X-Git-Url: https://git.rrq.au/?a=blobdiff_plain;f=main.asm;h=d0c43917ae4971d8961f3d6b02d9030b95c90753;hb=1013709c277143b58deafcf9a1faccd3e9b799a2;hp=834872cd54a756fef046f7a67c0dd1463fd8f86b;hpb=be8f6fc600856043dede3a5da6aa9aded19fcb34;p=rrq%2Fjonasforth.git

diff --git a/main.asm b/main.asm
index 834872c..d0c4391 100644
--- a/main.asm
+++ b/main.asm
@@ -56,30 +56,34 @@ macro forth_asm label, name {
 ;; 'dq' statements.)
 macro forth label, name {
   header label, name
-  dq docol
+  dq DOCOL
 }
 
 segment readable executable
 
+entry main
+
+include "impl.asm"
+
 main:
   cld                        ; Clear direction flag so LODSQ does the right thing.
   mov rbp, return_stack_top  ; Initialize return stack
 
-  mov rsi, program
-  next
+  mov rax, MAIN
+  jmp qword [rax]
 
 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
@@ -95,42 +99,31 @@ 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]
   next
 
+;; Given an entry in the dictionary, return a pointer to the codeword of that
+;; entry.
+forth_asm TCFA, '>CFA'
+  pop rax
+  add rax, 8                    ; [rax] = length of name
+  movzx rbx, byte [rax]
+  inc rax
+  add rax, rbx                  ; [rax] = codeword
+  push rax
+  next
+
 ;; BRANCH is the fundamental mechanism for branching. BRANCH reads the next word
 ;; as a signed integer literal and jumps by that offset.
 forth_asm BRANCH, 'BRANCH'
@@ -149,6 +142,16 @@ forth_asm ZBRANCH, '0BRANCH'
   add rsi, 8     ; We need to skip over the next word, which contains the offset.
   next
 
+;; Duplicate the top of the stack.
+forth_asm DUP_, 'DUP'
+  push qword [rsp]
+  next
+
+;; Execute the codeword at the given address.
+forth_asm EXEC, 'EXEC'
+  pop rax
+  jmp qword [rax]
+
 ;; Expects a character on the stack and prints it to standard output.
 forth_asm EMIT, 'EMIT'
   pushr rsi
@@ -179,98 +182,25 @@ forth SPACE, 'SPACE'
 ;; 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
-
-.read_alpha:
-  mov al, [.char_buffer]
-  movzx rbx, [.length]
-  mov rsi, .buffer
-  add rsi, rbx
-  mov [rsi], al
-  inc [.length]
-
-  mov rax, 0
-  mov rdi, 0
-  mov rsi, .char_buffer
-  mov rdx, 1
-  syscall
-
-  cmp [.char_buffer], ' '
-  je .end
-  cmp [.char_buffer], $A
-  jne .read_alpha
-
-.end:
-  push .buffer
-  movzx rax, [.length]
-  push rax
+  call read_word
+  push rdi                      ; Buffer
+  push rdx                      ; Length
 
   mov rsi, [.rsi]
-  mov rax, [.rax]
-
   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:
-
-  ;; 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
+  pop [parse_number.length]     ; Length
+  pop [parse_number.buffer]     ; String pointer
 
-  dec rcx
-  jnz .loop
-
-  push r8
+  push rsi
+  call parse_number
+  pop rsi
 
+  push rax                      ; Result
   next
 
 forth READ_NUMBER, 'READ-NUMBER'
@@ -310,6 +240,55 @@ forth HELLO, 'HELLO'
   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
+
+;; 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
+
 ;; .U prints the value on the stack as an unsigned integer in hexadecimal.
 forth_asm DOTU, '.U'
   mov [.length], 0
@@ -371,39 +350,150 @@ forth_asm DOTU, '.U'
   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
+
+;; 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
+
+;; 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
+;; 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.
+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
+
+;; 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], rcx                ; 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 MAIN, 'MAIN'
   dq HELLO
-  dq READ_WORD, FIND, DOTU, NEWLINE
-  dq BRANCH, -8 * 5
+  dq INTERPRET
+  dq BRANCH, -8 * 2
   dq TERMINATE
 
 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
 
-SPACE_string db 'SPACE'
-.length = $ - SPACE_string
-HELLO_string db 'HELLO'
-.length = $ - HELLO_string
-DOTU_string db '.U'
-.length = $ - DOTU_string
-HELLA_string db 'HELLA'
-.length = $ - HELLA_string
+;; The STATE variable is 0 when the interpreter is executing, and non-zero when
+;; it is compiling.
+var_STATE dq 0
 
-
-you_typed_string db 'You typed: '
-.length = $ - you_typed_string
-
-FIND.search_length dq ?
-FIND.search_buffer dq ?
 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
@@ -411,7 +501,9 @@ 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 $2000
 
 ;; Return stack
 rq $2000