[Bug rtl-optimization/71118] New: [5 Regression] ftois instruction not emitted for float -> int bitcast

[mattst88 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71118

            Bug ID: 71118
           Summary: [5 Regression] ftois instruction not emitted for float
                    -> int bitcast
           Product: gcc
           Version: 5.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mattst88 at gmail dot com
  Target Milestone: ---

Created attachment 38490
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38490&action=edit
ftois.c

For the attached ftois.c, gcc-4.9.3 -O2 -mcpu=ev67 emits

0000000000000000 <f2i>:
   0:   01 0f 1f 72     ftois   $f16,t0
   4:   00 00 e1 43     sextl   t0,v0
   8:   01 80 fa 6b     ret
   c:   00 00 fe 2f     unop

0000000000000010 <f2l>:
  10:   01 0f 1f 72     ftois   $f16,t0
  14:   20 f6 21 48     zapnot  t0,0xf,v0
  18:   01 80 fa 6b     ret
  1c:   00 00 fe 2f     unop


while gcc-5.3.0 -O2 -mcpu=ev67 emits

0000000000000000 <f2i>:
   0:   f0 ff de 23     lda     sp,-16(sp)
   4:   00 00 1e 9a     sts     $f16,0(sp)
   8:   00 00 1e a0     ldl     v0,0(sp)
   c:   10 00 de 23     lda     sp,16(sp)
  10:   01 80 fa 6b     ret
  14:   00 00 fe 2f     unop
  18:   1f 04 ff 47     nop
  1c:   00 00 fe 2f     unop

0000000000000020 <f2l>:
  20:   01 0f 1f 72     ftois   $f16,t0
  24:   20 f6 21 48     zapnot  t0,0xf,v0
  28:   01 80 fa 6b     ret
  2c:   00 00 fe 2f     unop

In fact, the alpha architecture reference says that ftois is exactly equivalent
to an sts/ldl sequence.

f2i should have used ftois, as with gcc-4.9.3.