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
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
;; 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 [.search_length]
+ pop [.search_buffer]
+
+ ;; RSI contains the entry we are currently looking at
+ mov rsi, [latest_entry] ; Start with the last added word
+
+.loop:
+ movzx rcx, byte [rsi + 8] ; Length of word being looked at
+ cmp rcx, [.search_length]
+ jne .next ; If the words don't have the same length, we have the wrong word
+
+ ;; Otherwise, we need to compare strings
+ lea rdx, [rsi + 8 + 1] ; Location of character being compared in entry
+ mov rdi, [.search_buffer] ; Location of character being compared in search buffer
+.compare_char:
+ mov al, [rdx]
+ mov ah, [rdi]
+ cmp al, ah
+ jne .next ; They don't match; try again
+ inc rdx ; These characters match; look at the next ones
+ inc rdi
+ loop .compare_char
+
+ jmp .found ; They match! We are done.
+
+.next:
+ mov rsi, [rsi] ; Look at the previous entry
+ cmp rsi, 0
+ jnz .loop ; If there is no previous word, exit and return 0
+
+.found:
+ 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
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
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: ; 400170
- dq .start
-.start:
+forth_asm READ_WORD, 'READ-WORD'
mov [.rsi], rsi
mov [.rax], rax
next
+;; Takes a string on the stack and replaces it with the decimal number that the
+;; string represents.
+forth_asm PARSE_NUMBER, 'PARSE-NUMBER'
+ 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
+
+ dec rcx
+ jnz .loop
+
+ push r8
+
+ 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.
-TELL:
- dq .start
-.start:
+forth_asm TELL, 'TELL'
mov rbx, rsi
mov rcx, rax
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
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
push rsi ; Save value of RSI
pop rsi
next
-MAIN:
- dq docol
+forth MAIN, 'MAIN'
dq HELLO
- dq LIT, 1234567890, DOTU, NEWLINE
- dq LIT, $ABCD, DOTU, NEWLINE
- dq LIT, $1234ABCD5678EFAB, DOTU, NEWLINE
+ dq READ_WORD, FIND, TCFA, EXEC
+ dq BRANCH, -8 * 5
dq TERMINATE
segment readable writable
+latest_entry dq initial_latest_entry
+
+SPACE_string db 'SPACE'
+.length = $ - SPACE_string
+HELLO_string db 'HELLO'
+.length = $ - HELLO_string
+DOTU_string db '.U'
+.length = $ - DOTU_string
+HELLA_string db 'HELLA'
+.length = $ - HELLA_string
+
+
you_typed_string db 'You typed: '
.length = $ - you_typed_string
+FIND.search_length dq ?
+FIND.search_buffer dq ?
+FIND.rsi dq ?
+
READ_WORD.rsi dq ?
READ_WORD.rax dq ?
READ_WORD.max_size = $FF
DOTU.length dq ?
DOTU.printed_length dq ?
+PARSE_NUMBER.length dq ?
+
;; Return stack
rq $2000
return_stack_top: