Add distributed benchmarks for SpMV, Solver, Multi-Vector #1204

MarcelKoch · 2022-11-18T16:38:43Z

This PR adds distributed benchmarks for SpMV, Solver, and Multi-Vector operations.

To reduce code duplication, the non-distributed versions of these benchmarks have been refactored. In each case, a common core has been extracted. Additionally, creating vectors/matrices has been delegated to helper types for the sequential and distributed case. The common code uses templating with duck-typing to be generalizable across sequential and distributed benchmarks.

Also, this enables generated stencil matrices for all benchmarks.

The distributed multi-vector benchmark does not support BLAS level 2 or level 3 operations. The non-distributed still supports the same operations as before.

The distributed solver benchmark supports only the currently available distributed solvers.

Both in the distributed solver and spmv benchmark, the local and non-local format can be specified via the CLI.

upsj

First pass - a few small issues, but otherwise really nice work!

Are there any notable changes in the operations I should be aware of (except for supporting the generators)? Reviewing moved code is always a bit of a challenge.

benchmark/blas/blas.cpp

benchmark/utils/loggers.hpp

benchmark/utils/distributed_helpers.hpp

benchmark/utils/generator.hpp

MarcelKoch · 2022-11-23T15:31:04Z

@upsj I don't think I've changed anything significant in the moved code, except for using the generator to create matrices and vectors.

upsj

I think there might be an issue with the boundary nodes in the stencil matrix setup, otherwise no major concerns.

benchmark/blas/blas_common.hpp

benchmark/utils/generator.hpp

benchmark/utils/stencil_matrix.hpp

sonarcloud · 2022-12-19T16:25:12Z

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities
1 Security Hotspot
98 Code Smells

0.0% Coverage
6.4% Duplication

pratikvn

I really like the generalization of the matrix and vector generate functions. LGTM!

pratikvn · 2023-01-26T10:59:17Z

benchmark/utils/generator.hpp

+ }
+ }
+
+ static std::shared_ptr<gko::LinOp> generate_matrix(


I think being explicit doesnt hurt.

Suggested change

static std::shared_ptr<gko::LinOp> generate_matrix(

static std::shared_ptr<gko::LinOp> generate_matrix_optimal_format(

benchmark/utils/stencil_matrix.hpp

pratikvn · 2023-01-26T12:11:38Z

benchmark/utils/generator.hpp

+ }
+ }
+
+ std::shared_ptr<gko::LinOp> generate_matrix(


Suggested change

std::shared_ptr<gko::LinOp> generate_matrix(

std::shared_ptr<gko::LinOp> generate_matrix_with_optimal_format(

pratikvn · 2023-01-26T12:12:51Z

benchmark/utils/generator.hpp

+ auto part = gko::experimental::distributed::
+ Partition<itype, global_itype>::build_from_global_size_uniform(
+ exec, comm.size(), static_cast<global_itype>(data.size[0]));


Maybe this Partition should be a parameter as well, allowing users to benchmark different kinds of partitions?

I think instead of a function parameter, a better way to customize the partition would be to derive from the class, and override this function. I'm not sure how else the partition can be passed along.
So I will leave it in its current state.

pratikvn · 2023-01-26T12:13:32Z

benchmark/utils/generator.hpp

+ auto part = gko::experimental::distributed::
+ Partition<itype, global_itype>::build_from_global_size_uniform(
+ exec, comm.size(), static_cast<global_itype>(size[0]));


- remove dead code - rename strided vector creation function - merge logged/unlogged matrix creation Co-authored-by: Tobias Ribizel <[email protected]>

- renaming - documentation Co-authored-by: Tobias Ribizel <[email protected]>

- renaming - refactor stencil neighbor check Co-authored-by: Tobias Ribizel <[email protected]>

it can be generated from the already exisiting functions by using a 1D domain decomposition

- renaming - consistent ordering of slow to fast indices Co-authored-by: Pratik Nayak <[email protected]>

MarcelKoch · 2023-01-27T09:32:47Z

format!

Co-authored-by: Marcel Koch <[email protected]>

sonarcloud · 2023-01-30T14:35:56Z

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities
1 Security Hotspot
98 Code Smells

0.0% Coverage
6.4% Duplication

codecov · 2023-01-30T14:40:56Z

Codecov Report

Base: 91.49% // Head: 91.49% // Increases project coverage by +0.00% 🎉

Coverage data is based on head (613acb3) compared to base (d53acb2).
Patch has no changes to coverable lines.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #1204   +/-   ##
========================================
  Coverage    91.49%   91.49%           
========================================
  Files          556      556           
  Lines        47465    47467    +2     
========================================
+ Hits         43428    43432    +4     
+ Misses        4037     4035    -2

Impacted Files	Coverage Δ
include/ginkgo/core/base/mpi.hpp	`89.63% <0.00%> (-1.65%)`	⬇️
reference/base/index_set_kernels.cpp	`94.11% <0.00%> (-0.09%)`	⬇️
omp/reorder/rcm_kernels.cpp	`98.13% <0.00%> (+3.69%)`	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

Release 1.6.0 of Ginkgo. The Ginkgo team is proud to announce the new Ginkgo minor release 1.6.0. This release brings new features such as: - Several building blocks for GPU-resident sparse direct solvers like symbolic and numerical LU and Cholesky factorization, ..., - A distributed Schwarz preconditioner, - New FGMRES and GCR solvers, - Distributed benchmarks for the SpMV operation, solvers, ... - Support for non-default streams in the CUDA and HIP backends, - Mixed precision support for the CSR SpMV, - A new profiling logger which integrates with NVTX, ROCTX, TAU and VTune to provide internal Ginkgo knowledge to most HPC profilers! 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 Clang: 14.0 is tested. Earlier versions might also work. + NVHPC: 22.7+ + Cray Compiler: 14.0.1+ + CUDA module: CUDA 9.2+ or NVHPC 22.7+ + HIP module: ROCm 4.5+ + DPC++ module: Intel OneAPI 2021.3+ with oneMKL and oneDPL. Set the CXX compiler to `dpcpp`. + Windows + MinGW: GCC 5.5+ + Microsoft Visual Studio: VS 2019+ + CUDA module: CUDA 9.2+, Microsoft Visual Studio + OpenMP module: MinGW. ### Version Support Changes + ROCm 4.0+ -> 4.5+ after [#1303](#1303) + Removed Cygwin pipeline and support [#1283](#1283) ### Interface Changes + Due to internal changes, `ConcreteExecutor::run` will now always throw if the corresponding module for the `ConcreteExecutor` is not build [#1234](#1234) + The constructor of `experimental::distributed::Vector` was changed to only accept local vectors as `std::unique_ptr` [#1284](#1284) + The default parameters for the `solver::MultiGrid` were improved. In particular, the smoother defaults to one iteration of `Ir` with `Jacobi` preconditioner, and the coarse grid solver uses the new direct solver with LU factorization. [#1291](#1291) [#1327](#1327) + The `iteration_complete` event gained a more expressive overload with additional parameters, the old overloads were deprecated. [#1288](#1288) [#1327](#1327) ### Deprecations + Deprecated less expressive `iteration_complete` event. Users are advised to now implement the function `void iteration_complete(const LinOp* solver, const LinOp* b, const LinOp* x, const size_type& it, const LinOp* r, const LinOp* tau, const LinOp* implicit_tau_sq, const array<stopping_status>* status, bool stopped)` [#1288](#1288) ### Added Features + A distributed Schwarz preconditioner. [#1248](#1248) + A GCR solver [#1239](#1239) + Flexible Gmres solver [#1244](#1244) + Enable Gmres solver for distributed matrices and vectors [#1201](#1201) + An example that uses Kokkos to assemble the system matrix [#1216](#1216) + A symbolic LU factorization allowing the `gko::experimental::factorization::Lu` and `gko::experimental::solver::Direct` classes to be used for matrices with non-symmetric sparsity pattern [#1210](#1210) + A numerical Cholesky factorization [#1215](#1215) + Symbolic factorizations in host-side operations are now wrapped in a host-side `Operation` to make their execution visible to loggers. This means that profiling loggers and benchmarks are no longer missing a separate entry for their runtime [#1232](#1232) + Symbolic factorization benchmark [#1302](#1302) + The `ProfilerHook` logger allows annotating the Ginkgo execution (apply, operations, ...) for profiling frameworks like NVTX, ROCTX and TAU. [#1055](#1055) + `ProfilerHook::created_(nested_)summary` allows the generation of a lightweight runtime profile over all Ginkgo functions written to a user-defined stream [#1270](#1270) for both host and device timing functionality [#1313](#1313) + It is now possible to enable host buffers for MPI communications at runtime even if the compile option `GINKGO_FORCE_GPU_AWARE_MPI` is set. [#1228](#1228) + A stencil matrices generator (5-pt, 7-pt, 9-pt, and 27-pt) for benchmarks [#1204](#1204) + Distributed benchmarks (multi-vector blas, SpMV, solver) [#1204](#1204) + Benchmarks for CSR sorting and lookup [#1219](#1219) + A timer for MPI benchmarks that reports the longest time [#1217](#1217) + A `timer_method=min|max|average|median` flag for benchmark timing summary [#1294](#1294) + Support for non-default streams in CUDA and HIP executors [#1236](#1236) + METIS integration for nested dissection reordering [#1296](#1296) + SuiteSparse AMD integration for fillin-reducing reordering [#1328](#1328) + Csr mixed-precision SpMV support [#1319](#1319) + A `with_loggers` function for all `Factory` parameters [#1337](#1337) ### Improvements + Improve naming of kernel operations for loggers [#1277](#1277) + Annotate solver iterations in `ProfilerHook` [#1290](#1290) + Allow using the profiler hooks and inline input strings in benchmarks [#1342](#1342) + Allow passing smart pointers in place of raw pointers to most matrix functions. This means that things like `vec->compute_norm2(x.get())` or `vec->compute_norm2(lend(x))` can be simplified to `vec->compute_norm2(x)` [#1279](#1279) [#1261](#1261) + Catch overflows in prefix sum operations, which makes Ginkgo's operations much less likely to crash. This also improves the performance of the prefix sum kernel [#1303](#1303) + Make the installed GinkgoConfig.cmake file relocatable and follow more best practices [#1325](#1325) ### Fixes + Fix OpenMPI version check [#1200](#1200) + Fix the mpi cxx type binding by c binding [#1306](#1306) + Fix runtime failures for one-sided MPI wrapper functions observed on some OpenMPI versions [#1249](#1249) + Disable thread pinning with GPU executors due to poor performance [#1230](#1230) + Fix hwloc version detection [#1266](#1266) + Fix PAPI detection in non-implicit include directories [#1268](#1268) + Fix PAPI support for newer PAPI versions: [#1321](#1321) + Fix pkg-config file generation for library paths outside prefix [#1271](#1271) + Fix various build failures with ROCm 5.4, CUDA 12, and OneAPI 6 [#1214](#1214), [#1235](#1235), [#1251](#1251) + Fix incorrect read for skew-symmetric MatrixMarket files with explicit diagonal entries [#1272](#1272) + Fix handling of missing diagonal entries in symbolic factorizations [#1263](#1263) + Fix segmentation fault in benchmark matrix construction [#1299](#1299) + Fix the stencil matrix creation for benchmarking [#1305](#1305) + Fix the additional residual check in IR [#1307](#1307) + Fix the cuSPARSE CSR SpMM issue on single strided vector when cuda >= 11.6 [#1322](#1322) [#1331](#1331) + Fix Isai generation for large sparsity powers [#1327](#1327) + Fix Ginkgo compilation and test with NVHPC >= 22.7 [#1331](#1331) + Fix Ginkgo compilation of 32 bit binaries with MSVC [#1349](#1349)

MarcelKoch added the 1:ST:need-feedback The PR is somewhat ready but feedback on a blocking topic is required before a proper review. label Nov 18, 2022

MarcelKoch self-assigned this Nov 18, 2022

ginkgo-bot added reg:benchmarking This is related to benchmarking. reg:build This is related to the build system. type:preconditioner This is related to the preconditioners type:solver This is related to the solvers labels Nov 18, 2022

MarcelKoch requested review from a team November 18, 2022 16:39

upsj reviewed Nov 22, 2022

View reviewed changes

upsj requested changes Nov 23, 2022

View reviewed changes

MarcelKoch added 1:ST:ready-for-review This PR is 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 Nov 28, 2022

MarcelKoch requested review from a team November 28, 2022 14:00

MarcelKoch force-pushed the distributed-benchmark branch from d70f984 to 0ec33af Compare November 29, 2022 16:07

upsj approved these changes Nov 30, 2022

View reviewed changes

benchmark/utils/stencil_matrix.hpp Outdated Show resolved Hide resolved

benchmark/utils/stencil_matrix.hpp Outdated Show resolved Hide resolved

benchmark/utils/stencil_matrix.hpp Outdated Show resolved Hide resolved

MarcelKoch requested a review from a team December 7, 2022 08:54

upsj requested a review from a team December 8, 2022 21:44

MarcelKoch force-pushed the distributed-benchmark branch from 633f35c to 512fc17 Compare December 19, 2022 08:46

pratikvn approved these changes Jan 26, 2023

View reviewed changes

MarcelKoch added 5 commits January 27, 2023 10:31

add distributed spmv scaling benchmark

8f2b9dd

initializes executor in solver benchmark correctly

3ac2a09

extracts stencil matrix generation into separate file

659863b

introduces intermediate matrix_type_factory map

51adeef

simplifies stencil generation call

46b711c

MarcelKoch and others added 18 commits January 27, 2023 10:31

refactors blas and adds distributed blas

241e11a

removes old distributed benchmarks

c11ec4c

restricts printing to mpi rank 0

2cd198c

fixup namespaces after rebase

dd3187d

extracts input broadcasting

b3aaa68

removes some code duplication

d68472c

fixup rebase issues

8aafd82

throws if can't generate matrix data

acee19b

interprets input size as global size

a5da60a

fixes stencil generation

04729f1

fixes missing include

530acab

review updates:

40c2cf5

- remove dead code - rename strided vector creation function - merge logged/unlogged matrix creation Co-authored-by: Tobias Ribizel <[email protected]>

fixes distributed stencil matrix generation

67ca237

review updates:

ae25fd0

- renaming - documentation Co-authored-by: Tobias Ribizel <[email protected]>

fixes return type in lambda

53e7506

review updates:

a089931

- renaming - refactor stencil neighbor check Co-authored-by: Tobias Ribizel <[email protected]>

removes special case to generate matrix with optimal comm

05e3d24

it can be generated from the already exisiting functions by using a 1D domain decomposition

review updates:

ffd82c6

- renaming - consistent ordering of slow to fast indices Co-authored-by: Pratik Nayak <[email protected]>

MarcelKoch force-pushed the distributed-benchmark branch from 7c8edae to ffd82c6 Compare January 27, 2023 09:32

Format files

613acb3

Co-authored-by: Marcel Koch <[email protected]>

MarcelKoch 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 Jan 30, 2023

MarcelKoch merged commit 3f56698 into develop Jan 30, 2023

MarcelKoch deleted the distributed-benchmark branch January 30, 2023 14:43

MarcelKoch mentioned this pull request Jul 7, 2023

Scaling example #894

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add distributed benchmarks for SpMV, Solver, Multi-Vector #1204

Add distributed benchmarks for SpMV, Solver, Multi-Vector #1204

MarcelKoch commented Nov 18, 2022

upsj left a comment

MarcelKoch commented Nov 23, 2022

upsj left a comment

sonarcloud bot commented Dec 19, 2022

pratikvn left a comment

pratikvn Jan 26, 2023

pratikvn Jan 26, 2023

pratikvn Jan 26, 2023

MarcelKoch Jan 26, 2023

pratikvn Jan 26, 2023

MarcelKoch commented Jan 27, 2023

sonarcloud bot commented Jan 30, 2023

codecov bot commented Jan 30, 2023

	static std::shared_ptr<gko::LinOp> generate_matrix(
	static std::shared_ptr<gko::LinOp> generate_matrix_optimal_format(

	std::shared_ptr<gko::LinOp> generate_matrix(
	std::shared_ptr<gko::LinOp> generate_matrix_with_optimal_format(

Add distributed benchmarks for SpMV, Solver, Multi-Vector #1204

Add distributed benchmarks for SpMV, Solver, Multi-Vector #1204

Conversation

MarcelKoch commented Nov 18, 2022

upsj left a comment

Choose a reason for hiding this comment

MarcelKoch commented Nov 23, 2022

upsj left a comment

Choose a reason for hiding this comment

sonarcloud bot commented Dec 19, 2022

pratikvn left a comment

Choose a reason for hiding this comment

pratikvn Jan 26, 2023

Choose a reason for hiding this comment

pratikvn Jan 26, 2023

Choose a reason for hiding this comment

pratikvn Jan 26, 2023

Choose a reason for hiding this comment

MarcelKoch Jan 26, 2023

Choose a reason for hiding this comment

pratikvn Jan 26, 2023

Choose a reason for hiding this comment

MarcelKoch commented Jan 27, 2023

sonarcloud bot commented Jan 30, 2023

codecov bot commented Jan 30, 2023

Codecov Report