3 format ELF64 executable
7 ;; [NOTE] Volatile registers Linux (syscalls) vs UEFI
9 ;; Linux syscalls: RAX, RCX, R11
10 ;; UEFI: RAX, RCX, R11, RDX, R8, R9, R10
12 ;; We are in the process of replacing our dependency on Linux with a dependency
13 ;; on UEFI. The following macros attempt to isolate what would be syscalls in
14 ;; Linux; thus, we will be able to replace these with UEFI-based implementations,
15 ;; and in theory we should expect the program to work.
17 ;; Print a string of a given length.
20 ;; - RCX = Pointer to buffer
21 ;; - RDX = Buffer length
23 ;; Clobbers: RAX, RCX, R11, RDI, RSI
24 macro sys_print_string {
31 ;; Read a character from the user into the given buffer.
34 ;; - RSI = Character buffer
37 ;; - BYTE [RSI] = Character
39 ;; Clobbers: RAX, RCX, R11, RDI, RSI, RDX
47 macro sys_terminate code {
55 ;; The code in this macro is placed at the end of each Forth word. When we are
56 ;; executing a definition, this code is what causes execution to resume at the
57 ;; next word in that definition.
59 ;; RSI points to the address of the definition of the next word to execute.
60 lodsq ; Load value at RSI into RAX and increment RSI
61 ;; Now RAX contains the location of the next word to execute. The first 8
62 ;; bytes of this word is the address of the codeword, which is what we want
64 jmp qword [rax] ; Jump to the codeword of the current word
67 ;; pushr and popr work on the return stack, whose location is stored in the
78 ;; The following macro generates the dictionary header. It updates the
79 ;; initial_latest_entry variable, which is used as the initial value of the
80 ;; latest_entry variable that is made available at runtime.
82 ;; The header contains a link to the previous entry, the length of the name of
83 ;; the word and the word itself as a string literal.
85 ;; This macro also defines a label LABEL_entry.
86 initial_latest_entry = 0
87 macro header label, name, immediate {
91 dq initial_latest_entry
97 db .string_end - ($ + 1)
102 initial_latest_entry = label#_entry
105 ;; Define a Forth word that is implemented in assembly. See 'header' for details.
106 macro forth_asm label, name, immediate {
107 header label, name, immediate
112 segment readable executable
116 include "impl.asm" ; Misc. subroutines
117 include "bootstrap.asm" ; Forth words encoded in Assembly
120 cld ; Clear direction flag so LODSQ does the right thing.
121 mov rbp, return_stack_top ; Initialize return stack
128 ;; The codeword is the code that will be executed at the beginning of a forth
129 ;; word. It needs to save the old RSI and update it to point to the next word to
131 header DOCOL, 'DOCOL'
132 pushr rsi ; Save old value of RSI on return stack; we will continue execution there after we are done executing this word
133 lea rsi, [rax + 8] ; RAX currently points to the address of the codeword, so we want to continue at RAX+8
134 next ; Execute word pointed to by RSI
136 ;; This word is called at the end of a Forth definition. It just needs to
137 ;; restore the old value of RSI (saved by 'DOCOL') and resume execution.
138 forth_asm EXIT, 'EXIT'
142 ;; LIT is a special word that reads the next "word pointer" and causes it to be
143 ;; placed on the stack rather than executed.
149 ;; Given a string (a pointer following by a size), return the location of the
150 ;; dictionary entry for that word. If no such word exists, return 0.
151 forth_asm FIND, 'FIND'
154 pop [find.search_length]
155 pop [find.search_buffer]
156 mov rsi, [latest_entry] ; Start with the last added word
167 ;; Given an entry in the dictionary, return a pointer to the codeword of that
169 forth_asm TCFA, '>CFA'
171 add rax, 8 + 1 ; [rax] = length of name
172 movzx rbx, byte [rax]
174 add rax, rbx ; [rax] = codeword
178 ;; BRANCH is the fundamental mechanism for branching. BRANCH reads the next word
179 ;; as a signed integer literal and jumps by that offset.
180 forth_asm BRANCH, 'BRANCH'
181 add rsi, [rsi] ; [RSI], which is the next word, contains the offset; we add this to the instruction pointer.
182 next ; Then, we can just continue execution as normal
184 ;; 0BRANCH is like BRANCH, but it jumps only if the top of the stack is zero.
185 forth_asm ZBRANCH, '0BRANCH'
186 ;; Compare top of stack to see if we should branch
193 add rsi, 8 ; We need to skip over the next word, which contains the offset.
196 ;; Duplicate the top of the stack.
197 forth_asm DUP_, 'DUP'
201 ;; Execute the codeword at the given address.
202 forth_asm EXEC, 'EXEC'
206 ;; Expects a character on the stack and prints it to standard output.
207 forth_asm EMIT, 'EMIT'
220 ;; Read a word and push it onto the stack as a pointer and a size. The pointer
221 ;; is valid until the next call to READ_WORD.
222 forth_asm READ_WORD, 'READ-WORD'
223 ;; Are we reading from user input or from the input buffer?
224 cmp [input_buffer], 0
227 ;; Reading user input
238 ;; Reading from buffer
241 mov rsi, [input_buffer]
242 mov rcx, [input_buffer_length]
246 mov [input_buffer], rsi ; Updated buffer
247 mov [input_buffer_length], rcx ; Length of updated buffer
248 push rdi ; Word buffer
249 push rdx ; Length of word buffer
254 ;; Takes a string on the stack and replaces it with the decimal number that the
255 ;; string represents.
256 forth_asm PARSE_NUMBER, 'PARSE-NUMBER'
258 pop rdi ; String pointer
267 ;; Takes a string (in the form of a pointer and a length on the stack) and
268 ;; prints it to standard output.
269 forth_asm TELL, 'TELL'
281 ;; Exit the program cleanly.
282 forth_asm TERMINATE, 'TERMINATE'
285 ;; Duplicate a pair of elements.
286 forth_asm PAIRDUP, '2DUP'
295 ;; Swap the top two elements on the stack.
296 forth_asm SWAP, 'SWAP'
303 ;; Remove the top element from the stack.
304 forth_asm DROP, 'DROP'
308 forth_asm NOT_, 'NOT'
319 ;; .U prints the value on the stack as an unsigned integer in hexadecimal.
322 mov [.printed_length], 1
323 pop rax ; RAX = value to print
324 push rsi ; Save value of RSI
326 ;; We start by constructing the buffer to print in reverse
331 div rbx ; Put remainer in RDX and quotient in RAX
333 ;; Place the appropriate character in the buffer
342 ;; .printed_length is the number of characters that we ulitmately want to
343 ;; print. If we have printed a non-zero character, then we should update
346 je .skip_updating_real_length
348 mov [.printed_length], rbx
349 .skip_updating_real_length:
354 ;; Flip buffer around, since it is currently reversed
355 mov rcx, [.printed_length]
363 add rdi, [.printed_length]
371 mov rdx, [.printed_length]
374 ;; Restore RSI and continue execution
378 ;; Takes a value and an address, and stores the value at the given address.
385 ;; Takes an address and returns the value at the given address.
392 forth_asm PUT_BYTE, 'C!'
398 forth_asm GET_BYTE, 'C@'
400 movzx rax, byte [rax]
404 ;; Add two integers on the stack.
412 ;; Calculate difference between two integers on the stack. The second number is
413 ;; subtracted from the first.
421 ;; Given two integers a and b on the stack, pushes the quotient and remainder of
422 ;; division of a by b.
423 forth_asm TIMESMOD, '/MOD'
432 ;; Read input until next " character is found. Push a string containing the
433 ;; input on the stack as (buffer length). Note that the buffer is only valid
434 ;; until the next call to S" and that no more than 255 characters can be read.
435 forth_asm READ_STRING, 'S"'
436 ;; If the input buffer is set, we should read from there instead.
437 cmp [input_buffer], 0
438 jne read_string_buffer
445 mov rsi, .char_buffer
448 mov al, [.char_buffer]
469 ;; We borrow READ_STRING's buffer. They won't mind.
470 mov [READ_STRING.length], 0
473 mov rbx, [input_buffer]
478 mov rdx, READ_STRING.buffer
479 add rdx, [READ_STRING.length]
481 inc [READ_STRING.length]
484 dec [input_buffer_length]
493 dec [input_buffer_length]
495 push READ_STRING.buffer
496 push [READ_STRING.length]
500 ;; CREATE inserts a new header in the dictionary, and updates LATEST so that it
501 ;; points to the header. To compile a word, the user can then call ',' to
502 ;; continue to append data after the header.
504 ;; It takes the name of the word as a string (address length) on the stack.
505 forth_asm CREATE, 'CREATE'
506 pop rcx ; Word string length
507 pop rdx ; Word string pointer
509 mov rdi, [here] ; rdi = Address at which to insert this entry
510 mov rax, [latest_entry] ; rax = Address of the previous entry
511 mov [rdi], rax ; Insert link to previous entry
512 mov [latest_entry], rdi ; Update LATEST to point to this word
515 mov [rdi], byte 0 ; Insert immediate flag
518 mov [rdi], byte cl ; Insert length
520 ;; Insert word string
524 mov rsi, rdx ; rsi = Word string pointer
547 forth_asm PICK, 'PICK'
549 lea rax, [rsp + 8 * rax]
573 ;; Built-in variables:
579 forth LATEST, 'LATEST'
587 forth SYSCODE, 'SYSCODE'
592 forth INPUT_BUFFER, 'INPUT-BUFFER'
596 forth INPUT_LENGTH, 'INPUT-LENGTH'
597 dq LIT, input_buffer_length
600 segment readable writable
602 ;; The LATEST variable holds a pointer to the word that was last added to the
603 ;; dictionary. This pointer is updated as new words are added, and its value is
604 ;; used by FIND to look up words.
605 latest_entry dq initial_latest_entry
607 ;; The STATE variable is 0 when the interpreter is executing, and non-zero when
611 ;; The interpreter can read either from standard input or from a buffer. When
612 ;; input-buffer is set (non-null), words like READ-WORD and S" will use this
613 ;; buffer instead of reading user input.
615 input_buffer_length dq 0
622 READ_STRING.char_buffer db ?
623 READ_STRING.buffer rb $FF
624 READ_STRING.length dq ?
626 DOTU.chars db '0123456789ABCDEF'
627 DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex
630 DOTU.printed_length dq ?
632 ;; Reserve space for compiled words, accessed through HERE.
642 ;; We store some Forth code in sys.f that defined common words that the user
643 ;; would expect to have available at startup. To execute these words, we just
644 ;; include the file directly in the binary, and then interpret it at startup.