NVIDIACUDA编程指南.pdf_cuda编程

需积分: 50 937 浏览量更新于2023-05-24 评论 5 收藏 6.43MB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

资源详情

资源评论

资源推荐

NVIDIA CUDA 编程指南

NVIDIA

技术文档（全译文）

Jan

2008

Version 1.1

- 2 -

GPU

系列技术文档

.....................................................................................................................1

NVIDIA CUDA

编程指南

.........................................................................................................................1

Chapter1

介绍

CUDA…….....................................................................................................................11

1.1

作为一个并行数据计算设备的图形处理器单元

………………………….............................................11

1.2 CUDA

：一个在

GPU

上计算的新架构

..............................................................................................12

Chapter2

编程模型

............................................................................................................................... 15

2.1

一个超多线程协处理器

.....................................................................................................................15

2.2

线程批处理

.......................................................................................................................................15

2.2.1

线程块

..........................................................................................................................................16

2.2.2

线程块栅格

...................................................................................................................................16

2.3

内存模型

...........................................................................................................................................17

Chapter3

硬件实现

................................................................................................................................18

3.1

一组带有

on-chip

共享内存的

SIMD

多处理器

....................................................................................18

3.2

执行模式

...........................................................................................................................................19

3.3

计算兼容性

........................................................................................................................................20

3.4

多设备

...............................................................................................................................................20

3.5

模式切换

...........................................................................................................................................20

Chapter4

应用程序编程接口（

API

）

...................................................................................................21

4.1

一个

语言的扩展

..............................................................................................................................21

4.2

语言扩展

...........................................................................................................................................21

4.2.1

函数类型限定词

......................................................................................................................... 22

4.2.1.1 __device__..................................................................................................................22

4.2.1.2 __global__..........................................................................................................22

4.2.1.3 __host__............................................................................................................22

4.2.1.4

限定

....................................................................................................................22

4.2.2

变量类型限定词

........................................................................................................................23

4.2.2.1 __device__..................................................................................................23

4.2.2.2 __constant__ .......................................................................................................23

4.2.2.3 __shared__............................................................................................................23

4.2.2.4

限定

.............................................................................................................................24

4.2.3

执行配置

...........................................................................................................................25

4.2.4

内置变量

..............................................................................................................................26

4.2.4.1 gridDim.............................................................................................................................. 26

4.2.4.2 blockIdx...........................................................................................................................26

4.2.4.3 blockDim............................................................................................................................26

- 3 -

4.2.4.4 threadIdx......................................................................................................................... 26

4.2.4.5

限定

..................................................................................................................................... 26

4.2.5 NVCC

编译

.................................................................................................................................. 26

4.2.5.1 __noinline__ ...................................................................................................................27

4.2.5.2 #pragmaunroll .................................................................................................................27

4.3

公共

Runtime

组件

........................................................................................................................... 28

4.3.1

内置矢量类型

............................................................................................................................. 28

4.3.1.1

char1,uchar1,char2,uchar2,char3,uchar3,char4,uchar4,short1,ushort1,short2,us

hort2,short3,ushort3,short4,ushort4,int1,uint1,int2,uint2,int3,uint3,int4,ui

nt4,long1,ulong1,long2,ulong2,long3,ulong3,long4,ulong4,float1,float2,float3

,float4

......................................................................................................................................... 28

4.3.1.2 dim3

类型

........................................................................................................................... 28

4.3.2

数学函数

..................................................................................................................................... 28

4.3.3

时间函数

..................................................................................................................................... 28

4.3.4

纹理类型

..................................................................................................................................... 29

4.3.4.1 Texture Reference

声明

.......................................................................................................29

4.3.4.2 RuntimeTexture Reference

属性

.........................................................................................30

4.3.4.3

线性内存纹理操作对比

CUDA

数组

.......................................................................................31

4.4

设备

Runtime

组件

........................................................................................................................... 31

4.4.1

数学函数

..................................................................................................................................... 31

4.4.2

同步函数

..................................................................................................................................... 31

4.4.3

类型转换函数

............................................................................................................................. 32

4.4.4 TypeCasting

函数

........................................................................................................................32

4.4.5

纹理函数

..................................................................................................................................... 33

4.4.5.1

设备内存纹理操作

................................................................................................................33

4.4.5.2 CUDA

数组纹理操作

............................................................................................................33

4.4.6

原子函数

...................................................................................................................................... 34

4.5

主机

Runtime

组件

........................................................................................................................... 34

4.5.1

公共概念

..................................................................................................................................... 35

4.5.1.1

设备

..................................................................................................................................... 35

4.5.1.2

内存

..................................................................................................................................... 35

4.5.1.3 OpenGL Interoperability ...................................................................................................... 36

4.5.1.4 Direct3D Interoperability ..................................................................................................... 36

4.5.1.5

异步的并发执行

...................................................................................................................37

- 4 -

4.5.2 RuntimeAPI.................................................................................................................................... 38

4.5.2.1

初始化

..................................................................................................................................... 38

4.5.2.2

设备管理

.................................................................................................................................. 38

4.5.2.3

内存管理

.................................................................................................................................. 39

4.5.2.4

流管理

..................................................................................................................................... 40

4.5.2.5

事件管理

.................................................................................................................................. 41

4.5.2.6 Texture Reference

管理

...........................................................................................................42

4.5.2.7 OpenGL Interoperability .......................................................................................................... 44

4.5.2.8 Direct3D Interoperability ......................................................................................................... 44

4.5.2.9

使用设备仿真方式调试

.............................................................................................................45

4.5.3

驱动

API ........................................................................................................................................ 47

4.5.3.1

初始化

..................................................................................................................................... 47

4.5.3.2

设备管理

................................................................................................................................. 47

4.5.3.3 Context

管理

............................................................................................................................. 47

4.5.3.4

模块管理

................................................................................................................................. 48

4.5.3.5

执行控制

................................................................................................................................. 49

4.5.3.6

内存管理

................................................................................................................................. 49

4.5.3.7

流管理

..................................................................................................................................... 51

4.5.3.8

事件管理

................................................................................................................................. 51

4.5.3.9 Texture Reference

管理

..........................................................................................................52

4.5.3.10 OpenGL Interoperability ...................................................................................................... 53

4.5.3.11 Direct3D Interoperability ...................................................................................................... 53

Chapter5

性能指导

................................................................................................................................ 54

5.1

指令性能

........................................................................................................................................... 54

5.1.1

指令吞吐量

................................................................................................................................. 54

5.1.1.1

算术指令

............................................................................................................................. 54

5.1.1.2

控制流指令

.......................................................................................................................... 55

5.1.1.3

内存指令

............................................................................................................................. 56

5.1.1.4

同步指令

............................................................................................................................. 56

5.1.2

内存带宽

.................................................................................................................................... 56

5.1.2.1

全局内存

............................................................................................................................. 57

5.1.2.2

常量内存

............................................................................................................................. 62

5.1.2.3

纹理内存

............................................................................................................................. 63

5.1.2.4

共享内存

............................................................................................................................. 63

5.1.2.5

寄存器

................................................................................................................................. 70

- 5 -

5.2

每个块的线程数量

.......................................................................................................................... 70

5.3

主机与设备的数据传输

....................................................................................................................71

5.4 Texture Fetch

对比全局或常驻内存读取

.........................................................................................71

5.5

性能优化策略总结

.......................................................................................................................... 72

Chapter6

矩阵乘法的例子

....................................................................................................................74

6.1

概要

............................................................................................................................................... 74

6.2

源代码

........................................................................................................................................... 76

6.3

源代码解释

.................................................................................................................................... 78

6.3.1 Mul().................................................................................................................................. 78

6.3.2 Muld()................................................................................................................................ 79

附录

技术规格

................................................................................................................................... 80

A.1

通用规格

....................................................................................................................................... 81

A.2

浮点数标准

.................................................................................................................................... 82

附录

数学函数

................................................................................................................................... 83

B.1

公共

runtime

组件

............................................................................................................................. 83

B.2

设备

runtime

组件

........................................................................................................................... 86

附录

原子函数

................................................................................................................................... 88

C.1

算法函数

....................................................................................................................................... 88

C.1.1 atomicAdd() ................................................................................................................... 88

C.1.2 atomicSub() ................................................................................................................... 88

C.1.3 atomicExch() ................................................................................................................. 88

C.1.4 atomicMin() ................................................................................................................... 88

C.1.5 atomicMax() ................................................................................................................... 89

C.1.6 atomicInc() ....................................................................................................................89

C.1.7 atomicDec() ................................................................................................................... 89

C.1.8 atomicCAS() ................................................................................................................... 89

C.2

位操作函数

.................................................................................................................................... 90

C.2.1 atomicAnd() ................................................................................................................... 90

C.2.2 atomicOr()..................................................................................................................….. 90

C.2.3 atomicXor() ................................................................................................................... 90

附录

D Runtime API Reference ............................................................................................................ 91

D.1

设备管理

........................................................................................................................................ 91

D.1.1 cudaGetDeviceCount().......................................................................................................91

D.1.2 cudaSetDevice()..................................................................................................................91

D.1.3 cudaGetDevice()..................................................................................................................91

D.1.4 cudaGetDeviceProperties() .............................................................................91

D.1.5 cudaChooseDevice() .......................................................................................................93

剩余135页未读，继续阅读

泸州月

粉丝: 314
资源: 27

会员权益专享

NVIDIA CUDA编程指南.pdf

评论0

会员权益专享

最新资源

NVIDIA CUDA编程指南.pdf

评论0

ubuntu安装NVIDIA CUDA

MPI CUDA编程.pdf+英文版+MPI程序讲解

NVIDIA 3D Settings Server 31.0.15.1654NVIDIA CUDA 11.7.99 driver

Jetson-TX2手动安装CUDA和Cudnn.pdf

torch.cuda.amp.autocast

怎么看 libtorch_cuda_cu.so 所在的目录在 LD_LIBRARY_PATH 的路径中

cuda_10.0.130_411.31_win10

CUDA.by.Example.An.Introduction.to.General-Purpose.GPU.Programming

Ubuntu 18.04 安装 NVIDIA显卡驱动及CUDA包.pdf

如何将open3d.cuda.pybind.geometry.RGBDImage转换成numpy

torch.cuda.amp.gradscaler

torch.cuda.amp.autocast()

CUDA编程指南5.0中文版

cat /usr/local/cuda/version.txt

args.cuda = not args.no_cuda and torch.cuda.is_available() 是什么意思

torch.cuda.amp.gradscaler()

libtorch_cuda_cu.so

if torch.cuda.is_available(): torch.set_default_tensor_type(torch.cuda.FloatTensor) print("using cuda:", torch.cuda.get_device_name(0)) pass device = torch.device("cuda" if torch.cuda.is_available() else "cpu")这段代码是什么意思

0004、51单片机12864大液晶屏proteus仿真.zip

小红书测试用例0529

会员权益专享

最新资源