-format ELF64 executable
+;; vim: syntax=fasm
+
+;; At compile-time we load the module given by the environment variable
+;; OS_INCLUDE. All of the following these procedures should preserve the value
+;; of RSI and RSP. They may use other registers as they like.
+;;
+;; The module should provide the following:
+;;
+;; os_code_section
+;; Macro to start the text segment.
+;;
+;; os_data_section
+;; Macro to start the data segment.
+;;
+;; os_initialize
+;; Called at initialization.
+;;
+;; os_print_string
+;; Takes a string buffer in RCX and the length in RDX, and prints the string
+;; to the console.
+;;
+;; os_read_char
+;; Wait for the user to type a key, and then put the corresponding ASCII byte
+;; into RAX.
+;;
+;; os_terminate
+;; Shut down the system, returning the error code given in RAX.
+include '%OS_INCLUDE%'
;; The code in this macro is placed at the end of each Forth word. When we are
;; executing a definition, this code is what causes execution to resume at the
;;
;; This macro also defines a label LABEL_entry.
initial_latest_entry = 0
-macro header label, name {
+macro header label, name, immediate {
local .string_end
label#_entry:
dq initial_latest_entry
+ if immediate eq
+ db 0
+ else
+ db 1
+ end if
db .string_end - ($ + 1)
db name
.string_end:
}
;; Define a Forth word that is implemented in assembly. See 'header' for details.
-macro forth_asm label, name {
- header label, name
+macro forth_asm label, name, immediate {
+ header label, name, immediate
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
+os_code_section
-entry main
-
-include "impl.asm"
+include "impl.asm" ; Misc. subroutines
+include "bootstrap.asm" ; Forth words encoded in Assembly
main:
cld ; Clear direction flag so LODSQ does the right thing.
mov rbp, return_stack_top ; Initialize return stack
+ call os_initialize
+
mov rax, MAIN
jmp qword [rax]
;; The codeword is the code that will be executed at the beginning of a forth
;; word. It needs to save the old RSI and update it to point to the next word to
;; execute.
-docol:
+header DOCOL, 'DOCOL'
pushr rsi ; Save old value of RSI on return stack; we will continue execution there after we are done executing this word
lea rsi, [rax + 8] ; RAX currently points to the address of the codeword, so we want to continue at RAX+8
next ; Execute word pointed to by RSI
;; 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.
+;; restore the old value of RSI (saved by 'DOCOL') and resume execution.
forth_asm EXIT, 'EXIT'
popr rsi
next
push rax
next
+;; When LITSTRING is encountered while executing a word, it instead reads a
+;; string from the definition of that word, and places that string on the stack
+;; as (buffer, length).
+forth_asm LITSTRING, 'LITSTRING'
+ lodsb
+ push rsi ; Buffer
+ movzx rax, al
+ push rax ; Length
+ add rsi, rax ; Skip over string before resuming execution
+ 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'
;; entry.
forth_asm TCFA, '>CFA'
pop rax
- add rax, 8 ; [rax] = length of name
+ add rax, 8 + 1 ; [rax] = length of name
movzx rbx, byte [rax]
inc rax
add rax, rbx ; [rax] = codeword
forth_asm EMIT, 'EMIT'
pushr rsi
pushr rax
- mov rax, 1
- mov rdi, 1
- lea rsi, [rsp]
+
+ lea rcx, [rsp]
mov rdx, 1
- syscall
+ call os_print_string
+
add rsp, 8
popr rax
popr rsi
next
-;; Prints a newline to standard output.
-forth NEWLINE, 'NEWLINE'
- dq LIT, $A
- dq EMIT
- dq EXIT
+;; Read a single character from the current input stream. Usually, this will wait
+;; for the user to press a key, and then return the corresponding character. When
+;; reading from a special buffer, it will instead return the next characater from
+;; that buffer.
+;;
+;; The ASCII character code is placed on the stack.
+forth_asm KEY, 'KEY'
+ call .impl
+ push rax
+ next
-;; Prints a space to standard output.
-forth SPACE, 'SPACE'
- dq LIT, ' '
- dq EMIT
- dq EXIT
+;; Result in RAX
+.impl:
+ ;; Are we reading from user input or from the input buffer?
+ cmp [input_buffer], 0
+ jne .from_buffer
-;; 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.
+ ;; Reading user input
+ call os_read_char
+ ret
+
+.from_buffer:
+ ;; Reading from buffer
+ mov rax, [input_buffer]
+ movzx rax, byte [rax]
+
+ inc [input_buffer]
+ dec [input_buffer_length]
+ ret
+
+;; Read a word and push it onto the stack as a pointer and a size. The pointer
+;; is valid until the next call to READ_WORD.
forth_asm READ_WORD, 'READ-WORD'
- mov [.rsi], rsi
+ push rsi
+.skip_whitespace:
+ ;; Read characters until one of them is not whitespace.
+ call KEY.impl
+ ;; We consider newlines and spaces to be whitespace.
+ cmp al, ' '
+ je .skip_whitespace
+ cmp al, $A
+ je .skip_whitespace
+
+ ;; We got a character that wasn't whitespace. Now read the actual word.
+ mov [.length], 0
+
+.read_alpha:
+ movzx rbx, [.length]
+ mov rsi, .buffer
+ add rsi, rbx
+ mov [rsi], al
+ inc [.length]
- call read_word
- push rdi ; Buffer
- push rdx ; Length
+ call KEY.impl
+
+ cmp al, ' '
+ je .end
+ cmp al, $A
+ jne .read_alpha
+
+.end:
+ pop rsi
+ push .buffer
+ movzx rax, [.length]
+ push rax
- mov rsi, [.rsi]
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 [parse_number.length] ; Length
- pop [parse_number.buffer] ; String pointer
+ pop rcx ; Length
+ pop rdi ; String pointer
push rsi
call parse_number
push rax ; Result
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.
forth_asm TELL, 'TELL'
- mov rbx, rsi
- mov rcx, rax
+ pushr rax
+ pushr rsi
- mov rax, 1
- mov rdi, 1
- pop rdx ; Length
- pop rsi ; Buffer
- syscall
+ pop rdx ; Length
+ pop rcx ; Buffer
+ call os_print_string
- mov rax, rcx
- mov rsi, rbx
+ popr rsi
+ popr rax
next
;; Exit the program cleanly.
forth_asm TERMINATE, 'TERMINATE'
- mov rax, $3C
- mov rdi, 0
- syscall
-
-forth HELLO, 'HELLO'
- dq LIT, 'H', EMIT
- dq LIT, 'e', EMIT
- dq LIT, 'l', EMIT
- dq LIT, 'l', EMIT
- dq LIT, 'o', EMIT
- dq LIT, '!', EMIT
- dq NEWLINE
- dq EXIT
+ mov rax, 0
+ call os_terminate
;; Duplicate a pair of elements.
forth_asm PAIRDUP, '2DUP'
add rsp, 8
next
-;; The INTERPRET word reads and interprets user input. It's behavior depends on
-;; the current STATE. It provides special handling for integers. (TODO)
-forth INTERPRET, 'INTERPRET'
- ;; Read word
- dq READ_WORD
- dq PAIRDUP
- ;; Stack is (word length word length).
- dq FIND ; Try to find word
- dq DUP_
- dq ZBRANCH, 8 * 8 ; Check if word is found
-
- ;; Word is found, execute it
- dq TCFA
- ;; Stack is (word length addr)
- dq SWAP, DROP
- dq SWAP, DROP
- ;; Stack is (addr)
- dq EXEC
- dq EXIT
-
- ;; No word is found, assume it is an integer literal
- ;; Stack is (word length addr)
- dq DROP
- dq PARSE_NUMBER
- dq EXIT
+forth_asm NOT_, 'NOT'
+ pop rax
+ cmp rax, 0
+ jz .false
+.true:
+ push 0
+ next
+.false:
+ push 1
+ next
;; .U prints the value on the stack as an unsigned integer in hexadecimal.
forth_asm DOTU, '.U'
loop .flip
;; Print the buffer
- mov rax, 1
- mov rdi, 1
- mov rsi, .buffer
+ mov rcx, .buffer
mov rdx, [.printed_length]
- syscall
+ call os_print_string
;; Restore RSI and continue execution
pop rsi
push rax
next
+forth_asm PUT_BYTE, 'C!'
+ pop rbx
+ pop rax ; Value
+ mov [rbx], al
+ next
+
+forth_asm GET_BYTE, 'C@'
+ pop rax
+ movzx rax, byte [rax]
+ push rax
+ next
+
;; Add two integers on the stack.
forth_asm PLUS, '+'
pop rax
push rbx
next
-;; Get the location of the STATE variable. It can be set with '!' and read with
-;; '@'.
+;; Given two integers a and b on the stack, pushes the quotient and remainder of
+;; division of a by b.
+forth_asm TIMESMOD, '/MOD'
+ pop rbx ; b
+ pop rax ; a
+ mov rdx, 0
+ div rbx
+ push rax ; a / b
+ push rdx ; a % b
+ next
+
+;; Read input until next " character is found. Push a string containing the
+;; input on the stack as (buffer length). Note that the buffer is only valid
+;; until the next call to S" and that no more than 255 characters can be read.
+forth_asm READ_STRING, 'S"'
+ ;; If the input buffer is set, we should read from there instead.
+ cmp [input_buffer], 0
+ jne read_string_buffer
+
+ push rsi
+
+ mov [.length], 0
+
+.read_char:
+ call os_read_char
+ cmp al, '"'
+ je .done
+
+ mov rdx, .buffer
+ add rdx, [.length]
+ mov [rdx], al
+ inc [.length]
+ jmp .read_char
+
+.done:
+ pop rsi
+
+ push .buffer
+ push [.length]
+
+ next
+
+read_string_buffer:
+ push rsi
+
+ ;; We borrow READ_STRING's buffer. They won't mind.
+ mov [READ_STRING.length], 0
+
+.read_char:
+ mov rbx, [input_buffer]
+ mov al, [rbx]
+ cmp al, '"'
+ je .done
+
+ mov rdx, READ_STRING.buffer
+ add rdx, [READ_STRING.length]
+ mov [rdx], al
+ inc [READ_STRING.length]
+
+ inc [input_buffer]
+ dec [input_buffer_length]
+
+ jmp .read_char
+
+.done:
+ pop rsi
+
+ ;; Skip closing "
+ inc [input_buffer]
+ dec [input_buffer_length]
+
+ push READ_STRING.buffer
+ push [READ_STRING.length]
+
+ next
+
+;; CREATE inserts a new header in the dictionary, and updates LATEST so that it
+;; points to the header. To compile a word, the user can then call ',' to
+;; continue to append data after the header.
+;;
+;; It takes the name of the word as a string (address length) on the stack.
+forth_asm CREATE, 'CREATE'
+ pop rcx ; Word string length
+ pop rdx ; Word string pointer
+
+ mov rdi, [here] ; rdi = Address at which to insert this entry
+ mov rax, [latest_entry] ; rax = Address of the previous entry
+ mov [rdi], rax ; Insert link to previous entry
+ mov [latest_entry], rdi ; Update LATEST to point to this word
+
+ add rdi, 8
+ mov [rdi], byte 0 ; Insert immediate flag
+
+ add rdi, 1
+ mov [rdi], byte cl ; Insert length
+
+ ;; Insert word string
+ add rdi, 1
+
+ push rsi
+ mov rsi, rdx ; rsi = Word string pointer
+ rep movsb
+ pop rsi
+
+ ;; Update HERE
+ mov [here], rdi
+
+ next
+
+forth_asm TICK, "'"
+ lodsq
+ push rax
+ next
+
+forth_asm ROT, 'ROT'
+ pop rax
+ pop rbx
+ pop rdx
+ push rax
+ push rdx
+ push rbx
+ next
+
+forth_asm PICK, 'PICK'
+ pop rax
+ lea rax, [rsp + 8 * rax]
+ mov rax, [rax]
+ push rax
+ next
+
+forth_asm EQL, '='
+ pop rax
+ pop rbx
+ cmp rax, rbx
+ je .eq
+.noteq:
+ push 0
+ next
+.eq:
+ push 1
+ next
+
+forth MAIN, 'MAIN'
+ dq SYSCODE
+ dq INTERPRET_STRING
+ dq INTERPRET
+ dq BRANCH, -8 * 2
+ dq TERMINATE
+
+;; Built-in variables:
+
forth STATE, 'STATE'
dq LIT, var_STATE
dq EXIT
-;; Get the location of the LATEST variable. It can be set with '!' and read with
-;; '@'.
forth LATEST, 'LATEST'
dq LIT, latest_entry
dq EXIT
-;; Get the location at which compiled words are expected to be added. This
-;; pointer is usually modified automatically when calling ',', but we can also
-;; read it manually with 'HERE'.
forth HERE, 'HERE'
dq LIT, here
dq EXIT
-forth COMMA, ','
- dq HERE, GET, PUT ; Set the memory at the address pointed to by HERE
- dq HERE, GET, LIT, 8, PLUS ; Calculate new address for HERE to point to
- dq HERE, PUT ; Update HERE to point to the new address
+forth SYSCODE, 'SYSCODE'
+ dq LIT, sysf
+ dq LIT, sysf.len
dq EXIT
-forth MAIN, 'MAIN'
- dq HELLO
- dq INTERPRET
- dq BRANCH, -8 * 2
- dq TERMINATE
+forth INPUT_BUFFER, 'INPUT-BUFFER'
+ dq LIT, input_buffer
+ dq EXIT
-segment readable writable
+forth INPUT_LENGTH, 'INPUT-LENGTH'
+ dq LIT, input_buffer_length
+ dq EXIT
+
+os_data_section
;; The LATEST variable holds a pointer to the word that was last added to the
;; dictionary. This pointer is updated as new words are added, and its value is
;; it is compiling.
var_STATE dq 0
+;; The interpreter can read either from standard input or from a buffer. When
+;; input-buffer is set (non-null), words like READ-WORD and S" will use this
+;; buffer instead of reading user input.
+input_buffer dq 0
+input_buffer_length dq 0
+
FIND.rsi dq ?
READ_WORD.rsi dq ?
READ_WORD.rbp dq ?
+READ_STRING.char_buffer db ?
+READ_STRING.buffer rb $FF
+READ_STRING.length dq ?
+
DOTU.chars db '0123456789ABCDEF'
DOTU.buffer rq 16 ; 64-bit number has no more than 16 digits in hex
DOTU.rbuffer rq 16
DOTU.length dq ?
DOTU.printed_length dq ?
+KEY.buffer dq ?
+
+READ_WORD.buffer rb $FF
+READ_WORD.length db ?
+
;; Reserve space for compiled words, accessed through HERE.
here dq here_top
-here_top rq $2000
+here_top rq $4000
;; Return stack
rq $2000
return_stack_top:
+
+;; We store some Forth code in sys.f that defined common words that the user
+;; would expect to have available at startup. To execute these words, we just
+;; include the file directly in the binary, and then interpret it at startup.
+sysf:
+file 'sys.f'
+file 'example.f'
+sysf.len = $ - sysf
+