2012年11月份的资料,之前ARMv8手冊还没公布,我想办法从编译器的binutils中提取出了全部ARMv8指令的二进制编码,之前不能随便发,如今相当于解禁了^_^。
问题1:提取ARMv8的指令编码
答:ARMv8指令的opcode能够在支持ARMv8的编译器binutils中找到,该工具能在网上找到,见參考资料[1],或者在[1]中搜索aarch64_opcode_table。
为了说明指令编码细节。现将网页中部分内容摘录例如以下:
1208
struct aarch64_opcode aarch64_opcode_table[] =
1209 {
1210 /* Add/subtract (with carry). */
1211 {"adc", 0x1a000000, 0x7fe0fc00, addsub_carry, 0, CORE, OP3 (Rd, Rn, Rm), QL_I3SAMER, F_SF},
p=jk/arm64/binutils.git;a=blob;f=opcodes/aarch64-tbl.h;h=d360b1406718257da86050f5b3a760cd02196250;hb=aarch64#l1212">1212
以上结构体中的每一项相应某条指令的操作码等信息。当中aarch64_opcode结构体(见參考文献[2])在。内容摘录例如以下:
structaarch64_opcode
{
/*The name of the mnemonic. */
const char *name;
/*The opcode itself. Those bits which willbe filled in with operands are zeroes. */
aarch64_insn opcode;
/*The opcode mask. This is used by thedisassembler. This is a mask containingones indicating those bits which must match the opcode field, and zeroesindicating those bits which need not match (and are presumably filled in byoperands). */
aarch64_insn mask;
/*Instruction class.
*/
enum aarch64_insn_class iclass;
/*Enumerator identifier. */
enum aarch64_op op;
/*Which architecture variant provides this instruction. */
const aarch64_feature_set *avariant;
/*An array of operand codes. Each code isan index into the operand table. Theyappear in the order which the operands must appear in assembly code, and are terminatedby a zero. */
enum aarch64_opnd operands[AARCH64_MAX_OPND_NUM];
/*A list of operand qualifier code sequence. Each operand qualifier code qualifies the corresponding operandcode. Each operand qualifier sequencespecifies a valid opcode variant and related constraint on operands. */
aarch64_opnd_qualifier_seq_t qualifiers_list[AARCH64_MAX_QLF_SEQ_NUM];
/*Flags providing information about this instruction */
uint32_t flags;
}
依照以上信息解释aarch64_opcode_table中的结构体例如以下表:
"adc", |
0x1a000000 |
0x7fe0fc00 |
addsub_carry, |
0 |
CORE |
OP3 (Rd, Rn, Rm) |
QL_I3SAMER, |
F_SF |
助记符 |
操作码 |
操作码掩码 |
指令所属类 |
枚举器标识符 |
提供这样的指令的结构体变量 |
操作数编码数组 |
操作数限定符的代码序列列表 |
指令信息标志位 |
1:Aarch64-opc.c中有关于指令位域的定义。
const aarch64_field fields[] =
{
{ 0, 0 }, /*NIL. */
{ 0, 4 }, /*cond2: condition in truly conditional-executed inst. */
{ 0, 4 }, /*nzcv: flag bit specifier, encoded in the "nzcv" field. */
{ 5, 5 }, /*defgh: d:e:f:g:h bits in AdvSIMD modified immediate. */
{16, 3 }, /*abc: a:b:c bits in AdvSIMD modified immediate. */
{ 5, 19 }, /* imm19: e.g. in CBZ. */
{ 5, 19 }, /* immhi: e.g. in ADRP. */
{29, 2 }, /*immlo: e.g. in ADRP. */
{22, 2 }, /*size: in most AdvSIMD and floating-point instructions. */
{10, 2 }, /*vldst_size: size field in the AdvSIMD load/store inst. */
{29, 1 }, /*op: in AdvSIMD modified immediate instructions. */
{30, 1 }, /*Q: in most AdvSIMD instructions. */
{ 0, 5 }, /*Rt: in load/store instructions. */
{ 0, 5 }, /*Rd: in many integer instructions. */
{ 5, 5 }, /*Rn: in many integer instructions. */
{10, 5 }, /*Rt2: in load/store pair instructions. */
{10, 5 }, /*Ra: in fp instructions. */
{ 5, 3 }, /*op2: in the system instructions. */
{ 8, 4 }, /*CRm: in the system instructions. */
{12, 4 }, /*CRn: in the system instructions. */
{16, 3 }, /*op1: in the system instructions. */
{19, 2 }, /*op0: in the system instructions. */
{10, 3 }, /*imm3: in add/sub extended reg instructions. */
{12, 4 }, /*cond: condition flags as a source operand. */
{12, 4 }, /*opcode: in advsimd load/store instructions. */
{12, 4 }, /*cmode: in advsimd modified immediate instructions. */
{13, 3 }, /*asisdlso_opcode: opcode in advsimd ld/st single element. */
{13, 2 }, /*len: in advsimd tbl/tbx instructions. */
{16, 5 }, /*Rm: in ld/st reg offset and some integer inst. */
{16, 5 }, /*Rs: in load/store exclusive instructions. */
{13, 3 }, /*option: in ld/st reg offset + add/sub extended reg inst. */
{12, 1 }, /*S: in load/store reg offset instructions. */
{21, 2 }, /*hw: in move wide constant instructions. */
{22, 2 }, /*opc: in load/store reg offset instructions. */
{23, 1 }, /*opc1: in load/store reg offset instructions. */
{22, 2 }, /*shift: in add/sub reg/imm shifted instructions. */
{22, 2 }, /*type: floating point type field in fp data inst. */
{30, 2 }, /*ldst_size: size field in ld/st reg offset inst. */
{10, 6 }, /*imm6: in add/sub reg shifted instructions. */
{11, 4 }, /*imm4: in advsimd ext and advsimd ins instructions. */
{16, 5 }, /*imm5: in conditional compare (immediate) instructions. */
{15, 7 }, /*imm7: in load/store pair pre/post index instructions. */
{13, 8 }, /*imm8: in floating-point scalar move immediate inst. */
{12, 9 }, /*imm9: in load/store pre/post index instructions. */
{10, 12 }, /* imm12: in ld/stunsigned imm or add/sub shifted inst. */
{ 5, 14 }, /* imm14: in test bit and branch instructions. */
{ 5, 16 }, /* imm16: in exception instructions. */
{ 0, 26 }, /* imm26: in unconditional branch instructions. */
{ 10, 6 }, /* imms: in bitfield andlogical immediate instructions. */
{16, 6 }, /*immr: in bitfield and logical immediate instructions. */
{16, 3 }, /*immb: in advsimd shift by immediate instructions. */
{19, 4 }, /*immh: in advsimd shift by immediate instructions. */
{22, 1 }, /*N: in logical (immediate) instructions. */
{11, 1 }, /*index: in ld/st inst deciding the pre/post-index. */
{24, 1 }, /*index2: in ld/st pair inst deciding the pre/post-index. */
{31, 1 }, /*sf: in integer data processing instructions. */
{11, 1 }, /*H: in advsimd scalar x indexed element instructions. */
{21, 1 }, /*L: in advsimd scalar x indexed element instructions. */
{20, 1 }, /*M: in advsimd scalar x indexed element instructions. */
{31, 1 }, /*b5: in the test bit and branch instructions. */
{19, 5 }, /*b40: in the test bit and branch instructions. */
{10, 6 }, /*scale: in the fixed-point scalar to fp converting inst. */
};
2、函数aarch64_opcode_encode用于将操作数插入opcode中
參考资料
[1] ARMv8的编译器binutils,结构体aarch64_opcode_table定义见line 1208
[2] 结构体aarch64_opcode定义,line451
[3] 《ARMv8 InstructionSet Overview 》page11