simple question relevant to format

[yugn]

I paste the code first:

open(5,file='read-test.txt',status='old')
read(5,'(F8.2)') rd
write(6,'(F8.2)') rd

content in read-test.txt:
.0028 .0197 270.25 40.0


here's the question:
why the read command whose format is '(F8.2)', reads all 4bit fraction , while the write command strictly output only 2bit fraction? I'm confused by the different behaviors between read and write.

[Steven_L_Intel1]

You should read the section of the Compaq Fortran Language Reference Manual pertaining to F format (11.2.4.1)

Briefly, on input, the only effect the .2 has is if the input value has no decimal point - then it assumes the rightmost two digits are the fraction.

It is often the case that formats do different things on input and output. This is explained in the LRM.

Steve

[yugn]

thank you, steve
It's always a good time to post question here!

What LRM stands for?

[Intel_C_Intel]

Steve pointed to chapter 11.2.4.1 (F editing) in Language Reference Manual (LRM). It's I/O formatting (pages 11-16 in the printed version or 291 in electronic version (pdf)). Examples from there (separated by semicolon)
Format; Input; Value
F8.5; 123456789; 123.45678
F8.5; -1234.567; -1234.56
F8.5; 24.77E+2; 2477.0
F5.2; 1234567.89; 123.45

Format; Value; Output
F8.5; 2.3547188; ^2.35472
F9.3; 8789.7361; ^8789.736
F2.1; 51.44; **
F10.4; -23.24352; ^^-23.2435
F5.2; 325.013; ******
F5.2; -.2; -0.20

A.

[Steven_L_Intel1]

Hmm - the examples in the manual don't seem to have taken rounding into account. I'll look into that.

[Intel_C_Intel]

Steve,
I checked it, it's OK (I mean LRM) for double precision - 1234.45678 (no rounding) while for single it's rounded to 123.4568.

A.

[yugn]

thanks, arturguzik, I finally found the topic in Online Document(with this keyword F8.5). I didnt find LRM's pdf file.

I think, it is not necessary for read to take rounding into account, but it is necessary for write.

[Intel_C_Intel]

All docs in html and pdf format you can download from here.

A.

[Steven_L_Intel1]

Read rounds the value to the closest representable value - the .n part of the format doesn't matter for this.

Steve