1 format ELF64 executable
3 ;; The code in this macro is placed at the end of each Forth word. When we are
4 ;; executing a definition, this code is what causes execution to resume at the
5 ;; next word in that definition.
7 ;; RSI points to the address of the definition of the next word to execute.
8 lodsq ; Load value at RSI into RAX and increment RSI
9 ;; Now RAX contains the location of the next word to execute. The first 8
10 ;; bytes of this word is the address of the codeword, which is what we want
12 jmp qword [rax] ; Jump to the codeword of the current word
15 ;; pushr and popr work on the return stack, whose location is stored in the
26 ;; The following macro generates the dictionary header. It updates the
27 ;; initial_latest_entry variable, which is used as the initial value of the
28 ;; latest_entry variable that is made available at runtime.
30 ;; The header contains a link to the previous entry, the length of the name of
31 ;; the word and the word itself as a string literal.
33 ;; This macro also defines a label LABEL_entry.
34 initial_latest_entry = 0
35 macro header label, name, immediate {
39 dq initial_latest_entry
41 db .string_end - ($ + 1)
46 initial_latest_entry = label#_entry
49 ;; Define a Forth word that is implemented in assembly. See 'header' for details.
50 macro forth_asm label, name, immediate {
51 header label, name, immediate
56 ;; Define a Forth word that is implemented in Forth. (The body will be a list of
58 macro forth label, name, immediate {
59 header label, name, immediate
63 segment readable executable
70 cld ; Clear direction flag so LODSQ does the right thing.
71 mov rbp, return_stack_top ; Initialize return stack
78 ;; The codeword is the code that will be executed at the beginning of a forth
79 ;; word. It needs to save the old RSI and update it to point to the next word to
82 pushr rsi ; Save old value of RSI on return stack; we will continue execution there after we are done executing this word
83 lea rsi, [rax + 8] ; RAX currently points to the address of the codeword, so we want to continue at RAX+8
84 next ; Execute word pointed to by RSI
86 ;; This word is called at the end of a Forth definition. It just needs to
87 ;; restore the old value of RSI (saved by 'DOCOL') and resume execution.
88 forth_asm EXIT, 'EXIT'
92 ;; LIT is a special word that reads the next "word pointer" and causes it to be
93 ;; placed on the stack rather than executed.
99 ;; Given a string (a pointer following by a size), return the location of the
100 ;; dictionary entry for that word. If no such word exists, return 0.
101 forth_asm FIND, 'FIND'
104 pop [find.search_length]
105 pop [find.search_buffer]
106 mov rsi, [latest_entry] ; Start with the last added word
117 ;; Given an entry in the dictionary, return a pointer to the codeword of that
119 forth_asm TCFA, '>CFA'
121 add rax, 16 ; [rax] = length of name
122 movzx rbx, byte [rax]
124 add rax, rbx ; [rax] = codeword
128 ;; BRANCH is the fundamental mechanism for branching. BRANCH reads the next word
129 ;; as a signed integer literal and jumps by that offset.
130 forth_asm BRANCH, 'BRANCH'
131 add rsi, [rsi] ; [RSI], which is the next word, contains the offset; we add this to the instruction pointer.
132 next ; Then, we can just continue execution as normal
134 ;; 0BRANCH is like BRANCH, but it jumps only if the top of the stack is zero.
135 forth_asm ZBRANCH, '0BRANCH'
136 ;; Compare top of stack to see if we should branch
143 add rsi, 8 ; We need to skip over the next word, which contains the offset.
146 ;; Duplicate the top of the stack.
147 forth_asm DUP_, 'DUP'
151 ;; Execute the codeword at the given address.
152 forth_asm EXEC, 'EXEC'
156 ;; Expects a character on the stack and prints it to standard output.
157 forth_asm EMIT, 'EMIT'
170 ;; Prints a newline to standard output.
171 forth NEWLINE, 'NEWLINE'
176 ;; Prints a space to standard output.
182 ;; Read a word from standard input and push it onto the stack as a pointer and a
183 ;; size. The pointer is valid until the next call to READ_WORD.
184 forth_asm READ_WORD, 'READ-WORD'
194 ;; Takes a string on the stack and replaces it with the decimal number that the
195 ;; string represents.
196 forth_asm PARSE_NUMBER, 'PARSE-NUMBER'
197 pop [parse_number.length] ; Length
198 pop [parse_number.buffer] ; String pointer
207 forth READ_NUMBER, 'READ-NUMBER'
212 ;; Takes a string (in the form of a pointer and a length on the stack) and
213 ;; prints it to standard output.
214 forth_asm TELL, 'TELL'
228 ;; Exit the program cleanly.
229 forth_asm TERMINATE, 'TERMINATE'
244 ;; Duplicate a pair of elements.
245 forth_asm PAIRDUP, '2DUP'
254 ;; Swap the top two elements on the stack.
255 forth_asm SWAP, 'SWAP'
262 ;; Remove the top element from the stack.
263 forth_asm DROP, 'DROP'
267 ;; The INTERPRET word reads and interprets user input. It's behavior depends on
268 ;; the current STATE. It provides special handling for integers. (TODO)
269 forth INTERPRET, 'INTERPRET'
273 ;; Stack is (word length word length).
274 dq FIND ; Try to find word
276 dq ZBRANCH, 8 * 8 ; Check if word is found
278 ;; Word is found, execute it
280 ;; Stack is (word length addr)
287 ;; No word is found, assume it is an integer literal
288 ;; Stack is (word length addr)
293 ;; .U prints the value on the stack as an unsigned integer in hexadecimal.
296 mov [.printed_length], 1
297 pop rax ; RAX = value to print
298 push rsi ; Save value of RSI
300 ;; We start by constructing the buffer to print in reverse
305 div rbx ; Put remainer in RDX and quotient in RAX
307 ;; Place the appropriate character in the buffer
316 ;; .printed_length is the number of characters that we ulitmately want to
317 ;; print. If we have printed a non-zero character, then we should update
320 je .skip_updating_real_length
322 mov [.printed_length], rbx
323 .skip_updating_real_length:
328 ;; Flip buffer around, since it is currently reversed
329 mov rcx, [.printed_length]
337 add rdi, [.printed_length]
347 mov rdx, [.printed_length]
350 ;; Restore RSI and continue execution
354 ;; Takes a value and an address, and stores the value at the given address.
361 ;; Takes an address and returns the value at the given address.
368 ;; Add two integers on the stack.
376 ;; Calculate difference between two integers on the stack. The second number is
377 ;; subtracted from the first.
385 ;; Get the location of the STATE variable. It can be set with '!' and read with
391 ;; Get the location of the LATEST variable. It can be set with '!' and read with
393 forth LATEST, 'LATEST'
397 ;; Get the location at which compiled words are expected to be added. This
398 ;; pointer is usually modified automatically when calling ',', but we can also
399 ;; read it manually with 'HERE'.
405 dq HERE, GET, PUT ; Set the memory at the address pointed to by HERE
406 dq HERE, GET, LIT, 8, PLUS ; Calculate new address for HERE to point to
407 dq HERE, PUT ; Update HERE to point to the new address
410 ;; Read user input until next " character is found. Push a string containing the
411 ;; input on the stack as (buffer length). Note that the buffer is only valid
412 ;; until the next call to S" and that no more than 255 character can be read.
413 forth_asm READ_STRING, 'S"'
421 mov rsi, .char_buffer
425 mov al, [.char_buffer]
443 ;; CREATE inserts a new header in the dictionary, and updates LATEST so that it
444 ;; points to the header. To compile a word, the user can then call ',' to
445 ;; continue to append data after the header.
447 ;; It takes the name of the word as a string (address length) on the stack.
448 forth_asm CREATE, 'CREATE'
449 pop rcx ; Word string length
450 pop rdx ; Word string pointer
452 mov rdi, [here] ; rdi = Address at which to insert this entry
453 mov rax, [latest_entry] ; rax = Address of the previous entry
454 mov [rdi], rax ; Insert link to previous entry
455 mov [latest_entry], rdi ; Update LATEST to point to this word
459 mov [rdi], rax ; Set immediate flag
462 mov [rdi], rcx ; Insert length
464 ;; Insert word string
468 mov rsi, rdx ; rsi = Word string pointer
477 ;; Mark the last added word as immediate.
478 forth IMMEDIATE, 'IMMEDIATE', 1
485 ;; Return 0 if the given word is not immediate.
486 forth IS_IMMEDIATE, 'IMMEDIATE?'
498 segment readable writable
500 ;; The LATEST variable holds a pointer to the word that was last added to the
501 ;; dictionary. This pointer is updated as new words are added, and its value is
502 ;; used by FIND to look up words.
503 latest_entry dq initial_latest_entry
505 ;; The STATE variable is 0 when the interpreter is executing, and non-zero when
514 READ_STRING.char_buffer db ?
515 READ_STRING.buffer rb $FF
516 READ_STRING.length dq ?
518 DOTU.chars db '0123456789ABCDEF'
519 DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex
522 DOTU.printed_length dq ?
524 ;; Reserve space for compiled words, accessed through HERE.