X-Git-Url: https://git.rrq.au/?a=blobdiff_plain;f=main.asm;h=8d9c51b77573f2da465b50914b6574acb1f29679;hb=6345a1a92698a72f5865e8e04dbc4bd14858e747;hp=f466a104d0b7caaba7e2f9f5f7ddfcf284a53e23;hpb=38ab2f16dd683ac65c0567f3d7a079dcdc3edf2f;p=rrq%2Fjonasforth.git diff --git a/main.asm b/main.asm index f466a10..8d9c51b 100644 --- a/main.asm +++ b/main.asm @@ -23,14 +23,56 @@ macro popr x { add rbp, 8 } +;; The following macro generates the dictionary header. It updates the +;; initial_latest_entry variable, which is used as the initial value of the +;; latest_entry variable that is made available at runtime. +;; +;; The header contains a link to the previous entry, the length of the name of +;; the word and the word itself as a string literal. +;; +;; This macro also defines a label LABEL_entry. +initial_latest_entry = 0 +macro header label, name { + local .string_end + +label#_entry: + dq initial_latest_entry + db .string_end - ($ + 1) + db name + .string_end: +label: + +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 + 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 +} + + + segment readable executable +entry main + +include "impl.asm" + main: cld ; Clear direction flag so LODSQ does the right thing. mov rbp, return_stack_top ; Initialize return stack - mov rsi, program - next + mov rax, MAIN + jmp qword [rax] program: dq MAIN @@ -44,33 +86,54 @@ docol: ;; 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. -EXIT: - dq .start -.start: +forth_asm EXIT, 'EXIT' popr rsi next ;; LIT is a special word that reads the next "word pointer" and causes it to be ;; placed on the stack rather than executed. -LIT: - dq .start -.start: +forth_asm LIT, 'LIT' lodsq push rax next +;; Given a string (a pointer following by a size), return the location of the +;; dictionary entry for that word. If no such word exists, return 0. +forth_asm FIND, 'FIND' + mov [.rsi], rsi + + pop [find.search_length] + pop [find.search_buffer] + mov rsi, [latest_entry] ; Start with the last added word + call find + push rsi + + mov rsi, [.rsi] + next + push rsi + + mov rsi, [.rsi] + next + +;; Given an entry in the dictionary, return a pointer to the codeword of that +;; entry. +forth_asm TCFA, '>CFA' + pop rax + add rax, 8 ; [rax] = length of name + movzx rbx, byte [rax] + inc rax + add rax, rbx ; [rax] = codeword + push rax + next + ;; BRANCH is the fundamental mechanism for branching. BRANCH reads the next word ;; as a signed integer literal and jumps by that offset. -BRANCH: - dq .start -.start: +forth_asm BRANCH, 'BRANCH' add rsi, [rsi] ; [RSI], which is the next word, contains the offset; we add this to the instruction pointer. next ; Then, we can just continue execution as normal ;; 0BRANCH is like BRANCH, but it jumps only if the top of the stack is zero. -ZBRANCH: - dq .start -.start: +forth_asm ZBRANCH, '0BRANCH' ;; Compare top of stack to see if we should branch pop rax cmp rax, 0 @@ -81,10 +144,18 @@ ZBRANCH: add rsi, 8 ; We need to skip over the next word, which contains the offset. next +;; Duplicate the top of the stack. +forth_asm DUP_, 'DUP' + push qword [rsp] + next + +;; Execute the codeword at the given address. +forth_asm EXEC, 'EXEC' + pop rax + jmp qword [rax] + ;; Expects a character on the stack and prints it to standard output. -EMIT: - dq .start -.start: +forth_asm EMIT, 'EMIT' pushr rsi pushr rax mov rax, 1 @@ -98,125 +169,50 @@ EMIT: next ;; Prints a newline to standard output. -NEWLINE: - dq docol +forth NEWLINE, 'NEWLINE' dq LIT, $A dq EMIT dq EXIT +;; Prints a space to standard output. +forth SPACE, 'SPACE' + dq LIT, ' ' + dq EMIT + dq EXIT + ;; Read a word from standard input and push it onto the stack as a pointer and a ;; size. The pointer is valid until the next call to READ_WORD. -READ_WORD: - dq .start -.start: +forth_asm READ_WORD, 'READ-WORD' mov [.rsi], rsi - mov [.rax], rax - -.skip_whitespace: - ;; Read characters into .char_buffer until one of them is not whitespace. - mov rax, 0 - mov rdi, 0 - mov rsi, .char_buffer - mov rdx, 1 - syscall - cmp [.char_buffer], ' ' - je .skip_whitespace - cmp [.char_buffer], $A - je .skip_whitespace - -.alpha: - ;; We got a character that wasn't whitespace. Now read the actual word. - mov [.length], 0 - -.read_alpha: - mov al, [.char_buffer] - movzx rbx, [.length] - mov rsi, .buffer - add rsi, rbx - mov [rsi], al - inc [.length] - - mov rax, 0 - mov rdi, 0 - mov rsi, .char_buffer - mov rdx, 1 - syscall - - cmp [.char_buffer], ' ' - je .end - cmp [.char_buffer], $A - jne .read_alpha - -.end: - push .buffer - movzx rax, [.length] - push rax + call read_word + push rdi ; Buffer + push rdx ; Length mov rsi, [.rsi] - mov rax, [.rax] - next ;; Takes a string on the stack and replaces it with the decimal number that the ;; string represents. -PARSE_NUMBER: - dq .start -.start: - pop [.length] ; Length - pop rdi ; String pointer - mov r8, 0 ; Result - - ;; Add (10^(rcx-1) * parse_char(rdi[length - rcx])) to the accumulated value - ;; for each rcx. - mov rcx, [.length] -.loop: - ;; First, calcuate 10^(rcx - 1) - mov rax, 1 - - mov r9, rcx - .exp_loop: - dec r9 - jz .break - mov rbx, 10 - mul rbx - jmp .exp_loop - .break: - - ;; Now, rax = 10^(rcx - 1). - - ;; We need to calulate the value of the character at rdi[length - rcx]. - mov rbx, rdi - add rbx, [.length] - sub rbx, rcx - movzx rbx, byte [rbx] - sub rbx, '0' - - ;; Multiply this value by rax to get (10^(rcx-1) * parse_char(rdi[length - rcx])), - ;; then add this to the result. - mul rbx - - ;; Add that value to r8 - add r8, rax +forth_asm PARSE_NUMBER, 'PARSE-NUMBER' + pop [parse_number.length] ; Length + pop [parse_number.buffer] ; String pointer - dec rcx - jnz .loop - - push r8 + push rsi + call parse_number + pop rsi + push rax ; Result next -READ_NUMBER: - dq docol +forth READ_NUMBER, 'READ-NUMBER' dq READ_WORD dq PARSE_NUMBER dq EXIT ;; Takes a string (in the form of a pointer and a length on the stack) and ;; prints it to standard output. -TELL: - dq .start -.start: +forth_asm TELL, 'TELL' mov rbx, rsi mov rcx, rax @@ -231,32 +227,12 @@ TELL: next ;; Exit the program cleanly. -TERMINATE: - dq .start -.start: +forth_asm TERMINATE, 'TERMINATE' mov rax, $3C mov rdi, 0 syscall -PUSH_HELLO_CHARS: - dq docol - dq LIT, $A - dq LIT, 'o' - dq LIT, 'l' - dq LIT, 'l' - dq LIT, 'e' - dq LIT, 'H' - dq EXIT - -PUSH_YOU_TYPED: - dq .start -.start: - push you_typed_string - push you_typed_string.length - next - -HELLO: - dq docol +forth HELLO, 'HELLO' dq LIT, 'H', EMIT dq LIT, 'e', EMIT dq LIT, 'l', EMIT @@ -266,10 +242,17 @@ HELLO: dq NEWLINE dq EXIT +;; The INTERPRET word reads and interprets user input. It's behavior depends on +;; the current STATE. It provides special handling for integers. (TODO) +forth INTERPRET, 'INTERPRET' + dq READ_WORD + dq FIND + dq TCFA + dq EXEC + dq EXIT + ;; .U prints the value on the stack as an unsigned integer in hexadecimal. -DOTU: - dq .start -.start: +forth_asm DOTU, '.U' mov [.length], 0 mov [.printed_length], 1 pop rax ; RAX = value to print @@ -329,24 +312,20 @@ DOTU: pop rsi next -MAIN: - dq docol +forth MAIN, 'MAIN' dq HELLO - dq READ_NUMBER, DOTU, NEWLINE - dq BRANCH, -8 * 4 + dq INTERPRET + dq BRANCH, -8 * 2 dq TERMINATE segment readable writable -you_typed_string db 'You typed: ' -.length = $ - you_typed_string +latest_entry dq initial_latest_entry + +FIND.rsi dq ? READ_WORD.rsi dq ? -READ_WORD.rax dq ? -READ_WORD.max_size = $FF -READ_WORD.buffer rb READ_WORD.max_size -READ_WORD.length db ? -READ_WORD.char_buffer db ? +READ_WORD.rbp dq ? DOTU.chars db '0123456789ABCDEF' DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex @@ -354,8 +333,6 @@ DOTU.rbuffer rq 16 DOTU.length dq ? DOTU.printed_length dq ? -PARSE_NUMBER.length dq ? - ;; Return stack rq $2000 return_stack_top: