comp-1 and comp-2 values

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

Hi

The PIC clause for comp-1 and comp-2 are :

COMP-1 : SV9(8)ES99

COMP-2: SV9(17)ES99

I want to know how 'S' is represented in the exponential part ( ES99) . Under what scenario will it show negative ( - ) ? Or will it be +ve always (i.e.blanks).

I mean will it ever show something like ( -.12300000E-03) , for an input say -123 moved to comp-1 and comp-2 ?

dbzTHEdinosauer · Posted: Tue Jul 20, 2010 1:20 pm

Internal representation

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

I had seen this link earlier also,but couldn't figure out how the comp-1 internal representation came to 43 4D 20 00 for +1234.

could you please help me with it ?

dbzTHEdinosauer · Posted: Tue Jul 20, 2010 4:34 pm

well, you need to start at basics.

floating-point

COMP-1 and COMP-2 are internal-floating point. They do not have picture clauses

discussions on external floating point are referenced in the above link.

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

Hi

Thanks for the links . I did go through them and it helped me.

I could understand the internal representation of value +1234 in binary is 04 D2 . But i couldn't figure out how the represenation for -1234 would be determined ?

Could you please help.

Robert Sample · Posted: Wed Jul 21, 2010 3:56 pm

Hex C3 4D 20 00 -- the sign bit is turned on to denote the negative value. The 43 4D 20 00 remains exactly the same.

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

This is were i am confused.

For Binary , value of +1234 has internal representation as 04 D2 , coz the binary value would be 0000 0100 1101 0010.

But for value -1234 how is it FB 2E.

Secondly for same value in Comp-1 how is it represented as 43 4D 20 00 ?

dick scherrer · Posted: Thu Jul 22, 2010 9:14 am

Hello,

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

ya i understood now how -1234 is FB2E , but couldn't understand how the Comp-1 representation is 43 4D 20 00

Robert Sample · Posted: Thu Jul 22, 2010 5:00 pm

Since you obviously have a reading problem (this is explained in the COBOL manuals as you have been repeatedly told), I'll elaborate:

The first bit is the sign bit -- 0 means positive, 1 means negative.

The next seven bits are the exponent. For technical reasons, IBM stores the exponent in excess-64 notation (meaning they add 64 or hex 40 to the actual exponent). Hence 43 minus 40 is 3 (even in hex), so the exponent is 3.

The remaining 24 bits represent the mantissa of the floating point number. So the 4D2000 represents the 1234. The decimal point is implied after the first digit -- as IBM explicitly states in the manual.

So positive 1.234 times 10 raised to the third power is ... 1234.

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

Thanks Robert for the explanantion.

You said this is explained in the COBOL manuals ,

Could you please let me know which COBOL manual should be referred.

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

Hi Robert ,

In your above reply you have mentioned For technical reasons, IBM stores the exponent in excess-64 notation (meaning they add 64 or hex 40 to the actual exponent). Hence 43 minus 40 is 3 (even in hex), so the exponent is 3.

Could you please explain how 43 is obtained ?

Robert Sample · Posted: Tue Sep 28, 2010 7:25 pm

Start with the value. Shift the decimal point left or right until the decimal point is behind the first digit. The number of shifts is the exponent value. If you shifted left, the value is positive. If you shifted right, the value is negative.

1234 requires 3 shifts to be 1.234, hence the 3 -- and since the shifts were to the left, it is a positive three.

.0012345 requires 3 shifts to be 1.234, so again a 3 -- but this time the shifts were to the right, so the value is negative 3.

IBM adds the 40 (hex) automatically -- so +3 added to 40 becomes 43. 40 hex minus 3 is 3D so .001234 expressed as a COMP-1 value is 3D 4D 20 00

Ambili S · Active User Joined: 06 Sep 2007 Posts: 112 Location: India

That was really helpful Robert.Thanks a lot