btest , what's wrong...

[lparedi]

do k=1,10
ibip=0
do j=0,15
if(btest(k,j))ibip=ibip+1
enddo
write(6,*)k,ibip
enddo

stop
end



Version 12.0.X give wrong results...version 11 is ok.

[forcpp]

The maximum bit positions for maximum integer, k=10, will be 4 that is j=0,3. This is the possible reason because of which you are getting wrong answer. I checked your code by ifort 12.0.3.167and it is giving wrong result because of wrong declaration. But, the gfortran compiler behaves differently and gives right answer with the present form of the code.

[Steven_L_Intel1]

Same program as here - where is this coming from?

[mecej4]

> it is giving wrong result because of wrong declaration

What "wrong declaration"? The second argument to the intrinsic function btest(i, pos) is required to satisfy 0 pos < bit_size(i). Do you think there is a violation of this requirement in the test program?

[forcpp]

This may be bug as I see some older posts. But this can be circumvented if we declare j = 0, 3. Moreover, I also noticed that the results are coming right up to j=0, 10. If we give j=0, 11 results go wrong.

[Steven_L_Intel1]

I don't see anything wrong with the source as posted. This is pretty much the exact same program as in the other thread. Clearly both of you read this somewhere - where?