Files
elna/boot/stage2.elna
2025-08-30 01:29:00 +02:00

856 lines
14 KiB
Plaintext

# This Source Code Form is subject to the terms of the Mozilla Public License,
# v. 2.0. If a copy of the MPL was not distributed with this file, You can
# obtain one at https://mozilla.org/MPL/2.0/.
# Stage2 compiler.
#
# It supports declaring and calling procedures without arguments.
# A procedure name should start with an underscore.
.section .rodata
.type keyword_equ, @object
keyword_equ: .ascii ".equ"
.equ KEYWORD_EQU_SIZE, 4
.type keyword_section, @object
keyword_section: .ascii ".section"
.equ KEYWORD_SECTION_SIZE, 8
.type keyword_type, @object
keyword_type: .ascii ".type"
.equ KEYWORD_TYPE_SIZE, 5
.type keyword_ret, @object
keyword_ret: .ascii "ret"
.equ KEYWORD_RET_SIZE, 3
.type keyword_global, @object
keyword_global: .ascii ".globl"
.equ KEYWORD_GLOBAL_SIZE, 6
.type keyword_proc, @object
keyword_proc: .ascii "proc "
.equ KEYWORD_PROC_SIZE, 5
.type keyword_end, @object
keyword_end: .ascii "end"
.equ KEYWORD_END_SIZE, 3
.type keyword_begin, @object
keyword_begin: .ascii "begin"
.equ KEYWORD_BEGIN_SIZE, 5
.type keyword_var, @object
keyword_var: .ascii "var"
.equ KEYWORD_VAR_SIZE, 3
.type asm_prologue, @object
asm_prologue: .string "\taddi sp, sp, -32\n\tsw ra, 28(sp)\n\tsw s0, 24(sp)\n\taddi s0, sp, 32\n"
.type asm_epilogue, @object
asm_epilogue: .string "\tlw ra, 28(sp)\n\tlw s0, 24(sp)\n\taddi sp, sp, 32\n\tret\n"
.type asm_type_directive, @object
asm_type_directive: .string ".type "
.type asm_type_function, @object
asm_type_function: .string ", @function\n"
.type asm_colon, @object
asm_colon: .string ":\n"
.type asm_call, @object
asm_call: .string "\tcall "
.type asm_j, @object
asm_j: .string "\tj "
.type asm_li, @object
asm_li: .string "\tli "
.type asm_lw, @object
asm_lw: .string "\tlw "
.type asm_t0, @object
asm_t0: .string "t0"
.type asm_a0, @object
asm_a0: .string "a0"
.type asm_comma, @object
asm_comma: .string ", "
.type asm_sp, @object
asm_sp: .string "(sp)"
.section .bss
.equ SOURCE_BUFFER_SIZE, 81920
.type source_code, @object
source_code: .zero SOURCE_BUFFER_SIZE
.section .data
.type source_code_position, @object
source_code_position: .word source_code
.section .text
# Reads standard input into a buffer.
# a0 - Buffer pointer.
# a1 - Buffer size.
#
# Returns the amount of bytes written in a0.
proc _read_file();
begin
mv a2, a1
mv a1, a0
# STDIN.
li a0, 0
li a7, 63 # SYS_READ.
ecall
end;
# Writes to the standard output.
#
# Parameters:
# a0 - Buffer.
# a1 - Buffer length.
proc _write();
begin
mv a2, a1
mv a1, a0
# STDOUT.
li a0, 1
li a7, 64 # SYS_WRITE.
ecall
end;
# Writes a character from a0 into the standard output.
proc _write_c();
begin
sb a0, 20(sp)
addi a0, sp, 20
li a1, 1
_write();
end;
# Write null terminated string.
#
# Parameters:
# a0 - String.
proc _write_z();
begin
sw a0, 20(sp)
.write_z_loop:
# Check for 0 character.
lb a0, (a0)
beqz a0, .write_z_end
# Print a character.
lw a0, 20(sp)
lb a0, (a0)
_write_c();
# Advance the input string by one byte.
lw a0, 20(sp)
addi a0, a0, 1
sw a0, 20(sp)
j .write_z_loop
.write_z_end:
end;
# Detects if a0 is an uppercase character. Sets a0 to 1 if so, otherwise to 0.
proc _is_upper();
begin
li t0, 'A' - 1
sltu t1, t0, a0 # t1 = a0 >= 'A'
sltiu t2, a0, 'Z' + 1 # t2 = a0 <= 'Z'
and a0, t1, t2 # t1 = a0 >= 'A' & a0 <= 'Z'
end;
# Detects if a0 is an lowercase character. Sets a0 to 1 if so, otherwise to 0.
proc _is_lower();
begin
li t0, 'a' - 1
sltu t2, t0, a0 # t2 = a0 >= 'a'
sltiu t3, a0, 'z' + 1 # t3 = a0 <= 'z'
and a0, t2, t3 # t2 = a0 >= 'a' & a0 <= 'z'
end;
# Detects if the passed character is a 7-bit alpha character or an underscore.
#
# Paramters:
# a0 - Tested character.
#
# Sets a0 to 1 if the character is an alpha character or underscore, sets it to 0 otherwise.
proc _is_alpha();
begin
sw a0, 20(sp)
_is_upper();
sw a0, 16(sp)
lw a0, 20(sp)
_is_lower();
lw t0, 20(sp)
xori t1, t0, '_'
seqz t1, t1
lw t0, 16(sp)
or a0, a0, t0
or a0, a0, t1
end;
# Detects whether the passed character is a digit
# (a value between 0 and 9).
#
# Parameters:
# a0 - Exemined value.
#
# Sets a0 to 1 if it is a digit, to 0 otherwise.
proc _is_digit();
begin
li t0, '0' - 1
sltu t1, t0, a0 # t1 = a0 >= '0'
sltiu t2, a0, '9' + 1 # t2 = a0 <= '9'
and a0, t1, t2
end;
# Reads the next token.
#
# Returns token length in a0.
proc _read_token();
begin
la t0, source_code_position # Token pointer.
lw t0, (t0)
sw t0, 20(sp) # Current token position.
sw zero, 16(sp) # Token length.
.read_token_loop:
lb t0, (t0) # Current character.
# First we try to read a derictive.
# A derictive can contain a dot and characters.
li t1, '.'
beq t0, t1, .read_token_next
lw a0, 20(sp)
lb a0, (a0)
_is_alpha();
bnez a0, .read_token_next
lw a0, 20(sp)
lb a0, (a0)
_is_digit();
bnez a0, .read_token_next
j .read_token_end
.read_token_next:
# Advance the source code position and token length.
lw t0, 16(sp)
addi t0, t0, 1
sw t0, 16(sp)
lw t0, 20(sp)
addi t0, t0, 1
sw t0, 20(sp)
j .read_token_loop
.read_token_end:
lw a0, 16(sp)
end;
# a0 - First pointer.
# a1 - Second pointer.
# a2 - The length to compare.
#
# Returns 0 in a0 if memory regions are equal.
proc _memcmp();
begin
mv t0, a0
li a0, 0
.Lmemcmp_loop:
beqz a2, .Lmemcmp_end
lbu t1, (t0)
lbu t2, (a1)
sub a0, t1, t2
bnez a0, .Lmemcmp_end
addi t0, t0, 1
addi a1, a1, 1
addi a2, a2, -1
j .Lmemcmp_loop
.Lmemcmp_end:
end;
# Advances the token stream by a0 bytes.
proc _advance_token();
begin
# Skip the .equ directive.
la t0, source_code_position
lw t1, (t0)
add t1, t1, a0
sw t1, (t0)
end;
# Prints the current token.
#
# Parameters:
# a0 - Token length.
#
# Returns a0 unchanged.
proc _write_token();
begin
sw a0, 20(sp)
la a0, source_code_position
lw a0, (a0)
lw a1, 20(sp)
_write();
lw a0, 20(sp)
end;
proc _compile_section();
begin
# Print and skip the .section directive and a space after it.
li a0, KEYWORD_SECTION_SIZE + 1
_write_token();
_advance_token();
# Read the section name.
_read_token();
addi a0, a0, 1
_write_token();
_advance_token();
end;
# Prints and skips a line.
proc _skip_comment();
begin
la t0, source_code_position
lw t1, (t0)
.skip_comment_loop:
# Check for newline character.
lb t2, (t1)
li t3, '\n'
beq t2, t3, .skip_comment_end
# Advance the input string by one byte.
addi t1, t1, 1
sw t1, (t0)
j .skip_comment_loop
.skip_comment_end:
# Skip the newline.
addi t1, t1, 1
sw t1, (t0)
end;
# Prints and skips a line.
proc _compile_line();
begin
.compile_line_loop:
la a0, source_code_position
lw a1, (a0)
lb t0, (a1)
li t1, '\n'
beq t0, t1, .compile_line_end
# Print a character.
lw a0, (a1)
_write_c();
# Advance the input string by one byte.
li a0, 1
_advance_token();
j .compile_line_loop
.compile_line_end:
li a0, '\n'
_write_c();
li a0, 1
_advance_token();
end;
proc _compile_integer_literal();
begin
la a0, asm_li
_write_z();
la a0, asm_a0
_write_z();
la a0, asm_comma
_write_z();
_read_token();
_write_token();
_advance_token();
li a0, '\n'
_write_c();
end;
proc _compile_character_literal();
begin
la a0, asm_li
_write_z();
la a0, asm_a0
_write_z();
la a0, asm_comma
_write_z();
.compile_character_literal_loop:
la a0, source_code_position
lw a0, (a0)
li a1, 1
_write();
li a0, 1
_advance_token();
la t0, source_code_position
lw t0, (t0)
lb a0, (t0)
li t1, '\''
beq a0, t1, .compile_character_literal_end
j .compile_character_literal_loop
.compile_character_literal_end:
li a0, '\''
_write_c();
li a0, '\n'
_write_c();
li a0, 1
_advance_token();
end;
proc _compile_variable_expression();
begin
la a0, asm_lw
_write_z();
la a0, asm_a0
_write_z();
la a0, asm_comma
_write_z();
la a0, source_code_position
lw a0, (a0)
addi a0, a0, 1
li a1, 2
_write();
la a0, asm_sp
_write_z();
li a0, '\n'
_write_c();
li a0, 3
_advance_token();
end;
proc _compile_expression();
begin
la t0, source_code_position
lw t0, (t0)
lb a0, (t0)
li t1, '\''
beq a0, t1, .compile_expression_character_literal
li t1, 'v'
beq a0, t1, .compile_expression_variable
_is_digit();
bnez a0, .compile_expression_integer_literal
j .compile_expression_end
.compile_expression_character_literal:
_compile_character_literal();
j .compile_expression_end
.compile_expression_integer_literal:
_compile_integer_literal();
j .compile_expression_end
.compile_expression_variable:
_compile_variable_expression();
j .compile_expression_end;
.compile_expression_end:
end;
proc _compile_call();
begin
_read_token();
sw a0, 20(sp)
la t0, source_code_position
lw t0, (t0)
sw t0, 16(sp)
# Skip the identifier and left paren.
addi a0, a0, 1
_advance_token();
la t0, source_code_position
lw t0, (t0)
lb t0, (t0)
li t1, ')'
beq t0, t1, .compile_call_finalize
_compile_expression();
.compile_call_finalize:
la a0, asm_call
_write_z();
lw a0, 16(sp)
lw a1, 20(sp)
_write();
# Skip the right paren.
li a0, 1
_advance_token();
end;
proc _compile_goto();
begin
li a0, 5
_advance_token();
_read_token();
sw a0, 20(sp)
la a0, asm_j
_write_z();
lw a0, 20(sp)
_write_token();
_advance_token();
end;
proc _compile_statement();
begin
# This is a call if the statement starts with an underscore.
la t0, source_code_position
lw t0, (t0)
# First character after alignment tab.
addi t0, t0, 1
lb t0, (t0)
li t1, '_'
beq t0, t1, .compile_statement_call
li t1, 'g'
beq t0, t1, .compile_statement_goto
_compile_line();
j .compile_statement_end
.compile_statement_call:
li a0, 1
_advance_token();
_compile_call();
j .compile_statement_semicolon
.compile_statement_goto:
li a0, 1
_advance_token();
_compile_goto();
j .compile_statement_semicolon
.compile_statement_semicolon:
li a0, 2
_advance_token();
li a0, '\n'
_write_c();
.compile_statement_end:
end;
proc _compile_procedure_body();
begin
.compile_procedure_body_loop:
la a0, source_code_position
lw a0, (a0)
la a1, keyword_end
li a2, KEYWORD_END_SIZE
_memcmp();
beqz a0, .compile_procedure_body_epilogue
_compile_statement();
j .compile_procedure_body_loop
.compile_procedure_body_epilogue:
end;
proc _compile_procedure();
begin
# Skip "proc ".
li a0, KEYWORD_PROC_SIZE
_advance_token();
_read_token();
sw a0, 20(sp) # Save the procedure name length.
# Write .type _procedure_name, @function.
la a0, asm_type_directive
_write_z();
lw a0, 20(sp)
_write_token();
la a0, asm_type_function
_write_z();
# Write procedure label, _procedure_name:
lw a0, 20(sp)
_write_token();
la a0, asm_colon
_write_z();
# Skip the function name and trailing parens, semicolon, "begin" and newline.
lw a0, 20(sp)
addi a0, a0, KEYWORD_BEGIN_SIZE + 1 + 4
_advance_token();
la a0, asm_prologue
_write_z();
_compile_procedure_body();
# Write the epilogue.
la a0, asm_epilogue
_write_z();
li a0, KEYWORD_END_SIZE + 2
_advance_token();
end;
proc _compile_type();
begin
# Print and skip the .type directive and a space after it.
li a0, KEYWORD_TYPE_SIZE + 1
_write_token();
_advance_token();
# Read and print the symbol name.
_read_token();
sw a0, 20(sp)
# Print and skip the symbol name, comma, space and @.
lw a0, 20(sp)
addi a0, a0, 3
_write_token();
_advance_token();
# Read the symbol type.
_read_token();
sw a0, 16(sp)
la t0, source_code_position
lw t0, (t0)
sw t0, 12(sp)
# Print the symbol type and newline.
lw a0, 16(sp)
addi a0, a0, 1
_write_token();
_advance_token();
# Write the object definition itself.
_compile_line();
.compile_type_end:
end;
proc _compile_equ();
begin
# Print and skip the .equ directive and a space after it.
li a0, KEYWORD_EQU_SIZE + 1
_write_token();
_advance_token();
# Read and print the constant name.
_read_token();
sw a0, 20(sp)
# Print and skip the constant name, comma and space.
lw a0, 20(sp)
addi a0, a0, 2
_write_token();
_advance_token();
# Read the constant value.
_read_token();
sw a0, 16(sp)
# Print and skip the constant value and newline.
lw a0, 16(sp)
addi a0, a0, 1
_write_token();
_advance_token();
end;
proc _skip_newlines();
begin
# Skip newlines.
la t0, source_code_position
lw t1, (t0)
.skip_newlines_loop:
lb t2, (t1)
li t3, '\n'
bne t2, t3, .skip_newlines_end
beqz t2, .skip_newlines_end
addi t1, t1, 1
sw t1, (t0)
j .skip_newlines_loop
.skip_newlines_end:
end;
# Process the source code and print the generated code.
proc _compile();
begin
.compile_loop:
_skip_newlines();
la t0, source_code_position
lw t0, (t0)
lb t0, (t0)
beqz t0, .compile_end
li t1, '#'
beq t0, t1, .compile_comment
la a0, source_code_position
lw a0, (a0)
la a1, keyword_equ
li a2, KEYWORD_EQU_SIZE
_memcmp();
beqz a0, .compile_equ
la a0, source_code_position
lw a0, (a0)
la a1, keyword_section
li a2, KEYWORD_SECTION_SIZE
_memcmp();
beqz a0, .compile_section
la a0, source_code_position
lw a0, (a0)
la a1, keyword_type
li a2, KEYWORD_TYPE_SIZE
_memcmp();
beqz a0, .compile_type
la a0, source_code_position
lw a0, (a0)
la a1, keyword_proc
li a2, KEYWORD_PROC_SIZE
_memcmp();
beqz a0, .compile_procedure
la a0, source_code_position
lw a0, (a0)
la a1, keyword_global
li a2, KEYWORD_GLOBAL_SIZE
_memcmp();
beqz a0, .compile_global
# Not a known token, exit.
j .compile_end
.compile_equ:
_compile_equ();
j .compile_loop
.compile_section:
_compile_section();
j .compile_loop
.compile_type:
_compile_type();
j .compile_loop
.compile_global:
_compile_line();
j .compile_loop
.compile_comment:
_skip_comment();
j .compile_loop
.compile_procedure:
_compile_procedure();
j .compile_loop
.compile_end:
end;
# Entry point.
.globl _start
proc _start();
begin
# Read the source from the standard input.
la a0, source_code
li a1, SOURCE_BUFFER_SIZE # Buffer size.
_read_file();
_compile();
# Call exit.
li a0, 0 # Use 0 return code.
li a7, 93 # SYS_EXIT.
ecall
end;