没有合适的资源？快使用搜索试试~ 我知道了~

首页CUDA编程指南：C/C++接口与6.0/6.1/6.2更新

CUDA编程指南：C/C++接口与6.0/6.1/6.2更新

CUDA

需积分: 13 5 下载量 163 浏览量更新于2024-07-20 收藏 4.5MB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

试读

275页

《CUDA编程指南》(CUDA C Programming Guide)是一份由NVIDIA发布的文档，版本为PG-02829-001_v8.0，发布日期为2016年9月。这份设计指南主要针对CUDA技术，这是一种专门为NVIDIA GPU设计的并行计算平台和编程模型，旨在加速科学计算、工程分析和机器学习等领域的高性能计算任务。本指南的核心内容包括CUDA的最新发展，特别强调了对CUDA Compute Capability 6.0, 6.1, 和 6.2的支持。这些新版本更新了表12，增加了对64位浮点原子操作的支持，并在Table 13中添加了相应设备的列。对于6.x能力等级，还引入了对`atomicAdd()`函数的64位浮点支持。指南分为五个主要章节： 1. **介绍**：阐述了CUDA从图形处理单元（GPU）向通用目的并行计算的转变，强调了CUDA作为一个可扩展的编程模型，以及文档结构的设计。 2. **编程模型**：深入探讨了CUDA的编程模型，涉及核心概念如内核（Kernels）、线程层次结构、内存层次划分，以及如何实现异构编程，同时介绍了每个版本的兼容性——Compute Capability。 3. **编程接口**：讲解了如何使用NVCC编译器进行CUDA代码的编译流程，包括编译步骤和关键概念。 4. **CUDA编程实践**：这部分可能包含实际编程示例，展示如何创建和管理线程，以及如何利用CUDA的内存管理和同步机制。 5. **统一内存编程**：随着CUDA的发展，统一内存（Unified Memory）成为一项关键特性，这一章可能涵盖了如何优化数据访问和内存分配，以充分利用GPU和CPU之间的共享内存空间。《CUDA编程指南》作为开发者的重要参考资料，提供了CUDA语言的细节，以及如何将其应用于各种并行计算场景。无论是初学者还是经验丰富的开发者，都可以通过这份文档深入了解CUDA编程的各个方面，从而提升GPU性能和应用程序的效率。

资源详情

资源推荐

Introduction

www.nvidia.com

CUDA C Programming Guide PG-02829-001_v8.0|2

Figure2 Memory Bandwidth for the CPU and GPU

The reason behind the discrepancy in floating-point capability between the CPU and the

GPU is that the GPU is specialized for compute-intensive, highly parallel computation

- exactly what graphics rendering is about - and therefore designed such that more

transistors are devoted to data processing rather than data caching and flow control, as

schematically illustrated by Figure 3.

Cache

ALU

Cont rol

ALU

DRAM

CPU

DRAM

GPU

Figure3 The GPU Devotes More Transistors to Data Processing

More specifically, the GPU is especially well-suited to address problems that can be

expressed as data-parallel computations - the same program is executed on many data

elements in parallel - with high arithmetic intensity - the ratio of arithmetic operations

to memory operations. Because the same program is executed for each data element,

Introduction

www.nvidia.com

CUDA C Programming Guide PG-02829-001_v8.0|3

there is a lower requirement for sophisticated flow control, and because it is executed on

many data elements and has high arithmetic intensity, the memory access latency can be

hidden with calculations instead of big data caches.

Data-parallel processing maps data elements to parallel processing threads. Many

applications that process large data sets can use a data-parallel programming model

to speed up the computations. In 3D rendering, large sets of pixels and vertices are

mapped to parallel threads. Similarly, image and media processing applications such as

post-processing of rendered images, video encoding and decoding, image scaling, stereo

vision, and pattern recognition can map image blocks and pixels to parallel processing

threads. In fact, many algorithms outside the field of image rendering and processing

are accelerated by data-parallel processing, from general signal processing or physics

simulation to computational finance or computational biology.

1.2.CUDA

: A General-Purpose Parallel Computing

Platform and Programming Model

In November 2006, NVIDIA introduced CUDA

, a general purpose parallel computing

platform and programming model that leverages the parallel compute engine in

NVIDIA GPUs to solve many complex computational problems in a more efficient way

than on a CPU.

CUDA comes with a software environment that allows developers to use C as a high-

level programming language. As illustrated by Figure 4, other languages, application

programming interfaces, or directives-based approaches are supported, such as

FORTRAN, DirectCompute, OpenACC.

Introduction

www.nvidia.com

CUDA C Programming Guide PG-02829-001_v8.0|4

Figure4 GPU Computing Applications

CUDA is designed to support various languages and application programming interfaces.

1.3.A Scalable Programming Model

The advent of multicore CPUs and manycore GPUs means that mainstream processor

chips are now parallel systems. Furthermore, their parallelism continues to scale

with Moore's law. The challenge is to develop application software that transparently

scales its parallelism to leverage the increasing number of processor cores, much as

3D graphics applications transparently scale their parallelism to manycore GPUs with

widely varying numbers of cores.

The CUDA parallel programming model is designed to overcome this challenge while

maintaining a low learning curve for programmers familiar with standard programming

languages such as C.

At its core are three key abstractions - a hierarchy of thread groups, shared memories,

and barrier synchronization - that are simply exposed to the programmer as a minimal

set of language extensions.

Introduction

www.nvidia.com

CUDA C Programming Guide PG-02829-001_v8.0|5

These abstractions provide fine-grained data parallelism and thread parallelism,

nested within coarse-grained data parallelism and task parallelism. They guide the

programmer to partition the problem into coarse sub-problems that can be solved

independently in parallel by blocks of threads, and each sub-problem into finer pieces

that can be solved cooperatively in parallel by all threads within the block.

This decomposition preserves language expressivity by allowing threads to cooperate

when solving each sub-problem, and at the same time enables automatic scalability.

Indeed, each block of threads can be scheduled on any of the available multiprocessors

within a GPU, in any order, concurrently or sequentially, so that a compiled CUDA

program can execute on any number of multiprocessors as illustrated by Figure 5, and

only the runtime system needs to know the physical multiprocessor count.

This scalable programming model allows the GPU architecture to span a wide market

range by simply scaling the number of multiprocessors and memory partitions: from

the high-performance enthusiast GeForce GPUs and professional Quadro and Tesla

computing products to a variety of inexpensive, mainstream GeForce GPUs (see CUDA-

Enabled GPUs for a list of all CUDA-enabled GPUs).

Introduction

www.nvidia.com

CUDA C Programming Guide PG-02829-001_v8.0|6

GPU w ith 2 SMs

SM 1

SM 0

GPU w ith 4 SMs

SM 1

SM 0

SM 3

SM 2

Block

Mult ithreaded CUDA Program

Block 0

Block

Block 3

Block 4

Block

Block 7

Block

Block 0

Block

Block 6

Block

Block 4

Block

Block 7

A GPU is built around an array of Streaming Multiprocessors (SMs) (see Hardware

Implementation for more details). A multithreaded program is partitioned into blocks

of threads that execute independently from each other, so that a GPU with more

multiprocessors will automatically execute the program in less time than a GPU with

fewer multiprocessors.

Figure5 Automatic Scalability

1.4.Document Structure

This document is organized into the following chapters:

‣

Chapter Introduction is a general introduction to CUDA.

‣

Chapter Programming Model outlines the CUDA programming model.

‣

Chapter Programming Interface describes the programming interface.

‣

Chapter Hardware Implementation describes the hardware implementation.

‣

Chapter Performance Guidelines gives some guidance on how to achieve maximum

performance.

‣

Appendix CUDA-Enabled GPUs lists all CUDA-enabled devices.

‣

Appendix C Language Extensions is a detailed description of all extensions to the C

language.

剩余274页未读，继续阅读

解释：if(CUDA_FOUND) message(STATUS "Found CUDA Toolkit v${CUDA_VERSION_STRING}") enable_language(CUDA) set(HAVE_CUDA TRUE) if (CMAKE_CUDA_COMPILER_ID STREQUAL "NVIDIA") if(${CUDA_VERSION_STRING} VERSION_GREATER_EQUAL "11.1") execute_process(COMMAND ${CMAKE_CUDA_COMPILER} --list-gpu-code RESULT_VARIABLE EXIT_CODE OUTPUT_VARIABLE OUTPUT_VAL) if(EXIT_CODE EQUAL 0) #Remove sm_ string(REPLACE "sm_" "" OUTPUT_VAL ${OUTPUT_VAL}) #Convert to list string(REPLACE "\n" ";" __CUDA_ARCH_BIN ${OUTPUT_VAL}) #Remove last empty entry list(REMOVE_AT __CUDA_ARCH_BIN -1) else() message(FATAL_ERROR "Failed to run NVCC to get list of GPU codes: ${EXIT_CODE}") endif() elseif(${CUDA_VERSION_STRING} VERSION_GREATER_EQUAL "11.0") set(__CUDA_ARCH_BIN "35;37;50;52;53;60;61;62;70;72;75;80") elseif(${CUDA_VERSION_STRING} VERSION_GREATER_EQUAL "10.0") set(__CUDA_ARCH_BIN "30;32;35;37;50;52;53;60;61;62;70;72;75") elseif(${CUDA_VERSION_STRING} VERSION_GREATER_EQUAL "9.1") set(__CUDA_ARCH_BIN "30;32;35;37;50;52;53;60;61;62;70;72") else() set(__CUDA_ARCH_BIN "30;32;35;37;50;52;53;60;61;62;70") endif() else() message(FATAL_ERROR "Unsupported CUDA compiler ${CMAKE_CUDA_COMPILER_ID}.") endif() set(CUDA_ARCH_BIN ${__CUDA_ARCH_BIN} CACHE STRING "Specify 'real' GPU architectures to build binaries for") if(POLICY CMP0104) cmake_policy(SET CMP0104 NEW) set(CMAKE_CUDA_ARCHITECTURES ${CUDA_ARCH_BIN}) message(STATUS "CMAKE_CUDA_ARCHITECTURES: ${CMAKE_CUDA_ARCHITECTURES}") #Add empty project as its not required with newer CMake add_library(pcl_cuda INTERFACE) else() # Generate SASS set(CMAKE_CUDA_ARCHITECTURES ${CUDA_ARCH_BIN}) # Generate PTX for last architecture list(GET CUDA_ARCH_BIN -1 ver) set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode arch=compute_${ver},code=compute_${ver}") message(STATUS "CMAKE_CUDA_FLAGS: ${CMAKE_CUDA_FLAGS}") add_library(pcl_cuda INTERFACE) target_include_directories(pcl_cuda INTERFACE ${CUDA_TOOLKIT_INCLUDE}) endif () endif()

这段代码是CMake脚本，主要用于检测CUDA是否安装，并设置CUDA相关的编译器标志和选项。首先，它检查CUDA是否被找到，如果有就输出CUDA Toolkit的版本号，并启用CUDA编程语言。然后，它设置了一个名为HAVE_CUDA的...

runtimeerror: cuda error: cublas_status_execution_failed when calling `cublasgemmex( handle, opa, opb, m, n, k, &falpha, a, cuda_r_16f, lda, b, cuda_r_16f, ldb, &fbeta, c, cuda_r_16f, ldc, cuda_r_32f, cublas_gemm_dfalt_tensor_op)`

运行时错误：调用`cublasgemmex（handle，opa，opb，m，n，k，＆falpha，a，cuda_r_16f，lda，b，cuda_r_16f，ldb，＆fbeta，c，cuda_r_16f，ldc，cuda_r_32f，cublas_gemm_dfalt_tensor_op）`时出现cuda错误：...

cuda_visible_devices无效

cuda_visible_devices是一个环境变量，用于指定在使用CUDA时可见的设备。它的作用是在一个具有多个GPU的系统中，让用户能够选择使用哪个GPU来运行CUDA应用程序。如果在设置cuda_visible_devices后发现无效，可能...

debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions. '.

这个错误信息是在使用PyTorch时出现的，它意味着在CUDA kernel执行时出现了问题。其中，`CUDA_LAUNCH_BLOCKING=1`选项可以启用CUDA kernel的同步执行，以便更容易地调试代码。另外，`TORCH_USE_CUDA_DSA`是一个...

粉丝: 258
资源: 42

上传资源快速赚钱

我的内容管理展开

我的资源快来上传第一个资源

我的收益

登录查看自己的收益

我的积分登录查看自己的积分

我的C币登录后查看C币余额

我的收藏

我的下载

下载帮助

CUDA编程指南：C/C++接口与6.0/6.1/6.2更新

系统分析师教程2008版part4

CUDA C Programming Guide v9.0

CUDA C Programming Guide v8.0

Please use cuda_runtime_api.h or cuda_runtime.h instead."

For debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertion

CUDA_VISIBLE_DEVICES=0 : 无法将“CUDA_VISIBLE_DEVICES=0”项识别为 cmdlet、函数、脚本文件或可运行程序的名称。

NotImplementedError: CUDA_HOME is None. Please set environment variable CUDA_HOME.

CMAKE_CUDA_ARCHITECTURES

os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" os.environ["CUDA_VISIBLE_DEVICES"] = "0"

os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" os.environ["CUDA_VISIBLE_DEVICES"] = "0" CUDA_VISIBLE_DEVICES=1

CUDA_HOME示例

runtimeerror: cuda error: cublas_status_execution_failed when calling `cublasgemmex( handle, opa, opb, m, n, k, &falpha, a, cuda_r_16f, lda, b, cuda_r_16f, ldb, &fbeta, c, cuda_r_16f, ldc, cuda_r_32f, cublas_gemm_dfalt_tensor_op)`

cuda_visible_devices无效

debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions. '.

CUDA_VISIBLE_DEVICES‘ 不是内部或外部命令，也不是可运行的程序或批处理文件

CUDA_HOME environment variable is not set.

我应该如何设置PYTORCH_CUDA_ALLOC_CONF

CMAKE_CUDA_ARCHITECTURES如何设置

CUDA_VISIBLE_DEVICES

最新资源