This is the mail archive of the gcc-patches@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]

[PATCH][committed] Powerpc patch #5, Add power7/vsx support (fix bugs in patch #4)


On Thu, Jul 30, 2009 at 11:53:13AM -0400, David Edelsohn wrote:
> On Tue, Jul 28, 2009 at 6:31 PM, Michael
> Meissner<meissner@linux.vnet.ibm.com> wrote:
> > This patch is a respin of patch #4, that fixes the V2DImode problem on altivec
> > that I noticed after submitting the patch, and to address the comments David
> > made about patch #4 (using CONST0_RTX instead of const_int 0, and changing the
> > UNSPEC number so it doesn't collide with rs6000.md).
> >
> > Note, the patch is getting long, so I split out the documentation and testsuite
> > patches for patch #6.
> >
> > Is this ok to commit to the tree? ?The branch ibm/power7-tmp has been sync'ed
> > up to subversion id 150177, if people just want to check out the branch to try
> > it out.
> 
> This patch is okay with Uros's comments addressed.
> 
> Thanks, David

The patch has been commited, along with this patch that removes handling of
abs/min/max in rs6000_builtin_vectorized_function, since they are handled in
machine independent code.

--- gcc/config/rs6000/rs6000.c.~2~	2009-07-30 13:39:18.000000000 -0400
+++ gcc/config/rs6000/rs6000.c	2009-07-30 14:06:56.000000000 -0400
@@ -3135,53 +3135,8 @@ rs6000_builtin_vectorized_function (unsi
           && in_mode == SFmode && in_n == 4)
         return rs6000_builtin_decls[VSX_BUILTIN_XVRSPIC];
       break;
-    case BUILT_IN_FABS:
-      if (VECTOR_UNIT_VSX_P (V2DFmode)
-	  && out_mode == DFmode && out_n == 2
-	  && in_mode == DFmode && in_n == 2)
-	return rs6000_builtin_decls[VSX_BUILTIN_XVABSDP];
-      break;
-    case BUILT_IN_FABSF:
-      if (out_mode != SFmode || out_n != 4
-	  || in_mode != SFmode || in_n != 4)
-	break;
-      if (VECTOR_UNIT_VSX_P (V4SFmode))
-	return rs6000_builtin_decls[VSX_BUILTIN_XVABSSP];
-      if (VECTOR_UNIT_ALTIVEC_P (V4SFmode))
-	return rs6000_builtin_decls[ALTIVEC_BUILTIN_ABS_V4SF];
-      break;
-    case BUILT_IN_FMAX:
-      if (VECTOR_UNIT_VSX_P (V2DFmode)
-	  && out_mode == DFmode && out_n == 2
-	  && in_mode == DFmode && in_n == 2)
-	return rs6000_builtin_decls[VSX_BUILTIN_XVMAXDP];
-      break;
-    case BUILT_IN_FMAXF:
-      if (out_mode != SFmode || out_n != 4
-	  || in_mode != SFmode || in_n != 4)
-	break;
-      if (VECTOR_UNIT_VSX_P (V4SFmode))
-	return rs6000_builtin_decls[VSX_BUILTIN_XVMAXSP];
-      if (VECTOR_UNIT_ALTIVEC_P (V4SFmode))
-	return rs6000_builtin_decls[ALTIVEC_BUILTIN_VMAXFP];
-      break;
-    case BUILT_IN_FMIN:
-      if (VECTOR_UNIT_VSX_P (V2DFmode)
-	  && out_mode == DFmode && out_n == 2
-	  && in_mode == DFmode && in_n == 2)
-	return rs6000_builtin_decls[VSX_BUILTIN_XVMINDP];
-      break;
-    case BUILT_IN_FMINF:
-      if (out_mode != SFmode || out_n != 4
-	  || in_mode != SFmode || in_n != 4)
-	break;
-      if (VECTOR_UNIT_VSX_P (V4SFmode))
-	return rs6000_builtin_decls[VSX_BUILTIN_XVMINSP];
-      if (VECTOR_UNIT_ALTIVEC_P (V4SFmode))
-	return rs6000_builtin_decls[ALTIVEC_BUILTIN_VMINFP];
-      break;
     default:
-      ;
+      break;
     }
   return NULL_TREE;
 }

-- 
Michael Meissner, IBM
4 Technology Place Drive, MS 2203A, Westford, MA, 01886, USA
meissner@linux.vnet.ibm.com


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