Loss of accuracy

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

Hi

I am using the pgi compiler and I am decomposing a large number (index) into various separate smaller indexs.

It starts working ok, but then it starts making mistakes after about 16m calculations.

e.g. correct output to start:

1 1 1 0 1 1
2 1 1 0 1 1
3 1 1 0 1 1
4 1 1 0 1 1
5 1 1 0 1 1
6 1 1 0 1 1
7 1 1 0 1 1
8 1 1 0 1 1
9 1 1 0 1 1
10 1 1 0 1 1
11 1 1 0 1 1
12 1 1 0 1 1
13 1 1 0 1 1
14 1 1 0 1 1
15 1 1 0 1 1
16 1 1 0 1 1
17 1 1 0 1 1
18 1 1 0 1 1
19 1 1 0 1 1
20 1 1 0 1 1

then later it breaks and the numbers dont count smoothly anymore. It isnt the maths behind it, so I am assuming it is the compiler?
e.g. broken number list

1 16 6 1 30 11
1 17 6 1 30 11
1 17 6 1 30 11
3 17 6 1 30 11
5 17 6 1 30 11
5 17 6 1 30 11
5 17 6 1 30 11
7 17 6 1 30 11
9 17 6 1 30 11
9 17 6 1 30 11
9 17 6 1 30 11
11 17 6 1 30 11
13 17 6 1 30 11
13 17 6 1 30 11
13 17 6 1 30 11
15 17 6 1 30 11
17 17 6 1 30 11
17 17 6 1 30 11
17 17 6 1 30 11
19 17 6 1 30 11

Anyone have any idea what could be causing it?
Thanks
I have attached my code bellow:

Index = 12*CAPEq*CAPD*CAPBM*(CAPCR+1)*CAPIRCs

DO L = 1, Index
	!! extract the categories
   floorn = L-1
   Eqitr = MOD(FLOOR(floorn), CAPEq)+1

   floorn = (L-1)/CAPEq
   Debtitr = MOD(FLOOR(floorn), CAPD)+1

   floorn = (L-1)/(CAPEq*CAPD)
   BMitr = MOD(FLOOR(floorn), CAPBM)+1

   floorn = (L-1)/(CAPEq*CAPD*CAPBM)
   K = MOD(FLOOR(floorn), (CAPCR+1))

   floorn = (L-1)/(CAPEq*CAPD*CAPBM*(CAPCR+1))
   IRcat2 = MOD(FLOOR(floorn), CAPIRCs)+1

   floorn = (L-1)/(CAPEq*CAPDebt*CAPBM*(CAPCR+1)*CAPIRCs)
   R = MOD(FLOOR(floorn), 12)+1

   WRITE(3,'(6I3)') Eqitr, Debtitr, BMitr, K, IRcat2, R
ENDDO
2

Please provide a complete program and information on how you

compiled it. The details of data declarations, etc, are where the issues
are found.

Make sure your all of your variables and functions are specifically
declared (integer8/integer4 instead of integer, etc). Often it is a mixing of integer8 with integer4 where inaccuracy can happen.
Avoid using -i4 or -i8 to make everything one type, since this can
cause constants and functions to be mis-declared.

dave