This page represents examples 69, 70, and 71 of Halstead's "Machine-Independent Computer Programming".
Comments in italics, as well as the disassembly and Algol hand translation, have been added by
MikeVanEmmerik.
Again, I find the above difficult to read. Here is my translation to Algol,
which Neliac is supposed to be a dialect of, using the procedure of Halstead's
book, page 22:
EXAMPLE 69. Address ff jkb yyyyy Address ff jkb yyyyy 10100 00 000 00000 10142 07 000 00036 10101 00 000 00000 10143 03 000 00036 10102 00 000 00000 10144 23 030 10106 10103 00 000 00000 10145 14 030 10107 10104 00 000 00000 10146 61 010 10131 10105 00 000 00000 10147 10 030 10103 10106 00 000 00000 10150 26 030 10100 10107 00 000 00000 10151 22 030 10106 10110 00 000 00000 10152 14 030 10107 10111 00 000 00000 10153 10 030 10100 10112 00 000 00000 10154 27 630 10103 10113 00 000 00000 I0155 65 000 10113 10114 10 030 10103 10156 11 030 10100 10115 35 030 10100 10157 10 030 10106 10116 61 010 10113 10160 27 000 00001 10117 00 000 00000 10161 04 430 10103 10120 10 000 00005 10162 61 000 10165 10121 26 030 10100 10163 65 000 10117 10122 27 730 10106 10164 61 000 10166 10123 61 000 10126 10165 65 000 10131 10124 65 000 10113 10166 61 000 10167 10125 61 000 10130 10167 12 100 00000 10126 10 030 10103 10170 10 000 00170 10127 34 030 10100 10171 40 031 10103 10130 61 010 10117 10172 07 000 00073 10131 00 000 00000 10173 41 031 10100 10132 10 030 10106 10174 07 000 00072 10133 26 030 10103 10175 14 031 10107 10134 07 000 00036 10176 71 100 00003 10135 03 000 00036 10177 61 000 10170 10136 23 030 10106 10200 10 030 10103 10137 26 030 10107 10201 14 030 10106 10140 22 030 10100 10202 61 400 10000 10141 26 030 10103The above is obviously quite hard to read. Here is my attempt at disassembling it:
10100 AB: .word 0,0,0
10103 AA: .word 0,0,0
10106 AC: .word 0
10107 AD: .word 0,0,0
10112 .word 0
10113 SUB_B: .word 0 ; Space for return address
10114 ld q,AA
10115 sub_st q,AB ; Subtract and store
10116 jmp *SUBC ; Jump indirect on return address
10117 SUB_A: .word 0
10120 ld q,#5
10121 add q,AB
10122 sub <,q,AC ; Subtract and skip next instruction
; if result < 0
10123 jmp ENT_D
10124 call SUB_B
10125 jmp SUB_A ; Call SUB_A then return
10126 ENT_D: ld q,AA
10127 add_st q,AB ; Add and store
10130 ENT_C: jmp *SUB_A ; return from SUB_A
10131 SUB_C: .word 0
10132 ld q,AC
10133 add q,AA
; Note: registers on the Univac M460 are 30 bits
10134 shl aq,#30 ; shift left the a and q registers
10135 shr aq,#30 ; shift right aq
10136 div AC
10137 add q,AD
10140 mul AB
10141 add q,AA
10142 shl aq,#30
10143 shr aq,#30
10144 div AC
10145 st q,AD
10146 jmp *SUBC ; return
10147 START: ld q,AA ; note no return address storage
10150 add q,AB
10151 mul AC
10152 st q,AD
10153 ld q,AB
10154 sub >=,q,AA ; subtract, skip if result >= 0
10155 call SUB_B
10156 ld a,AB
10157 ld q,AC
10160 sub q,#1
10161 cmp aq,AA ; skip if a < OP <= q
10162 jmp ENT_E
10163 call SUB_A
10164 jmp ENT_A
10165 ENT_E: call SUB_C
10166 ENT_A: jmp ENT_B
10167 ENT_B: ld b,#0 ; don't know why j=1
10170 @1: ld q,#120
10171 ld aq,AA(b) ; "enter logical product"
10172 shl aq,#59
10173 add aq,AB(b)
10174 shl aq,#58
10175 st q,AD(b)
10176 cmp =,b,#2 ; compare b with 2 and skip if equal?
10177 jmp @1
10200 ld q,AA
10201 st q,AC
10202 halt
The program of Example 69 was then accepted by the Decompiler
of Appendix D, which found all of the absolute addresses used as
nouns or verbs, and assigned arbitrary names to them. The program
logic itself was then decoded into the equivalent Neliac statements,
as shown in Example 70.
Again, I find the above difficult to read. Here is my translation to Algol,
which Neliac is supposed to be a dialect of, using the procedure of Halstead's
book, page 22:
Comments This is MikeVanEmmerik's hand translation to Algol of Example 70
in "Machine-Independent Computer Programming".
integer procedure START();
integer A STORE, Q STORE, AA(3), AB(3), AC, AD(4);
begin
AD := (AA + AB) * AC;
if AB - AA >= 0 ;
else SUB B();
A STORE := AB;
Q STORE := AC-1;
if Q STORE >= AA and if A STORE < AA;
else goto ENT_E;
SUB A();
goto ENT_A;
ENT E:
SUB C();
ENT A:
goto ENT B;
ENT B:
i := 0;
for i := i step 1 until 2 do
begin AD[i] := AA[i](3..6) + AB[i](2..5) end
AC := AA;
end..
integer procedure SUB C();
begin
AD := (((AC + AA)/AC+AD)*AB+AA)/AC;
end
integer procedure SUB B();
begin
AB := AB - AA;
end
integer procedure SUB A();
begin
if 5 + AB - AC < 0;
else goto ENT D;
SUB B();
goto ENT_C;
ENT D:
AB := AB + AA;
ENT C:
end
As you can see, there are no parameters to worry about (most code seems to use global variables).
The semantics of the skip instructions are obvious, with the blank if clause always followed by an else clause.
I don't know the semantics for bit selection in Algol, if it has any.
No attempt has been made to structure the gotos in the program.
However, a for loop has been recovered, complete with indexes and bit ranges.
In addition to producing the Neliac version solely from the
machine-coded version of the program, the decompiler also pro-
duced a listing of the various names which it assigned, as shown
in Example 71.
EXAMPLE 71.
SUB A 0 10117
ENT A 0 10166
ENT B 0 10167
SUB B 0 10113
ENT C 0 10130
ENT D 0 10126
ENT E 0 10165
SUB C 0 10131
START 0 10147
AA 3 10103
AB 3 10100
AC 3 10106
AD 3 10107
where an initial zero denotes a verb and a 3 indicates a noun.
It may be noted that in a case of this kind there is no inherent
meaning in the names assigned to nouns or verbs, except perhaps
to the first and last of the verbs. If the original documentation has
been adequate, so that the comments accompanying the original
program has been sufficiently informative to allow a person to provide
meaningful names for given addresses, the decompiler will
accept and use them.
CategoryDecompilationRevision: r1.1 - 23 Aug 2004 - 04:00 - MikeVanEmmerik