Past GPU Memory Limits with Unified Memory On Pascal > 자유게시판

• 목록
• 답변
• 글쓰기
• 게시판 리스트 옵션
  • 수정
  • 삭제

Modern pc architectures have a hierarchy of memories of varying measurement and performance. GPU architectures are approaching a terabyte per second memory bandwidth that, coupled with excessive-throughput computational cores, creates a great device for information-intensive duties. Nevertheless, everyone knows that quick memory is expensive. Modern purposes striving to solve bigger and bigger issues could be restricted by GPU memory capacity. Since the capacity of GPU memory is significantly decrease than system memory, it creates a barrier for developers accustomed to programming just one memory area. With the legacy GPU programming mannequin there is no such thing as a simple way to "just run" your software when you’re oversubscribing GPU memory. Even in case your dataset is just barely larger than the obtainable capacity, you would nonetheless must manage the energetic working set in GPU memory. Unified Memory is a way more intelligent memory administration system that simplifies GPU growth by offering a single memory house immediately accessible by all GPUs and Memory Wave CPUs within the system, with computerized web page migration for knowledge locality.

Migration of pages allows the accessing processor Memory Wave Workshop to learn from L2 caching and the lower latency of native memory. Furthermore, migrating pages to GPU memory ensures GPU kernels make the most of the very high bandwidth of GPU memory (e.g. 720 GB/s on a Tesla P100). And web page migration is all utterly invisible to the developer: the system automatically manages all information motion for Memory Wave you. Sounds nice, proper? With the Pascal GPU architecture Unified Memory is even more highly effective, thanks to Pascal’s larger digital memory deal with house and Page Migration Engine, enabling true virtual memory demand paging. It’s additionally worth noting that manually managing memory movement is error-prone, which affects productiveness and delays the day when you'll be able to finally run your entire code on the GPU to see those great speedups that others are bragging about. Builders can spend hours debugging their codes because of memory coherency issues. Unified memory brings big advantages for developer productiveness. In this post I will present you how Pascal can enable purposes to run out-of-the-box with bigger memory footprints and obtain great baseline performance.

For a second you can completely neglect about GPU memory limitations whereas developing your code. Unified Memory was launched in 2014 with CUDA 6 and the Kepler structure. This comparatively new programming mannequin allowed GPU purposes to use a single pointer in both CPU features and GPU kernels, which drastically simplified memory administration. CUDA eight and the Pascal architecture significantly improves Unified Memory performance by adding 49-bit digital addressing and on-demand page migration. The large 49-bit digital addresses are sufficient to enable GPUs to access your complete system memory plus the memory of all GPUs within the system. The Page Migration engine permits GPU threads to fault on non-resident memory accesses so the system can migrate pages from wherever within the system to the GPUs memory on-demand for environment friendly processing. In different phrases, Unified Memory transparently allows out-of-core computations for any code that's using Unified Memory for allocations (e.g. `cudaMallocManaged()`). It "just works" with none modifications to the application.

CUDA 8 also adds new methods to optimize knowledge locality by providing hints to the runtime so it remains to be doable to take full control over knowledge migrations. Nowadays it’s exhausting to discover a high-performance workstation with just one GPU. Two-, four- and eight-GPU techniques have gotten common in workstations as well as massive supercomputers. The NVIDIA DGX-1 is one example of a excessive-performance built-in system for deep studying with 8 Tesla P100 GPUs. For those who thought it was difficult to manually handle data between one CPU and one GPU, now you could have eight GPU memory spaces to juggle between. Unified Memory Wave Workshop is essential for such systems and it enables more seamless code improvement on multi-GPU nodes. Whenever a selected GPU touches data managed by Unified Memory, this data may migrate to native memory of the processor or the driver can establish a direct entry over the available interconnect (PCIe or NVLINK).