[Bug target/23892] New: -msse causes SSE code generation without -mfpmath=sse

[corey dot taylor at gmail dot com]

I cannot find anything that says this should happen.  It is causing SIGILL on
the host system.

When enabling -msse or -msse2 I expect the intrinsics to be enabled.  This would
then be useful for creating SSE and non-SSE functions based off of CPUID.  

However with these enabled, other code is generated such as floating point and
cast optimizations.  

This system is not x86-64 and does not have -mfpmath=sse enabled.  I don't know
how to simply enable the intrinsics and not change the code generation.  

corey

-- 
   Summary: -msse causes SSE code generation without -mfpmath=sse
   Product: gcc
   Version: 4.0.1
Status: UNCONFIRMED
  Severity: normal
  Priority: P2
 Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: corey dot taylor at gmail dot com
CC: gcc-bugs at gcc dot gnu dot org
  GCC host triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23892

[Bug target/23892] -msse causes SSE code generation without -mfpmath=sse

[corey dot taylor at gmail dot com]

--- Additional Comments From corey dot taylor at gmail dot com  2005-09-15 
02:36 ---
Just the answer I was looking for.  I tried a search but didn't find anything
with my keywords.

corey

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23892

[Bug c++/22354] g++ accepts specializiation without declaration

[corey dot taylor at gmail dot com]

--- Additional Comments From corey dot taylor at gmail dot com  2005-07-08 
05:30 ---
The reason we decided to post this is many front ends were giving different
results: some errored out with diagnostic, some accepted, some accepted but
wouldn't allow proper declaration (considered it a redefinition which is against
pargraph 15 of the same section).

Considering that this implementation currently allows either method without
portability issues then the enhancement is definately optional as long as it
works properly as-is.

Just wanted to see if this (and we sent to other projects the same issue) is how
implementation was intended.  I was part of the discussion on the list.

corey

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22354