X-Git-Url: https://git.rrq.au/?a=blobdiff_plain;f=main.asm;h=ae3f9913eae15ea9f26a86c1b3d04a05215e2a8c;hb=055f2efc3011a2102125cd0001adcff1cc6feab9;hp=8d9c51b77573f2da465b50914b6574acb1f29679;hpb=6345a1a92698a72f5865e8e04dbc4bd14858e747;p=rrq%2Fjonasforth.git diff --git a/main.asm b/main.asm index 8d9c51b..ae3f991 100644 --- a/main.asm +++ b/main.asm @@ -1,5 +1,35 @@ +;; vim: syntax=fasm + format ELF64 executable +;; "Syscalls" {{{ + +;; [NOTE] Volatile registers Linux (syscalls) vs UEFI +;; +;; Linux syscalls: RAX, RCX, R11 +;; UEFI: RAX, RCX, R11, RDX, R8, R9, R10 + +;; We are in the process of replacing our dependency on Linux with a dependency +;; on UEFI. The following macros attempt to isolate what would be syscalls in +;; Linux; thus, we will be able to replace these with UEFI-based implementations, +;; and in theory we should expect the program to work. + +;; Print a string of a given length. +;; +;; Input: +;; - RCX = Pointer to buffer +;; - RDX = Buffer length +;; +;; Clobbers: RAX, RCX, R11, RDI, RSI +macro sys_print_string { + mov rax, 1 + mov rdi, 1 + mov rsi, rcx + syscall +} + +;; }}} + ;; The code in this macro is placed at the end of each Forth word. When we are ;; executing a definition, this code is what causes execution to resume at the ;; next word in that definition. @@ -32,11 +62,16 @@ macro popr x { ;; ;; This macro also defines a label LABEL_entry. initial_latest_entry = 0 -macro header label, name { +macro header label, name, immediate { local .string_end label#_entry: dq initial_latest_entry + if immediate eq + db 0 + else + db 1 + end if db .string_end - ($ + 1) db name .string_end: @@ -46,21 +81,19 @@ initial_latest_entry = label#_entry } ;; Define a Forth word that is implemented in assembly. See 'header' for details. -macro forth_asm label, name { - header label, name +macro forth_asm label, name, immediate { + header label, name, immediate dq .start .start: } ;; Define a Forth word that is implemented in Forth. (The body will be a list of ;; 'dq' statements.) -macro forth label, name { - header label, name - dq docol +macro forth label, name, immediate { + header label, name, immediate + dq DOCOL } - - segment readable executable entry main @@ -79,13 +112,13 @@ program: dq MAIN ;; The codeword is the code that will be executed at the beginning of a forth ;; word. It needs to save the old RSI and update it to point to the next word to ;; execute. -docol: +header DOCOL, 'DOCOL' pushr rsi ; Save old value of RSI on return stack; we will continue execution there after we are done executing this word lea rsi, [rax + 8] ; RAX currently points to the address of the codeword, so we want to continue at RAX+8 next ; Execute word pointed to by RSI ;; This word is called at the end of a Forth definition. It just needs to -;; restore the old value of RSI (saved by 'docol') and resume execution. +;; restore the old value of RSI (saved by 'DOCOL') and resume execution. forth_asm EXIT, 'EXIT' popr rsi next @@ -119,7 +152,7 @@ forth_asm FIND, 'FIND' ;; entry. forth_asm TCFA, '>CFA' pop rax - add rax, 8 ; [rax] = length of name + add rax, 8 + 1 ; [rax] = length of name movzx rbx, byte [rax] inc rax add rax, rbx ; [rax] = codeword @@ -158,11 +191,11 @@ forth_asm EXEC, 'EXEC' forth_asm EMIT, 'EMIT' pushr rsi pushr rax - mov rax, 1 - mov rdi, 1 - lea rsi, [rsp] + + lea rcx, [rsp] mov rdx, 1 - syscall + sys_print_string + add rsp, 8 popr rax popr rsi @@ -195,8 +228,8 @@ forth_asm READ_WORD, 'READ-WORD' ;; Takes a string on the stack and replaces it with the decimal number that the ;; string represents. forth_asm PARSE_NUMBER, 'PARSE-NUMBER' - pop [parse_number.length] ; Length - pop [parse_number.buffer] ; String pointer + pop rcx ; Length + pop rdi ; String pointer push rsi call parse_number @@ -213,17 +246,15 @@ forth READ_NUMBER, 'READ-NUMBER' ;; Takes a string (in the form of a pointer and a length on the stack) and ;; prints it to standard output. forth_asm TELL, 'TELL' - mov rbx, rsi - mov rcx, rax + pushr rax + pushr rsi - mov rax, 1 - mov rdi, 1 - pop rdx ; Length - pop rsi ; Buffer - syscall + pop rdx ; Length + pop rcx ; Buffer + sys_print_string - mov rax, rcx - mov rsi, rbx + popr rsi + popr rax next ;; Exit the program cleanly. @@ -232,25 +263,89 @@ forth_asm TERMINATE, 'TERMINATE' mov rdi, 0 syscall -forth HELLO, 'HELLO' - dq LIT, 'H', EMIT - dq LIT, 'e', EMIT - dq LIT, 'l', EMIT - dq LIT, 'l', EMIT - dq LIT, 'o', EMIT - dq LIT, '!', EMIT - dq NEWLINE - dq EXIT +;; Duplicate a pair of elements. +forth_asm PAIRDUP, '2DUP' + pop rbx + pop rax + push rax + push rbx + push rax + push rbx + next + +;; Swap the top two elements on the stack. +forth_asm SWAP, 'SWAP' + pop rax + pop rbx + push rax + push rbx + next + +;; Remove the top element from the stack. +forth_asm DROP, 'DROP' + add rsp, 8 + next + +forth_asm NOT_, 'NOT' + pop rax + cmp rax, 0 + jz .false +.true: + push 0 + next +.false: + push 1 + next ;; The INTERPRET word reads and interprets user input. It's behavior depends on -;; the current STATE. It provides special handling for integers. (TODO) +;; the current STATE. It provides special handling for integers. forth INTERPRET, 'INTERPRET' + ;; Read word dq READ_WORD - dq FIND + dq PAIRDUP + ;; Stack is (word length word length). + dq FIND ; Try to find word + dq DUP_ + dq ZBRANCH, 8 * 22 ; Check if word is found + + ;; - Word is found - + + dq STATE, GET, ZBRANCH, 8 * 11 ; Check whether we are in compilation or immediate mode + + ;; (Word found, compilation mode) + dq DUP_, IS_IMMEDIATE, NOT_, ZBRANCH, 8 * 6 ; If the word is immediate, continue as we would in immediate mode + + ;; Otherwise, we want to compile this word + dq TCFA + dq COMMA + dq DROP, DROP + dq EXIT + + ;; (Word found, immediate mode) + ;; Execute word dq TCFA + ;; Stack is (word length addr) + dq SWAP, DROP + dq SWAP, DROP + ;; Stack is (addr) dq EXEC dq EXIT + ;; - No word is found, assume it is an integer literal - + ;; Stack is (word length addr) + dq DROP + dq PARSE_NUMBER + + dq STATE, GET, ZBRANCH, 8 * 5 ; Check whether we are in compilation or immediate mode + + ;; (Number, compilation mode) + dq LIT, LIT, COMMA + dq COMMA + dq EXIT + + ;; (Number, immediate mode) + dq EXIT + ;; .U prints the value on the stack as an unsigned integer in hexadecimal. forth_asm DOTU, '.U' mov [.length], 0 @@ -302,37 +397,251 @@ forth_asm DOTU, '.U' loop .flip ;; Print the buffer - mov rax, 1 - mov rdi, 1 - mov rsi, .buffer + mov rcx, .buffer mov rdx, [.printed_length] - syscall + sys_print_string ;; Restore RSI and continue execution pop rsi next +;; Takes a value and an address, and stores the value at the given address. +forth_asm PUT, '!' + pop rbx ; Address + pop rax ; Value + mov [rbx], rax + next + +;; Takes an address and returns the value at the given address. +forth_asm GET, '@' + pop rax + mov rax, [rax] + push rax + next + +forth_asm PUT_BYTE, 'C!' + pop rbx + pop rax ; Value + mov [rbx], al + next + +forth_asm GET_BYTE, 'C@' + pop rax + movzx rax, byte [rax] + push rax + next + +;; Add two integers on the stack. +forth_asm PLUS, '+' + pop rax + pop rbx + add rax, rbx + push rax + next + +;; Calculate difference between two integers on the stack. The second number is +;; subtracted from the first. +forth_asm MINUS, '-' + pop rax + pop rbx + sub rbx, rax + push rbx + next + +;; Given two integers a and b on the stack, pushes the quotient and remainder of +;; division of a by b. +forth_asm TIMESMOD, '/MOD' + pop rbx ; b + pop rax ; a + mov rdx, 0 + div rbx + push rax ; a / b + push rdx ; a % b + next + +;; Get the location of the STATE variable. It can be set with '!' and read with +;; '@'. +forth STATE, 'STATE' + dq LIT, var_STATE + dq EXIT + +;; Get the location of the LATEST variable. It can be set with '!' and read with +;; '@'. +forth LATEST, 'LATEST' + dq LIT, latest_entry + dq EXIT + +;; Get the location at which compiled words are expected to be added. This +;; pointer is usually modified automatically when calling ',', but we can also +;; read it manually with 'HERE'. +forth HERE, 'HERE' + dq LIT, here + dq EXIT + +forth COMMA, ',' + dq HERE, GET, PUT ; Set the memory at the address pointed to by HERE + dq HERE, GET, LIT, 8, PLUS ; Calculate new address for HERE to point to + dq HERE, PUT ; Update HERE to point to the new address + dq EXIT + +;; Read user input until next " character is found. Push a string containing the +;; input on the stack as (buffer length). Note that the buffer is only valid +;; until the next call to S" and that no more than 255 character can be read. +forth_asm READ_STRING, 'S"' + push rsi + + mov [.length], 0 + +.read_char: + mov rax, 0 + mov rdi, 0 + mov rsi, .char_buffer + mov rdx, 1 + syscall + + mov al, [.char_buffer] + cmp al, '"' + je .done + + mov rdx, .buffer + add rdx, [.length] + mov [rdx], al + inc [.length] + jmp .read_char + +.done: + pop rsi + + push .buffer + push [.length] + + next + +;; CREATE inserts a new header in the dictionary, and updates LATEST so that it +;; points to the header. To compile a word, the user can then call ',' to +;; continue to append data after the header. +;; +;; It takes the name of the word as a string (address length) on the stack. +forth_asm CREATE, 'CREATE' + pop rcx ; Word string length + pop rdx ; Word string pointer + + mov rdi, [here] ; rdi = Address at which to insert this entry + mov rax, [latest_entry] ; rax = Address of the previous entry + mov [rdi], rax ; Insert link to previous entry + mov [latest_entry], rdi ; Update LATEST to point to this word + + add rdi, 8 + mov [rdi], byte 0 ; Insert immediate flag + + add rdi, 1 + mov [rdi], byte cl ; Insert length + + ;; Insert word string + add rdi, 1 + + push rsi + mov rsi, rdx ; rsi = Word string pointer + rep movsb + pop rsi + + ;; Update HERE + mov [here], rdi + + next + +;; Mark the last added word as immediate. +forth IMMEDIATE, 'IMMEDIATE', 1 + dq LIT, 1 + dq LATEST, GET + dq LIT, 8, PLUS + dq PUT_BYTE + dq EXIT + +;; Given the address of a word, return 0 if the given word is not immediate. +forth IS_IMMEDIATE, 'IMMEDIATE?' + dq LIT, 8, PLUS + dq GET_BYTE + dq EXIT + +;; Enter immediate mode, immediately +forth INTO_IMMEDIATE, '[', 1 + dq LIT, 0, STATE, PUT_BYTE + dq EXIT + +;; Enter compilation mode +forth OUTOF_IMMEDIATE, ']' + dq LIT, 1, STATE, PUT_BYTE + dq EXIT + +forth_asm TICK, "'" + lodsq + push rax + next + +forth_asm ROT, 'ROT' + pop rax + pop rbx + pop rdx + push rax + push rdx + push rbx + next + +forth_asm PICK, 'PICK' + pop rax + lea rax, [rsp + 8 * rax] + mov rax, [rax] + push rax + next + +forth_asm EQL, '=' + pop rax + pop rbx + cmp rax, rbx + je .eq +.noteq: + push 0 + next +.eq: + push 1 + next + forth MAIN, 'MAIN' - dq HELLO dq INTERPRET dq BRANCH, -8 * 2 dq TERMINATE segment readable writable +;; The LATEST variable holds a pointer to the word that was last added to the +;; dictionary. This pointer is updated as new words are added, and its value is +;; used by FIND to look up words. latest_entry dq initial_latest_entry +;; The STATE variable is 0 when the interpreter is executing, and non-zero when +;; it is compiling. +var_STATE dq 0 + FIND.rsi dq ? READ_WORD.rsi dq ? READ_WORD.rbp dq ? +READ_STRING.char_buffer db ? +READ_STRING.buffer rb $FF +READ_STRING.length dq ? + DOTU.chars db '0123456789ABCDEF' DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex DOTU.rbuffer rq 16 DOTU.length dq ? DOTU.printed_length dq ? +;; Reserve space for compiled words, accessed through HERE. +here dq here_top +here_top rq $4000 + ;; Return stack rq $2000 return_stack_top: