target-i386: move eflags computation closer to gen_op_set_cc_op

This ensures 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.

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 <pbonzini@redhat.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
This commit is contained in:
Paolo Bonzini 2012-10-06 00:18:55 +02:00 committed by Richard Henderson
parent 52320e15db
commit 0ff6addd92

View File

@ -1597,14 +1597,16 @@ static void gen_rot_rm_T1(DisasContext *s, int ot, int op1,
gen_op_mov_reg_v(ot, op1, t0); 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) if (s->cc_op != CC_OP_DYNAMIC)
gen_op_set_cc_op(s->cc_op); 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(); label2 = gen_new_label();
tcg_gen_brcondi_tl(TCG_COND_EQ, t1, 0, label2); 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_andi_tl(cpu_cc_src, cpu_cc_src, ~(CC_O | CC_C));
tcg_gen_xor_tl(cpu_tmp0, t2, t0); tcg_gen_xor_tl(cpu_tmp0, t2, t0);
tcg_gen_lshift(cpu_tmp0, cpu_tmp0, 11 - (data_bits - 1)); tcg_gen_lshift(cpu_tmp0, cpu_tmp0, 11 - (data_bits - 1));
@ -1615,12 +1617,8 @@ static void gen_rot_rm_T1(DisasContext *s, int ot, int op1,
} }
tcg_gen_andi_tl(t0, t0, CC_C); tcg_gen_andi_tl(t0, t0, CC_C);
tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, t0); 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); gen_set_label(label2);
s->cc_op = CC_OP_DYNAMIC; /* cannot predict flags after */
tcg_temp_free(t0); tcg_temp_free(t0);
tcg_temp_free(t1); tcg_temp_free(t1);
@ -1684,6 +1682,9 @@ static void gen_rot_rm_im(DisasContext *s, int ot, int op1, int op2,
gen_op_set_cc_op(s->cc_op); gen_op_set_cc_op(s->cc_op);
gen_compute_eflags(cpu_cc_src); 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_andi_tl(cpu_cc_src, cpu_cc_src, ~(CC_O | CC_C));
tcg_gen_xor_tl(cpu_tmp0, t1, t0); tcg_gen_xor_tl(cpu_tmp0, t1, t0);
tcg_gen_lshift(cpu_tmp0, cpu_tmp0, 11 - (data_bits - 1)); tcg_gen_lshift(cpu_tmp0, cpu_tmp0, 11 - (data_bits - 1));
@ -1694,10 +1695,6 @@ static void gen_rot_rm_im(DisasContext *s, int ot, int op1, int op2,
} }
tcg_gen_andi_tl(t0, t0, CC_C); tcg_gen_andi_tl(t0, t0, CC_C);
tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, t0); 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); tcg_temp_free(t0);