[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PATCH 05/14] i386: move eflags computation closer to gen_o
From: |
Paolo Bonzini |
Subject: |
[Qemu-devel] [PATCH 05/14] i386: move eflags computation closer to gen_op_set_cc_op |
Date: |
Sat, 6 Oct 2012 14:30:12 +0200 |
In some cases this is just simple code movement, ensuring the invariant
that cpu_cc_op matches s->cc_op when calling the helpers. The next patches
need this because gen_compute_eflags and gen_compute_eflags_c will take
care of setting cpu_cc_op.
Also, for shifts, always compute EFLAGS first since it is needed whenever
the shift is non-zero, i.e. most of the time. This makes it possible
to remove some writes of CC_OP_EFLAGS to cpu_cc_op and more importantly
removes cases where s->cc_op becomes CC_OP_DYNAMIC. These are slow and
we want to avoid them: CC_OP_EFLAGS is quite efficient once we paid the
initial cost of computing the flags.
Finally, always follow gen_compute_eflags(cpu_cc_src) by setting s->cc_op
and discarding cpu_cc_dst.
Signed-off-by: Paolo Bonzini <address@hidden>
---
target-i386/translate.c | 32 +++++++++++++++-----------------
1 file modificato, 15 inserzioni(+), 17 rimozioni(-)
diff --git a/target-i386/translate.c b/target-i386/translate.c
index 38f62eb..0821468 100644
--- a/target-i386/translate.c
+++ b/target-i386/translate.c
@@ -1363,6 +1363,7 @@ static void gen_inc(DisasContext *s1, int ot, int d, int
c)
gen_op_ld_T0_A0(ot + s1->mem_index);
if (s1->cc_op != CC_OP_DYNAMIC)
gen_op_set_cc_op(s1->cc_op);
+ gen_compute_eflags_c(cpu_cc_src);
if (c > 0) {
tcg_gen_addi_tl(cpu_T[0], cpu_T[0], 1);
s1->cc_op = CC_OP_INCB + ot;
@@ -1374,7 +1375,6 @@ static void gen_inc(DisasContext *s1, int ot, int d, int
c)
gen_op_mov_reg_T0(ot, d);
else
gen_op_st_T0_A0(ot + s1->mem_index);
- gen_compute_eflags_c(cpu_cc_src);
tcg_gen_mov_tl(cpu_cc_dst, cpu_T[0]);
}
@@ -1587,14 +1587,16 @@ static void gen_rot_rm_T1(DisasContext *s, int ot, int
op1,
gen_op_mov_reg_v(ot, op1, t0);
}
- /* update eflags */
+ /* update eflags. It is needed anyway most of the time, do it always. */
if (s->cc_op != CC_OP_DYNAMIC)
gen_op_set_cc_op(s->cc_op);
+ gen_compute_eflags(cpu_cc_src);
+ tcg_gen_discard_tl(cpu_cc_dst);
+ s->cc_op = CC_OP_EFLAGS;
label2 = gen_new_label();
tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, label2);
- gen_compute_eflags(cpu_cc_src);
tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, ~(CC_O | CC_C));
tcg_gen_xor_tl(cpu_tmp0, t2, t0);
tcg_gen_lshift(cpu_tmp0, cpu_tmp0, 11 - (data_bits - 1));
@@ -1605,12 +1607,8 @@ static void gen_rot_rm_T1(DisasContext *s, int ot, int
op1,
}
tcg_gen_andi_tl(t0, t0, CC_C);
tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, t0);
-
- tcg_gen_discard_tl(cpu_cc_dst);
- tcg_gen_movi_i32(cpu_cc_op, CC_OP_EFLAGS);
-
+
gen_set_label(label2);
- s->cc_op = CC_OP_DYNAMIC; /* cannot predict flags after */
tcg_temp_free(t0);
tcg_temp_free(t1);
@@ -1674,6 +1672,9 @@ static void gen_rot_rm_im(DisasContext *s, int ot, int
op1, int op2,
gen_op_set_cc_op(s->cc_op);
gen_compute_eflags(cpu_cc_src);
+ tcg_gen_discard_tl(cpu_cc_dst);
+ s->cc_op = CC_OP_EFLAGS;
+
tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, ~(CC_O | CC_C));
tcg_gen_xor_tl(cpu_tmp0, t1, t0);
tcg_gen_lshift(cpu_tmp0, cpu_tmp0, 11 - (data_bits - 1));
@@ -1684,10 +1685,6 @@ static void gen_rot_rm_im(DisasContext *s, int ot, int
op1, int op2,
}
tcg_gen_andi_tl(t0, t0, CC_C);
tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, t0);
-
- tcg_gen_discard_tl(cpu_cc_dst);
- tcg_gen_movi_i32(cpu_cc_op, CC_OP_EFLAGS);
- s->cc_op = CC_OP_EFLAGS;
}
tcg_temp_free(t0);
@@ -1703,6 +1700,9 @@ static void gen_rotc_rm_T1(DisasContext *s, int ot, int
op1,
if (s->cc_op != CC_OP_DYNAMIC)
gen_op_set_cc_op(s->cc_op);
+ gen_compute_eflags(cpu_cc_src);
+ tcg_gen_discard_tl(cpu_cc_dst);
+ s->cc_op = CC_OP_EFLAGS;
/* load */
if (op1 == OR_TMP0)
@@ -1756,11 +1756,7 @@ static void gen_rotc_rm_T1(DisasContext *s, int ot, int
op1,
tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_cc_tmp, -1, label1);
tcg_gen_mov_tl(cpu_cc_src, cpu_cc_tmp);
- tcg_gen_discard_tl(cpu_cc_dst);
- tcg_gen_movi_i32(cpu_cc_op, CC_OP_EFLAGS);
-
gen_set_label(label1);
- s->cc_op = CC_OP_DYNAMIC; /* cannot predict flags after */
}
/* XXX: add faster immediate case */
@@ -6501,10 +6497,12 @@ static target_ulong disas_insn(DisasContext *s,
target_ulong pc_start)
if (s->cc_op != CC_OP_DYNAMIC)
gen_op_set_cc_op(s->cc_op);
gen_compute_eflags(cpu_cc_src);
+ tcg_gen_discard_tl(cpu_cc_dst);
+ s->cc_op = CC_OP_EFLAGS;
+
tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, CC_O);
tcg_gen_andi_tl(cpu_T[0], cpu_T[0], CC_S | CC_Z | CC_A | CC_P | CC_C);
tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, cpu_T[0]);
- s->cc_op = CC_OP_EFLAGS;
break;
case 0x9f: /* lahf */
if (CODE64(s) && !(s->cpuid_ext3_features & CPUID_EXT3_LAHF_LM))
--
1.7.12.1
- Re: [Qemu-devel] [PATCH 07/14] i386: add helper functions to get other flags, (continued)
- [Qemu-devel] [PATCH 06/14] i386: factor gen_op_set_cc_op/tcg_gen_discard_tl around computing flags, Paolo Bonzini, 2012/10/06
- [Qemu-devel] [PATCH 09/14] i386: do not call helper to compute ZF/SF, Paolo Bonzini, 2012/10/06
- [Qemu-devel] [PATCH 08/14] i386: do not compute eflags multiple times consecutively, Paolo Bonzini, 2012/10/06
- [Qemu-devel] [PATCH 05/14] i386: move eflags computation closer to gen_op_set_cc_op,
Paolo Bonzini <=
- [Qemu-devel] [PATCH 10/14] i386: use inverted setcond when computing NS or NZ, Paolo Bonzini, 2012/10/06
- [Qemu-devel] [PATCH 13/14] i386: optimize setbe, Paolo Bonzini, 2012/10/06
- [Qemu-devel] [PATCH 12/14] i386: change gen_setcc_slow_T0 to gen_setcc_slow, Paolo Bonzini, 2012/10/06