-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
.start:
}
-;; Define a Forth word that is implemented in Forth. (The body will be a list of
-;; 'dq' statements.)
-macro forth label, name, immediate {
- header label, name, immediate
- dq DOCOL
-}
-
-segment readable executable
+include "impl.asm" ; Misc. subroutines
+include "bootstrap.asm" ; Forth words encoded in Assembly
-entry main
-
-include "impl.asm"
+os_code_section
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]
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'
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
+
+ ;; 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 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.
+;; 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
- call read_word
- push rdi ; Buffer
- push rdx ; Length
+.read_alpha:
+ movzx rbx, [.length]
+ mov rsi, .buffer
+ add rsi, rbx
+ mov [rsi], al
+ inc [.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'
push 1
next
-;; The INTERPRET word reads and interprets user input. It's behavior depends on
-;; the current STATE. It provides special handling for integers.
-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 * 20 ; Check if word is found
-
- ;; - Word is found -
-
- dq STATE, GET, ZBRANCH, 8 * 9 ; Check whether we are in compilation or immediate mode
-
- ;; (Word found, compilation mode)
- dq DUP_, IS_IMMEDIATE, NOT_, ZBRANCH, 8 * 4 ; If the word is immediate, continue as we would in immediate mode
-
- ;; Otherwise, we want to compile this word
- dq TCFA
- dq COMMA
- dq EXIT
-
- ;; (Word found, immediate mode)
- ;; Execute word
- 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 STATE, GET, ZBRANCH, 8 * 5 ; Check whether we are in compilation or immediate mode
-
- ;; (Number, compilation mode)
- dq LIT, LIT, COMMA
- dq COMMA
- dq EXIT
-
- ;; (Number, immediate mode)
- dq EXIT
-
;; .U prints the value on the stack as an unsigned integer in hexadecimal.
forth_asm DOTU, '.U'
mov [.length], 0
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 rdx ; a % b
next
-;; Get the location of the STATE variable. It can be set with '!' and read with
-;; '@'.
-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
- dq EXIT
-
-;; Read user input until next " character is found. Push a string containing the
+;; 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 character can be read.
+;; 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:
- mov rax, 0
- mov rdi, 0
- mov rsi, .char_buffer
- mov rdx, 1
- syscall
-
- mov al, [.char_buffer]
+ call os_read_char
cmp al, '"'
je .done
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.
next
-;; Mark the last added word as immediate.
-forth IMMEDIATE, 'IMMEDIATE', 1
- dq LIT, 1
- dq LATEST, GET
- dq LIT, 8, PLUS
- dq PUT_BYTE
- dq EXIT
+forth_asm TICK, "'"
+ lodsq
+ push rax
+ next
-;; Given the address of a word, return 0 if the given word is not immediate.
-forth IS_IMMEDIATE, 'IMMEDIATE?'
- dq LIT, 8, PLUS
- dq GET_BYTE
- dq EXIT
+forth_asm ROT, 'ROT'
+ pop rax
+ pop rbx
+ pop rdx
+ push rax
+ push rdx
+ push rbx
+ next
-;; Enter immediate mode, immediately
-forth INTO_IMMEDIATE, '[', 1
- dq LIT, 0, STATE, PUT_BYTE
- dq EXIT
+forth_asm PICK, 'PICK'
+ pop rax
+ lea rax, [rsp + 8 * rax]
+ mov rax, [rax]
+ push rax
+ next
-;; Enter compilation mode
-forth OUTOF_IMMEDIATE, ']'
- dq LIT, 1, STATE, PUT_BYTE
- dq EXIT
+forth_asm EQL, '='
+ pop rax
+ pop rbx
+ cmp rax, rbx
+ je .eq
+.noteq:
+ push 0
+ next
+.eq:
+ push 1
+ next
forth MAIN, 'MAIN'
- dq HELLO
+ dq SYSCODE
+ dq INTERPRET_STRING
dq INTERPRET
dq BRANCH, -8 * 2
dq TERMINATE
-segment readable writable
+;; EFI:
+
+forth EFI_SYSTEM_TABLE_CONSTANT, 'SystemTable'
+ dq LIT, system_table, GET
+ dq EXIT
+
+forth_asm EFICALL2, 'EFICALL2'
+ pop rax ; function pointer
+ pop rdx ; 2nd argument
+ pop rcx ; 1st argument
+
+ sub rsp, 32
+ call rax
+ add rsp, 32
+
+ next
+
+forth_asm EFICALL3, 'EFICALL3'
+ pop rax ; function pointer
+ pop r8 ; 3rd argument
+ pop rdx ; 2nd argument
+ pop rcx ; 1st argument
+
+ sub rsp, 32
+ call rax
+ add rsp, 32
+
+ push rax
+
+ next
+
+forth_asm EFICALL10, 'EFICALL10'
+ pop rax ; function pointer
+
+ mov rcx, [rsp + 8 * 9]
+ mov rdx, [rsp + 8 * 8]
+ mov r8, [rsp + 8 * 7]
+ mov r9, [rsp + 8 * 6]
+
+ ;; Reverse order of stack arguments
+ mov r10, [rsp + 8 * 5]
+ mov r11, [rsp + 8 * 0]
+ mov [rsp + 8 * 5], r11
+ mov [rsp + 8 * 0], r10
+
+ mov r10, [rsp + 8 * 4]
+ mov r11, [rsp + 8 * 1]
+ mov [rsp + 8 * 4], r11
+ mov [rsp + 8 * 1], r10
+
+ mov r10, [rsp + 8 * 3]
+ mov r11, [rsp + 8 * 2]
+ mov [rsp + 8 * 3], r11
+ mov [rsp + 8 * 2], r10
+
+ sub rsp, 32
+ call rax
+ add rsp, 32 + 8 * 10
+
+ push rax
+
+ next
+
+;; Built-in variables:
+
+forth STATE, 'STATE'
+ dq LIT, var_STATE
+ dq EXIT
+
+forth LATEST, 'LATEST'
+ dq LIT, latest_entry
+ dq EXIT
+
+forth HERE, 'HERE'
+ dq LIT, here
+ dq EXIT
+
+forth SYSCODE, 'SYSCODE'
+ dq LIT, sysf
+ dq LIT, sysf.len
+ dq EXIT
+
+forth INPUT_BUFFER, 'INPUT-BUFFER'
+ dq LIT, input_buffer
+ dq EXIT
+
+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 ?
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 'uefi.f'
+sysf.len = $ - sysf
+