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 {
39 dq initial_latest_entry
40 db .string_end - ($ + 1)
45 initial_latest_entry = label#_entry
48 ;; Define a Forth word that is implemented in assembly. See 'header' for details.
49 macro forth_asm label, name {
55 ;; Define a Forth word that is implemented in Forth. (The body will be a list of
57 macro forth label, name {
62 segment readable executable
65 cld ; Clear direction flag so LODSQ does the right thing.
66 mov rbp, return_stack_top ; Initialize return stack
73 ;; The codeword is the code that will be executed at the beginning of a forth
74 ;; word. It needs to save the old RSI and update it to point to the next word to
77 pushr rsi ; Save old value of RSI on return stack; we will continue execution there after we are done executing this word
78 lea rsi, [rax + 8] ; RAX currently points to the address of the codeword, so we want to continue at RAX+8
79 next ; Execute word pointed to by RSI
81 ;; This word is called at the end of a Forth definition. It just needs to
82 ;; restore the old value of RSI (saved by 'docol') and resume execution.
83 forth_asm EXIT, 'EXIT'
87 ;; LIT is a special word that reads the next "word pointer" and causes it to be
88 ;; placed on the stack rather than executed.
94 ;; Given a string (a pointer following by a size), return the location of the
95 ;; dictionary entry for that word. If no such word exists, return 0.
96 forth_asm FIND, 'FIND'
101 ;; RSI contains the entry we are currently looking at
102 mov rsi, [latest_entry] ; Start with the last added word
105 movzx rcx, byte [rsi + 8] ; Length of word being looked at
106 cmp rcx, [.search_length]
107 jne .next ; If the words don't have the same length, we have the wrong word
109 ;; Otherwise, we need to compare strings
110 lea rdx, [rsi + 8 + 1] ; Location of character being compared in entry
111 mov rdi, [.search_buffer] ; Location of character being compared in search buffer
116 jne .next ; They don't match; try again
117 inc rdx ; These characters match; look at the next ones
121 jmp .found ; They match! We are done.
124 mov rsi, [rsi] ; Look at the previous entry
126 jnz .loop ; If there is no previous word, exit and return 0
134 ;; Given an entry in the dictionary, return a pointer to the codeword of that
136 forth_asm TCFA, '>CFA'
138 add rax, 8 ; [rax] = length of name
139 movzx rbx, byte [rax]
141 add rax, rbx ; [rax] = codeword
145 ;; BRANCH is the fundamental mechanism for branching. BRANCH reads the next word
146 ;; as a signed integer literal and jumps by that offset.
147 forth_asm BRANCH, 'BRANCH'
148 add rsi, [rsi] ; [RSI], which is the next word, contains the offset; we add this to the instruction pointer.
149 next ; Then, we can just continue execution as normal
151 ;; 0BRANCH is like BRANCH, but it jumps only if the top of the stack is zero.
152 forth_asm ZBRANCH, '0BRANCH'
153 ;; Compare top of stack to see if we should branch
160 add rsi, 8 ; We need to skip over the next word, which contains the offset.
163 ;; Duplicate the top of the stack.
164 forth_asm DUP_, 'DUP'
168 ;; Execute the codeword at the given address.
169 forth_asm EXEC, 'EXEC'
173 ;; Expects a character on the stack and prints it to standard output.
174 forth_asm EMIT, 'EMIT'
187 ;; Prints a newline to standard output.
188 forth NEWLINE, 'NEWLINE'
193 ;; Prints a space to standard output.
199 ;; Read a word from standard input and push it onto the stack as a pointer and a
200 ;; size. The pointer is valid until the next call to READ_WORD.
201 forth_asm READ_WORD, 'READ-WORD'
206 ;; Read characters into .char_buffer until one of them is not whitespace.
209 mov rsi, .char_buffer
213 cmp [.char_buffer], ' '
215 cmp [.char_buffer], $A
219 ;; We got a character that wasn't whitespace. Now read the actual word.
223 mov al, [.char_buffer]
232 mov rsi, .char_buffer
236 cmp [.char_buffer], ' '
238 cmp [.char_buffer], $A
251 ;; Takes a string on the stack and replaces it with the decimal number that the
252 ;; string represents.
253 forth_asm PARSE_NUMBER, 'PARSE-NUMBER'
254 pop [.length] ; Length
255 pop rdi ; String pointer
258 ;; Add (10^(rcx-1) * parse_char(rdi[length - rcx])) to the accumulated value
262 ;; First, calcuate 10^(rcx - 1)
274 ;; Now, rax = 10^(rcx - 1).
276 ;; We need to calulate the value of the character at rdi[length - rcx].
280 movzx rbx, byte [rbx]
283 ;; Multiply this value by rax to get (10^(rcx-1) * parse_char(rdi[length - rcx])),
284 ;; then add this to the result.
287 ;; Add that value to r8
297 forth READ_NUMBER, 'READ-NUMBER'
302 ;; Takes a string (in the form of a pointer and a length on the stack) and
303 ;; prints it to standard output.
304 forth_asm TELL, 'TELL'
318 ;; Exit the program cleanly.
319 forth_asm TERMINATE, 'TERMINATE'
334 ;; .U prints the value on the stack as an unsigned integer in hexadecimal.
337 mov [.printed_length], 1
338 pop rax ; RAX = value to print
339 push rsi ; Save value of RSI
341 ;; We start by constructing the buffer to print in reverse
346 div rbx ; Put remainer in RDX and quotient in RAX
348 ;; Place the appropriate character in the buffer
357 ;; .printed_length is the number of characters that we ulitmately want to
358 ;; print. If we have printed a non-zero character, then we should update
361 je .skip_updating_real_length
363 mov [.printed_length], rbx
364 .skip_updating_real_length:
369 ;; Flip buffer around, since it is currently reversed
370 mov rcx, [.printed_length]
378 add rdi, [.printed_length]
388 mov rdx, [.printed_length]
391 ;; Restore RSI and continue execution
397 dq READ_WORD, FIND, TCFA, EXEC
401 segment readable writable
403 latest_entry dq initial_latest_entry
405 FIND.search_length dq ?
406 FIND.search_buffer dq ?
411 READ_WORD.max_size = $FF
412 READ_WORD.buffer rb READ_WORD.max_size
413 READ_WORD.length db ?
414 READ_WORD.char_buffer db ?
416 DOTU.chars db '0123456789ABCDEF'
417 DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex
420 DOTU.printed_length dq ?
422 PARSE_NUMBER.length dq ?