使用 accelerator 和 accelerator_view 对象

可以使用 acceleratoraccelerator_view 类指定要运行 C++ AMP 代码的设备或仿真器。 系统可能有多个设备或仿真器,它们在内存量、共享内存支持、调试支持或双精度支持方面有所不同。 C++ Accelerated Massive Parallelism (C++ AMP) 提供可用于检查可用加速器、将一个加速器设置为默认加速器、为多个 parallel_for_each 调用指定多个 accelerator_view 并执行特殊调试任务的 API。

注意

从 Visual Studio 2022 版本 17.0 开始,已弃用 C++ AMP 标头。 包含任何 AMP 标头都会导致生成错误。 应在包含任何 AMP 标头之前定义 _SILENCE_AMP_DEPRECATION_WARNINGS,以使警告静音。

使用默认加速器

除非编写代码以选取特定加速器,否则 C++ AMP 运行时会选取默认加速器。 运行时按如下所示选择默认加速器:

  1. 如果应用在调试模式下运行,则是支持调试的加速器。

  2. 否则是由 CPPAMP_DEFAULT_ACCELERATOR 环境变量(如果已设置)指定的加速器。

  3. 否则是非仿真设备。

  4. 否则是具有最大可用内存量的设备。

  5. 否则是未附加到显示的设备。

此外,运行时会为默认加速器将 access_type 指定为 access_type_auto。 这意味着,如果共享内存受支持并且其性能特征(带宽和延迟)已知与专用(非共享)内存相同,则默认加速器会使用共享内存。

可以通过构造默认加速器并检查其属性来确定默认加速器的属性。 下面的代码示例会打印默认加速器的路径、加速器内存量、共享内存支持、双精度支持以及有限双精度支持。

void default_properties() {
    accelerator default_acc;
    std::wcout << default_acc.device_path << "\n";
    std::wcout << default_acc.dedicated_memory << "\n";
    std::wcout << (accs[i].supports_cpu_shared_memory ?
        "CPU shared memory: true" : "CPU shared memory: false") << "\n";
    std::wcout << (accs[i].supports_double_precision ?
        "double precision: true" : "double precision: false") << "\n";
    std::wcout << (accs[i].supports_limited_double_precision ?
        "limited double precision: true" : "limited double precision: false") << "\n";
}

CPPAMP_DEFAULT_ACCELERATOR 环境变量

可以设置 CPPAMP_DEFAULT_ACCELERATOR 环境变量以指定默认加速器的 accelerator::device_path。 路径依赖于硬件。 以下代码使用 accelerator::get_all 函数检索可用加速器的列表,然后显示每个加速器的路径和特征。

void list_all_accelerators()
{
    std::vector<accelerator> accs = accelerator::get_all();

    for (int i = 0; i <accs.size(); i++) {
        std::wcout << accs[i].device_path << "\n";
        std::wcout << accs[i].dedicated_memory << "\n";
        std::wcout << (accs[i].supports_cpu_shared_memory ?
            "CPU shared memory: true" : "CPU shared memory: false") << "\n";
        std::wcout << (accs[i].supports_double_precision ?
            "double precision: true" : "double precision: false") << "\n";
        std::wcout << (accs[i].supports_limited_double_precision ?
            "limited double precision: true" : "limited double precision: false") << "\n";
    }
}

选择加速器

若要选择加速器,请使用 accelerator::get_all 方法检索可用加速器的列表,然后基于其属性选择一个加速器。 此示例演示如何选取内存最多的加速器:

void pick_with_most_memory()
{
    std::vector<accelerator> accs = accelerator::get_all();
    accelerator acc_chosen = accs[0];

    for (int i = 0; i <accs.size(); i++) {
        if (accs[i].dedicated_memory> acc_chosen.dedicated_memory) {
            acc_chosen = accs[i];
        }
    }

    std::wcout << "The accelerator with the most memory is "
        << acc_chosen.device_path << "\n"
        << acc_chosen.dedicated_memory << ".\n";
}

注意

accelerator::get_all 返回的加速器之一是 CPU 加速器。 无法在 CPU 加速器上执行代码。 若要筛选出 CPU 加速器,请将 accelerator::get_all 返回的加速器的 device_path 属性值与 accelerator::cpu_accelerator 的值进行比较。 有关详细信息,请参阅本文中的“特殊加速器”部分。

Shared Memory

共享内存是 CPU 和加速器都可以访问的内存。 使用共享内存可以消除或显著减少在 CPU 和加速器之间复制数据的开销。 尽管内存是共享的,但 CPU 和加速器不能同时访问它,因为这样做会导致未定义的行为。 如果加速器支持共享内存,则加速器属性 supports_cpu_shared_memory 返回 true,并且 default_cpu_access_type 属性会获取在 accelerator 上分配的内存的默认 access_type(例如,与在 accelerator 上访问的 acceleratorarray_view 对象关联的数组)。

C++ AMP 运行时会自动为每个 accelerator 选择最佳默认 access_type,但是在从 CPU 读取、从 CPU 写入或同时执行两种操作时,共享内存的性能特征(带宽和延迟)比专用(非共享)加速器内存的性能特征更差。 如果对于从 CPU 读取和写入,共享内存的性能与专用内存一样好,则运行时默认为 access_type_read_write;否则,运行时会选择更保守的默认值 access_type,并允许应用在计算内核的内存访问模式受益于其他 access_type 时替代它。

下面的代码示例演示如何确定默认加速器是否支持共享内存,然后替代其默认访问类型并从中创建 accelerator_view

#include <amp.h>
#include <iostream>

using namespace Concurrency;

int main()
{
    accelerator acc = accelerator(accelerator::default_accelerator);

    // Early out if the default accelerator doesn't support shared memory.
    if (!acc.supports_cpu_shared_memory)
    {
        std::cout << "The default accelerator does not support shared memory" << std::endl;
        return 1;
    }

    // Override the default CPU access type.
    acc.set_default_cpu_access_type(access_type_read_write);

    // Create an accelerator_view from the default accelerator. The
    // accelerator_view reflects the default_cpu_access_type of the
    // accelerator it's associated with.
    accelerator_view acc_v = acc.default_view;
}

accelerator_view 始终反映与其关联的 acceleratordefault_cpu_access_type,它不提供任何接口来替代或更改其 access_type

更改默认加速器

可以通过调用 accelerator::set_default 方法来更改默认加速器。 每次应用执行时只能更改默认加速器一次,并且必须先更改它,然后才能在 GPU 上执行任何代码。 用于更改加速器的任何后续函数调用会返回 false。 如果要在 parallel_for_each 调用中使用其他加速器,请阅读本文中的“使用多个加速器”部分。 下面的代码示例将默认加速器设置为未仿真、未连接到显示器且支持双精度的加速器。

bool pick_accelerator()
{
    std::vector<accelerator> accs = accelerator::get_all();
    accelerator chosen_one;

    auto result = std::find_if(accs.begin(), accs.end(),
        [] (const accelerator& acc) {
            return !acc.is_emulated &&
                acc.supports_double_precision &&
                !acc.has_display;
        });

    if (result != accs.end()) {
        chosen_one = *(result);
    }

    std::wcout <<chosen_one.description <<std::endl;
    bool success = accelerator::set_default(chosen_one.device_path);
    return success;
}

使用多个加速器

可通过两种方法在应用中使用多个加速器:

  • 可以将 accelerator_view 对象传递给对 parallel_for_each 方法的调用。

  • 可以使用特定 accelerator_view 对象构造数组对象。 C+AMP 运行时会从 lambda 表达式中捕获的数组对象中选取 accelerator_view 对象。

特殊加速器

三个特殊加速器的设备路径可用作 accelerator 类的属性:

互操作性

C++ AMP 运行时支持 accelerator_view 类与 Direct3D ID3D11Device 接口之间的互操作性。 create_accelerator_view 方法采用 IUnknown 接口,并返回 accelerator_view 对象。 get_device 方法采用 accelerator_view 对象,并返回 IUnknown 接口。

另请参阅

C++ AMP (C++ Accelerated Massive Parallelism)
调试 GPU 代码
accelerator_view 类