b9f...b11ca27. Read the comment docs.

upsj

LGTM! Only a comment on relocatability

accessor/accessor_references.hpp

accessor/accessor_helper.hpp

Slaedr

I think it mostly looks good. Thanks for taking care of the block col major as well. I just have a few doubts:

What's the purpose of the acc::range class? Just some backward compatibility? The original range class in the include directory is still there, so I'm not sure what's the need for this.
I see that the sub-range assignment tests were removed for mulitdimensional row-major and block-column-major. It's not obvious to me why that is.

And there a couple of other points/suggestions.

Slaedr · 2021-03-11T13:13:59Z

accessor/index_span.hpp

+    friend GKO_ACC_ATTRIBUTES constexpr bool operator<=(
+        const index_span &first, const index_span &second)
+    {
+        return first.end <= second.begin;


How about

Suggested change

return first.end <= second.begin;

return first.begin <= second.begin && first.end <= second.end;

I'm not sure, but maybe this is more 'natural'.

This is a 1:1 copy from the gko::span. I think making it something completely different to operator < might be even more confusing. The current way defines a partial order, if we make this change, it is no longer partially ordered.

Actually even my suggestion would define a partial order. In fact, I think mine would define a total order, because then for any two spans A and B, at least one of A <= B or B <= A is necessarily true. Iff both are true, A = B.

The existing definition is only a partial order, true, because not all spans are comparable with <=. Also it can never happen that both A <= B and B <= A, unless both A and B refer to a single point.

But I understand if you want to keep it the way it is.

accessor/range.hpp

Slaedr · 2021-03-11T13:26:16Z

accessor/accessor_references.hpp



 // CUDA TOOLKIT < 11 does not support constexpr in combination with
 // thrust::complex, which is why constexpr is only present in later versions
 #if defined(__CUDA_ARCH__) && defined(__CUDACC_VER_MAJOR__) && \
    (__CUDACC_VER_MAJOR__ < 11)

-#define GKO_ENABLE_REFERENCE_CONSTEXPR
+#define GKO_ACC_ENABLE_REFERENCE_CONSTEXPR


This is minor and it's up to you, but you could reduce the verbosity:

Suggested change

#define GKO_ACC_ENABLE_REFERENCE_CONSTEXPR

#define GKO_ACC_REFERENCE_CONSTEXPR

accessor/accessor_references.hpp

core/test/base/range_accessors.cpp

Slaedr

LGTM!

Move the accessor {scaled,}reduced_row_major to `accessor/` in order for others to use it without compiling Ginkgo. RowMajor was not copied yet (it must not be moved since it is part of the public interface).

Includes: - Add documentation to index_span - Update documentation of range - Give accessor helper functions more descriptive names

Co-authored-by: Thomas Grützmacher <thoasm@users.noreply.github.com>

Only the new multidimensional row_major is moved, the 2D row_major that is part of the public interface remains. Additionally, unnecessary code from public accessors was removed. The code that was removed was never in the release and originally intended for usage inside the accessors. Since they moved to an independent directory, that code is no longer necessary. This removal prevents it from ever becomming part of the public interface.

Inside the accessor folder, header are now included without the `accessor/` prefix. Co-authored-by: Tobias Ribizel <ribizel@kit.edu>

Changes `std::array<const size_type, dim>` to `std::array<size_type, dim>` in order to have an easier user-interface (and const-ness is preserved because all length_types and stride_types are const themselves). Additionally, add a dim_type in {scaled_,}reduced_row_major accessors.

This step was forgotten previously. Additionally, `scaled_reduced_row_major` now statically rejects complex types (since they don't work). Co-authored-by: Aditya Kashi <aditya.kashi@kit.edu>

Simply always using double braces ({{}}) for std::array initialization resolved the issue.

Now, the size is always first in the constructor, and the data pointer is followed by the corresponding stride (if existing).

pratikvn

LGTM!

Can you also maybe check if it is indeed independent by creating something like a test_install for this independent accessor and using it by not including any ginkgo code, but only the accessor headers ?

thoasm · 2021-03-15T14:15:39Z

@pratikvn what about writing the tests without including any other Ginkgo code? I only need to modify the reduced_row_major test (make 2 separate tests out of it) because I need gko::half internally.
After that, the tests are basically the test_install without checking if any public headers are included.

pratikvn · 2021-03-15T14:42:03Z

@thoasm , if it would be possible to do that, it would be great. But I think you will need core/test/utils.hpp, but maybe you need to copy that over and not include any ginkgo headers in that.

thoasm · 2021-03-15T14:54:58Z

@pratikvn I am currently setting up a custom create_independent_test, which does not link to Ginkgo in any way and only adds Ginkgo_SOURCE_DIR as the include directory. That way, if I include any public header anywhere, it fails to compile.

Thank you very much for that input, having a check if it is truly independent is quite important here (so far, I only checked that with the accessor benchmark).

The accessor tests now make sure that no dependency (on public header) exists inside the accessor headers by only having the source directory of Ginkgo as an include directory (and no longer linking against Ginkgo). Only one test (`reduced_row_major_ginkgo`) still depends on Ginkgo since it uses `gko::half`.

sonarcloud · 2021-03-16T15:52:01Z

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities
0 Security Hotspots
16 Code Smells

24.5% Coverage
1.9% Duplication

Ginkgo release 1.4.0 The Ginkgo team is proud to announce the new Ginkgo minor release 1.4.0. This release brings most of the Ginkgo functionality to the Intel DPC++ ecosystem which enables Intel-GPU and CPU execution. The only Ginkgo features which have not been ported yet are some preconditioners. Ginkgo's mixed-precision support is greatly enhanced thanks to: 1. The new Accessor concept, which allows writing kernels featuring on-the-fly memory compression, among other features. The accessor can be used as header-only, see the [accessor BLAS benchmarks repository](https://github.com/ginkgo-project/accessor-BLAS/tree/develop) as a usage example. 2. All LinOps now transparently support mixed-precision execution. By default, this is done through a temporary copy which may have a performance impact but already allows mixed-precision research. Native mixed-precision ELL kernels are implemented which do not see this cost. The accessor is also leveraged in a new CB-GMRES solver which allows for performance improvements by compressing the Krylov basis vectors. Many other features have been added to Ginkgo, such as reordering support, a new IDR solver, Incomplete Cholesky preconditioner, matrix assembly support (only CPU for now), machine topology information, and more! Supported systems and requirements: + For all platforms, cmake 3.13+ + C++14 compliant compiler + Linux and MacOS + gcc: 5.3+, 6.3+, 7.3+, all versions after 8.1+ + clang: 3.9+ + Intel compiler: 2018+ + Apple LLVM: 8.0+ + CUDA module: CUDA 9.0+ + HIP module: ROCm 3.5+ + DPC++ module: Intel OneAPI 2021.3. Set the CXX compiler to `dpcpp`. + Windows + MinGW and Cygwin: gcc 5.3+, 6.3+, 7.3+, all versions after 8.1+ + Microsoft Visual Studio: VS 2019 + CUDA module: CUDA 9.0+, Microsoft Visual Studio + OpenMP module: MinGW or Cygwin. Algorithm and important feature additions: + Add a new DPC++ Executor for SYCL execution and other base utilities [#648](#648), [#661](#661), [#757](#757), [#832](#832) + Port matrix formats, solvers and related kernels to DPC++. For some kernels, also make use of a shared kernel implementation for all executors (except Reference). [#710](#710), [#799](#799), [#779](#779), [#733](#733), [#844](#844), [#843](#843), [#789](#789), [#845](#845), [#849](#849), [#855](#855), [#856](#856) + Add accessors which allow multi-precision kernels, among other things. [#643](#643), [#708](#708) + Add support for mixed precision operations through apply in all LinOps. [#677](#677) + Add incomplete Cholesky factorizations and preconditioners as well as some improvements to ILU. [#672](#672), [#837](#837), [#846](#846) + Add an AMGX implementation and kernels on all devices but DPC++. [#528](#528), [#695](#695), [#860](#860) + Add a new mixed-precision capability solver, Compressed Basis GMRES (CB-GMRES). [#693](#693), [#763](#763) + Add the IDR(s) solver. [#620](#620) + Add a new fixed-size block CSR matrix format (for the Reference executor). [#671](#671), [#730](#730) + Add native mixed-precision support to the ELL format. [#717](#717), [#780](#780) + Add Reverse Cuthill-McKee reordering [#500](#500), [#649](#649) + Add matrix assembly support on CPUs. [#644](#644) + Extends ISAI from triangular to general and spd matrices. [#690](#690) Other additions: + Add the possibility to apply real matrices to complex vectors. [#655](#655), [#658](#658) + Add functions to compute the absolute of a matrix format. [#636](#636) + Add symmetric permutation and improve existing permutations. [#684](#684), [#657](#657), [#663](#663) + Add a MachineTopology class with HWLOC support [#554](#554), [#697](#697) + Add an implicit residual norm criterion. [#702](#702), [#818](#818), [#850](#850) + Row-major accessor is generalized to more than 2 dimensions and a new "block column-major" accessor has been added. [#707](#707) + Add an heat equation example. [#698](#698), [#706](#706) + Add ccache support in CMake and CI. [#725](#725), [#739](#739) + Allow tuning and benchmarking variables non intrusively. [#692](#692) + Add triangular solver benchmark [#664](#664) + Add benchmarks for BLAS operations [#772](#772), [#829](#829) + Add support for different precisions and consistent index types in benchmarks. [#675](#675), [#828](#828) + Add a Github bot system to facilitate development and PR management. [#667](#667), [#674](#674), [#689](#689), [#853](#853) + Add Intel (DPC++) CI support and enable CI on HPC systems. [#736](#736), [#751](#751), [#781](#781) + Add ssh debugging for Github Actions CI. [#749](#749) + Add pipeline segmentation for better CI speed. [#737](#737) Changes: + Add a Scalar Jacobi specialization and kernels. [#808](#808), [#834](#834), [#854](#854) + Add implicit residual log for solvers and benchmarks. [#714](#714) + Change handling of the conjugate in the dense dot product. [#755](#755) + Improved Dense stride handling. [#774](#774) + Multiple improvements to the OpenMP kernels performance, including COO, an exclusive prefix sum, and more. [#703](#703), [#765](#765), [#740](#740) + Allow specialization of submatrix and other dense creation functions in solvers. [#718](#718) + Improved Identity constructor and treatment of rectangular matrices. [#646](#646) + Allow CUDA/HIP executors to select allocation mode. [#758](#758) + Check if executors share the same memory. [#670](#670) + Improve test install and smoke testing support. [#721](#721) + Update the JOSS paper citation and add publications in the documentation. [#629](#629), [#724](#724) + Improve the version output. [#806](#806) + Add some utilities for dim and span. [#821](#821) + Improved solver and preconditioner benchmarks. [#660](#660) + Improve benchmark timing and output. [#669](#669), [#791](#791), [#801](#801), [#812](#812) Fixes: + Sorting fix for the Jacobi preconditioner. [#659](#659) + Also log the first residual norm in CGS [#735](#735) + Fix BiCG and HIP CSR to work with complex matrices. [#651](#651) + Fix Coo SpMV on strided vectors. [#807](#807) + Fix segfault of extract_diagonal, add short-and-fat test. [#769](#769) + Fix device_reset issue by moving counter/mutex to device. [#810](#810) + Fix `EnableLogging` superclass. [#841](#841) + Support ROCm 4.1.x and breaking HIP_PLATFORM changes. [#726](#726) + Decreased test size for a few device tests. [#742](#742) + Fix multiple issues with our CMake HIP and RPATH setup. [#712](#712), [#745](#745), [#709](#709) + Cleanup our CMake installation step. [#713](#713) + Various simplification and fixes to the Windows CMake setup. [#720](#720), [#785](#785) + Simplify third-party integration. [#786](#786) + Improve Ginkgo device arch flags management. [#696](#696) + Other fixes and improvements to the CMake setup. [#685](#685), [#792](#792), [#705](#705), [#836](#836) + Clarification of dense norm documentation [#784](#784) + Various development tools fixes and improvements [#738](#738), [#830](#830), [#840](#840) + Make multiple operators/constructors explicit. [#650](#650), [#761](#761) + Fix some issues, memory leaks and warnings found by MSVC. [#666](#666), [#731](#731) + Improved solver memory estimates and consistent iteration counts [#691](#691) + Various logger improvements and fixes [#728](#728), [#743](#743), [#754](#754) + Fix for ForwardIterator requirements in iterator_factory. [#665](#665) + Various benchmark fixes. [#647](#647), [#673](#673), [#722](#722) + Various CI fixes and improvements. [#642](#642), [#641](#641), [#795](#795), [#783](#783), [#793](#793), [#852](#852) Related PR: #857

Release 1.4.0 to master The Ginkgo team is proud to announce the new Ginkgo minor release 1.4.0. This release brings most of the Ginkgo functionality to the Intel DPC++ ecosystem which enables Intel-GPU and CPU execution. The only Ginkgo features which have not been ported yet are some preconditioners. Ginkgo's mixed-precision support is greatly enhanced thanks to: 1. The new Accessor concept, which allows writing kernels featuring on-the-fly memory compression, among other features. The accessor can be used as header-only, see the [accessor BLAS benchmarks repository](https://github.com/ginkgo-project/accessor-BLAS/tree/develop) as a usage example. 2. All LinOps now transparently support mixed-precision execution. By default, this is done through a temporary copy which may have a performance impact but already allows mixed-precision research. Native mixed-precision ELL kernels are implemented which do not see this cost. The accessor is also leveraged in a new CB-GMRES solver which allows for performance improvements by compressing the Krylov basis vectors. Many other features have been added to Ginkgo, such as reordering support, a new IDR solver, Incomplete Cholesky preconditioner, matrix assembly support (only CPU for now), machine topology information, and more! Supported systems and requirements: + For all platforms, cmake 3.13+ + C++14 compliant compiler + Linux and MacOS + gcc: 5.3+, 6.3+, 7.3+, all versions after 8.1+ + clang: 3.9+ + Intel compiler: 2018+ + Apple LLVM: 8.0+ + CUDA module: CUDA 9.0+ + HIP module: ROCm 3.5+ + DPC++ module: Intel OneAPI 2021.3. Set the CXX compiler to `dpcpp`. + Windows + MinGW and Cygwin: gcc 5.3+, 6.3+, 7.3+, all versions after 8.1+ + Microsoft Visual Studio: VS 2019 + CUDA module: CUDA 9.0+, Microsoft Visual Studio + OpenMP module: MinGW or Cygwin. Algorithm and important feature additions: + Add a new DPC++ Executor for SYCL execution and other base utilities [#648](#648), [#661](#661), [#757](#757), [#832](#832) + Port matrix formats, solvers and related kernels to DPC++. For some kernels, also make use of a shared kernel implementation for all executors (except Reference). [#710](#710), [#799](#799), [#779](#779), [#733](#733), [#844](#844), [#843](#843), [#789](#789), [#845](#845), [#849](#849), [#855](#855), [#856](#856) + Add accessors which allow multi-precision kernels, among other things. [#643](#643), [#708](#708) + Add support for mixed precision operations through apply in all LinOps. [#677](#677) + Add incomplete Cholesky factorizations and preconditioners as well as some improvements to ILU. [#672](#672), [#837](#837), [#846](#846) + Add an AMGX implementation and kernels on all devices but DPC++. [#528](#528), [#695](#695), [#860](#860) + Add a new mixed-precision capability solver, Compressed Basis GMRES (CB-GMRES). [#693](#693), [#763](#763) + Add the IDR(s) solver. [#620](#620) + Add a new fixed-size block CSR matrix format (for the Reference executor). [#671](#671), [#730](#730) + Add native mixed-precision support to the ELL format. [#717](#717), [#780](#780) + Add Reverse Cuthill-McKee reordering [#500](#500), [#649](#649) + Add matrix assembly support on CPUs. [#644](#644) + Extends ISAI from triangular to general and spd matrices. [#690](#690) Other additions: + Add the possibility to apply real matrices to complex vectors. [#655](#655), [#658](#658) + Add functions to compute the absolute of a matrix format. [#636](#636) + Add symmetric permutation and improve existing permutations. [#684](#684), [#657](#657), [#663](#663) + Add a MachineTopology class with HWLOC support [#554](#554), [#697](#697) + Add an implicit residual norm criterion. [#702](#702), [#818](#818), [#850](#850) + Row-major accessor is generalized to more than 2 dimensions and a new "block column-major" accessor has been added. [#707](#707) + Add an heat equation example. [#698](#698), [#706](#706) + Add ccache support in CMake and CI. [#725](#725), [#739](#739) + Allow tuning and benchmarking variables non intrusively. [#692](#692) + Add triangular solver benchmark [#664](#664) + Add benchmarks for BLAS operations [#772](#772), [#829](#829) + Add support for different precisions and consistent index types in benchmarks. [#675](#675), [#828](#828) + Add a Github bot system to facilitate development and PR management. [#667](#667), [#674](#674), [#689](#689), [#853](#853) + Add Intel (DPC++) CI support and enable CI on HPC systems. [#736](#736), [#751](#751), [#781](#781) + Add ssh debugging for Github Actions CI. [#749](#749) + Add pipeline segmentation for better CI speed. [#737](#737) Changes: + Add a Scalar Jacobi specialization and kernels. [#808](#808), [#834](#834), [#854](#854) + Add implicit residual log for solvers and benchmarks. [#714](#714) + Change handling of the conjugate in the dense dot product. [#755](#755) + Improved Dense stride handling. [#774](#774) + Multiple improvements to the OpenMP kernels performance, including COO, an exclusive prefix sum, and more. [#703](#703), [#765](#765), [#740](#740) + Allow specialization of submatrix and other dense creation functions in solvers. [#718](#718) + Improved Identity constructor and treatment of rectangular matrices. [#646](#646) + Allow CUDA/HIP executors to select allocation mode. [#758](#758) + Check if executors share the same memory. [#670](#670) + Improve test install and smoke testing support. [#721](#721) + Update the JOSS paper citation and add publications in the documentation. [#629](#629), [#724](#724) + Improve the version output. [#806](#806) + Add some utilities for dim and span. [#821](#821) + Improved solver and preconditioner benchmarks. [#660](#660) + Improve benchmark timing and output. [#669](#669), [#791](#791), [#801](#801), [#812](#812) Fixes: + Sorting fix for the Jacobi preconditioner. [#659](#659) + Also log the first residual norm in CGS [#735](#735) + Fix BiCG and HIP CSR to work with complex matrices. [#651](#651) + Fix Coo SpMV on strided vectors. [#807](#807) + Fix segfault of extract_diagonal, add short-and-fat test. [#769](#769) + Fix device_reset issue by moving counter/mutex to device. [#810](#810) + Fix `EnableLogging` superclass. [#841](#841) + Support ROCm 4.1.x and breaking HIP_PLATFORM changes. [#726](#726) + Decreased test size for a few device tests. [#742](#742) + Fix multiple issues with our CMake HIP and RPATH setup. [#712](#712), [#745](#745), [#709](#709) + Cleanup our CMake installation step. [#713](#713) + Various simplification and fixes to the Windows CMake setup. [#720](#720), [#785](#785) + Simplify third-party integration. [#786](#786) + Improve Ginkgo device arch flags management. [#696](#696) + Other fixes and improvements to the CMake setup. [#685](#685), [#792](#792), [#705](#705), [#836](#836) + Clarification of dense norm documentation [#784](#784) + Various development tools fixes and improvements [#738](#738), [#830](#830), [#840](#840) + Make multiple operators/constructors explicit. [#650](#650), [#761](#761) + Fix some issues, memory leaks and warnings found by MSVC. [#666](#666), [#731](#731) + Improved solver memory estimates and consistent iteration counts [#691](#691) + Various logger improvements and fixes [#728](#728), [#743](#743), [#754](#754) + Fix for ForwardIterator requirements in iterator_factory. [#665](#665) + Various benchmark fixes. [#647](#647), [#673](#673), [#722](#722) + Various CI fixes and improvements. [#642](#642), [#641](#641), [#795](#795), [#783](#783), [#793](#793), [#852](#852) Related PR: #866

thoasm self-assigned this Feb 21, 2021

ginkgo-bot added mod:core This is related to the core module. reg:testing This is related to testing. labels Feb 21, 2021

thoasm added the 1:ST:WIP This PR is a work in progress. Not ready for review. label Feb 21, 2021

thoasm force-pushed the independent_accessor branch from 0709e82 to 799d5ce Compare February 22, 2021 11:30

thoasm mentioned this pull request Feb 22, 2021

Multidimensional row-major and column-major accessors #707

Merged

thoasm force-pushed the independent_accessor branch from 799d5ce to d0b63e3 Compare February 24, 2021 06:03

thoasm force-pushed the independent_accessor branch 2 times, most recently from 09a4ffa to f42bf15 Compare February 24, 2021 22:49

thoasm added 1:ST:ready-for-review This PR is ready for review and removed 1:ST:WIP This PR is a work in progress. Not ready for review. labels Mar 9, 2021

thoasm requested review from tcojean, pratikvn, fritzgoebel, hartwiganzt, Slaedr, upsj and yhmtsai March 9, 2021 10:06

upsj approved these changes Mar 9, 2021

View reviewed changes

accessor/accessor_references.hpp Outdated Show resolved Hide resolved

accessor/accessor_helper.hpp Outdated Show resolved Hide resolved

hartwiganzt approved these changes Mar 9, 2021

View reviewed changes

fritzgoebel mentioned this pull request Mar 10, 2021

Mixed Precision ISAI #719

Closed

3 tasks

thoasm force-pushed the independent_accessor branch 2 times, most recently from 92af985 to e8c413f Compare March 10, 2021 19:00

Slaedr reviewed Mar 11, 2021

View reviewed changes

thoasm force-pushed the independent_accessor branch from 04a440d to b529750 Compare March 12, 2021 09:30

thoasm changed the title ~~Make reduced storage Accessors independent of Ginkgo~~ Make new Accessors independent of Ginkgo Mar 12, 2021

Slaedr approved these changes Mar 12, 2021

View reviewed changes

thoasm added 1:ST:ready-to-merge This PR is ready to merge. and removed 1:ST:ready-for-review This PR is ready for review labels Mar 15, 2021

Thomas Grützmacher and others added 13 commits March 15, 2021 11:37

Make reduced_row_major independent of Ginkgo

4ae4842

Move the accessor {scaled,}reduced_row_major to `accessor/` in order for others to use it without compiling Ginkgo. RowMajor was not copied yet (it must not be moved since it is part of the public interface).

Adopt the CB-GMRES to the accessor changes

3494cec

Small improvements to the accessor directory

9152d8f

Includes: - Add documentation to index_span - Update documentation of range - Give accessor helper functions more descriptive names

Move the accessor tests to their own directory

b938453

Sort and re-format contributors.txt

0d76469

Format files

4103930

Co-authored-by: Thomas Grützmacher <thoasm@users.noreply.github.com>

Inside accessor, include header locally

d26fc36

Inside the accessor folder, header are now included without the `accessor/` prefix. Co-authored-by: Tobias Ribizel <ribizel@kit.edu>

Prevent narrowing casting in row_major accessor

ffd954f

Remove copy_from functionality from Accessors

fc5e6d4

This step was forgotten previously. Additionally, `scaled_reduced_row_major` now statically rejects complex types (since they don't work). Co-authored-by: Aditya Kashi <aditya.kashi@kit.edu>

Resolve compiler error with GCC 5.4

09a3a79

Simply always using double braces ({{}}) for std::array initialization resolved the issue.

Change Accessor constructor order

ab10e82

Now, the size is always first in the constructor, and the data pointer is followed by the corresponding stride (if existing).

thoasm force-pushed the independent_accessor branch from f7f8540 to ab10e82 Compare March 15, 2021 10:37

pratikvn approved these changes Mar 15, 2021

View reviewed changes

thoasm merged commit 3e31cc0 into develop Mar 16, 2021

thoasm deleted the independent_accessor branch March 16, 2021 17:58

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make new Accessors independent of Ginkgo #708

Make new Accessors independent of Ginkgo #708

thoasm commented Feb 21, 2021 •

edited

Loading

thoasm commented Feb 21, 2021

Slaedr commented Feb 23, 2021

thoasm commented Feb 23, 2021

thoasm commented Feb 24, 2021

codecov bot commented Feb 24, 2021 •

edited

Loading

upsj left a comment

Slaedr left a comment

Slaedr Mar 11, 2021

thoasm Mar 12, 2021

Slaedr Mar 12, 2021 •

edited

Loading

Slaedr Mar 11, 2021

Slaedr left a comment

pratikvn left a comment

thoasm commented Mar 15, 2021

pratikvn commented Mar 15, 2021

thoasm commented Mar 15, 2021

sonarcloud bot commented Mar 16, 2021

	return first.end <= second.begin;
	return first.begin <= second.begin && first.end <= second.end;

	#define GKO_ACC_ENABLE_REFERENCE_CONSTEXPR
	#define GKO_ACC_REFERENCE_CONSTEXPR

Make new Accessors independent of Ginkgo #708

Make new Accessors independent of Ginkgo #708

Conversation

thoasm commented Feb 21, 2021 • edited Loading

thoasm commented Feb 21, 2021

Slaedr commented Feb 23, 2021

thoasm commented Feb 23, 2021

thoasm commented Feb 24, 2021

codecov bot commented Feb 24, 2021 • edited Loading

Codecov Report

upsj left a comment

Choose a reason for hiding this comment

Slaedr left a comment

Choose a reason for hiding this comment

Slaedr Mar 11, 2021

Choose a reason for hiding this comment

thoasm Mar 12, 2021

Choose a reason for hiding this comment

Slaedr Mar 12, 2021 • edited Loading

Choose a reason for hiding this comment

Slaedr Mar 11, 2021

Choose a reason for hiding this comment

Slaedr left a comment

Choose a reason for hiding this comment

pratikvn left a comment

Choose a reason for hiding this comment

thoasm commented Mar 15, 2021

pratikvn commented Mar 15, 2021

thoasm commented Mar 15, 2021

sonarcloud bot commented Mar 16, 2021

thoasm commented Feb 21, 2021 •

edited

Loading

codecov bot commented Feb 24, 2021 •

edited

Loading

Slaedr Mar 12, 2021 •

edited

Loading