Starting on asm, implementing extern function call so progress is visible

37 files changed, 1469 insertions, 359 deletions

diff --git a/README.md b/README.md
index c02a2d2..f4f228c 100644
--- a/README.md
+++ b/README.md
@@ -6,8 +6,15 @@ Amalgam is not meant to be a replacement for any other language but rather a new
 with gpu without writing an external gpu program (glsl/hlsl).
 
 Files have to be in utf-8 format and can optionally have utf-8 BOM.
+# Fast compilation
+Every stage of the compiler is multithreaded and data copy is kept to a minimal, for example tokenization
+is done without storing tokens in a list. Almost all allocation is done using an arena allocator that
+is only cleaned up once (when the compiler is finished), and the data is allocated sequentially.
 # TODO
-Build with -nostdlib and replace use of libc with syscalls (on linux)
+Build with -nostdlib and replace use of libc with syscalls (on linux).
+Don't parse files unless the variable they are assigned to (with @import) is used. This is useful when only using small parts of a library.
+Align machine code to word boundary for the start of functions. No need to pad with NOP, as functions return before the padding.
+Use const to cleanup ANSI C style variable declarations, since const allows you to declare and assign variables on the same line.
 # Documents
 Documents are located under doc. The file doc/Documentation.md is generated from source files by running doc/doc_extract.py
-but there is no need to run this script unless you are modifying documentation in the source.
\ No newline at end of file
+but there is no need to run this script unless you are modifying documentation in the source.
diff --git a/build.sh b/build.sh
index 214769a..8fd07b8 100755
--- a/build.sh
+++ b/build.sh
@@ -23,41 +23,88 @@ if [ ! -z "$PEDANTIC" ]; then
     CFLAGS+="-DAMAL_PEDANTIC -pedantic "
 fi
 
-CFLAGS+="-Wall -Wextra -Werror -g -O0 -DDEBUG -std=c89 -D_GNU_SOURCE"
-LIBS+="-pthread"
-
-BUILD_ARGS="$source_files $CFLAGS $LIBS -shared -fpic -o "$this_script_dir/libamalgam.so""
-set -x
-time $CC $BUILD_ARGS
-if [ ! -z "$SCAN_BUILD" ]; then
-    scan-build $CC $BUILD_ARGS
-fi
-set +x
+build_test() {
+    CFLAGS+="-Wall -Wextra -Werror -Wno-format-security -g -O0 -DDEBUG -std=c89 -D_GNU_SOURCE"
+    LIBS+="-pthread"
 
-if [ -z "$NO_TEST" ]; then
-    source_files_tests=$(readlink -f $(find "$this_script_dir/tests" -name "*.c"))
+    BUILD_ARGS="$source_files $CFLAGS $LIBS -shared -fpic -o "$this_script_dir/libamalgam.so""
     set -x
-    time $CC $source_files_tests $CFLAGS $LIBS -o test "$this_script_dir/libamalgam.so"
-fi
+    time $CC $BUILD_ARGS
+    if [ ! -z "$SCAN_BUILD" ]; then
+        scan-build $CC $BUILD_ARGS
+    fi
+    set +x
+
+    if [ -z "$NO_TEST" ]; then
+        source_files_tests=$(readlink -f $(find "$this_script_dir/tests" -name "*.c"))
+        set -x
+        time $CC $source_files_tests $CFLAGS $LIBS -o test "$this_script_dir/libamalgam.so"
+    fi
 
-set +x
-compile_commands=$(
-first=0
-echo "["
-for source_file in $source_files $source_files_tests; do
-    if [ $first == 1 ]; then
-        echo ","
+    set +x
+    compile_commands=$(
+    first=0
+    echo "["
+    for source_file in $source_files $source_files_tests; do
+        if [ $first == 1 ]; then
+            echo "  ,"
+        fi
+        first=1
+        o_file="${source_file}.o"
+        echo "  {"
+        echo "    \"file\": \"$source_file\","
+        echo "    \"directory\": \"$this_script_dir\","
+        echo "    \"command\": \"$CC -o $o_file $CFLAGS $LIBS -c $source_file\""
+        echo "  }"
+    done
+    echo "]")
+    echo "$compile_commands" > "$this_script_dir/compile_commands.json"
+}
+
+build_release() {
+    CFLAGS+="-Wall -Wextra -Werror -Wno-format-security -O2 -DNDEBUG -std=c89 -D_GNU_SOURCE -s"
+    LIBS+="-pthread"
+
+    BUILD_ARGS="$source_files $CFLAGS $LIBS -shared -fpic -o "$this_script_dir/libamalgam.so""
+    set -x
+    time $CC $BUILD_ARGS
+    if [ ! -z "$SCAN_BUILD" ]; then
+        scan-build $CC $BUILD_ARGS
     fi
-    first=1
-    o_file="${source_file}.o"
-    echo "  {"
-    echo "    \"file\": \"$source_file\","
-    echo "    \"directory\": \"$this_script_dir\","
-    echo "    \"command\": \"$CC -o $o_file $CFLAGS $LIBS -c $source_file\","
-    echo "    \"output\": \"$o_file\""
-    echo "  }"
-done
-echo "]")
-echo "$compile_commands" > "$this_script_dir/compile_commands.json"
-
-echo "Finished building"
+    set +x
+
+    set +x
+    compile_commands=$(
+    first=0
+    echo "["
+    for source_file in $source_files; do
+        if [ $first == 1 ]; then
+            echo "  ,"
+        fi
+        first=1
+        o_file="${source_file}.o"
+        echo "  {"
+        echo "    \"file\": \"$source_file\","
+        echo "    \"directory\": \"$this_script_dir\","
+        echo "    \"command\": \"$CC -o $o_file $CFLAGS $LIBS -c $source_file\""
+        echo "  }"
+    done
+    echo "]")
+    echo "$compile_commands" > "$this_script_dir/compile_commands.json"
+}
+
+case "$1" in
+    "test")
+        build_test
+        ;;
+    "release")
+        build_release
+        ;;
+    *)
+        echo "usage: build.sh COMMAND"
+        echo "COMMANDS:"
+        echo "  test       Build tests"
+        echo "  release    Build as a release package, enabling all optimizations and stripping of binary"
+        exit 1
+        ;;
+esac
diff --git a/doc/Documentation.md b/doc/Documentation.md
index 5564b05..3e60994 100644
--- a/doc/Documentation.md
+++ b/doc/Documentation.md
@@ -2,19 +2,21 @@
 Variable length opcodes. Sizes range from 1 to 4 bytes.
 ## Instruction formats
 Instructions can be in 6 different formats:
-1. 1 byte: Opcode
-2. 2 bytes: Opcode + register
-3. 3 bytes: Opcode + register + register
+1. 1 byte: Opcode(u8)
+2. 2 bytes: Opcode(u8) + register(u8)
+3. 3 bytes: Opcode(u8) + register(u8) + register(u8)
 4. 3 bytes:\
-4.1 Opcode + intermediate\
-4.2 Opcode + data\
-4.3 Opcode + index\
-4.4 Opcode + offset
-5. 4 bytes: Opcode + register + register + register
+4.1 Opcode(u8) + intermediate(u16)\
+4.2 Opcode(u8) + data(u16)\
+4.3 Opcode(u8) + offset(i16)\
+4.4 Opcode(u8) + num_reg(u16)\
+4.5 Opcode(u8) + register(u8) + num_args(u8)
+5. 4 bytes: Opcode(u8) + register(u8) + register(u8) + register(u8)
 6. 4 bytes:\
-6.1 Opcode + register + offset\
-6.2 Opcode + register + intermediate\
-6.3 Opcode + register + data
+6.1 Opcode(u8) + register(u8) + offset(i16)\
+6.2 Opcode(u8) + register(u8) + intermediate(u16)\
+6.3 Opcode(u8) + register(u8) + data(u16)
+7. 4 bytes: Opcode(u8) + index(u16) + num_args(u8)
 
 # Compiler flow
 (Tokenize&parse -> Resolve AST -> Generate SSA -> Generate bytecode) -> Generate program\
diff --git a/include/asm/x86_64.h b/include/asm/x86_64.h
new file mode 100644
index 0000000..92de96b
--- /dev/null
+++ b/include/asm/x86_64.h
@@ -0,0 +1,136 @@
+#ifndef AMAL_ASM_X86_64_H
+#define AMAL_ASM_X86_64_H
+
+#include "../std/misc.h"
+#include "../std/types.h"
+
+typedef struct {
+    void *code;
+    u8 *code_it;
+    usize size;
+} Asm;
+
+typedef enum {
+    EAX,
+    ECX,
+    EDX,
+    EBX,
+    ESP,
+    EBP,
+    ESI,
+    EDI
+} Reg32;
+
+typedef enum {
+    RAX,
+    RCX,
+    RDX,
+    RBX,
+    RSP,
+    RBP,
+    RSI,
+    RDI
+} Reg64;
+
+typedef struct {
+    Reg64 base;
+    Reg64 index;
+    i32 disp;
+    u8 scale;
+} AsmPtr;
+
+void asm_ptr_init(AsmPtr *self, Reg64 base);
+void asm_ptr_init_index(AsmPtr *self, Reg64 base, Reg64 index);
+void asm_ptr_init_disp(AsmPtr *self, Reg64 base, i32 disp);
+void asm_ptr_init_index_disp(AsmPtr *self, Reg64 base, Reg64 index, i32 disp);
+
+CHECK_RESULT int asm_init(Asm *self);
+void asm_deinit(Asm *self);
+
+CHECK_RESULT int asm_execute(Asm *self);
+
+CHECK_RESULT int asm_nop(Asm *self);
+
+
+
+
+
+
+
+
+
+CHECK_RESULT int asm_mov_mi(Asm *self, AsmPtr *dst, i32 immediate);
+CHECK_RESULT int asm_mov_mr(Asm *self, AsmPtr *dst, Reg64 src);
+CHECK_RESULT int asm_mov_rm(Asm *self, Reg64 dst, AsmPtr *src);
+CHECK_RESULT int asm_mov_ri(Asm *self, Reg64 dst, i64 immediate);
+CHECK_RESULT int asm_mov_rr(Asm *self, Reg64 dst, Reg64 src);
+
+CHECK_RESULT int asm_add_rr(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_sub_rr(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_imul_rr(Asm *self, Reg64 dst, Reg64 src);
+
+CHECK_RESULT int asm_pushr(Asm *self, Reg64 reg);
+CHECK_RESULT int asm_popr(Asm *self, Reg64 reg);
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+CHECK_RESULT int asm_mov_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_add_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_sub_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_and_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_or_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_xor_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_cmp_rm32(Asm *self, Reg32 dst, Reg32 src);
+CHECK_RESULT int asm_add_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_or_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_adc_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_sbb_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_and_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_sub_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_xor_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_cmp_rm32_imm(Asm *self, Reg32 reg, i32 immediate);
+CHECK_RESULT int asm_rol_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+CHECK_RESULT int asm_ror_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+CHECK_RESULT int asm_rcl_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+CHECK_RESULT int asm_rcr_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+CHECK_RESULT int asm_shl_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+CHECK_RESULT int asm_shr_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+CHECK_RESULT int asm_sar_rm32_imm(Asm *self, Reg32 reg, i8 immediate);
+
+CHECK_RESULT int asm_mov_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_add_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_sub_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_and_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_or_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_xor_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_cmp_rm64(Asm *self, Reg64 dst, Reg64 src);
+CHECK_RESULT int asm_add_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_or_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_adc_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_sbb_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_and_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_sub_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_xor_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_cmp_rm64_imm(Asm *self, Reg64 reg, i32 immediate);
+CHECK_RESULT int asm_rol_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+CHECK_RESULT int asm_ror_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+CHECK_RESULT int asm_rcl_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+CHECK_RESULT int asm_rcr_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+CHECK_RESULT int asm_shl_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+CHECK_RESULT int asm_shr_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+CHECK_RESULT int asm_sar_rm64_imm(Asm *self, Reg64 reg, i8 immediate);
+
+CHECK_RESULT int asm_ret(Asm *self, u16 bytes);
+
+#endif
diff --git a/include/ast.h b/include/ast.h
index f056db7..4c54766 100644
--- a/include/ast.h
+++ b/include/ast.h
@@ -6,7 +6,7 @@
 #include "std/buffer_view.h"
 #include "std/buffer.h"
 #include "std/misc.h"
-#include "std/scoped_allocator.h"
+#include "std/arena_allocator.h"
 #include "std/hash_map.h"
 #include "binop_type.h"
 #include "ssa/ssa.h"
@@ -106,6 +106,7 @@ struct Variable {
 
 struct FunctionSignature {
     int params; /* TODO: Implement signatures */
+    bool resolved;
 };
 
 struct FunctionDecl {
@@ -218,12 +219,13 @@ typedef struct {
     Scope *scope;
 } AstCompilerContext;
 
-CHECK_RESULT int ast_create(ScopedAllocator *allocator, void *value, AstType type, Ast **result);
+CHECK_RESULT int ast_create(ArenaAllocator *allocator, void *value, AstType type, Ast **result);
 BufferView ast_get_name(Ast *self);
 
-CHECK_RESULT int funcdecl_init(FunctionDecl *self, FunctionSignature *signature, Scope *parent, ScopedAllocator *allocator);
-CHECK_RESULT int funccall_init(FunctionCall *self, BufferView name, ScopedAllocator *allocator);
-CHECK_RESULT int structdecl_init(StructDecl *self, Scope *parent, ScopedAllocator *allocator);
+void function_signature_init(FunctionSignature *self);
+CHECK_RESULT int funcdecl_init(FunctionDecl *self, FunctionSignature *signature, Scope *parent, ArenaAllocator *allocator);
+CHECK_RESULT int funccall_init(FunctionCall *self, BufferView name, ArenaAllocator *allocator);
+CHECK_RESULT int structdecl_init(StructDecl *self, Scope *parent, ArenaAllocator *allocator);
 void structfield_init(StructField *self, BufferView name, BufferView type_name);
 void lhsexpr_init(LhsExpr *self, bool is_extern, bool is_pub, bool is_const, BufferView var_name);
 void assignmentexpr_init(AssignmentExpr *self, Ast *lhs_expr, Ast *rhs_expr);
@@ -232,12 +234,12 @@ CHECK_RESULT int string_init(String *self, BufferView str);
 void number_init(Number *self, i64 value, bool is_integer, BufferView code_ref);
 void variable_init(Variable *self, BufferView name);
 void binop_init(Binop *self);
-CHECK_RESULT int if_statement_init(IfStatement *self, Scope *parent, ScopedAllocator *allocator);
-CHECK_RESULT int else_if_statement_init(ElseIfStatement *self, Scope *parent, ScopedAllocator *allocator);
-CHECK_RESULT int while_statement_init(WhileStatement *self, Scope *parent, ScopedAllocator *allocator);
+CHECK_RESULT int if_statement_init(IfStatement *self, Scope *parent, ArenaAllocator *allocator);
+CHECK_RESULT int else_if_statement_init(ElseIfStatement *self, Scope *parent, ArenaAllocator *allocator);
+CHECK_RESULT int while_statement_init(WhileStatement *self, Scope *parent, ArenaAllocator *allocator);
 
-CHECK_RESULT int scope_init(Scope *self, Scope *parent, ScopedAllocator *allocator);
-CHECK_RESULT int file_scope_reference_init(FileScopeReference *self, BufferView canonical_path, ScopedAllocator *allocator);
+CHECK_RESULT int scope_init(Scope *self, Scope *parent, ArenaAllocator *allocator);
+CHECK_RESULT int file_scope_reference_init(FileScopeReference *self, BufferView canonical_path, ArenaAllocator *allocator);
 CHECK_RESULT int scope_add_child(Scope *self, Ast *child);
 /* longjump to compiler env on failure */
 void scope_resolve(Scope *self, AstCompilerContext *context);
diff --git a/include/bytecode/bytecode.h b/include/bytecode/bytecode.h
index b571383..32488cc 100644
--- a/include/bytecode/bytecode.h
+++ b/include/bytecode/bytecode.h
@@ -12,41 +12,45 @@
     Variable length opcodes. Sizes range from 1 to 4 bytes.
     # Instruction formats
     Instructions can be in 6 different formats:
-    1. 1 byte: Opcode
-    2. 2 bytes: Opcode + register
-    3. 3 bytes: Opcode + register + register
+    1. 1 byte: Opcode(u8)
+    2. 2 bytes: Opcode(u8) + register(u8)
+    3. 3 bytes: Opcode(u8) + register(u8) + register(u8)
     4. 3 bytes:\
-    4.1 Opcode + intermediate\
-    4.2 Opcode + data\
-    4.3 Opcode + index\
-    4.4 Opcode + offset
-    5. 4 bytes: Opcode + register + register + register
+    4.1 Opcode(u8) + intermediate(u16)\
+    4.2 Opcode(u8) + data(u16)\
+    4.3 Opcode(u8) + offset(i16)\
+    4.4 Opcode(u8) + num_reg(u16)\
+    4.5 Opcode(u8) + register(u8) + num_args(u8)
+    5. 4 bytes: Opcode(u8) + register(u8) + register(u8) + register(u8)
     6. 4 bytes:\
-    6.1 Opcode + register + offset\
-    6.2 Opcode + register + intermediate\
-    6.3 Opcode + register + data
+    6.1 Opcode(u8) + register(u8) + offset(i16)\
+    6.2 Opcode(u8) + register(u8) + intermediate(u16)\
+    6.3 Opcode(u8) + register(u8) + data(u16)
+    7. 4 bytes: Opcode(u8) + index(u16) + num_args(u8)
 */
 typedef enum {
-    AMAL_OP_NOP,         /* No operation. This can be used for patching */
+    AMAL_OP_NOP,         /* No operation (do nothing). This can be used for patching code */
     AMAL_OP_SETZ,        /* setz reg             - Set register value to 0 */
-    AMAL_OP_MOV,         /* mov dst, src         - move src register to dst register */
-    AMAL_OP_MOVI,        /* movi dst, src        - move src intermediate to dst register */
-    AMAL_OP_MOVD,        /* movd dst, src        - move src data to dst register */
-    AMAL_OP_ADD,         /* add dst, reg1, reg2 */
-    AMAL_OP_SUB,         /* sub dst, reg1, reg2 */
-    AMAL_OP_MUL,         /* mul dst, reg1, reg2 */
-    AMAL_OP_DIV,         /* div dst, reg1, reg2 */
+    AMAL_OP_MOV,         /* mov dst, src         - Move src register to dst register */
+    AMAL_OP_MOVI,        /* movi dst, src        - Move src intermediate to dst register */
+    AMAL_OP_MOVD,        /* movd dst, src        - Move src data to dst register */
+    AMAL_OP_ADD,         /* add dst, reg1, reg2  - Add reg1 and reg2 and put the result in dst */
+    AMAL_OP_SUB,         /* sub dst, reg1, reg2  - Substract reg2 from reg1 and put the result in dst */
+    AMAL_OP_IMUL,        /* imul dst, reg1, reg2 - Signed multiplication */
+    AMAL_OP_MUL,         /* mul dst, reg1, reg2  - Unsigned multiplication */
+    AMAL_OP_IDIV,        /* idiv dst, reg1, reg2 - Signed division */
+    AMAL_OP_DIV,         /* div dst, reg1, reg2  - Unsigned division */
     AMAL_OP_PUSH,        /* push reg             - Push register onto stack */
     AMAL_OP_PUSHI,       /* pushi int            - Push intermediate onto stack */
     AMAL_OP_PUSHD,       /* pushd data           - Push data onto stack */
-    AMAL_OP_CALL,        /* call fi              - Call a function using function index (fi). fi is u16 */
-    AMAL_OP_CALLR,       /* callr reg            - Call a function using a register. Used for function pointers */
+    AMAL_OP_CALL,        /* call fi, num_args    - Call a function using function index (fi) and num_args arguments. fi is u16 and num_args is u8 */
+    AMAL_OP_CALLR,       /* callr reg, num_args  - Call a function using a register. Used for function pointers. num_args is u8 */
     AMAL_OP_CMP,         /* cmp dst, reg1, reg2  - Set dst to 1 if reg1 equals reg2, otherwise set it to 0 */
-    AMAL_OP_JZ,          /* jz reg, offset       - jump to offset if reg is zero. offset is i16 */
-    AMAL_OP_JMP,         /* jmp offset           - unconditional jump to offset. offset is i16 */
-    AMAL_OP_RET,         /* ret */
-    AMAL_OP_FUNC_START,  /* func_start */
-    AMAL_OP_FUNC_END     /* func_end */
+    AMAL_OP_JZ,          /* jz reg, offset       - Jump to offset if reg is zero. offset is i16 */
+    AMAL_OP_JMP,         /* jmp offset           - Unconditional jump to offset. offset is i16 */
+    AMAL_OP_RET,         /* ret                  - Return from the function */
+    AMAL_OP_FUNC_START,  /* func_start num_reg   - Start of a function which has num_reg registers allocated. num_reg is a u16 */
+    AMAL_OP_FUNC_END     /* func_end             - End of a function. Implementation should do a ret here */
 } AmalOpcode;
 
 typedef u8 AmalOpcodeType;
@@ -61,7 +65,7 @@ typedef struct {
     Parser *parser; /* borrowed */
 } BytecodeCompilerContext;
 
-CHECK_RESULT int bytecode_init(Bytecode *self, ScopedAllocator *allocator);
+CHECK_RESULT int bytecode_init(Bytecode *self, ArenaAllocator *allocator);
 
 /* longjump to self->env on failure */
 void generate_bytecode_from_ssa(BytecodeCompilerContext *self);
diff --git a/include/compiler.h b/include/compiler.h
index 0ae8532..a0ac17a 100644
--- a/include/compiler.h
+++ b/include/compiler.h
@@ -4,7 +4,7 @@
 #include "std/misc.h"
 #include "std/buffer.h"
 #include "std/buffer_view.h"
-#include "std/scoped_allocator.h"
+#include "std/arena_allocator.h"
 #include "std/thread.h"
 #include "compiler_options.h"
 #include "ast.h"
@@ -14,27 +14,38 @@
 /* General error */
 #define AMAL_COMPILER_ERR -1
 
+#define NUM_ARITHMETIC_TYPES 10
+
+typedef struct {
+    LhsExpr lhs_expr;
+    bool is_signed;
+} amal_default_type;
+
 typedef struct {
-    LhsExpr *i8;
-    LhsExpr *i16;
-    LhsExpr *i32;
-    LhsExpr *i64;
-    LhsExpr *u8;
-    LhsExpr *u16;
-    LhsExpr *u32;
-    LhsExpr *u64;
-    LhsExpr *isize;
-    LhsExpr *usize;
-    LhsExpr *f32;
-    LhsExpr *f64;
-    LhsExpr *str;
+    amal_default_type *i8;
+    amal_default_type *i16;
+    amal_default_type *i32;
+    amal_default_type *i64;
+    amal_default_type *u8;
+    amal_default_type *u16;
+    amal_default_type *u32;
+    amal_default_type *u64;
+    amal_default_type *isize;
+    amal_default_type *usize;
+    amal_default_type *f32;
+    amal_default_type *f64;
+    amal_default_type *str;
+
+    amal_default_type *arithmetic_types[NUM_ARITHMETIC_TYPES];
 } amal_default_types;
 
+bool is_arithmetic_type(LhsExpr *expr, amal_compiler *compiler);
+
 struct amal_compiler {
     amal_default_types default_types;
     amal_compiler_options options;
     amal_program *program;
-    ScopedAllocator allocator;
+    ArenaAllocator allocator;
     Scope root_scope;
     Buffer/*<Parser*>*/ parsers;
     Buffer/*<FileScopeReference*>*/ queued_files;
diff --git a/include/parser.h b/include/parser.h
index bc205f0..4427cc8 100644
--- a/include/parser.h
+++ b/include/parser.h
@@ -2,7 +2,7 @@
 #define AMALGAM_PARSER_H
 
 #include "std/buffer_view.h"
-#include "std/scoped_allocator.h"
+#include "std/arena_allocator.h"
 #include "std/thread.h"
 #include "bytecode/bytecode.h"
 #include "tokenizer.h"
@@ -25,7 +25,7 @@ typedef enum {
 struct ParserThreadData {
     amal_thread thread;
     ParserThreadStatus status;
-    ScopedAllocator allocator;
+    ArenaAllocator allocator;
 };
 
 typedef enum {
@@ -41,7 +41,7 @@ struct Parser {
     Scope *current_scope;
     bool has_func_parent;
     /* Borrowed from @compiler for faster access to allocator. The allocator is thread-specific */
-    ScopedAllocator *allocator;
+    ArenaAllocator *allocator;
     amal_compiler *compiler;
     Ssa *ssa;
     bool started;
@@ -56,7 +56,7 @@ void parser_thread_data_deinit(ParserThreadData *self);
 CHECK_RESULT int parser_thread_data_start(ParserThreadData *self, AmalThreadCallbackFunc callback_func, void *userdata);
 CHECK_RESULT int parser_thread_data_join(ParserThreadData *self, void **result);
 
-CHECK_RESULT int parser_init(Parser *self, amal_compiler *compiler, ScopedAllocator *allocator);
+CHECK_RESULT int parser_init(Parser *self, amal_compiler *compiler, ArenaAllocator *allocator);
 
 /*
 @buffer_name will be the path to the file when using parser_parse_file and when parsing a buffer
diff --git a/include/program.h b/include/program.h
index f251fbc..3fc69fa 100644
--- a/include/program.h
+++ b/include/program.h
@@ -3,6 +3,7 @@
 
 #include "std/buffer.h"
 #include "bytecode/bytecode.h"
+#include "asm/x86_64.h"
 
 #define AMAL_PROGRAM_OK 0
 #define AMAL_PROGRAM_INVALID_HEADER -1
@@ -14,18 +15,36 @@
 #define AMAL_PROGRAM_INVALID_STRINGS_SIZE -7
 #define AMAL_PROGRAM_STRING_ALLOC_FAILURE -8
 #define AMAL_PROGRAM_INVALID_INSTRUCTIONS_SIZE -9
+#define AMAL_PROGRAM_INVALID_INSTRUCTION -10
+#define AMAL_PROGRAM_INSTRUCTION_INVALID_INTERMEDIATE_INDEX -11
+#define AMAL_PROGRAM_INSTRUCTION_INVALID_DATA_INDEX -12
+#define AMAL_PROGRAM_INSTRUCTION_STACK_OVERFLOW -13
+#define AMAL_PROGRAM_INSTRUCTION_STACK_OOM -14
+#define AMAL_PROGRAM_INSTRUCTION_ILLEGAL_JUMP_TARGET -15
 
 #define AMAL_PROGRAM_MAGIC_NUMBER (u32)0xdec05eba
 #define AMAL_PROGRAM_MAJOR_VERSION 1
 #define AMAL_PROGRAM_MINOR_VERSION 0
 #define AMAL_PROGRAM_PATCH_VERSION 0
 
+#define AMAL_PROGRAM_NUM_REGISTERS 256
+
 typedef struct {
     Buffer/*<...>*/ data;
-    Buffer/*<u32>*/ string_indices;
-    char *intermediates_start;
-    char *strings_start;
+    u32 *string_indices;
+    char *intermediates_start;  /* Reference inside @data */
+    char *strings_start;        /* Reference inside @data */
     usize read_index;
+
+    u16 num_strings;
+    u16 num_intermediates;
+
+    u64 reg[AMAL_PROGRAM_NUM_REGISTERS];
+    u64 *stack;
+    usize stack_size;
+    usize stack_index;
+
+    Asm asm;
 } amal_program;
 
 CHECK_RESULT int amal_program_init(amal_program *self);
diff --git a/include/ssa/ssa.h b/include/ssa/ssa.h
index 15a9f78..1d4c612 100644
--- a/include/ssa/ssa.h
+++ b/include/ssa/ssa.h
@@ -14,7 +14,9 @@ typedef enum {
     SSA_ASSIGN_REG,
     SSA_ADD,
     SSA_SUB,
+    SSA_IMUL,
     SSA_MUL,
+    SSA_IDIV,
     SSA_DIV,
     SSA_EQUALS,
     SSA_FUNC_START,
@@ -69,7 +71,7 @@ typedef struct {
 
 typedef struct {
     SsaFuncIndex func_index;
-    u8 num_args;
+    u16 num_registers;
 } SsaInsFuncStart;
 
 typedef struct {
@@ -93,7 +95,7 @@ SsaNumber create_ssa_float(f64 value);
 SsaNumber ssa_get_intermediate(Ssa *self, SsaIntermediateIndex index);
 BufferView ssa_get_string(Ssa *self, SsaStringIndex index);
 
-CHECK_RESULT int ssa_init(Ssa *self, ScopedAllocator *allocator);
+CHECK_RESULT int ssa_init(Ssa *self, ArenaAllocator *allocator);
 
 typedef struct {
     jmp_buf env;
diff --git a/include/std/arena_allocator.h b/include/std/arena_allocator.h
new file mode 100644
index 0000000..30147c8
--- /dev/null
+++ b/include/std/arena_allocator.h
@@ -0,0 +1,29 @@
+#ifndef AMALGAM_ARENA_ALLOCATOR_H
+#define AMALGAM_ARENA_ALLOCATOR_H
+
+#include "defs.h"
+#include "misc.h"
+#include "types.h"
+#include "buffer.h"
+
+struct ArenaAllocatorNode {
+    char *data;
+    usize size;
+    ArenaAllocatorNode *next;
+};
+
+struct ArenaAllocator {
+    ArenaAllocatorNode head;
+    ArenaAllocatorNode *current;
+    Buffer/*<void*>*/ mems;
+};
+
+CHECK_RESULT int arena_allocator_node_init(ArenaAllocatorNode *self);
+void arena_allocator_node_deinit(ArenaAllocatorNode *self);
+
+CHECK_RESULT int arena_allocator_init(ArenaAllocator *self);
+void arena_allocator_deinit(ArenaAllocator *self);
+CHECK_RESULT int arena_allocator_alloc(ArenaAllocator *self, usize size, void **mem);
+CHECK_RESULT int arena_allocator_add_mem(ArenaAllocator *self, usize *result_index);
+
+#endif
diff --git a/include/std/buffer.h b/include/std/buffer.h
index 7a74d33..ac722b1 100644
--- a/include/std/buffer.h
+++ b/include/std/buffer.h
@@ -13,11 +13,11 @@ typedef struct {
     usize size;
     usize capacity;
 
-    ScopedAllocator *allocator;
+    ArenaAllocator *allocator;
     usize allocator_index;
 } Buffer;
 
-CHECK_RESULT int buffer_init(Buffer *self, struct ScopedAllocator *allocator);
+CHECK_RESULT int buffer_init(Buffer *self, struct ArenaAllocator *allocator);
 void buffer_deinit(Buffer *self);
 
 /*
diff --git a/include/std/defs.h b/include/std/defs.h
index ee049d4..8a63a69 100644
--- a/include/std/defs.h
+++ b/include/std/defs.h
@@ -4,7 +4,7 @@
 typedef struct amal_thread amal_thread;
 typedef struct amal_mutex amal_mutex;
 
-typedef struct ScopedAllocatorNode ScopedAllocatorNode;
-typedef struct ScopedAllocator ScopedAllocator;
+typedef struct ArenaAllocatorNode ArenaAllocatorNode;
+typedef struct ArenaAllocator ArenaAllocator;
 
 #endif
diff --git a/include/std/hash_map.h b/include/std/hash_map.h
index 9b31213..b9b90c6 100644
--- a/include/std/hash_map.h
+++ b/include/std/hash_map.h
@@ -13,7 +13,7 @@ typedef int(*HashMapCompare)(const void *a, const void *b);
 typedef usize(*HashMapHash)(const u8 *data, usize size);
 
 struct HashMap {
-    ScopedAllocator *allocator; /* borrowed */
+    ArenaAllocator *allocator; /* borrowed */
     Buffer/*HashMapBucket*/ buckets;
     usize value_type_size;
     usize num_elements;
@@ -21,7 +21,7 @@ struct HashMap {
     HashMapHash hash_func;
 };
 
-CHECK_RESULT int hash_map_init(HashMap *self, ScopedAllocator *allocator, usize value_type_size, HashMapCompare compare_func, HashMapHash hash_func);
+CHECK_RESULT int hash_map_init(HashMap *self, ArenaAllocator *allocator, usize value_type_size, HashMapCompare compare_func, HashMapHash hash_func);
 /*
 Not thread-safe.
 Expected @value size to be @self->value_type_size.
diff --git a/include/std/mem.h b/include/std/mem.h
index ddd2b31..2afce0d 100644
--- a/include/std/mem.h
+++ b/include/std/mem.h
@@ -9,4 +9,6 @@ int am_memcmp(const void *lhs, const void *rhs, usize size);
 bool am_memeql(const void *lhs, const void *rhs, usize size);
 void am_memset(void *dest, int value, usize size);
 
+long am_pagesize();
+
 #endif
diff --git a/include/std/misc.h b/include/std/misc.h
index b1932d5..2fbe92c 100644
--- a/include/std/misc.h
+++ b/include/std/misc.h
@@ -23,7 +23,7 @@
     do { \
         int return_if_result; \
         return_if_result = (result); \
-        if((return_if_result) != 0) { \
+        if(return_if_result != 0) { \
             return_if_debug_msg; \
             return return_if_result; \
         } \
diff --git a/include/std/scoped_allocator.h b/include/std/scoped_allocator.h
deleted file mode 100644
index f037ea5..0000000
--- a/include/std/scoped_allocator.h
+++ /dev/null
@@ -1,29 +0,0 @@
-#ifndef AMALGAM_SCOPED_ALLOCATOR_H
-#define AMALGAM_SCOPED_ALLOCATOR_H
-
-#include "defs.h"
-#include "misc.h"
-#include "types.h"
-#include "buffer.h"
-
-struct ScopedAllocatorNode {
-    char *data;
-    usize size;
-    ScopedAllocatorNode *next;
-};
-
-struct ScopedAllocator {
-    ScopedAllocatorNode head;
-    ScopedAllocatorNode *current;
-    Buffer/*<void*>*/ mems;
-};
-
-CHECK_RESULT int scoped_allocator_node_init(ScopedAllocatorNode *self);
-void scoped_allocator_node_deinit(ScopedAllocatorNode *self);
-
-CHECK_RESULT int scoped_allocator_init(ScopedAllocator *self);
-void scoped_allocator_deinit(ScopedAllocator *self);
-CHECK_RESULT int scoped_allocator_alloc(ScopedAllocator *self, usize size, void **mem);
-CHECK_RESULT int scoped_allocator_add_mem(ScopedAllocator *self, usize *result_index);
-
-#endif
diff --git a/include/tokenizer.h b/include/tokenizer.h
index 4018e3a..f624816 100644
--- a/include/tokenizer.h
+++ b/include/tokenizer.h
@@ -60,7 +60,7 @@ typedef struct {
         BinopType binop_type;
     } value;
     bool number_is_integer;
-    ScopedAllocator *allocator; /* borrowed */
+    ArenaAllocator *allocator; /* borrowed */
     const amal_compiler_options *compiler_options; /* borrowed */
 } Tokenizer;
 
@@ -69,7 +69,7 @@ typedef struct {
     char* str;
 } TokenizerError;
 
-CHECK_RESULT int tokenizer_init(Tokenizer *self, ScopedAllocator *allocator, BufferView code, BufferView code_name, const amal_compiler_options *compiler_options);
+CHECK_RESULT int tokenizer_init(Tokenizer *self, ArenaAllocator *allocator, BufferView code, BufferView code_name, const amal_compiler_options *compiler_options);
 CHECK_RESULT int tokenizer_accept(Tokenizer *self, Token expected_token);
 /* 
    @result is set to 0 if the next token is equal to @expected_token,
diff --git a/src/asm/x86_64.c b/src/asm/x86_64.c
new file mode 100644
index 0000000..2cbeead
--- /dev/null
+++ b/src/asm/x86_64.c
@@ -0,0 +1,383 @@
+#include "../../include/asm/x86_64.h"
+
+#include "../../include/std/mem.h"
+#include "../../include/std/log.h"
+#include <errno.h>
+#include <assert.h>
+#include <stdio.h>
+
+#include <sys/mman.h>
+
+#define REX_W 0x48
+
+void asm_ptr_init(AsmPtr *self, Reg64 base) {
+    self->base = base;
+    self->index = -1;
+    self->disp = 0;
+    self->scale = 0;
+}
+
+void asm_ptr_init_index(AsmPtr *self, Reg64 base, Reg64 index) {
+    self->base = base;
+    self->index = index;
+    self->disp = 0;
+    self->scale = 0;
+}
+
+void asm_ptr_init_disp(AsmPtr *self, Reg64 base, i32 disp) {
+    self->base = base;
+    self->index = -1;
+    self->disp = disp;
+    self->scale = 0;
+}
+
+void asm_ptr_init_index_disp(AsmPtr *self, Reg64 base, Reg64 index, i32 disp) {
+    self->base = base;
+    self->index = index;
+    self->disp = disp;
+    self->scale = 0;
+}
+
+int asm_init(Asm *self) {
+    self->size = am_pagesize();
+    amal_log_debug("asm: page size: %u", self->size);
+    self->code = mmap(NULL, self->size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+    if(self->code == MAP_FAILED)
+        return -errno;
+    self->code_it = self->code;
+    return 0;
+}
+
+void asm_deinit(Asm *self) {
+    if(self->code)
+        munmap(self->code, self->size);
+    self->code = NULL;
+    self->code_it = NULL;
+    self->size = 0;
+}
+
+static void asm_print_code_hex(Asm *self) {
+    u8 *ptr;
+    int off;
+    ptr = self->code;
+    off = 0;
+    while(ptr != self->code_it) {
+        printf("%02x", *ptr);
+        ++ptr;
+        ++off;
+        if(off == 8) {
+            putc('\n', stdout);
+            off = 0;
+        } else {
+            putc(' ', stdout);
+        }
+    }
+    if(off != 0)
+        putc('\n', stdout);
+}
+
+int asm_execute(Asm *self) {
+    void (*func)();
+    if(mprotect(self->code, self->size, PROT_READ | PROT_EXEC) != 0)
+        return -errno;
+
+    asm_print_code_hex(self);
+
+    /* TODO: Verify if this is valid on all platforms. According to ISO C standard it isn't? */
+    *(void**)(&func) = self->code;
+    func();
+    return 0;
+}
+
+/* TODO: See how this can be optimized */
+static CHECK_RESULT int asm_ensure_capacity(Asm *self, usize size) {
+    usize current_offset;
+    current_offset = (u8*)self->code_it - (u8*)self->code;
+    if(current_offset + size > self->size) {
+        void *new_mem;
+        usize new_size;
+        new_size = self->size + am_pagesize();
+        new_mem = mmap(NULL, new_size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+        if(self->code == MAP_FAILED)
+            return -errno;
+
+        am_memcpy(new_mem, self->code, self->size);
+        self->code = new_mem;
+        self->size = new_size;
+        self->code_it = (u8*)self->code + current_offset;
+    }
+    return 0;
+}
+
+static isize asm_get_capacity_left(Asm *self) {
+    return (isize)self->size - (isize)((u8*)self->code_it - (u8*)self->code);
+}
+
+int asm_nop(Asm *self) {
+    return_if_error(asm_ensure_capacity(self, 1));
+    *self->code_it++ = 0x90;
+    return 0;
+}
+
+static i32 abs_i32(i32 value) {
+    return value >= 0 ? value : -value;
+}
+
+/*
+TODO: Implement 1 and 2 byte displacement?
+There has to be at least 6 bytes left in the asm buffer before calling this function.
+*/
+static void asm_rm(Asm *self, AsmPtr *mem, Reg64 reg) {
+    u8 rm_byte;
+    u8 disp_bytes;
+    assert(asm_get_capacity_left(self) >= 6);
+    if((int)mem->index != -1) {
+        u8 sib_offset;
+        if(mem->disp == 0) {
+            rm_byte = 0x04;
+            disp_bytes = 0;
+        } else if(abs_i32(mem->disp) <= INT8_MAX) {
+            rm_byte = 0x44;
+            disp_bytes = 1;
+        } else {
+            rm_byte = 0x84;
+            disp_bytes = 4;
+        }
+
+        #ifdef DEBUG
+        if(mem->scale != 0 && mem->scale != 2 && mem->scale != 4 && mem->scale != 8) {
+            amal_log_error("Invalid scale %d, expected 0, 2, 4, or 8", mem->scale);
+            assert(bool_false);
+        }
+        #endif
+        assert(mem->base != RBP && "TODO: Implement RBP base for sib byte. RBP is special and requires different logic");
+        sib_offset = (mem->scale << 5) + 8*mem->index + mem->base;
+
+        *self->code_it++ = rm_byte;
+        *self->code_it++ = sib_offset;
+    } else {
+        if(mem->disp == 0) {
+            if(mem->base == RBP) {
+                rm_byte = 0x45;
+                disp_bytes = 1;
+            } else {
+                rm_byte = mem->base;
+                disp_bytes = 0;
+            }
+        } else if(abs_i32(mem->disp) <= INT8_MAX) {
+            rm_byte = 0x40 + mem->base;
+            disp_bytes = 1;
+        } else {
+            rm_byte = 0x80 + mem->base;
+            disp_bytes = 4;
+        }
+        *self->code_it++ = (reg << 3) | rm_byte;
+    }
+
+    am_memcpy(self->code_it, &mem->disp, disp_bytes);
+    self->code_it += disp_bytes;
+}
+
+/* There has to be at least 1 byte left in the asm buffer before calling this function. */
+static void asm_rr(Asm *self, Reg64 dst, Reg64 src) {
+    assert(asm_get_capacity_left(self) >= 1);
+    *self->code_it++ = 0xC0 + dst + 8*src;
+}
+
+/* TODO: Implement 1 and 2 byte immediate? */
+int asm_mov_mi(Asm *self, AsmPtr *dst, i32 immediate) {
+    /* 12 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 12));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0xC7;
+    asm_rm(self, dst, 0);
+    am_memcpy(self->code_it, &immediate, sizeof(immediate));
+    self->code_it += sizeof(immediate);
+    return 0;
+}
+
+int asm_mov_mr(Asm *self, AsmPtr *dst, Reg64 src) {
+    /* 8 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 8));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0x89;
+    asm_rm(self, dst, src);
+    return 0;
+}
+
+int asm_mov_rm(Asm *self, Reg64 dst, AsmPtr *src) {
+    /* 8 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 8));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0x8B;
+    asm_rm(self, src, dst);
+    return 0;
+}
+
+int asm_mov_ri(Asm *self, Reg64 dst, i64 immediate) {
+    /* 10 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 10));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0xB8 + dst;
+    am_memcpy(self->code_it, &immediate, sizeof(immediate));
+    self->code_it += sizeof(immediate);
+    return 0;
+}
+
+int asm_mov_rr(Asm *self, Reg64 dst, Reg64 src) {
+    /* 3 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 3));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0x89;
+    asm_rr(self, dst, src);
+    return 0;
+}
+
+int asm_add_rr(Asm *self, Reg64 dst, Reg64 src) {
+    /* 3 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 3));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0x01;
+    asm_rr(self, dst, src);
+    return 0;
+}
+
+int asm_sub_rr(Asm *self, Reg64 dst, Reg64 src) {
+    /* 3 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 3));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0x29;
+    asm_rr(self, dst, src);
+    return 0;
+}
+
+int asm_imul_rr(Asm *self, Reg64 dst, Reg64 src) {
+    /* 3 bytes is the maximum size of the instruction. We don't how how large it will be so we prepare for the largest size */
+    return_if_error(asm_ensure_capacity(self, 4));
+    *self->code_it++ = REX_W;
+    *self->code_it++ = 0x0F;
+    *self->code_it++ = 0xAF;
+    asm_rr(self, dst, src);
+    return 0;
+}
+
+int asm_pushr(Asm *self, Reg64 reg) {
+    return_if_error(asm_ensure_capacity(self, 1));
+    *self->code_it++ = 0x50 + reg;
+    return 0;
+}
+
+int asm_popr(Asm *self, Reg64 reg) {
+    return_if_error(asm_ensure_capacity(self, 1));
+    *self->code_it++ = 0x58 + reg;
+    return 0;
+}
+
+/* /r */
+#define DEFINE_INS_RM(mnemonic, opcode)                         \
+int asm_##mnemonic##_rm32(Asm *self, Reg32 dst, Reg32 src) {    \
+    return_if_error(asm_ensure_capacity(self, 2));              \
+    *self->code_it++ = opcode;                             \
+    *self->code_it++ = 0xC0 + 8*dst + src;                 \
+    return 0;                                                   \
+}                                                               \
+                                                                \
+int asm_##mnemonic##_rm64(Asm *self, Reg64 dst, Reg64 src) {    \
+    return_if_error(asm_ensure_capacity(self, 1));              \
+    *self->code_it++ = REX_W;                               \
+    return asm_##mnemonic##_rm32(self, (Reg32)dst, (Reg32)src); \
+}
+
+DEFINE_INS_RM(mov, 0x8B)
+DEFINE_INS_RM(add, 0x03)
+DEFINE_INS_RM(sub, 0x2B)
+DEFINE_INS_RM(and, 0x23)
+DEFINE_INS_RM(or, 0x0B)
+DEFINE_INS_RM(xor, 0x33)
+DEFINE_INS_RM(cmp, 0x3B)
+
+/*
+    /number
+    The number is called the extension, a number from 0 to 7;
+    It's a number used to extend the opcode type, since the instruction only uses
+    one register the other register can be encoded for that.
+*/
+#define DEFINE_INS_EXT_IMM(mnemonic, extension)                      \
+int asm_##mnemonic##_rm32_imm(Asm *self, Reg32 reg, i32 immediate) { \
+    if(abs_i32(immediate) <= INT8_MAX) {                                 \
+        return_if_error(asm_ensure_capacity(self, 3));               \
+        *self->code_it++ = 0x83;                                \
+        *self->code_it++ = 0xC0 + 8*extension + reg;            \
+        *self->code_it++ = (u8)immediate;                       \
+    } else {                                                         \
+        return_if_error(asm_ensure_capacity(self, 6));               \
+        *self->code_it++ = 0x81;                                \
+        *self->code_it++ = 0xC0 + 8*extension + reg;            \
+        am_memcpy(self->code_it, &immediate, sizeof(immediate));     \
+        self->code_it += sizeof(immediate);                          \
+    }                                                                \
+    return 0;                                                        \
+}                                                                    \
+                                                                     \
+int asm_##mnemonic##_rm64_imm(Asm *self, Reg64 reg, i32 immediate) { \
+    return_if_error(asm_ensure_capacity(self, 1));                   \
+    *self->code_it++ = REX_W;                                    \
+    return asm_##mnemonic##_rm32_imm(self, (Reg32)reg, immediate);   \
+}
+
+DEFINE_INS_EXT_IMM(add, 0)
+DEFINE_INS_EXT_IMM(or, 1)
+DEFINE_INS_EXT_IMM(adc, 2)
+DEFINE_INS_EXT_IMM(sbb, 3)
+DEFINE_INS_EXT_IMM(and, 4)
+DEFINE_INS_EXT_IMM(sub, 5)
+DEFINE_INS_EXT_IMM(xor, 6)
+DEFINE_INS_EXT_IMM(cmp, 7)
+
+/*
+    /number
+    The number is called the extension, a number from 0 to 7;
+    It's a number used to extend the opcode type, since the instruction only uses
+    one register the other register can be encoded for that.
+*/
+#define DEFINE_INS_SHIFT_IMM8(mnemonic, extension)                  \
+int asm_##mnemonic##_rm32_imm(Asm *self, Reg32 reg, i8 immediate) { \
+    if(immediate == 1) {                                            \
+        return_if_error(asm_ensure_capacity(self, 2));              \
+        *self->code_it++ = 0xC1;                               \
+        *self->code_it++ = 0xC0 + 8*reg + extension;           \
+    } else {                                                        \
+        return_if_error(asm_ensure_capacity(self, 3));              \
+        *self->code_it++ = 0xD1;                               \
+        *self->code_it++ = 0xC0 + 8*reg + extension;           \
+        *self->code_it++ = immediate;                          \
+    }                                                               \
+    return 0;                                                       \
+}                                                                   \
+                                                                    \
+int asm_##mnemonic##_rm64_imm(Asm *self, Reg64 reg, i8 immediate) { \
+    return_if_error(asm_ensure_capacity(self, 1));                  \
+    *self->code_it++ = REX_W;                                   \
+    return asm_##mnemonic##_rm32_imm(self, (Reg32)reg, immediate);  \
+}
+
+DEFINE_INS_SHIFT_IMM8(rol, 0)
+DEFINE_INS_SHIFT_IMM8(ror, 1)
+DEFINE_INS_SHIFT_IMM8(rcl, 2)
+DEFINE_INS_SHIFT_IMM8(rcr, 3)
+DEFINE_INS_SHIFT_IMM8(shl, 4)
+DEFINE_INS_SHIFT_IMM8(shr, 5)
+/*DEFINE_INS_SHIFT_IMM8(shl, 6)*/
+DEFINE_INS_SHIFT_IMM8(sar, 7)
+
+int asm_ret(Asm *self, u16 bytes) {
+    if(bytes == 0) {
+        return_if_error(asm_ensure_capacity(self, 1));
+        *self->code_it++ = 0xC3;
+    } else {
+        return_if_error(asm_ensure_capacity(self, 3));
+        *self->code_it++ = 0xC2;
+        am_memcpy(self->code_it, &bytes, sizeof(bytes));
+    }
+    return 0;
+}
\ No newline at end of file
diff --git a/src/ast.c b/src/ast.c
index 69243f8..35ece2e 100644
--- a/src/ast.c
+++ b/src/ast.c
@@ -23,8 +23,8 @@ static void resolve_data_init(AstResolveData *self) {
     self->type = NULL;
 }
 
-int ast_create(ScopedAllocator *allocator, void *value, AstType type, Ast **result) {
-    return_if_error(scoped_allocator_alloc(allocator, sizeof(Ast), (void**)result));
+int ast_create(ArenaAllocator *allocator, void *value, AstType type, Ast **result) {
+    return_if_error(arena_allocator_alloc(allocator, sizeof(Ast), (void**)result));
     (*result)->value.data = value;
     (*result)->type = type;
     resolve_data_init(&(*result)->resolve_data);
@@ -75,18 +75,23 @@ static BufferView ast_get_code_reference(Ast *self) {
     return ast_get_name(self);
 }
 
-int funcdecl_init(FunctionDecl *self, FunctionSignature *signature, Scope *parent, ScopedAllocator *allocator) {
+void function_signature_init(FunctionSignature *self) {
+    self->params = 0;
+    self->resolved = bool_false;
+}
+
+int funcdecl_init(FunctionDecl *self, FunctionSignature *signature, Scope *parent, ArenaAllocator *allocator) {
     self->signature = signature;
     self->ssa_func_index = 0;
     return scope_init(&self->body, parent, allocator);
 }
 
-int funccall_init(FunctionCall *self, BufferView name, ScopedAllocator *allocator) {
+int funccall_init(FunctionCall *self, BufferView name, ArenaAllocator *allocator) {
     variable_init(&self->func, name);
     return buffer_init(&self->args, allocator);
 }
 
-int structdecl_init(StructDecl *self, Scope *parent, ScopedAllocator *allocator) {
+int structdecl_init(StructDecl *self, Scope *parent, ArenaAllocator *allocator) {
     return scope_init(&self->body, parent, allocator);
 }
 
@@ -139,24 +144,24 @@ void binop_init(Binop *self) {
     self->grouped = bool_false;
 }
 
-int if_statement_init(IfStatement *self, Scope *parent, ScopedAllocator *allocator) {
+int if_statement_init(IfStatement *self, Scope *parent, ArenaAllocator *allocator) {
     self->condition = NULL;
     self->else_if_stmt = NULL;
     return scope_init(&self->body, parent, allocator);
 }
 
-int else_if_statement_init(ElseIfStatement *self, Scope *parent, ScopedAllocator *allocator) {
+int else_if_statement_init(ElseIfStatement *self, Scope *parent, ArenaAllocator *allocator) {
     self->condition = NULL;
     self->next_else_if_stmt = NULL;
     return scope_init(&self->body, parent, allocator);
 }
 
-int while_statement_init(WhileStatement *self, Scope *parent, ScopedAllocator *allocator) {
+int while_statement_init(WhileStatement *self, Scope *parent, ArenaAllocator *allocator) {
     self->condition = NULL;
     return scope_init(&self->body, parent, allocator);
 }
 
-int scope_init(Scope *self, Scope *parent, ScopedAllocator *allocator) {
+int scope_init(Scope *self, Scope *parent, ArenaAllocator *allocator) {
     return_if_error(buffer_init(&self->ast_objects, allocator));
     return_if_error(hash_map_init(&self->named_objects, allocator, sizeof(Ast*), hash_map_compare_string, amal_hash_string));
     self->parent = parent;
@@ -164,7 +169,7 @@ int scope_init(Scope *self, Scope *parent, ScopedAllocator *allocator) {
     return 0;
 }
 
-int file_scope_reference_init(FileScopeReference *self, BufferView canonical_path, ScopedAllocator *allocator) {
+int file_scope_reference_init(FileScopeReference *self, BufferView canonical_path, ArenaAllocator *allocator) {
     char null_terminator;
     null_terminator = '\0';
     self->parser = NULL;
@@ -273,15 +278,17 @@ static Ast* scope_get_resolved_variable(Scope *self, AstCompilerContext *context
 
 static void function_signature_resolve(FunctionSignature *self, AstCompilerContext *context) {
     /* TODO: Implement */
+    if(self->resolved)
+        return;
+    self->resolved = bool_true;
     (void)self;
     (void)context;
 }
 
 static void variable_resolve(Variable *self, AstCompilerContext *context, AstResolveData *resolve_data) {
-    if(!self->resolved_var) {
+    if(!self->resolved_var)
         self->resolved_var = scope_get_resolved_variable(context->scope, context, self->name);
-        resolve_data->type = self->resolved_var->resolve_data.type;
-    }
+    resolve_data->type = self->resolved_var->resolve_data.type;
 }
 
 static LhsExpr* variable_get_resolved_type(Variable *self) {
@@ -289,15 +296,18 @@ static LhsExpr* variable_get_resolved_type(Variable *self) {
     return self->resolved_var->value.lhs_expr;
 }
 
-static void variable_type_resolve(VariableType *self, AstCompilerContext *context, AstResolveData *resolve_data) {
-    switch(self->type) {
+static void variable_type_resolve(LhsExpr *self, AstCompilerContext *context, AstResolveData *resolve_data) {
+    VariableType *lhs_expr_type;
+    lhs_expr_type = &self->type;
+    switch(lhs_expr_type->type) {
         case VARIABLE_TYPE_NONE:
             return;
         case VARIABLE_TYPE_VARIABLE:
-            variable_resolve(self->value.variable, context, resolve_data);
+            variable_resolve(lhs_expr_type->value.variable, context, resolve_data);
             break;
         case VARIABLE_TYPE_SIGNATURE:
-            function_signature_resolve(self->value.signature, context);
+            function_signature_resolve(lhs_expr_type->value.signature, context);
+            resolve_data->type = self;
             break;
     }
 }
@@ -333,13 +343,11 @@ static LhsExpr* lhsexpr_resolve_rhs(LhsExpr *self, AstCompilerContext *context)
 
 static void lhsexpr_resolve(Ast *ast, AstCompilerContext *context) {
     LhsExpr *self;
-    LhsExpr *rhs_resolve_type;
 
     assert(ast->type == AST_LHS);
     self = ast->value.lhs_expr;
-    rhs_resolve_type = NULL;
 
-    variable_type_resolve(&self->type, context, &ast->resolve_data);
+    variable_type_resolve(self, context, &ast->resolve_data);
 
     /*
     TODO: When parameters and return types are implemented, AST_RESOLVE_END should be set after
@@ -347,8 +355,22 @@ static void lhsexpr_resolve(Ast *ast, AstCompilerContext *context) {
     be allowed but recursive function calls still require parameters and return types to be known.
     */
     if(self->rhs_expr) {
-        if(self->rhs_expr->type == AST_FUNCTION_DECL)
+        LhsExpr *rhs_resolve_type;
+        if(self->rhs_expr->type == AST_FUNCTION_DECL) {
+            /*
+                The function declaration itself always resolves the signature, but we also do it here because we
+                want to have the signature solved before setting the lhs expr as solved. Also function signatures can exist
+                without lhs expr (anonymous function).
+            */
+            function_signature_resolve(self->rhs_expr->value.func_decl->signature, context);
             ast->resolve_data.status = AST_RESOLVED;
+            /*
+                If rhs is a function declaration then there is no need to wait until it has been resolved before setting the type as the type
+                is @self (the lhs). We still need to continue after this, so rhs can be resolved.
+            */
+            if(!ast->resolve_data.type)
+                ast->resolve_data.type = self;
+        }
 
         rhs_resolve_type = lhsexpr_resolve_rhs(self, context);
 
@@ -365,10 +387,8 @@ static void lhsexpr_resolve(Ast *ast, AstCompilerContext *context) {
                 "Variable type and variable assignment type (right-hand side) do not match");
             throw(AST_ERR);
         }
-    }
-
-    if(rhs_resolve_type)
         ast->resolve_data.type = rhs_resolve_type;
+    }
 }
 
 static LhsExpr* binop_get_lhs_expr(Binop *self) {
@@ -443,18 +463,20 @@ static void import_resolve(Ast *ast, AstCompilerContext *context) {
 
 static Scope* lhsexpr_get_scope(LhsExpr *self) {
     AstValue value;
-    value = self->rhs_expr->value;
-    switch(self->rhs_expr->type) {
-        case AST_FUNCTION_DECL:
-            return &value.func_decl->body;
-        case AST_STRUCT_DECL:
-            return &value.struct_decl->body;
-        case AST_IMPORT:
-            return &value.import->file_scope->parser->struct_decl.body;
-        default:
-            break;
+    if(self->rhs_expr) {
+        value = self->rhs_expr->value;
+        switch(self->rhs_expr->type) {
+            case AST_FUNCTION_DECL:
+                return &value.func_decl->body;
+            case AST_STRUCT_DECL:
+                return &value.struct_decl->body;
+            case AST_IMPORT:
+                return &value.import->file_scope->parser->struct_decl.body;
+            default:
+                break;
+        }
     }
-    assert(bool_false && "Expected lhsexpr_get_scope to only be called for function decl, struct decl and import");
+    assert(bool_false && "Expected lhsexpr_get_scope to only be called for non-extern function declaration, struct declaration and import");
     return NULL;
 }
 
@@ -466,9 +488,20 @@ static Parser* get_resolved_type_parser(Ast *self) {
 
 static void funcdecl_resolve(FunctionDecl *self, AstCompilerContext *context) {
     /* TODO: Implement parameters and return types */
+    function_signature_resolve(self->signature, context);
     scope_resolve(&self->body, context);
 }
 
+/*
+    Dont need to check if @self is resolved, since it will always be partially resolved when called from @funccall_resolve.
+    Meaning the resolve status wont be set to solved but the resolve type will be set.
+*/
+static bool is_func_decl(Ast *self) {
+    const LhsExpr *resolved_type = self->resolve_data.type;
+    return (resolved_type->rhs_expr && resolved_type->rhs_expr->type == AST_FUNCTION_DECL) ||
+            resolved_type->type.type == VARIABLE_TYPE_SIGNATURE;
+}
+
 static void funccall_resolve(Ast *self, AstCompilerContext *context) {
     FunctionCall *func_call;
     Ast **ast;
@@ -477,7 +510,7 @@ static void funccall_resolve(Ast *self, AstCompilerContext *context) {
     func_call = self->value.func_call;
     variable_resolve(&func_call->func, context, &self->resolve_data);
     /* Attemping to use call syntax (variable_name ( ) ) with a variable that is not a function */
-    if(self->resolve_data.type->rhs_expr->type != AST_FUNCTION_DECL) {
+    if(!is_func_decl(self)) {
         Parser *caller_parser;
         Parser *callee_parser;
         BufferView callee_code_ref;
@@ -514,6 +547,12 @@ static void structfield_resolve(Ast *self, AstCompilerContext *context) {
     variable_resolve(&struct_field->type, context, &self->resolve_data);
 }
 
+static bool is_struct_decl(Ast *self) {
+    const LhsExpr *resolved_type = self->resolve_data.type;
+    assert(self->resolve_data.status == AST_RESOLVED);
+    return resolved_type->rhs_expr && resolved_type->rhs_expr->type == AST_STRUCT_DECL;
+}
+
 static void binop_resolve_dot_access(Ast *ast, AstCompilerContext *context) {
     Binop *self;
     Scope *lhs_scope;
@@ -541,7 +580,7 @@ static void binop_resolve_dot_access(Ast *ast, AstCompilerContext *context) {
     caller_code_ref = ast_get_code_reference(self->rhs);
     callee_code_ref = self->rhs->resolve_data.type->var_name;
 
-    if(self->lhs->resolve_data.type->rhs_expr->type != AST_STRUCT_DECL) {
+    if(!is_struct_decl(self->lhs)) {
         parser_print_error(caller_parser, caller_code_ref.data, "Can only access field of structs");
         /* TODO: use tokenizer_print_note, once it has been added */
         /* TODO: Print type */
@@ -583,6 +622,8 @@ static void binop_resolve(Ast *ast, AstCompilerContext *context) {
             /*
             TODO: For this first error, only print the line without a reference to code.
             This requires change in tokenizer_print_error to be able to take a line as reference.
+
+            TODO: Use note for the additional information instead of error.
             */
             Parser *parser;
             parser = scope_get_parser(context->scope);
@@ -597,6 +638,14 @@ static void binop_resolve(Ast *ast, AstCompilerContext *context) {
                 "Right-hand side is of type %.*s",
                     self->rhs->resolve_data.type->var_name.size, self->rhs->resolve_data.type->var_name.data);
             throw(AST_ERR);
+        } else if(!is_arithmetic_type(self->lhs->resolve_data.type, context->compiler)) { /* TODO: Optimize this? store arithmetic type in the LhsExpr itself? */
+            /* TODO: Point the error at the binop instead of LHS */
+            Parser *parser;
+            parser = scope_get_parser(context->scope);
+            parser_print_error(parser, ast_get_code_reference(self->lhs).data,
+                "Arithmetic operation can only be performed with the types i8, u8, i16, u16, i32, u32, i64, u64, isize and usize",
+                    self->lhs->resolve_data.type->var_name.size, self->lhs->resolve_data.type->var_name.data);
+            throw(AST_ERR);
         }
         ast->resolve_data.type = self->lhs->resolve_data.type;
     }
@@ -655,9 +704,9 @@ void ast_resolve(Ast *self, AstCompilerContext *context) {
             number = self->value.number;
             /* TODO: Support other number types */
             if(number->is_integer)
-                self->resolve_data.type = context->compiler->default_types.i64;
+                self->resolve_data.type = (LhsExpr*)context->compiler->default_types.i64;
             else
-                self->resolve_data.type = context->compiler->default_types.f64;
+                self->resolve_data.type = (LhsExpr*)context->compiler->default_types.f64;
             break;
         }
         case AST_FUNCTION_DECL:
@@ -684,7 +733,7 @@ void ast_resolve(Ast *self, AstCompilerContext *context) {
             break;
         case AST_STRING:
             /* TODO: Convert special combinations. For example \n to newline */
-            self->resolve_data.type = context->compiler->default_types.str;
+            self->resolve_data.type = (LhsExpr*)context->compiler->default_types.str;
             break;
         case AST_VARIABLE:
             variable_resolve(self->value.variable, context, &self->resolve_data);
diff --git a/src/bytecode/bytecode.c b/src/bytecode/bytecode.c
index 9bf5f24..ffcd2e0 100644
--- a/src/bytecode/bytecode.c
+++ b/src/bytecode/bytecode.c
@@ -18,7 +18,7 @@
             throw(return_if_result); \
     } while(0)
 
-int bytecode_init(Bytecode *self, ScopedAllocator *allocator) {
+int bytecode_init(Bytecode *self, ArenaAllocator *allocator) {
     return buffer_init(&self->data, allocator);
 }
 
@@ -37,8 +37,8 @@ static CHECK_RESULT usize ssa_extract_form2(u8 *instruction_data, SsaInsForm2 *r
 
 static CHECK_RESULT usize ssa_extract_func_start(u8 *instruction_data, SsaInsFuncStart *result) {
     am_memcpy(&result->func_index, instruction_data, sizeof(result->func_index));
-    am_memcpy(&result->num_args, instruction_data + sizeof(result->func_index), sizeof(result->num_args));
-    return sizeof(result->func_index) + sizeof(result->num_args);
+    am_memcpy(&result->num_registers, instruction_data + sizeof(result->func_index), sizeof(result->num_registers));
+    return sizeof(result->func_index) + sizeof(result->num_registers);
 }
 
 static CHECK_RESULT usize ssa_extract_func_call(u8 *instruction_data, SsaInsFuncCall *result) {
@@ -109,7 +109,7 @@ void add_strings(BytecodeCompilerContext *self) {
     throw_if_error(buffer_append(instructions, &strings_size, sizeof(u32)));
     for(; string != strings_end; ++string) {
         throw_if_error(buffer_append(instructions, &string->size, sizeof(u16)));
-        throw_if_error(buffer_append(instructions, &string->data, string->size));
+        throw_if_error(buffer_append(instructions, string->data, string->size));
     }
 }
 
@@ -189,6 +189,20 @@ static void add_ins6(BytecodeCompilerContext *self, AmalOpcode opcode, u8 dst_re
     fputc('\n', stderr);
 }
 
+static void add_ins7(BytecodeCompilerContext *self, AmalOpcode opcode, u16 idx, u8 arg, const char *fmt) {
+    Buffer *instructions;
+    size_t index;
+    instructions = &self->bytecode.data;
+    index = instructions->size;
+    
+    throw_if_error(buffer_append_empty(instructions, sizeof(AmalOpcodeType) + sizeof(idx) + sizeof(arg)));
+    instructions->data[index] = opcode;
+    memcpy(instructions->data + index + sizeof(AmalOpcodeType), &idx, sizeof(idx));
+    instructions->data[index + sizeof(AmalOpcodeType) + sizeof(idx)] = arg;
+    fprintf(stderr, fmt, idx, arg);
+    fputc('\n', stderr);
+}
+
 #if 0
 #define NUM_MAX_REGS 256
 #define NUM_MAX_FUNC_ARGS 32
@@ -397,11 +411,21 @@ static void add_instructions(BytecodeCompilerContext *self) {
                 add_ins5(self, AMAL_OP_SUB, ssa_ins_form2.result, ssa_ins_form2.lhs, ssa_ins_form2.rhs, "sub r%d, r%d, r%d");
                 break;
             }
+            case SSA_IMUL: {
+                instruction += ssa_extract_form2(instruction, &ssa_ins_form2);
+                add_ins5(self, AMAL_OP_IMUL, ssa_ins_form2.result, ssa_ins_form2.lhs, ssa_ins_form2.rhs, "imul r%d, r%d, r%d");
+                break;
+            }
             case SSA_MUL: {
                 instruction += ssa_extract_form2(instruction, &ssa_ins_form2);
                 add_ins5(self, AMAL_OP_MUL, ssa_ins_form2.result, ssa_ins_form2.lhs, ssa_ins_form2.rhs, "mul r%d, r%d, r%d");
                 break;
             }
+            case SSA_IDIV: {
+                instruction += ssa_extract_form2(instruction, &ssa_ins_form2);
+                add_ins5(self, AMAL_OP_IDIV, ssa_ins_form2.result, ssa_ins_form2.lhs, ssa_ins_form2.rhs, "idiv r%d, r%d, r%d");
+                break;
+            }
             case SSA_DIV: {
                 instruction += ssa_extract_form2(instruction, &ssa_ins_form2);
                 add_ins5(self, AMAL_OP_DIV, ssa_ins_form2.result, ssa_ins_form2.lhs, ssa_ins_form2.rhs, "div r%d, r%d, r%d");
@@ -414,11 +438,11 @@ static void add_instructions(BytecodeCompilerContext *self) {
             }
             case SSA_FUNC_START: {
                 instruction += ssa_extract_func_start(instruction, &ssa_ins_func_start);
-                add_ins1(self, AMAL_OP_FUNC_START, "func_start");
+                add_ins4(self, AMAL_OP_FUNC_START, ssa_ins_func_start.num_registers, "func_start %u");
                 break;
             }
             case SSA_FUNC_END: {
-                add_ins1(self, AMAL_OP_FUNC_START, "func_end");
+                add_ins1(self, AMAL_OP_FUNC_END, "func_end");
                 break;
             }
             case SSA_PUSH: {
@@ -429,7 +453,6 @@ static void add_instructions(BytecodeCompilerContext *self) {
                 break;
             }
             case SSA_CALL: {
-                /* TODO: Add args, using number of bytes to pop after function call. */
                 /* 
                     TODO: Pass return register to function. The register should be a pointer that
                     has the size of the function return values so the return values can fit in it.
@@ -443,7 +466,8 @@ static void add_instructions(BytecodeCompilerContext *self) {
                     is defined as the size of all previous files' number of functions.
                 */
                 instruction += ssa_extract_func_call(instruction, &ssa_ins_func_call);
-                add_ins4(self, AMAL_OP_CALL, ssa_ins_func_call.func_decl->ssa_func_index, "call %d");
+                /* TODO: Replace 0 with the number of arguments */
+                add_ins7(self, AMAL_OP_CALL, ssa_ins_func_call.func_decl->ssa_func_index, 0, "call %d, %d");
                 break;
             }
             case SSA_JUMP_ZERO: {
@@ -456,9 +480,6 @@ static void add_instructions(BytecodeCompilerContext *self) {
                 add_ins4(self, AMAL_OP_JMP, ssa_ins_jump.jump_offset, "jmp %d");
                 break;
             }
-            default:
-                amal_log_error("Instruction not yet implemented: %d", ins);
-                assert(bool_false && "Instruction not yet implemented");
         }
     }
 
diff --git a/src/compiler.c b/src/compiler.c
index 05fae78..8c3266c 100644
--- a/src/compiler.c
+++ b/src/compiler.c
@@ -29,10 +29,10 @@ static CHECK_RESULT int create_default_type(amal_compiler *compiler, const char
     StructDecl *struct_decl;
     Ast *expr;
 
-    return_if_error(scoped_allocator_alloc(&compiler->allocator, sizeof(StructDecl), (void**)&struct_decl));
+    return_if_error(arena_allocator_alloc(&compiler->allocator, sizeof(StructDecl), (void**)&struct_decl));
     return_if_error(structdecl_init(struct_decl, &compiler->root_scope, &compiler->allocator));
 
-    return_if_error(scoped_allocator_alloc(&compiler->allocator, sizeof(LhsExpr), (void**)lhs_expr));
+    return_if_error(arena_allocator_alloc(&compiler->allocator, sizeof(LhsExpr), (void**)lhs_expr));
     lhsexpr_init(*lhs_expr, bool_true, bool_true, bool_true, create_buffer_view(name, strnlen(name, PATH_MAX)));
     return_if_error(ast_create(&compiler->allocator, struct_decl, AST_STRUCT_DECL, &(*lhs_expr)->rhs_expr));
     return_if_error(ast_create(&compiler->allocator, *lhs_expr, AST_LHS, &expr));
@@ -42,22 +42,59 @@ static CHECK_RESULT int create_default_type(amal_compiler *compiler, const char
 }
 
 static CHECK_RESULT int init_default_types(amal_compiler *compiler) {
-    return_if_error(create_default_type(compiler, "i8", &compiler->default_types.i8));
-    return_if_error(create_default_type(compiler, "i16", &compiler->default_types.i16));
-    return_if_error(create_default_type(compiler, "i32", &compiler->default_types.i32));
-    return_if_error(create_default_type(compiler, "i64", &compiler->default_types.i64));
-    return_if_error(create_default_type(compiler, "u8", &compiler->default_types.u8));
-    return_if_error(create_default_type(compiler, "u16", &compiler->default_types.u16));
-    return_if_error(create_default_type(compiler, "u32", &compiler->default_types.u32));
-    return_if_error(create_default_type(compiler, "u64", &compiler->default_types.u64));
-    return_if_error(create_default_type(compiler, "isize", &compiler->default_types.isize));
-    return_if_error(create_default_type(compiler, "usize", &compiler->default_types.usize));
-    return_if_error(create_default_type(compiler, "f32", &compiler->default_types.f32));
-    return_if_error(create_default_type(compiler, "f64", &compiler->default_types.f64));
-    return_if_error(create_default_type(compiler, "str", &compiler->default_types.str));
+    return_if_error(create_default_type(compiler, "i8", (LhsExpr**)&compiler->default_types.i8));
+    return_if_error(create_default_type(compiler, "i16", (LhsExpr**)&compiler->default_types.i16));
+    return_if_error(create_default_type(compiler, "i32", (LhsExpr**)&compiler->default_types.i32));
+    return_if_error(create_default_type(compiler, "i64", (LhsExpr**)&compiler->default_types.i64));
+    return_if_error(create_default_type(compiler, "u8", (LhsExpr**)&compiler->default_types.u8));
+    return_if_error(create_default_type(compiler, "u16", (LhsExpr**)&compiler->default_types.u16));
+    return_if_error(create_default_type(compiler, "u32", (LhsExpr**)&compiler->default_types.u32));
+    return_if_error(create_default_type(compiler, "u64", (LhsExpr**)&compiler->default_types.u64));
+    return_if_error(create_default_type(compiler, "isize", (LhsExpr**)&compiler->default_types.isize));
+    return_if_error(create_default_type(compiler, "usize", (LhsExpr**)&compiler->default_types.usize));
+    return_if_error(create_default_type(compiler, "f32", (LhsExpr**)&compiler->default_types.f32));
+    return_if_error(create_default_type(compiler, "f64", (LhsExpr**)&compiler->default_types.f64));
+    return_if_error(create_default_type(compiler, "str", (LhsExpr**)&compiler->default_types.str));
+
+    compiler->default_types.arithmetic_types[0] = compiler->default_types.i8;
+    compiler->default_types.arithmetic_types[1] = compiler->default_types.u8;
+    compiler->default_types.arithmetic_types[2] = compiler->default_types.i16;
+    compiler->default_types.arithmetic_types[3] = compiler->default_types.u16;
+    compiler->default_types.arithmetic_types[4] = compiler->default_types.i32;
+    compiler->default_types.arithmetic_types[5] = compiler->default_types.u32;
+    compiler->default_types.arithmetic_types[6] = compiler->default_types.i64;
+    compiler->default_types.arithmetic_types[7] = compiler->default_types.u64;
+    compiler->default_types.arithmetic_types[8] = compiler->default_types.isize;
+    compiler->default_types.arithmetic_types[9] = compiler->default_types.usize;
+
+    compiler->default_types.i8->is_signed = bool_true;
+    compiler->default_types.u8->is_signed = bool_false;
+    compiler->default_types.i16->is_signed = bool_true;
+    compiler->default_types.u16->is_signed = bool_false;
+    compiler->default_types.i32->is_signed = bool_true;
+    compiler->default_types.u32->is_signed = bool_false;
+    compiler->default_types.i64->is_signed = bool_true;
+    compiler->default_types.u64->is_signed = bool_false;
+    compiler->default_types.isize->is_signed = bool_true;
+    compiler->default_types.usize->is_signed = bool_false;
+    compiler->default_types.f32->is_signed = bool_true;
+    compiler->default_types.f64->is_signed = bool_true;
+    compiler->default_types.str->is_signed = bool_false;
     return 0;
 }
 
+bool is_arithmetic_type(LhsExpr *expr, amal_compiler *compiler) {
+    usize i;
+    const amal_default_types *default_types;
+    default_types = &compiler->default_types;
+
+    for(i = 0; i < NUM_ARITHMETIC_TYPES; ++i) {
+        if(expr == (LhsExpr*)default_types->arithmetic_types[i])
+            return bool_true;
+    }
+    return bool_false;
+}
+
 void amal_compiler_options_init(amal_compiler_options *self) {
     self->error_callback = NULL;
     self->error_callback_userdata = NULL;
@@ -89,12 +126,12 @@ static CHECK_RESULT int amal_compiler_init(amal_compiler *self, const amal_compi
     self->generic_work_object_index = 0;
     amal_mutex_init(&self->mutex);
 
-    return_if_error(scoped_allocator_init(&self->allocator));
+    return_if_error(arena_allocator_init(&self->allocator));
     cleanup_if_error(scope_init(&self->root_scope, NULL, &self->allocator));
     cleanup_if_error(buffer_init(&self->parsers, &self->allocator));
     cleanup_if_error(buffer_init(&self->queued_files, &self->allocator));
     cleanup_if_error(hash_map_init(&self->file_scopes, &self->allocator, sizeof(FileScopeReference*), hash_map_compare_string, amal_hash_string));
-    cleanup_if_error(scoped_allocator_alloc(&self->allocator,
+    cleanup_if_error(arena_allocator_alloc(&self->allocator,
                                             self->usable_thread_count * sizeof(ParserThreadData),
                                             (void**)&self->threads));
     for(i = 0; i < self->usable_thread_count; ++i)
@@ -114,7 +151,7 @@ void amal_compiler_deinit(amal_compiler *self) {
     }
 
     amal_mutex_deinit(&self->mutex);
-    scoped_allocator_deinit(&self->allocator);
+    arena_allocator_deinit(&self->allocator);
 }
 
 typedef enum {
@@ -145,7 +182,7 @@ typedef struct {
     ThreadWorkType type;
 } ThreadWorkData;
 
-static CHECK_RESULT int amal_compiler_load_in_this_thread(amal_compiler *compiler, FileScopeReference *file_scope, ScopedAllocator *allocator) {
+static CHECK_RESULT int amal_compiler_load_in_this_thread(amal_compiler *compiler, FileScopeReference *file_scope, ArenaAllocator *allocator) {
     Parser *parser;
     int result;
     BufferView filepath;
@@ -153,7 +190,7 @@ static CHECK_RESULT int amal_compiler_load_in_this_thread(amal_compiler *compile
 
     filepath = create_buffer_view(file_scope->canonical_path.data, file_scope->canonical_path.size);
     amal_log_info("Started parsing %.*s", filepath.size, filepath.data);
-    return_if_error(scoped_allocator_alloc(allocator, sizeof(Parser), (void**)&parser));
+    return_if_error(arena_allocator_alloc(allocator, sizeof(Parser), (void**)&parser));
     return_if_error(parser_init(parser, compiler, allocator));
     file_scope->parser = parser;
     return_if_error(parser_parse_file(parser, filepath));
@@ -229,7 +266,7 @@ static CHECK_RESULT int thread_generate_ssa(Parser *parser) {
     SsaCompilerContext compiler_context;
     int result;
 
-    return_if_error(scoped_allocator_alloc(parser->allocator, sizeof(Ssa), (void**)&compiler_context.ssa));
+    return_if_error(arena_allocator_alloc(parser->allocator, sizeof(Ssa), (void**)&compiler_context.ssa));
     return_if_error(ssa_init(compiler_context.ssa, parser->allocator));
     compiler_context.compiler = parser->compiler;
     parser->ssa = compiler_context.ssa;
@@ -446,7 +483,10 @@ static CHECK_RESULT int amal_compiler_dispatch_generic(amal_compiler *self, Thre
 }
 
 static CHECK_RESULT int amal_compiler_generate_program(amal_compiler *self) {
-    /* TODO: Copying the bytecode to the program can be done using multiple threads */
+    /*
+        TODO: Copying the bytecode to the program can be done using multiple threads.
+        Use self->threads for that.
+    */
     Parser **parser;
     Parser **parser_end;
     parser = buffer_begin(&self->parsers);
@@ -472,7 +512,7 @@ static CHECK_RESULT int try_create_file_scope(amal_compiler *compiler, const cha
     path_view = create_buffer_view(result_path, result_path_size);
     cleanup_if_error(amal_mutex_lock(&compiler->mutex, "try_create_file_scope"));
     if(!hash_map_get(&compiler->file_scopes, path_view, file_scope)) {
-        cleanup_if_error(scoped_allocator_alloc(&compiler->allocator, sizeof(FileScopeReference), (void**)file_scope));
+        cleanup_if_error(arena_allocator_alloc(&compiler->allocator, sizeof(FileScopeReference), (void**)file_scope));
         /* @(*file_scope)->canonical_path won't change after this, so it's fine if allocator belongs to non-thread safe compiler instance */
         cleanup_if_error(file_scope_reference_init(*file_scope, path_view, &compiler->allocator));
         cleanup_if_error(hash_map_insert(&compiler->file_scopes, path_view, file_scope));
@@ -487,11 +527,11 @@ static CHECK_RESULT int try_create_file_scope(amal_compiler *compiler, const cha
 }
 
 int amal_compiler_load_file(amal_compiler_options *options, amal_program *program, const char *filepath) {
-    assert(program);
-    assert(filepath);
     amal_compiler compiler;
     FileScopeReference *file_scope;
     int result;
+    assert(program);
+    assert(filepath);
     return_if_error(amal_compiler_init(&compiler, options, program));
     result = amal_compiler_internal_load_file(&compiler, filepath, &file_scope);
     amal_compiler_deinit(&compiler);
diff --git a/src/parser.c b/src/parser.c
index b61a968..fdf34ce 100644
--- a/src/parser.c
+++ b/src/parser.c
@@ -30,12 +30,12 @@ int parser_thread_data_init(ParserThreadData *self) {
     am_memset(&self->allocator, 0, sizeof(self->allocator));
     am_memset(&self->thread, 0, sizeof(self->thread));
     self->status = PARSER_THREAD_STATUS_NEW;
-    return scoped_allocator_init(&self->allocator);
+    return arena_allocator_init(&self->allocator);
 }
 
 void parser_thread_data_deinit(ParserThreadData *self) {
     ignore_result_int(amal_thread_deinit(&self->thread));
-    scoped_allocator_deinit(&self->allocator);
+    arena_allocator_deinit(&self->allocator);
 }
 
 int parser_thread_data_start(ParserThreadData *self, AmalThreadCallbackFunc callback_func, void *userdata) {
@@ -51,7 +51,7 @@ int parser_thread_data_join(ParserThreadData *self, void **result) {
     return amal_thread_join(&self->thread, result);
 }
 
-int parser_init(Parser *self, amal_compiler *compiler, ScopedAllocator *allocator) {
+int parser_init(Parser *self, amal_compiler *compiler, ArenaAllocator *allocator) {
     self->allocator = allocator;
     self->compiler = compiler;
     self->ssa = NULL;
@@ -151,15 +151,15 @@ static CHECK_RESULT FunctionSignature* parser_parse_function_signature(Parser *s
         /* TODO: Parse return types */
     }
 
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(FunctionSignature), (void**)&signature));
-    signature->params = 0;
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(FunctionSignature), (void**)&signature));
+    function_signature_init(signature);
     return signature;
 }
 
 /*
 VAR_TYPE_DEF = ':' TOK_IDENTIFIER|FUNC_SIGNATURE
 */
-static CHECK_RESULT int parser_parse_var_type_def(Parser *self, VariableType *result) {
+static void parser_parse_var_type_def(Parser *self, VariableType *result) {
     bool match;
 
     result->type = VARIABLE_TYPE_NONE;
@@ -167,14 +167,14 @@ static CHECK_RESULT int parser_parse_var_type_def(Parser *self, VariableType *re
 
     throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_COLON, &match));
     if(!match)
-        return -1;
+        return;
 
     throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_IDENTIFIER, &match));
     if(match) {
         result->type = VARIABLE_TYPE_VARIABLE;
-        throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(Variable), (void**)&result->value.variable));
+        throw_if_error(arena_allocator_alloc(self->allocator, sizeof(Variable), (void**)&result->value.variable));
         variable_init(result->value.variable, self->tokenizer.value.identifier);
-        return 0;
+        return;
     }
 
     result->type = VARIABLE_TYPE_SIGNATURE;
@@ -185,7 +185,6 @@ static CHECK_RESULT int parser_parse_var_type_def(Parser *self, VariableType *re
                                             "Expected type or closure signature");
         throw(PARSER_UNEXPECTED_TOKEN);
     }
-    return 0;
 }
 
 /*
@@ -232,10 +231,10 @@ static CHECK_RESULT LhsExpr* parser_parse_declaration_lhs(Parser *self) {
 
     throw_if_error(tokenizer_accept(&self->tokenizer, TOK_IDENTIFIER));
     var_name = self->tokenizer.value.identifier;
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(LhsExpr), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(LhsExpr), (void**)&result));
     lhsexpr_init(result, is_extern, is_pub, is_const, var_name);
 
-    ignore_result_int(parser_parse_var_type_def(self, &result->type));
+    parser_parse_var_type_def(self, &result->type);
     return result;
 }
 
@@ -252,7 +251,7 @@ static CHECK_RESULT FunctionDecl* parser_parse_closure(Parser *self) {
     if(!signature)
         return NULL;
 
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(FunctionDecl), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(FunctionDecl), (void**)&result));
     throw_if_error(funcdecl_init(result, signature, self->current_scope, self->allocator));
 
     self->current_scope = &result->body;
@@ -277,7 +276,7 @@ static CHECK_RESULT StructDecl* parser_parse_struct_decl(Parser *self) {
     if(!match)
         return result;
 
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(StructDecl), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(StructDecl), (void**)&result));
     throw_if_error(structdecl_init(result, self->current_scope, self->allocator));
 
     self->current_scope = &result->body;
@@ -329,13 +328,13 @@ static CHECK_RESULT Ast* parser_parse_function_call_or_variable(Parser *self) {
     throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_OPEN_PAREN, &match));
     if(!match) {
         Variable *variable;
-        throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(Variable), (void**)&variable));
+        throw_if_error(arena_allocator_alloc(self->allocator, sizeof(Variable), (void**)&variable));
         variable_init(variable, identifier);
         throw_if_error(ast_create(self->allocator, variable, AST_VARIABLE, &result));
         return result;
     }
     
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(FunctionCall), (void**)&func_call));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(FunctionCall), (void**)&func_call));
     throw_if_error(funccall_init(func_call, self->tokenizer.value.identifier, self->allocator));
     throw_if_error(ast_create(self->allocator, func_call, AST_FUNCTION_CALL, &result));
     /* Ends after TOK_CLOSING_PAREN */
@@ -355,7 +354,7 @@ static CHECK_RESULT Import* parser_parse_import(Parser *self) {
     if(!match)
         return result;
     
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(Import), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(Import), (void**)&result));
     import_init(result, self->tokenizer.value.string);
     return result;
 }
@@ -371,7 +370,7 @@ static CHECK_RESULT ElseIfStatement* parser_parse_else_if_statement(Parser *self
     if(!match)
         return NULL;
 
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(ElseIfStatement), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(ElseIfStatement), (void**)&result));
     throw_if_error(else_if_statement_init(result, self->current_scope, self->allocator));
 
     throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_IF, &match));
@@ -417,7 +416,7 @@ static CHECK_RESULT IfStatement* parser_parse_if_statement(Parser *self) {
     if(!match)
         return NULL;
 
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(IfStatement), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(IfStatement), (void**)&result));
     throw_if_error(if_statement_init(result, self->current_scope, self->allocator));
 
     result->condition = parser_parse_rhs_binop(self);
@@ -446,7 +445,7 @@ static CHECK_RESULT WhileStatement* parser_parse_while_statement(Parser *self) {
     if(!match)
         return NULL;
 
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(WhileStatement), (void**)&result));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(WhileStatement), (void**)&result));
     throw_if_error(while_statement_init(result, self->current_scope, self->allocator));
 
     result->condition = parser_parse_rhs_binop(self);
@@ -473,7 +472,7 @@ static CHECK_RESULT Ast* parser_parse_number(Parser *self) {
     if(!match)
         return result;
     
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(Number), (void**)&number));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(Number), (void**)&number));
     number_init(number, self->tokenizer.value.integer, self->tokenizer.number_is_integer,
         create_buffer_view(self->tokenizer.code.data + self->tokenizer.prev_index, self->tokenizer.index - self->tokenizer.prev_index));
     throw_if_error(ast_create(self->allocator, number, AST_NUMBER, &result));
@@ -491,7 +490,7 @@ static Ast* parser_parse_rhs_single_expr(Parser *self) {
     throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_STRING, &match));
     if(match) {
         String *string;
-        throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(String), (void**)&string));
+        throw_if_error(arena_allocator_alloc(self->allocator, sizeof(String), (void**)&string));
         throw_if_error(string_init(string, self->tokenizer.value.string));
         throw_if_error(ast_create(self->allocator, string, AST_STRING, &result));
         return result;
@@ -549,7 +548,7 @@ Ast* parser_parse_rhs_binop(Parser *self) {
     binop_type = self->tokenizer.value.binop_type;
 
     rhs = parser_parse_rhs_binop(self);
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(Binop), (void**)&binop));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(Binop), (void**)&binop));
     binop_init(binop);
     binop->type = binop_type;
     binop->lhs = lhs;
@@ -645,11 +644,23 @@ Ast* parser_parse_body(Parser *self) {
         throw_if_error(ast_create(self->allocator, lhs_expr, AST_LHS, &result));
         if(lhs_expr->is_extern) {
             throw_if_error(tokenizer_accept(&self->tokenizer, TOK_SEMICOLON));
+            if (lhs_expr->type.type == VARIABLE_TYPE_NONE) {
+                self->error = tokenizer_create_error(&self->tokenizer, self->tokenizer.prev_index, "A variable can't be declared without a type or assignment");
+                throw(PARSER_UNEXPECTED_TOKEN);
+            }
             return result;
         } else {
             throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_SEMICOLON, &match));
-            if(match)
+            if(match) {
+                if(lhs_expr->type.type == VARIABLE_TYPE_SIGNATURE) {
+                    self->error = tokenizer_create_error(&self->tokenizer, self->tokenizer.prev_index, "Expected function declaration. Only extern functions can have empty declarations.");
+                    throw(PARSER_UNEXPECTED_TOKEN);
+                } else if (lhs_expr->type.type == VARIABLE_TYPE_NONE) {
+                    self->error = tokenizer_create_error(&self->tokenizer, self->tokenizer.prev_index, "A variable can't be declared without a type or assignment");
+                    throw(PARSER_UNEXPECTED_TOKEN);
+                }
                 return result;
+            }
         }
 
         throw_if_error(tokenizer_accept(&self->tokenizer, TOK_EQUALS));
@@ -683,14 +694,16 @@ Ast* parser_parse_body(Parser *self) {
     self->error_context = ERROR_CONTEXT_RHS_STANDALONE;
     rhs_expr = parser_parse_rhs(self);
     self->error_context = ERROR_CONTEXT_NONE;
+    /* Variable declaration with lhs and rhs */
     if(lhs_expr) {
         lhs_expr->rhs_expr = rhs_expr;
     } else {
         bool match;
         throw_if_error(tokenizer_consume_if(&self->tokenizer, TOK_EQUALS, &match));
+        /* Variable assignment */
         if(match) {
             AssignmentExpr *assign_expr;
-            throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(AssignmentExpr), (void**)&assign_expr));
+            throw_if_error(arena_allocator_alloc(self->allocator, sizeof(AssignmentExpr), (void**)&assign_expr));
             throw_if_error(ast_create(self->allocator, assign_expr, AST_ASSIGN, &result));
             assign_expr->lhs_expr = rhs_expr;
 
@@ -698,6 +711,7 @@ Ast* parser_parse_body(Parser *self) {
             assign_expr->rhs_expr = parser_parse_rhs(self);
             self->error_context = ERROR_CONTEXT_NONE;
         } else {
+            /* Only rhs (for example function call, binop etc...) */
             result = rhs_expr;
         }
     }
@@ -721,7 +735,7 @@ Ast* parser_parse_struct_body(Parser *self) {
     throw_if_error(tokenizer_accept(&self->tokenizer, TOK_IDENTIFIER));
     type_name = self->tokenizer.value.identifier;
     throw_if_error(tokenizer_accept(&self->tokenizer, TOK_SEMICOLON));
-    throw_if_error(scoped_allocator_alloc(self->allocator, sizeof(LhsExpr), (void**)&struct_field));
+    throw_if_error(arena_allocator_alloc(self->allocator, sizeof(LhsExpr), (void**)&struct_field));
     structfield_init(struct_field, var_name, type_name);
     throw_if_error(ast_create(self->allocator, struct_field, AST_STRUCT_FIELD, &result));
     return result;
diff --git a/src/program.c b/src/program.c
index 167f4c4..5e63c04 100644
--- a/src/program.c
+++ b/src/program.c
@@ -2,6 +2,7 @@
 #include "../include/std/mem.h"
 #include "../include/std/alloc.h"
 #include "../include/std/log.h"
+#include "../include/std/buffer_view.h"
 #include <stdio.h>
 #include <errno.h>
 #include <assert.h>
@@ -19,13 +20,12 @@ typedef union {
     f64 floating;
 } NumberUnion;
 
-int amal_program_init(amal_program *self) {
-    ignore_result_int(buffer_init(&self->data, NULL));
-    ignore_result_int(buffer_init(&self->string_indices, NULL));
-    self->intermediates_start = NULL;
-    self->strings_start = NULL;
-    self->read_index = 0;
+typedef struct {
+    NumberType type;
+    NumberUnion value;
+} Number;
 
+static CHECK_RESULT int amal_program_append_header(amal_program *self) {
     /*doc(Bytecode header)
         # Header layout
         |Size|Name         |Description                                                                 |
@@ -50,9 +50,37 @@ int amal_program_init(amal_program *self) {
     return 0;
 }
 
+int amal_program_init(amal_program *self) {
+    ignore_result_int(buffer_init(&self->data, NULL));
+    self->string_indices = NULL;
+    self->intermediates_start = NULL;
+    self->strings_start = NULL;
+    self->read_index = 0;
+    self->num_strings = 0;
+    self->num_intermediates = 0;
+    am_memset(self->reg, 0, sizeof(self->reg));
+    self->stack = NULL;
+    self->stack_size = 4096;
+    am_memset(&self->asm, 0, sizeof(self->asm));
+    cleanup_if_error(am_malloc(self->stack_size, (void**)&self->stack));
+    cleanup_if_error(asm_init(&self->asm));
+    self->stack_index = 0;
+
+    cleanup_if_error(amal_program_append_header(self));
+    return 0;
+
+    cleanup:
+    amal_program_deinit(self);
+    return -1;
+}
+
 void amal_program_deinit(amal_program *self) {
     buffer_deinit(&self->data);
-    buffer_deinit(&self->string_indices);
+    am_free(self->string_indices);
+    self->string_indices = NULL;
+    am_free(self->stack);
+    self->stack = NULL;
+    asm_deinit(&self->asm);
 }
 
 int amal_program_append_bytecode(amal_program *self, Bytecode *bytecode) {
@@ -113,6 +141,8 @@ static CHECK_RESULT int amal_program_read_intermediates(amal_program *self) {
     }
 
     self->intermediates_start = &self->data.data[self->read_index];
+    if(intermediates_size > 0)
+        self->num_intermediates = intermediates_size / (sizeof(u8) + sizeof(u64));
     /*
     read_end = self->read_index + intermediates_size;
     while(self->read_index < read_end) {
@@ -140,10 +170,11 @@ static CHECK_RESULT int amal_program_read_strings(amal_program *self) {
 
     am_memcpy(&num_strings, &self->data.data[self->read_index], sizeof(num_strings));
     self->read_index += sizeof(num_strings);
+    self->num_strings = num_strings;
 
-    if(buffer_append_empty(&self->string_indices, num_strings) != 0)
+    if(am_malloc(sizeof(u32) * num_strings, (void**)&self->string_indices) != 0)
         return AMAL_PROGRAM_STRING_ALLOC_FAILURE;
-    string_index_ptr = (u32*)self->string_indices.data;
+    string_index_ptr = self->string_indices;
 
     if(bytes_left_to_read(self) < sizeof(strings_size))
         return AMAL_PROGRAM_INVALID_STRINGS;
@@ -178,9 +209,58 @@ static CHECK_RESULT int amal_program_read_strings(amal_program *self) {
     return AMAL_PROGRAM_OK;
 }
 
+static CHECK_RESULT int amal_program_get_intermediate_by_index(amal_program *self, u16 index, Number *result) {
+    if(index >= self->num_intermediates)
+        return AMAL_PROGRAM_INSTRUCTION_INVALID_INTERMEDIATE_INDEX;
+    
+    am_memcpy(&result->type, &self->intermediates_start[(sizeof(u8) + sizeof(u64)) * (usize)index], sizeof(u8));
+    am_memcpy(&result->value, &self->intermediates_start[(sizeof(u8) + sizeof(u64)) * (usize)index + sizeof(u8)], sizeof(u64));
+    return 0;
+}
+
+static CHECK_RESULT int amal_program_get_data_by_index(amal_program *self, u16 index, char **result) {
+    char *str_ptr;
+
+    if(index >= self->num_strings) {
+        amal_log_error("Data index %ld is out of range (%ld)", index, self->num_strings);
+        return AMAL_PROGRAM_INSTRUCTION_INVALID_DATA_INDEX;
+    }
+    
+    str_ptr = self->strings_start + self->string_indices[index];
+    am_memcpy(result, &str_ptr, sizeof(char**));
+    return 0;
+}
+
+static CHECK_RESULT int ensure_stack_capacity_for_push(amal_program *self) {
+    if(self->stack_index >= self->stack_size) {
+        self->stack_size *= 2;
+        /* 4MB */
+        if(self->stack_size >= (1<<22))
+            return AMAL_PROGRAM_INSTRUCTION_STACK_OVERFLOW;
+        if(am_realloc(self->stack, self->stack_size, (void**)&self->stack) != 0)
+            return AMAL_PROGRAM_INSTRUCTION_STACK_OOM;
+    }
+    return 0;
+}
+
+static i64 abs_i64(i64 value) {
+    return value >= 0 ? value : -value;
+}
+
+static int assert_reg_outside_stack() {
+    assert(bool_false && "Register outside stack!");
+    return 0;
+}
+
 static CHECK_RESULT int amal_program_read_instructions(amal_program *self) {
     u32 instructions_size;
+    u32 read_start;
     u32 read_end;
+    bool inside_func;
+    u16 func_num_registers;
+
+    func_num_registers = 0;
+    inside_func = bool_false;
 
     if(bytes_left_to_read(self) < sizeof(instructions_size))
         return AMAL_PROGRAM_INVALID_INSTRUCTIONS_SIZE;
@@ -191,68 +271,296 @@ static CHECK_RESULT int amal_program_read_instructions(amal_program *self) {
     if(bytes_left_to_read(self) < instructions_size)
         return AMAL_PROGRAM_INVALID_INSTRUCTIONS_SIZE;
 
-    read_end = self->read_index + instructions_size;
+    /*
+        TODO: self->reg should be of type Number and each arithmetic operation should operate
+        on the type of the register.
+
+        TODO: Currently almost all operations are performed on memory. This should be optimized
+        to take advantage of registers.
+
+        TODO: Operations with memory registers could access outside the stack. Should this be checked?
+    */
+
+    #ifdef DEBUG
+    #define get_register_at_offset(offset) \
+        (self->data.data[self->read_index + (offset)] < func_num_registers ? self->data.data[self->read_index + (offset)] * (int)sizeof(usize) + (int)sizeof(usize) : assert_reg_outside_stack())
+    #else
+    #define get_register_at_offset(offset) (self->data.data[self->read_index + (offset)] * (int)sizeof(usize) + (int)sizeof(usize))
+    #endif
+
+    read_start = self->read_index;
+    read_end = read_start + instructions_size;
     while(self->read_index < read_end) {
         AmalOpcode opcode;
         opcode = self->data.data[self->read_index];
         self->read_index += sizeof(AmalOpcodeType);
         switch(opcode) {
-            case AMAL_OP_NOP:
+            case AMAL_OP_NOP: {
+                return_if_error(asm_nop(&self->asm));
                 break;
-            case AMAL_OP_SETZ:
+            }
+            case AMAL_OP_SETZ: {
+                AsmPtr dst;
+                asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                return_if_error(asm_mov_mi(&self->asm, &dst, 0));
+                self->reg[(u8)self->data.data[self->read_index]] = 0;
                 self->read_index += 1;
                 break;
-            case AMAL_OP_MOV:
+            }
+            case AMAL_OP_MOV: {
+                AsmPtr ptr;
+                asm_ptr_init_disp(&ptr, RBP, -(i32)get_register_at_offset(1));
+                return_if_error(asm_mov_rm(&self->asm, RAX, &ptr));
+
+                asm_ptr_init_disp(&ptr, RBP, -(i32)get_register_at_offset(0));
+                return_if_error(asm_mov_mr(&self->asm, &ptr, RAX));
+
+                self->reg[(u8)self->data.data[self->read_index]] = self->reg[(u8)self->data.data[self->read_index + 1]];
                 self->read_index += 2;
                 break;
-            case AMAL_OP_MOVI:
+            }
+            case AMAL_OP_MOVI: {
+                u8 dst_reg;
+                u16 intermediate_index;
+                Number number;
+
+                dst_reg = self->reg[(u8)self->data.data[self->read_index]];
+                am_memcpy(&intermediate_index, &self->data.data[self->read_index + sizeof(u8)], sizeof(intermediate_index));
+
+                return_if_error(amal_program_get_intermediate_by_index(self, intermediate_index, &number));
+                self->reg[dst_reg] = number.value.integer;
+
+                /* TODO: if @number is a float then use float instructions */
+                if(abs_i64(number.value.integer) <= INT32_MAX) {
+                    AsmPtr dst;
+                    asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                    return_if_error(asm_mov_mi(&self->asm, &dst, number.value.integer));
+                } else {
+                    AsmPtr dst;
+                    asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                    return_if_error(asm_mov_ri(&self->asm, RAX, number.value.integer));
+                    return_if_error(asm_mov_mr(&self->asm, &dst, RAX));
+                }
+
+                self->read_index += 3;
+                break;
+            }
+            case AMAL_OP_MOVD: {
+                u8 dst_reg;
+                u16 data_index;
+                char *data_ptr;
+                AsmPtr dst;
+
+                dst_reg = self->reg[(u8)self->data.data[self->read_index]];
+                am_memcpy(&data_index, &self->data.data[self->read_index + sizeof(u8)], sizeof(data_index));
+
+                return_if_error(amal_program_get_data_by_index(self, data_index, &data_ptr));
+                self->reg[dst_reg] = (uintptr_t)data_ptr;
+
+                asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                return_if_error(asm_mov_ri(&self->asm, RAX, (uintptr_t)data_ptr));
+                return_if_error(asm_mov_mr(&self->asm, &dst, RAX));
+
+                self->read_index += 3;
+                break;
+            }
+            case AMAL_OP_ADD: {
+                AsmPtr dst;
+                AsmPtr reg1;
+                AsmPtr reg2;
+
+                asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                asm_ptr_init_disp(&reg1, RBP, -(i32)get_register_at_offset(1));
+                asm_ptr_init_disp(&reg2, RBP, -(i32)get_register_at_offset(2));
+
+                return_if_error(asm_mov_rm(&self->asm, RAX, &reg1));
+                return_if_error(asm_mov_rm(&self->asm, RCX, &reg2));
+                return_if_error(asm_add_rr(&self->asm, RAX, RCX));
+                return_if_error(asm_mov_mr(&self->asm, &dst, RAX));
+
                 self->read_index += 3;
                 break;
-            case AMAL_OP_MOVD:
+            }
+            case AMAL_OP_SUB: {
+                AsmPtr dst;
+                AsmPtr reg1;
+                AsmPtr reg2;
+
+                asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                asm_ptr_init_disp(&reg1, RBP, -(i32)get_register_at_offset(1));
+                asm_ptr_init_disp(&reg2, RBP, -(i32)get_register_at_offset(2));
+
+                return_if_error(asm_mov_rm(&self->asm, RAX, &reg1));
+                return_if_error(asm_mov_rm(&self->asm, RCX, &reg2));
+                return_if_error(asm_sub_rr(&self->asm, RAX, RCX));
+                return_if_error(asm_mov_mr(&self->asm, &dst, RAX));
+
                 self->read_index += 3;
                 break;
-            case AMAL_OP_ADD:
+            }
+            case AMAL_OP_IMUL: {
+                AsmPtr dst;
+                AsmPtr reg1;
+                AsmPtr reg2;
+
+                asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                asm_ptr_init_disp(&reg1, RBP, -(i32)get_register_at_offset(1));
+                asm_ptr_init_disp(&reg2, RBP, -(i32)get_register_at_offset(2));
+
+                return_if_error(asm_mov_rm(&self->asm, RAX, &reg1));
+                return_if_error(asm_mov_rm(&self->asm, RCX, &reg2));
+                return_if_error(asm_imul_rr(&self->asm, RAX, RCX));
+                return_if_error(asm_mov_mr(&self->asm, &dst, RAX));
+                
                 self->read_index += 3;
                 break;
-            case AMAL_OP_SUB:
+            }
+            case AMAL_OP_MUL: {
+                #if 0
+                AsmPtr dst;
+                AsmPtr reg1;
+                AsmPtr reg2;
+
+                asm_ptr_init_disp(&dst, RBP, -(i32)get_register_at_offset(0));
+                asm_ptr_init_disp(&reg1, RBP, -(i32)get_register_at_offset(1));
+                asm_ptr_init_disp(&reg2, RBP, -(i32)get_register_at_offset(2));
+
+                return_if_error(asm_mov_rm(&self->asm, RAX, &reg1));
+                return_if_error(asm_mov_rm(&self->asm, RCX, &reg2));
+                return_if_error(asm_mul_rr(&self->asm, RAX, RCX));
+                return_if_error(asm_mov_mr(&self->asm, &dst, RAX));
+                #endif
+
                 self->read_index += 3;
                 break;
-            case AMAL_OP_MUL:
+            }
+            case AMAL_OP_IDIV: {
                 self->read_index += 3;
                 break;
-            case AMAL_OP_DIV:
+            }
+            case AMAL_OP_DIV: {
                 self->read_index += 3;
                 break;
-            case AMAL_OP_PUSH:
+            }
+            case AMAL_OP_PUSH: {
+                #if 0
+                AsmPtr reg;
+                asm_ptr_init_disp(&reg, RBP, -(i32)get_register_at_offset(0));
+
+                return_if_error(ensure_stack_capacity_for_push(self));
+                self->stack[self->stack_index] = self->reg[(u8)self->data.data[self->read_index]];
+                ++self->stack_index;
+                return_if_error(asm_pushm(&self->asm, &reg));
+                #endif
                 self->read_index += 1;
                 break;
-            case AMAL_OP_PUSHI:
+            }
+            case AMAL_OP_PUSHI: {
+                u16 intermediate_index;
+                Number number;
+
+                am_memcpy(&intermediate_index, &self->data.data[self->read_index], sizeof(intermediate_index));
+                return_if_error(amal_program_get_intermediate_by_index(self, intermediate_index, &number));
+                return_if_error(ensure_stack_capacity_for_push(self));
+
+                self->stack[self->stack_index] = number.value.integer;
+                ++self->stack_index;
                 self->read_index += 2;
                 break;
-            case AMAL_OP_PUSHD:
+            }
+            case AMAL_OP_PUSHD: {
+                u16 data_index;
+                char *data_ptr;
+
+                am_memcpy(&data_index, &self->data.data[self->read_index], sizeof(data_index));
+                return_if_error(amal_program_get_data_by_index(self, data_index, &data_ptr));
+                return_if_error(ensure_stack_capacity_for_push(self));
+
+                self->stack[self->stack_index] = (uintptr_t)data_ptr;
+                ++self->stack_index;
                 self->read_index += 2;
                 break;
+            }
             case AMAL_OP_CALL:
-                self->read_index += 2;
+                /*assert(bool_false && "TODO: Implement CALL");*/
+                self->read_index += 3;
                 break;
             case AMAL_OP_CALLR:
-                self->read_index += 1;
+                /*assert(bool_false && "TODO: Implement CALLR");*/
+                self->read_index += 2;
                 break;
-            case AMAL_OP_CMP:
+            case AMAL_OP_CMP: {
+                self->reg[(u8)self->data.data[self->read_index]] = 
+                    self->reg[(u8)self->data.data[self->read_index + 1]] == self->reg[(u8)self->data.data[self->read_index + 2]];
                 self->read_index += 3;
                 break;
-            case AMAL_OP_JZ:
+            }
+            case AMAL_OP_JZ: {
+                #if 0
+                u8 reg;
+                i16 jump_offset;
+                u32 jump_target;
+
+                reg = (u8)self->data.data[self->read_index];
+                am_memcpy(&jump_offset, &self->data.data[self->read_index + 1], sizeof(jump_offset));
+                jump_target = (isize)self->read_index + jump_offset;
+                if(jump_target < read_start || jump_target >= read_end)
+                    return AMAL_PROGRAM_INSTRUCTION_ILLEGAL_JUMP_TARGET;
+                #endif
                 self->read_index += 3;
                 break;
-            case AMAL_OP_JMP:
+            }
+            case AMAL_OP_JMP: {
+                #if 0
+                i16 jump_offset;
+                u32 jump_target;
+                
+                am_memcpy(&jump_offset, &self->data.data[self->read_index], sizeof(jump_offset));
+                jump_target = (isize)self->read_index + jump_offset;
+                if(jump_target < read_start || jump_target >= read_end)
+                    return AMAL_PROGRAM_INSTRUCTION_ILLEGAL_JUMP_TARGET;
+                #endif
                 self->read_index += 2;
                 break;
+            }
             case AMAL_OP_RET:
+                /* return_if_error(asm_ret(&self->asm, 0)); */
+                assert(bool_false && "TODO: Implement RET. RET needs to restore the stack before returning");
                 break;
-            case AMAL_OP_FUNC_START:
+            case AMAL_OP_FUNC_START: {
+                assert(!inside_func);
+                inside_func = bool_true;
+                am_memcpy(&func_num_registers, &self->data.data[self->read_index], sizeof(func_num_registers));
+                /*
+                    TODO: Validate stack size, or maybe remove all validation? do we really need validation?
+                    If we need security, we could fork the process instead.
+                */
+
+                /*
+                    Some registers need to be preserved before entering a function scope and these registers are different on different platforms.
+                    32-bit: EBX, ESI, EDI, EBP
+                    64-bit Windows: RBX, RSI, RDI, RBP, R12-R15, XMM6-XMM15
+                    64-bit Linux,BSD,Mac: RBX, RBP, R12-R15
+                */
+                return_if_error(asm_pushr(&self->asm, RBX));
+                return_if_error(asm_pushr(&self->asm, RBP));
+                return_if_error(asm_mov_rr(&self->asm, RBP, RSP));
+                return_if_error(asm_sub_rm64_imm(&self->asm, RSP, func_num_registers * sizeof(usize)));
+                self->read_index += 2;
                 break;
-            case AMAL_OP_FUNC_END:
+            }
+            case AMAL_OP_FUNC_END: {
+                assert(inside_func);
+                inside_func = bool_false;
+                /*assert(bool_false && "TODO: Implement FUNC_END");*/
+                /* TODO: Validate FUNC_END is called for every FUNC_START, otherwise stack will be corrupted */
+                /* TODO: Use mov_rr(RSP, RBP) instead? why doesn't gcc do this? */
+                return_if_error(asm_mov_rr(&self->asm, RSP, RBP));
+                return_if_error(asm_popr(&self->asm, RBP));
+                return_if_error(asm_popr(&self->asm, RBX));
+                return_if_error(asm_ret(&self->asm, 0));
                 break;
+            }
         }
     }
 
@@ -266,7 +574,7 @@ int amal_program_run(amal_program *self) {
         return_if_error(amal_program_read_strings(self));
         return_if_error(amal_program_read_instructions(self));
     }
-    return AMAL_PROGRAM_OK;
+    return asm_execute(&self->asm);
 }
 
 int amal_program_save(amal_program *self, const char *filepath) {
diff --git a/src/ssa/ssa.c b/src/ssa/ssa.c
index 6e61bf1..b95cd10 100644
--- a/src/ssa/ssa.c
+++ b/src/ssa/ssa.c
@@ -50,7 +50,7 @@ SsaNumber create_ssa_float(f64 value) {
     return result;
 }
 
-int ssa_init(Ssa *self, ScopedAllocator *allocator) {
+int ssa_init(Ssa *self, ArenaAllocator *allocator) {
     return_if_error(buffer_init(&self->instructions, allocator));
     return_if_error(hash_map_init(&self->intermediates_map, allocator, sizeof(SsaIntermediateIndex), compare_number, hash_number));
     return_if_error(buffer_init(&self->intermediates, allocator));
@@ -207,7 +207,7 @@ static CHECK_RESULT int ssa_ins_binop(Ssa *self, SsaInstruction binop_type, SsaR
     return ssa_add_ins_form2(self, binop_type, lhs, rhs, result);
 }
 
-static CHECK_RESULT int ssa_ins_func_start(Ssa *self, u8 num_args, SsaFuncIndex *result) {
+static CHECK_RESULT int ssa_ins_func_start(Ssa *self, SsaFuncIndex *result, usize *func_metadata_index) {
     usize index;
     index = self->instructions.size;
 
@@ -215,12 +215,13 @@ static CHECK_RESULT int ssa_ins_func_start(Ssa *self, u8 num_args, SsaFuncIndex
     if(self->func_counter + 1 < self->func_counter)
         return -1;
 
-    return_if_error(buffer_append_empty(&self->instructions, sizeof(u8) + sizeof(SsaFuncIndex) + sizeof(u8)));
+    return_if_error(buffer_append_empty(&self->instructions, sizeof(u8) + sizeof(SsaFuncIndex) + sizeof(u16)));
     *result = self->func_counter++;
     self->instructions.data[index + 0] = SSA_FUNC_START;
     am_memcpy(self->instructions.data + index + 1, result, sizeof(SsaFuncIndex));
-    self->instructions.data[index + 1 + sizeof(SsaFuncIndex)] = num_args;
-    amal_log_debug("FUNC_START f%u(%u)", *result, num_args);
+    *func_metadata_index = index + 1 + sizeof(SsaFuncIndex);
+    /* No need to add data to instructions.data here, it can contain undefined data until we set it (@ the caller) */
+    amal_log_debug("FUNC_START f%u", *result);
     return 0;
 }
 
@@ -363,9 +364,9 @@ static CHECK_RESULT SsaRegister lhsexpr_generate_ssa(Ast *self, SsaCompilerConte
     } else {
         /* TODO: Do not assign if we dont want default value */
         SsaNumber number;
-        if(self->resolve_data.type == context->compiler->default_types.i64)
+        if(self->resolve_data.type == (LhsExpr*)context->compiler->default_types.i64)
             number = create_ssa_integer(0);
-        else if(self->resolve_data.type == context->compiler->default_types.f64)
+        else if(self->resolve_data.type == (LhsExpr*)context->compiler->default_types.f64)
             number = create_ssa_float(0.0);
         else
             assert(bool_false && "TODO: assign default value to reg depending on LhsExpr type");
@@ -399,23 +400,24 @@ static CHECK_RESULT SsaRegister funcdecl_generate_ssa(FunctionDecl *self, SsaCom
     that is reset after function end
     */
     SsaRegister prev_reg_counter;
+    usize func_metadata_index;
     prev_reg_counter = context->ssa->reg_counter;
     context->ssa->reg_counter = 0;
 
     amal_log_debug("SSA funcdecl %p", self);
-    throw_if_error(ssa_ins_func_start(context->ssa, 0, &self->ssa_func_index));
+    throw_if_error(ssa_ins_func_start(context->ssa, &self->ssa_func_index, &func_metadata_index));
     scope_generate_ssa(&self->body, context);
     throw_if_error(ssa_ins_func_end(context->ssa));
 
+    /* Add the number of registers used to the function metadata (FUNC_START) */
+    am_memcpy(&context->ssa->instructions.data[func_metadata_index], &context->ssa->reg_counter, sizeof(u16));
     context->ssa->reg_counter = prev_reg_counter;
-    /*assert(bool_false);*/
     return 0;
 }
 
 static CHECK_RESULT SsaRegister funccall_generate_ssa(Ast *self, SsaCompilerContext *context) {
     /* TODO: Implement */
     FunctionCall *func_call;
-    FunctionDecl *func_to_call;
     Ast **ast;
     Ast **ast_end;
     SsaRegister reg;
@@ -430,14 +432,19 @@ static CHECK_RESULT SsaRegister funccall_generate_ssa(Ast *self, SsaCompilerCont
         throw_if_error(ssa_ins_push(context->ssa, arg_reg));
     }
 
-    assert(self->resolve_data.type->rhs_expr->type == AST_FUNCTION_DECL);
-    func_to_call = self->resolve_data.type->rhs_expr->value.func_decl;
-    /*
-    TODO: Implement func reference instead of using 0. Perhaps the best way is to use function declaration pointer value?
-    then there is no need for mutex locks.
-    */
-    amal_log_debug("SSA funccall %.*s, func index ptr: %p", func_call->func.name.size, func_call->func.name.data, func_to_call);
-    throw_if_error(ssa_ins_call(context->ssa, func_to_call, &reg));
+    assert((self->resolve_data.type->rhs_expr && self->resolve_data.type->rhs_expr->type == AST_FUNCTION_DECL) ||
+            self->resolve_data.type->type.type == VARIABLE_TYPE_SIGNATURE);
+    if(self->resolve_data.type->is_extern) {
+        amal_log_error("TODO: Implement extern function call (extern function %.*s was called)", func_call->func.name.size, func_call->func.name.data);
+        reg = 0;
+        assert(bool_false && "TODO: Implement extern function call!");
+    } else {
+        FunctionDecl *func_to_call;
+        func_to_call = self->resolve_data.type->rhs_expr->value.func_decl;
+        amal_log_debug("SSA funccall %.*s, func index ptr: %p", func_call->func.name.size, func_call->func.name.data, func_to_call);
+        throw_if_error(ssa_ins_call(context->ssa, func_to_call, &reg));
+    }
+
     return reg;
 }
 
@@ -470,16 +477,16 @@ static CHECK_RESULT SsaRegister variable_generate_ssa(Variable *self, SsaCompile
     return ast_generate_ssa(self->resolved_var, context);
 }
 
-static SsaInstruction binop_type_to_ssa_type(BinopType binop_type) {
+static SsaInstruction binop_type_to_ssa_type(BinopType binop_type, amal_default_type *type) {
     switch(binop_type) {
         case BINOP_ADD:
             return SSA_ADD;
         case BINOP_SUB:
             return SSA_SUB;
         case BINOP_MUL:
-            return SSA_MUL;
+            return type->is_signed ? SSA_IMUL : SSA_MUL;
         case BINOP_DIV:
-            return SSA_DIV;
+            return type->is_signed ? SSA_IDIV : SSA_DIV;
         case BINOP_DOT:
             assert(bool_false && "Binop dot not valid for arithmetic operation and requires special functionality");
             return 0;
@@ -494,18 +501,23 @@ static CHECK_RESULT SsaRegister binop_generate_ssa(Binop *self, SsaCompilerConte
     SsaRegister rhs_reg;
     SsaRegister reg;
 
+    /*
+    const std = @import("std.amal");
+    std.printf
+    */
     if(self->type == BINOP_DOT && self->rhs->resolve_data.type->rhs_expr->type == AST_FUNCTION_DECL) {
         reg = ast_generate_ssa(self->rhs, context);
     } else {
         lhs_reg = ast_generate_ssa(self->lhs, context);
         rhs_reg = ast_generate_ssa(self->rhs, context);
-        throw_if_error(ssa_ins_binop(context->ssa, binop_type_to_ssa_type(self->type), lhs_reg, rhs_reg, &reg));
+        throw_if_error(ssa_ins_binop(context->ssa, binop_type_to_ssa_type(self->type, (amal_default_type*)self->lhs->resolve_data.type), lhs_reg, rhs_reg, &reg));
     }
     return reg;
 }
 
 static void else_if_statement_generate_ssa(ElseIfStatement *else_if_stmt, SsaCompilerContext *context) {
     usize jump_ins_index;
+    jump_ins_index = 0;
     if(else_if_stmt->condition) {
         SsaRegister condition_reg;
         condition_reg = ast_generate_ssa(else_if_stmt->condition, context);
diff --git a/src/std/scoped_allocator.c b/src/std/arena_allocator.c
index d8acbf6..20f0394 100644
--- a/src/std/scoped_allocator.c
+++ b/src/std/arena_allocator.c
@@ -1,4 +1,4 @@
-#include "../../include/std/scoped_allocator.h"
+#include "../../include/std/arena_allocator.h"
 #include "../../include/std/alloc.h"
 #include "../../include/std/thread.h"
 #include "../../include/std/log.h"
@@ -6,31 +6,31 @@
 
 #define ALLOC_NODE_SIZE 4096
 
-int scoped_allocator_node_init(ScopedAllocatorNode *self) {
+int arena_allocator_node_init(ArenaAllocatorNode *self) {
     self->data = NULL;
     self->size = 0;
     self->next = NULL;
     return am_malloc(ALLOC_NODE_SIZE, (void**)&self->data);
 }
 
-void scoped_allocator_node_deinit(ScopedAllocatorNode *self) {
+void arena_allocator_node_deinit(ArenaAllocatorNode *self) {
     am_free(self->data);
     self->data = NULL;
     self->size = 0;
     if(self->next) {
-        scoped_allocator_node_deinit(self->next);
+        arena_allocator_node_deinit(self->next);
         am_free(self->next);
         self->next = NULL;
     }
 }
 
-int scoped_allocator_init(ScopedAllocator *self) {
-    return_if_error(scoped_allocator_node_init(&self->head));
+int arena_allocator_init(ArenaAllocator *self) {
+    return_if_error(arena_allocator_node_init(&self->head));
     self->current = &self->head;
     return buffer_init(&self->mems, NULL);
 }
 
-static void scoped_allocator_deinit_buffers(ScopedAllocator *self) {
+static void arena_allocator_deinit_buffers(ArenaAllocator *self) {
     void **mem;
     void **mems_end;
     mem = buffer_begin(&self->mems);
@@ -42,19 +42,19 @@ static void scoped_allocator_deinit_buffers(ScopedAllocator *self) {
     buffer_deinit(&self->mems);
 }
 
-void scoped_allocator_deinit(ScopedAllocator *self) {
+void arena_allocator_deinit(ArenaAllocator *self) {
     self->current = NULL;
-    scoped_allocator_deinit_buffers(self);
-    scoped_allocator_node_deinit(&self->head);
+    arena_allocator_deinit_buffers(self);
+    arena_allocator_node_deinit(&self->head);
 }
 
-static CHECK_RESULT int scoped_allocator_ensure_capacity_for(ScopedAllocator *self, usize size) {
+static CHECK_RESULT int arena_allocator_ensure_capacity_for(ArenaAllocator *self, usize size) {
     void *new_node;
     new_node = NULL;
     
     if(self->current->size + size > ALLOC_NODE_SIZE) {
-        return_if_error(am_malloc(sizeof(ScopedAllocatorNode), &new_node));
-        cleanup_if_error(scoped_allocator_node_init(new_node));
+        return_if_error(am_malloc(sizeof(ArenaAllocatorNode), &new_node));
+        cleanup_if_error(arena_allocator_node_init(new_node));
         self->current->next = new_node;
         self->current = new_node;
     }
@@ -78,19 +78,19 @@ static usize align_ptr_ceil_offset(void *ptr, uintptr_t alignment) {
 
 #define SCOPED_ALLOC_ALIGNMENT 8
 
-int scoped_allocator_alloc(ScopedAllocator *self, usize size, void **mem) {
-    ScopedAllocatorNode *current;
+int arena_allocator_alloc(ArenaAllocator *self, usize size, void **mem) {
+    ArenaAllocatorNode *current;
     usize alloc_size;
     assert(self->current);
     current = self->current;
 
     if(size >= ALLOC_NODE_SIZE) {
-        amal_log_error("scoped_allocator_alloc: tried to alloc memory of size %lu. Max allowed alloc size is %lu", size, ALLOC_NODE_SIZE);
+        amal_log_error("arena_allocator_alloc: tried to alloc memory of size %lu. Max allowed alloc size is %lu", size, ALLOC_NODE_SIZE);
         return -1;
     }
 
     alloc_size = size + align_ptr_ceil_offset(self->current->data + self->current->size, SCOPED_ALLOC_ALIGNMENT);
-    return_if_error(scoped_allocator_ensure_capacity_for(self, alloc_size));
+    return_if_error(arena_allocator_ensure_capacity_for(self, alloc_size));
     /* Reallocated (new node created) */
     if(self->current != current) {
         *mem = self->current->data;
@@ -102,7 +102,7 @@ int scoped_allocator_alloc(ScopedAllocator *self, usize size, void **mem) {
     return 0;
 }
 
-int scoped_allocator_add_mem(ScopedAllocator *self, usize *result_index) {
+int arena_allocator_add_mem(ArenaAllocator *self, usize *result_index) {
     void *null_data;
     null_data = NULL;
     *result_index = buffer_get_size(&self->mems, sizeof(void*));
diff --git a/src/std/buffer.c b/src/std/buffer.c
index f4e93e5..0e4ca89 100644
--- a/src/std/buffer.c
+++ b/src/std/buffer.c
@@ -1,16 +1,16 @@
 #include "../../include/std/buffer.h"
 #include "../../include/std/alloc.h"
 #include "../../include/std/mem.h"
-#include "../../include/std/scoped_allocator.h"
+#include "../../include/std/arena_allocator.h"
 #include <assert.h>
 
-int buffer_init(Buffer *self, struct ScopedAllocator *allocator) {
+int buffer_init(Buffer *self, struct ArenaAllocator *allocator) {
     self->data = NULL;
     self->size = 0;
     self->capacity = 0;
     self->allocator = allocator;
     if(allocator) {
-        return scoped_allocator_add_mem(allocator, &self->allocator_index);
+        return arena_allocator_add_mem(allocator, &self->allocator_index);
     } else {
         self->allocator_index = ~(usize)0;
         return 0;
diff --git a/src/std/hash_map.c b/src/std/hash_map.c
index 1ad0dea..bcb43eb 100644
--- a/src/std/hash_map.c
+++ b/src/std/hash_map.c
@@ -1,5 +1,5 @@
 #include "../../include/std/hash_map.h"
-#include "../../include/std/scoped_allocator.h"
+#include "../../include/std/arena_allocator.h"
 #include "../../include/std/mem.h"
 #include <assert.h>
 
@@ -73,7 +73,7 @@ static void* bucket_node_get_value(HashMapBucketNode *self) {
     return value;
 }
 
-int hash_map_init(HashMap *self, ScopedAllocator *allocator, usize value_type_size, 
+int hash_map_init(HashMap *self, ArenaAllocator *allocator, usize value_type_size, 
                   HashMapCompare compare_func, HashMapHash hash_func) {
     assert(compare_func);
     assert(hash_func);
@@ -91,7 +91,7 @@ int hash_map_init(HashMap *self, ScopedAllocator *allocator, usize value_type_si
 
 static CHECK_RESULT int hash_map_bucket_add(HashMap *self, HashMapBucket *bucket, BufferView key, void *value, usize hash) {
     HashMapBucketNode *new_bucket_node;
-    return_if_error(scoped_allocator_alloc(self->allocator, 
+    return_if_error(arena_allocator_alloc(self->allocator, 
                                            sizeof(HashMapBucketNode*) + sizeof(hash) + sizeof(u32) + key.size + self->value_type_size,
                                            (void**)&new_bucket_node));
     bucket_node_set_next(new_bucket_node, bucket->start);
diff --git a/src/std/mem.c b/src/std/mem.c
index f406176..95edcb9 100644
--- a/src/std/mem.c
+++ b/src/std/mem.c
@@ -1,5 +1,6 @@
 #include "../../include/std/mem.h"
 #include <string.h>
+#include <unistd.h>
 
 void am_memcpy(void *dest, const void *src, usize size) {
     memcpy(dest, src, size);
@@ -16,3 +17,8 @@ bool am_memeql(const void *lhs, const void *rhs, usize size) {
 void am_memset(void *dest, int value, usize size) {
     memset(dest, value, size);
 }
+
+long am_pagesize() {
+    return sysconf(_SC_PAGESIZE);
+}
+
diff --git a/src/tokenizer.c b/src/tokenizer.c
index bce386d..7620fc0 100644
--- a/src/tokenizer.c
+++ b/src/tokenizer.c
@@ -2,7 +2,7 @@
 #include "../include/std/mem.h"
 #include "../include/std/log.h"
 #include "../include/std/thread.h"
-#include "../include/std/scoped_allocator.h"
+#include "../include/std/arena_allocator.h"
 #include <assert.h>
 #include <limits.h>
 #include <stdio.h>
@@ -25,7 +25,7 @@ static int tokenizer_get_end_of_line_from_index(Tokenizer *self, int index);
 /* Returns -1 if end of multiline comment was not found */
 static int tokenizer_get_end_of_multiline_comment(Tokenizer *self, int index);
 
-int tokenizer_init(Tokenizer *self, ScopedAllocator *allocator, BufferView code, BufferView code_name, const amal_compiler_options *compiler_options) {
+int tokenizer_init(Tokenizer *self, ArenaAllocator *allocator, BufferView code, BufferView code_name, const amal_compiler_options *compiler_options) {
     assert(code.size <= INT_MAX);
     assert(compiler_options);
     /* Skip UTF-8 BOM */
@@ -415,6 +415,7 @@ static const char* binop_to_string(BinopType binop_type) {
 
 static BufferView tokenizer_expected_token_as_string(Token token) {
     const char *str;
+    str = "";
     switch(token) {
         case TOK_NONE:
             str = "none";
@@ -737,7 +738,7 @@ TokenizerError tokenizer_create_error(Tokenizer *self, int index, const char *fm
 
     result.index = index;
     result.str = NULL;
-    ignore_result_int(scoped_allocator_alloc(self->allocator, bytes_copied + 1, (void**)&result.str));
+    ignore_result_int(arena_allocator_alloc(self->allocator, bytes_copied + 1, (void**)&result.str));
     if(result.str && bytes_copied > 0)
         am_memcpy(result.str, buffer, bytes_copied + 1);
     return result;
diff --git a/tests/bytecode.amal b/tests/bytecode.amal
index c5b2cb3..3484b21 100644
--- a/tests/bytecode.amal
+++ b/tests/bytecode.amal
@@ -1,12 +1,14 @@
 extern const printf: fn;
 
-const print = fn {
-
-}
-
 const main = fn {
     var value = 23;
     value = 34;
+    const value2: i64 = 23;
+    const value3 = 2 + 5 - 1 * 10 / 2;
     const str_value = "hello, world";
-    print();
+    printf();
+}
+
+const print = fn {
+
 }
\ No newline at end of file
diff --git a/tests/errors/arithmetic_incompatible_types.amal b/tests/errors/arithmetic_incompatible_types.amal
new file mode 100644
index 0000000..367e5ed
--- /dev/null
+++ b/tests/errors/arithmetic_incompatible_types.amal
@@ -0,0 +1,5 @@
+const main = fn {
+    const value = 23;
+    const value2 = "sadsa";
+    const value3 = value + value2;
+}
\ No newline at end of file
diff --git a/tests/errors/declaration_no_type.amal b/tests/errors/declaration_no_type.amal
new file mode 100644
index 0000000..fc7d4f3
--- /dev/null
+++ b/tests/errors/declaration_no_type.amal
@@ -0,0 +1 @@
+const a;
\ No newline at end of file
diff --git a/tests/errors/declaration_no_type_extern.amal b/tests/errors/declaration_no_type_extern.amal
new file mode 100644
index 0000000..1d5c677
--- /dev/null
+++ b/tests/errors/declaration_no_type_extern.amal
@@ -0,0 +1 @@
+extern const a;
\ No newline at end of file
diff --git a/tests/errors/empty_body_regular_func.amal b/tests/errors/empty_body_regular_func.amal
new file mode 100644
index 0000000..afc3363
--- /dev/null
+++ b/tests/errors/empty_body_regular_func.amal
@@ -0,0 +1 @@
+const func: fn;
\ No newline at end of file
diff --git a/tests/errors/non_arithmetic_type_arithmetic.amal b/tests/errors/non_arithmetic_type_arithmetic.amal
new file mode 100644
index 0000000..6635d06
--- /dev/null
+++ b/tests/errors/non_arithmetic_type_arithmetic.amal
@@ -0,0 +1,4 @@
+const main = fn {
+    const value = "hello, world";
+    const value2 = value + value;
+}
\ No newline at end of file
diff --git a/tests/main.c b/tests/main.c
index 1394860..5b8bad0 100644
--- a/tests/main.c
+++ b/tests/main.c
@@ -20,14 +20,14 @@ static int num_tests_run = 0;
 }while(0)
 
 static CHECK_RESULT int test_hash_map() {
-    ScopedAllocator scoped_allocator;
+    ArenaAllocator arena_allocator;
     HashMap hash_map;
     int value;
     bool has_key;
     unsigned char i;
 
-    return_if_error(scoped_allocator_init(&scoped_allocator));
-    cleanup_if_error(hash_map_init(&hash_map, &scoped_allocator, sizeof(int), hash_map_compare_string, amal_hash_string));
+    return_if_error(arena_allocator_init(&arena_allocator));
+    cleanup_if_error(hash_map_init(&hash_map, &arena_allocator, sizeof(int), hash_map_compare_string, amal_hash_string));
 
     value = 34;
     return_if_error(hash_map_insert(&hash_map, create_buffer_view("hello", 5), &value));
@@ -43,7 +43,7 @@ static CHECK_RESULT int test_hash_map() {
     REQUIRE_EQ_INT(34, value);
 
     cleanup:
-    scoped_allocator_deinit(&scoped_allocator);
+    arena_allocator_deinit(&arena_allocator);
     return 0;
 }
 
@@ -200,6 +200,55 @@ static void test_load_error(const char *filepath, const char *expected_error) {
     free(expected_data.expected_error);
 }
 
+static void run_all_tests() {
+    test_load("tests/main.amal");
+    test_load("tests/utf8bom.amal");
+    test_load("tests/bytecode.amal");
+
+    test_load_error("tests/errors/duplicate_declaration.amal", 
+        "2:7: error: Variable with the name main was declared twice in the same scope\n"
+        "const main = fn {}\n"
+        "      ^\n");
+    test_load_error("tests/errors/pub_in_closure.amal", 
+        "2:5: error: Only declarations in global structs can be public\n"
+        "    pub const num = 45;\n"
+        "    ^\n");
+    test_load_error("tests/errors/closure_no_lhs.amal", 
+        "1:1: error: Expected string, variable, closure, struct, function call or import\n"
+        "fn {}\n"
+        "^\n");
+    test_load_error("tests/errors/const_assign.amal", 
+        "3:5: error: Can't assign to a const value\n"
+        "    value = 34;\n"
+        "    ^\n");
+    test_load_error("tests/errors/arithmetic_incompatible_types.amal", 
+        "4:28: error: Can't cast type \"str\" to type \"i64\"\n"
+        "    const value3 = value + value2;\n"
+        "                           ^\n"
+        "4:20: error: Left-hand side is of type i64\n"
+        "    const value3 = value + value2;\n"
+        "          ^\n"
+        "4:28: error: Right-hand side is of type str\n"
+        "    const value3 = value + value2;\n"
+        "                           ^\n");
+    test_load_error("tests/errors/non_arithmetic_type_arithmetic.amal", 
+        "3:20: error: Arithmetic operation can only be performed with the types i8, u8, i16, u16, i32, u32, i64, u64, isize and usize\n"
+        "   const value2 = value + value;\n"
+        "                  ^\n");
+    test_load_error("tests/errors/empty_body_regular_func.amal",
+        "1:15: error: Expected function declaration. Only extern functions can have empty declarations.\n"
+        "    const func: fn;\n"
+        "                  ^\n");
+    test_load_error("tests/errors/declaration_no_type.amal",
+        "1:8: error: A variable can't be declared without a type or assignment\n"
+        "   const a;\n"
+        "          ^\n");
+    test_load_error("tests/errors/declaration_no_type_extern.amal",
+        "1:15: error: A variable can't be declared without a type or assignment\n"
+        "   extern const a;\n"
+        "                 ^\n");
+}
+
 /* TODO: Restrict variables in global scope to const */
 int main(int argc, char **argv) {
     int result;
@@ -212,34 +261,14 @@ int main(int argc, char **argv) {
         exit(1);
     }
 
+    /* Sanity check */
     return_if_error(test_hash_map());
 
     /* Run all tests */
     if(argc == 1) {
-        test_load("tests/main.amal");
-        test_load("tests/utf8bom.amal");
-        test_load("tests/bytecode.amal");
-
-        test_load_error("tests/errors/duplicate_declaration.amal", 
-            "2:7: error: Variable with the name main was declared twice in the same scope\n"
-            "const main = fn {}\n"
-            "      ^\n");
-        test_load_error("tests/errors/pub_in_closure.amal", 
-            "2:5: error: Only declarations in global structs can be public\n"
-            "    pub const num = 45;\n"
-            "    ^\n");
-        test_load_error("tests/errors/closure_no_lhs.amal", 
-            "1:1: error: Expected string, variable, closure, struct, function call or import\n"
-            "fn {}\n"
-            "^\n");
-        test_load_error("tests/errors/const_assign.amal", 
-            "3:5: error: Can't assign to a const value\n"
-            "    value = 34;\n"
-            "    ^\n");
-        fprintf(stderr, "##### %d/%d tests succeeded #####\n", num_successful_tests, num_tests_run);
+        run_all_tests();
     } else if(argc == 2) {
         test_load(argv[1]);
-        fprintf(stderr, "##### %d/%d tests succeeded #####\n", num_successful_tests, num_tests_run);
     } else {
         fprintf(stderr, "usage: test [test-file-path]\n");
         fprintf(stderr, "If you run test without any files then all tests will run.\n");
@@ -249,5 +278,6 @@ int main(int argc, char **argv) {
         exit(1);
     }
 
+    fprintf(stderr, "##### %d/%d tests succeeded #####\n", num_successful_tests, num_tests_run);
     return 0;
-}
\ No newline at end of file
+}