GPU-side data structure methods? Distributed Data Structures on GPUsPassing Objects into CUDA Kernels - Passing an object by value into a CUDA kernel results in a copy - Object likely destroyed before copied to GPU GPU Kernel Executed (Asynchronously) Destructor calledPassing Objects into CUDA Kernels Copy Constructor Invoked (on Host) New object trivially copied to GPU GPU Kernel Executed ... BCL::cuda::HashMap map(100); kernel<<<1, 100>>>(map);Passing Objects into CUDA Kernels Copy Constructor Invoked (on Host) New object trivially copied to GPU GPU Kernel Executed

kernel_a1<<<32, 256, 0, stream>>>(); 10 }); // synchronize 11 }); CUDA stream for offloading GPU kernels 32Taro’s Programming Model Taro: https://github.com/dian-lun-lin/taro A B Callback Wait Polling 256, 0, stream>>>(); 17 }); // suspend and multitask 18 }); CUDA stream for offloading GPU kernels 37Taro’s Programming Model – Example 1 Taro: https://github.com/dian-lun-lin/taro A B Callback store suspended tasks Low-priority queue (LPQ): store new tasks Worker 1 1. Offload GPU kernels in task A 2. Suspend task A 3. Go to sleep 62Taro’s Scheduler Taro: https://github.com/dian-lun-lin/taro

https://aka.ms/cpp/code Thu 10/28 – 2pm An Editor Can Do That? Debugging Assembly Language and GPU Kernels in Visual Studio Code Julia Reid – _3 Visual Studio CppCon 2020 Visual Studio 2019 Time Zones in MSVC – Miya Natsuhara • An Editor Can Do That? Debugging Assembly Language and GPU Kernels in Visual Studio Code – Julia Reid • Why does std::format do that? – Charlie Barto • Finding bugs

kernel We use C because kernels use C Baremetal embedded is really doing what the kernel does, so just be like the kernel right?Break free of the kernel We use C because kernels use C Baremetal embedded

can access the memory ○ No implied synchronization for simultaneous writes from two different kernels 56 Work Item Private Memory Work Item Private Memory Work Item Private Memory Work Item Very close to regular C++ programming ● Accessors ○ Implicitly builds data dependency DAG between kernels 69Device Copyable 70Device Copyable ● How can we copy objects between a host or a device and another

Time Zones in MSVC – Miya Natsuhara • An Editor Can Do That? Debugging Assembly Language and GPU Kernels in Visual Studio Code – Julia Reid • Why does std::format do that? – Charlie Barto • Finding bugs

Time Zones in MSVC – Miya Natsuhara • An Editor Can Do That? Debugging Assembly Language and GPU Kernels in Visual Studio Code – Julia Reid • Why does std::format do that? – Charlie Barto • Finding bugs

Clang-Tidy, makefile, CMake, GitHub and More CppCon 2021 - Debugging Assembly Language and GPU Kernels in Visual Studio Code CppCon 2020 - Collaborative C++ Development with Visual Studio CodePopular

Time Zones in MSVC – Miya Natsuhara • An Editor Can Do That? Debugging Assembly Language and GPU Kernels in Visual Studio Code – Julia Reid • Why does std::format do that? – Charlie Barto • Finding bugs

和 gridDim ，看起来非常方便。 本方法出自英伟达官方博客： https://developer.nvidia.com/blog/cuda-pro-tip-write-flexible-kernels-grid-stride-loops/ 第 4 章： C++ 封装 std::vector 的秘密：第二模板参数 • 你知道吗？ std::vector 作为模板类，其实有两个模板参数： std::vector

0 码力 | 142 页 | 13.52 MB | 1 年前

3