diff --git a/Gemfile b/Gemfile
deleted file mode 100644
index 97bdde4..0000000
--- a/Gemfile
+++ /dev/null
@@ -1,9 +0,0 @@
-# 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/.
-# frozen_string_literal: true
-
-source 'https://rubygems.org'
-
-gem 'term-ansicolor', '~> 1.2'
-gem 'rake', '~> 13.2'
diff --git a/Gemfile.lock b/Gemfile.lock
deleted file mode 100644
index 3521a74..0000000
--- a/Gemfile.lock
+++ /dev/null
@@ -1,22 +0,0 @@
-GEM
-  remote: https://rubygems.org/
-  specs:
-    bigdecimal (3.1.9)
-    rake (13.2.1)
-    sync (0.5.0)
-    term-ansicolor (1.11.2)
-      tins (~> 1.0)
-    tins (1.38.0)
-      bigdecimal
-      sync
-
-PLATFORMS
-  ruby
-  x86_64-linux
-
-DEPENDENCIES
-  rake (~> 13.2)
-  term-ansicolor (~> 1.2)
-
-BUNDLED WITH
-   2.6.7
diff --git a/Rakefile b/Rakefile
index e6dc163..2b61d70 100644
--- a/Rakefile
+++ b/Rakefile
@@ -6,8 +6,6 @@
 require 'open3'
 require 'rake/clean'
 
-SYSROOT = '../eugenios/build/sysroot'
-QEMU = 'qemu-riscv32'
 STAGES = Dir.glob('boot/stage*.elna').collect { |stage| File.basename stage, '.elna' }.sort
 
 CLEAN.include 'build/boot', 'build/valid'
@@ -19,6 +17,10 @@ def compile(input, output)
   sh ENV.fetch('CC', 'gcc'), '-nostdlib', '-o', output, input
 end
 
+def run(exe)
+  ENV.fetch('QEMU', '').split << exe
+end
+
 task default: :boot
 
 desc 'Final stage'
@@ -31,9 +33,8 @@ task boot: "boot/#{STAGES.last}.elna" do |t|
   source = groupped['.elna']
 
   cat_arguments = ['cat', source]
-  compiler_arguments = [QEMU, '-L', SYSROOT, exe]
   diff_arguments = ['diff', '-Nur', '--text', expected, '-']
-  Open3.pipeline(cat_arguments, compiler_arguments, diff_arguments)
+  Open3.pipeline(cat_arguments, run(exe), diff_arguments)
 end
 
 desc 'Convert previous stage language into the current stage language'
@@ -60,8 +61,7 @@ STAGES.each do |stage|
     exe, source = t.prerequisites
 
     cat_arguments = ['cat', source]
-    compiler_arguments = [QEMU, '-L', SYSROOT, exe]
-    last_stdout, wait_threads = Open3.pipeline_r(cat_arguments, compiler_arguments)
+    last_stdout, wait_threads = Open3.pipeline_r(cat_arguments, run(exe))
 
     IO.copy_stream last_stdout, t.name
   end
@@ -70,8 +70,7 @@ STAGES.each do |stage|
     exe, source = t.prerequisites
 
     cat_arguments = ['cat', source]
-    compiler_arguments = [QEMU, '-L', SYSROOT, exe]
-    last_stdout, wait_threads = Open3.pipeline_r(cat_arguments, compiler_arguments)
+    last_stdout, wait_threads = Open3.pipeline_r(cat_arguments, run(exe))
 
     IO.copy_stream last_stdout, t.name
   end
@@ -85,8 +84,7 @@ file 'build/valid/stage1.s' => ['build/boot/stage1', 'boot/stage1.s', 'build/val
   source, exe, = t.prerequisites.partition { |prerequisite| prerequisite.end_with? '.s' }
 
   cat_arguments = ['cat', *source]
-  compiler_arguments = [QEMU, '-L', SYSROOT, *exe]
-  last_stdout, wait_threads = Open3.pipeline_r(cat_arguments, compiler_arguments)
+  last_stdout, wait_threads = Open3.pipeline_r(cat_arguments, run(exe.first))
 
   IO.copy_stream last_stdout, t.name
 end
diff --git a/boot/stage1.s b/boot/stage1.s
index b4dbfed..870952b 100644
--- a/boot/stage1.s
+++ b/boot/stage1.s
@@ -2,7 +2,7 @@
 # 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/.
 
-# Stage1 compiler.
+# Stage 1 compiler.
 #
 # It supports declaring and calling procedures without arguments.
 # A procedure name should start with an underscore.
diff --git a/boot/stage2.elna b/boot/stage2.elna
index ce9de8a..0423b3b 100644
--- a/boot/stage2.elna
+++ b/boot/stage2.elna
@@ -2,7 +2,7 @@
 # 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.
+# Stage 2 compiler.
 #
 # - Procedures without none or one argument.
 # - Goto statements.
diff --git a/boot/stage3.elna b/boot/stage3.elna
index 2c79295..aec9832 100644
--- a/boot/stage3.elna
+++ b/boot/stage3.elna
@@ -2,7 +2,7 @@
 # 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/.
 
-# Stage3 compiler.
+# Stage 3 compiler.
 #
 # - Procedures with multiple arguments.
 # - Character literals with and without escaping.
diff --git a/boot/stage4.elna b/boot/stage4.elna
index 1d98c3a..d873b9a 100644
--- a/boot/stage4.elna
+++ b/boot/stage4.elna
@@ -2,7 +2,7 @@
 # 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/.
 
-# Stage4 compiler.
+# Stage 4 compiler.
 #
 # - Taking value of local and global variables. Variables that doesn't begin
 #   with "v" are considered global.
diff --git a/boot/stage5.elna b/boot/stage5.elna
index 88b407f..69623db 100644
--- a/boot/stage5.elna
+++ b/boot/stage5.elna
@@ -2,7 +2,7 @@
 # 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/.
 
-# Stage5 compiler.
+# Stage 5 compiler.
 #
 # - Stack size increased to 128 bytes per procedure.
 #   7 words on the stack, 92 - 120, are reversed for procedure arguments (caller side).
diff --git a/boot/stage6.elna b/boot/stage6.elna
index f412cc6..7d426f9 100644
--- a/boot/stage6.elna
+++ b/boot/stage6.elna
@@ -1,9 +1,10 @@
-# ihis Source Code Form is subject to the terms of the Mozilla Public License,
+# 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/.
 
-# Stage6 compiler.
+# Stage 6 compiler.
 #
+# - String literals.
 
 .section .rodata
 
@@ -43,6 +44,9 @@ asm_type_directive: .string ".type "
 .type asm_type_function, @object
 asm_type_function: .string ", @function\n"
 
+.type asm_type_object, @object
+asm_type_object: .string ", @object\n"
+
 .type asm_colon, @object
 asm_colon: .string ":\n"
 
@@ -112,19 +116,110 @@ asm_comma: .string ", "
 .type asm_sp, @object
 asm_sp: .string "sp"
 
+.type asm_rodata, @object
+asm_rodata: .string ".section .rodata\n"
+
+.type asm_strings, @object
+asm_strings: .string "strings"
+
+.type asm_ascii, @object
+asm_ascii: .string " .ascii "
+
 .section .bss
 
 # When modifiying also change the read size in the entry point procedure.
 .type source_code, @object
 source_code: .zero 81920
 
+.type compiler_strings, @object
+compiler_strings: .zero 8192
+
 .section .data
 
+.type compiler_strings_position, @object
+compiler_strings_position: .word compiler_strings
+
+.type compiler_strings_length, @object
+compiler_strings_length: .word 0
+
 .type source_code_position, @object
 source_code_position: .word source_code
 
 .section .text
 
+# Calculates and returns the string token length between quotes, including the
+# escaping slash characters.
+#
+# Parameters:
+# a0 - String token pointer.
+#
+# Returns the length in a0.
+proc _string_length();
+begin
+	# Reset the counter.
+	v0 := 0;
+
+.string_length_loop:
+	v88 := v88 + 1;
+
+	lw t0, 88(sp)
+	lb t0, (t0)
+
+	li t1, '"'
+	beq t0, t1, .string_length_end
+
+	v0 := v0 + 1;
+	goto .string_length_loop;
+
+.string_length_end:
+	return v0
+end;
+
+# Adds a string to the global, read-only string storage.
+#
+# Parameters:
+# a0 - String token.
+#
+# Returns the offset from the beginning of the storage to the new string in a0.
+proc _add_string();
+begin
+	v0 := v88 + 1;
+	v4 := compiler_strings_length;
+
+.add_string_loop:
+	lw t0, 0(sp)
+	lb t1, (t0)
+	li t2, '"'
+
+	beq t1, t2, .add_string_end
+
+	la t2, compiler_strings_position
+	lw t3, (t2)
+	sb t1, (t3)
+
+	addi t3, t3, 1
+	sw t3, (t2)
+
+	addi t0, t0, 1
+	sw t0, 0(sp)
+
+	li t2, '\\'
+	bne t1, t2, .add_string_increment
+
+	goto .add_string_loop;
+
+.add_string_increment:
+	la t2, compiler_strings_length
+	lw t4, (t2)
+	addi t4, t4, 1
+	sw t4, (t2)
+
+	goto .add_string_loop;
+
+.add_string_end:
+	return v4
+end;
+
 # Reads standard input into a buffer.
 # a0 - Buffer pointer.
 # a1 - Buffer size.
@@ -595,6 +690,37 @@ begin
 
 end;
 
+proc _compile_string_literal();
+begin
+	_string_length(source_code_position);
+	sw a0, 0(sp)
+
+	_add_string(source_code_position);
+	sw a0, 4(sp)
+
+	_advance_token(v0 + 2);
+
+	_write_z(@asm_la);
+	_write_register('t', 0);
+	_write_z(@asm_comma);
+	_write_z(@asm_strings);
+	_write_c('\n');
+
+	_write_z(@asm_li);
+	_write_register('t', 1);
+	_write_z(@asm_comma);
+	_write_i(v4);
+	_write_c('\n');
+
+	_write_z(@asm_add);
+	_write_register('t', 0);
+	_write_z(@asm_comma);
+	_write_register('t', 0);
+	_write_z(@asm_comma);
+	_write_register('t', 1);
+	_write_c('\n');
+end;
+
 proc _compile_term();
 begin
 	la t0, source_code_position
@@ -614,6 +740,9 @@ begin
 	li t1, '~'
 	beq a0, t1, .compile_term_not
 
+	li t1, '"'
+	beq a0, t1, .compile_term_string_literal
+
 	_is_digit(v0);
 	bnez a0, .compile_term_integer_literal
 
@@ -639,6 +768,10 @@ begin
 	_compile_not_expression();
 	goto .compile_term_end;
 
+.compile_term_string_literal:
+	_compile_string_literal();
+	goto .compile_term_end;
+
 .compile_term_variable:
 	_compile_variable_expression();
 	goto .compile_term_end;
@@ -1336,63 +1469,99 @@ begin
 end;
 
 # Process the source code and print the generated code.
-proc _compile();
+proc _compile_module();
 begin
-.compile_loop:
+.compile_module_loop:
 	_skip_newlines();
 
 	la t0, source_code_position
 	lw t0, (t0)
 	lb t0, (t0)
-	beqz t0, .compile_end
+	beqz t0, .compile_module_end
 	li t1, '#'
-	beq t0, t1, .compile_comment
+	beq t0, t1, .compile_module_comment
 
 	# 8 is ".section" length.
 	_memcmp(source_code_position, @keyword_section, 8);
-	beqz a0, .compile_section
+	beqz a0, .compile_module_section
 
 	# 5 is ".type" length.
 	_memcmp(source_code_position, @keyword_type, 5);
-	beqz a0, .compile_type
+	beqz a0, .compile_module_type
 
 	# 5 is "proc " length. Space is needed to distinguish from "procedure".
 	_memcmp(source_code_position, @keyword_proc, 5);
-	beqz a0, .compile_procedure
+	beqz a0, .compile_module_procedure
 
 	# 6 is ".globl" length.
 	_memcmp(source_code_position, @keyword_global, 6);
-	beqz a0, .compile_global
+	beqz a0, .compile_module_global
 
 	# Not a known token, exit.
-	goto .compile_end;
+	goto .compile_module_end;
 
-.compile_section:
+.compile_module_section:
 	_compile_section();
 
-	goto .compile_loop;
+	goto .compile_module_loop;
 
-.compile_type:
+.compile_module_type:
 	_compile_type();
 
-	goto .compile_loop;
+	goto .compile_module_loop;
 
-.compile_global:
+.compile_module_global:
 	_compile_line();
 
-	goto .compile_loop;
+	goto .compile_module_loop;
 
-.compile_comment:
+.compile_module_comment:
 	_skip_comment();
 
-	goto .compile_loop;
+	goto .compile_module_loop;
 
-.compile_procedure:
+.compile_module_procedure:
 	_compile_procedure();
 
-	goto .compile_loop;
+	goto .compile_module_loop;
+
+.compile_module_end:
+end;
+
+proc _compile();
+begin
+	_compile_module();
+
+	_write_z(@asm_rodata);
+	_write_z(@asm_type_directive);
+	_write_z(@asm_strings);
+	_write_z(@asm_type_object);
+	_write_z(@asm_strings);
+	_write_c(':');
+	_write_z(@asm_ascii);
+	_write_c('"');
+
+	la t0, compiler_strings
+	sw t0, 0(sp)
+
+.compile_loop:
+	lw t0, 0(sp)
+	la t1, compiler_strings_position
+	lw t1, (t1)
+	bge t0, t1, .compile_end
+
+	lb a0, (t0)
+
+	addi t0, t0, 1
+	sw t0, 0(sp)
+
+	_write_c();
+
+	j .compile_loop
 
 .compile_end:
+	_write_c('"');
+	_write_c('\n');
 end;
 
 # Terminates the program. a0 contains the return code.
diff --git a/boot/stage7.elna b/boot/stage7.elna
new file mode 100644
index 0000000..6973963
--- /dev/null
+++ b/boot/stage7.elna
@@ -0,0 +1,1260 @@
+# 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/.
+
+# Stage 7 compiler.
+#
+# - String literals.
+
+.section .bss
+
+# When modifiying also change the read size in the entry point procedure.
+.type source_code, @object
+source_code: .zero 81920
+
+.type compiler_strings, @object
+compiler_strings: .zero 8192
+
+.section .data
+
+.type compiler_strings_position, @object
+compiler_strings_position: .word compiler_strings
+
+.type compiler_strings_length, @object
+compiler_strings_length: .word 0
+
+.type source_code_position, @object
+source_code_position: .word source_code
+
+.section .text
+
+# Calculates and returns the string token length between quotes, including the
+# escaping slash characters.
+#
+# Parameters:
+# a0 - String token pointer.
+#
+# Returns the length in a0.
+proc _string_length();
+begin
+	# Reset the counter.
+	v0 := 0;
+
+.string_length_loop:
+	v88 := v88 + 1;
+
+	lw t0, 88(sp)
+	lb t0, (t0)
+
+	li t1, '"'
+	beq t0, t1, .string_length_end
+
+	v0 := v0 + 1;
+	goto .string_length_loop;
+
+.string_length_end:
+	return v0
+end;
+
+# Adds a string to the global, read-only string storage.
+#
+# Parameters:
+# a0 - String token.
+#
+# Returns the offset from the beginning of the storage to the new string in a0.
+proc _add_string();
+begin
+	v0 := v88 + 1;
+	v4 := compiler_strings_length;
+
+.add_string_loop:
+	lw t0, 0(sp)
+	lb t1, (t0)
+	li t2, '"'
+
+	beq t1, t2, .add_string_end
+
+	la t2, compiler_strings_position
+	lw t3, (t2)
+	sb t1, (t3)
+
+	addi t3, t3, 1
+	sw t3, (t2)
+
+	addi t0, t0, 1
+	sw t0, 0(sp)
+
+	li t2, '\\'
+	bne t1, t2, .add_string_increment
+
+	goto .add_string_loop;
+
+.add_string_increment:
+	la t2, compiler_strings_length
+	lw t4, (t2)
+	addi t4, t4, 1
+	sw t4, (t2)
+
+	goto .add_string_loop;
+
+.add_string_end:
+	return v4
+end;
+
+# 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_s();
+begin
+	mv a2, a1
+	mv a1, a0
+	# STDOUT.
+	li a0, 1
+	li a7, 64 # SYS_WRITE.
+	ecall
+end;
+
+# Writes a number to a string buffer.
+#
+# t0 - Local buffer.
+# t1 - Constant 10.
+# t2 - Current character.
+# t3 - Whether the number is negative.
+#
+# Parameters:
+# a0 - Whole number.
+# a1 - Buffer pointer.
+#
+# Sets a0 to the length of the written number.
+proc _print_i();
+begin
+	li t1, 10
+	addi t0, s0, -9
+
+	li t3, 0
+	bgez a0, .print_i_digit10
+	li t3, 1
+	neg a0, a0
+
+.print_i_digit10:
+	rem t2, a0, t1
+	addi t2, t2, '0'
+	sb t2, 0(t0)
+	div a0, a0, t1
+	addi t0, t0, -1
+	bne zero, a0, .print_i_digit10
+
+	beq zero, t3, .print_i_write_call
+	addi t2, zero, '-'
+	sb t2, 0(t0)
+	addi t0, t0, -1
+
+.print_i_write_call:
+	mv a0, a1
+	addi a1, t0, 1
+	sub a2, s0, t0
+	addi a2, a2, -9
+	sw a2, 0(sp)
+
+	_memcpy();
+
+	return v0
+end;
+
+# Writes a number to the standard output.
+#
+# Parameters:
+# a0 - Whole number.
+proc _write_i();
+begin
+	_print_i(v88, @v0);
+
+	mv a1, a0
+	addi a0, sp, 0
+	_write_s();
+
+end;
+
+# Writes a character from a0 into the standard output.
+proc _write_c();
+begin
+	_write_s(@v88, 1);
+end;
+
+# Write null terminated string.
+#
+# Parameters:
+# a0 - String.
+proc _write_z();
+begin
+.write_z_loop:
+	# Check for 0 character.
+	lw a0, 88(sp)
+	lb a0, (a0)
+	beqz a0, .write_z_end
+
+	# Print a character.
+	_write_c();
+
+	# Advance the input string by one byte.
+	v88 := v88 + 1;
+
+	goto .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
+	v0 := v88 >= 'A';
+	v4 := v88 <= 'Z';
+
+	return v0 & v4
+
+end;
+
+# Detects if a0 is an lowercase character. Sets a0 to 1 if so, otherwise to 0.
+proc _is_lower();
+begin
+	v0 := v88 >= 'a';
+	v4 := v88 <= 'z';
+
+	return v0 & v4
+
+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, 0(sp)
+
+	_is_upper();
+	sw a0, 4(sp)
+
+	_is_lower(v0);
+
+	lw t0, 0(sp)
+	xori t1, t0, '_'
+	seqz t1, t1
+
+	lw t0, 4(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
+	v0 := v88 >= '0';
+	v4 := v88 <= '9';
+
+	return v0 & v4
+end;
+
+proc _is_alnum();
+begin
+	sw a0, 4(sp)
+
+	_is_alpha();
+	sw a0, 0(sp)
+
+	_is_digit(v4);
+
+	lw a1, 0(sp)
+	or a0, a0, a1
+end;
+
+# Reads the next token.
+#
+# Returns token length in a0.
+proc _read_token();
+begin
+	# Current token position.
+	v0 := source_code_position;
+	# Token length.
+	v4 := 0;
+
+.read_token_loop:
+	lw t0, 0(sp)
+	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, 0(sp)
+	lb a0, (a0)
+	_is_alnum();
+	bnez a0, .read_token_next
+
+	goto .read_token_end;
+
+.read_token_next:
+	# Advance the source code position and token length.
+	v4 := v4 + 1;
+	v0 := v0 + 1;
+
+	goto .read_token_loop;
+
+.read_token_end:
+	return v4
+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
+
+.memcmp_loop:
+	beqz a2, .memcmp_end
+
+	lbu t1, (t0)
+	lbu t2, (a1)
+	sub a0, t1, t2
+
+	bnez a0, .memcmp_end
+
+	addi t0, t0, 1
+	addi a1, a1, 1
+	addi a2, a2, -1
+
+	goto .memcmp_loop;
+
+.memcmp_end:
+end;
+
+# Copies memory.
+#
+# Parameters:
+# a0 - Destination.
+# a1 - Source.
+# a2 - Size.
+#
+# Preserves a0.
+proc _memcpy();
+begin
+	mv t0, a0
+
+.memcpy_loop:
+	beqz a2, .memcpy_end
+
+	lbu t1, (a1)
+	sb t1, (a0)
+
+	addi a0, a0, 1
+	addi a1, a1, 1
+	addi a2, a2, -1
+
+	goto .memcpy_loop
+
+.memcpy_end:
+	mv a0, t0
+end;
+
+# Advances the token stream by a0 bytes.
+proc _advance_token();
+begin
+	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
+	_write_s(source_code_position, v88);
+	return v88
+end;
+
+proc _compile_section();
+begin
+	# Print and skip the ".section" (8 characters) directive and a space after it.
+	_write_token(9);
+	_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)
+
+	goto .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.
+	_advance_token(1);
+
+	goto .compile_line_loop;
+
+.compile_line_end:
+	_write_c('\n');
+
+	_advance_token(1);
+end;
+
+proc _compile_integer_literal();
+begin
+	_write_z("\tli t0, \0");
+
+	_read_token();
+	_write_token();
+	_advance_token();
+
+	_write_c('\n');
+end;
+
+proc _compile_character_literal();
+begin
+	_write_z("\tli t0, \0");
+
+	_write_c('\'');
+	_advance_token(1);
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+	li t1, '\\'
+	bne a0, t1, .compile_character_literal_end
+	
+	_write_c('\\');
+	_advance_token(1);
+
+.compile_character_literal_end:
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+	_write_c();
+
+	_write_c('\'');
+	_write_c('\n');
+
+	_advance_token(2);
+
+end;
+
+proc _compile_variable_expression();
+begin
+	_compile_designator();
+	_write_z("\tlw t0, (t0)\n\0");
+end;
+
+proc _compile_address_expression();
+begin
+	# Skip the "@" sign.
+	_advance_token(1);
+	_compile_designator();
+
+end;
+
+proc _compile_negate_expression();
+begin
+	# Skip the "-" sign.
+	_advance_token(1);
+	_compile_term();
+
+	_write_z("\tneg t0, t0\n\0");
+end;
+
+proc _compile_not_expression();
+begin
+	# Skip the "~" sign.
+	_advance_token(1);
+	_compile_term();
+
+	_write_z("\tnot t0, t0\n\0");
+end;
+
+proc _compile_string_literal();
+begin
+	_string_length(source_code_position);
+	sw a0, 0(sp)
+
+	_add_string(source_code_position);
+	sw a0, 4(sp)
+
+	_advance_token(v0 + 2);
+	_write_z("\tla t0, strings\n\0");
+
+	_write_z("\tli t1, \0");
+	_write_i(v4);
+	_write_c('\n');
+
+	_write_z("\tadd t0, t0, t1\n\0");
+end;
+
+proc _compile_term();
+begin
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+	sw a0, 0(sp)
+
+	li t1, '\''
+	beq a0, t1, .compile_term_character_literal
+
+	li t1, '@'
+	beq a0, t1, .compile_term_address
+
+	li t1, '-'
+	beq a0, t1, .compile_term_negation
+
+	li t1, '~'
+	beq a0, t1, .compile_term_not
+
+	li t1, '"'
+	beq a0, t1, .compile_term_string_literal
+
+	_is_digit(v0);
+	bnez a0, .compile_term_integer_literal
+
+	goto .compile_term_variable;
+
+.compile_term_character_literal:
+	_compile_character_literal();
+	goto .compile_term_end;
+
+.compile_term_integer_literal:
+	_compile_integer_literal();
+	goto .compile_term_end;
+
+.compile_term_address:
+	_compile_address_expression();
+	goto .compile_term_end;
+
+.compile_term_negation:
+	_compile_negate_expression();
+	goto .compile_term_end;
+
+.compile_term_not:
+	_compile_not_expression();
+	goto .compile_term_end;
+
+.compile_term_string_literal:
+	_compile_string_literal();
+	goto .compile_term_end;
+
+.compile_term_variable:
+	_compile_variable_expression();
+	goto .compile_term_end;
+
+.compile_term_end:
+end;
+
+proc _compile_binary_rhs();
+begin
+	# Skip the whitespace after the binary operator.
+	_advance_token(1);
+	_compile_term();
+
+	# Load the left expression from the stack;
+	_write_z("\tlw t1, 24(sp)\n\0");
+end;
+
+proc _compile_expression();
+begin
+	_compile_term();
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+
+	li t1, ' '
+	bne a0, t1, .compile_expression_end
+
+	# It is a binary expression.
+
+	# Save the value of the left expression on the stack.
+	_write_z("sw t0, 24(sp)\n\0");
+
+	# Skip surrounding whitespace in front of the operator.
+	_advance_token(1);
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, '+'
+	beq t0, t1, .compile_expression_add
+
+	li t1, '*'
+	beq t0, t1, .compile_expression_mul
+
+	li t1, '&'
+	beq t0, t1, .compile_expression_and
+
+	li t1, 'o'
+	beq t0, t1, .compile_expression_or
+
+	li t1, 'x'
+	beq t0, t1, .compile_expression_xor
+
+	li t1, '='
+	beq t0, t1, .compile_expression_equals
+
+	li t1, '<'
+	beq t0, t1, .compile_expression_less
+
+	li t1, '>'
+	beq t0, t1, .compile_expression_greater
+
+	# Unknown binary operator.
+	unimp
+
+.compile_expression_add:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("add t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_mul:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tmul t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_and:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tand t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_or:
+	_advance_token(2);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("or t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_xor:
+	_advance_token(3);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("xor t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_equals:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("xor t0, t0, t1\nseqz t0, t0\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_less:
+	_advance_token(1);
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, '>'
+	beq t0, t1, .compile_expression_not_equal
+
+	li t1, '='
+	beq t0, t1, .compile_expression_less_equal
+
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("slt t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_not_equal:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\txor t0, t0, t1\nsnez t0, t0\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_less_equal:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tslt t0, t0, t1\nxori t0, t0, 1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_greater:
+	_advance_token(1);
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, '='
+	beq t0, t1, .compile_expression_greater_equal
+
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tslt t0, t1, t0\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_greater_equal:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tslt t0, t1, t0\nxori t0, t0, 1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_end:
+end;
+
+proc _compile_call();
+begin
+	# Stack variables:
+	# v0 - Procedure name length.
+	# v4 - Procedure name pointer. 
+	# v8 - Argument count.
+
+	_read_token();
+	sw a0, 0(sp)
+	v4 := source_code_position;
+	v8 := 0;
+
+	# Skip the identifier and left paren.
+	_advance_token(v0 + 1);
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, ')'
+	beq t0, t1, .compile_call_finalize
+
+.compile_call_loop:
+	_compile_expression();
+
+	# Save the argument on the stack.
+	_write_z("\tsw t0, \0");
+
+	# Calculate the stack offset: 116 - (4 * argument_counter)
+	v12 := v8 * 4;
+	v12 := 116 + -v12;
+	_write_i(v12);
+
+	_write_z("(sp)\n\0");
+
+	# Add one to the argument counter.
+	v8 := v8 + 1;
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, ','
+	bne t0, t1, .compile_call_finalize
+
+	_advance_token(2);
+	goto .compile_call_loop;
+
+.compile_call_finalize:
+	# Load the argument from the stack.
+
+	lw t0, 8(sp)
+	beqz t0, .compile_call_end
+
+	# Decrement the argument counter.
+	v8 := v8 + -1;
+
+	_write_z("\tlw a\0");
+	_write_i(v8);
+
+	_write_z(", \0");
+
+	# Calculate the stack offset: 116 - (4 * argument_counter)
+	v12 := v8 * 4;
+	v12 := 116 + -v12;
+	_write_i(v12);
+
+	_write_z("(sp)\n\0");
+
+	goto .compile_call_finalize;
+
+.compile_call_end:
+	_write_z("\tcall \0");
+	_write_s(v4, v0);
+
+	# Skip the right paren.
+	_advance_token(1);
+end;
+
+proc _compile_goto();
+begin
+	_advance_token(5);
+
+	_read_token();
+	sw a0, 0(sp)
+
+	_write_z("\tj \0");
+
+	_write_token(v0);
+	_advance_token();
+end;
+
+proc _compile_local_designator();
+begin
+	# Skip "v" in the local variable name.
+	_advance_token(1);
+	_write_z("\t addi t0, sp, \0");
+
+	# Read local variable stack offset and save it.
+	_read_token();
+	_write_token();
+	_advance_token();
+	_write_c('\n');
+end;
+
+proc _compile_global_designator();
+begin
+	_write_z("\tla t0, \0");
+
+	_read_token();
+	_write_token();
+	_advance_token();
+
+	_write_c('\n');
+end;
+
+proc _compile_designator();
+begin
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+
+	li t1, 'v'
+	beq a0, t1, .compile_designator_local
+
+	goto .compile_designator_global;
+
+.compile_designator_local:
+	_compile_local_designator();
+	goto .compile_designator_end;
+
+.compile_designator_global:
+	_compile_global_designator();
+	goto .compile_designator_end;
+
+.compile_designator_end:
+end;
+
+proc _compile_assignment();
+begin
+	_compile_designator();
+
+	# Save the assignee address on the stack.
+	_write_z("\tsw t0, 20(sp)\n\0");
+
+	# Skip the assignment sign (:=) with surrounding whitespaces.
+	_advance_token(4);
+
+	# Compile the assignment.
+	_compile_expression();
+
+	_write_z("\tlw t1, 20(sp)\nsw t0, (t1)\n\0");
+end;
+
+proc _compile_return_statement();
+begin
+	# Skip "return" keyword and whitespace after it.
+	_advance_token(7);
+	_compile_expression();
+
+	_write_z("mv a0, t0\n\0");
+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
+
+	li t1, 'v'
+	beq t0, t1, .compile_statement_assignment
+
+	# keyword_ret contains "\tret", so it's 4 bytes long.
+	_memcmp(source_code_position, "\treturn", 7);
+	beqz a0, .compile_statement_return
+
+	_compile_line();
+	goto .compile_statement_end;
+
+.compile_statement_call:
+	_advance_token(1);
+	_compile_call();
+
+	goto .compile_statement_semicolon;
+
+.compile_statement_goto:
+	_advance_token(1);
+	_compile_goto();
+
+	goto .compile_statement_semicolon;
+
+.compile_statement_assignment:
+	_advance_token(1);
+	_compile_assignment();
+
+	goto .compile_statement_semicolon;
+
+.compile_statement_return:
+	_advance_token(1);
+	_compile_return_statement();
+	_write_c('\n');
+
+	goto .compile_statement_end;
+
+.compile_statement_semicolon:
+	_advance_token(2);
+	_write_c('\n');
+
+.compile_statement_end:
+end;
+
+proc _compile_procedure_body();
+begin
+.compile_procedure_body_loop:
+	# 3 is "end" length.
+	_memcmp(source_code_position, "end", 3);
+
+	beqz a0, .compile_procedure_body_epilogue
+
+	_compile_statement();
+	goto .compile_procedure_body_loop;
+
+.compile_procedure_body_epilogue:
+end;
+
+# Writes a regster name to the standard output.
+#
+# Parameters:
+# a0 - Register character.
+# a1 - Register number.
+proc _write_register();
+begin
+	_write_c(v88);
+	v84 := v84 + '0';
+	_write_c(v84);
+end;
+
+proc _compile_procedure_prologue();
+begin
+	_write_z("\taddi sp, sp, -128\n\tsw ra, 124(sp)\n\tsw s0, 120(sp)\n\taddi s0, sp, 128\n\0");
+	v0 := 0;
+
+.compile_procedure_prologue_loop:
+	_write_z("\tsw a\0");
+	_write_i(v0);
+	_write_z(", \0");
+
+	# Calculate the stack offset: 88 - (4 * parameter_counter)
+	v4 := v0 * 4;
+	v4 := 88 + -v4;
+	_write_i(v4);
+
+	_write_z("(sp)\n\0");
+
+	v0 := v0 + 1;
+	lw a0, 0(sp)
+
+	li t0, 8
+	bne a0, t0, .compile_procedure_prologue_loop
+end;
+
+proc _compile_procedure();
+begin
+	# Skip "proc ".
+	_advance_token(5);
+
+	_read_token();
+	sw a0, 0(sp) # Save the procedure name length.
+
+	# Write .type _procedure_name, @function.
+	_write_z(".type \0");
+
+	_write_token(v0);
+	_write_z(", @function\n\0");
+
+	# Write procedure label, _procedure_name:
+	_write_token(v0);
+	_write_z(":\n\0");
+
+	# Skip the function name and trailing parens, semicolon, "begin" and newline.
+	_advance_token(v0 + 10);
+
+	_compile_procedure_prologue();
+	_compile_procedure_body();
+
+	# Write the epilogue.
+	_write_z("\tlw ra, 124(sp)\n\tlw s0, 120(sp)\n\taddi sp, sp, 128\n\tret\n\0");
+
+	# Skip the "end" keyword, semicolon and newline.
+	_advance_token(5);
+end;
+
+proc _compile_type();
+begin
+	# Print and skip the ".type" (5 characters) directive and a space after it.
+	_write_token(6);
+	_advance_token();
+
+	# Read and print the symbol name.
+	_read_token();
+
+	# Print and skip the symbol name, comma, space and @.
+	addi a0, a0, 3
+	_write_token();
+	_advance_token();
+
+	# Read the symbol type.
+	_read_token();
+
+	# Print the symbol type and newline.
+	addi a0, a0, 1
+	_write_token();
+	_advance_token();
+
+	# Write the object definition itself.
+	_compile_line();
+
+.compile_type_end:
+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)
+
+	goto .skip_newlines_loop;
+
+.skip_newlines_end:
+end;
+
+# Process the source code and print the generated code.
+proc _compile_module();
+begin
+.compile_module_loop:
+	_skip_newlines();
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+	beqz t0, .compile_module_end
+	li t1, '#'
+	beq t0, t1, .compile_module_comment
+
+	# 8 is ".section" length.
+	_memcmp(source_code_position, ".section", 8);
+	beqz a0, .compile_module_section
+
+	# 5 is ".type" length.
+	_memcmp(source_code_position, ".type", 5);
+	beqz a0, .compile_module_type
+
+	# 5 is "proc " length. Space is needed to distinguish from "procedure".
+	_memcmp(source_code_position, "proc ", 5);
+	beqz a0, .compile_module_procedure
+
+	# 6 is ".globl" length.
+	_memcmp(source_code_position, ".globl", 6);
+	beqz a0, .compile_module_global
+
+	# Not a known token, exit.
+	goto .compile_module_end;
+
+.compile_module_section:
+	_compile_section();
+
+	goto .compile_module_loop;
+
+.compile_module_type:
+	_compile_type();
+
+	goto .compile_module_loop;
+
+.compile_module_global:
+	_compile_line();
+
+	goto .compile_module_loop;
+
+.compile_module_comment:
+	_skip_comment();
+
+	goto .compile_module_loop;
+
+.compile_module_procedure:
+	_compile_procedure();
+
+	goto .compile_module_loop;
+
+.compile_module_end:
+end;
+
+proc _compile();
+begin
+	_compile_module();
+
+	_write_z(".section .rodata\n.type strings, @object\nstrings: .ascii \0");
+	_write_c('"');
+
+	la t0, compiler_strings
+	sw t0, 0(sp)
+
+.compile_loop:
+	lw t0, 0(sp)
+	la t1, compiler_strings_position
+	lw t1, (t1)
+	bge t0, t1, .compile_end
+
+	lb a0, (t0)
+
+	addi t0, t0, 1
+	sw t0, 0(sp)
+
+	_write_c();
+
+	j .compile_loop
+
+.compile_end:
+	_write_c('"');
+	_write_c('\n');
+end;
+
+# Terminates the program. a0 contains the return code.
+#
+# Parameters:
+# a0 - Status code.
+proc _exit();
+begin
+	li a7, 93 # SYS_EXIT
+	ecall
+end;
+
+# Entry point.
+.globl _start
+proc _start();
+begin
+	# Read the source from the standard input.
+	# Second argument is buffer size. Modifying update the source_code definition.
+	_read_file(@source_code, 81920);
+	_compile();
+
+	_exit(0);
+
+end;
diff --git a/boot/stage8.elna b/boot/stage8.elna
new file mode 100644
index 0000000..6973963
--- /dev/null
+++ b/boot/stage8.elna
@@ -0,0 +1,1260 @@
+# 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/.
+
+# Stage 7 compiler.
+#
+# - String literals.
+
+.section .bss
+
+# When modifiying also change the read size in the entry point procedure.
+.type source_code, @object
+source_code: .zero 81920
+
+.type compiler_strings, @object
+compiler_strings: .zero 8192
+
+.section .data
+
+.type compiler_strings_position, @object
+compiler_strings_position: .word compiler_strings
+
+.type compiler_strings_length, @object
+compiler_strings_length: .word 0
+
+.type source_code_position, @object
+source_code_position: .word source_code
+
+.section .text
+
+# Calculates and returns the string token length between quotes, including the
+# escaping slash characters.
+#
+# Parameters:
+# a0 - String token pointer.
+#
+# Returns the length in a0.
+proc _string_length();
+begin
+	# Reset the counter.
+	v0 := 0;
+
+.string_length_loop:
+	v88 := v88 + 1;
+
+	lw t0, 88(sp)
+	lb t0, (t0)
+
+	li t1, '"'
+	beq t0, t1, .string_length_end
+
+	v0 := v0 + 1;
+	goto .string_length_loop;
+
+.string_length_end:
+	return v0
+end;
+
+# Adds a string to the global, read-only string storage.
+#
+# Parameters:
+# a0 - String token.
+#
+# Returns the offset from the beginning of the storage to the new string in a0.
+proc _add_string();
+begin
+	v0 := v88 + 1;
+	v4 := compiler_strings_length;
+
+.add_string_loop:
+	lw t0, 0(sp)
+	lb t1, (t0)
+	li t2, '"'
+
+	beq t1, t2, .add_string_end
+
+	la t2, compiler_strings_position
+	lw t3, (t2)
+	sb t1, (t3)
+
+	addi t3, t3, 1
+	sw t3, (t2)
+
+	addi t0, t0, 1
+	sw t0, 0(sp)
+
+	li t2, '\\'
+	bne t1, t2, .add_string_increment
+
+	goto .add_string_loop;
+
+.add_string_increment:
+	la t2, compiler_strings_length
+	lw t4, (t2)
+	addi t4, t4, 1
+	sw t4, (t2)
+
+	goto .add_string_loop;
+
+.add_string_end:
+	return v4
+end;
+
+# 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_s();
+begin
+	mv a2, a1
+	mv a1, a0
+	# STDOUT.
+	li a0, 1
+	li a7, 64 # SYS_WRITE.
+	ecall
+end;
+
+# Writes a number to a string buffer.
+#
+# t0 - Local buffer.
+# t1 - Constant 10.
+# t2 - Current character.
+# t3 - Whether the number is negative.
+#
+# Parameters:
+# a0 - Whole number.
+# a1 - Buffer pointer.
+#
+# Sets a0 to the length of the written number.
+proc _print_i();
+begin
+	li t1, 10
+	addi t0, s0, -9
+
+	li t3, 0
+	bgez a0, .print_i_digit10
+	li t3, 1
+	neg a0, a0
+
+.print_i_digit10:
+	rem t2, a0, t1
+	addi t2, t2, '0'
+	sb t2, 0(t0)
+	div a0, a0, t1
+	addi t0, t0, -1
+	bne zero, a0, .print_i_digit10
+
+	beq zero, t3, .print_i_write_call
+	addi t2, zero, '-'
+	sb t2, 0(t0)
+	addi t0, t0, -1
+
+.print_i_write_call:
+	mv a0, a1
+	addi a1, t0, 1
+	sub a2, s0, t0
+	addi a2, a2, -9
+	sw a2, 0(sp)
+
+	_memcpy();
+
+	return v0
+end;
+
+# Writes a number to the standard output.
+#
+# Parameters:
+# a0 - Whole number.
+proc _write_i();
+begin
+	_print_i(v88, @v0);
+
+	mv a1, a0
+	addi a0, sp, 0
+	_write_s();
+
+end;
+
+# Writes a character from a0 into the standard output.
+proc _write_c();
+begin
+	_write_s(@v88, 1);
+end;
+
+# Write null terminated string.
+#
+# Parameters:
+# a0 - String.
+proc _write_z();
+begin
+.write_z_loop:
+	# Check for 0 character.
+	lw a0, 88(sp)
+	lb a0, (a0)
+	beqz a0, .write_z_end
+
+	# Print a character.
+	_write_c();
+
+	# Advance the input string by one byte.
+	v88 := v88 + 1;
+
+	goto .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
+	v0 := v88 >= 'A';
+	v4 := v88 <= 'Z';
+
+	return v0 & v4
+
+end;
+
+# Detects if a0 is an lowercase character. Sets a0 to 1 if so, otherwise to 0.
+proc _is_lower();
+begin
+	v0 := v88 >= 'a';
+	v4 := v88 <= 'z';
+
+	return v0 & v4
+
+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, 0(sp)
+
+	_is_upper();
+	sw a0, 4(sp)
+
+	_is_lower(v0);
+
+	lw t0, 0(sp)
+	xori t1, t0, '_'
+	seqz t1, t1
+
+	lw t0, 4(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
+	v0 := v88 >= '0';
+	v4 := v88 <= '9';
+
+	return v0 & v4
+end;
+
+proc _is_alnum();
+begin
+	sw a0, 4(sp)
+
+	_is_alpha();
+	sw a0, 0(sp)
+
+	_is_digit(v4);
+
+	lw a1, 0(sp)
+	or a0, a0, a1
+end;
+
+# Reads the next token.
+#
+# Returns token length in a0.
+proc _read_token();
+begin
+	# Current token position.
+	v0 := source_code_position;
+	# Token length.
+	v4 := 0;
+
+.read_token_loop:
+	lw t0, 0(sp)
+	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, 0(sp)
+	lb a0, (a0)
+	_is_alnum();
+	bnez a0, .read_token_next
+
+	goto .read_token_end;
+
+.read_token_next:
+	# Advance the source code position and token length.
+	v4 := v4 + 1;
+	v0 := v0 + 1;
+
+	goto .read_token_loop;
+
+.read_token_end:
+	return v4
+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
+
+.memcmp_loop:
+	beqz a2, .memcmp_end
+
+	lbu t1, (t0)
+	lbu t2, (a1)
+	sub a0, t1, t2
+
+	bnez a0, .memcmp_end
+
+	addi t0, t0, 1
+	addi a1, a1, 1
+	addi a2, a2, -1
+
+	goto .memcmp_loop;
+
+.memcmp_end:
+end;
+
+# Copies memory.
+#
+# Parameters:
+# a0 - Destination.
+# a1 - Source.
+# a2 - Size.
+#
+# Preserves a0.
+proc _memcpy();
+begin
+	mv t0, a0
+
+.memcpy_loop:
+	beqz a2, .memcpy_end
+
+	lbu t1, (a1)
+	sb t1, (a0)
+
+	addi a0, a0, 1
+	addi a1, a1, 1
+	addi a2, a2, -1
+
+	goto .memcpy_loop
+
+.memcpy_end:
+	mv a0, t0
+end;
+
+# Advances the token stream by a0 bytes.
+proc _advance_token();
+begin
+	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
+	_write_s(source_code_position, v88);
+	return v88
+end;
+
+proc _compile_section();
+begin
+	# Print and skip the ".section" (8 characters) directive and a space after it.
+	_write_token(9);
+	_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)
+
+	goto .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.
+	_advance_token(1);
+
+	goto .compile_line_loop;
+
+.compile_line_end:
+	_write_c('\n');
+
+	_advance_token(1);
+end;
+
+proc _compile_integer_literal();
+begin
+	_write_z("\tli t0, \0");
+
+	_read_token();
+	_write_token();
+	_advance_token();
+
+	_write_c('\n');
+end;
+
+proc _compile_character_literal();
+begin
+	_write_z("\tli t0, \0");
+
+	_write_c('\'');
+	_advance_token(1);
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+	li t1, '\\'
+	bne a0, t1, .compile_character_literal_end
+	
+	_write_c('\\');
+	_advance_token(1);
+
+.compile_character_literal_end:
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+	_write_c();
+
+	_write_c('\'');
+	_write_c('\n');
+
+	_advance_token(2);
+
+end;
+
+proc _compile_variable_expression();
+begin
+	_compile_designator();
+	_write_z("\tlw t0, (t0)\n\0");
+end;
+
+proc _compile_address_expression();
+begin
+	# Skip the "@" sign.
+	_advance_token(1);
+	_compile_designator();
+
+end;
+
+proc _compile_negate_expression();
+begin
+	# Skip the "-" sign.
+	_advance_token(1);
+	_compile_term();
+
+	_write_z("\tneg t0, t0\n\0");
+end;
+
+proc _compile_not_expression();
+begin
+	# Skip the "~" sign.
+	_advance_token(1);
+	_compile_term();
+
+	_write_z("\tnot t0, t0\n\0");
+end;
+
+proc _compile_string_literal();
+begin
+	_string_length(source_code_position);
+	sw a0, 0(sp)
+
+	_add_string(source_code_position);
+	sw a0, 4(sp)
+
+	_advance_token(v0 + 2);
+	_write_z("\tla t0, strings\n\0");
+
+	_write_z("\tli t1, \0");
+	_write_i(v4);
+	_write_c('\n');
+
+	_write_z("\tadd t0, t0, t1\n\0");
+end;
+
+proc _compile_term();
+begin
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+	sw a0, 0(sp)
+
+	li t1, '\''
+	beq a0, t1, .compile_term_character_literal
+
+	li t1, '@'
+	beq a0, t1, .compile_term_address
+
+	li t1, '-'
+	beq a0, t1, .compile_term_negation
+
+	li t1, '~'
+	beq a0, t1, .compile_term_not
+
+	li t1, '"'
+	beq a0, t1, .compile_term_string_literal
+
+	_is_digit(v0);
+	bnez a0, .compile_term_integer_literal
+
+	goto .compile_term_variable;
+
+.compile_term_character_literal:
+	_compile_character_literal();
+	goto .compile_term_end;
+
+.compile_term_integer_literal:
+	_compile_integer_literal();
+	goto .compile_term_end;
+
+.compile_term_address:
+	_compile_address_expression();
+	goto .compile_term_end;
+
+.compile_term_negation:
+	_compile_negate_expression();
+	goto .compile_term_end;
+
+.compile_term_not:
+	_compile_not_expression();
+	goto .compile_term_end;
+
+.compile_term_string_literal:
+	_compile_string_literal();
+	goto .compile_term_end;
+
+.compile_term_variable:
+	_compile_variable_expression();
+	goto .compile_term_end;
+
+.compile_term_end:
+end;
+
+proc _compile_binary_rhs();
+begin
+	# Skip the whitespace after the binary operator.
+	_advance_token(1);
+	_compile_term();
+
+	# Load the left expression from the stack;
+	_write_z("\tlw t1, 24(sp)\n\0");
+end;
+
+proc _compile_expression();
+begin
+	_compile_term();
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+
+	li t1, ' '
+	bne a0, t1, .compile_expression_end
+
+	# It is a binary expression.
+
+	# Save the value of the left expression on the stack.
+	_write_z("sw t0, 24(sp)\n\0");
+
+	# Skip surrounding whitespace in front of the operator.
+	_advance_token(1);
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, '+'
+	beq t0, t1, .compile_expression_add
+
+	li t1, '*'
+	beq t0, t1, .compile_expression_mul
+
+	li t1, '&'
+	beq t0, t1, .compile_expression_and
+
+	li t1, 'o'
+	beq t0, t1, .compile_expression_or
+
+	li t1, 'x'
+	beq t0, t1, .compile_expression_xor
+
+	li t1, '='
+	beq t0, t1, .compile_expression_equals
+
+	li t1, '<'
+	beq t0, t1, .compile_expression_less
+
+	li t1, '>'
+	beq t0, t1, .compile_expression_greater
+
+	# Unknown binary operator.
+	unimp
+
+.compile_expression_add:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("add t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_mul:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tmul t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_and:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tand t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_or:
+	_advance_token(2);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("or t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_xor:
+	_advance_token(3);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("xor t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_equals:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("xor t0, t0, t1\nseqz t0, t0\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_less:
+	_advance_token(1);
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, '>'
+	beq t0, t1, .compile_expression_not_equal
+
+	li t1, '='
+	beq t0, t1, .compile_expression_less_equal
+
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("slt t0, t0, t1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_not_equal:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\txor t0, t0, t1\nsnez t0, t0\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_less_equal:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tslt t0, t0, t1\nxori t0, t0, 1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_greater:
+	_advance_token(1);
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, '='
+	beq t0, t1, .compile_expression_greater_equal
+
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tslt t0, t1, t0\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_greater_equal:
+	_advance_token(1);
+	_compile_binary_rhs();
+
+	# Execute the operation.
+	_write_z("\tslt t0, t1, t0\nxori t0, t0, 1\n\0");
+
+	goto .compile_expression_end;
+
+.compile_expression_end:
+end;
+
+proc _compile_call();
+begin
+	# Stack variables:
+	# v0 - Procedure name length.
+	# v4 - Procedure name pointer. 
+	# v8 - Argument count.
+
+	_read_token();
+	sw a0, 0(sp)
+	v4 := source_code_position;
+	v8 := 0;
+
+	# Skip the identifier and left paren.
+	_advance_token(v0 + 1);
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, ')'
+	beq t0, t1, .compile_call_finalize
+
+.compile_call_loop:
+	_compile_expression();
+
+	# Save the argument on the stack.
+	_write_z("\tsw t0, \0");
+
+	# Calculate the stack offset: 116 - (4 * argument_counter)
+	v12 := v8 * 4;
+	v12 := 116 + -v12;
+	_write_i(v12);
+
+	_write_z("(sp)\n\0");
+
+	# Add one to the argument counter.
+	v8 := v8 + 1;
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+
+	li t1, ','
+	bne t0, t1, .compile_call_finalize
+
+	_advance_token(2);
+	goto .compile_call_loop;
+
+.compile_call_finalize:
+	# Load the argument from the stack.
+
+	lw t0, 8(sp)
+	beqz t0, .compile_call_end
+
+	# Decrement the argument counter.
+	v8 := v8 + -1;
+
+	_write_z("\tlw a\0");
+	_write_i(v8);
+
+	_write_z(", \0");
+
+	# Calculate the stack offset: 116 - (4 * argument_counter)
+	v12 := v8 * 4;
+	v12 := 116 + -v12;
+	_write_i(v12);
+
+	_write_z("(sp)\n\0");
+
+	goto .compile_call_finalize;
+
+.compile_call_end:
+	_write_z("\tcall \0");
+	_write_s(v4, v0);
+
+	# Skip the right paren.
+	_advance_token(1);
+end;
+
+proc _compile_goto();
+begin
+	_advance_token(5);
+
+	_read_token();
+	sw a0, 0(sp)
+
+	_write_z("\tj \0");
+
+	_write_token(v0);
+	_advance_token();
+end;
+
+proc _compile_local_designator();
+begin
+	# Skip "v" in the local variable name.
+	_advance_token(1);
+	_write_z("\t addi t0, sp, \0");
+
+	# Read local variable stack offset and save it.
+	_read_token();
+	_write_token();
+	_advance_token();
+	_write_c('\n');
+end;
+
+proc _compile_global_designator();
+begin
+	_write_z("\tla t0, \0");
+
+	_read_token();
+	_write_token();
+	_advance_token();
+
+	_write_c('\n');
+end;
+
+proc _compile_designator();
+begin
+	la t0, source_code_position
+	lw t0, (t0)
+	lb a0, (t0)
+
+	li t1, 'v'
+	beq a0, t1, .compile_designator_local
+
+	goto .compile_designator_global;
+
+.compile_designator_local:
+	_compile_local_designator();
+	goto .compile_designator_end;
+
+.compile_designator_global:
+	_compile_global_designator();
+	goto .compile_designator_end;
+
+.compile_designator_end:
+end;
+
+proc _compile_assignment();
+begin
+	_compile_designator();
+
+	# Save the assignee address on the stack.
+	_write_z("\tsw t0, 20(sp)\n\0");
+
+	# Skip the assignment sign (:=) with surrounding whitespaces.
+	_advance_token(4);
+
+	# Compile the assignment.
+	_compile_expression();
+
+	_write_z("\tlw t1, 20(sp)\nsw t0, (t1)\n\0");
+end;
+
+proc _compile_return_statement();
+begin
+	# Skip "return" keyword and whitespace after it.
+	_advance_token(7);
+	_compile_expression();
+
+	_write_z("mv a0, t0\n\0");
+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
+
+	li t1, 'v'
+	beq t0, t1, .compile_statement_assignment
+
+	# keyword_ret contains "\tret", so it's 4 bytes long.
+	_memcmp(source_code_position, "\treturn", 7);
+	beqz a0, .compile_statement_return
+
+	_compile_line();
+	goto .compile_statement_end;
+
+.compile_statement_call:
+	_advance_token(1);
+	_compile_call();
+
+	goto .compile_statement_semicolon;
+
+.compile_statement_goto:
+	_advance_token(1);
+	_compile_goto();
+
+	goto .compile_statement_semicolon;
+
+.compile_statement_assignment:
+	_advance_token(1);
+	_compile_assignment();
+
+	goto .compile_statement_semicolon;
+
+.compile_statement_return:
+	_advance_token(1);
+	_compile_return_statement();
+	_write_c('\n');
+
+	goto .compile_statement_end;
+
+.compile_statement_semicolon:
+	_advance_token(2);
+	_write_c('\n');
+
+.compile_statement_end:
+end;
+
+proc _compile_procedure_body();
+begin
+.compile_procedure_body_loop:
+	# 3 is "end" length.
+	_memcmp(source_code_position, "end", 3);
+
+	beqz a0, .compile_procedure_body_epilogue
+
+	_compile_statement();
+	goto .compile_procedure_body_loop;
+
+.compile_procedure_body_epilogue:
+end;
+
+# Writes a regster name to the standard output.
+#
+# Parameters:
+# a0 - Register character.
+# a1 - Register number.
+proc _write_register();
+begin
+	_write_c(v88);
+	v84 := v84 + '0';
+	_write_c(v84);
+end;
+
+proc _compile_procedure_prologue();
+begin
+	_write_z("\taddi sp, sp, -128\n\tsw ra, 124(sp)\n\tsw s0, 120(sp)\n\taddi s0, sp, 128\n\0");
+	v0 := 0;
+
+.compile_procedure_prologue_loop:
+	_write_z("\tsw a\0");
+	_write_i(v0);
+	_write_z(", \0");
+
+	# Calculate the stack offset: 88 - (4 * parameter_counter)
+	v4 := v0 * 4;
+	v4 := 88 + -v4;
+	_write_i(v4);
+
+	_write_z("(sp)\n\0");
+
+	v0 := v0 + 1;
+	lw a0, 0(sp)
+
+	li t0, 8
+	bne a0, t0, .compile_procedure_prologue_loop
+end;
+
+proc _compile_procedure();
+begin
+	# Skip "proc ".
+	_advance_token(5);
+
+	_read_token();
+	sw a0, 0(sp) # Save the procedure name length.
+
+	# Write .type _procedure_name, @function.
+	_write_z(".type \0");
+
+	_write_token(v0);
+	_write_z(", @function\n\0");
+
+	# Write procedure label, _procedure_name:
+	_write_token(v0);
+	_write_z(":\n\0");
+
+	# Skip the function name and trailing parens, semicolon, "begin" and newline.
+	_advance_token(v0 + 10);
+
+	_compile_procedure_prologue();
+	_compile_procedure_body();
+
+	# Write the epilogue.
+	_write_z("\tlw ra, 124(sp)\n\tlw s0, 120(sp)\n\taddi sp, sp, 128\n\tret\n\0");
+
+	# Skip the "end" keyword, semicolon and newline.
+	_advance_token(5);
+end;
+
+proc _compile_type();
+begin
+	# Print and skip the ".type" (5 characters) directive and a space after it.
+	_write_token(6);
+	_advance_token();
+
+	# Read and print the symbol name.
+	_read_token();
+
+	# Print and skip the symbol name, comma, space and @.
+	addi a0, a0, 3
+	_write_token();
+	_advance_token();
+
+	# Read the symbol type.
+	_read_token();
+
+	# Print the symbol type and newline.
+	addi a0, a0, 1
+	_write_token();
+	_advance_token();
+
+	# Write the object definition itself.
+	_compile_line();
+
+.compile_type_end:
+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)
+
+	goto .skip_newlines_loop;
+
+.skip_newlines_end:
+end;
+
+# Process the source code and print the generated code.
+proc _compile_module();
+begin
+.compile_module_loop:
+	_skip_newlines();
+
+	la t0, source_code_position
+	lw t0, (t0)
+	lb t0, (t0)
+	beqz t0, .compile_module_end
+	li t1, '#'
+	beq t0, t1, .compile_module_comment
+
+	# 8 is ".section" length.
+	_memcmp(source_code_position, ".section", 8);
+	beqz a0, .compile_module_section
+
+	# 5 is ".type" length.
+	_memcmp(source_code_position, ".type", 5);
+	beqz a0, .compile_module_type
+
+	# 5 is "proc " length. Space is needed to distinguish from "procedure".
+	_memcmp(source_code_position, "proc ", 5);
+	beqz a0, .compile_module_procedure
+
+	# 6 is ".globl" length.
+	_memcmp(source_code_position, ".globl", 6);
+	beqz a0, .compile_module_global
+
+	# Not a known token, exit.
+	goto .compile_module_end;
+
+.compile_module_section:
+	_compile_section();
+
+	goto .compile_module_loop;
+
+.compile_module_type:
+	_compile_type();
+
+	goto .compile_module_loop;
+
+.compile_module_global:
+	_compile_line();
+
+	goto .compile_module_loop;
+
+.compile_module_comment:
+	_skip_comment();
+
+	goto .compile_module_loop;
+
+.compile_module_procedure:
+	_compile_procedure();
+
+	goto .compile_module_loop;
+
+.compile_module_end:
+end;
+
+proc _compile();
+begin
+	_compile_module();
+
+	_write_z(".section .rodata\n.type strings, @object\nstrings: .ascii \0");
+	_write_c('"');
+
+	la t0, compiler_strings
+	sw t0, 0(sp)
+
+.compile_loop:
+	lw t0, 0(sp)
+	la t1, compiler_strings_position
+	lw t1, (t1)
+	bge t0, t1, .compile_end
+
+	lb a0, (t0)
+
+	addi t0, t0, 1
+	sw t0, 0(sp)
+
+	_write_c();
+
+	j .compile_loop
+
+.compile_end:
+	_write_c('"');
+	_write_c('\n');
+end;
+
+# Terminates the program. a0 contains the return code.
+#
+# Parameters:
+# a0 - Status code.
+proc _exit();
+begin
+	li a7, 93 # SYS_EXIT
+	ecall
+end;
+
+# Entry point.
+.globl _start
+proc _start();
+begin
+	# Read the source from the standard input.
+	# Second argument is buffer size. Modifying update the source_code definition.
+	_read_file(@source_code, 81920);
+	_compile();
+
+	_exit(0);
+
+end;