ART世界探险(17) - 中层中间代码MIR
Dalvik字节码的指令格式
指令格式分类
Dalvik指令，根据需要的寄存器数目的不同，长度也有所不同。
如下面的结构所示，有下面的这些情况：
enum Format {
k10x, // op
k12x, // op vA, vB
k11n, // op vA, #+B
k11x, // op vAA
k10t, // op +AA
k20t, // op +AAAA
k22x, // op vAA, vBBBB
k21t, // op vAA, +BBBB
k21s, // op vAA, #+BBBB
k21h, // op vAA, #+BBBB00000[00000000]
k21c, // op vAA, thing@BBBB
k23x, // op vAA, vBB, vCC
k22b, // op vAA, vBB, #+CC
k22t, // op vA, vB, +CCCC
k22s, // op vA, vB, #+CCCC
k22c, // op vA, vB, thing@CCCC
k32x, // op vAAAA, vBBBB
k30t, // op +AAAAAAAA
k31t, // op vAA, +BBBBBBBB
k31i, // op vAA, #+BBBBBBBB
k31c, // op vAA, thing@BBBBBBBB
k35c, // op {vC, vD, vE, vF, vG}, thing@BBBB (B: count, A: vG)
k3rc, // op {vCCCC .. v(CCCC+AA-1)}, meth@BBBB
k51l, // op vAA, #+BBBBBBBBBBBBBBBB
};
计算指令格式长度
MIRGraph的ParseInsn方法就是用来计算指令长度的。
/*
* Parse an instruction, return the length of the instruction
*/
int MIRGraph::ParseInsn(const uint16_t* code_ptr, MIR::DecodedInstruction* decoded_instruction) {
const Instruction* inst = Instruction::At(code_ptr);
decoded_instruction->opcode = inst->Opcode();
decoded_instruction->vA = inst->HasVRegA() ? inst->VRegA() : 0;
decoded_instruction->vB = inst->HasVRegB() ? inst->VRegB() : 0;
decoded_instruction->vB_wide = inst->HasWideVRegB() ? inst->WideVRegB() : 0;
decoded_instruction->vC = inst->HasVRegC() ? inst->VRegC() : 0;
if (inst->HasVarArgs()) {
inst->GetVarArgs(decoded_instruction->arg);
}
return inst->SizeInCodeUnits();
}
MIR中使用的Dalvik指令集
我们之前用了8讲的内容专门讲指令。
下面我们将这些指令的中间细节列一下，大家从下面的表中可以查到每一条指令的格式。
const uint64_t MIRGraph::oat_data_flow_attributes_[kMirOpLast] = {
// 00 NOP
DF_NOP,
// 01 MOVE vA, vB
DF_DA | DF_UB | DF_IS_MOVE,
// 02 MOVE_FROM16 vAA, vBBBB
DF_DA | DF_UB | DF_IS_MOVE,
// 03 MOVE_16 vAAAA, vBBBB
DF_DA | DF_UB | DF_IS_MOVE,
// 04 MOVE_WIDE vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_IS_MOVE,
// 05 MOVE_WIDE_FROM16 vAA, vBBBB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_IS_MOVE,
// 06 MOVE_WIDE_16 vAAAA, vBBBB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_IS_MOVE,
// 07 MOVE_OBJECT vA, vB
DF_DA | DF_UB | DF_NULL_TRANSFER_0 | DF_IS_MOVE | DF_REF_A | DF_REF_B,
// 08 MOVE_OBJECT_FROM16 vAA, vBBBB
DF_DA | DF_UB | DF_NULL_TRANSFER_0 | DF_IS_MOVE | DF_REF_A | DF_REF_B,
// 09 MOVE_OBJECT_16 vAAAA, vBBBB
DF_DA | DF_UB | DF_NULL_TRANSFER_0 | DF_IS_MOVE | DF_REF_A | DF_REF_B,
// 0A MOVE_RESULT vAA
DF_DA,
// 0B MOVE_RESULT_WIDE vAA
DF_DA | DF_A_WIDE,
// 0C MOVE_RESULT_OBJECT vAA
DF_DA | DF_REF_A,
// 0D MOVE_EXCEPTION vAA
DF_DA | DF_REF_A | DF_NON_NULL_DST,
// 0E RETURN_VOID
DF_NOP,
// 0F RETURN vAA
DF_UA,
// 10 RETURN_WIDE vAA
DF_UA | DF_A_WIDE,
// 11 RETURN_OBJECT vAA
DF_UA | DF_REF_A,
// 12 CONST_4 vA, #+B
DF_DA | DF_SETS_CONST,
// 13 CONST_16 vAA, #+BBBB
DF_DA | DF_SETS_CONST,
// 14 CONST vAA, #+BBBBBBBB
DF_DA | DF_SETS_CONST,
// 15 CONST_HIGH16 VAA, #+BBBB0000
DF_DA | DF_SETS_CONST,
// 16 CONST_WIDE_16 vAA, #+BBBB
DF_DA | DF_A_WIDE | DF_SETS_CONST,
// 17 CONST_WIDE_32 vAA, #+BBBBBBBB
DF_DA | DF_A_WIDE | DF_SETS_CONST,
// 18 CONST_WIDE vAA, #+BBBBBBBBBBBBBBBB
DF_DA | DF_A_WIDE | DF_SETS_CONST,
// 19 CONST_WIDE_HIGH16 vAA, #+BBBB000000000000
DF_DA | DF_A_WIDE | DF_SETS_CONST,
// 1A CONST_STRING vAA, string@BBBB
DF_DA | DF_REF_A | DF_NON_NULL_DST,
// 1B CONST_STRING_JUMBO vAA, string@BBBBBBBB
DF_DA | DF_REF_A | DF_NON_NULL_DST,
// 1C CONST_CLASS vAA, type@BBBB
DF_DA | DF_REF_A | DF_NON_NULL_DST,
// 1D MONITOR_ENTER vAA
DF_UA | DF_NULL_CHK_A | DF_REF_A,
// 1E MONITOR_EXIT vAA
DF_UA | DF_NULL_CHK_A | DF_REF_A,
// 1F CHK_CAST vAA, type@BBBB
DF_UA | DF_REF_A | DF_CHK_CAST | DF_UMS,
// 20 INSTANCE_OF vA, vB, type@CCCC
DF_DA | DF_UB | DF_CORE_A | DF_REF_B | DF_UMS,
// 21 ARRAY_LENGTH vA, vB
DF_DA | DF_UB | DF_NULL_CHK_B | DF_CORE_A | DF_REF_B,
// 22 NEW_INSTANCE vAA, type@BBBB
DF_DA | DF_NON_NULL_DST | DF_REF_A | DF_UMS,
// 23 NEW_ARRAY vA, vB, type@CCCC
DF_DA | DF_UB | DF_NON_NULL_DST | DF_REF_A | DF_CORE_B | DF_UMS,
// 24 FILLED_NEW_ARRAY {vD, vE, vF, vG, vA}
DF_FORMAT_35C | DF_NON_NULL_RET | DF_UMS,
// 25 FILLED_NEW_ARRAY_RANGE {vCCCC .. vNNNN}, type@BBBB
DF_FORMAT_3RC | DF_NON_NULL_RET | DF_UMS,
// 26 FILL_ARRAY_DATA vAA, +BBBBBBBB
DF_UA | DF_REF_A | DF_UMS,
// 27 THROW vAA
DF_UA | DF_REF_A | DF_UMS,
// 28 GOTO
DF_NOP,
// 29 GOTO_16
DF_NOP,
// 2A GOTO_32
DF_NOP,
// 2B PACKED_SWITCH vAA, +BBBBBBBB
DF_UA | DF_CORE_A,
// 2C SPARSE_SWITCH vAA, +BBBBBBBB
DF_UA | DF_CORE_A,
// 2D CMPL_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_B | DF_FP_C | DF_CORE_A,
// 2E CMPG_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_B | DF_FP_C | DF_CORE_A,
// 2F CMPL_DOUBLE vAA, vBB, vCC
DF_DA | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_B | DF_FP_C | DF_CORE_A,
// 30 CMPG_DOUBLE vAA, vBB, vCC
DF_DA | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_B | DF_FP_C | DF_CORE_A,
// 31 CMP_LONG vAA, vBB, vCC
DF_DA | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 32 IF_EQ vA, vB, +CCCC
DF_UA | DF_UB | DF_SAME_TYPE_AB,
// 33 IF_NE vA, vB, +CCCC
DF_UA | DF_UB | DF_SAME_TYPE_AB,
// 34 IF_LT vA, vB, +CCCC
DF_UA | DF_UB | DF_SAME_TYPE_AB,
// 35 IF_GE vA, vB, +CCCC
DF_UA | DF_UB | DF_SAME_TYPE_AB,
// 36 IF_GT vA, vB, +CCCC
DF_UA | DF_UB | DF_SAME_TYPE_AB,
// 37 IF_LE vA, vB, +CCCC
DF_UA | DF_UB | DF_SAME_TYPE_AB,
// 38 IF_EQZ vAA, +BBBB
DF_UA,
// 39 IF_NEZ vAA, +BBBB
DF_UA,
// 3A IF_LTZ vAA, +BBBB
DF_UA,
// 3B IF_GEZ vAA, +BBBB
DF_UA,
// 3C IF_GTZ vAA, +BBBB
DF_UA,
// 3D IF_LEZ vAA, +BBBB
DF_UA,
// 3E UNUSED_3E
DF_NOP,
// 3F UNUSED_3F
DF_NOP,
// 40 UNUSED_40
DF_NOP,
// 41 UNUSED_41
DF_NOP,
// 42 UNUSED_42
DF_NOP,
// 43 UNUSED_43
DF_NOP,
// 44 AGET vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 45 AGET_WIDE vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 46 AGET_OBJECT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_A | DF_REF_B | DF_CORE_C | DF_LVN,
// 47 AGET_BOOLEAN vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 48 AGET_BYTE vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 49 AGET_CHAR vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 4A AGET_SHORT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 4B APUT vAA, vBB, vCC
DF_UA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 4C APUT_WIDE vAA, vBB, vCC
DF_UA | DF_A_WIDE | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 4D APUT_OBJECT vAA, vBB, vCC
DF_UA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_A | DF_REF_B | DF_CORE_C | DF_LVN,
// 4E APUT_BOOLEAN vAA, vBB, vCC
DF_UA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 4F APUT_BYTE vAA, vBB, vCC
DF_UA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 50 APUT_CHAR vAA, vBB, vCC
DF_UA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 51 APUT_SHORT vAA, vBB, vCC
DF_UA | DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 52 IGET vA, vB, field@CCCC
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 53 IGET_WIDE vA, vB, field@CCCC
DF_DA | DF_A_WIDE | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 54 IGET_OBJECT vA, vB, field@CCCC
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_A | DF_REF_B | DF_IFIELD | DF_LVN,
// 55 IGET_BOOLEAN vA, vB, field@CCCC
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 56 IGET_BYTE vA, vB, field@CCCC
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 57 IGET_CHAR vA, vB, field@CCCC
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 58 IGET_SHORT vA, vB, field@CCCC
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 59 IPUT vA, vB, field@CCCC
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 5A IPUT_WIDE vA, vB, field@CCCC
DF_UA | DF_A_WIDE | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 5B IPUT_OBJECT vA, vB, field@CCCC
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_A | DF_REF_B | DF_IFIELD | DF_LVN,
// 5C IPUT_BOOLEAN vA, vB, field@CCCC
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 5D IPUT_BYTE vA, vB, field@CCCC
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 5E IPUT_CHAR vA, vB, field@CCCC
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 5F IPUT_SHORT vA, vB, field@CCCC
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// 60 SGET vAA, field@BBBB
DF_DA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 61 SGET_WIDE vAA, field@BBBB
DF_DA | DF_A_WIDE | DF_SFIELD | DF_CLINIT | DF_UMS,
// 62 SGET_OBJECT vAA, field@BBBB
DF_DA | DF_REF_A | DF_SFIELD | DF_CLINIT | DF_UMS,
// 63 SGET_BOOLEAN vAA, field@BBBB
DF_DA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 64 SGET_BYTE vAA, field@BBBB
DF_DA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 65 SGET_CHAR vAA, field@BBBB
DF_DA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 66 SGET_SHORT vAA, field@BBBB
DF_DA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 67 SPUT vAA, field@BBBB
DF_UA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 68 SPUT_WIDE vAA, field@BBBB
DF_UA | DF_A_WIDE | DF_SFIELD | DF_CLINIT | DF_UMS,
// 69 SPUT_OBJECT vAA, field@BBBB
DF_UA | DF_REF_A | DF_SFIELD | DF_CLINIT | DF_UMS,
// 6A SPUT_BOOLEAN vAA, field@BBBB
DF_UA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 6B SPUT_BYTE vAA, field@BBBB
DF_UA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 6C SPUT_CHAR vAA, field@BBBB
DF_UA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 6D SPUT_SHORT vAA, field@BBBB
DF_UA | DF_SFIELD | DF_CLINIT | DF_UMS,
// 6E INVOKE_VIRTUAL {vD, vE, vF, vG, vA}
DF_FORMAT_35C | DF_NULL_CHK_OUT0 | DF_UMS,
// 6F INVOKE_SUPER {vD, vE, vF, vG, vA}
DF_FORMAT_35C | DF_NULL_CHK_OUT0 | DF_UMS,
// 70 INVOKE_DIRECT {vD, vE, vF, vG, vA}
DF_FORMAT_35C | DF_NULL_CHK_OUT0 | DF_UMS,
// 71 INVOKE_STATIC {vD, vE, vF, vG, vA}
DF_FORMAT_35C | DF_CLINIT | DF_UMS,
// 72 INVOKE_INTERFACE {vD, vE, vF, vG, vA}
DF_FORMAT_35C | DF_NULL_CHK_OUT0 | DF_UMS,
// 73 RETURN_VOID_NO_BARRIER
DF_NOP,
// 74 INVOKE_VIRTUAL_RANGE {vCCCC .. vNNNN}
DF_FORMAT_3RC | DF_NULL_CHK_OUT0 | DF_UMS,
// 75 INVOKE_SUPER_RANGE {vCCCC .. vNNNN}
DF_FORMAT_3RC | DF_NULL_CHK_OUT0 | DF_UMS,
// 76 INVOKE_DIRECT_RANGE {vCCCC .. vNNNN}
DF_FORMAT_3RC | DF_NULL_CHK_OUT0 | DF_UMS,
// 77 INVOKE_STATIC_RANGE {vCCCC .. vNNNN}
DF_FORMAT_3RC | DF_CLINIT | DF_UMS,
// 78 INVOKE_INTERFACE_RANGE {vCCCC .. vNNNN}
DF_FORMAT_3RC | DF_NULL_CHK_OUT0 | DF_UMS,
// 79 UNUSED_79
DF_NOP,
// 7A UNUSED_7A
DF_NOP,
// 7B NEG_INT vA, vB
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// 7C NOT_INT vA, vB
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// 7D NEG_LONG vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// 7E NOT_LONG vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// 7F NEG_FLOAT vA, vB
DF_DA | DF_UB | DF_FP_A | DF_FP_B,
// 80 NEG_DOUBLE vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// 81 INT_TO_LONG vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_CORE_A | DF_CORE_B,
// 82 INT_TO_FLOAT vA, vB
DF_DA | DF_UB | DF_FP_A | DF_CORE_B,
// 83 INT_TO_DOUBLE vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_FP_A | DF_CORE_B,
// 84 LONG_TO_INT vA, vB
DF_DA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// 85 LONG_TO_FLOAT vA, vB
DF_DA | DF_UB | DF_B_WIDE | DF_FP_A | DF_CORE_B,
// 86 LONG_TO_DOUBLE vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_FP_A | DF_CORE_B,
// 87 FLOAT_TO_INT vA, vB
DF_DA | DF_UB | DF_FP_B | DF_CORE_A,
// 88 FLOAT_TO_LONG vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_FP_B | DF_CORE_A,
// 89 FLOAT_TO_DOUBLE vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_FP_A | DF_FP_B,
// 8A DOUBLE_TO_INT vA, vB
DF_DA | DF_UB | DF_B_WIDE | DF_FP_B | DF_CORE_A,
// 8B DOUBLE_TO_LONG vA, vB
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_FP_B | DF_CORE_A,
// 8C DOUBLE_TO_FLOAT vA, vB
DF_DA | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// 8D INT_TO_BYTE vA, vB
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// 8E INT_TO_CHAR vA, vB
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// 8F INT_TO_SHORT vA, vB
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// 90 ADD_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 91 SUB_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 92 MUL_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 93 DIV_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 94 REM_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 95 AND_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 96 OR_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 97 XOR_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 98 SHL_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 99 SHR_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 9A USHR_INT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 9B ADD_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 9C SUB_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 9D MUL_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 9E DIV_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// 9F REM_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A0 AND_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A1 OR_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A2 XOR_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A3 SHL_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A4 SHR_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A5 USHR_LONG vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_CORE_A | DF_CORE_B | DF_CORE_C,
// A6 ADD_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_A | DF_FP_B | DF_FP_C,
// A7 SUB_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_A | DF_FP_B | DF_FP_C,
// A8 MUL_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_A | DF_FP_B | DF_FP_C,
// A9 DIV_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_A | DF_FP_B | DF_FP_C,
// AA REM_FLOAT vAA, vBB, vCC
DF_DA | DF_UB | DF_UC | DF_FP_A | DF_FP_B | DF_FP_C,
// AB ADD_DOUBLE vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_A | DF_FP_B | DF_FP_C,
// AC SUB_DOUBLE vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_A | DF_FP_B | DF_FP_C,
// AD MUL_DOUBLE vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_A | DF_FP_B | DF_FP_C,
// AE DIV_DOUBLE vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_A | DF_FP_B | DF_FP_C,
// AF REM_DOUBLE vAA, vBB, vCC
DF_DA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_UC | DF_C_WIDE | DF_FP_A | DF_FP_B | DF_FP_C,
// B0 ADD_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B1 SUB_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B2 MUL_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B3 DIV_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B4 REM_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B5 AND_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B6 OR_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B7 XOR_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B8 SHL_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// B9 SHR_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// BA USHR_INT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// BB ADD_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// BC SUB_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// BD MUL_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// BE DIV_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// BF REM_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// C0 AND_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// C1 OR_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// C2 XOR_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// C3 SHL_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// C4 SHR_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// C5 USHR_LONG_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_CORE_A | DF_CORE_B,
// C6 ADD_FLOAT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// C7 SUB_FLOAT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// C8 MUL_FLOAT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// C9 DIV_FLOAT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// CA REM_FLOAT_2ADDR vA, vB
DF_DA | DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// CB ADD_DOUBLE_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// CC SUB_DOUBLE_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// CD MUL_DOUBLE_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// CE DIV_DOUBLE_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// CF REM_DOUBLE_2ADDR vA, vB
DF_DA | DF_A_WIDE | DF_UA | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// D0 ADD_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D1 RSUB_INT vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D2 MUL_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D3 DIV_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D4 REM_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D5 AND_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D6 OR_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D7 XOR_INT_LIT16 vA, vB, #+CCCC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D8 ADD_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// D9 RSUB_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// DA MUL_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// DB DIV_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// DC REM_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// DD AND_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// DE OR_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// DF XOR_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// E0 SHL_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// E1 SHR_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// E2 USHR_INT_LIT8 vAA, vBB, #+CC
DF_DA | DF_UB | DF_CORE_A | DF_CORE_B,
// E3 IGET_QUICK
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// E4 IGET_WIDE_QUICK
DF_DA | DF_A_WIDE | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// E5 IGET_OBJECT_QUICK
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_A | DF_REF_B | DF_IFIELD | DF_LVN,
// E6 IPUT_QUICK
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// E7 IPUT_WIDE_QUICK
DF_UA | DF_A_WIDE | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// E8 IPUT_OBJECT_QUICK
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_A | DF_REF_B | DF_IFIELD | DF_LVN,
// E9 INVOKE_VIRTUAL_QUICK
DF_FORMAT_35C | DF_NULL_CHK_OUT0 | DF_UMS,
// EA INVOKE_VIRTUAL_RANGE_QUICK
DF_FORMAT_3RC | DF_NULL_CHK_OUT0 | DF_UMS,
// EB IPUT_BOOLEAN_QUICK vA, vB, index
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// EC IPUT_BYTE_QUICK vA, vB, index
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// ED IPUT_CHAR_QUICK vA, vB, index
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// EE IPUT_SHORT_QUICK vA, vB, index
DF_UA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// EF IGET_BOOLEAN_QUICK vA, vB, index
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// F0 IGET_BYTE_QUICK vA, vB, index
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// F1 IGET_CHAR_QUICK vA, vB, index
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// F2 IGET_SHORT_QUICK vA, vB, index
DF_DA | DF_UB | DF_NULL_CHK_B | DF_REF_B | DF_IFIELD | DF_LVN,
// F3 UNUSED_F3
DF_NOP,
// F4 UNUSED_F4
DF_NOP,
// F5 UNUSED_F5
DF_NOP,
// F6 UNUSED_F6
DF_NOP,
// F7 UNUSED_F7
DF_NOP,
// F8 UNUSED_F8
DF_NOP,
// F9 UNUSED_F9
DF_NOP,
// FA UNUSED_FA
DF_NOP,
// FB UNUSED_FB
DF_NOP,
// FC UNUSED_FC
DF_NOP,
// FD UNUSED_FD
DF_NOP,
// FE UNUSED_FE
DF_NOP,
// FF UNUSED_FF
DF_NOP,
// Beginning of extended MIR opcodes
// 100 MIR_PHI
DF_DA | DF_NULL_TRANSFER_N,
// 101 MIR_COPY
DF_DA | DF_UB | DF_IS_MOVE,
// 102 MIR_FUSED_CMPL_FLOAT
DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// 103 MIR_FUSED_CMPG_FLOAT
DF_UA | DF_UB | DF_FP_A | DF_FP_B,
// 104 MIR_FUSED_CMPL_DOUBLE
DF_UA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// 105 MIR_FUSED_CMPG_DOUBLE
DF_UA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_FP_A | DF_FP_B,
// 106 MIR_FUSED_CMP_LONG
DF_UA | DF_A_WIDE | DF_UB | DF_B_WIDE | DF_CORE_A | DF_CORE_B,
// 107 MIR_NOP
DF_NOP,
// 108 MIR_NULL_CHECK
DF_UA | DF_REF_A | DF_NULL_CHK_A | DF_LVN,
// 109 MIR_RANGE_CHECK
0,
// 10A MIR_DIV_ZERO_CHECK
0,
// 10B MIR_CHECK
0,
// 10D MIR_SELECT
DF_DA | DF_UB,
// 10E MirOpConstVector
0,
// 10F MirOpMoveVector
0,
// 110 MirOpPackedMultiply
0,
// 111 MirOpPackedAddition
0,
// 112 MirOpPackedSubtract
0,
// 113 MirOpPackedShiftLeft
0,
// 114 MirOpPackedSignedShiftRight
0,
// 115 MirOpPackedUnsignedShiftRight
0,
// 116 MirOpPackedAnd
0,
// 117 MirOpPackedOr
0,
// 118 MirOpPackedXor
0,
// 119 MirOpPackedAddReduce
DF_FORMAT_EXTENDED,
// 11A MirOpPackedReduce
DF_FORMAT_EXTENDED,
// 11B MirOpPackedSet
DF_FORMAT_EXTENDED,
// 11C MirOpReserveVectorRegisters
0,
// 11D MirOpReturnVectorRegisters
0,
// 11E MirOpMemBarrier
0,
// 11F MirOpPackedArrayGet
DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 120 MirOpPackedArrayPut
DF_UB | DF_UC | DF_NULL_CHK_B | DF_RANGE_CHK_C | DF_REF_B | DF_CORE_C | DF_LVN,
// 121 MirOpMaddInt
DF_FORMAT_EXTENDED,
// 122 MirOpMsubInt
DF_FORMAT_EXTENDED,
// 123 MirOpMaddLong
DF_FORMAT_EXTENDED,
// 124 MirOpMsubLong
DF_FORMAT_EXTENDED,
};
MIR
生成一条MIR
// Allocate a new MIR.
MIR* MIRGraph::NewMIR() {
MIR* mir = new (arena_) MIR();
return mir;
}
代码块 - BasicBlock
我们都知道，Java中的代码是由一个个代码块所组成的。
代码块的类型 - BBType
enum BBType {
kNullBlock,
kEntryBlock,
kDalvikByteCode,
kExitBlock,
kExceptionHandling,
kDead,
};
代码块列表类型
enum BlockListType {
kNotUsed = 0,
kCatch,
kPackedSwitch,
kSparseSwitch,
};
基本块的数据流
// Dataflow attributes of a basic block.
struct BasicBlockDataFlow {
ArenaBitVector* use_v;
ArenaBitVector* def_v;
ArenaBitVector* live_in_v;
int32_t* vreg_to_ssa_map_exit;
};
基本代码块的创建 - CreateNewBB方法
- 首先根据block_list_中已有的基本代码块的数目生成新的BasicBlockId。
- 然后调用NewMemBB去真正分配空间。
- 生成之后，再压到block_list_向量中。
BasicBlock* MIRGraph::CreateNewBB(BBType block_type) {
BasicBlockId id = static_cast(block_list_.size());
BasicBlock* res = NewMemBB(block_type, id);
block_list_.push_back(res);
return res;
}
分配一个基本代码块
主要是预留出相应的空间来。
另外还要到block_id_map_中备个案。
// Allocate a new basic block.
BasicBlock* MIRGraph::NewMemBB(BBType block_type, int block_id) {
BasicBlock* bb = new (arena_) BasicBlock(block_id, block_type, arena_);
// TUNING: better estimate of the exit block predecessors?
bb->predecessors.reserve((block_type == kExitBlock) ? 2048 : 2);
block_id_map_.Put(block_id, block_id);
return bb;
}
代码项
Dex文件中的代码项，用CodeItem结构体来表示：
struct CodeItem {
uint16_t registers_size_;
uint16_t ins_size_;
uint16_t outs_size_;
uint16_t tries_size_;
uint32_t debug_info_off_; // file offset to debug info stream
uint32_t insns_size_in_code_units_; // size of the insns array, in 2 byte code units
uint16_t insns_[1];
private:
DISALLOW_COPY_AND_ASSIGN(CodeItem);
};
调用类型
调用方法的类型有以下几种：
- kStatic：调用静态方法
- kDirect:调用普通方法
- kVirtual：调用虚方法
- kSuper：调用父类方法
- kInterface：调用实现的接口中的方法
enum InvokeType {
kStatic, // <>
kDirect, // <>
kVirtual, // <>
kSuper, // <>
kInterface, // <>
kMaxInvokeType = kInterface
};
最后，我们上一张MIRGraph的大图. 后面几节我们主要都是跟它打交道了
MIRGraph