Micro Language Compiler
Assignment 1: Micro Language Compiler
1 Introduction
In this assignment, you are required to design and implement a compiler frontend for Micro
language which transforms the Micro Program into corresponding LLVM Intermediate Representation (IR) and finally translated to RISC-V assembly code and executable with the help of
LLVM optimizer and its RISC-V backend. After that, we can execute the compiled program on our
RISC-V docker container to verify the correctness of your compiler.
Since it is a senior major elective course, we don’t want to set any limitation for you. You are strongly
recommended to use Lex/Flex and Yacc/Bison taught in tutorial 3 to design your compiler frontend,
but it is not forcible. You can choose Any Programming Language you like for this assignment,
but the RISC-V container we use only has C/C++ toolchain installed and you need to provide me a
Dockerfile to run your compiler and execute the RISC-V program, which may need some extra effort.
Some languages also provide tools like Lex and Yacc, and you are free to use them. It is also OK if
you want to design the scanner and parser by hand instead of using tools.
2 Micro Language
Before we move on to the compiler design, it is necessary to have an introduction to Micro Language,
that serves as the input of our compiler. Actually, it is a very simple language, with limited number
of tokens and production rules of context-free grammar (CFG):
• Only integers(i32); No float numbers
• No Declarations
• Variable consists of a-z, A-Z, 0-9, at most 32 characters long, must start with character and are
initialized as 0
• Comments begin with ”−−” and end with end-of-line(EOL)
• Three kind of statements:
– assignments, e.g. a:=b+c
– read(list of IDs), e.g. read(a, b)
– write(list of Expressions), e.g. write (a, b, a+b)
• BEGIN, END, READ, WRITE are reserved words
• Tokens may not extend to the following line
1
2.1 Tokens & Regular Expression
Micro Language has 14 Tokens, and the regular expression for each token is listed below. Since BEGIN
is a reserved word in C/C++, we need to use the alternative BEGIN as the token class.
1. BEGIN : begin
2. END: end
3. READ: read
4. WRITE: write
5. LPAREN: (
6. RPAREN: )
7. SEMICOLON: ;
8. COMMA: ,
9. ASSIGNOP: :=
10. PLUSOP: +
11. MINUSOP: −
12. ID: [a−zA−Z][a−zA−Z0−9 ]{0,31}
13. INTLITERAL: −?[0−9]+
14. SCANEOF: <
2.2 Context Free Grammar
Here is the extended context-free grammar (CFG) of Micro Language:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Note: {} means the content inside can appear 0, 1 or multiple times.
2.3 How to Run Micro Compiler
Here is a very simple Micro program that we are going to use as the sample program throughout this
instruction. SCANEOF is the end of line and we do not need to explicitly write it in the program.
−− Expected Output: 30
begin
A := 10;
B := A + 20;
write (B);
end
We can use our compiler to compile, optimize and execute this program to get expected output 30.
Note: The exact command to run your compiler is up to you. Just specify out in your report how
to compile and execute your compiler to get the expected output.
118010200@c2d52c9b1339:˜/A1$ ./compiler ./testcases/test0.m
118010200@c2d52c9b1339:˜/A1$ llc −march=riscv64 ./program.ll −o ./program.s
118010200@c2d52c9b1339:˜/A1$ riscv64−unknown−linux−gnu−gcc ./program.s −o ./program
118010200@c2d52c9b1339:˜/A1$ qemu−riscv64 −L /opt/riscv/sysroot ./program
30
2
3 Compiler Design
Figure 1 shows the overall structure of a compiler. In this assignment, your task is to implement the
frontend only, which contains scanner, parser and intermediate code generator and form a whole →
compiler with LLVM for Micro language.
Figure 1: Compiler Structure
3.1 Scanner
Scanner takes input character stream and extracts out a series of tokens, and your scanner should
be able to print out both token class and lexeme for each token. Figure ?? shows an example of the
sample Micro program.
118010200@c2d52c9b1339:˜/A1$ ./compiler ./testcases/test0.m −−scan−only
BEGIN begin
ID A
ASSIGNOP :=
INTLITERAL 10
SEMICOLON ;
ID B
ASSIGNOP :=
ID A
PLUOP +
INTLITERAL 20
SEMICOLON ;
WRITE write
LPAREN (
ID B
RPAREN )
SEMICOLON ;
END end
SCANEOF
3
3.2 Parser
Parser receives the tokens extracted from scanner, and generates a parse tree (or concrete syntax tree),
and futhermore, the abstract syntax tree (AST) based on the CFG. Your compiler should be able to
print out both the parse tree and the abstract syntax tree, and visualize them with graphviz.
3.2.1 Parse Tree (Concrete Syntax Tree)
Figure 2 shows an example of the concrete syntax tree generated from the sample program:
Figure 2: Concrete Syntax Tree of Sample Program
3.2.2 Abstract Syntax Tree
Figure 3 shows an example of the abstract syntax tree generated from the sample program:
Figure 3: Abstract Syntax Tree of Sample Program
4
3.3 Intermediate Code Generator
For all the assignments in this course, the intermediate representation (IR) of the compiler should be
the LLVM IR. There are mainly two reasons why LLVM IR is chosen. One is that LLVM IR can take
advantage of the powerful LLVM optimizer and make up for the missing backend part of compiler in
this course. The other is that LLVM IR can be easier translated into assembly code for any target
machine, no matter MIPS, x86 64, ARM, or RISC-V. This functionality can make our compiler more
compatible to machines with different architecture. The following shows the LLVM IR generated for
the sample Micro program:
; Declare printf
declare i32 @printf (i8 ∗, ...)
; Declare scanf
declare i32 @scanf(i8 ∗, ...)
define i32 @main() {
% ptr0 = alloca i32
store i32 10, i32∗ % ptr0
%A = load i32, i32∗ % ptr0
% 1 = add i32 %A, 20
store i32 % 1, i32∗ % ptr0
%B = load i32, i32∗ % ptr0
% scanf format0 = alloca [4 x i8 ]
store [4 x i8 ] c”%d\0A\00”, [4 x i8]∗ % scanf format0
% scanf str0 = getelementptr [4 x i8 ], [4 x i8]∗ % scanf format0, i32 0, i32 0
call i32 (i8 ∗, ...) @printf (i8∗ % scanf str0 , i32 %B)
ret i32 0
}
3.4 Bonus (Extra Credits 10%)
If you are interested and want to make your compiler better, you may try the following options:
• Add robust syntax error report
• Add a symbol table to make your compiler more complete
• Generate LLVM IR with LLVM C/C++ API instead of simply generating the string
• Optimize the LLVM IR generation plan for more efficient IR
• Any other you can think about...
4 Submission and Grading
4.1 Grading Scheme
• Scanner: 20%
• Parser: 40% (20% for parse tree generator and 20% for AST generation)
• Intermediate Code Generator: 30%
We have prepared 10 test cases, and the points for each section will be graded according to the
number of testcases you passed.
• Technical Report: 10%
If your report properly covers the three required aspects and the format is clean, you will get 10
points.
5
• Bonus: 10%
Refer to section 3.4 for more details. The grading of this part will be very flexible and highly
depend on the TA’s own judgement. Please specify clearly what you have done for the bonus
part so that he do not miss anything.
4.2 Submission with Source Code
If you want to submit source C/C++ program that is executable in our RISC-V docker container,
your submission should look like:
csc4180−a1−118010200.zip
|−
|−−− csc4180−a1−118010200−report.pdf
|−
|−−− testcases
|−
|−−− src
|−
|−−−Makefile
|−−−ir generator.cpp
|−−−ir generator.hpp
|−−−node.cpp
|−−−node.hpp
|−−−scanner.l
|−−−parser.y
|−−−Other possible files
4.3 Submission with Docker
If you want to submit your program in a docker, your submission should look like:
csc4180−a1−118010200.zip
|−
|−−− csc4180−a1−118010200.Dockerfile
|−
|−−− csc4180−a1−118010200−report.pdf
|−
|−−− src
|−
|−−−Makefile
|−
|−−−run compiler.sh
|−
|−−−Your Code Files
4.4 Technical Report
Please answer the following questions in your report:
• How to execute your compiler to get expected output?
• How do you design the Scanner?
• How do you design the Parser?
• How do you design the Intermediate Code Generator?
• Some other things you have done in this assignment?
The report doesn’t need to be very long, but the format should be clean. As long as the three WX：codehelp