.start:
}
-;; Define a Forth word that is implemented in Forth. (The body will be a list of
-;; 'dq' statements.)
-macro forth label, name, immediate {
- header label, name, immediate
- dq DOCOL
-}
-
segment readable executable
entry main
-include "impl.asm"
+include "impl.asm" ; Misc. subroutines
+include "bootstrap.asm" ; Forth words encoded in Assembly
main:
cld ; Clear direction flag so LODSQ does the right thing.
popr rsi
next
-;; Prints a newline to standard output.
-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 a word and push it onto the stack as a pointer and a size. The pointer
+;; is valid until the next call to READ_WORD.
forth_asm READ_WORD, 'READ-WORD'
+ ;; Are we reading from user input or from the input buffer?
+ cmp [input_buffer], 0
+ jne .from_buffer
+
+ ;; Reading user input
mov [.rsi], rsi
call read_word
mov rsi, [.rsi]
next
+.from_buffer:
+ ;; Reading from buffer
+ mov [.rsi], rsi
+
+ mov rsi, [input_buffer]
+ mov rcx, [input_buffer_length]
+
+ call pop_word
+
+ mov [input_buffer], rsi ; Updated buffer
+ mov [input_buffer_length], rcx ; Length of updated buffer
+ push rdi ; Word buffer
+ push rdx ; Length of word buffer
+
+ mov rsi, [.rsi]
+ next
+
;; Takes a string on the stack and replaces it with the decimal number that the
;; string represents.
forth_asm PARSE_NUMBER, 'PARSE-NUMBER'
push rax ; Result
next
-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.
forth_asm TELL, 'TELL'
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.
-forth INTERPRET, 'INTERPRET'
- ;; Read word
- dq READ_WORD
- 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
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
+;; Read 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.
+;; until the next call to S" and that no more than 255 characters can be read.
forth_asm READ_STRING, 'S"'
+ ;; If the input buffer is set, we should read from there instead.
+ cmp [input_buffer], 0
+ jne read_string_buffer
+
push rsi
mov [.length], 0
next
+read_string_buffer:
+ push rsi
+
+ ;; We borrow READ_STRING's buffer. They won't mind.
+ mov [READ_STRING.length], 0
+
+ ;; Skip space ([TODO]: Shouldn't we do this while parsing instead?)
+ inc [input_buffer]
+ dec [input_buffer_length]
+
+.read_char:
+ mov rbx, [input_buffer]
+ mov al, [rbx]
+ cmp al, '"'
+ je .done
+
+ mov rdx, READ_STRING.buffer
+ add rdx, [READ_STRING.length]
+ mov [rdx], al
+ inc [READ_STRING.length]
+
+ inc [input_buffer]
+ dec [input_buffer_length]
+
+ jmp .read_char
+
+.done:
+ pop rsi
+
+ ;; Skip closing "
+ inc [input_buffer]
+ dec [input_buffer_length]
+
+ push READ_STRING.buffer
+ push [READ_STRING.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.
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 MAIN, 'MAIN'
+ dq SYSCODE
+ dq INTERPRET_STRING
dq INTERPRET
dq BRANCH, -8 * 2
dq TERMINATE
+;; Built-in variables:
+
+forth STATE, 'STATE'
+ dq LIT, var_STATE
+ dq EXIT
+
+forth LATEST, 'LATEST'
+ dq LIT, latest_entry
+ dq EXIT
+
+forth HERE, 'HERE'
+ dq LIT, here
+ dq EXIT
+
+forth SYSCODE, 'SYSCODE'
+ dq LIT, sysf
+ dq LIT, sysf.len
+ dq EXIT
+
+forth INPUT_BUFFER, 'INPUT-BUFFER'
+ dq LIT, input_buffer
+ dq EXIT
+
+forth INPUT_LENGTH, 'INPUT-LENGTH'
+ dq LIT, input_buffer_length
+ dq EXIT
+
segment readable writable
;; The LATEST variable holds a pointer to the word that was last added to the
;; it is compiling.
var_STATE dq 0
+;; The interpreter can read either from standard input or from a buffer. When
+;; input-buffer is set (non-null), words like READ-WORD and S" will use this
+;; buffer instead of reading user input.
+input_buffer dq 0
+input_buffer_length dq 0
+
FIND.rsi dq ?
READ_WORD.rsi dq ?
;; Return stack
rq $2000
return_stack_top:
+
+segment readable
+
+;; We store some Forth code in sys.f that defined common words that the user
+;; would expect to have available at startup. To execute these words, we just
+;; include the file directly in the binary, and then interpret it at startup.
+sysf file 'sys.f'
+sysf.len = $ - sysf
+