from-scratch/stages/as0/as1.asm
2026-08-04 18:20:58 +02:00

386 lines
8.6 KiB
NASM

section .bss
token_buf resb 0x100
buf resb 0x100
labels resb 0x1000
section .text
extern panic_abort
;; returns the next character from the source without advancing.
extern peekc
;; returns the next character from the source and advances.
extern getc
;; advances past the next character and peeks the next.
extern consuming_peekc
;; calculates a quick hash of a byte sequence given by rdi (pointer) and rsi (length)
fasthash:
mov eax, 0x811c9dc5 ; FNV offset basis
.loop:
test rsi, rsi
jz .done
movzx ecx, byte [rdi] ; get next byte
xor eax, ecx
imul eax, 0x1000193 ; FNV prime
inc rdi
dec rsi
jmp .loop
.done:
ret
;; @param lhs: (rdi, rsi)
;; @param rhs: (rdx, rcx)
;; @return al
strcmp:
cmp rcx, rsi
cmovb rsi, rcx ; if rhs is shorter, use its length for the loop
xor eax, eax
.loop:
cmp rsi, rax
je .equal
movzx ecx, byte [rdx + rax]
cmp byte [rdi + rax], cl
lea rax, [rax + 1]
je .loop
seta al ; al = lhs > rhs
sbb al, 0 ; al = al - CF
ret
.equal:
xor eax, eax
ret
memcpy:
xor rax, rax
.loop:
cmp rax, rdx
je .done
mov al, byte [rdi + rax]
mov byte [rsi + rax], al
lea rax, [rax + 1]
jmp .loop
.done:
ret
;; returns 1 if the result of `peekc()` is $sil
;; treats all characters less than ' ' as spaces.
is_ch:
push rsi
call peekc
pop rsi
cmp al, ' '
setbe cl ; cl = al <= ' '
movzx ecx, cl
mov edx, ' '
mul ecx, edx ; cl = (al < ' ') ? ' ' : 0
cmp sil, ' '
cmove ax, cx ; al = (sil < ' ') ? cl : al
cmp al, sil
setz al ; al = al == sil
ret
;; struct packed LabelEntry {
;; hash: u64,
;; file_offset: Option<NonMaxU32>,
;; }
;; push the label pointed at by rdi with length rsi, and the current file offset onto the labels array.
push_lbl:
push r12
lea r12, [rel labels]
mov rax, dword [r12 + 0x1000 - 4] ; get current label count
cmp rax, 341
jge panic_abort
lea rax, [rax + rax*2]
shl rax, 2 ; $rax = num_labels + size_of::<label_entry>()
add r12, rax
call fasthash
mov qword [r12], rax ; store hash in labels array
mov dword [r12 + 8], -1 ; initialize file offset to -1 (invalid)
inc dword [r12 + 0x1000 - 4] ; increment label count
pop r12
ret
;; searches for the offset of the label with the name pointed at by rdi with length rsi, and returns it in rax. If the label is not found, returns 0 in rax.
find_lbl_ptr:
call fasthash
lea rdx, [r15 + 0x5010] ; rdx = pointer to start of labels array
mov ebx, dword [r15 + 0x6000]
.loop:
test rbx, rbx
jz .not_found
dec rbx
cmp qword [rdx], rax
je .found
add rdx, 12 ; move to next label entry
jmp .loop
.found:
mov rax, rdx
ret
.not_found:
xor rax, rax
ret
push_or_find_lbl:
call fasthash
push rax
mov rdi, rax
call find_lbl_ptr
test rax, rax
jz .push
add rsp, 8
ret
.push:
pop rdi
call push_lbl
ret
find_lbl_offset:
call find_lbl_ptr
test rax, rax
jz .not_found
mov eax, dword [rax + 8] ; get file offset of label
ret
.not_found:
xor rdi, rdi
call panic_abort
set_lbl_offset:
push rsi
call find_lbl_ptr
pop rsi
mov dword [rax + 8], esi ; set file offset of label
ret
is_alpha:
movzx edi, dil
and edi, 0x1fffdf ; ignore bit 5 (case)
add edi, -65 ; subtract 'A'
cmp edi, 26 ; check if in range 0-25
setb al
ret
is_digit:
movzx edi, dil
lea eax, [rdi - 48] ; subtract '0'
cmp eax, 10 ; check if in range 0-9
setb al
ret
;; returns 1 if $dil is a valid identifier character (alphanumeric, '-' or '_'), and 0 otherwise.
;; clobbers rdi, rax, rcx
is_id_cont:
movzx rdi, dil
; is_digit {
lea eax, [rdi - 48]
cmp eax, 10
setb al
; } || is_alpha {
mov ecx, edi
and ecx, 0x1fffdf
add ecx, -65
cmp ecx, 26
setb cl
; } is_alpha
or al, cl
cmp edi, 45 ; || '-'
sete cl
or al, cl
cmp edi, 95 ; || '_'
sete cl
or al, cl
ret
;; converts char $dil to a digit with radix $rsi, returning it in $edx.
;; $al is set to 1 if the char is a valid digit, and 0 otherwise.
to_digit:
lea eax, [rsi - 2]
cmp eax, 35
jae .invalid
movzx rdi, dil
lea edx, [rdi - 65] ; 'A' = 65
and edx, -33 ; convert to uppercase
add edx, 10 ; 'A' should map to 10
lea eax, [rdi - 48] ; '0' = 48
cmp esi, 11
cmovb edx, eax ; if radix <= 10, then take the difference from '0'
cmp edi, 58
cmovb edx, eax ; or if char < '9', then take the difference from '0'
xor eax, eax
cmp edx, esi
setb al ; al = edx < radix
ret
.invalid:
xor eax, eax
ret
;; reads from $rdi and parses digits with radix $rsi until a non-digit is encountered.
parse_digits:
sub rsp, 24
mov qword [rsp + 16], rsi ; save the radix
mov qword [rsp + 8], rdi ; save the source iterator
mov qword [rsp], 0
.loop:
mov rdi, qword [rsp + 8] ; restore the source iterator
call peekc
mov dil, al
mov rsi, qword [rsp + 16] ; restore the radix
call to_digit
test al, al
jz .done
mov rax, qword [rsp]
mov rsi, qword [rsp + 16] ; restore the radix
imul rax, rsi
add rax, rdx
mov qword [rsp], rax
mov rdi, qword [rsp + 8] ; restore the source iterator
call getc
jmp .loop
.done:
mov rax, [rsp]
add rsp, 24
ret
parse_num:
sub rsp, 24
mov qword [rsp], 0 ; acc
mov qword [rsp + 8], rdi ; source iterator
mov dword [rsp + 16], 10 ; radix
call peekc
cmp al, `-`
jne .skip_sign
mov qword [rsp], -1 ; acc = -1
call consuming_peekc
.skip_sign:
cmp al, `0`
jne .skip_radix
call consuming_peekc
cmp al, `x`
jne .skip_radix
mov dword [rsp + 16], 16 ; radix = 16
call consuming_peekc
.skip_radix:
mov dil, al
call to_digit
test al, al
jz .done
mov rax, qword [rsp] ; acc
mov rsi, qword [rsp + 16] ; radix
mov rcx, rdx
imul rax, rsi
add rax, rcx
mov qword [rsp], rax ; acc = acc * radix + digit
mov rdi, qword [rsp + 8] ; source iterator
call consuming_peekc
jmp .skip_radix
.done:
mov rax, qword [rsp] ; move the result into rax
add rsp, 24
ret
parse_label:
push r12
lea r12, [rel buf]
sub rsp, 8
mov qword [rsp], rdi ; source iterator
call consuming_peekc
mov dil, al
call is_alpha
test al, al
jz .done
mov byte [r12], al
inc r12
mov rdi, qword [rsp] ; restore source iterator
call consuming_peekc
mov dil, al
call is_id_cont
test al, al
jz .done
jmp .loop
.done:
lea rdi, [rel buf]
mov rsi, r12
sub rsi, rdi
call push_or_find_lbl
add rsp, 8
pop r12
ret
;; reads the next token into `token_buf`
next_token:
sub rsp, 8
mov qword [rsp], 0
.skip_spaces:
mov rsi, ' '
call is_ch
test al, al
jz .skipped_spaces:
call getc
jmp .skip_spaces
.skipped_spaces:
; leading [ is a memory access
mov rsi, `[`
call is_ch
test al, al
jnz .mem_access
; leading @ is label-def
mov rsi, '@'
call is_ch
test al, al
jnz .label_def
; leading ' is a label-ref
mov rsi, `'`
call is_ch
test al, al
jnz .label_ref
; leading digit is a number
call peekc
mov rdi, rax
call is_digit
test al, al
jnz .number
; leading ; or # is a comment
mov rsi, `;`
call is_ch
test al, al
jnz .comment
mov rsi, `#`
call is_ch
test al, al
jnz .comment
; leading , is a separator
mov rsi, `,`
call is_ch
test al, al
jz .ident
; just skip the comma since its not important for parsing
call getc
jmp .skip_spaces
.ident:
; now we have some identifier
; (mov rax (byte rdi))
;; registers:
;; r0/rax
;; r1/rcx
;; r2/rdx
;; r3/rbx
;; r4/rsp
;; r5/rbp
;; r6/rsi
;; r7/rdi
;; r8
;; r9
;; r10
;; r11
;; r12
;; r13
;; r14
;; r15
;;
;; registers of the form `rNUM` may have an optional suffix b, w, d, or q
;; legacy registers may also be referenced by their 8-bit, 16-bit, or 32-bit names (e.g. al, ax, eax, rax or spl, sp, esp, rsp)
;; high-byte registers (ah, bh, ch, dh) are not supported