[Bug target/51337] New: SH Target: Various testsuite ICEs for -m2a -O0

[oleg.endo@t-online.de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51337

             Bug #: 51337
           Summary: SH Target: Various testsuite ICEs for -m2a -O0
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: oleg.endo@t-online.de
                CC: kkojima@gcc.gnu.org
            Target: sh2a-*-*


There are various testsuite ICEs for -m2a -mb -O0, which look similar to the
following:

gcc.c-torture/compile/20000923-1.c:

...: error: insn does not satisfy its constraints:
(insn 142 34 35 (set (mem/c:SI (plus:SI (reg/f:SI 14 r14)
                (const_int 36 [0x24])) [0 %sfp+-16 S4 A32])
        (reg:SI 150 fpul)) ... {movsi_ie}
     (nil))
...: internal compiler error: in extract_constrain_insn_cached, at recog.c:2052

The problem seems to be the movsi_ie insn, for which reload thinks it is
possible to load the FPUL register from a stack allocated reg using SImode
displacement addressing, which is actually not possible.

The following patch seems to help the issues...

--- gcc/config/sh/sh.c.orig    2011-11-28 10:03:04.000000000 +0900
+++ gcc/config/sh/sh.c    2011-11-28 15:09:01.000000000 +0900
@@ -12432,6 +12432,10 @@ sh_secondary_reload (bool in_p, rtx x, r
   if (rclass != GENERAL_REGS && REG_P (x)
       && TARGET_REGISTER_P (REGNO (x)))
     return GENERAL_REGS;
+  /* If here fall back to loading FPUL register through general regs. 
+     Happens when FPUL has to be loaded from a reg allocated on the stack.  */
+  if (rclass == FPUL_REGS && !REG_P (x))
+    return GENERAL_REGS;
   return NO_REGS;
 }

However, I think the following should be a better way to check the condition in
question...

+  if (rclass == FPUL_REGS && true_regnum (x) == -1)
+    return GENERAL_REGS;

I'll submit a patch once testing is completed.