topic Re: CVF 6 to Intel 9.1 Format changes #.###e+001 vs #.###e+01 in Intel® Fortran Compiler

CVF 6 to Intel 9.1 Format changes #.###e+001 vs #.###e+01

carlls — Tue, 11 Aug 2009 17:15:28 GMT

Greeetings:

I have a mix of CVF 6 and Intel 9.1 FORTRAN codes. Seems CVF put out numbers with e+## format while the newer Intel puts out e+###. So reading old file in new code is no issue but not all codes are ported yet.

Is there a way to have Intel 9.1 FORTRAN put out e+## range?

Regards
Carl

Re: CVF 6 to Intel 9.1 Format changes #.###e+001 vs #.###e+01

Steven_L_Intel1 — Tue, 11 Aug 2009 17:26:25 GMT

Please supply a test program that shows this behavior. I don't know of any reason why simply changing compilers should change the formatted output - especially as you're saying that a character is being dropped.

Also, I'll comment that 9.1 is quite old at this point.

Re: CVF 6 to Intel 9.1 Format changes #.###e+001 vs #.###e+01

carlls — Wed, 12 Aug 2009 11:27:44 GMT

Quoting - Steve Lionel (Intel)

Better than that I will solve my own problem ....

Seems the E format will take on optional format character E#so E15.6E2 will do the trick .... so the last digit sets the range of values.

As an alternative a "simple" sed script could be use to change the files ....

#!/bin/bash
cp $1 $1.bak
cat $1.bak | sed -e "s/([0-9.]*e[+-])0([0-9][0-9])/12 /g" >$1

Hope this helps someone else....

Regards
Carl

Re: CVF 6 to Intel 9.1 Format changes #.###e+001 vs #.###e+01

Steven_L_Intel1 — Wed, 12 Aug 2009 13:15:05 GMT

I am concerned about the behavior change, which should not be there. Please show me an example that illustrates the difference.