Summary

test code sets: this link todo the file structure for this project

Project 4

# t5 := b - a
  #++ 'ensureReg           '  (b) <-> $v0
  #++ 'ensureReg           '  (a) <-> $v0
  #++ 'seizeReg            '
  #++ 'spillReg            '
  #++ 'print_spill_absorb  '
  sw   $v0, -16($fp)
  #++ 'print_spill_absorb  '
  lw   $v0, -12($fp)
  #++ 'ensureReg           '  (t5) <-> $a0
  #++ 'init_addr_descriptor'  OFFSET -56
  addi $sp, $sp, -4
  #++ 'seizeReg            '
  #++ 'spillReg            '
  #++ 'print_spill_absorb  '
  sw   $a0, -24($fp)
  sub  $a0, $v0, $v0

I used MARS(MIPS Assembler and Runtime Simulator) external tools setting up options

adjust the sequence of arguments in Compiler.c

Assumptions:

• assume all variable will be saved in at most one register at any given time.
• assume all variables will be saved before exiting the basic block

counter.arg will be reset in caller every time function call returns from callee.

learn to use:

• asprintf
• qsort, bsearch
• generic custom function such as reverseArray in util.c
• typeof

Project 3

Date: 29 Nov 2024

State: Finish all basic functionalities as well as the second option which is multi-dimension array.

In the duration of writing code, I refer to this repo. Thanks Unkrible!

Compared with the last project where I used a myriad of malloc to manually create data. This time, I made use of initializers and 'compound literals' powered by C itself. What is also highly recognizable is that I leveraged a huge amount of macros which season the code to be more expressive, efficient as well as tasty:).

Project 2

Date: 14 Nov 2024

State: Finished all functionality related to mandatory and the second optional task (scoping).

I managed to lower the complexity and improve efficiency in the process of writing code. Before starting this project, migrating to CMake from Makefile was an unavoidable hurdle, as I missed the convenient functionalities provided by CLion. So, I started learning it the hard way.

After achieving that, I coincidentally noticed dynamic analysis tools nested in the depths of CLion settings. I went a little astray for a while, but the benefits of mastering sanitizers gradually appeared when dealing with pointers. I am now reassuringly peaceful even though lagging behind.

• Redirecting stdout to a file is helpful for reference.
• Using soft links to solve include errors.
• Adding const and assert statements, no matter how tiresome, prevents the program from derailing.
• A new way of constructing a linked list using double pointers:

(*t)->array.elemType = NULL; // Initialize elemType to NULL
t = &(*t)->array.elemType;   // Update t to point to elemType

• Inside switch statements, initialization cannot be done directly without curly braces.
• A seemingly trivial error: I originally evaluated Exp ASSIGNOP EXP inside resolveDef. Alas, at this time variables haven't been inserted into vMap. So I squeeze through it by using a cumbersome function checkInit. Hope I will roll with it later.

Project 1

Data: 7 Oct 2024

实现的功能 必选与选项一, 所要求的功能均已实现

总结

先从变成工具的使用说起，往常C语言的开发，我都在Clion中进行。但是这次，因为没有提供cmakelist，我只好在vscode 中进行。而恰好是这次，让我能够更深入地了解vscode。我学会了写task.json，了解到vscode还支持profile等等。我使用git进行历史记录的保留。这些都是非常宝贵的尝试。

前期，主要面临工具使用不熟练的问题。一方面，表现在上述的问题；另一方面，表现为不熟悉flex和bison的使用。对此二者编程几乎无工具可言，只能一个字符一个字符地敲上。这种绝无仅有的体验，让我感受到了纯粹痛苦之后的快乐。

让我印象较深的一个bug是，我在bison的文件中指定INT返回类型为int， FLOAT返回类型为float，而其他则返回node。这么定义是方便flex返回值，比如检测到一个int，就返回对应的int值。后来，我发现这样不对，此举会让bison错误判断INT在bison文件中的返回值。具体而言，从flex到bison传递的是int类型的值，在bison中从INT到包含有INT的表达式的值依旧是int类型，然而本该是node类型，只有这样才能构建语法树。回望千钧一发之际，我不胜悲凉，原来debug是要靠直觉的……