IBM 7090 - 5.3.04 FLOATING-POINT ARITHMETIC INSTRUCTIONS

190 pages

To Next Page

To Next Page

To Previous Page

To Previous Page

Loading...

In

the

7090,

multiplication

follows

this

procedure

closely.

Starting

at

the

right

of

the

multiplier

in

the

MQ,

only

one

position

of

the

multiplier

is

scanned

at

a

time.

In

the

7090

there

is

no

means

of

remembering

partial

products

so

they

may

be

added

atter

each

position

of

the

multiplier

has

been

used.

Therefore,

if

the

multiplier

position

be-

ing

scanned

is

a

one,

the

multiplicand

is

added

to

any

previous

partial

product

that

is

in

the

AC. In

the

7090,

multiplication

is

done

as

follows:

0110

Multiplicand

(SR)

1011

Multiplier

(MQ)

0000

Start

with

zeros

in

AC

+0110

Right

MQ

position

scanned

0110

Partial

Product

+0110

Next

MQ

position

scanned

10010

Partial

Product

0000

Next

MQ

positions

scanned

010010

Partial

Product

+0110

Next

MQ

position

scanned

1000010

Product

(AC

and

MQ)

In

this

example,

each

time

the

multiplicand

is

added

to

the

previous

partial

product,

the

multiplicand

must

be

moved

to

the

left

to

correctly

line

up

the

bit

positions.

In

the

7090,

the

partial

product

is

moved

to

the

right

rather

than

the

multiplicand

moving

to

the

left.

The

end

result

is

the

same.

Multiplication

in

the

7090

breaks

to

two

major

procedures:

1.

Addition

of SR

and

AC

contents.

2.

Shifting

the

contents

of

the

AC

and

MQ

right.

Figure

5.3-17

shows

a

MPY

operation

in

the

7090.

During

the

E

cycle,

in

addition

to

bringing

the

multiplicand

to

the

SR,

the

AC

is

cleared

and

the

sign

is

set.

This

allows

the

MPY

to

start

with

the

AC

containing

zeros.

Also

during

the

E

cycle

the

word

coming

from

storage

is

tested

to

see

if

all

positions

contain

a

zero.

If

the

multiplicand

is

zero

the

product

is

to

be

zero.

To

save

machine

time,

the

MQ

is

cleared

and

the

instruction

ends

operation;

438 (3510)

is

put

into

the

SC.

The

SC

will

be

used

to

indi-

cate

when

all

bits

of

the

MQ

have

been

tested.

Also,

because

the

storage

word

was

not

zero,

the

computer

is

put

into

L

time

for

multiplication.

Now MQ(35)

is

tested

for

a

bit.

If

there

is

a

bit

in

MQ(35),

the

SR

and

AC

contents

are

added

and

the

result

is

put

into

the

AC.

If

MQ(35)

had

no

bit,

this

addition

would

not

have

been

done

because

the

sum

of

zero

and

a

number

equals

the

same

number.

The

AC

and

MQ

are

shifted

right

to

put

the

next

position

of

the

multiplier

in

MQ(35)

to

be

tested,

and

put

the

partial

product

in

correct

alignment

with

the

multiplicand.

This

type

of

shift

is

called

a

slow-speed

shift.

High-speed

shifting

(shifting

once

for

each

clock

time)

will

occur

if

MQ(34)

contains

zero.

This

is

done

to

speed

the

multiplica-

tion

operation.

Each

time

the

AC

and

MQ

are

shifted,

the

shift

counter

is

stepped

and,

as

long

as

the

shift

counter

does

not

equal

zero,

MQ(34)

is

once

more

tested

and

the

process

starts

again.

When

the

SC

has

been

stepped

down

to

zero,

the

computer

signals

that

multiplication

is

complete.

Because

all

positions

of

the

multiplier

have

been

scanned

and

all

additions

have

been

completed,

MPY

then

ends

operation.

62

Table of Contents