realization of the following circuit

[sun_ray]

How can I realize the following circuit ?

out = In x 1600


What can be the RTL for it?

 

[dll_fpga]

How can I realize the following circuit ?

out = In x 1600


What can be the RTL for it?

write the RTL such that input is shifted log2 1600 to the left ....append the remaining with ZEros.....enjoy!!

 

[Alosevskoy]

Hi! I think it'll be like this

module hard_mult_by_1600(In, Out);
input [7:0] In;
output [18:0] Out;

// Out = (64 + 512 + 1024) * In = (2^6 + 2^9 + 2^10) * In

wire [13:0] prt_mult_by_64 = In << 6;
wire [16:0] prt_mult_by_512 = In << 9;
wire [17:0] prt_mult_by_1024 = In << 10;

assign Out = prt_mult_by_64 + prt_mult_by_512 + prt_mult_by_1024;

endmodule

 

[alexan_e]

write the RTL such that input is shifted log2 1600 to the left ....append the remaining with ZEros.....enjoy!!

1600 is not a power of 2

 

[sun_ray]

Hi dll_fpga,

I am unable to understand your basic principal. Can you please elaborate?

Do you think Alosevskoy has provided the correct solution?

 

[Alosevskoy]

Emm.... Do we speak about logarithm of 2?? I thought it's about input signal In multiblied by 1600...

 

[sun_ray]

Hi Alosevskoy

You are correct that it was an input signal. Here In is an input signal. Can you suggest any other principal to solve the above problem because your principle may require big size adder.

Regards

 

[rca]

adder will used less ressource than multiplier, the Alosevsko proposal seem interesting.

the other question is why to multiply by 1600 in hardware?, what do you do with the "out" signal?

 

[Alosevskoy]

I've never done this before but I think one can use smaller adder to calculate small portions of sum sequentially and keep the intermediate results in a dedicated register... but it'll take more time (in term of clock cycles) than pure comb logic as I've posted above.

 

[sun_ray]

The question was more to know thee way a multiplication can be done by any number rather than specifically by 1600. The 1600 was chosen to frame the question.

 

[rca]

fix shift does not "consume" time.

 

[sun_ray]

What do you want to mean by your above sentence rca?

 

[Alosevskoy]

sun_ray, You've chosen a good number :-D I think only a multiplier can solve this task...

 

[rca]

<<6 is shifter that mean, only move the bit position, so just need to rediret the bit 0 to position 6, to be added with the other shift....

 

[sun_ray]

Hi Alosevskoy

Your rtl code is correctly implementing multiplication by 1600 without multiplier. How do you say that you think think only a multiplier can solve this task?

 

[Alosevskoy]

sun_ray, I mean for the task of multiplication In by any number a general purpose multiplier is needed....

 

[FvM]

The question was more to know thee way a multiplication can be done by any number rather than specifically by 1600. The 1600 was chosen to frame the question.

So you mean a constant multiplier? To make it a reasonable RTL question, you should mention bit lengths respectively ranges for all involved terms.

As long as you don't intend a bit serial multiplier, you can just write In*constant and leave the implementation to the synthesis tool.

 

[sun_ray]

fvm

I can allow as much bits as required for 1600 and for in and for out also. Where
is the problem? Can u please clarify?

 

[FvM]

I don't see a problem. To synthesize hardware, you need to define the bit widths.

 

[sun_ray]

What do you mean by "As long as you don't intend a bit serial multiplier,"? Do you want to mean that depending on the code style it can be a serial multiplier or other type of multiplier? How does the code look like then?