decoding of fsubrp vs fsubp mixed up

While looking at #2175 (closed) I noticed a discrepancy in DR's decoder vs gdb on fsubrp vs fsubp:

  0xf7566960  de e1                fsubrp %st0 %st1 -> %st1
  0xf7566962  c1 e2 02             shl    $0x00000002 %edx -> %edx
  0xf7566965  de ea                fsubp  %st0 %st2 -> %st2

   0xf7542960 <+48>:    fsubp  %st,%st(1)
   0xf7542962 <+50>:    shl    $0x2,%edx
   0xf7542965 <+53>:    fsubrp %st,%st(2)