+/*
+ * note:
+ * temp registers must be eax-edx due to use of SETcc.
+ * 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 IOP_JMP 0xeb
+#define IOP_JO 0x70
+#define IOP_JNO 0x71
+#define IOP_JB 0x72
+#define IOP_JAE 0x73
#define IOP_JE 0x74
#define IOP_JNE 0x75
#define IOP_JBE 0x76
#define IOP_JA 0x77
#define IOP_JS 0x78
#define IOP_JNS 0x79
+#define IOP_JL 0x7c
+#define IOP_JGE 0x7d
#define IOP_JLE 0x7e
+#define IOP_JG 0x7f
// unified conditions (we just use rel8 jump instructions for x86)
#define DCOND_EQ IOP_JE
#define DCOND_NE IOP_JNE
#define DCOND_MI IOP_JS // MInus
#define DCOND_PL IOP_JNS // PLus or zero
+#define DCOND_HI IOP_JA // higher (unsigned)
+#define DCOND_HS IOP_JAE // higher || same (unsigned)
+#define DCOND_LO IOP_JB // lower (unsigned)
+#define DCOND_LS IOP_JBE // lower || same (unsigned)
+#define DCOND_GE IOP_JGE // greater || equal (signed)
+#define DCOND_GT IOP_JG // greater (signed)
+#define DCOND_LE IOP_JLE // less || equal (signed)
+#define DCOND_LT IOP_JL // less (signed)
+#define DCOND_VS IOP_JO // oVerflow Set
+#define DCOND_VC IOP_JNO // oVerflow Clear
#define EMIT_PTR(ptr, val, type) \
*(type *)(ptr) = val
#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) { \
EMIT_OP(op); \
EMIT_MODRM(mod, r, rm); \
EMIT_PTR(ptr, op, 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_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)
+ 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_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) { \
EMIT_OP_MODRM(0x81, 3, op, r); \
EMIT(imm, u32); \
}
-// 2 - adc, 3 - sbb, 6 - xor, 7 - cmp
+// 2 - adc, 3 - sbb
#define emith_add_r_imm(r, imm) \
emith_arith_r_imm(0, 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) { \
EMIT_OP_MODRM(0xf7, 3, 0, r); \
EMIT(imm, u32); \
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_arith_r_imm(0, r, imm); \
+ emith_add_r_imm(r, imm); \
}
#define emith_or_r_imm_c(cond, r, imm) { \
(void)(cond); \
- emith_arith_r_imm(1, r, imm); \
+ emith_or_r_imm(r, imm); \
+}
+
+#define emith_eor_r_imm_c(cond, r, imm) { \
+ (void)(cond); \
+ emith_eor_r_imm(r, imm); \
}
#define emith_sub_r_imm_c(cond, r, imm) { \
(void)(cond); \
- emith_arith_r_imm(5, r, imm); \
+ emith_sub_r_imm(r, imm); \
+}
+
+#define emith_bic_r_imm_c(cond, r, imm) { \
+ (void)(cond); \
+ emith_bic_r_imm(r, imm); \
+}
+
+// _r_r_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
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_lsl(d, s, cnt) \
- emith_shift(4, 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) \
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(d, s1, s2) { \
+#define emith_mul_(op, dlo, dhi, s1, s2) { \
int rmr; \
- if (d != xAX) \
+ 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) \
emith_move_r_r(xAX, s1); \
rmr = s2; \
} \
- emith_push(xDX); \
- EMIT_OP_MODRM(0xf7, 3, 4, rmr); /* MUL rmr */ \
- emith_pop(xDX); \
- if (d != xAX) { \
- emith_move_r_r(d, xAX); \
+ 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
// XXX: offs is 8bit only
-#define emith_ctx_read(r, offs) { \
+#define emith_ctx_read(r, offs) do { \
EMIT_OP_MODRM(0x8b, 1, r, xBP); \
EMIT(offs, u8); /* mov tmp, [ebp+#offs] */ \
-}
+} while (0)
+
+#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(r, offs) { \
+#define emith_ctx_write(r, offs) do { \
EMIT_OP_MODRM(0x89, 1, r, xBP); \
EMIT(offs, u8); /* mov [ebp+#offs], tmp */ \
-}
+} 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)
#define emith_jump(ptr) { \
u32 disp = (u32)ptr - ((u32)tcache_ptr + 5); \
EMIT(disp, u32); \
}
-// "simple" or "short" jump
-#define EMITH_SJMP_START(cond) { \
+#define emith_call_cond(cond, ptr) \
+ emith_call(ptr)
+
+#define emith_jump_reg(r) \
+ EMIT_OP_MODRM(0xff, 3, 4, r)
+
+#define EMITH_JMP_START(cond) { \
u8 *cond_ptr; \
JMP8_POS(cond_ptr)
-#define EMITH_SJMP_END(cond) \
+#define EMITH_JMP_END(cond) \
JMP8_EMIT(cond, cond_ptr); \
}
+// "simple" jump (no more then a few insns)
+#define EMITH_SJMP_START EMITH_JMP_START
+#define EMITH_SJMP_END EMITH_JMP_END
+
#define host_arg2reg(rd, arg) \
switch (arg) { \
case 0: rd = xAX; 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); \
+ EMIT_OP(0xc3); /* ret */\
+}
+
#define emith_sh2_test_t() { \
int t = rcache_get_reg(SHR_SR, RC_GR_READ); \
- EMIT_OP_MODRM(0xf6, 3, 0, t); \
- EMIT(0x01, u8); /* test <reg>, byte 1 */ \
+ EMIT(0x66, u8); \
+ EMIT_OP_MODRM(0xf7, 3, 0, t); \
+ EMIT(0x01, u16); /* test <reg>, word 1 */ \
}
#define emith_sh2_dtbf_loop() { \
u8 *jmp0; /* negative cycles check */ \
u8 *jmp1; /* unsinged overflow check */ \
int cr, rn; \
- tmp = rcache_get_tmp(); \
+ 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); \
+ emith_asr(tmp_, cr, 2+12); \
JMP8_POS(jmp0); /* no negative cycles */ \
- emith_move_r_imm(tmp, 0); \
+ emith_move_r_imm(tmp_, 0); \
JMP8_EMIT(IOP_JNS, jmp0); \
emith_and_r_imm(cr, 0xffe); \
- emith_subf_r_r(rn, tmp); \
+ 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_imm(cr, 1); \
emith_move_r_imm(rn, 0); \
JMP8_EMIT(IOP_JA, jmp1); \
- rcache_free_tmp(tmp); \
+ rcache_free_tmp(tmp_); \
+}
+
+#define emith_write_sr(srcr) { \
+ int tmp_ = rcache_get_tmp(); \
+ int srr = rcache_get_reg(SHR_SR, RC_GR_RMW); \
+ emith_clear_msb(tmp_, srcr, 20); \
+ emith_bic_r_imm(srr, 0xfff); \
+ emith_or_r_r(srr, 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(IOP_JE, jmp0); /* do_sub: */ \
+ emith_sub_r_r(rn, rm); \
+ JMP8_EMIT(IOP_JMP, jmp1);/* done: */ \
+ emith_setc(tmp_); \
+ EMIT_OP_MODRM(0x31, 3, tmp_, sr); /* T = Q1 ^ Q2 */ \
+ rcache_free_tmp(tmp_); \
}