Patchwork [SH] Backport PR 33135 fix to 4.7 and 4.6

login
register
mail settings
Submitter Oleg Endo
Date July 22, 2012, 3:58 p.m.
Message ID <1342972733.2373.2.camel@yam-132-YW-E178-FTW>
Download mbox | patch
Permalink /patch/172521/
State New
Headers show

Comments

Oleg Endo - July 22, 2012, 3:58 p.m.
Hello,

The attached patches fix PR 33135 on the 4.7 and 4.6 branches.
Tested on both branches with 'make all' and 'make info dvi pdf'.

OK to install?

Cheers,
Oleg

ChangeLog:

	PR target/33135
	* config/sh/sh.opt (mieee): Use Var instead of Mask.  Correct 
	description.
	* config/sh/sh.c (sh_option_override): Do not change 
	flag_finite_math_only.  Set TARGET_IEEE to complement of
	flag_finite_math_only.
	* doc/invoke.texi (SH options): Add mno-ieee.  Correct
	description of mieee and mno-ieee behavior.
Kaz Kojima - July 22, 2012, 10 p.m.
Oleg Endo <oleg.endo@t-online.de> wrote:
> The attached patches fix PR 33135 on the 4.7 and 4.6 branches.
> Tested on both branches with 'make all' and 'make info dvi pdf'.
> 
> OK to install?

OK.

Regards,
	kaz

Patch

Index: gcc/config/sh/sh.opt
===================================================================
--- gcc/config/sh/sh.opt	(revision 189755)
+++ gcc/config/sh/sh.opt	(working copy)
@@ -270,8 +270,8 @@ 
 Follow Renesas (formerly Hitachi) / SuperH calling conventions
 
 mieee
-Target Report Mask(IEEE)
-Increase the IEEE compliance for floating-point code
+Target Var(TARGET_IEEE)
+Increase the IEEE compliance for floating-point comparisons
 
 mindexed-addressing
 Target Report Mask(ALLOW_INDEXED_ADDRESS) Condition(SUPPORT_ANY_SH5_32MEDIA)
Index: gcc/config/sh/sh.c
===================================================================
--- gcc/config/sh/sh.c	(revision 189755)
+++ gcc/config/sh/sh.c	(working copy)
@@ -763,11 +763,6 @@ 
   SUBTARGET_OVERRIDE_OPTIONS;
   if (optimize > 1 && !optimize_size)
     target_flags |= MASK_SAVE_ALL_TARGET_REGS;
-  if (flag_finite_math_only == 2)
-    flag_finite_math_only
-      = !flag_signaling_nans && TARGET_SH2E && ! TARGET_IEEE;
-  if (TARGET_SH2E && !flag_finite_math_only)
-    target_flags |= MASK_IEEE;
   sh_cpu = PROCESSOR_SH1;
   assembler_dialect = 0;
   if (TARGET_SH2)
@@ -1007,6 +1002,11 @@ 
 	align_functions = min_align;
     }
 
+  /* If the -mieee option was not explicitly set by the user, turn it on
+     unless -ffinite-math-only was specified.  See also PR 33135.  */
+  if (! global_options_set.x_TARGET_IEEE)
+    TARGET_IEEE = ! flag_finite_math_only;
+
   if (sh_fixed_range_str)
     sh_fix_range (sh_fixed_range_str);
 
Index: gcc/doc/invoke.texi
===================================================================
--- gcc/doc/invoke.texi	(revision 189755)
+++ gcc/doc/invoke.texi	(working copy)
@@ -853,8 +853,8 @@ 
 -m5-compact  -m5-compact-nofpu @gol
 -mb  -ml  -mdalign  -mrelax @gol
 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
--mieee  -mbitops  -misize  -minline-ic_invalidate -mpadstruct  -mspace @gol
--mprefergot  -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
+-mieee -mno-ieee -mbitops  -misize  -minline-ic_invalidate -mpadstruct @gol
+-mspace -mprefergot  -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
 -maccumulate-outgoing-args -minvalid-symbols}
@@ -16938,13 +16938,15 @@ 
 @option{-mhitachi} is given.
 
 @item -mieee
+@item -mno-ieee
 @opindex mieee
-Increase IEEE-compliance of floating-point code.
-At the moment, this is equivalent to @option{-fno-finite-math-only}.
-When generating 16 bit SH opcodes, getting IEEE-conforming results for
-comparisons of NANs / infinities incurs extra overhead in every
-floating point comparison, therefore the default is set to
-@option{-ffinite-math-only}.
+@opindex mnoieee
+Control the IEEE compliance of floating-point comparisons, which affects the
+handling of cases where the result of a comparison is unordered.  By default
+@option{-mieee} is implicitly enabled.  If @option{-ffinite-math-only} is
+enabled @option{-mno-ieee} is implicitly set, which results in faster
+floating-point greater-equal and less-equal comparisons.  The implcit settings
+can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
 
 @item -minline-ic_invalidate
 @opindex minline-ic_invalidate