void asm_x64_cmp_i32_with_r32(asm_x64_t* as, int src_i32, int src_r32) {
if (SIGNED_FIT8(src_i32)) {
asm_x64_write_byte_2(as, OPCODE_CMP_I8_WITH_RM32, MODRM_R64(7) | MODRM_RM_REG | MODRM_RM_R64(src_r32));
asm_x64_write_byte_1(as, src_i32 & 0xff);
} else {
asm_x64_write_byte_2(as, OPCODE_CMP_I32_WITH_RM32, MODRM_R64(7) | MODRM_RM_REG | MODRM_RM_R64(src_r32));
asm_x64_write_word32(as, src_i32);
}
}
STATIC void asm_x86_add_i32_to_r32(asm_x86_t *as, int src_i32, int dest_r32) {
if (SIGNED_FIT8(src_i32)) {
asm_x86_write_byte_2(as, OPCODE_ADD_I8_TO_RM32, MODRM_R32(0) | MODRM_RM_REG | MODRM_RM_R32(dest_r32));
asm_x86_write_byte_1(as, src_i32 & 0xff);
} else {
asm_x86_write_byte_2(as, OPCODE_ADD_I32_TO_RM32, MODRM_R32(0) | MODRM_RM_REG | MODRM_RM_R32(dest_r32));
asm_x86_write_word32(as, src_i32);
}
}
STATIC void asm_x64_write_r64_disp(asm_x64_t *as, int r64, int disp_r64, int disp_offset) {
assert(disp_r64 != ASM_X64_REG_RSP);
if (disp_r64 == ASM_X64_REG_R12) {
// special case for r12; not fully implemented
assert(SIGNED_FIT8(disp_offset));
asm_x64_write_byte_3(as, MODRM_R64(r64) | MODRM_RM_DISP8 | MODRM_RM_R64(disp_r64), 0x24, IMM32_L0(disp_offset));
return;
}
if (disp_offset == 0 && disp_r64 != ASM_X64_REG_RBP && disp_r64 != ASM_X64_REG_R13) {
asm_x64_write_byte_1(as, MODRM_R64(r64) | MODRM_RM_DISP0 | MODRM_RM_R64(disp_r64));
} else if (SIGNED_FIT8(disp_offset)) {
asm_x64_write_byte_2(as, MODRM_R64(r64) | MODRM_RM_DISP8 | MODRM_RM_R64(disp_r64), IMM32_L0(disp_offset));
} else {
asm_x64_write_byte_1(as, MODRM_R64(r64) | MODRM_RM_DISP32 | MODRM_RM_R64(disp_r64));
asm_x64_write_word32(as, disp_offset);
}
}
void asm_x64_sub_i32_from_r64(asm_x64_t* as, int src_i32, int dest_r64) {
if (SIGNED_FIT8(src_i32)) {
// use REX prefix for 64 bit operation
asm_x64_write_byte_3(as, REX_PREFIX | REX_W, OPCODE_SUB_I8_FROM_RM64, MODRM_R64(5) | MODRM_RM_REG | MODRM_RM_R64(dest_r64));
asm_x64_write_byte_1(as, src_i32 & 0xff);
} else {
// use REX prefix for 64 bit operation
asm_x64_write_byte_3(as, REX_PREFIX | REX_W, OPCODE_SUB_I32_FROM_RM64, MODRM_R64(5) | MODRM_RM_REG | MODRM_RM_R64(dest_r64));
asm_x64_write_word32(as, src_i32);
}
}
void asm_x64_sub_i32_from_r32(asm_x64_t* as, int src_i32, int dest_r32) {
if (SIGNED_FIT8(src_i32)) {
// defaults to 32 bit operation
asm_x64_write_byte_2(as, OPCODE_SUB_I8_FROM_RM64, MODRM_R64(5) | MODRM_RM_REG | MODRM_RM_R64(dest_r32));
asm_x64_write_byte_1(as, src_i32 & 0xff);
} else {
// defaults to 32 bit operation
asm_x64_write_byte_2(as, OPCODE_SUB_I32_FROM_RM64, MODRM_R64(5) | MODRM_RM_REG | MODRM_RM_R64(dest_r32));
asm_x64_write_word32(as, src_i32);
}
}
STATIC void asm_x86_sub_r32_i32(asm_x86_t *as, int dest_r32, int src_i32) {
if (SIGNED_FIT8(src_i32)) {
// defaults to 32 bit operation
asm_x86_write_byte_2(as, OPCODE_SUB_I8_FROM_RM32, MODRM_R32(5) | MODRM_RM_REG | MODRM_RM_R32(dest_r32));
asm_x86_write_byte_1(as, src_i32 & 0xff);
} else {
// defaults to 32 bit operation
asm_x86_write_byte_2(as, OPCODE_SUB_I32_FROM_RM32, MODRM_R32(5) | MODRM_RM_REG | MODRM_RM_R32(dest_r32));
asm_x86_write_word32(as, src_i32);
}
}
static void asm_x64_write_r64_disp(asm_x64_t* as, int r64, int disp_r64, int disp_offset) {
assert(disp_r64 != REG_RSP);
if (disp_offset == 0 && disp_r64 != REG_RBP) {
asm_x64_write_byte_1(as, MODRM_R64(r64) | MODRM_RM_DISP0 | MODRM_RM_R64(disp_r64));
} else if (SIGNED_FIT8(disp_offset)) {
asm_x64_write_byte_2(as, MODRM_R64(r64) | MODRM_RM_DISP8 | MODRM_RM_R64(disp_r64), IMM32_L0(disp_offset));
} else {
asm_x64_write_byte_1(as, MODRM_R64(r64) | MODRM_RM_DISP32 | MODRM_RM_R64(disp_r64));
asm_x64_write_word32(as, disp_offset);
}
}
STATIC void asm_x86_write_r32_disp(asm_x86_t *as, int r32, int disp_r32, int disp_offset) {
assert(disp_r32 != ASM_X86_REG_ESP);
if (disp_offset == 0 && disp_r32 != ASM_X86_REG_EBP) {
asm_x86_write_byte_1(as, MODRM_R32(r32) | MODRM_RM_DISP0 | MODRM_RM_R32(disp_r32));
} else if (SIGNED_FIT8(disp_offset)) {
asm_x86_write_byte_2(as, MODRM_R32(r32) | MODRM_RM_DISP8 | MODRM_RM_R32(disp_r32), IMM32_L0(disp_offset));
} else {
asm_x86_write_byte_1(as, MODRM_R32(r32) | MODRM_RM_DISP32 | MODRM_RM_R32(disp_r32));
asm_x86_write_word32(as, disp_offset);
}
}
void asm_x64_add_i32_to_r32(asm_x64_t* as, int src_i32, int dest_r32)
{
assert(dest_r32 != REG_RSP); // in this case i think src_i32 must be 64 bits
if (SIGNED_FIT8(src_i32))
{
asm_x64_write_byte_2(as, OPCODE_ADD_I8_TO_RM32, MODRM_R64(0) | MODRM_RM_REG | MODRM_RM_R64(dest_r32));
asm_x64_write_byte_1(as, src_i32 & 0xff);
}
else
{
asm_x64_write_byte_2(as, OPCODE_ADD_I32_TO_RM32, MODRM_R64(0) | MODRM_RM_REG | MODRM_RM_R64(dest_r32));
asm_x64_write_word32(as, src_i32);
}
}
void asm_xtensa_exit(asm_xtensa_t *as) {
// restore registers
for (int i = NUM_REGS_SAVED - 1; i >= 1; --i) {
asm_xtensa_op_l32i_n(as, ASM_XTENSA_REG_A11 + i, ASM_XTENSA_REG_A1, i);
}
asm_xtensa_op_l32i_n(as, ASM_XTENSA_REG_A0, ASM_XTENSA_REG_A1, 0);
// restore stack-pointer and return
if (SIGNED_FIT8(as->stack_adjust)) {
asm_xtensa_op_addi(as, ASM_XTENSA_REG_A1, ASM_XTENSA_REG_A1, as->stack_adjust);
} else {
asm_xtensa_op_movi(as, ASM_XTENSA_REG_A9, as->stack_adjust);
asm_xtensa_op_add_n(as, ASM_XTENSA_REG_A1, ASM_XTENSA_REG_A1, ASM_XTENSA_REG_A9);
}
asm_xtensa_op_ret_n(as);
}
void asm_xtensa_entry(asm_xtensa_t *as, int num_locals) {
// jump over the constants
asm_xtensa_op_j(as, as->num_const * WORD_SIZE + 4 - 4);
mp_asm_base_get_cur_to_write_bytes(&as->base, 1); // padding/alignment byte
as->const_table = (uint32_t*)mp_asm_base_get_cur_to_write_bytes(&as->base, as->num_const * 4);
// adjust the stack-pointer to store a0, a12, a13, a14, a15 and locals, 16-byte aligned
as->stack_adjust = (((NUM_REGS_SAVED + num_locals) * WORD_SIZE) + 15) & ~15;
if (SIGNED_FIT8(-as->stack_adjust)) {
asm_xtensa_op_addi(as, ASM_XTENSA_REG_A1, ASM_XTENSA_REG_A1, -as->stack_adjust);
} else {
asm_xtensa_op_movi(as, ASM_XTENSA_REG_A9, as->stack_adjust);
asm_xtensa_op_sub(as, ASM_XTENSA_REG_A1, ASM_XTENSA_REG_A1, ASM_XTENSA_REG_A9);
}
// save return value (a0) and callee-save registers (a12, a13, a14, a15)
asm_xtensa_op_s32i_n(as, ASM_XTENSA_REG_A0, ASM_XTENSA_REG_A1, 0);
for (int i = 1; i < NUM_REGS_SAVED; ++i) {
asm_xtensa_op_s32i_n(as, ASM_XTENSA_REG_A11 + i, ASM_XTENSA_REG_A1, i);
}
}
void asm_x64_jcc_label(asm_x64_t *as, int jcc_type, int label) {
int dest = get_label_dest(as, label);
int rel = dest - as->code_offset;
if (dest >= 0 && rel < 0) {
// is a backwards jump, so we know the size of the jump on the first pass
// calculate rel assuming 8 bit relative jump
rel -= 2;
if (SIGNED_FIT8(rel)) {
asm_x64_write_byte_2(as, OPCODE_JCC_REL8 | jcc_type, rel & 0xff);
} else {
rel += 2;
goto large_jump;
}
} else {
// is a forwards jump, so need to assume it's large
large_jump:
rel -= 6;
asm_x64_write_byte_2(as, OPCODE_JCC_REL32_A, OPCODE_JCC_REL32_B | jcc_type);
asm_x64_write_word32(as, rel);
}
}
void asm_x64_jmp_label(asm_x64_t *as, mp_uint_t label) {
mp_uint_t dest = get_label_dest(as, label);
mp_int_t rel = dest - as->base.code_offset;
if (dest != (mp_uint_t)-1 && rel < 0) {
// is a backwards jump, so we know the size of the jump on the first pass
// calculate rel assuming 8 bit relative jump
rel -= 2;
if (SIGNED_FIT8(rel)) {
asm_x64_write_byte_2(as, OPCODE_JMP_REL8, rel & 0xff);
} else {
rel += 2;
goto large_jump;
}
} else {
// is a forwards jump, so need to assume it's large
large_jump:
rel -= 5;
asm_x64_write_byte_1(as, OPCODE_JMP_REL32);
asm_x64_write_word32(as, rel);
}
}