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, rs6000] ELFv2 ABI preparation: Refactor rs6000_arg_partial_bytes


Hello,

this is the final patch preparing for the new ABI.

The logic in rs6000_arg_partial_bytes is a bit complex, since it apparently
still contains remnants from the time where this routine was used even for
arguments that are returned both in GPRs and FPRs/VRs (*and* memory).

These days, all such cases are handled completely in rs6000_function_arg
so rs6000_arg_partial_bytes should always return 0; which it *does*, even
though in a somewhat complex way.

This complex logic would make handling homogeneous structs more difficult
than it needs to be.  Therefore, this patch simplifies the logic by making
explicit the fact that rs6000_arg_partial_bytes does not actually need to
handle the cases described above.

No change in generated code expected.

Tested on powerpc64-linux and powerpc64le-linux.

OK for mainline?

Bye,
Ulrich


ChangeLog:

2013-11-11  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>

	* config/rs6000/rs6000.c (rs6000_arg_partial_bytes): Simplify logic
	by making use of the fact that for vector / floating point arguments
	passed both in VRs/FPRs and in the fixed parameter area, the partial
	bytes mechanism is in fact not used.

Index: gcc/gcc/config/rs6000/rs6000.c
===================================================================
--- gcc.orig/gcc/config/rs6000/rs6000.c
+++ gcc/gcc/config/rs6000/rs6000.c
@@ -9835,15 +9835,25 @@ rs6000_arg_partial_bytes (cumulative_arg
 			  tree type, bool named)
 {
   CUMULATIVE_ARGS *cum = get_cumulative_args (cum_v);
+  bool passed_in_gprs = true;
   int ret = 0;
   int align_words;
 
   if (DEFAULT_ABI == ABI_V4)
     return 0;
 
-  if (USE_ALTIVEC_FOR_ARG_P (cum, mode, named)
-      && cum->nargs_prototype >= 0)
-    return 0;
+  if (USE_ALTIVEC_FOR_ARG_P (cum, mode, named))
+    {
+      /* If we are passing this arg in the fixed parameter save area
+         (gprs or memory) as well as VRs, we do not use the partial
+	 bytes mechanism; instead, rs6000_function_arg wil return a
+	 PARALLEL including a memory element as necessary.  */
+      if (TARGET_64BIT && ! cum->prototype)
+	return 0;
+
+      /* Otherwise, we pass in VRs only.  No partial copy possible.  */
+      passed_in_gprs = false;
+    }
 
   /* In this complicated case we just disable the partial_nregs code.  */
   if (TARGET_MACHO && rs6000_darwin64_struct_check_p (mode, type))
@@ -9853,24 +9863,27 @@ rs6000_arg_partial_bytes (cumulative_arg
 
   if (USE_FP_FOR_ARG_P (cum, mode))
     {
+      unsigned long n_fpreg = (GET_MODE_SIZE (mode) + 7) >> 3;
+
       /* If we are passing this arg in the fixed parameter save area
-	 (gprs or memory) as well as fprs, then this function should
-	 return the number of partial bytes passed in the parameter
-	 save area rather than partial bytes passed in fprs.  */
+         (gprs or memory) as well as FPRs, we do not use the partial
+	 bytes mechanism; instead, rs6000_function_arg wil return a
+	 PARALLEL including a memory element as necessary.  */
       if (type
 	  && (cum->nargs_prototype <= 0
 	      || (DEFAULT_ABI == ABI_AIX
 		  && TARGET_XL_COMPAT
 		  && align_words >= GP_ARG_NUM_REG)))
 	return 0;
-      else if (cum->fregno + ((GET_MODE_SIZE (mode) + 7) >> 3)
-	       > FP_ARG_MAX_REG + 1)
+
+      /* Otherwise, we pass in FPRs only.  Check for partial copies.  */
+      passed_in_gprs = false;
+      if (cum->fregno + n_fpreg > FP_ARG_MAX_REG + 1)
 	ret = (FP_ARG_MAX_REG + 1 - cum->fregno) * 8;
-      else if (cum->nargs_prototype >= 0)
-	return 0;
     }
 
-  if (align_words < GP_ARG_NUM_REG
+  if (passed_in_gprs
+      && align_words < GP_ARG_NUM_REG
       && GP_ARG_NUM_REG < align_words + rs6000_arg_size (mode, type))
     ret = (GP_ARG_NUM_REG - align_words) * (TARGET_32BIT ? 4 : 8);
 
-- 
  Dr. Ulrich Weigand
  GNU/Linux compilers and toolchain
  Ulrich.Weigand@de.ibm.com


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