1    /*
2     * Long integer shift.  This is different from the generic 32/64-bit
3     * binary operations because vAA/vBB are 64-bit but vCC (the shift
4     * distance) is 32-bit.  Also, Dalvik requires us to mask off the low
5     * 6 bits of the shift distance.
6     */
7    /* ushr-long vAA, vBB, vCC */
8    FETCH r0, 1                         @ r0<- CCBB
9    mov     r9, rINST, lsr #8           @ r9<- AA
10    and     r3, r0, #255                @ r3<- BB
11    mov     r0, r0, lsr #8              @ r0<- CC
12    VREG_INDEX_TO_ADDR r3, r3           @ r3<- &fp[BB]
13    GET_VREG r2, r0                     @ r2<- vCC
14    ldmia   r3, {r0-r1}                 @ r0/r1<- vBB/vBB+1
15    CLEAR_SHADOW_PAIR r9, lr, ip        @ Zero out the shadow regs
16    and     r2, r2, #63                 @ r0<- r0 & 0x3f
17    VREG_INDEX_TO_ADDR r9, r9           @ r9<- &fp[AA]
18    mov     r0, r0, lsr r2              @ r0<- r2 >> r2
19    rsb     r3, r2, #32                 @ r3<- 32 - r2
20    orr     r0, r0, r1, asl r3          @ r0<- r0 | (r1 << (32-r2))
21    subs    ip, r2, #32                 @ ip<- r2 - 32
22    movpl   r0, r1, lsr ip              @ if r2 >= 32, r0<-r1 >>> (r2-32)
23    FETCH_ADVANCE_INST 2                @ advance rPC, load rINST
24    mov     r1, r1, lsr r2              @ r1<- r1 >>> r2
25    GET_INST_OPCODE ip                  @ extract opcode from rINST
26    stmia   r9, {r0-r1}                 @ vAA/vAA+1<- r0/r1
27    GOTO_OPCODE ip                      @ jump to next instruction
28