word works with negative numbers now.

symbols moved to head of machine code.

thun/asm-dump.txt

@@ -1,97 +1,111 @@
 [word(0),
-do_offset(_7910),
+do_offset(_2170),
-allocate(_7920,20),
+allocate(_2180,20),
-label(_7910),
+label(_2170),
 mov_imm(0,0),
 store_word(0,0,0),
 mov_imm(0,4096),
-mov_imm(1,_7982),
+mov_imm(1,_2242),
 mov_imm(2,0),
 mov_imm(3,0),
 store_word(2,0,0),
-label(_8030),
+labelv_imm(3,0),
+store_word(2,0,0),
+label(_2290),
 sub_imm(1,1,0),
-eq_offset(_8042),
+eq_offset(_2302),
 load_word(4,1,0),
 lsl_imm(5,4,2),
 asr_imm(5,5,17),
-eq_offset(_8158),
+eq_offset(_24346,4,17),
+asr_imm(6,6,17),
+eq_offset(),
 add(5,5,1),
-label(_8158),
+label(_2434),
 lsl_imm(6,4,17),
 asr_imm(6,6,17),
-eq_offset(_8220),
+eq_offset(_2502),
 add(6,6,1),
-label(_8220),
+label(_2502),
 load_word(3,5,0),
 mov(1,6),
-asr_imm(6,3,30),
+asr_imm(6,3,t(6,16383),
+ior_imm(6,6,65535),
+and(30),
 and_imm(6,6,2),
 sub_imm(6,6,2),
-ne_offset(_8322),
+ne_offset(_2564),
 mov_imm_with_shift(6,16383),
 ior_imm(6,6,65535),
 and(6,6,3),
 do(6),
-label(_8322),
+label(_2564),
-sub_imm(0,0,4),
+sub_imm(0(_2810),
+store_word(2,0,0),
+do_offse,0,4),
 sub(2,5,0),
-hi_offset(_8424),
+hi_offset(_2748),
 and_imm(2,2,32767),
-label(_8424),
+label(_2748),
 lsl_imm(2,2,15),
 ior_imm(2,2,4),
-label(_8486),
+label(_2810),
 store_word(2,0,0),
-do_offset(_8030),
+do_offset,lsl_imm(2,2,17),
-label(_8042),
+asr_imm(2,2,17),
-do_offset(_8042),
+e(_2290),
-label(_8540),
+label(_2302),
+do_offset(_2302),
+label(_2864),
+symbol(_2864),
 lsl_imm(6,2,2),
 asr_imm(6,6,17),
-eq_offset(_8656),
+eq_offset(_3064),
 add(6,6,0),
-label(_8656),
+label(_3064),
-lsl_imm(2,2,17),
+9,7,17),
+asr_imm(9,9,17),
+eq_offset(lsl_imm(2,2,17),
 asr_imm(2,2,17),
-eq_offset(_8718),
+eq_offset(_3132),
 add(2,2,0),
-label(_8718),
+label(_3132),
 load_word(7,2,0),
 lsl_imm(8,7,2),
 asr_imm(8,8,17),
-eq_offset(_8794),
+eq_offset(_3214)q_offset(_3394),
-add(8,8,2),
+sub(6,6,0),
-label(_8794),
+and_imm,add(8,8,2),
+label(_3214),
 lsl_imm(9,7,17),
 asr_imm(9,9,17),
-eq_offset(_8856),
+eq_offset(_3282),
 add(9,9,2),
-label(_8856),
+label(_3282),
 sub_imm(0,0,4),
 sub_imm(8,8,0),
-eq_offset(_8900),
+eq_offset,9,32767),
+label(_3518),
+mov_imm_wit(_3326),
 sub(8,8,0),
 and_imm(8,8,32767),
-label(_8900),
+label(_3326),
 sub_imm(6,6,0),
-eq_offset(_8968),
+eq_offset(_3394),
 sub(6,6,0),
 and_imm(6,6,32767),
-label(_8968),
+label(_3394),
 lsl_imm(8,8,15),
 ior(8,8,6),
 store_word(8,0,0),
 sub_imm(0,0,4),
 sub_imm(9,9,0),
-eq_offset(_9114),
+eq_offset(_3518),
 sub(9,9,0),
 and_imm(9,9,32767),
-label(_9114),
+label(_3518),
 mov_imm_with_shift(2,2),
 ior(2,2,9),
-do_offset(_8486),
+do_offset(_2810),
-label(_7982),
+label(_2242),
-expr_cell(_9232,0),
+expr_cell(_2864,0)]
-label(_9232),
-symbol(_8540)]

thun/compiler.markII.pl

@@ -85,7 +85,7 @@ Mark II
 % ======================================
 ],⟐([
 
-    label(Cons),  % Let's cons.
+    definition(Cons),  % Let's cons.
 
     unpack_pair(TOS, TEMP0, TOS, SP),
     % TEMP0 = Address of the list to which to append.
@@ -107,8 +107,7 @@ Mark II
     jump(Done)  % Rely on mainloop::Done to write TOS to RAM.
 ]),[
     label(Expression),
-    expr_cell(ConsSym, 0),
+    expr_cell(Cons, 0)
-    label(ConsSym), symbol(Cons)
 ].
 
 
@@ -172,6 +171,8 @@ language.
     [mov_imm_with_shift(In, 2),  % In := 4 << 15
      ior(In, In, Term)].
 
+⟐(definition(Name)) --> [label(Name), symbol(Name)].
+
 
 do :-
     compile_program(Binary),
@@ -218,9 +219,13 @@ asm([(N, Instruction)|Rest]) --> !, asm(N, Instruction), asm(Rest).
 asm(Here, expr_cell(Func, NextCell)) --> !,
     {Data is ((Func - Here) << 15) \/ NextCell}, asm(Here, word(Data)).
 
-asm(_, symbol(Sym)) --> !, {Data is Sym \/ 0x80000000}, asm(_, word(Data)).
+asm(_, symbol(Sym)) --> !, {Data is (Sym + 4) \/ 0x80000000}, asm(_, word(Data)).
+% The symbol is at the beginning of the function machine code, so the pointer it
+% holds to that code has to be one word beyond the pointer/label Sym that points
+% to the symbol itself (one word before the machine code.)  The symbol's address
+% is used in expressions.
 
-asm(_, word(Word)) --> !, {binary_number(Bits, Word)}, collect(32, Bits).
+asm(_, word(Word)) --> !, encode_int(32, Word).
 
 asm(_,  load_word(A, B, Offset)) --> !, instruction_format_F2(0, 0, A, B, Offset).
 asm(_,  load_byte(A, B, Offset)) --> !, instruction_format_F2(0, 1, A, B, Offset).

thun/markII.rst

@@ -295,6 +295,23 @@ useful for appending lists to other lists.
 
 Yeah, I think that's the way to go...
 
+---------------------------------------------
+
+There's soething fishey with the symbols now that they are in the header
+of the machine code.  Each symbol's pointer field points to the next cell,
+which seems really redundant.  We need some symbol record to
+differentiate from lists and ints.  Maybe if more information was in the
+header? Like the name of the function?  It would make more sense?
+
+Right now I'm assuming that the eventual parser would be looking up
+symbols at parse-time and reusing the header symbols rather than
+allocating cells for new ones.  If the symbols were kept apart from the
+machine code then it makes sense for them to have pointers?
+
+Maybe I can dispense with symbol records by modifying the
+is-this-a-symbol code to just check if the address is below the end of
+the library code.
+
 
 
 PC == 0