X86-32/64, SSE & AVX编程全攻略：从基础到高级技术详解

需积分: 5 5 浏览量更新于2024-07-17 收藏 7.63MB PDF 举报

现代X86汇编语言编程是一本深入讲解X86架构及其不同版本（32位和64位）的专业书籍，特别关注SSE和AVX等高级指令集扩展。该书结构清晰，共分为15章，旨在帮助读者理解和掌握在不同平台上的低级编程技术。第1章至第4章涵盖了基础的X86-32核心架构和编程，包括X86-32处理器的基本工作原理、指令集和内存管理。第1章介绍了X86-32的核心架构，让读者对处理器的体系结构有初步了解；第2章则深入到具体的编程实践，涉及内存操作、控制转移和数据类型等。接下来的章节转向更复杂的计算能力增强特性。第5章和第6章讲述了MMX（MultiMedia Extension）技术，这是早期用于多媒体处理的增强指令集，支持128位数据处理。第7章进一步扩展到了Streaming SIMD Extensions（SSE），该部分着重于单精度浮点数和向量运算，包括SSE编程的各个方面。第8至第11章深入探讨了SSE的高级应用，分别涵盖了单精度和双精度浮点数的向量操作、整数打包以及文本字符串处理。这些章节详细介绍了如何利用SSE优化性能，尤其是在处理大量数据时。最后，本书的高潮部分是第12至第15章，重点讲解了Advanced Vector Extensions（AVX）。AVX是X86架构的一个重大飞跃，它支持256位的向量操作，提供了更高的并行性和计算效率。章节依次介绍了AVX在单精度和双精度浮点数、整数打包以及可能的应用场景中的编程方法。这本教材适合对底层汇编编程感兴趣或者从事计算机科学、软件开发的专业人士，特别是那些需要优化性能、利用最新硬件特性进行高精度计算的开发者。通过学习，读者将能掌握如何在现代X86平台上编写高效、可移植的代码，充分利用SSE和AVX提供的强大计算能力。

Chapter 1 ■ X86-32 Core arChiteCture

Scalar data types are used to perform calculations with discrete values. The x86

platform supports a set of scalar data types that resemble the basic data types available

in C/C++. These are illustrated in Table 1-2. The x86-32 instruction set intrinsically

supports operations on 8-, 16-, and 32-bit scalar integers, both signed and unsigned.

A few instructions are also capable of manipulating 64-bit values. Comprehensive

support for 64-bit values, however, requires x86-64 mode.

Table 1-2. X86 Numerical Data Types

Type Size in Bits Equivalent C/C++ Type

Signed integers 8 char

16 short

32 int, long

64 long long

Unsigned integers 8 unsigned char

16 unsigned short

32 unsigned int, unsigned long

64 unsigned long long

Floating-point 32 float

64 double

80 long double

The x87 FPU supports three different scalar floating-point encodings ranging in

length from 32 to 80 bits. X86 assembly language functions can readily use any of the

supported encodings. It should be noted, however, that C/C++ support for the 80-bit

double extended-precision floating-point data encoding is not universal. Some compilers

use the 64-bit encoding for both double and long double. Chapter 3 examines the x87

FPU and its supported data types in greater detail.

Packed Data Types

The x86 platform supports a variety of packed data types, which are employed to perform

SIMD calculations using either integers or floating-point values. For example, a 64-bit

wide packed data type can be used to hold eight 8-bit integers, four 16-bit integers, or

two 32-bit integers. A 256-bit wide packed data type can hold a variety of data elements

including 32 8-bit integers, 8 single-precision floating-point values, or 4 double-precision

floating-point values. Table 1-3 lists the valid packed data type sizes along with the

corresponding data element types and maximum possible number of data elements.

Chapter 1 ■ X86-32 Core arChiteCture

The remainder of this section examines the x86-32 core execution unit in greater

detail. It starts an exploration of the unit’s register sets, including its segment registers,

general-purpose registers, status flags register, and instruction pointer. This is followed

by a discussion of instruction operands and memory addressing modes. The remaining

execution units are examined later in this book. Chapter 3 explores the internal

architecture of the x87 FPU, while Chapters 5, 7, and 12 delve into the architectural

intricacies of MMX, x86-SSE, and x86-AVX, respectively.

Segment Registers

The x86-32 core execution unit uses segment registers to define a logical memory model

for program execution and data storage. An x86 processor contains six segment registers

that designate blocks of memory for code, data, and stack space. When operating in

x86-32 protected mode, a segment register contains a segment selector, which is used

as an index into a segment descriptor table that defines the segment’s operational

characteristics. A segment’s operational characteristics include its size, type (code or

data), and access rights (read or write). Segment register initialization and management

is normally handled by the operating system. Most x86-32 application programs are

written without any direct knowledge of how the segment registers are programmed.

General-Purpose Registers

The x86-32 core execution unit contains eight 32-bit general-purpose registers. These

registers are primarily used to perform logical, arithmetic, and address calculations. They

also can be employed for temporary storage and as pointers to data items that are stored

in memory. Figure 1-3 shows the complete set of general-purpose registers along with the

names that are used to specify a register as an instruction operand. Besides supporting

32-bit operands, the general-purpose registers also can perform calculations using 8-bit

or 16-bit operands. For example, a function can use registers AL, BL, CL, and DL to

perform 8-bit calculations in the low-order bytes of registers EAX, EBX, ECX, and EDX,

respectively. Similarly, the registers AX, BX, CX, and DX can be used to carry out 16-bit

calculations in the low-order words.

Chapter 1 ■ X86-32 Core arChiteCture

Despite their designation as general-purpose registers, the x86-32 instruction set

imposes some noteworthy restrictions on how they can be used. Many instructions either

require or implicitly use specific registers as operands. For example, some variations of

the imul (Signed Multiply) and idiv (Signed Divide) instructions use the EDX register to

hold the high-order doubleword of a product or dividend. The string instructions require

that the addresses of the source and destination operands be placed in the ESI and EDI

registers, respectively. String instructions that include a repeat prefix must use ECX as

the count register, while variable bit shift and rotate instructions must load the bit count

value into the CL register.

The processor uses the ESP register to support stack-related operations such as

function calls and returns. The stack itself is simply a contiguous block of memory that

is assigned to a process or thread by the operating system. Application programs can

also use the stack to pass function arguments and store temporary data. Register ESP

always points to the stack’s top-most item. While it is possible to use the ESP register as a

general-purpose register, such use is impractical and strongly discouraged. Register EBP

is typically used as a base pointer to access data items that are stored on the stack (ESP

can also be used as a base pointer to access data items on the stack). When not employed

as a base pointer, EBP can be used as a general-purpose register.

The mandatory or implicit use of specific registers by some instructions is a legacy

design pattern that dates back to the 8086, ostensibly to improve code density. What this

means from a modern programing perspective is that certain register usage conventions

tend be observed when writing x86-32 assembly code. Table 1-4 lists the general-purpose

registers and their conventional uses.

EAX

EBX

ECX

EDX

ESI

EDI

EBP

ESP

BLBH

CLCH

DLDH

ALAH

32-Bit Registers

031

8-Bit and 16-Bit

Registers

07815

Figure 1-3. X86-32 general-purpose registers

剩余684页未读，继续阅读

GPU大哥

粉丝: 2
资源: 11

X86-32/64, SSE & AVX编程全攻略：从基础到高级技术详解

Modern X86 Assembly Language Programming 32-bit, 64-bit, SSE and AVX 无水印pdf

Modern X86 Assembly Language Programming 32-bit, 64-bit, SSE

modern-x86-assembly-language-programming-2e:Daniel Kusswurm的“现代X86汇编语言编程”源代码-e language source code

现代X86汇编语言程序设计Modern X86 Assembly Language Programming

modern-cpp-programming-cookbook-2nd

Modern-CPP-Programming-Cookbook-Second-Edition-master.zip

modern_x86_assembly_language_programming_x86_Asm_

《The C Programming Language》---2nd---英文PDF

ARM Assembly Language Programming.pdf

Modern-CPP-Programming-Cookbook-Second-Edition:Packt出版的《现代C ++编程手册》，第二版

最新资源