See Test.java of test/compiler/6431242, the function:
  public void test_copy_longs_store_reversed(long[] dst, long[] src) {
    for(int i=0; i<src.length; i++) {
      dst[i] = Long.reverseBytes(1 + src[i]);
    }
  }

-XX:PrintOptoAssembly :
02c   B3: #     B16 B4 <- B2 B6         Loop: B3-B6 inner stride: not constant pre of N164 Freq: 13469.202c +   SLL    R
_L5,#2,R_L2
030 +   CMP    R_L5,R_L1        ! unsigned
034     BPuge  icc_U,B16  P=0.000001 C=-1.000000
034     ADD    R_I2,R_L2,R_L0
034
03c   B4: #     B15 B5 <- B3  Freq: 13469.2
03c +   LDUW   [R_L0 + #12],R_L0
040 +   ADD    R_L0,#1,R_L3
044 +   STW    R_L3,[SP + #64]  !stk
048 +   LDUWA  [SP + #64], R_L4 !asi
050     LDUW   [R_I1 + #8],R_L3 ! range
054     NullCheck R_I1
054
054   B5: #     B14 B6 <- B4  Freq: 13469.1
054 +   CMP    R_L5,R_L3        ! unsigned
058     BPuge  icc_U,B14  P=0.000001 C=-1.000000
058
060   B6: #     B3 B7 <- B5  Freq: 13469.1
060 +   ADD    R_I1,R_L2,R_L2
064 +   ADD    R_L5,#1,R_L5
068 +   CMP    R_L5,#1
06c     BPlt   ccr,B3   ! Loop end  P=0.888820 C=7131.000000
06c     STW    R_L4,[R_L2 + #12]

The problem for pre- and post- loop where generated code still spill on stack first and read back reversely. The expected result is take the addI(P) result and store byte reversely into destination.
test_copy_ints_store_reversed and test_copy_long_store_reversed have same result.