X-Git-Url: https://git.rrq.au/?a=blobdiff_plain;f=main.asm;h=a4db7329bb1f8ed39570b1196f0df2556f75e761;hb=f0df53c35124c9a5e2b045aba60b513faa344567;hp=3187d043f0c730006a7ebf97ccf1c4587844272b;hpb=2f4b8a792fbf943f813bbcf09bff6ddd9f38855b;p=rrq%2Fjonasforth.git

diff --git a/main.asm b/main.asm
index 3187d04..a4db732 100644
--- a/main.asm
+++ b/main.asm
@@ -1,5 +1,35 @@
+;; vim: syntax=fasm
+
 format ELF64 executable
 
+;; "Syscalls" {{{
+
+;; [NOTE] Volatile registers Linux (syscalls) vs UEFI
+;;
+;;   Linux syscalls: RAX, RCX, R11
+;;   UEFI:           RAX, RCX, R11, RDX, R8, R9, R10
+
+;; We are in the process of replacing our dependency on Linux with a dependency
+;; on UEFI. The following macros attempt to isolate what would be syscalls in
+;; Linux; thus, we will be able to replace these with UEFI-based implementations,
+;; and in theory we should expect the program to work.
+
+;; Print a string of a given length.
+;;
+;; Input:
+;; - RCX = Pointer to buffer
+;; - RDX = Buffer length
+;;
+;; Clobbers: RAX, RCX, R11, RDI, RSI
+macro sys_print_string {
+  mov rax, 1
+  mov rdi, 1
+  mov rsi, rcx
+  syscall
+}
+
+;; }}}
+
 ;; 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
 ;; next word in that definition.
@@ -32,11 +62,16 @@ macro popr x {
 ;;
 ;; 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:
@@ -46,21 +81,19 @@ initial_latest_entry = label#_entry
 }
 
 ;; 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
 
+entry main
+
+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
@@ -73,13 +106,13 @@ program: dq MAIN
 ;; 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
@@ -95,37 +128,15 @@ forth_asm LIT, 'LIT'
 ;; dictionary entry for that word. If no such word exists, return 0.
 forth_asm FIND, 'FIND'
   mov [.rsi], rsi
-  pop [.search_length]
-  pop [.search_buffer]
 
-  ;; RSI contains the entry we are currently looking at
+  pop [find.search_length]
+  pop [find.search_buffer]
   mov rsi, [latest_entry]       ; Start with the last added word
+  call find
+  push rsi
 
-.loop:
-  movzx rcx, byte [rsi + 8]     ; Length of word being looked at
-  cmp rcx, [.search_length]
-  jne .next    ; If the words don't have the same length, we have the wrong word
-
-  ;; Otherwise, we need to compare strings
-  lea rdx, [rsi + 8 + 1]        ; Location of character being compared in entry
-  mov rdi, [.search_buffer]     ; Location of character being compared in search buffer
-.compare_char:
-  mov al, [rdx]
-  mov ah, [rdi]
-  cmp al, ah
-  jne .next                     ; They don't match; try again
-  inc rdx                       ; These characters match; look at the next ones
-  inc rdi
-  loop .compare_char
-
-  jmp .found                    ; They match! We are done.
-
-.next:
-  mov rsi, [rsi]                ; Look at the previous entry
-  cmp rsi, 0
-  jnz .loop                    ; If there is no previous word, exit and return 0
-
-.found:
+  mov rsi, [.rsi]
+  next
   push rsi
 
   mov rsi, [.rsi]
@@ -135,7 +146,7 @@ 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
@@ -174,145 +185,72 @@ forth_asm EXEC, 'EXEC'
 forth_asm EMIT, 'EMIT'
   pushr rsi
   pushr rax
-  mov rax, 1
-  mov rdi, 1
-  lea rsi, [rsp]
+
+  lea rcx, [rsp]
   mov rdx, 1
-  syscall
+  sys_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
-
-;; Prints a space to standard output.
-forth SPACE, 'SPACE'
-  dq LIT, ' '
-  dq EMIT
-  dq EXIT
-
 ;; 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.
 forth_asm READ_WORD, 'READ-WORD'
   mov [.rsi], rsi
-  mov [.rax], rax
-
-.skip_whitespace:
-  ;; Read characters into .char_buffer until one of them is not whitespace.
-  mov rax, 0
-  mov rdi, 0
-  mov rsi, .char_buffer
-  mov rdx, 1
-  syscall
-
-  cmp [.char_buffer], ' '
-  je .skip_whitespace
-  cmp [.char_buffer], $A
-  je .skip_whitespace
 
-.alpha:
-  ;; 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:
-  mov al, [.char_buffer]
-  movzx rbx, [.length]
-  mov rsi, .buffer
-  add rsi, rbx
-  mov [rsi], al
-  inc [.length]
+  mov rsi, [.rsi]
+  next
 
-  mov rax, 0
-  mov rdi, 0
-  mov rsi, .char_buffer
-  mov rdx, 1
-  syscall
+;; Read a word from a buffer. Expects (buffer buffer-length) on the stack.
+;; Updates buffer and buffer-length, such that the word has been removed from
+;; the buffer. Appends (word-buffer word-buffer-length) to the stack.
+forth_asm POP_WORD, 'POP-WORD'
+  pushr rsi
 
-  cmp [.char_buffer], ' '
-  je .end
-  cmp [.char_buffer], $A
-  jne .read_alpha
+  pop rcx ; Length
+  pop rsi ; Buffer
 
-.end:
-  push .buffer
-  movzx rax, [.length]
-  push rax
+  call pop_word
 
-  mov rsi, [.rsi]
-  mov rax, [.rax]
+  push rsi ; Updated buffer
+  push rcx ; Length of updated buffer
+  push rdi ; Word buffer
+  push rdx ; Length of word buffer
 
+  popr 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 [.length]                 ; Length
-  pop rdi                       ; String pointer
-  mov r8, 0                     ; Result
-
-  ;; Add (10^(rcx-1) * parse_char(rdi[length - rcx])) to the accumulated value
-  ;; for each rcx.
-  mov rcx, [.length]
-.loop:
-  ;; First, calcuate 10^(rcx - 1)
-  mov rax, 1
-
-  mov r9, rcx
-  .exp_loop:
-    dec r9
-    jz .break
-    mov rbx, 10
-    mul rbx
-    jmp .exp_loop
-  .break:
+  pop rcx     ; Length
+  pop rdi     ; String pointer
 
-  ;; Now, rax = 10^(rcx - 1).
-
-  ;; We need to calulate the value of the character at rdi[length - rcx].
-  mov rbx, rdi
-  add rbx, [.length]
-  sub rbx, rcx
-  movzx rbx, byte [rbx]
-  sub rbx, '0'
-
-  ;; Multiply this value by rax to get (10^(rcx-1) * parse_char(rdi[length - rcx])),
-  ;; then add this to the result.
-  mul rbx
-
-  ;; Add that value to r8
-  add r8, rax
-
-  dec rcx
-  jnz .loop
-
-  push r8
+  push rsi
+  call parse_number
+  pop rsi
 
+  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
+  sys_print_string
 
-  mov rax, rcx
-  mov rsi, rbx
+  popr rsi
+  popr rax
   next
 
 ;; Exit the program cleanly.
@@ -321,15 +259,39 @@ forth_asm TERMINATE, 'TERMINATE'
   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
+;; Duplicate a pair of elements.
+forth_asm PAIRDUP, '2DUP'
+  pop rbx
+  pop rax
+  push rax
+  push rbx
+  push rax
+  push rbx
+  next
+
+;; Swap the top two elements on the stack.
+forth_asm SWAP, 'SWAP'
+  pop rax
+  pop rbx
+  push rax
+  push rbx
+  next
+
+;; Remove the top element from the stack.
+forth_asm DROP, 'DROP'
+  add rsp, 8
+  next
+
+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'
@@ -382,36 +344,264 @@ 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
+  sys_print_string
 
   ;; Restore RSI and continue execution
   pop rsi
   next
 
+;; Takes a value and an address, and stores the value at the given address.
+forth_asm PUT, '!'
+  pop rbx                       ; Address
+  pop rax                       ; Value
+  mov [rbx], rax
+  next
+
+;; Takes an address and returns the value at the given address.
+forth_asm GET, '@'
+  pop rax
+  mov rax, [rax]
+  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
+  pop rbx
+  add rax, rbx
+  push rax
+  next
+
+;; Calculate difference between two integers on the stack. The second number is
+;; subtracted from the first.
+forth_asm MINUS, '-'
+  pop rax
+  pop rbx
+  sub rbx, rax
+  push rbx
+  next
+
+;; 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 user 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.
+;;
+;; [TODO] We want to be able to use this when reading from buffers (e.g. in
+;; INTERPRET-STRING) too!
+forth_asm READ_STRING, 'S"'
+  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]
+  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
+
+;; BUF" works a bit like S, but it reads the string from the current input
+;; buffer described in INPUT-BUFFER and INPUT-LENGTH. We use this fucntion in
+;; sys.f to store strings.
+forth_asm BUF_READ_STRING, 'BUF"'
+  push rsi
+
+  ;; We borrow READ_STRING's buffer. They won't mind.
+  mov [READ_STRING.length], 0
+
+  ;; Skip space ([TODO]: Shouldn't we do this while parsing instead?)
+  inc [input_buffer]
+  dec [input_buffer_length]
+
+.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 HELLO
-  dq READ_WORD, FIND, TCFA, EXEC
-  dq BRANCH, -8 * 5
+  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
+
+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
+
 segment readable writable
 
+;; 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
+;; used by FIND to look up words.
 latest_entry dq initial_latest_entry
 
-FIND.search_length dq ?
-FIND.search_buffer dq ?
+;; The STATE variable is 0 when the interpreter is executing, and non-zero when
+;; it is compiling.
+var_STATE dq 0
+
 FIND.rsi dq ?
 
 READ_WORD.rsi dq ?
-READ_WORD.rax dq ?
-READ_WORD.max_size = $FF
-READ_WORD.buffer rb READ_WORD.max_size
-READ_WORD.length db ?
-READ_WORD.char_buffer db ?
+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
@@ -419,8 +609,25 @@ DOTU.rbuffer rq 16
 DOTU.length dq ?
 DOTU.printed_length dq ?
 
-PARSE_NUMBER.length dq ?
+;; Reserve space for compiled words, accessed through HERE.
+here dq here_top
+here_top rq $4000
+
+;; Pointer to input buffer and its length. Used as local variable in
+;; INTERPRET-STRING (see bootstrap.asm). [TODO] The code organization is a bit
+;; awkward here.
+input_buffer dq ?
+input_buffer_length dq ?
 
 ;; Return stack
 rq $2000
 return_stack_top:
+
+segment readable
+
+;; 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'
+sysf.len = $ - sysf
+