October 4, 1963
Revised October 17, 1963
Revised October 23, 1963

FP6000 Project Memorandum #29

Notes on the Order Code

Revisions have been underlined.

W. R. Whittall
Technical Services
Data Systems Department.

DISTRIBUTION LIST

L. R. WoodEngineeringTechnical Services
B. Z. de FerrantiG. CollinsL. Albert
E. StrainJ. D. Buck
B. OvenellJ. Chapman
H. RedderingL. A. Cragg
A. JacksonB. CeilierA. G. Henry
G. CalverleyK. BowerL. F. Lawry
J. DavidsonF. M. Longstaff
R. F. Johnston
M. J. Marcotty
R. Moore
E. McDorman
J. W. McSherry
D. E. Nuttall
D. K. Ritchie
R. Sadana
I. P. Sharp
A. Sharp
D. K. Smith
V. Taylor
W. R. Whittall
File
B. Daly
P. Boyd

NOTES ON THE FP6000 ORDER CODE

1. FUNCTIONS 004, 014, 104 )
) (additions)
005, 015, 105 )

These functions first perform the specified addition on the 24 bit
signed operands, set carry if the result is negative and then clear the sign
bit of the result before storing.

2. FUNCTIONS 005, 016, 106 )
) (subtractions)
07, 017, 107 )

These functions first perform the specified subtraction on the 24 bit
operands, set carry if there is an overall borrow (treating the sign bits as
having positive weight) and then clear the sign bit of the result before storing.

NOTE. Function 027 is identical to function 007 except that the value of
x is left unchanged.

3. MULTIPLICATION 040, 041, 042

(a) Overflow can only be set when n = x = -1 and result is then -1/

(b) In the 042 instruction x* may be positive or negative.

4. MULTIPLY by 10 043, 047

The sign bits of x and x* are ignored and left clear except
when a non-numeric character is used with the 043.

5. DIVISION 044, 045, 043

(a) If n = 0 the division is not performed but overflow is set.
x: is left undisturbed.

(b) The sign bit of x* is completely ignored.

(c) It is possible for rounding in the 045 instruction to cause overflow,
or to recover overflow set by the division process.

(d) The remainder always satisfies the equation r = x' = x - nx*'

6. SUBROUTINE ENTRY 070

It is the "programmers I.N.R. " which is stored in X. i. e. in normal
mode datum is subtracted from the contents of the "machine I.N.R. before
storage. (The I.N.R. consists of V, C, 7 mode bits plus 15 address bits).
V and C are left clear but the mode bits are unchanged.

7.SUBROUTINE EXIT 072

N is datumed. The whole of the I.N.R. is restored but V is left set
if already set. C is cleared.

8. SHIFT INSTRUCTIONS 110, 111, 112, 113

The number of places shifted is equal to Ns if Ns is ≤ 127 or
equal to Ns (mod. 64) + 65 if Ns > 127.

{RDM:The preceding remark on shift counts is only useful in calculating
instruction execution times as the maximum meaningful shift was either 24 or 48 bits.}

9. SINGLE LENGTH SPECIAL RIGHT SHIFT 112

Overflow is always left clear even if rounding “overflows“. There
is only one value of x for which this can arise, i. e. if V is set and X contains
all ones the result of a single shift is -1.0 with V clear.

10. NORMALIZE 114, 115

(i) N = 0 or N(mod 4096) ≥ 2048
x: is set equal to zero and V is cleared.

(ii) Not case (i), V clear
x: is shifted left (arithmetically) and the ten least significant
bits of N are counted down once per shift until either:-
1. The sign bit of x: is different from the 2-1 bit.
2. N = 0 (ten least significant bits).
3. The six least significant bits of N become zero for the second
time. i. e. x: must have been equal to zero.

In case 1 the nine least significant bits of the new N replace the
nine least significant bits of x* and V is set if the tenth bit of N is not zero.
In cases 2 and 3, x: is set equal to zero and V cannot be set.

(ill) Not case (i) V Set
First x: is shifted one place right (special shift). Then the nine
least significant bits of N replace the nine least significant bits of x* and one
is added to x*. V will be cleared unless there is a carry from the ninth
bit of x* into the tenth (from the least significant end) or unless the tenth
bit of N is not zero.

NOTE

(a) In the case of the 114 instruction x* is first set equal to zero
and the shifts are single length without rounding.

(b) In all cases the sign bit of x* is ignored. The result in x*
is always positive except when there is a carry into the sign
bit as a result of adding one to x* after a right shift. V will
always be set if this happens.

11. NO OPERATION 123

This instruction does not clear the carry flip-flop.

W. R. Whittall