Error observed in using bitsll and bitsra

When utilizing bitsll and bitsra, an error was noticed.

I have the following code and intend to utilize Embedded.fi [fixed point ] as double implementation show in case -1.

Can someone please show me how to code correctly and help me understand what I'm missing in cases 2 and 3?

It would be nice if I could do cases 2 and 3 in a single function/ single solution.

Thank you very much.

% case -1 : double

k = linspace(-12,12,10); 
for i = 1 : length(k)
    a =  2 ^ k(i)
end

a = 2.4414e-04

a = 0.0016

a = 0.0098

a = 0.0625

a = 0.3969

a = 2.5198

a = 16

a = 101.5937

a = 645.0796

a = 4096

% case -2: Embdedded fi ,fixed point - right shift

k = fi(linspace(-12,12,10),1,32,27);
a = int8(1);
for i = 1 : length(k)
%   fxpOut = bitsra(a,k(i))
  disp(bin(bitsra(a,k(i))))
end

Error using bitsra
K must be numeric, scalar, real, integer-valued, and greater than or equal to zero in BITSRA(A,K).

Error in embedded.fi/bitsra (line 38)
y = bitsra(x, double(kin));

% case - 3: Embdedded fi ,fixed point - left shift

k = fi(linspace(-12,12,10),1,32,27);
a = int8(1);
for i = 1 : length(k)
  fxpOut = bitsll(a,k(i))
  disp(bin(bitsll(a,k(i))))
end

[Richard McCormack]

Hi @Life is Wonderful,

As th error messages indicates, the issue is that linspace(-12,12,10) results in k containing negative values. bitsll and bitsra do not support shifting with a negative value. Can you explain what you like the behavior to be in the case where k(i) is negative?

[Life is Wonderful]

@Richard McCormack,

Thank you!!.

As you indicated, bitsll and bitsra do not support negative values; do we have an alternative way that uses I can do an in fixed point calculation?

kn = [-12]

kn = -12

kp = [12]

kp = 12

dblout  = 2 ^ kn

dblout = 2.4414e-04

fpxout  = 2 ^ kp

fpxout = 4096

In my head, I am carrying out the following procedure.

   K = [-12,12]

K = 1×2

-12 12

for i = 1:length(K)

if sign(K(i)) < 0 % Bitsliceget to know bitposition and do a bitand for sign

% Negative

b = fi(bitsra(1,abs(K(i))));

disp(bin(b))

else

% Positive

b_i = fi(bitsra(1,abs(K(i))));

disp(bin(b_i))

end

end

say

k =  -12

k = -12

fpx_neg = bitsra(1,abs(k))

fpx_neg = 2.4414e-04

k =  12

k = 12

fpx_pos = bitsll(1,abs(k))

fpx_pos = 4096

I'm confident that you'll have more ideas than I did.

Thank you very much

[Richard McCormack]

Hi @Life is Wonderful,

I am happy that you made progress!

I made some adjustments to your code to make it work for code generation.

But maybe there is a reason that you can't use abs. If that is the case, please let me know why you can't use abs.

function scratch
coder.extrinsic('bin');
K = [-12,12];
coder.unroll;
for i = 1:length(K)
    if sign(K(i)) < 0  % Bitsliceget to know bitposition and do a bitand for sign
        % Negative
        b = fi(bitsra(1,abs(K(i))));
        disp(bin(b))
    else
        % Positive
        b_i = fi(bitsra(1,abs(K(i))));
        disp(bin(b_i))
    end
end
end

[Life is Wonderful]

Thanks a lot @Richard McCormack.

I am fine in using abs. I will try out your suggested code and make some test.