[PATCH 1/2] rs6000: Another PIC LRA fix

[PATCH 1/2] rs6000: Another PIC LRA fix

[Segher Boessenkool]

This one for TARGET_TOC && TARGET_MINIMAL_TOC.  Without it, r30 is not
saved in the prologue for functions that do not use r30, but the register
is set later in the prologue anyhow.  This made all java tests fail.

Tested on powerpc64-linux, --enable-languages=all,ada,go,obj-c++ and test
variants {-m32/-mno-lra,-m32/-mlra,-m32/-mpowerpc64,-m64/-mno-lra,-m64/-mlra};
and on powerpc64le-linux, everything default.  Both also with bootstrapping
with LRA defaulted on.

Okay for trunk?


Segher


2015-10-31  Segher Boessenkool  <segher@kernel.crashing.org>

	* config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Use the
	same condition for testing whether RS6000_PIC_OFFSET_TABLE_REGNUM is
	live as for using it elsewhere, for TARGET_MINIMAL_TOC.

---
 gcc/config/rs6000/rs6000.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 9610625..1c55244 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -21793,7 +21793,9 @@ rs6000_reg_live_or_pic_offset_p (int reg)
           || (reg == RS6000_PIC_OFFSET_TABLE_REGNUM
 	      && !TARGET_SINGLE_PIC_BASE
               && ((DEFAULT_ABI == ABI_V4 && flag_pic != 0)
-                  || (DEFAULT_ABI == ABI_DARWIN && flag_pic))));
+                  || (DEFAULT_ABI == ABI_DARWIN && flag_pic)
+		  || (TARGET_TOC && TARGET_MINIMAL_TOC
+		      && get_pool_size () != 0))));
 }
 
 /* Return the first fixed-point register that is required to be
-- 
1.9.3

[PATCH 2/2] rs6000: Rewrite rs6000_reg_live_or_pic_offset_p

[Segher Boessenkool]

This function is quite a puzzle; untangle it.  No functional change.

Tested etc.; okay for trunk?


Segher


2015-10-31  Segher Boessenkool  <segher@kernel.crashing.org>

	* config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Rewrite.

---
 gcc/config/rs6000/rs6000.c | 35 ++++++++++++++++++++---------------
 1 file changed, 20 insertions(+), 15 deletions(-)

diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 1c55244..37abf88 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -21780,22 +21780,27 @@ save_reg_p (int r)
 static bool
 rs6000_reg_live_or_pic_offset_p (int reg)
 {
-  /* If the function calls eh_return, claim used all the registers that would
-     be checked for liveness otherwise.  This is required for the PIC offset
-     register with -mminimal-toc on AIX, as it is advertised as "fixed" for
-     register allocation purposes in this case.  */
+  /* We need to mark the PIC offset register live for the same conditions
+     as it is set up, or otherwise it won't be saved before we clobber it.  */
 
-  return (((crtl->calls_eh_return || df_regs_ever_live_p (reg))
-           && (!call_used_regs[reg]
-               || (reg == RS6000_PIC_OFFSET_TABLE_REGNUM
-		   && !TARGET_SINGLE_PIC_BASE
-                   && TARGET_TOC && TARGET_MINIMAL_TOC)))
-          || (reg == RS6000_PIC_OFFSET_TABLE_REGNUM
-	      && !TARGET_SINGLE_PIC_BASE
-              && ((DEFAULT_ABI == ABI_V4 && flag_pic != 0)
-                  || (DEFAULT_ABI == ABI_DARWIN && flag_pic)
-		  || (TARGET_TOC && TARGET_MINIMAL_TOC
-		      && get_pool_size () != 0))));
+  if (reg == RS6000_PIC_OFFSET_TABLE_REGNUM && !TARGET_SINGLE_PIC_BASE)
+    {
+      if (TARGET_TOC && TARGET_MINIMAL_TOC
+	  && (crtl->calls_eh_return
+	      || df_regs_ever_live_p (reg)
+	      || get_pool_size ()))
+	return true;
+
+      if ((DEFAULT_ABI == ABI_V4 || DEFAULT_ABI == ABI_DARWIN)
+	  && flag_pic)
+	return true;
+    }
+
+  /* If the function calls eh_return, claim used all the registers that would
+     be checked for liveness otherwise.  */
+
+  return ((crtl->calls_eh_return || df_regs_ever_live_p (reg))
+	  && !call_used_regs[reg]);
 }
 
 /* Return the first fixed-point register that is required to be
-- 
1.9.3

Re: [PATCH 1/2] rs6000: Another PIC LRA fix

[David Edelsohn]

On Sat, Oct 31, 2015 at 1:39 PM, Segher Boessenkool
<segher@kernel.crashing.org> wrote:
> This one for TARGET_TOC && TARGET_MINIMAL_TOC.  Without it, r30 is not
> saved in the prologue for functions that do not use r30, but the register
> is set later in the prologue anyhow.  This made all java tests fail.
>
> Tested on powerpc64-linux, --enable-languages=all,ada,go,obj-c++ and test
> variants {-m32/-mno-lra,-m32/-mlra,-m32/-mpowerpc64,-m64/-mno-lra,-m64/-mlra};
> and on powerpc64le-linux, everything default.  Both also with bootstrapping
> with LRA defaulted on.
>
> Okay for trunk?
>
>
> Segher
>
>
> 2015-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
>
>         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Use the
>         same condition for testing whether RS6000_PIC_OFFSET_TABLE_REGNUM is
>         live as for using it elsewhere, for TARGET_MINIMAL_TOC.

Okay.

Thanks, David

Re: [PATCH 2/2] rs6000: Rewrite rs6000_reg_live_or_pic_offset_p

[David Edelsohn]

On Sat, Oct 31, 2015 at 1:39 PM, Segher Boessenkool
<segher@kernel.crashing.org> wrote:
> This function is quite a puzzle; untangle it.  No functional change.
>
> Tested etc.; okay for trunk?
>
>
> Segher
>
>
> 2015-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
>
>         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Rewrite.

Okay.

Thanks, David