[prev in list] [next in list] [prev in thread] [next in thread] 

List:       pcc-list
Subject:    Re: [Pcc] =?utf-8?q?Side_effects_of_=5F=5Fbuiltin=5Funimpl=28=29?=
From:       "Gregory McGarry" <greg () bitlynx ! com>
Date:       2014-06-04 8:31:57
Message-ID: 20140604013156.ab62c74ef4fdac7a72bd4f4cde8f70c1.54c4e1b875.wbe () email03 ! secureserver ! net
[Download RAW message or body]

Is the amd64 backend working for you on OS X?  I've only just switched
from using the i386 backend on OS X.


> -------- Original Message --------
> Subject: Re: [Pcc] Side effects of __builtin_unimpl()
> From: Anders Magnusson <ragge@ludd.ltu.se>
> Date: Wed, June 04, 2014 12:46 am
> To: Gregory McGarry <greg@bitlynx.com>, pcc@lists.ludd.ltu.se
> 
> 
> Eh, where is this function declared?  grep in /usr/include do not find 
> it... (OSX 10.9).
> Anyway, the code (despite ugly) should work, but it may need some 
> trickery, can you please add it to Jira?
> 
> -- Ragge
> 
> 
> Gregory McGarry skrev 2014-06-04 08:38:
> > For unimplemented builtins, (I think) a local symbol name is allocated
> > and the builtin is mapped to the version in libc.
> >
> > This causes a problem when there is already a local variable of the same
> > name.  For example, the apple headers have the following code which
> > fails to build:
> >
> > static __inline__  int __inline_isnormalf(float __x ) {
> >           float fabsf = __builtin_fabsf(__x);
> >           if( __x != __x ) return 0;
> >           return fabsf < __builtin_inff() && fabsf >= __FLT_MIN__;
> > }
> >
> > _______________________________________________
> > Pcc mailing list
> > Pcc@lists.ludd.ltu.se
> > http://lists.ludd.ltu.se/cgi-bin/mailman/listinfo/pcc
_______________________________________________
Pcc mailing list
Pcc@lists.ludd.ltu.se
http://lists.ludd.ltu.se/cgi-bin/mailman/listinfo/pcc
[prev in list] [next in list] [prev in thread] [next in thread]