NKU computer architecture lab. An instruction level mips simulator.

指令级 MIPS 模拟器。

The instructions implemented by this simulator are listed below.

该模拟器实现的所有指令如下：

First, a simple decoder to extract fields from instructions is implemented by several functions.

首先，使用几个不同的函数实现了一个简单的指令解码的功能。

uint32_t extract_op(uint32_t inst) { return inst >> 26; }
uint32_t extract_rs(uint32_t inst) { return (inst >> 21) & 0x1f; }
uint32_t extract_rt(uint32_t inst) { return (inst >> 16) & 0x1f; }
uint32_t extract_rd(uint32_t inst) { return (inst >> 11) & 0x1f; }
uint32_t extract_target(uint32_t inst) { return inst & 0x3ffffff; }
uint32_t extract_imm(uint32_t inst) { return inst & 0xffff; }
uint32_t extract_shamt(uint32_t inst) { return (inst >> 6) & 0x1f; }
uint32_t extract_funct(uint32_t inst) { return inst & 0x3f; }

All the fields are extracted before the instruction is executed.

在指令执行前，先提取所有指令中的字段。

uint32_t op = extract_op(inst);
uint32_t rs = extract_rs(inst);
uint32_t rt = extract_rt(inst);
uint32_t imm = extract_imm(inst);
uint32_t rd = extract_rd(inst);
uint32_t shamt = extract_shamt(inst);
uint32_t funct = extract_funct(inst);

And some helper functions for sign/zero extensions. Because of the variety of load instructions, extensions for word, half word and byte are implemented.

此外，还实现了一些用于符号扩展和零扩展的辅助函数。因为加载指令的种类很多，所以分别实现了对 word，half word 和 byte 的扩展。此处的符号扩展使用指针和类型转换的方式。

/// Sign extend the 16 bits immediate
uint32_t sign_ext(uint32_t imm) {
    int32_t signed_imm = *((int16_t*)&imm);
    uint32_t extended_imm = *((uint32_t*)&signed_imm);
    return extended_imm;
}

/// Sign extend a byte to 32 bits
uint32_t sign_ext_byte(uint8_t imm) {
    int32_t signed_imm = *((int8_t*)&imm);
    uint32_t extended_imm = *((uint32_t*)&signed_imm);
    return extended_imm;
}

uint32_t sign_ext_half(uint16_t imm) {
    int32_t signed_imm = *((int16_t*)&imm);
    uint32_t extended_imm = *((uint32_t*)&signed_imm);
    return extended_imm;
}

uint32_t zero_ext(uint32_t imm) { return imm; }

uint32_t zero_ext_byte(uint8_t imm) { return imm; }

uint32_t zero_ext_half(uint16_t imm) { return imm; }

Then in process_instruction, an instruction is firstly fetched from memory using program counter.

在 process_instruction中，首先从内存中根据程序计数器的值来获取一条指令。

uint32_t inst = mem_read_32(CURRENT_STATE.PC);

Then a large switch statement is adopted to handle all the different types of instructions. and funct is switch-ed if necessary.

Note that to handle the branch instruction in the simulator without the delay slot, the PC is incremented by an additional 4.

之后使用一整个 switch 语句根据 op 来处理所有不同的指令类型。如果需要，funct 也会 switch 处理。

在分支指令中，由于模拟器并不实现延迟槽的功能，每一次分支时程序计数器也都需要额外加四。

// BLEZ
uint32_t offset = sign_ext(imm) << 2;

if (rt == 0) {
    if ((CURRENT_STATE.REGS[rs] & 0x80000000) != 0 ||
        CURRENT_STATE.REGS[rs] == 0) {
        NEXT_STATE.PC = CURRENT_STATE.PC + offset + 4;
    } else {
        NEXT_STATE.PC = CURRENT_STATE.PC + 4;
    }
} else {
    // Illegal instruction
    printf("Illegal rt in BLEZ.\n");
}
break;

And if the branch is not taken, the PC is incremented by 4 as normal.

若分支没有执行则程序计数器仍然照常加四。

Other instructions are implemented according to the ISA manual.

其余算数运算、位运算和跳转指令按照 ISA 手册进行编写，具体实现可参见代码。

In load/store instructions, the mem_read_32 and mem_write_32 are used to read and write memory. In case of a write instruction with less than 32 bits, the memory is read first and then the value is concatenated with the original value.

访存指令中使用 mem_read_32 读内存，mem_write_32 写内存，对于写入小于 32 位数的指令（如 sb, sh）采用先读取内存内容并进行拼接的方式。

// SB

uint32_t addr = sign_ext(imm) + CURRENT_STATE.REGS[rs];

uint32_t val = (mem_read_32(addr) & 0xffffff00) |
                (CURRENT_STATE.REGS[rt] & 0xff);

mem_write_32(addr, val);
NEXT_STATE.PC = CURRENT_STATE.PC + 4;
break;

Note that the program counter is incremented by 4 after each instruction.

此外，在每个指令执行之后将 PC 加 4.

NEXT_STATE.PC = CURRENT_STATE.PC + 4;

Spim is buggy and the latest version has poor support for pseudo instructions. So mars is used to assemble the code and test the simulator. Assemble scripts using spim is under tools, but note that pseudo instructions (e.g. large immediates) are not supported by spim. Also the mars is also under the tools folder. Mars seems not available through command line, so the .x files are generated by hand. The output of sim is compared with the output of mars.

由于 Spim 存在一些 Bug 并且最新版本对伪指令的支持不好，所以使用 mars 来进行汇编和测试。使用 spim 进行汇编的脚本在 tools 文件夹下，但是注意伪指令（比如大立即数）在 spim 中不支持。同时，mars 这个工具也在 tools 文件夹下。由于没有找到通过命令行直接操作 mars 的方法，所以采用了手动汇编的方式生成了 .x 文件。测试时直接对比 sim 寄存器的输出和 mars 运行结果。

Additional testcases are under tests folder. The results of the simulator is the same as the results of mars.

额外的测试用例在 tests 文件夹下。模拟器的输出结果和 mars 的输出结果相同。