Peter Kremsner wrote:
> What programming language do you use?
> Maybe the function name is to_long, or you could transform the 64Bit
> vector to 32bit, this truncates the information in the lower 32Bit but
> maybe the precision is enough for your purpose.
>
> Why would you transform to integer datatype, if your datatype is double
> just round it and keep the double datatype


Thanks for your quickly response,

I use VHDL language (xilinx ISE),
The program concerns manipulation of image block data in order to
achieve a motion estimation, I use a register (32bits) for saving (or
delaying) the data, thus the program works when data size is less than
32-bits but in my case I need 64-bits, so how can I do this?

Thank you,

Peter Kremsner wrote:
> Truncate the 64 Bit to 32 Bit or use 2 32Bit Register for saving the
> data.
> 1 for the low 32Bit the other for the high 32Bit and perform the
> operation with two registers.


Thank you again for your valuable comments,
I did as you mentioned, I split my data into two parts, once using 
32-bit registers, and the second part another 32-bit registers, only I 
hope it wont be a blend of data :)

Thank you all,
Best regards,

Mira Miyou wrote:
> I did as you mentioned, I split my data into two parts
And what do you do with that segmented value afterwards?

My way would be to use a unsigned(63 downto 0) vector and the 
numeric_std for the calculations with it...

Lothar Miller wrote:
> And what do you do with that segmented value afterwards?
>
> My way would be to use a unsigned(63 downto 0) vector and the
> numeric_std for the calculations with it...

Actually I reduced the size of register at 32bits and I changed the way 
in which the data are presented in my program so that it goes with the 
32bits, since anyway I can't go beyond.

concerning the function unsigned, I don't use it because before storing 
the value in the register I use an absolute value like this:
abs (to_integer (signed (data)))
thank you again for the time you spent to write me,

berndl wrote:
> Mira Miyou wrote:
>> abs (to_integer (signed (data)))
>
> you should accept the fact that VHDL has integers (that's signed by the
> way!) of only 32 bits. For anything beyond this, you might follow
> Lothars advice and go to 'signed' and 'unsigned' (which are unlimited in
> bit-width)...

Ok, thank's for your advice :)
Regards,