Data representation question

[FX]

> Since none of us here are really knowledgeable about Fortran, this is
> just a follow-up query to see whether we should expect any problems from
> this different choice of representation.  It seems clear that the use of
> TRANSFER is extremely target-dependent, and that code making assumptions
> about bit representations when using TRANSFER should expect some porting
> problems.

I don’t expect you’ll have much trouble. The front-end code should be totally clean, i.e. it translates directly from the MPFR-represented values (obtained by parsing and front-end simplification passes) to the middle-end trees, without any assumptions about format. [1] The library is probably also safe, because we don’t manipulate the individual bits, we use libc functions instead.

I expect the testsuite might have a few more occurrences of bit-pattern comparison, though. I’ve looked on the gcc-testresults archive but couldn’t find test results for your target (powerpc64le), though!


FX


[1] In fact, the only case that currently needs special handling is the IEEE extended double, in gfc_init_kinds(). But it’ll work whether you’re LE or BE.