Move long-displacement Jcc, SETcc and CMOVcc to the new decoder.
While filling in the tables makes the code seem longer, the new
emitters are all just one line of code.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
target/i386/tcg/decode-new.h | 1 +
target/i386/tcg/translate.c | 2 +-
target/i386/tcg/decode-new.c.inc | 56 ++++++++++++++++++++++++++++++++
target/i386/tcg/emit.c.inc | 10 ++++++
4 files changed, 68 insertions(+), 1 deletion(-)
diff --git a/target/i386/tcg/decode-new.h b/target/i386/tcg/decode-new.h
index 77bb31eb143..cd7ceca21e8 100644
--- a/target/i386/tcg/decode-new.h
+++ b/target/i386/tcg/decode-new.h
@@ -106,6 +106,7 @@ typedef enum X86CPUIDFeature {
X86_FEAT_AVX2,
X86_FEAT_BMI1,
X86_FEAT_BMI2,
+ X86_FEAT_CMOV,
X86_FEAT_CMPCCXADD,
X86_FEAT_F16C,
X86_FEAT_FMA,
diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c
index b94d9504090..a80021930bf 100644
--- a/target/i386/tcg/translate.c
+++ b/target/i386/tcg/translate.c
@@ -3206,7 +3206,7 @@ static bool disas_insn(DisasContext *s, CPUState *cpu)
#ifndef CONFIG_USER_ONLY
use_new &= b <= limit;
#endif
- if (use_new && 0) {
+ if (use_new && (b >= 0x138 && b <= 0x19f)) {
disas_insn_new(s, cpu, b);
return true;
}
diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index a47ecab6dd4..7528e9e4f07 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -993,6 +993,15 @@ static const X86OpEntry opcodes_0F[256] = {
/* Incorrectly listed as Mq,Vq in the manual */
[0x17] = X86_OP_ENTRY3(VMOVHPx_st, M,q, None,None, V,dq, vex5 p_00_66),
+ [0x40] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x41] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x42] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x43] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x44] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x45] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x46] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x47] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+
[0x50] = X86_OP_ENTRY3(MOVMSK, G,y, None,None, U,x, vex7 p_00_66),
[0x51] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), /* sqrtps */
[0x52] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex4_rep5 p_00_f3), /* rsqrtps */
@@ -1020,6 +1029,24 @@ static const X86OpEntry opcodes_0F[256] = {
[0x76] = X86_OP_ENTRY3(PCMPEQD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
[0x77] = X86_OP_GROUP0(0F77),
+ [0x80] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x81] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x82] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x83] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x84] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x85] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x86] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x87] = X86_OP_ENTRYr(Jcc, J,z_f64),
+
+ [0x90] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x91] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x92] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x93] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x94] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x95] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x96] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x97] = X86_OP_ENTRYw(SETcc, E,b),
+
[0x28] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex1 p_00_66), /* MOVAPS */
[0x29] = X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1 p_00_66), /* MOVAPS */
[0x2A] = X86_OP_GROUP0(0F2A),
@@ -1032,6 +1059,15 @@ static const X86OpEntry opcodes_0F[256] = {
[0x38] = X86_OP_GROUP0(0F38),
[0x3a] = X86_OP_GROUP0(0F3A),
+ [0x48] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x49] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x4a] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x4b] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x4c] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x4d] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x4e] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+ [0x4f] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)),
+
[0x58] = X86_OP_ENTRY3(VADD, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
[0x59] = X86_OP_ENTRY3(VMUL, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
[0x5a] = X86_OP_GROUP0(0F5A),
@@ -1057,6 +1093,24 @@ static const X86OpEntry opcodes_0F[256] = {
[0x7e] = X86_OP_GROUP0(0F7E),
[0x7f] = X86_OP_GROUP0(0F7F),
+ [0x88] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x89] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x8a] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x8b] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x8c] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x8d] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x8e] = X86_OP_ENTRYr(Jcc, J,z_f64),
+ [0x8f] = X86_OP_ENTRYr(Jcc, J,z_f64),
+
+ [0x98] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x99] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x9a] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x9b] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x9c] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x9d] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x9e] = X86_OP_ENTRYw(SETcc, E,b),
+ [0x9f] = X86_OP_ENTRYw(SETcc, E,b),
+
[0xae] = X86_OP_GROUP0(group15),
[0xc2] = X86_OP_ENTRY4(VCMP, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
@@ -1918,6 +1972,8 @@ static bool has_cpuid_feature(DisasContext *s, X86CPUIDFeature cpuid)
switch (cpuid) {
case X86_FEAT_None:
return true;
+ case X86_FEAT_CMOV:
+ return (s->cpuid_features & CPUID_CMOV);
case X86_FEAT_F16C:
return (s->cpuid_ext_features & CPUID_EXT_F16C);
case X86_FEAT_FMA:
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index ffe458b80f9..a48ff1536a4 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -1386,6 +1386,11 @@ static void gen_CMC(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
tcg_gen_xori_tl(cpu_cc_src, cpu_cc_src, CC_C);
}
+static void gen_CMOVcc(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
+{
+ gen_cmovcc1(s, decode->b & 0xf, s->T0, s->T1);
+}
+
static void gen_CMPccXADD(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGLabel *label_top = gen_new_label();
@@ -3298,6 +3303,11 @@ static void gen_SCAS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
}
}
+static void gen_SETcc(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
+{
+ gen_setcc1(s, decode->b & 0xf, s->T0);
+}
+
static void gen_SHA1NEXTE(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_helper_sha1nexte(OP_PTR0, OP_PTR1, OP_PTR2);
--
2.45.0