[picodrive.git] / cpu / drc / emit_x86.c

/*
 * note:
 *  temp registers must be eax-edx due to use of SETcc and r/w 8/16.
 * note about silly things like emith_eor_r_r_r:
 *  these are here because the compiler was designed
 *  for ARM as it's primary target.
 */
#include <stdarg.h>

enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };

#define CONTEXT_REG xBP

#define ICOND_JO  0x00
#define ICOND_JNO 0x01
#define ICOND_JB  0x02
#define ICOND_JAE 0x03
#define ICOND_JE  0x04
#define ICOND_JNE 0x05
#define ICOND_JBE 0x06
#define ICOND_JA  0x07
#define ICOND_JS  0x08
#define ICOND_JNS 0x09
#define ICOND_JL  0x0c
#define ICOND_JGE 0x0d
#define ICOND_JLE 0x0e
#define ICOND_JG  0x0f

#define IOP_JMP   0xeb

// unified conditions (we just use rel8 jump instructions for x86)
#define DCOND_EQ ICOND_JE
#define DCOND_NE ICOND_JNE
#define DCOND_MI ICOND_JS      // MInus
#define DCOND_PL ICOND_JNS     // PLus or zero
#define DCOND_HI ICOND_JA      // higher (unsigned)
#define DCOND_HS ICOND_JAE     // higher || same (unsigned)
#define DCOND_LO ICOND_JB      // lower (unsigned)
#define DCOND_LS ICOND_JBE     // lower || same (unsigned)
#define DCOND_GE ICOND_JGE     // greater || equal (signed)
#define DCOND_GT ICOND_JG      // greater (signed)
#define DCOND_LE ICOND_JLE     // less || equal (signed)
#define DCOND_LT ICOND_JL      // less (signed)
#define DCOND_VS ICOND_JO      // oVerflow Set
#define DCOND_VC ICOND_JNO     // oVerflow Clear

#define EMIT_PTR(ptr, val, type) \
	*(type *)(ptr) = val

#define EMIT(val, type) { \
	EMIT_PTR(tcache_ptr, val, type); \
	tcache_ptr += sizeof(type); \
}

#define EMIT_OP(op) { \
	COUNT_OP; \
	EMIT(op, u8); \
}

#define EMIT_MODRM(mod,r,rm) \
	EMIT(((mod)<<6) | ((r)<<3) | (rm), u8)

#define EMIT_SIB(scale,index,base) \
	EMIT(((scale)<<6) | ((index)<<3) | (base), u8)

#define EMIT_OP_MODRM(op,mod,r,rm) do { \
	EMIT_OP(op); \
	EMIT_MODRM(mod, r, rm); \
} while (0)

#define JMP8_POS(ptr) \
	ptr = tcache_ptr; \
	tcache_ptr += 2

#define JMP8_EMIT(op, ptr) \
	EMIT_PTR(ptr, 0x70|(op), u8); \
	EMIT_PTR(ptr + 1, (tcache_ptr - (ptr+2)), u8)

#define JMP8_EMIT_NC(ptr) \
	EMIT_PTR(ptr, IOP_JMP, u8); \
	EMIT_PTR(ptr + 1, (tcache_ptr - (ptr+2)), u8)

// _r_r
#define emith_move_r_r(dst, src) \
	EMIT_OP_MODRM(0x8b, 3, dst, src)

#define emith_add_r_r(d, s) \
	EMIT_OP_MODRM(0x01, 3, s, d)

#define emith_sub_r_r(d, s) \
	EMIT_OP_MODRM(0x29, 3, s, d)

#define emith_adc_r_r(d, s) \
	EMIT_OP_MODRM(0x11, 3, s, d)

#define emith_sbc_r_r(d, s) \
	EMIT_OP_MODRM(0x19, 3, s, d) /* SBB */

#define emith_or_r_r(d, s) \
	EMIT_OP_MODRM(0x09, 3, s, d)

#define emith_and_r_r(d, s) \
	EMIT_OP_MODRM(0x21, 3, s, d)

#define emith_eor_r_r(d, s) \
	EMIT_OP_MODRM(0x31, 3, s, d) /* XOR */

#define emith_tst_r_r(d, s) \
	EMIT_OP_MODRM(0x85, 3, s, d) /* TEST */

#define emith_cmp_r_r(d, s) \
	EMIT_OP_MODRM(0x39, 3, s, d)

// fake teq - test equivalence - get_flags(d ^ s)
#define emith_teq_r_r(d, s) { \
	emith_push(d); \
	emith_eor_r_r(d, s); \
	emith_pop(d); \
}

#define emith_mvn_r_r(d, s) { \
	if (d != s) \
		emith_move_r_r(d, s); \
	EMIT_OP_MODRM(0xf7, 3, 2, d); /* NOT d */ \
}

#define emith_negc_r_r(d, s) { \
	int tmp_ = rcache_get_tmp(); \
	emith_move_r_imm(tmp_, 0); \
	emith_sbc_r_r(tmp_, s); \
	emith_move_r_r(d, tmp_); \
	rcache_free_tmp(tmp_); \
}

#define emith_neg_r_r(d, s) { \
	if (d != s) \
		emith_move_r_r(d, s); \
	EMIT_OP_MODRM(0xf7, 3, 3, d); /* NEG d */ \
}

// _r_r_r
#define emith_eor_r_r_r(d, s1, s2) { \
	if (d == s1) { \
		emith_eor_r_r(d, s2); \
	} else if (d == s2) { \
		emith_eor_r_r(d, s1); \
	} else { \
		emith_move_r_r(d, s1); \
		emith_eor_r_r(d, s2); \
	} \
}

// _r_r_shift
#define emith_or_r_r_lsl(d, s, lslimm) { \
	int tmp_ = rcache_get_tmp(); \
	emith_lsl(tmp_, s, lslimm); \
	emith_or_r_r(d, tmp_); \
	rcache_free_tmp(tmp_); \
}

// d != s
#define emith_eor_r_r_lsr(d, s, lsrimm) { \
	emith_push(s); \
	emith_lsr(s, s, lsrimm); \
	emith_eor_r_r(d, s); \
	emith_pop(s); \
}

// _r_imm
#define emith_move_r_imm(r, imm) { \
	EMIT_OP(0xb8 + (r)); \
	EMIT(imm, u32); \
}

#define emith_move_r_imm_s8(r, imm) \
	emith_move_r_imm(r, (u32)(signed int)(signed char)(imm))

#define emith_arith_r_imm(op, r, imm) do { \
	EMIT_OP_MODRM(0x81, 3, op, r); \
	EMIT(imm, u32); \
} while (0)

#define emith_add_r_imm(r, imm) \
	emith_arith_r_imm(0, r, imm)

#define emith_or_r_imm(r, imm) \
	emith_arith_r_imm(1, r, imm)

#define emith_adc_r_imm(r, imm) \
	emith_arith_r_imm(2, r, imm)

#define emith_sbc_r_imm(r, imm) \
	emith_arith_r_imm(3, r, imm) // sbb

#define emith_and_r_imm(r, imm) \
	emith_arith_r_imm(4, r, imm)

#define emith_sub_r_imm(r, imm) \
	emith_arith_r_imm(5, r, imm)

#define emith_eor_r_imm(r, imm) \
	emith_arith_r_imm(6, r, imm)

#define emith_cmp_r_imm(r, imm) \
	emith_arith_r_imm(7, r, imm)

#define emith_tst_r_imm(r, imm) do { \
	EMIT_OP_MODRM(0xf7, 3, 0, r); \
	EMIT(imm, u32); \
} while (0)

// fake
#define emith_bic_r_imm(r, imm) \
	emith_arith_r_imm(4, r, ~(imm))

// fake conditionals (using SJMP instead)
#define emith_move_r_imm_c(cond, r, imm) { \
	(void)(cond); \
	emith_move_r_imm(r, imm); \
}

#define emith_add_r_imm_c(cond, r, imm) { \
	(void)(cond); \
	emith_add_r_imm(r, imm); \
}

#define emith_sub_r_imm_c(cond, r, imm) { \
	(void)(cond); \
	emith_sub_r_imm(r, imm); \
}

#define emith_or_r_imm_c(cond, r, imm) \
	emith_or_r_imm(r, imm)
#define emith_eor_r_imm_c(cond, r, imm) \
	emith_eor_r_imm(r, imm)
#define emith_bic_r_imm_c(cond, r, imm) \
	emith_bic_r_imm(r, imm)
#define emith_ror_c(cond, d, s, cnt) \
	emith_ror(d, s, cnt)

#define emith_read_r_r_offs_c(cond, r, rs, offs) \
	emith_read_r_r_offs(r, rs, offs)
#define emith_write_r_r_offs_c(cond, r, rs, offs) \
	emith_write_r_r_offs(r, rs, offs)
#define emith_read8_r_r_offs_c(cond, r, rs, offs) \
	emith_read8_r_r_offs(r, rs, offs)
#define emith_write8_r_r_offs_c(cond, r, rs, offs) \
	emith_write8_r_r_offs(r, rs, offs)
#define emith_read16_r_r_offs_c(cond, r, rs, offs) \
	emith_read16_r_r_offs(r, rs, offs)
#define emith_write16_r_r_offs_c(cond, r, rs, offs) \
	emith_write16_r_r_offs(r, rs, offs)
#define emith_jump_reg_c(cond, r) \
	emith_jump_reg(r)
#define emith_jump_ctx_c(cond, offs) \
	emith_jump_ctx(offs)
#define emith_ret_c(cond) \
	emith_ret()

// _r_r_imm
#define emith_add_r_r_imm(d, s, imm) { \
	if (d != s) \
		emith_move_r_r(d, s); \
	emith_add_r_imm(d, imm); \
}

#define emith_and_r_r_imm(d, s, imm) { \
	if (d != s) \
		emith_move_r_r(d, s); \
	emith_and_r_imm(d, imm); \
}

// shift
#define emith_shift(op, d, s, cnt) { \
	if (d != s) \
		emith_move_r_r(d, s); \
	EMIT_OP_MODRM(0xc1, 3, op, d); \
	EMIT(cnt, u8); \
}

#define emith_lsl(d, s, cnt) \
	emith_shift(4, d, s, cnt)

#define emith_lsr(d, s, cnt) \
	emith_shift(5, d, s, cnt)

#define emith_asr(d, s, cnt) \
	emith_shift(7, d, s, cnt)

#define emith_rol(d, s, cnt) \
	emith_shift(0, d, s, cnt)

#define emith_ror(d, s, cnt) \
	emith_shift(1, d, s, cnt)

#define emith_rolc(r) \
	EMIT_OP_MODRM(0xd1, 3, 2, r)

#define emith_rorc(r) \
	EMIT_OP_MODRM(0xd1, 3, 3, r)

// misc
#define emith_push(r) \
	EMIT_OP(0x50 + (r))

#define emith_push_imm(imm) { \
	EMIT_OP(0x68); \
	EMIT(imm, u32); \
}

#define emith_pop(r) \
	EMIT_OP(0x58 + (r))

#define emith_neg_r(r) \
	EMIT_OP_MODRM(0xf7, 3, 3, r)

#define emith_clear_msb(d, s, count) { \
	u32 t = (u32)-1; \
	t >>= count; \
	if (d != s) \
		emith_move_r_r(d, s); \
	emith_and_r_imm(d, t); \
}

#define emith_clear_msb_c(cond, d, s, count) { \
	(void)(cond); \
	emith_clear_msb(d, s, count); \
}

#define emith_sext(d, s, bits) { \
	emith_lsl(d, s, 32 - (bits)); \
	emith_asr(d, d, 32 - (bits)); \
}

#define emith_setc(r) { \
	EMIT_OP(0x0f); \
	EMIT_OP_MODRM(0x92, 3, 0, r); /* SETC r */ \
}

// XXX: stupid mess
#define emith_mul_(op, dlo, dhi, s1, s2) { \
	int rmr; \
	if (dlo != xAX && dhi != xAX) \
		emith_push(xAX); \
	if (dlo != xDX && dhi != xDX) \
		emith_push(xDX); \
	if ((s1) == xAX) \
		rmr = s2; \
	else if ((s2) == xAX) \
		rmr = s1; \
	else { \
		emith_move_r_r(xAX, s1); \
		rmr = s2; \
	} \
	EMIT_OP_MODRM(0xf7, 3, op, rmr); /* xMUL rmr */ \
	/* XXX: using push/pop for the case of edx->eax; eax->edx */ \
	if (dhi != xDX && dhi != -1) \
		emith_push(xDX); \
	if (dlo != xAX) \
		emith_move_r_r(dlo, xAX); \
	if (dhi != xDX && dhi != -1) \
		emith_pop(dhi); \
	if (dlo != xDX && dhi != xDX) \
		emith_pop(xDX); \
	if (dlo != xAX && dhi != xAX) \
		emith_pop(xAX); \
}

#define emith_mul_u64(dlo, dhi, s1, s2) \
	emith_mul_(4, dlo, dhi, s1, s2) /* MUL */

#define emith_mul_s64(dlo, dhi, s1, s2) \
	emith_mul_(5, dlo, dhi, s1, s2) /* IMUL */

#define emith_mul(d, s1, s2) \
	emith_mul_(4, d, -1, s1, s2)

// (dlo,dhi) += signed(s1) * signed(s2)
#define emith_mula_s64(dlo, dhi, s1, s2) { \
	emith_push(dhi); \
	emith_push(dlo); \
	emith_mul_(5, dlo, dhi, s1, s2); \
	EMIT_OP_MODRM(0x03, 0, dlo, 4); \
	EMIT_SIB(0, 4, 4); /* add dlo, [esp] */ \
	EMIT_OP_MODRM(0x13, 1, dhi, 4); \
	EMIT_SIB(0, 4, 4); \
	EMIT(4, u8); /* adc dhi, [esp+4] */ \
	emith_add_r_imm(xSP, 4*2); \
}

// "flag" instructions are the same
#define emith_subf_r_imm emith_sub_r_imm
#define emith_addf_r_r   emith_add_r_r
#define emith_subf_r_r   emith_sub_r_r
#define emith_adcf_r_r   emith_adc_r_r
#define emith_sbcf_r_r   emith_sbc_r_r
#define emith_eorf_r_r   emith_eor_r_r
#define emith_negcf_r_r  emith_negc_r_r

#define emith_lslf  emith_lsl
#define emith_lsrf  emith_lsr
#define emith_asrf  emith_asr
#define emith_rolf  emith_rol
#define emith_rorf  emith_ror
#define emith_rolcf emith_rolc
#define emith_rorcf emith_rorc

#define emith_deref_op(op, r, rs, offs) do { \
	/* mov r <-> [ebp+#offs] */ \
	if ((offs) >= 0x80) { \
		EMIT_OP_MODRM(op, 2, r, rs); \
		EMIT(offs, u32); \
	} else { \
		EMIT_OP_MODRM(op, 1, r, rs); \
		EMIT(offs, u8); \
	} \
} while (0)

#define is_abcdx(r) (xAX <= (r) && (r) <= xDX)

#define emith_read_r_r_offs(r, rs, offs) \
	emith_deref_op(0x8b, r, rs, offs)

#define emith_write_r_r_offs(r, rs, offs) \
	emith_deref_op(0x89, r, rs, offs)

// note: don't use prefixes on this
#define emith_read8_r_r_offs(r, rs, offs) do { \
	int r_ = r; \
	if (!is_abcdx(r)) \
		r_ = rcache_get_tmp(); \
	emith_deref_op(0x8a, r_, rs, offs); \
	if ((r) != r_) { \
		emith_move_r_r(r, r_); \
		rcache_free_tmp(r_); \
	} \
} while (0)

#define emith_write8_r_r_offs(r, rs, offs) do {\
	int r_ = r; \
	if (!is_abcdx(r)) { \
		r_ = rcache_get_tmp(); \
		emith_move_r_r(r_, r); \
	} \
	emith_deref_op(0x88, r_, rs, offs); \
	if ((r) != r_) \
		rcache_free_tmp(r_); \
} while (0)

#define emith_read16_r_r_offs(r, rs, offs) { \
	EMIT(0x66, u8); /* operand override */ \
	emith_read_r_r_offs(r, rs, offs); \
}

#define emith_write16_r_r_offs(r, rs, offs) { \
	EMIT(0x66, u8); \
	emith_write_r_r_offs(r, rs, offs); \
}

#define emith_ctx_read(r, offs) \
	emith_read_r_r_offs(r, CONTEXT_REG, offs)

#define emith_ctx_write(r, offs) \
	emith_write_r_r_offs(r, CONTEXT_REG, offs)

#define emith_ctx_read_multiple(r, offs, cnt, tmpr) do { \
	int r_ = r, offs_ = offs, cnt_ = cnt;     \
	for (; cnt_ > 0; r_++, offs_ += 4, cnt_--) \
		emith_ctx_read(r_, offs_);        \
} while (0)

#define emith_ctx_write_multiple(r, offs, cnt, tmpr) do { \
	int r_ = r, offs_ = offs, cnt_ = cnt;     \
	for (; cnt_ > 0; r_++, offs_ += 4, cnt_--) \
		emith_ctx_write(r_, offs_);       \
} while (0)

// assumes EBX is free
#define emith_ret_to_ctx(offs) { \
	emith_pop(xBX); \
	emith_ctx_write(xBX, offs); \
}

#define emith_jump(ptr) { \
	u32 disp = (u32)(ptr) - ((u32)tcache_ptr + 5); \
	EMIT_OP(0xe9); \
	EMIT(disp, u32); \
}

#define emith_jump_patchable(target) \
	emith_jump(target)

#define emith_jump_cond(cond, ptr) { \
	u32 disp = (u32)(ptr) - ((u32)tcache_ptr + 6); \
	EMIT(0x0f, u8); \
	EMIT_OP(0x80 | (cond)); \
	EMIT(disp, u32); \
}

#define emith_jump_cond_patchable(cond, target) \
	emith_jump_cond(cond, target)

#define emith_jump_patch(ptr, target) do { \
	u32 disp_ = (u32)(target) - ((u32)(ptr) + 4); \
	u32 offs_ = (*(u8 *)(ptr) == 0x0f) ? 2 : 1; \
	EMIT_PTR((u8 *)(ptr) + offs_, disp_ - offs_, u32); \
} while (0)

#define emith_jump_at(ptr, target) { \
	u32 disp_ = (u32)(target) - ((u32)(ptr) + 5); \
	EMIT_PTR(ptr, 0xe9, u8); \
	EMIT_PTR((u8 *)(ptr) + 1, disp_, u32); \
}

#define emith_call(ptr) { \
	u32 disp = (u32)(ptr) - ((u32)tcache_ptr + 5); \
	EMIT_OP(0xe8); \
	EMIT(disp, u32); \
}

#define emith_call_cond(cond, ptr) \
	emith_call(ptr)

#define emith_call_reg(r) \
	EMIT_OP_MODRM(0xff, 3, 2, r)

#define emith_call_ctx(offs) { \
	EMIT_OP_MODRM(0xff, 2, 2, CONTEXT_REG); \
	EMIT(offs, u32); \
}

#define emith_ret() \
	EMIT_OP(0xc3)

#define emith_jump_reg(r) \
	EMIT_OP_MODRM(0xff, 3, 4, r)

#define emith_jump_ctx(offs) { \
	EMIT_OP_MODRM(0xff, 2, 4, CONTEXT_REG); \
	EMIT(offs, u32); \
}

#define emith_push_ret()

#define emith_pop_and_ret() \
	emith_ret()

#define EMITH_JMP_START(cond) { \
	u8 *cond_ptr; \
	JMP8_POS(cond_ptr)

#define EMITH_JMP_END(cond) \
	JMP8_EMIT(cond, cond_ptr); \
}

#define EMITH_JMP3_START(cond) { \
	u8 *cond_ptr, *else_ptr; \
	JMP8_POS(cond_ptr)

#define EMITH_JMP3_MID(cond) \
	JMP8_POS(else_ptr); \
	JMP8_EMIT(cond, cond_ptr);

#define EMITH_JMP3_END() \
	JMP8_EMIT_NC(else_ptr); \
}

// "simple" jump (no more then a few insns)
// ARM will use conditional instructions here
#define EMITH_SJMP_START EMITH_JMP_START
#define EMITH_SJMP_END EMITH_JMP_END

#define EMITH_SJMP3_START EMITH_JMP3_START
#define EMITH_SJMP3_MID EMITH_JMP3_MID
#define EMITH_SJMP3_END EMITH_JMP3_END

#define emith_pass_arg_r(arg, reg) { \
	int rd = 7; \
	host_arg2reg(rd, arg); \
	emith_move_r_r(rd, reg); \
}

#define emith_pass_arg_imm(arg, imm) { \
	int rd = 7; \
	host_arg2reg(rd, arg); \
	emith_move_r_imm(rd, imm); \
}

#define host_instructions_updated(base, end)

#define host_arg2reg(rd, arg) \
	switch (arg) { \
	case 0: rd = xAX; break; \
	case 1: rd = xDX; break; \
	case 2: rd = xCX; break; \
	}

/* SH2 drc specific */
#define emith_sh2_drc_entry() { \
	emith_push(xBX);        \
	emith_push(xBP);        \
	emith_push(xSI);        \
	emith_push(xDI);        \
}

#define emith_sh2_drc_exit() {  \
	emith_pop(xDI);         \
	emith_pop(xSI);         \
	emith_pop(xBP);         \
	emith_pop(xBX);         \
	emith_ret();            \
}

// assumes EBX is free temporary
#define emith_sh2_wcall(a, tab, ret_ptr) { \
	int arg2_; \
	host_arg2reg(arg2_, 2); \
	emith_lsr(xBX, a, SH2_WRITE_SHIFT); \
	EMIT_OP_MODRM(0x8b, 0, xBX, 4); \
	EMIT_SIB(2, xBX, tab); /* mov ebx, [tab + ebx * 4] */ \
	emith_ctx_read(arg2_, offsetof(SH2, is_slave)); \
	emith_push_imm((long)(ret_ptr)); \
	emith_jump_reg(xBX); \
}

#define emith_sh2_dtbf_loop() { \
	u8 *jmp0; /* negative cycles check */            \
	u8 *jmp1; /* unsinged overflow check */          \
	int cr, rn;                                      \
	int tmp_ = rcache_get_tmp();                     \
	cr = rcache_get_reg(SHR_SR, RC_GR_RMW);          \
	rn = rcache_get_reg((op >> 8) & 0x0f, RC_GR_RMW);\
	emith_sub_r_imm(rn, 1);                          \
	emith_sub_r_imm(cr, (cycles+1) << 12);           \
	cycles = 0;                                      \
	emith_asr(tmp_, cr, 2+12);                       \
	JMP8_POS(jmp0); /* no negative cycles */         \
	emith_move_r_imm(tmp_, 0);                       \
	JMP8_EMIT(ICOND_JNS, jmp0);                      \
	emith_and_r_imm(cr, 0xffe);                      \
	emith_subf_r_r(rn, tmp_);                        \
	JMP8_POS(jmp1); /* no overflow */                \
	emith_neg_r(rn); /* count left */                \
	emith_lsl(rn, rn, 2+12);                         \
	emith_or_r_r(cr, rn);                            \
	emith_or_r_imm(cr, 1);                           \
	emith_move_r_imm(rn, 0);                         \
	JMP8_EMIT(ICOND_JA, jmp1);                       \
	rcache_free_tmp(tmp_);                           \
}

#define emith_write_sr(sr, srcr) { \
	int tmp_ = rcache_get_tmp(); \
	emith_clear_msb(tmp_, srcr, 22); \
	emith_bic_r_imm(sr, 0x3ff); \
	emith_or_r_r(sr, tmp_); \
	rcache_free_tmp(tmp_); \
}

#define emith_tpop_carry(sr, is_sub) \
	emith_lsr(sr, sr, 1)

#define emith_tpush_carry(sr, is_sub) \
	emith_adc_r_r(sr, sr)

/*
 * if Q
 *   t = carry(Rn += Rm)
 * else
 *   t = carry(Rn -= Rm)
 * T ^= t
 */
#define emith_sh2_div1_step(rn, rm, sr) {         \
	u8 *jmp0, *jmp1;                          \
	int tmp_ = rcache_get_tmp();              \
	emith_eor_r_r(tmp_, tmp_);                \
	emith_tst_r_imm(sr, Q);  /* if (Q ^ M) */ \
	JMP8_POS(jmp0);          /* je do_sub */  \
	emith_add_r_r(rn, rm);                    \
	JMP8_POS(jmp1);          /* jmp done */   \
	JMP8_EMIT(ICOND_JE, jmp0); /* do_sub: */  \
	emith_sub_r_r(rn, rm);                    \
	JMP8_EMIT_NC(jmp1);      /* done: */      \
	emith_setc(tmp_);                         \
	EMIT_OP_MODRM(0x31, 3, tmp_, sr); /* T = Q1 ^ Q2 */ \
	rcache_free_tmp(tmp_);                    \
}
Commit	Line	Data
	1	/*
	2	* note:
	3	* temp registers must be eax-edx due to use of SETcc and r/w 8/16.
	4	* note about silly things like emith_eor_r_r_r:
	5	* these are here because the compiler was designed
	6	* for ARM as it's primary target.
	7	*/
	8	#include <stdarg.h>
	9
	10	enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
	11
	12	#define CONTEXT_REG xBP
	13
	14	#define ICOND_JO 0x00
	15	#define ICOND_JNO 0x01
	16	#define ICOND_JB 0x02
	17	#define ICOND_JAE 0x03
	18	#define ICOND_JE 0x04
	19	#define ICOND_JNE 0x05
	20	#define ICOND_JBE 0x06
	21	#define ICOND_JA 0x07
	22	#define ICOND_JS 0x08
	23	#define ICOND_JNS 0x09
	24	#define ICOND_JL 0x0c
	25	#define ICOND_JGE 0x0d
	26	#define ICOND_JLE 0x0e
	27	#define ICOND_JG 0x0f
	28
	29	#define IOP_JMP 0xeb
	30
	31	// unified conditions (we just use rel8 jump instructions for x86)
	32	#define DCOND_EQ ICOND_JE
	33	#define DCOND_NE ICOND_JNE
	34	#define DCOND_MI ICOND_JS // MInus
	35	#define DCOND_PL ICOND_JNS // PLus or zero
	36	#define DCOND_HI ICOND_JA // higher (unsigned)
	37	#define DCOND_HS ICOND_JAE // higher \|\| same (unsigned)
	38	#define DCOND_LO ICOND_JB // lower (unsigned)
	39	#define DCOND_LS ICOND_JBE // lower \|\| same (unsigned)
	40	#define DCOND_GE ICOND_JGE // greater \|\| equal (signed)
	41	#define DCOND_GT ICOND_JG // greater (signed)
	42	#define DCOND_LE ICOND_JLE // less \|\| equal (signed)
	43	#define DCOND_LT ICOND_JL // less (signed)
	44	#define DCOND_VS ICOND_JO // oVerflow Set
	45	#define DCOND_VC ICOND_JNO // oVerflow Clear
	46
	47	#define EMIT_PTR(ptr, val, type) \
	48	(type )(ptr) = val
	49
	50	#define EMIT(val, type) { \
	51	EMIT_PTR(tcache_ptr, val, type); \
	52	tcache_ptr += sizeof(type); \
	53	}
	54
	55	#define EMIT_OP(op) { \
	56	COUNT_OP; \
	57	EMIT(op, u8); \
	58	}
	59
	60	#define EMIT_MODRM(mod,r,rm) \
	61	EMIT(((mod)<<6) \| ((r)<<3) \| (rm), u8)
	62
	63	#define EMIT_SIB(scale,index,base) \
	64	EMIT(((scale)<<6) \| ((index)<<3) \| (base), u8)
	65
	66	#define EMIT_OP_MODRM(op,mod,r,rm) do { \
	67	EMIT_OP(op); \
	68	EMIT_MODRM(mod, r, rm); \
	69	} while (0)
	70
	71	#define JMP8_POS(ptr) \
	72	ptr = tcache_ptr; \
	73	tcache_ptr += 2
	74
	75	#define JMP8_EMIT(op, ptr) \
	76	EMIT_PTR(ptr, 0x70\|(op), u8); \
	77	EMIT_PTR(ptr + 1, (tcache_ptr - (ptr+2)), u8)
	78
	79	#define JMP8_EMIT_NC(ptr) \
	80	EMIT_PTR(ptr, IOP_JMP, u8); \
	81	EMIT_PTR(ptr + 1, (tcache_ptr - (ptr+2)), u8)
	82
	83	// _r_r
	84	#define emith_move_r_r(dst, src) \
	85	EMIT_OP_MODRM(0x8b, 3, dst, src)
	86
	87	#define emith_add_r_r(d, s) \
	88	EMIT_OP_MODRM(0x01, 3, s, d)
	89
	90	#define emith_sub_r_r(d, s) \
	91	EMIT_OP_MODRM(0x29, 3, s, d)
	92
	93	#define emith_adc_r_r(d, s) \
	94	EMIT_OP_MODRM(0x11, 3, s, d)
	95
	96	#define emith_sbc_r_r(d, s) \
	97	EMIT_OP_MODRM(0x19, 3, s, d) /* SBB */
	98
	99	#define emith_or_r_r(d, s) \
	100	EMIT_OP_MODRM(0x09, 3, s, d)
	101
	102	#define emith_and_r_r(d, s) \
	103	EMIT_OP_MODRM(0x21, 3, s, d)
	104
	105	#define emith_eor_r_r(d, s) \
	106	EMIT_OP_MODRM(0x31, 3, s, d) /* XOR */
	107
	108	#define emith_tst_r_r(d, s) \
	109	EMIT_OP_MODRM(0x85, 3, s, d) /* TEST */
	110
	111	#define emith_cmp_r_r(d, s) \
	112	EMIT_OP_MODRM(0x39, 3, s, d)
	113
	114	// fake teq - test equivalence - get_flags(d ^ s)
	115	#define emith_teq_r_r(d, s) { \
	116	emith_push(d); \
	117	emith_eor_r_r(d, s); \
	118	emith_pop(d); \
	119	}
	120
	121	#define emith_mvn_r_r(d, s) { \
	122	if (d != s) \
	123	emith_move_r_r(d, s); \
	124	EMIT_OP_MODRM(0xf7, 3, 2, d); /* NOT d */ \
	125	}
	126
	127	#define emith_negc_r_r(d, s) { \
	128	int tmp_ = rcache_get_tmp(); \
	129	emith_move_r_imm(tmp_, 0); \
	130	emith_sbc_r_r(tmp_, s); \
	131	emith_move_r_r(d, tmp_); \
	132	rcache_free_tmp(tmp_); \
	133	}
	134
	135	#define emith_neg_r_r(d, s) { \
	136	if (d != s) \
	137	emith_move_r_r(d, s); \
	138	EMIT_OP_MODRM(0xf7, 3, 3, d); /* NEG d */ \
	139	}
	140
	141	// _r_r_r
	142	#define emith_eor_r_r_r(d, s1, s2) { \
	143	if (d == s1) { \
	144	emith_eor_r_r(d, s2); \
	145	} else if (d == s2) { \
	146	emith_eor_r_r(d, s1); \
	147	} else { \
	148	emith_move_r_r(d, s1); \
	149	emith_eor_r_r(d, s2); \
	150	} \
	151	}
	152
	153	// _r_r_shift
	154	#define emith_or_r_r_lsl(d, s, lslimm) { \
	155	int tmp_ = rcache_get_tmp(); \
	156	emith_lsl(tmp_, s, lslimm); \
	157	emith_or_r_r(d, tmp_); \
	158	rcache_free_tmp(tmp_); \
	159	}
	160
	161	// d != s
	162	#define emith_eor_r_r_lsr(d, s, lsrimm) { \
	163	emith_push(s); \
	164	emith_lsr(s, s, lsrimm); \
	165	emith_eor_r_r(d, s); \
	166	emith_pop(s); \
	167	}
	168
	169	// _r_imm
	170	#define emith_move_r_imm(r, imm) { \
	171	EMIT_OP(0xb8 + (r)); \
	172	EMIT(imm, u32); \
	173	}
	174
	175	#define emith_move_r_imm_s8(r, imm) \
	176	emith_move_r_imm(r, (u32)(signed int)(signed char)(imm))
	177
	178	#define emith_arith_r_imm(op, r, imm) do { \
	179	EMIT_OP_MODRM(0x81, 3, op, r); \
	180	EMIT(imm, u32); \
	181	} while (0)
	182
	183	#define emith_add_r_imm(r, imm) \
	184	emith_arith_r_imm(0, r, imm)
	185
	186	#define emith_or_r_imm(r, imm) \
	187	emith_arith_r_imm(1, r, imm)
	188
	189	#define emith_adc_r_imm(r, imm) \
	190	emith_arith_r_imm(2, r, imm)
	191
	192	#define emith_sbc_r_imm(r, imm) \
	193	emith_arith_r_imm(3, r, imm) // sbb
	194
	195	#define emith_and_r_imm(r, imm) \
	196	emith_arith_r_imm(4, r, imm)
	197
	198	#define emith_sub_r_imm(r, imm) \
	199	emith_arith_r_imm(5, r, imm)
	200
	201	#define emith_eor_r_imm(r, imm) \
	202	emith_arith_r_imm(6, r, imm)
	203
	204	#define emith_cmp_r_imm(r, imm) \
	205	emith_arith_r_imm(7, r, imm)
	206
	207	#define emith_tst_r_imm(r, imm) do { \
	208	EMIT_OP_MODRM(0xf7, 3, 0, r); \
	209	EMIT(imm, u32); \
	210	} while (0)
	211
	212	// fake
	213	#define emith_bic_r_imm(r, imm) \
	214	emith_arith_r_imm(4, r, ~(imm))
	215
	216	// fake conditionals (using SJMP instead)
	217	#define emith_move_r_imm_c(cond, r, imm) { \
	218	(void)(cond); \
	219	emith_move_r_imm(r, imm); \
	220	}
	221
	222	#define emith_add_r_imm_c(cond, r, imm) { \
	223	(void)(cond); \
	224	emith_add_r_imm(r, imm); \
	225	}
	226
	227	#define emith_sub_r_imm_c(cond, r, imm) { \
	228	(void)(cond); \
	229	emith_sub_r_imm(r, imm); \
	230	}
	231
	232	#define emith_or_r_imm_c(cond, r, imm) \
	233	emith_or_r_imm(r, imm)
	234	#define emith_eor_r_imm_c(cond, r, imm) \
	235	emith_eor_r_imm(r, imm)
	236	#define emith_bic_r_imm_c(cond, r, imm) \
	237	emith_bic_r_imm(r, imm)
	238	#define emith_ror_c(cond, d, s, cnt) \
	239	emith_ror(d, s, cnt)
	240
	241	#define emith_read_r_r_offs_c(cond, r, rs, offs) \
	242	emith_read_r_r_offs(r, rs, offs)
	243	#define emith_write_r_r_offs_c(cond, r, rs, offs) \
	244	emith_write_r_r_offs(r, rs, offs)
	245	#define emith_read8_r_r_offs_c(cond, r, rs, offs) \
	246	emith_read8_r_r_offs(r, rs, offs)
	247	#define emith_write8_r_r_offs_c(cond, r, rs, offs) \
	248	emith_write8_r_r_offs(r, rs, offs)
	249	#define emith_read16_r_r_offs_c(cond, r, rs, offs) \
	250	emith_read16_r_r_offs(r, rs, offs)
	251	#define emith_write16_r_r_offs_c(cond, r, rs, offs) \
	252	emith_write16_r_r_offs(r, rs, offs)
	253	#define emith_jump_reg_c(cond, r) \
	254	emith_jump_reg(r)
	255	#define emith_jump_ctx_c(cond, offs) \
	256	emith_jump_ctx(offs)
	257	#define emith_ret_c(cond) \
	258	emith_ret()
	259
	260	// _r_r_imm
	261	#define emith_add_r_r_imm(d, s, imm) { \
	262	if (d != s) \
	263	emith_move_r_r(d, s); \
	264	emith_add_r_imm(d, imm); \
	265	}
	266
	267	#define emith_and_r_r_imm(d, s, imm) { \
	268	if (d != s) \
	269	emith_move_r_r(d, s); \
	270	emith_and_r_imm(d, imm); \
	271	}
	272
	273	// shift
	274	#define emith_shift(op, d, s, cnt) { \
	275	if (d != s) \
	276	emith_move_r_r(d, s); \
	277	EMIT_OP_MODRM(0xc1, 3, op, d); \
	278	EMIT(cnt, u8); \
	279	}
	280
	281	#define emith_lsl(d, s, cnt) \
	282	emith_shift(4, d, s, cnt)
	283
	284	#define emith_lsr(d, s, cnt) \
	285	emith_shift(5, d, s, cnt)
	286
	287	#define emith_asr(d, s, cnt) \
	288	emith_shift(7, d, s, cnt)
	289
	290	#define emith_rol(d, s, cnt) \
	291	emith_shift(0, d, s, cnt)
	292
	293	#define emith_ror(d, s, cnt) \
	294	emith_shift(1, d, s, cnt)
	295
	296	#define emith_rolc(r) \
	297	EMIT_OP_MODRM(0xd1, 3, 2, r)
	298
	299	#define emith_rorc(r) \
	300	EMIT_OP_MODRM(0xd1, 3, 3, r)
	301
	302	// misc
	303	#define emith_push(r) \
	304	EMIT_OP(0x50 + (r))
	305
	306	#define emith_push_imm(imm) { \
	307	EMIT_OP(0x68); \
	308	EMIT(imm, u32); \
	309	}
	310
	311	#define emith_pop(r) \
	312	EMIT_OP(0x58 + (r))
	313
	314	#define emith_neg_r(r) \
	315	EMIT_OP_MODRM(0xf7, 3, 3, r)
	316
	317	#define emith_clear_msb(d, s, count) { \
	318	u32 t = (u32)-1; \
	319	t >>= count; \
	320	if (d != s) \
	321	emith_move_r_r(d, s); \
	322	emith_and_r_imm(d, t); \
	323	}
	324
	325	#define emith_clear_msb_c(cond, d, s, count) { \
	326	(void)(cond); \
	327	emith_clear_msb(d, s, count); \
	328	}
	329
	330	#define emith_sext(d, s, bits) { \
	331	emith_lsl(d, s, 32 - (bits)); \
	332	emith_asr(d, d, 32 - (bits)); \
	333	}
	334
	335	#define emith_setc(r) { \
	336	EMIT_OP(0x0f); \
	337	EMIT_OP_MODRM(0x92, 3, 0, r); /* SETC r */ \
	338	}
	339
	340	// XXX: stupid mess
	341	#define emith_mul_(op, dlo, dhi, s1, s2) { \
	342	int rmr; \
	343	if (dlo != xAX && dhi != xAX) \
	344	emith_push(xAX); \
	345	if (dlo != xDX && dhi != xDX) \
	346	emith_push(xDX); \
	347	if ((s1) == xAX) \
	348	rmr = s2; \
	349	else if ((s2) == xAX) \
	350	rmr = s1; \
	351	else { \
	352	emith_move_r_r(xAX, s1); \
	353	rmr = s2; \
	354	} \
	355	EMIT_OP_MODRM(0xf7, 3, op, rmr); /* xMUL rmr */ \
	356	/* XXX: using push/pop for the case of edx->eax; eax->edx */ \
	357	if (dhi != xDX && dhi != -1) \
	358	emith_push(xDX); \
	359	if (dlo != xAX) \
	360	emith_move_r_r(dlo, xAX); \
	361	if (dhi != xDX && dhi != -1) \
	362	emith_pop(dhi); \
	363	if (dlo != xDX && dhi != xDX) \
	364	emith_pop(xDX); \
	365	if (dlo != xAX && dhi != xAX) \
	366	emith_pop(xAX); \
	367	}
	368
	369	#define emith_mul_u64(dlo, dhi, s1, s2) \
	370	emith_mul_(4, dlo, dhi, s1, s2) /* MUL */
	371
	372	#define emith_mul_s64(dlo, dhi, s1, s2) \
	373	emith_mul_(5, dlo, dhi, s1, s2) /* IMUL */
	374
	375	#define emith_mul(d, s1, s2) \
	376	emith_mul_(4, d, -1, s1, s2)
	377
	378	// (dlo,dhi) += signed(s1) * signed(s2)
	379	#define emith_mula_s64(dlo, dhi, s1, s2) { \
	380	emith_push(dhi); \
	381	emith_push(dlo); \
	382	emith_mul_(5, dlo, dhi, s1, s2); \
	383	EMIT_OP_MODRM(0x03, 0, dlo, 4); \
	384	EMIT_SIB(0, 4, 4); /* add dlo, [esp] */ \
	385	EMIT_OP_MODRM(0x13, 1, dhi, 4); \
	386	EMIT_SIB(0, 4, 4); \
	387	EMIT(4, u8); /* adc dhi, [esp+4] */ \
	388	emith_add_r_imm(xSP, 4*2); \
	389	}
	390
	391	// "flag" instructions are the same
	392	#define emith_subf_r_imm emith_sub_r_imm
	393	#define emith_addf_r_r emith_add_r_r
	394	#define emith_subf_r_r emith_sub_r_r
	395	#define emith_adcf_r_r emith_adc_r_r
	396	#define emith_sbcf_r_r emith_sbc_r_r
	397	#define emith_eorf_r_r emith_eor_r_r
	398	#define emith_negcf_r_r emith_negc_r_r
	399
	400	#define emith_lslf emith_lsl
	401	#define emith_lsrf emith_lsr
	402	#define emith_asrf emith_asr
	403	#define emith_rolf emith_rol
	404	#define emith_rorf emith_ror
	405	#define emith_rolcf emith_rolc
	406	#define emith_rorcf emith_rorc
	407
	408	#define emith_deref_op(op, r, rs, offs) do { \
	409	/* mov r <-> [ebp+#offs] */ \
	410	if ((offs) >= 0x80) { \
	411	EMIT_OP_MODRM(op, 2, r, rs); \
	412	EMIT(offs, u32); \
	413	} else { \
	414	EMIT_OP_MODRM(op, 1, r, rs); \
	415	EMIT(offs, u8); \
	416	} \
	417	} while (0)
	418
	419	#define is_abcdx(r) (xAX <= (r) && (r) <= xDX)
	420
	421	#define emith_read_r_r_offs(r, rs, offs) \
	422	emith_deref_op(0x8b, r, rs, offs)
	423
	424	#define emith_write_r_r_offs(r, rs, offs) \
	425	emith_deref_op(0x89, r, rs, offs)
	426
	427	// note: don't use prefixes on this
	428	#define emith_read8_r_r_offs(r, rs, offs) do { \
	429	int r_ = r; \
	430	if (!is_abcdx(r)) \
	431	r_ = rcache_get_tmp(); \
	432	emith_deref_op(0x8a, r_, rs, offs); \
	433	if ((r) != r_) { \
	434	emith_move_r_r(r, r_); \
	435	rcache_free_tmp(r_); \
	436	} \
	437	} while (0)
	438
	439	#define emith_write8_r_r_offs(r, rs, offs) do {\
	440	int r_ = r; \
	441	if (!is_abcdx(r)) { \
	442	r_ = rcache_get_tmp(); \
	443	emith_move_r_r(r_, r); \
	444	} \
	445	emith_deref_op(0x88, r_, rs, offs); \
	446	if ((r) != r_) \
	447	rcache_free_tmp(r_); \
	448	} while (0)
	449
	450	#define emith_read16_r_r_offs(r, rs, offs) { \
	451	EMIT(0x66, u8); /* operand override */ \
	452	emith_read_r_r_offs(r, rs, offs); \
	453	}
	454
	455	#define emith_write16_r_r_offs(r, rs, offs) { \
	456	EMIT(0x66, u8); \
	457	emith_write_r_r_offs(r, rs, offs); \
	458	}
	459
	460	#define emith_ctx_read(r, offs) \
	461	emith_read_r_r_offs(r, CONTEXT_REG, offs)
	462
	463	#define emith_ctx_write(r, offs) \
	464	emith_write_r_r_offs(r, CONTEXT_REG, offs)
	465
	466	#define emith_ctx_read_multiple(r, offs, cnt, tmpr) do { \
	467	int r_ = r, offs_ = offs, cnt_ = cnt; \
	468	for (; cnt_ > 0; r_++, offs_ += 4, cnt_--) \
	469	emith_ctx_read(r_, offs_); \
	470	} while (0)
	471
	472	#define emith_ctx_write_multiple(r, offs, cnt, tmpr) do { \
	473	int r_ = r, offs_ = offs, cnt_ = cnt; \
	474	for (; cnt_ > 0; r_++, offs_ += 4, cnt_--) \
	475	emith_ctx_write(r_, offs_); \
	476	} while (0)
	477
	478	// assumes EBX is free
	479	#define emith_ret_to_ctx(offs) { \
	480	emith_pop(xBX); \
	481	emith_ctx_write(xBX, offs); \
	482	}
	483
	484	#define emith_jump(ptr) { \
	485	u32 disp = (u32)(ptr) - ((u32)tcache_ptr + 5); \
	486	EMIT_OP(0xe9); \
	487	EMIT(disp, u32); \
	488	}
	489
	490	#define emith_jump_patchable(target) \
	491	emith_jump(target)
	492
	493	#define emith_jump_cond(cond, ptr) { \
	494	u32 disp = (u32)(ptr) - ((u32)tcache_ptr + 6); \
	495	EMIT(0x0f, u8); \
	496	EMIT_OP(0x80 \| (cond)); \
	497	EMIT(disp, u32); \
	498	}
	499
	500	#define emith_jump_cond_patchable(cond, target) \
	501	emith_jump_cond(cond, target)
	502
	503	#define emith_jump_patch(ptr, target) do { \
	504	u32 disp_ = (u32)(target) - ((u32)(ptr) + 4); \
	505	u32 offs_ = ((u8 )(ptr) == 0x0f) ? 2 : 1; \
	506	EMIT_PTR((u8 *)(ptr) + offs_, disp_ - offs_, u32); \
	507	} while (0)
	508
	509	#define emith_jump_at(ptr, target) { \
	510	u32 disp_ = (u32)(target) - ((u32)(ptr) + 5); \
	511	EMIT_PTR(ptr, 0xe9, u8); \
	512	EMIT_PTR((u8 *)(ptr) + 1, disp_, u32); \
	513	}
	514
	515	#define emith_call(ptr) { \
	516	u32 disp = (u32)(ptr) - ((u32)tcache_ptr + 5); \
	517	EMIT_OP(0xe8); \
	518	EMIT(disp, u32); \
	519	}
	520
	521	#define emith_call_cond(cond, ptr) \
	522	emith_call(ptr)
	523
	524	#define emith_call_reg(r) \
	525	EMIT_OP_MODRM(0xff, 3, 2, r)
	526
	527	#define emith_call_ctx(offs) { \
	528	EMIT_OP_MODRM(0xff, 2, 2, CONTEXT_REG); \
	529	EMIT(offs, u32); \
	530	}
	531
	532	#define emith_ret() \
	533	EMIT_OP(0xc3)
	534
	535	#define emith_jump_reg(r) \
	536	EMIT_OP_MODRM(0xff, 3, 4, r)
	537
	538	#define emith_jump_ctx(offs) { \
	539	EMIT_OP_MODRM(0xff, 2, 4, CONTEXT_REG); \
	540	EMIT(offs, u32); \
	541	}
	542
	543	#define emith_push_ret()
	544
	545	#define emith_pop_and_ret() \
	546	emith_ret()
	547
	548	#define EMITH_JMP_START(cond) { \
	549	u8 *cond_ptr; \
	550	JMP8_POS(cond_ptr)
	551
	552	#define EMITH_JMP_END(cond) \
	553	JMP8_EMIT(cond, cond_ptr); \
	554	}
	555
	556	#define EMITH_JMP3_START(cond) { \
	557	u8 cond_ptr, else_ptr; \
	558	JMP8_POS(cond_ptr)
	559
	560	#define EMITH_JMP3_MID(cond) \
	561	JMP8_POS(else_ptr); \
	562	JMP8_EMIT(cond, cond_ptr);
	563
	564	#define EMITH_JMP3_END() \
	565	JMP8_EMIT_NC(else_ptr); \
	566	}
	567
	568	// "simple" jump (no more then a few insns)
	569	// ARM will use conditional instructions here
	570	#define EMITH_SJMP_START EMITH_JMP_START
	571	#define EMITH_SJMP_END EMITH_JMP_END
	572
	573	#define EMITH_SJMP3_START EMITH_JMP3_START
	574	#define EMITH_SJMP3_MID EMITH_JMP3_MID
	575	#define EMITH_SJMP3_END EMITH_JMP3_END
	576
	577	#define emith_pass_arg_r(arg, reg) { \
	578	int rd = 7; \
	579	host_arg2reg(rd, arg); \
	580	emith_move_r_r(rd, reg); \
	581	}
	582
	583	#define emith_pass_arg_imm(arg, imm) { \
	584	int rd = 7; \
	585	host_arg2reg(rd, arg); \
	586	emith_move_r_imm(rd, imm); \
	587	}
	588
	589	#define host_instructions_updated(base, end)
	590
	591	#define host_arg2reg(rd, arg) \
	592	switch (arg) { \
	593	case 0: rd = xAX; break; \
	594	case 1: rd = xDX; break; \
	595	case 2: rd = xCX; break; \
	596	}
	597
	598	/* SH2 drc specific */
	599	#define emith_sh2_drc_entry() { \
	600	emith_push(xBX); \
	601	emith_push(xBP); \
	602	emith_push(xSI); \
	603	emith_push(xDI); \
	604	}
	605
	606	#define emith_sh2_drc_exit() { \
	607	emith_pop(xDI); \
	608	emith_pop(xSI); \
	609	emith_pop(xBP); \
	610	emith_pop(xBX); \
	611	emith_ret(); \
	612	}
	613
	614	// assumes EBX is free temporary
	615	#define emith_sh2_wcall(a, tab, ret_ptr) { \
	616	int arg2_; \
	617	host_arg2reg(arg2_, 2); \
	618	emith_lsr(xBX, a, SH2_WRITE_SHIFT); \
	619	EMIT_OP_MODRM(0x8b, 0, xBX, 4); \
	620	EMIT_SIB(2, xBX, tab); /* mov ebx, [tab + ebx * 4] */ \
	621	emith_ctx_read(arg2_, offsetof(SH2, is_slave)); \
	622	emith_push_imm((long)(ret_ptr)); \
	623	emith_jump_reg(xBX); \
	624	}
	625
	626	#define emith_sh2_dtbf_loop() { \
	627	u8 jmp0; / negative cycles check */ \
	628	u8 jmp1; / unsinged overflow check */ \
	629	int cr, rn; \
	630	int tmp_ = rcache_get_tmp(); \
	631	cr = rcache_get_reg(SHR_SR, RC_GR_RMW); \
	632	rn = rcache_get_reg((op >> 8) & 0x0f, RC_GR_RMW);\
	633	emith_sub_r_imm(rn, 1); \
	634	emith_sub_r_imm(cr, (cycles+1) << 12); \
	635	cycles = 0; \
	636	emith_asr(tmp_, cr, 2+12); \
	637	JMP8_POS(jmp0); /* no negative cycles */ \
	638	emith_move_r_imm(tmp_, 0); \
	639	JMP8_EMIT(ICOND_JNS, jmp0); \
	640	emith_and_r_imm(cr, 0xffe); \
	641	emith_subf_r_r(rn, tmp_); \
	642	JMP8_POS(jmp1); /* no overflow */ \
	643	emith_neg_r(rn); /* count left */ \
	644	emith_lsl(rn, rn, 2+12); \
	645	emith_or_r_r(cr, rn); \
	646	emith_or_r_imm(cr, 1); \
	647	emith_move_r_imm(rn, 0); \
	648	JMP8_EMIT(ICOND_JA, jmp1); \
	649	rcache_free_tmp(tmp_); \
	650	}
	651
	652	#define emith_write_sr(sr, srcr) { \
	653	int tmp_ = rcache_get_tmp(); \
	654	emith_clear_msb(tmp_, srcr, 22); \
	655	emith_bic_r_imm(sr, 0x3ff); \
	656	emith_or_r_r(sr, tmp_); \
	657	rcache_free_tmp(tmp_); \
	658	}
	659
	660	#define emith_tpop_carry(sr, is_sub) \
	661	emith_lsr(sr, sr, 1)
	662
	663	#define emith_tpush_carry(sr, is_sub) \
	664	emith_adc_r_r(sr, sr)
	665
	666	/*
	667	* if Q
	668	* t = carry(Rn += Rm)
	669	* else
	670	* t = carry(Rn -= Rm)
	671	* T ^= t
	672	*/
	673	#define emith_sh2_div1_step(rn, rm, sr) { \
	674	u8 jmp0, jmp1; \
	675	int tmp_ = rcache_get_tmp(); \
	676	emith_eor_r_r(tmp_, tmp_); \
	677	emith_tst_r_imm(sr, Q); /* if (Q ^ M) */ \
	678	JMP8_POS(jmp0); /* je do_sub */ \
	679	emith_add_r_r(rn, rm); \
	680	JMP8_POS(jmp1); /* jmp done */ \
	681	JMP8_EMIT(ICOND_JE, jmp0); /* do_sub: */ \
	682	emith_sub_r_r(rn, rm); \
	683	JMP8_EMIT_NC(jmp1); /* done: */ \
	684	emith_setc(tmp_); \
	685	EMIT_OP_MODRM(0x31, 3, tmp_, sr); /* T = Q1 ^ Q2 */ \
	686	rcache_free_tmp(tmp_); \
	687	}
	688