[Bug rtl-optimization/71119] New: [4.9 Regression] ftoit instruction not emitted for double -> long bitcast

[mattst88 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71119

            Bug ID: 71119
           Summary: [4.9 Regression] ftoit instruction not emitted for
                    double -> long bitcast
           Product: gcc
           Version: 4.9.3
            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 38491
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38491&action=edit
ftoit.c

For the attached ftoit.c, gcc-4.8.5 -O2 -mcpu=ev67 emits

0000000000000000 <f2i>:
   0:   00 0e 1f 72     ftoit   $f16,v0
   4:   00 00 e0 43     sextl   v0,v0
   8:   01 80 fa 6b     ret
   c:   00 00 fe 2f     unop

0000000000000010 <f2l>:
  10:   00 0e 1f 72     ftoit   $f16,v0
  14:   01 80 fa 6b     ret
  18:   1f 04 ff 47     nop
  1c:   00 00 fe 2f     unop

while gcc-4.9.3/gcc-5.3.0 -O2 -mcpu=ev67 emits

0000000000000000 <f2i>:
   0:   00 0e 1f 72     ftoit   $f16,v0
   4:   00 00 e0 43     sextl   v0,v0
   8:   01 80 fa 6b     ret
   c:   00 00 fe 2f     unop

0000000000000010 <f2l>:
  10:   f0 ff de 23     lda     sp,-16(sp)
  14:   00 00 1e 9e     stt     $f16,0(sp)
  18:   00 00 1e a4     ldq     v0,0(sp)
  1c:   10 00 de 23     lda     sp,16(sp)
  20:   01 80 fa 6b     ret
  24:   00 00 fe 2f     unop
  28:   1f 04 ff 47     nop
  2c:   00 00 fe 2f     unop

In fact, the alpha architecture reference says that ftoit is exactly equivalent
to an stt/ldq sequence.

f2l should have used ftoit, as with gcc-4.8.5.