Hi All,

I have an unformatted file, and I open it for reading in big
endian mode. The file has a repetitive structure (is a time step
report for a simulation): at the beginning of every time step, there
is this pseudo-declaration:

'PARAMS'
keyword number kind

where "keyword" is an 8 character string, "number" is an integer which
tells me how much data this time step contains, and "kind" is a 4
character string which tells me the data type (i.e., 'CHAR', 'REAL'
and so on). After this declaration, the data for the current time step
follows. As an example:

'PARAMS ' 4200 'REAL'
0.5372 3.7393 0.3273 .... (4200 real numbers)

Infinity77 wrote:

When describing an UNFORMATTED file, be sure to indicate the
record boundaries. Looking at the code, it seems that the
keyword number kind indicator is a separate record, but that isn't
obvious as described here.

On 2008-05-05, Infinity77 gmail.com> wrote:

If you're not going to read/write the file on multiple platforms with
different native endianness, get rid of the CONVERT= stuff. Converting
endianness on the fly has much more overhead than just streaming the
data directly into application memory.

On 2008-05-05, glen herrmannsfeldt ugcs.caltech.edu> wrote:

At least for gfortran the opposite is true. The implied do loop above
is handled by the frontend, IIRC converting it to (roughly) the
equivalent do loop like

do i = 1000*internalCount+1, 1000*(internalCount+1)
read (1, end=40) matrix(i,matrixIndex)
end do

Janne Blomqvist wrote:

I said one might be faster, but not which one.

Well, it can't be quite like that because of the record
boundaries, but one call to the I/O routine per array element
I do understand.