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 {
33 ;; The code in this macro is placed at the end of each Forth word. When we are
34 ;; executing a definition, this code is what causes execution to resume at the
35 ;; next word in that definition.
37 ;; RSI points to the address of the definition of the next word to execute.
38 lodsq ; Load value at RSI into RAX and increment RSI
39 ;; Now RAX contains the location of the next word to execute. The first 8
40 ;; bytes of this word is the address of the codeword, which is what we want
42 jmp qword [rax] ; Jump to the codeword of the current word
45 ;; pushr and popr work on the return stack, whose location is stored in the
56 ;; The following macro generates the dictionary header. It updates the
57 ;; initial_latest_entry variable, which is used as the initial value of the
58 ;; latest_entry variable that is made available at runtime.
60 ;; The header contains a link to the previous entry, the length of the name of
61 ;; the word and the word itself as a string literal.
63 ;; This macro also defines a label LABEL_entry.
64 initial_latest_entry = 0
65 macro header label, name, immediate {
69 dq initial_latest_entry
75 db .string_end - ($ + 1)
80 initial_latest_entry = label#_entry
83 ;; Define a Forth word that is implemented in assembly. See 'header' for details.
84 macro forth_asm label, name, immediate {
85 header label, name, immediate
90 segment readable executable
94 include "impl.asm" ; Misc. subroutines
95 include "bootstrap.asm" ; Forth words encoded in Assembly
98 cld ; Clear direction flag so LODSQ does the right thing.
99 mov rbp, return_stack_top ; Initialize return stack
106 ;; The codeword is the code that will be executed at the beginning of a forth
107 ;; word. It needs to save the old RSI and update it to point to the next word to
109 header DOCOL, 'DOCOL'
110 pushr rsi ; Save old value of RSI on return stack; we will continue execution there after we are done executing this word
111 lea rsi, [rax + 8] ; RAX currently points to the address of the codeword, so we want to continue at RAX+8
112 next ; Execute word pointed to by RSI
114 ;; This word is called at the end of a Forth definition. It just needs to
115 ;; restore the old value of RSI (saved by 'DOCOL') and resume execution.
116 forth_asm EXIT, 'EXIT'
120 ;; LIT is a special word that reads the next "word pointer" and causes it to be
121 ;; placed on the stack rather than executed.
127 ;; Given a string (a pointer following by a size), return the location of the
128 ;; dictionary entry for that word. If no such word exists, return 0.
129 forth_asm FIND, 'FIND'
132 pop [find.search_length]
133 pop [find.search_buffer]
134 mov rsi, [latest_entry] ; Start with the last added word
145 ;; Given an entry in the dictionary, return a pointer to the codeword of that
147 forth_asm TCFA, '>CFA'
149 add rax, 8 + 1 ; [rax] = length of name
150 movzx rbx, byte [rax]
152 add rax, rbx ; [rax] = codeword
156 ;; BRANCH is the fundamental mechanism for branching. BRANCH reads the next word
157 ;; as a signed integer literal and jumps by that offset.
158 forth_asm BRANCH, 'BRANCH'
159 add rsi, [rsi] ; [RSI], which is the next word, contains the offset; we add this to the instruction pointer.
160 next ; Then, we can just continue execution as normal
162 ;; 0BRANCH is like BRANCH, but it jumps only if the top of the stack is zero.
163 forth_asm ZBRANCH, '0BRANCH'
164 ;; Compare top of stack to see if we should branch
171 add rsi, 8 ; We need to skip over the next word, which contains the offset.
174 ;; Duplicate the top of the stack.
175 forth_asm DUP_, 'DUP'
179 ;; Execute the codeword at the given address.
180 forth_asm EXEC, 'EXEC'
184 ;; Expects a character on the stack and prints it to standard output.
185 forth_asm EMIT, 'EMIT'
198 ;; Read a word from standard input and push it onto the stack as a pointer and a
199 ;; size. The pointer is valid until the next call to READ_WORD.
200 forth_asm READ_WORD, 'READ-WORD'
210 ;; Read a word from a buffer. Expects (buffer buffer-length) on the stack.
211 ;; Updates buffer and buffer-length, such that the word has been removed from
212 ;; the buffer. Appends (word-buffer word-buffer-length) to the stack.
213 forth_asm POP_WORD, 'POP-WORD'
221 push rsi ; Updated buffer
222 push rcx ; Length of updated buffer
223 push rdi ; Word buffer
224 push rdx ; Length of word buffer
229 ;; Takes a string on the stack and replaces it with the decimal number that the
230 ;; string represents.
231 forth_asm PARSE_NUMBER, 'PARSE-NUMBER'
233 pop rdi ; String pointer
242 ;; Takes a string (in the form of a pointer and a length on the stack) and
243 ;; prints it to standard output.
244 forth_asm TELL, 'TELL'
256 ;; Exit the program cleanly.
257 forth_asm TERMINATE, 'TERMINATE'
262 ;; Duplicate a pair of elements.
263 forth_asm PAIRDUP, '2DUP'
272 ;; Swap the top two elements on the stack.
273 forth_asm SWAP, 'SWAP'
280 ;; Remove the top element from the stack.
281 forth_asm DROP, 'DROP'
285 forth_asm NOT_, 'NOT'
296 ;; .U prints the value on the stack as an unsigned integer in hexadecimal.
299 mov [.printed_length], 1
300 pop rax ; RAX = value to print
301 push rsi ; Save value of RSI
303 ;; We start by constructing the buffer to print in reverse
308 div rbx ; Put remainer in RDX and quotient in RAX
310 ;; Place the appropriate character in the buffer
319 ;; .printed_length is the number of characters that we ulitmately want to
320 ;; print. If we have printed a non-zero character, then we should update
323 je .skip_updating_real_length
325 mov [.printed_length], rbx
326 .skip_updating_real_length:
331 ;; Flip buffer around, since it is currently reversed
332 mov rcx, [.printed_length]
340 add rdi, [.printed_length]
348 mov rdx, [.printed_length]
351 ;; Restore RSI and continue execution
355 ;; Takes a value and an address, and stores the value at the given address.
362 ;; Takes an address and returns the value at the given address.
369 forth_asm PUT_BYTE, 'C!'
375 forth_asm GET_BYTE, 'C@'
377 movzx rax, byte [rax]
381 ;; Add two integers on the stack.
389 ;; Calculate difference between two integers on the stack. The second number is
390 ;; subtracted from the first.
398 ;; Given two integers a and b on the stack, pushes the quotient and remainder of
399 ;; division of a by b.
400 forth_asm TIMESMOD, '/MOD'
409 ;; Read user input until next " character is found. Push a string containing the
410 ;; input on the stack as (buffer length). Note that the buffer is only valid
411 ;; until the next call to S" and that no more than 255 character can be read.
412 forth_asm READ_STRING, 'S"'
420 mov rsi, .char_buffer
424 mov al, [.char_buffer]
442 ;; CREATE inserts a new header in the dictionary, and updates LATEST so that it
443 ;; points to the header. To compile a word, the user can then call ',' to
444 ;; continue to append data after the header.
446 ;; It takes the name of the word as a string (address length) on the stack.
447 forth_asm CREATE, 'CREATE'
448 pop rcx ; Word string length
449 pop rdx ; Word string pointer
451 mov rdi, [here] ; rdi = Address at which to insert this entry
452 mov rax, [latest_entry] ; rax = Address of the previous entry
453 mov [rdi], rax ; Insert link to previous entry
454 mov [latest_entry], rdi ; Update LATEST to point to this word
457 mov [rdi], byte 0 ; Insert immediate flag
460 mov [rdi], byte cl ; Insert length
462 ;; Insert word string
466 mov rsi, rdx ; rsi = Word string pointer
489 forth_asm PICK, 'PICK'
491 lea rax, [rsp + 8 * rax]
513 ;; Built-in variables:
519 forth LATEST, 'LATEST'
527 forth SYSCODE, 'SYSCODE'
532 segment readable writable
534 ;; The LATEST variable holds a pointer to the word that was last added to the
535 ;; dictionary. This pointer is updated as new words are added, and its value is
536 ;; used by FIND to look up words.
537 latest_entry dq initial_latest_entry
539 ;; The STATE variable is 0 when the interpreter is executing, and non-zero when
548 READ_STRING.char_buffer db ?
549 READ_STRING.buffer rb $FF
550 READ_STRING.length dq ?
552 DOTU.chars db '0123456789ABCDEF'
553 DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex
556 DOTU.printed_length dq ?
558 ;; Reserve space for compiled words, accessed through HERE.
568 ;; We store some Forth code in sys.f that defined common words that the user
569 ;; would expect to have available at startup. To execute these words, we just
570 ;; include the file directly in the binary, and then interpret it at startup.