https://github.com/kokkos/mdspangithub.com/griswaldbrooks/spanny2 goals ● deeper understanding of std::mdspan layouts and accessors ● how to write custom layouts and accessors ● dispel common misconceptions memory accessor ● improving memory access using asynchronicity 4 ● motivations for std::mdspan ● review std::mdspan declarationgithub.com/griswaldbrooks/spanny2 how ● layouts and their requirements memory accessor ● improving memory access using asynchronicity 5 ● motivations for std::mdspan ● review std::mdspan declarationgithub.com/griswaldbrooks/spanny2 how ● layouts and their requirements

Is std::mdspan a Zero-overhead Abstraction? Oleksandr Bacherikov Snap IncWhat is std::mdspan? It’s a view over a multi-dimensional array. It’s designed primarily to be used as a function parameter to C++ If the matrices are contiguous in memory, then we can treat them as ranges and simply use std::transform {-5, 5, -3, 1, 4, 0} + { 3, 2, 1, 3, -5, 4} = {-2, 7, -2, 4, -1, 4} 4*r8], xmm0 15Vectorization is great for performance, not great for slideware.clang16.0.4 -O1 -std=c++20 push rbx test rdi, rdi je .LBB0_6 mov rax, qword ptr [rsp + 24]

Limitations and Problems in std::function and similar constructs Amandeep Chawla | Sr. Computer Scientist II amandeep@adobe.com | adchawla@gmail.com Image by Bruno Tornielli© 2024 Adobe. All Rights whose type we can control. ▪ std::function© 2024 Adobe. All Rights Reserved. Adobe Confidential. Task Based Mechanism© 2024 Adobe. All Rights Reserved. Adobe Confidential. std::function ▪ Is a class template template ▪ It is a general-purpose polymorphic function wrapper ▪ Instances of std::function can store, copy, and invoke any CopyConstructible Callable target ▪ Uses type-erasure under the hood to gain

elements std::sort, std::stable_sort, ranges::sort, etc std::sort(begin, end); std::ranges::sort(cont); 4REMINDERS Sorting is the ordering of elements std::sort, std::stable_sort std::stable_sort, ranges::sort, etc std::sort(begin, end, comp); std::ranges::sort(cont, comp); 5QUICKSORT 6QUICKSORT Quick sort 6.1QUICKSORT Quick sort Take any element 6.2QUICKSORT __quick_sort_loop(first, last); 4 } 5 9Quick sort (STL version) 10STANDARD LIBRARY EVOLVED FROM STL 11STD::SORT 12STD::SORT 12.1Quicksort's worst case is Just pick the worst element 13GCC libstdc++ template

0 码力 | 182 页 | 7.65 MB | 6 月前

3

Center (DLR) Institute of Software Methods for Product Virtualization VECTORIZING A CFD CODE WITH STD::SIMD SUPPLEMENTED BY (ALMOST) TRANSPARENT LOADING AND STORINGMotivation: The Origin of the Talk deduction to load and store std::simd and scalar variables ▪ syntactically equalize scalar and vectorized code The talk: ▪ share experience with vectorization using std::simd ▪ introduce the SIMD_ACCESS once → Single Instruction Multiple Data (SIMD) For more details Matthias Kretz‘ Cppcon talk about std::simd: https://youtu.be/LAJ_hywLtMA 4 Olaf Krzikalla, DLR SP, 2024-09-17 𝑥1 𝑥2 𝑥3 𝑥4 + −∗

Can’t we just synthesize std::tuple_element from get? Jonathan Müller — @foonathanAdding structured bindings support to a custom tuple-like type template struct my_tuple { … }; Jonathan just synthesize std::tuple_element from get? CppCon 2024-09-18 1Adding structured bindings support to a custom tuple-like type template struct my_tuple { … }; template <std::size_t I, typename Tuple> requires instance_ofstd::remove_cvref_t> auto get(Tuple&& t) -> decltype(auto) { … } Jonathan Müller — @foonathan Can’t we just synthesize std::tuple_element from get? CppCon 2024-09-18

Extents Going beyond Conclusion A standardization effort On mdspan P0332: Relaxed Incomplete Multidimensional Array Type Declaration P0009: mdspan: A Non-Owning Multidimensional Array Reference P1684: mdarray: mdarray: An Owning Multidimensional Array Analog of mdspan On linear algebra in general P1417: Historical lessons for C++ linear algebra library standardization P1166: What do we need from a linear Standardization Design EDSL Extents Going beyond Conclusion The mdspan proposal: synopsis Proposed mdspan synopsis as of P0009R10 1 namespace std { 2 // [mdspan.extents], class template extents 3 template

0 码力 | 62 页 | 1.38 MB | 6 月前

3

12 std::linalg: Linear Algebra Coming to Standard C++ Mark Hoemmen, NVIDIA | CppCon 20233 Agenda • Motivating example: Matrix multiply • Where std::linalg fits in linear algebra’s layers • std::linalg std::linalg builds on the C++ Standard Library • std::linalg builds on the long history of the BLAS • Detailed example: Cholesky matrix factorization4 Motivating example: Updating matrix multiply 𝐶 := B.data_handle(), &LDB, A.data_handle(), &LDA, &beta, C.data_handle(), &LDC); // AFTER: Use std::linalg #include matrix_product(par_unseq, scaled(alpha, A), transposed(B), scaled(beta

(C++ templates, parallel STL, inlining and fusion, abstractions) Math, ML, Data Libraries; C++ Std, C, Python Libraries Libraries augment compiler optimizations for Performance Portable programs futures 21Moving with the Times 22#include int main(int argc, char *argv[]) { std::vector dA{ … }, dB{ … }, dO{ … }; try { sycl::queue gpuQueue{sycl::gpu_selector_v file has changed from CL/sycl.hpp #include int main(int argc, char *argv[]) { std::vector dA{ … }, dB{ … }, dO{ … }; try { sycl::queue gpuQueue{sycl::gpu_selector_v

CppCon 2024 7• mdspan (P0009), C++23 • A mutating multidimensional array reference (view)… • …of a container whose elements reside in contiguous memory ➢ std::vector ➢ Eigen::VectorXd ➢ std::mdarray (p1684) C++26 • A free function linear algebra interface based on the BLAS • Matrix represented by a 2-D mdspan • Default BLAS with major compilers/library vendors A Little History: C++23 and C++26 Daniel 2024 11• The Eigen::Matrix class supports the (usual) numerical types: • int • float • double • std::complex • Various aliases of the Matrix class • Fixed square dimensions (up to 4) ➢ Eigen::Matrix4i: