Simplify GMRES kernels #861

upsj · 2021-08-18T15:51:09Z

This PR separates the step1 and initialize2 kernels into individual reductions (norm and dot) and axpy/scale operations, which allows us to use the simple kernel setup for all of GMRES as well. This will also simplify the addition of CGS-Arnoldi to plain GMRES (and distributed GMRES later on)

The branch is based on simple_kernel_reduction, which is why the changes are a bit obscured. But the base isn't strictly necessary, so I could remove it.

TODO:

Add reference kernel tests
Fix DPC++
Fix complex GMRES (since we don't have real * complex scal operation yet - Add support for mixed real/complex BLAS operations #864 )

yhmtsai

about the performance check, do you also check for multiple right hand side?

core/solver/cb_gmres.cpp

common/unified/solver/common_gmres_kernels.cpp

common/unified/solver/gmres_kernels.cpp

yhmtsai · 2022-10-31T09:13:30Z

core/solver/gmres.cpp

-                .check(RelativeStoppingId, true, &stop_status, &one_changed)) {
+                .check(RelativeStoppingId, false, &stop_status, &one_changed)) {


Could you remind me why you change true to false?

core/solver/gmres.cpp

reference/solver/common_gmres_kernels.cpp

test/solver/gmres_kernels.cpp

yhmtsai · 2022-10-31T09:28:18Z

test/solver/gmres_kernels.cpp


    ref_solver->apply(b.get(), x.get());
    exec_solver->apply(d_b.get(), d_x.get());

-    GKO_ASSERT_MTX_NEAR(d_x, x, r<value_type>::value * 100);
+    GKO_ASSERT_MTX_NEAR(d_b, b, 0);
+    GKO_ASSERT_MTX_NEAR(d_x, x, r<value_type>::value * 1e3);


do you know which case needs larger criterion?

No, I just left the values in there it already had. I am also not sure why d_b and b is checked as it should be a direct copy.

sonarqubecloud · 2022-11-01T10:12:00Z

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities
0 Security Hotspots
65 Code Smells

66.1% Coverage
8.9% Duplication

thoasm · 2022-11-02T16:59:54Z

@yhmtsai I ran the same benchmark with 4 RHS:

Matrix name	develop iters	develop time [s]	gmres_simplify iters	gmres_simplify time [s]
G3_circuit	704	9.09	704	8.10
t2em	219	1.55	219	1.49
circuit5M_dc	42	0.65	42	0.55
audikw_1	17169	196.05	12747	140.18
Bump_2911	40862	1205.90	40368	1058.94
ecology1	1219	9.81	1219	9.28
ss	1089	16.14	1089	14.50
mc2depi	3345	13.10	3294	13.78

Only mc2depi seems to be slower, the rest is faster with this PR.

tcojean · 2022-11-02T17:11:16Z

Nice results, thanks Thomas.

yhmtsai

I do not review the answer of hessenbergqr yet.
should cb_gmres use workspace?
others look good to me

common/unified/solver/gmres_kernels.cpp

yhmtsai · 2022-11-02T18:35:33Z

core/solver/cb_gmres.cpp

+        auto hessenberg =
+            Vector::create(exec, dim<2>{krylov_dim + 1, krylov_dim * num_rhs});
+        auto buffer = Vector::create(exec, dim<2>{krylov_dim + 1, num_rhs});
+        auto givens_sin = Vector::create(exec, dim<2>{krylov_dim, num_rhs});
+        auto givens_cos = Vector::create(exec, dim<2>{krylov_dim, num_rhs});
+        auto residual_norm_collection =
+            Vector::create(exec, dim<2>{krylov_dim + 1, num_rhs});
+        auto residual_norm = VectorNorms::create(exec, dim<2>{1, num_rhs});


If I remember correctly, it was not added to the workspace because the at least the krylov_bases can't use the workspace (as the type is runtime dependent).

The logic behind not adopting CB-GMRES to the workspace was that it will be done when CB-GMRES is also getting the simplified kernels. Currently, only kernel names have been changed (and num_rhs is used instead of ->get_size()[1]).
I would like to adopt the workspace that in a separate PR (all at once including the krylov_bases).

reference/solver/common_gmres_kernels.cpp

reference/test/solver/gmres_kernels.cpp

dpcpp/solver/cb_gmres_kernels.dp.cpp

core/solver/gmres.cpp

yhmtsai

I finish the HessenbergQr answer check.
I am confusing of the documentation and updating iteration.
from the memory movement, it seems to d iteration for (0-(d-1) update + one restart).
from the updating iteration, it seems to 1 iteration contains (0-(d-1) update + one restart).?

reference/test/solver/gmres_kernels.cpp

thoasm · 2022-11-03T13:46:21Z

format-rebase!

ginkgo-bot · 2022-11-03T13:48:02Z

Formatting rebase introduced changes, see Artifacts here to review them

ginkgo-bot · 2022-11-03T13:58:33Z

Note: This PR changes the Ginkgo ABI:

Functions changes summary: 0 Removed, 0 Changed (16 filtered out), 6 Added functions
Variables changes summary: 0 Removed, 0 Changed, 0 Added variable

For details check the full ABI diff under Artifacts here

yhmtsai

The finalize is in multi_axpy now and the iteration count is still the same as develop.
LGTM.

Additionally, use temporary reduction array for norms in GMRES

Additionally, make the residual_norm parameter const in the restart kernel (as it was never written to).

Fix GMRES reference test initialization and improve memory read efficiency of hessenberg_qr. Co-authored-by: Pratik Nayak <pratikvn@protonmail.com> Co-authored-by: Yuhsiang M. Tsai <yhmtsai@gmail.com>

Add additional reference GMRES test, update parts of documentation Co-authored-by: Yuhsiang M. Tsai <yhmtsai@gmail.com>

ginkgo-bot · 2022-11-03T20:29:57Z

Error: PR already merged!

Advertise release 1.5.0 and last changes + Add changelog, + Update third party libraries + A small fix to a CMake file See PR: #1195 The Ginkgo team is proud to announce the new Ginkgo minor release 1.5.0. This release brings many important new features such as: - MPI-based multi-node support for all matrix formats and most solvers; - full DPC++/SYCL support, - functionality and interface for GPU-resident sparse direct solvers, - an interface for wrapping solvers with scaling and reordering applied, - a new algebraic Multigrid solver/preconditioner, - improved mixed-precision support, - support for device matrix assembly, and much more. If you face an issue, please first check our [known issues page](https://github.com/ginkgo-project/ginkgo/wiki/Known-Issues) and the [open issues list](https://github.com/ginkgo-project/ginkgo/issues) and if you do not find a solution, feel free to [open a new issue](https://github.com/ginkgo-project/ginkgo/issues/new/choose) or ask a question using the [github discussions](https://github.com/ginkgo-project/ginkgo/discussions). Supported systems and requirements: + For all platforms, CMake 3.13+ + C++14 compliant compiler + Linux and macOS + GCC: 5.5+ + clang: 3.9+ + Intel compiler: 2018+ + Apple LLVM: 8.0+ + NVHPC: 22.7+ + Cray Compiler: 14.0.1+ + CUDA module: CUDA 9.2+ or NVHPC 22.7+ + HIP module: ROCm 4.0+ + DPC++ module: Intel OneAPI 2021.3 with oneMKL and oneDPL. Set the CXX compiler to `dpcpp`. + Windows + MinGW and Cygwin: GCC 5.5+ + Microsoft Visual Studio: VS 2019 + CUDA module: CUDA 9.2+, Microsoft Visual Studio + OpenMP module: MinGW or Cygwin. Algorithm and important feature additions: + Add MPI-based multi-node for all matrix formats and solvers (except GMRES and IDR). ([#676](#676), [#908](#908), [#909](#909), [#932](#932), [#951](#951), [#961](#961), [#971](#971), [#976](#976), [#985](#985), [#1007](#1007), [#1030](#1030), [#1054](#1054), [#1100](#1100), [#1148](#1148)) + Porting the remaining algorithms (preconditioners like ISAI, Jacobi, Multigrid, ParILU(T) and ParIC(T)) to DPC++/SYCL, update to SYCL 2020, and improve support and performance ([#896](#896), [#924](#924), [#928](#928), [#929](#929), [#933](#933), [#943](#943), [#960](#960), [#1057](#1057), [#1110](#1110), [#1142](#1142)) + Add a Sparse Direct interface supporting GPU-resident numerical LU factorization, symbolic Cholesky factorization, improved triangular solvers, and more ([#957](#957), [#1058](#1058), [#1072](#1072), [#1082](#1082)) + Add a ScaleReordered interface that can wrap solvers and automatically apply reorderings and scalings ([#1059](#1059)) + Add a Multigrid solver and improve the aggregation based PGM coarsening scheme ([#542](#542), [#913](#913), [#980](#980), [#982](#982), [#986](#986)) + Add infrastructure for unified, lambda-based, backend agnostic, kernels and utilize it for some simple kernels ([#833](#833), [#910](#910), [#926](#926)) + Merge different CUDA, HIP, DPC++ and OpenMP tests under a common interface ([#904](#904), [#973](#973), [#1044](#1044), [#1117](#1117)) + Add a device_matrix_data type for device-side matrix assembly ([#886](#886), [#963](#963), [#965](#965)) + Add support for mixed real/complex BLAS operations ([#864](#864)) + Add a FFT LinOp for all but DPC++/SYCL ([#701](#701)) + Add FBCSR support for NVIDIA and AMD GPUs and CPUs with OpenMP ([#775](#775)) + Add CSR scaling ([#848](#848)) + Add array::const_view and equivalent to create constant matrices from non-const data ([#890](#890)) + Add a RowGatherer LinOp supporting mixed precision to gather dense matrix rows ([#901](#901)) + Add mixed precision SparsityCsr SpMV support ([#970](#970)) + Allow creating CSR submatrix including from (possibly discontinuous) index sets ([#885](#885), [#964](#964)) + Add a scaled identity addition (M <- aI + bM) feature interface and impls for Csr and Dense ([#942](#942)) Deprecations and important changes: + Deprecate AmgxPgm in favor of the new Pgm name. ([#1149](#1149)). + Deprecate specialized residual norm classes in favor of a common `ResidualNorm` class ([#1101](#1101)) + Deprecate CamelCase non-polymorphic types in favor of snake_case versions (like array, machine_topology, uninitialized_array, index_set) ([#1031](#1031), [#1052](#1052)) + Bug fix: restrict gko::share to rvalue references (*possible interface break*) ([#1020](#1020)) + Bug fix: when using cuSPARSE's triangular solvers, specifying the factory parameter `num_rhs` is now required when solving for more than one right-hand side, otherwise an exception is thrown ([#1184](#1184)). + Drop official support for old CUDA < 9.2 ([#887](#887)) Improved performance additions: + Reuse tmp storage in reductions in solvers and add a mutable workspace to all solvers ([#1013](#1013), [#1028](#1028)) + Add HIP unsafe atomic option for AMD ([#1091](#1091)) + Prefer vendor implementations for Dense dot, conj_dot and norm2 when available ([#967](#967)). + Tuned OpenMP SellP, COO, and ELL SpMV kernels for a small number of RHS ([#809](#809)) Fixes: + Fix various compilation warnings ([#1076](#1076), [#1183](#1183), [#1189](#1189)) + Fix issues with hwloc-related tests ([#1074](#1074)) + Fix include headers for GCC 12 ([#1071](#1071)) + Fix for simple-solver-logging example ([#1066](#1066)) + Fix for potential memory leak in Logger ([#1056](#1056)) + Fix logging of mixin classes ([#1037](#1037)) + Improve value semantics for LinOp types, like moved-from state in cross-executor copy/clones ([#753](#753)) + Fix some matrix SpMV and conversion corner cases ([#905](#905), [#978](#978)) + Fix uninitialized data ([#958](#958)) + Fix CUDA version requirement for cusparseSpSM ([#953](#953)) + Fix several issues within bash-script ([#1016](#1016)) + Fixes for `NVHPC` compiler support ([#1194](#1194)) Other additions: + Simplify and properly name GMRES kernels ([#861](#861)) + Improve pkg-config support for non-CMake libraries ([#923](#923), [#1109](#1109)) + Improve gdb pretty printer ([#987](#987), [#1114](#1114)) + Add a logger highlighting inefficient allocation and copy patterns ([#1035](#1035)) + Improved and optimized test random matrix generation ([#954](#954), [#1032](#1032)) + Better CSR strategy defaults ([#969](#969)) + Add `move_from` to `PolymorphicObject` ([#997](#997)) + Remove unnecessary device_guard usage ([#956](#956)) + Improvements to the generic accessor for mixed-precision ([#727](#727)) + Add a naive lower triangular solver implementation for CUDA ([#764](#764)) + Add support for int64 indices from CUDA 11 onward with SpMV and SpGEMM ([#897](#897)) + Add a L1 norm implementation ([#900](#900)) + Add reduce_add for arrays ([#831](#831)) + Add utility to simplify Dense View creation from an existing Dense vector ([#1136](#1136)). + Add a custom transpose implementation for Fbcsr and Csr transpose for unsupported vendor types ([#1123](#1123)) + Make IDR random initilization deterministic ([#1116](#1116)) + Move the algorithm choice for triangular solvers from Csr::strategy_type to a factory parameter ([#1088](#1088)) + Update CUDA archCoresPerSM ([#1175](#1116)) + Add kernels for Csr sparsity pattern lookup ([#994](#994)) + Differentiate between structural and numerical zeros in Ell/Sellp ([#1027](#1027)) + Add a binary IO format for matrix data ([#984](#984)) + Add a tuple zip_iterator implementation ([#966](#966)) + Simplify kernel stubs and declarations ([#888](#888)) + Simplify GKO_REGISTER_OPERATION with lambdas ([#859](#859)) + Simplify copy to device in tests and examples ([#863](#863)) + More verbose output to array assertions ([#858](#858)) + Allow parallel compilation for Jacobi kernels ([#871](#871)) + Change clang-format pointer alignment to left ([#872](#872)) + Various improvements and fixes to the benchmarking framework ([#750](#750), [#759](#759), [#870](#870), [#911](#911), [#1033](#1033), [#1137](#1137)) + Various documentation improvements ([#892](#892), [#921](#921), [#950](#950), [#977](#977), [#1021](#1021), [#1068](#1068), [#1069](#1069), [#1080](#1080), [#1081](#1081), [#1108](#1108), [#1153](#1153), [#1154](#1154)) + Various CI improvements ([#868](#868), [#874](#874), [#884](#884), [#889](#889), [#899](#899), [#903](#903), [#922](#922), [#925](#925), [#930](#930), [#936](#936), [#937](#937), [#958](#958), [#882](#882), [#1011](#1011), [#1015](#1015), [#989](#989), [#1039](#1039), [#1042](#1042), [#1067](#1067), [#1073](#1073), [#1075](#1075), [#1083](#1083), [#1084](#1084), [#1085](#1085), [#1139](#1139), [#1178](#1178), [#1187](#1187))

upsj self-assigned this Aug 18, 2021

upsj requested a review from a team August 18, 2021 15:51

upsj added this to the Ginkgo 1.5.0 milestone Aug 18, 2021

upsj mentioned this pull request Aug 19, 2021

Add support for mixed real/complex BLAS operations #864

Merged

upsj force-pushed the gmres_simplify branch from e1c255c to 1099dc7 Compare August 19, 2021 16:02

upsj added the 1:ST:need-feedback The PR is somewhat ready but feedback on a blocking topic is required before a proper review. label Oct 6, 2021

upsj force-pushed the gmres_simplify branch from 1099dc7 to a703e77 Compare December 14, 2021 12:55

upsj mentioned this pull request Dec 20, 2021

Problems compiling with the NVHPC suite #941

Closed

upsj mentioned this pull request Jan 23, 2022

Format conversion, SpMV and write improvements #905

Merged

upsj mentioned this pull request Feb 11, 2022

Compatibility to SYCL 2020 #943

Merged

upsj mentioned this pull request Feb 22, 2022

Add generic solver test #973

Merged

1 task

MarcelKoch mentioned this pull request Feb 24, 2022

Adds distributed support for several solvers #976

Merged

3 tasks

upsj added 1:ST:WIP This PR is a work in progress. Not ready for review. and removed 1:ST:need-feedback The PR is somewhat ready but feedback on a blocking topic is required before a proper review. labels Mar 10, 2022

upsj mentioned this pull request Apr 7, 2022

Fix (CB-)GMRES failing if initial guess is solution #1010

Closed

upsj force-pushed the gmres_simplify branch from a703e77 to f050cfd Compare April 26, 2022 10:21

upsj force-pushed the gmres_simplify branch 2 times, most recently from 5ef2d18 to e6f49ac Compare May 9, 2022 12:07

upsj 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 May 9, 2022

upsj requested review from a team and removed request for a team May 9, 2022 12:09

yhmtsai reviewed Oct 31, 2022

View reviewed changes

thoasm force-pushed the gmres_simplify branch from 2205766 to f230764 Compare October 31, 2022 16:49

yhmtsai reviewed Nov 2, 2022

View reviewed changes

thoasm added the 1:ST:run-full-test label Nov 3, 2022

yhmtsai reviewed Nov 3, 2022

View reviewed changes

reference/test/solver/gmres_kernels.cpp Show resolved Hide resolved

thoasm force-pushed the gmres_simplify branch from 72d158d to c095b76 Compare November 3, 2022 13:17

ginkgo-bot force-pushed the gmres_simplify branch from c095b76 to 42c245d Compare November 3, 2022 13:48

yhmtsai approved these changes Nov 3, 2022

View reviewed changes

thoasm added the 1:ST:ready-to-merge This PR is ready to merge. label Nov 3, 2022

upsj and others added 7 commits November 3, 2022 20:15

simplify GMRES kernels

963b1a3

rename cbgmres kernels

72bfffc

extract commonly used GMRES kernels

fc37ab6

Fix norm issue for complex GMRES solves

6124820

Additionally, use temporary reduction array for norms in GMRES

Add Kernel tests for reference GMRES

13c6544

Additionally, make the residual_norm parameter const in the restart kernel (as it was never written to).

Review updates

8fa5673

Fix GMRES reference test initialization and improve memory read efficiency of hessenberg_qr. Co-authored-by: Pratik Nayak <pratikvn@protonmail.com> Co-authored-by: Yuhsiang M. Tsai <yhmtsai@gmail.com>

Review update

2e70a8b

Add additional reference GMRES test, update parts of documentation Co-authored-by: Yuhsiang M. Tsai <yhmtsai@gmail.com>

thoasm force-pushed the gmres_simplify branch from 42c245d to 2e70a8b Compare November 3, 2022 19:29

thoasm merged commit 3902b86 into develop Nov 3, 2022

thoasm deleted the gmres_simplify branch November 3, 2022 19:32

MarcelKoch mentioned this pull request Nov 16, 2022

Enable distributed Gmres #1201

Merged

upsj mentioned this pull request Mar 22, 2023

fix the additional residual computation when passing residual criterion #1307

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Simplify GMRES kernels #861

Simplify GMRES kernels #861

upsj commented Aug 18, 2021 •

edited by thoasm

Loading

yhmtsai left a comment

yhmtsai Oct 31, 2022

yhmtsai Oct 31, 2022

thoasm Oct 31, 2022

sonarqubecloud bot commented Nov 1, 2022

thoasm commented Nov 2, 2022

tcojean commented Nov 2, 2022

yhmtsai left a comment •

edited

Loading

yhmtsai Nov 2, 2022

thoasm Nov 3, 2022

thoasm Nov 3, 2022

yhmtsai left a comment

thoasm commented Nov 3, 2022

ginkgo-bot commented Nov 3, 2022

ginkgo-bot commented Nov 3, 2022

yhmtsai left a comment

ginkgo-bot commented Nov 3, 2022

		.check(RelativeStoppingId, true, &stop_status, &one_changed)) {
		.check(RelativeStoppingId, false, &stop_status, &one_changed)) {

Simplify GMRES kernels #861

Simplify GMRES kernels #861

Conversation

upsj commented Aug 18, 2021 • edited by thoasm Loading

yhmtsai left a comment

Choose a reason for hiding this comment

yhmtsai Oct 31, 2022

Choose a reason for hiding this comment

yhmtsai Oct 31, 2022

Choose a reason for hiding this comment

thoasm Oct 31, 2022

Choose a reason for hiding this comment

sonarqubecloud bot commented Nov 1, 2022

thoasm commented Nov 2, 2022

tcojean commented Nov 2, 2022

yhmtsai left a comment • edited Loading

Choose a reason for hiding this comment

yhmtsai Nov 2, 2022

Choose a reason for hiding this comment

thoasm Nov 3, 2022

Choose a reason for hiding this comment

thoasm Nov 3, 2022

Choose a reason for hiding this comment

yhmtsai left a comment

Choose a reason for hiding this comment

thoasm commented Nov 3, 2022

ginkgo-bot commented Nov 3, 2022

ginkgo-bot commented Nov 3, 2022

yhmtsai left a comment

Choose a reason for hiding this comment

ginkgo-bot commented Nov 3, 2022

upsj commented Aug 18, 2021 •

edited by thoasm

Loading

yhmtsai left a comment •

edited

Loading