[RS6000] Fix ICE caused by rs6000_savres_strategy thinko

[Alan Modra]

On Tue, May 03, 2016 at 07:03:45AM -0500, Segher Boessenkool wrote:
> On Tue, May 03, 2016 at 09:02:26PM +0930, Alan Modra wrote:
> > Incidentally, the added testcase function shows a regression in -m32
> > -Os code quality that I'll fix sometime soon.
> 
> Could you tell more please?

This
000012dc <s_r31>:
    12dc:       94 21 ff b0     stwu    r1,-80(r1)
    12e0:       7c 08 02 a6     mflr    r0
    12e4:       93 e1 00 4c     stw     r31,76(r1)
    12e8:       90 01 00 54     stw     r0,84(r1)
    12ec:       3b e0 00 00     li      r31,0
    12f0:       39 61 00 50     addi    r11,r1,80
    12f4:       48 00 00 00     b       12f4 <s_r31+0x18>
                        12f4: R_PPC_REL24       _restgpr_31_x

vs. this from gcc-4.4.7
0000036c <s_r31>:
 36c:   94 21 ff c0     stwu    r1,-64(r1)
 370:   93 e1 00 3c     stw     r31,60(r1)
 374:   3b e0 00 00     li      r31,0
 378:   83 e1 00 3c     lwz     r31,60(r1)
 37c:   38 21 00 40     addi    r1,r1,64
 380:   4e 80 00 20     blr


-- 
Alan Modra
Australia Development Lab, IBM