When evaluating the expression Z = X * Y where X and Y is SINT32 you
can do it like this:
X0 = bitand(bitshift(x,-24),255);
X1 = bitand(bitshift(x,-16),255);
X2 = bitand(bitshift(x,-8),255);
X3 = bitand(y,255);
Y0 = bitand(bitshift(y,-24),255);
Y1 = bitand(bitshift(y,-16),255);
Y2 = bitand(bitshift(y,-8),255);
Y3 = bitand(y,255);
S0 = 2^24;
S1 = 2^16;
S2 = 2^8;
S3 = 2^0;
Z = (2^48)*X0*Y0 + (2^32)*X1*Y1 + (2^16)*X2*Y2 + X3*Y3 + (2^40)*X0*Y1
+ (2^40)*X1*Y0 + (2^32)*X0*Y2 + (2^32)*X2*Y0 + (2^24)*X0*Y3 +
(2^24)*X3*Y0 + (2^24)*X1*Y2 + (2^24)*X2*Y1 + (2^16)*X1*Y3 +
(2^16)*X3*Y1 + (2^8)*X2*Y3 + (2^8)*X3*Y2
Does this approach/method have a name ??
I would like to google for it and learn more, but I need to know what
I am looking for.
Thank you.
|
|
0
|
|
|
|
Reply
|
 John
|
2/7/2011 8:52:04 PM |
|
On 2/7/2011 12:52 PM, John McDermick wrote:
> When evaluating the expression Z = X * Y where X and Y is SINT32 you
> can do it like this:
>
> X0 = bitand(bitshift(x,-24),255);
> X1 = bitand(bitshift(x,-16),255);
> X2 = bitand(bitshift(x,-8),255);
> X3 = bitand(y,255);
>
> Y0 = bitand(bitshift(y,-24),255);
> Y1 = bitand(bitshift(y,-16),255);
> Y2 = bitand(bitshift(y,-8),255);
> Y3 = bitand(y,255);
>
> S0 = 2^24;
> S1 = 2^16;
> S2 = 2^8;
> S3 = 2^0;
>
> Z = (2^48)*X0*Y0 + (2^32)*X1*Y1 + (2^16)*X2*Y2 + X3*Y3 + (2^40)*X0*Y1
> + (2^40)*X1*Y0 + (2^32)*X0*Y2 + (2^32)*X2*Y0 + (2^24)*X0*Y3 +
> (2^24)*X3*Y0 + (2^24)*X1*Y2 + (2^24)*X2*Y1 + (2^16)*X1*Y3 +
> (2^16)*X3*Y1 + (2^8)*X2*Y3 + (2^8)*X3*Y2
>
>
>
> Does this approach/method have a name ??
>
> I would like to google for it and learn more, but I need to know what
> I am looking for.
>
> Thank you.
Crossproducts.
--
Rob Gaddi, Highland Technology
Email address is currently out of order
|
|
|
0
|
|
|
|
Reply
|
Rob
|
2/7/2011 8:55:03 PM
|
|
Rob Gaddi <rgaddi@technologyhighland.com> wrote:
>On 2/7/2011 12:52 PM, John McDermick wrote:
>> Does this approach/method have a name ??
>Crossproducts.
I would say "partial products".
S.
|
|
|
0
|
|
|
|
Reply
|
spope33
|
2/7/2011 10:22:42 PM
|
|
John McDermick <johnthedspguy@gmail.com> wrote:
> When evaluating the expression Z = X * Y where X and Y is SINT32 you
> can do it like this:
> X0 = bitand(bitshift(x,-24),255);
> X1 = bitand(bitshift(x,-16),255);
> X2 = bitand(bitshift(x,-8),255);
> X3 = bitand(y,255);
> Y0 = bitand(bitshift(y,-24),255);
> Y1 = bitand(bitshift(y,-16),255);
> Y2 = bitand(bitshift(y,-8),255);
> Y3 = bitand(y,255);
> S0 = 2^24;
> S1 = 2^16;
> S2 = 2^8;
> S3 = 2^0;
> Z = (2^48)*X0*Y0 + (2^32)*X1*Y1 + (2^16)*X2*Y2 + X3*Y3 + (2^40)*X0*Y1
> + (2^40)*X1*Y0 + (2^32)*X0*Y2 + (2^32)*X2*Y0 + (2^24)*X0*Y3 +
> (2^24)*X3*Y0 + (2^24)*X1*Y2 + (2^24)*X2*Y1 + (2^16)*X1*Y3 +
> (2^16)*X3*Y1 + (2^8)*X2*Y3 + (2^8)*X3*Y2
> Does this approach/method have a name ??
Note that this is fine for unsigned data, but you need a correction
for signed data. Google for "booth multiplication" or "booth recoding"
for some of the details on doing signed binary multiplication.
Otherwise, it is the way the partial products would be combined
in base 256 multiplication.
-- glen
|
|
|
0
|
|
|
|
Reply
|
glen
|
2/7/2011 11:11:16 PM
|
|
On Mon, 7 Feb 2011 12:52:04 -0800 (PST), John McDermick
<johnthedspguy@gmail.com> wrote:
>When evaluating the expression Z = X * Y where X and Y is SINT32 you
>can do it like this:
>
>X0 = bitand(bitshift(x,-24),255);
>X1 = bitand(bitshift(x,-16),255);
>X2 = bitand(bitshift(x,-8),255);
>X3 = bitand(y,255);
>
>Y0 = bitand(bitshift(y,-24),255);
>Y1 = bitand(bitshift(y,-16),255);
>Y2 = bitand(bitshift(y,-8),255);
>Y3 = bitand(y,255);
>
>S0 = 2^24;
>S1 = 2^16;
>S2 = 2^8;
>S3 = 2^0;
>
>Z = (2^48)*X0*Y0 + (2^32)*X1*Y1 + (2^16)*X2*Y2 + X3*Y3 + (2^40)*X0*Y1
>+ (2^40)*X1*Y0 + (2^32)*X0*Y2 + (2^32)*X2*Y0 + (2^24)*X0*Y3 +
>(2^24)*X3*Y0 + (2^24)*X1*Y2 + (2^24)*X2*Y1 + (2^16)*X1*Y3 +
>(2^16)*X3*Y1 + (2^8)*X2*Y3 + (2^8)*X3*Y2
>
>
>
>Does this approach/method have a name ??
>
>I would like to google for it and learn more, but I need to know what
>I am looking for.
>
>Thank you.
That is the definition of unsigned multiplication in base 256 ie. how
partial products are generated and combined if you only have 8 bit
multiplier (and much wider adders of course).
You can also look at it in a slightly different way:
X = 256^3 * X0 + 256^2 * X1 + 256 * X2 + X3
Y = 256^3 * Y0 + 256^2 * Y1 + 256 * Y2 + Y3
so X*Y = 256^6 * X0 * Y0 ...
(btw you have a typo in X3)
--
Muzaffer Kal
DSPIA INC.
ASIC/FPGA Design Services
http://www.dspia.com
|
|
|
0
|
|
|
|
Reply
|
Muzaffer
|
2/8/2011 5:50:26 AM
|
|
On 07/02/2011 21:52, John McDermick wrote:
> When evaluating the expression Z = X * Y where X and Y is SINT32 you
> can do it like this:
>
> X0 = bitand(bitshift(x,-24),255);
> X1 = bitand(bitshift(x,-16),255);
> X2 = bitand(bitshift(x,-8),255);
> X3 = bitand(y,255);
>
> Y0 = bitand(bitshift(y,-24),255);
> Y1 = bitand(bitshift(y,-16),255);
> Y2 = bitand(bitshift(y,-8),255);
> Y3 = bitand(y,255);
>
> S0 = 2^24;
> S1 = 2^16;
> S2 = 2^8;
> S3 = 2^0;
>
> Z = (2^48)*X0*Y0 + (2^32)*X1*Y1 + (2^16)*X2*Y2 + X3*Y3 + (2^40)*X0*Y1
> + (2^40)*X1*Y0 + (2^32)*X0*Y2 + (2^32)*X2*Y0 + (2^24)*X0*Y3 +
> (2^24)*X3*Y0 + (2^24)*X1*Y2 + (2^24)*X2*Y1 + (2^16)*X1*Y3 +
> (2^16)*X3*Y1 + (2^8)*X2*Y3 + (2^8)*X3*Y2
>
>
>
> Does this approach/method have a name ??
>
> I would like to google for it and learn more, but I need to know what
> I am looking for.
>
> Thank you.
At primary school we called this "long multiplication", though the
layout and ordering was slightly different.
And just like at school, it is easiest if you handle the sign separately
so that the multiplication is done with unsigned numbers.
|
|
|
0
|
|
|
|
Reply
|
David
|
2/8/2011 9:59:38 AM
|
|
|
5 Replies
221 Views
(page loaded in 0.099 seconds)
Similiar Articles: Multiplying two int32 numbers - comp.dspWhen evaluating the expression Z = X * Y where X and Y is SINT32 you can do it like this: X0 = bitand(bitshift(x,-24),255); X1 = bitand(bitshift(... How multiply two __int64 ... - comp.lang.cMultiplying two int32 numbers - comp.dsp How multiply two __int64 ... - comp.lang.c Multiplying two int32 numbers - comp.dsp How multiply two __int64 ... - comp.lang.c ... Rounded integer division - comp.lang.asm.x86Multiplying two int32 numbers - comp.dsp Rounded integer division - comp.lang.asm.x86 Multiplying two int32 numbers - comp.dsp Unsigned Integer Overflow on Multiplication ... Unsigned Integer Overflow on Multiplication and Division - comp ...Multiplying two int32 numbers - comp.dsp Unsigned Integer Overflow on Multiplication and Division - comp ... Multiplying two int32 numbers - comp.dsp That is the ... Rounding numbers - comp.databases.filemakerMultiplying two int32 numbers - comp.dsp Round-Function It however only rounds to an integer ... link see above): ---snip--- 12.2.4 Rounding Numbers ... scaling around the ... element-by-element multiplication - comp.soft-sys.matlab ...Multiplying two int32 numbers - comp.dsp element-by-element multiplication - comp.soft-sys.matlab ... Multiplying two int32 numbers - comp.dsp element-by-element ... how to use xlswrite for image files ? - comp.soft-sys.matlab ...Matlab xlswrite complex numbers - comp.soft-sys.matlab Hi im trying to use the xlswrite command to create a excel document from my ... Multiplying two int32 numbers - comp ... Suitable Integer Square Root Algorithm for 32-64-Bit Integers on ...... of larger integers have to be done by multiplying ... 2 + x<<(k+1) + 1<<(2*k) That's two ... site, untested) Here a C-routine for integer-square-rooting for numbers ... conventional wisdom how to upsample very large arrays, accurately ...It is well possible that zero-crossings can't be found > > by multiplying two consecutive numbers ... Since you are upsampling by 4 (an integer factor) you don't even need ... how to assign a NaN? - comp.lang.fortran... this will slow things down if numbers are ... such a NaN, being just another integer (or ...) from the compiler's point of view, may be generated by, say, multiplying two ... Multiplying two int32 numbers - comp.dsp | Computer GroupWhen evaluating the expression Z = X * Y where X and Y is SINT32 you can do it like this: X0 = bitand(bitshift(x,-24),255); X1 = bitand(bitshift(... IntegersMultiplying Integers. To multiply a pair of integers ... by looking at their positions on the number line. For any two different places on the number line, the integer on ... 7/28/2012 8:51:26 PM
|
Search |
Post Question |
Groups |
Stream |
About |
Register
19 followers. 