10 Oct
2006
10 Oct
'06
5:59 p.m.
Of course it's not decimal mode the way, say, a 6502 would do it but it's
interesting to see a major architecture bring back non-binary-based operations
(in this case base 10 calculations to avoid precision difficulties).
http://news.com.com/IBMs+Power6+gets+help+with+math%2C+multimedia/2100-1006…
--
--------------------------------- personal: http://www.armory.com/~spectre/ ---
Cameron Kaiser * Floodgap Systems * www.floodgap.com * ckaiser at floodgap.com
-- The best things in life are sold out. --------------------------------------
10 Oct
10 Oct
9:27 p.m.
On 10/10/2006 at 5:59 PM Cameron Kaiser wrote:
Of course it's not decimal mode the way, say, a
6502 would do it but it's
interesting to see a major architecture bring back non-binary-based
operations (in this case base 10 calculations to avoid precision
difficulties).
I don't understand what the hubbub is about. Can anyone explain for me?
Cheers,
Chuck
11 Oct
11 Oct
10:37 a.m.
On 10/10/2006 at 5:59 PM Cameron Kaiser wrote:
On Tue, 10 Oct 2006, Chuck Guzis wrote:
Of course it's not decimal mode the way, say,
a 6502 would do it but it's
interesting to see a major architecture bring back non-binary-based
operations (in this case base 10 calculations to avoid precision
difficulties).
I don't understand what the hubbub is about. Can
anyone explain for me?
Has somebody invented Binary Coded Decimal?
12:44 p.m.
As long as we're OT, here are two limericks I found on oedilf.com:
In the earliest Boolean days
All logic was built hardware-ways.
The math, now outmoded,
Was BCD-coded,
Most often when used with displays.
When base 2 and base 5 make a pair,
A bi-quinary number is there.
It's what abaci use,
And can roundly confuse
Round-eyed Yanks at a Chinatown fair.
Biquinary numbers derive
From two systems--base 2 and base 5.
Every
decimal n
Is expressible then
As a sum (if you're hip to the jive).
2:03 p.m.
On 10/11/2006 at 1:58 PM woodelf wrote:
What is XS3?
Sometimes called "excess three". Basically, a system of doing decimal
arithmetic on a binary computer by biasing decimal numbers by 3. Thus
decimal 0 = binary 0011; decimal 9 = 1100.
There are many other "BCD" codes, such as 5311, 2421. All of these (XS3,
5311, 2421) have the property that the binary one's complement of a number
is also the 9's complement--a property not enjoyed by 8421 decimal.
CDC 6000 COBOL used a type of XS3 for its decimal arithmetic--a rather
happy coincidence being that octal 33 was the display code for zero and 44
was the code for nine. It was possible to do addition and subtraction on
two 10-digit decimal numbers without resorting to branching--using only
shift, AND and ADD operations.
Cheers,
Chuck
3:24 p.m.
woodelf wrote:
I like the idea of a trinary computer myself
but all we have is binary memory. :(
Memory's cheap; stack it two bits wide and encode/decode between your memory
and data bus...
;-)
--
If you've ever wondered how you get triangles from a cow
You need buttermilk and cheese, and an equilateral chainsaw
12:08 p.m.
Fred Cisin wrote:
Has somebody invented Binary Coded Decimal?
Not
really. IBM has patented a way of packing 2 decimal
digits into 7 binary bits.Of course more hardware is needed
but that is patented too.
I think the thing that killed BCD computers is thae lack of
BCD memory addressing.
12:31 p.m.
On 10/11/2006 at 1:08 PM woodelf wrote:
Not really. IBM has patented a way of packing 2
decimal
digits into 7 binary bits.Of course more hardware is needed
but that is patented too.
Ah, so 00-63 hex is now proprietary. Better check my old code for
violations...
I think the thing that killed BCD computers is thae
lack of
BCD memory addressing.
You never used a 1620?
Cheers,
Chuck
1:11 p.m.
On 10/11/2006 at 1:57 PM woodelf wrote:
No it is the way you pack the data.
I can see how manipulating the data (and the circuitry involved therein)
might be a basis for a patent, but how can one legitimately patent a
representation of a number? How many of the required 100 states out of 128
possible would have to overlap before I'm sued for infringement? Am I
infringing more if I use Gray code, XS3, biquinary, or LFSR states to
represent my 100 states?
Cheers,
Chuck
1:30 p.m.
Chuck Guzis wrote:
On 10/11/2006 at 1:57 PM woodelf wrote:
Check here. This was not the orginal link I thought of.
All I know is with IBM they seem to have invented this
packing algorithom most likely patented for hardware.
http://www2.hursley.ibm.com/decimal/
No it is the way you pack the data.
I can see how manipulating the data (and the circuitry involved therein)
might be a basis for a patent, but how can one legitimately patent a
representation of a number? How many of the required 100 states out of 128
possible would have to overlap before I'm sued for infringement? Am I
infringing more if I use Gray code, XS3, biquinary, or LFSR states to
represent my 100 states?
Cheers,
Chuck
12:38 p.m.
On 10/11/2006 at 1:08 PM woodelf wrote:
Not really. IBM has patented a way of packing 2
decimal
digits into 7 binary bits.Of course more hardware is needed
but that is patented too.
You know, that's not particularly memory-efficient. Packing 6 digits into
20 bits is better.
Cheers,
Chuck
12:20 p.m.
I might be out of my mind (and mistaken too!), but I'm pretty sure that
POWER systems have always been BCD-enabled. At least since POWER2.
Doc
On 10/10/2006
at 5:59 PM Cameron Kaiser wrote:
On Tue, 10 Oct 2006, Chuck Guzis wrote:
Of course it's not decimal mode the way, say,
a 6502 would do it but
it's
interesting to see a major architecture bring back
non-binary-based
operations (in this case base 10 calculations to avoid precision
difficulties).
I don't understand what the hubbub is about.
Can anyone explain for
me?
Has somebody invented Binary Coded Decimal?
12:36 p.m.
doc at mdrconsult.com wrote:
I might be out of my mind (and mistaken too!), but
I'm pretty sure that
POWER systems have always been BCD-enabled. At least since POWER2.
I'm reasonably certain that you're right. Even with regards to the POWER1.
Peace... Sridhar
6647
days inactive
6647
days old
16 comments
8 participants
participants (8)
-
bfranchuk＠jetnet.ab.ca -
cclist＠sydex.com -
cisin＠xenosoft.com -
doc＠mdrconsult.com -
julesrichardsonuk＠yahoo.co.uk -
ploopster＠gmail.com -
spectre＠floodgap.com -
THX1138＠dakotacom.net