This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Bug 15718 dispute: showstopper or WONTFIX?


Joe Buck <jbuck@welsh-buck.org> writes:

| Joe Buck <Joe.Buck@synopsys.COM> writes:
| > | Since 3.3.4 shipped with this bug, we need a release note describing, in
| > | user-level terms, how to avoid the bug.
| > | 
| > | We can't write "don't pass a __complex__ in a register".  After all,
| 
| On Tue, 2004-06-01 at 19:49, Gabriel Dos Reis wrote:
| > Sure.
| > 
| >    Due to a calling convention bug in the Alpha backend, the compiler
| >    may not accept some constructs involving function calls using native
| >    complex numbers either as parameter values or return value. 
| 
| > That undoubtedly needs polishing.
| 
| It understates the issue considerably.  LAPACK 3.0 hits this bug.

In which ways does it understate the issue?  As far as I can tell, the 
description I gave covers the case you're mentioning.
I'm not pretending the wordsmithing was perfect, but I think saying it
understates the issue without providing better wording does not give
me any clue.

| I believe that *every* Fortran program using an inline function
                                         ^^^^^^^^^^^^^^^^^^^^^^^^
| with a complex argument hits the bug, though I'm not certain.
  ^^^^^^^^^^^^^^^^^^^^^^^

| 
| It's possible that, since it's a subreg bug, using double precision
| complex variables in certain places would avoid the bug.  I'll do
| some experiments tomorrow.
| 
| 

-- 
                                                       Gabriel Dos Reis 
                                           gdr@integrable-solutions.net


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]