Chapter One
冯诺依曼机：运算器，中央处理器，主存储器，输入，输出
原理——存储程序
总线；数据传输
总线传输——计算机性能的评级
显示器——选层
独立显卡：专门有块物理显存
内存显卡：从内存分出一块成为显存
显卡与显示器相连接传输：HDMI,VG5
软件与指令架构（硬件、软件的接口）Computer Performance Metrics: Response Time vs. Throughput
Response time takes up processor time: handler time and operating system time
cpu main frequency determines cycle
cpu time = number of instructions * cpu cycles (cpi) * cycle time = number of instructions * cpi / clock frequency
执行时间=秒/程序=指令数/程序 * 时钟周期数/指令数 * 秒/时钟周期数Power consumption = (The lowest point is called the power consumption wall)
Geometric mean is used when comparing two computers using SPEC scores
The calculation formula for the number of server-side java operations performed per watt is as follows:
spec score = reference execution time/execution time of test program
When calculating CPI (Cycles Per Instruction), the actual execution time of the program should be used instead of the reference execution time. Because CPI is an indicator used to measure the efficiency of computer instruction execution, it reflects the number of clock cycles required by the processor in an instruction execution
risc（risc_v,mips,arm）和cisc
Three major instructions (operation instructions (arithmetic, logic), transmission, control)
Instructions consist of two parts (opcode 7 bits, data)
The opcode is always in the seventh position
Operands include (registers (32 64-bit registers), momory, constants),
Why use a register to represent 0?
small segment memory
word alignment not required
Source operands rs, rc
destination operand rd
Register optimization
Instruction format: R shape: the operands are all registers (5 bits), the opcodes are the same, and they are all operation instructions
I form: only one operand is an immediate value, operation instruction, a series of load instructions
jlr
S-shape: separated immediate
SB shape: only even addresses can be jumped
UJ shape: only even addresses can be jumped
The return address is stored in x1