[PULL 49/53] target/i386: implement VLDMXCSR/VSTMXCSR

[Paolo Bonzini <pbonzini@redhat.com>]

These are exactly the same as the non-VEX version, but one has to be careful
that only VEX.L=0 is allowed.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/tcg/decode-new.c.inc | 25 +++++++++++++++++++++++++
 target/i386/tcg/emit.c.inc       | 20 ++++++++++++++++++++
 2 files changed, 45 insertions(+)

diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index a5d5428260..5f4268bdfb 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -82,6 +82,10 @@
 
 #define X86_OP_ENTRY2(op, op0, s0, op1, s1, ...)                  \
     X86_OP_ENTRY3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__)
+#define X86_OP_ENTRYw(op, op0, s0, ...)                           \
+    X86_OP_ENTRY3(op, op0, s0, None, None, None, None, ## __VA_ARGS__)
+#define X86_OP_ENTRYr(op, op0, s0, ...)                           \
+    X86_OP_ENTRY3(op, None, None, None, None, op0, s0, ## __VA_ARGS__)
 #define X86_OP_ENTRY0(op, ...)                                    \
     X86_OP_ENTRY3(op, None, None, None, None, None, None, ## __VA_ARGS__)
 
@@ -149,6 +153,25 @@ static inline const X86OpEntry *decode_by_prefix(DisasContext *s, const X86OpEnt
     }
 }
 
+static void decode_group15(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b)
+{
+    /* only includes ldmxcsr and stmxcsr, because they have AVX variants.  */
+    static const X86OpEntry group15_reg[8] = {
+    };
+
+    static const X86OpEntry group15_mem[8] = {
+        [2] = X86_OP_ENTRYr(LDMXCSR,    E,d, vex5),
+        [3] = X86_OP_ENTRYw(STMXCSR,    E,d, vex5),
+    };
+
+    uint8_t modrm = get_modrm(s, env);
+    if ((modrm >> 6) == 3) {
+        *entry = group15_reg[(modrm >> 3) & 7];
+    } else {
+        *entry = group15_mem[(modrm >> 3) & 7];
+    }
+}
+
 static void decode_group17(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b)
 {
     static const X86GenFunc group17_gen[8] = {
@@ -830,6 +853,8 @@ static const X86OpEntry opcodes_0F[256] = {
     [0x7e] = X86_OP_GROUP0(0F7E),
     [0x7f] = X86_OP_GROUP0(0F7F),
 
+    [0xae] = X86_OP_GROUP0(group15),
+
     [0xc2] = X86_OP_ENTRY4(VCMP,       V,x, H,x, W,x,       vex2_rep3 p_00_66_f3_f2),
     [0xc4] = X86_OP_ENTRY4(PINSRW,     V,dq,H,dq,E,w,       vex5 mmx p_00_66),
     [0xc5] = X86_OP_ENTRY3(PEXTRW,     G,d, U,dq,I,b,       vex5 mmx p_00_66),
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index 266e7499ad..d7c2290db1 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -1099,6 +1099,16 @@ static void gen_INSERTQ_r(DisasContext *s, CPUX86State *env, X86DecodedInsn *dec
     gen_helper_insertq_r(cpu_env, OP_PTR0, OP_PTR2);
 }
 
+static void gen_LDMXCSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
+{
+    if (s->vex_l) {
+        gen_illegal_opcode(s);
+        return;
+    }
+    tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T1);
+    gen_helper_ldmxcsr(cpu_env, s->tmp2_i32);
+}
+
 static void gen_MASKMOV(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
 {
     tcg_gen_mov_tl(s->A0, cpu_regs[R_EDI]);
@@ -1688,6 +1698,16 @@ static void gen_VAESKEYGEN(DisasContext *s, CPUX86State *env, X86DecodedInsn *de
     gen_helper_aeskeygenassist_xmm(cpu_env, OP_PTR0, OP_PTR1, imm);
 }
 
+static void gen_STMXCSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
+{
+    if (s->vex_l) {
+        gen_illegal_opcode(s);
+        return;
+    }
+    gen_helper_update_mxcsr(cpu_env);
+    tcg_gen_ld32u_tl(s->T0, cpu_env, offsetof(CPUX86State, mxcsr));
+}
+
 static void gen_VAESIMC(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
 {
     assert(!s->vex_l);
-- 
2.37.3