[Bug target/58160] Power8 fusion support has a bug that shows up in running spec 2006

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58160

Michael Meissner  changed:

   What|Removed |Added

 Status|ASSIGNED|RESOLVED
 Resolution|--- |FIXED

--- Comment #4 from Michael Meissner  ---
Fixed in subversion id 201792.

[Bug target/58160] Power8 fusion support has a bug that shows up in running spec 2006

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58160

--- Comment #3 from Michael Meissner  ---
I forgot to mention, with this patch, I have built and successfully run 403.gcc
and 435.gromacs using the -O2 -m32 -mcpu=power7 -mtune=power8 options that
broke 403.gcc, and also with the -O3 -funroll-loops -m32 -mcpu=power7
-mtune=power8 that broke 435.gromacs.

[Bug target/58160] Power8 fusion support has a bug that shows up in running spec 2006

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58160

--- Comment #2 from Michael Meissner  ---
Created attachment 30659
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30659&action=edit
Proposed patch to fix problem

[Bug target/58160] Power8 fusion support has a bug that shows up in running spec 2006

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58160

Michael Meissner  changed:

   What|Removed |Added

 Status|UNCONFIRMED |ASSIGNED
   Last reconfirmed||2013-08-14
   Assignee|unassigned at gcc dot gnu.org  |meissner at gcc dot 
gnu.org
 Ever confirmed|0   |1

--- Comment #1 from Michael Meissner  ---
In looking at the code for 403.gcc, the compiler spills the addis value to the
stack and then reloads it later.  It would be more optimal to just redo the
addis value.