[PATCH v2 36/37] target/i386: move 3DNow to the new decoder

[Paolo Bonzini]

This adds another kind of weirdness when you thought you had seen it all:
an opcode byte that comes _after_ the address, not before.  It's not
worth adding a new X86_SPECIAL_* constant for it, but it's actually
not unlike VCMP; so, forgive me for exploiting the similarity and just
deciding to dispatch to the right gen_helper_* call in a single code
generation function.

In fact, the old decoder had a bug where s->rip_offset should have
been set to 1 for 3DNow! instructions, and it's fixed now.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/tcg/decode-new.c.inc | 10 +++++
 target/i386/tcg/decode-new.h     |  1 +
 target/i386/tcg/emit.c.inc       | 61 +++++++++++++++++++++++++++
 target/i386/tcg/translate.c      | 71 +-------------------------------
 4 files changed, 74 insertions(+), 69 deletions(-)

diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index efe52504f2..e4de3bac0a 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -779,6 +779,14 @@ static void decode_0FE6(DisasContext *s, CPUX86State *env, 
X86OpEntry *entry, ui
 }
 
 static const X86OpEntry opcodes_0F[256] = {
+    [0x0E] = X86_OP_ENTRY0(EMMS,                              cpuid(3DNOW)), 
/* femms */
+    /*
+     * 3DNow!'s opcode byte comes *after* modrm and displacements, making it
+     * more like an Ib operand.  Dispatch to the right helper in a single gen_*
+     * function.
+     */
+    [0x0F] = X86_OP_ENTRY3(3dnow,       P,q, Q,q, I,b,        cpuid(3DNOW)),
+
     [0x10] = X86_OP_GROUP0(0F10),
     [0x11] = X86_OP_GROUP0(0F11),
     [0x12] = X86_OP_GROUP0(0F12),
@@ -1364,6 +1372,8 @@ static bool has_cpuid_feature(DisasContext *s, 
X86CPUIDFeature cpuid)
     case X86_FEAT_AVX:
         return (s->cpuid_ext_features & CPUID_EXT_AVX);
 
+    case X86_FEAT_3DNOW:
+        return (s->cpuid_ext2_features & CPUID_EXT2_3DNOW);
     case X86_FEAT_SSE4A:
         return (s->cpuid_ext3_features & CPUID_EXT3_SSE4A);
 
diff --git a/target/i386/tcg/decode-new.h b/target/i386/tcg/decode-new.h
index c74550e32b..789e1a3e1a 100644
--- a/target/i386/tcg/decode-new.h
+++ b/target/i386/tcg/decode-new.h
@@ -96,6 +96,7 @@ typedef enum X86OpSize {
 
 typedef enum X86CPUIDFeature {
     X86_FEAT_None,
+    X86_FEAT_3DNOW,
     X86_FEAT_ADX,
     X86_FEAT_AES,
     X86_FEAT_AVX,
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index be7e3797e8..ef4e3478fa 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -19,6 +19,7 @@
  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  */
 
+typedef void (*SSEFunc_0_epp)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b);
 typedef void (*SSEFunc_0_epppti)(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b,
                                  TCGv_ptr reg_c, TCGv a0, TCGv_i32 scale);
 
@@ -326,6 +327,66 @@ static void gen_store_sse(DisasContext *s, X86DecodedInsn 
*decode, int src_ofs)
     }
 }
 
+static void gen_helper_pavgusb(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b)
+{
+    gen_helper_pavgb_mmx(env, reg_a, reg_a, reg_b);
+}
+
+#define FN_3DNOW_MOVE ((SSEFunc_0_epp) (uintptr_t) 1)
+static const SSEFunc_0_epp fns_3dnow[] = {
+    [0x0c] = gen_helper_pi2fw,
+    [0x0d] = gen_helper_pi2fd,
+    [0x1c] = gen_helper_pf2iw,
+    [0x1d] = gen_helper_pf2id,
+    [0x8a] = gen_helper_pfnacc,
+    [0x8e] = gen_helper_pfpnacc,
+    [0x90] = gen_helper_pfcmpge,
+    [0x94] = gen_helper_pfmin,
+    [0x96] = gen_helper_pfrcp,
+    [0x97] = gen_helper_pfrsqrt,
+    [0x9a] = gen_helper_pfsub,
+    [0x9e] = gen_helper_pfadd,
+    [0xa0] = gen_helper_pfcmpgt,
+    [0xa4] = gen_helper_pfmax,
+    [0xa6] = FN_3DNOW_MOVE, /* PFRCPIT1; no need to actually increase 
precision */
+    [0xa7] = FN_3DNOW_MOVE, /* PFRSQIT1 */
+    [0xb6] = FN_3DNOW_MOVE, /* PFRCPIT2 */
+    [0xaa] = gen_helper_pfsubr,
+    [0xae] = gen_helper_pfacc,
+    [0xb0] = gen_helper_pfcmpeq,
+    [0xb4] = gen_helper_pfmul,
+    [0xb7] = gen_helper_pmulhrw_mmx,
+    [0xbb] = gen_helper_pswapd,
+    [0xbf] = gen_helper_pavgusb,
+};
+
+static void gen_3dnow(DisasContext *s, CPUX86State *env, X86DecodedInsn 
*decode)
+{
+    uint8_t b = decode->immediate;
+    SSEFunc_0_epp fn = b < ARRAY_SIZE(fns_3dnow) ? fns_3dnow[b] : NULL;
+
+    if (!fn) {
+        gen_illegal_opcode(s);
+        return;
+    }
+    if (s->flags & HF_TS_MASK) {
+        gen_NM_exception(s);
+        return;
+    }
+    if (s->flags & HF_EM_MASK) {
+        gen_illegal_opcode(s);
+        return;
+    }
+
+    gen_helper_enter_mmx(cpu_env);
+    if (fn == FN_3DNOW_MOVE) {
+       tcg_gen_ld_i64(s->tmp1_i64, cpu_env, decode->op[1].offset);
+       tcg_gen_st_i64(s->tmp1_i64, cpu_env, decode->op[0].offset);
+    } else {
+       fn(cpu_env, OP_PTR0, OP_PTR1);
+    }
+}
+
 /*
  * 00 = v*ps Vps, Hps, Wpd
  * 66 = v*pd Vpd, Hpd, Wps
diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c
index 2d93e52ae5..627d321f00 100644
--- a/target/i386/tcg/translate.c
+++ b/target/i386/tcg/translate.c
@@ -2930,7 +2930,6 @@ static bool first = true; static unsigned long limit;
 #define SSE_OPF_CMP       (1 << 1) /* does not write for first operand */
 #define SSE_OPF_BLENDV    (1 << 2) /* blendv* instruction */
 #define SSE_OPF_SPECIAL   (1 << 3) /* magic */
-#define SSE_OPF_3DNOW     (1 << 4) /* 3DNow! instruction */
 #define SSE_OPF_MMX       (1 << 5) /* MMX/integer/AVX2 instruction */
 #define SSE_OPF_SCALAR    (1 << 6) /* Has SSE scalar variants */
 #define SSE_OPF_SHUF      (1 << 9) /* pshufx/shufpx */
@@ -2964,13 +2963,9 @@ struct SSEOpHelper_table1 {
     SSEFuncs fn[4];
 };
 
-#define SSE_3DNOW { SSE_OPF_3DNOW }
 #define SSE_SPECIAL { SSE_OPF_SPECIAL }
 
 static const struct SSEOpHelper_table1 sse_op_table1[256] = {
-    /* 3DNow! extensions */
-    [0x0e] = SSE_SPECIAL, /* femms */
-    [0x0f] = SSE_3DNOW, /* pf... (sse_op_table5) */
     /* pure SSE operations */
     [0x10] = SSE_SPECIAL, /* movups, movupd, movss, movsd */
     [0x11] = SSE_SPECIAL, /* movups, movupd, movss, movsd */
@@ -3179,38 +3174,6 @@ static const SSEFunc_0_eppp sse_op_table4[8][4] = {
 };
 #undef SSE_CMP
 
-static void gen_helper_pavgusb(TCGv_ptr env, TCGv_ptr reg_a, TCGv_ptr reg_b)
-{
-    gen_helper_pavgb_mmx(env, reg_a, reg_a, reg_b);
-}
-
-static const SSEFunc_0_epp sse_op_table5[256] = {
-    [0x0c] = gen_helper_pi2fw,
-    [0x0d] = gen_helper_pi2fd,
-    [0x1c] = gen_helper_pf2iw,
-    [0x1d] = gen_helper_pf2id,
-    [0x8a] = gen_helper_pfnacc,
-    [0x8e] = gen_helper_pfpnacc,
-    [0x90] = gen_helper_pfcmpge,
-    [0x94] = gen_helper_pfmin,
-    [0x96] = gen_helper_pfrcp,
-    [0x97] = gen_helper_pfrsqrt,
-    [0x9a] = gen_helper_pfsub,
-    [0x9e] = gen_helper_pfadd,
-    [0xa0] = gen_helper_pfcmpgt,
-    [0xa4] = gen_helper_pfmax,
-    [0xa6] = gen_helper_movq, /* pfrcpit1; no need to actually increase 
precision */
-    [0xa7] = gen_helper_movq, /* pfrsqit1 */
-    [0xaa] = gen_helper_pfsubr,
-    [0xae] = gen_helper_pfacc,
-    [0xb0] = gen_helper_pfcmpeq,
-    [0xb4] = gen_helper_pfmul,
-    [0xb6] = gen_helper_movq, /* pfrcpit2 */
-    [0xb7] = gen_helper_pmulhrw_mmx,
-    [0xbb] = gen_helper_pswapd,
-    [0xbf] = gen_helper_pavgusb,
-};
-
 struct SSEOpHelper_table6 {
     SSEFuncs fn[2];
     uint32_t ext_mask;
@@ -3363,7 +3326,7 @@ static void gen_sse(CPUX86State *env, DisasContext *s, 
int b,
         b1 = 0;
     sse_op_flags = sse_op_table1[b].flags;
     sse_op_fn = sse_op_table1[b].fn[b1];
-    if ((sse_op_flags & (SSE_OPF_SPECIAL | SSE_OPF_3DNOW)) == 0
+    if ((sse_op_flags & SSE_OPF_SPECIAL) == 0
             && !sse_op_fn.op1) {
         goto unknown_op;
     }
@@ -3377,11 +3340,6 @@ static void gen_sse(CPUX86State *env, DisasContext *s, 
int b,
             is_xmm = 1;
         }
     }
-    if (sse_op_flags & SSE_OPF_3DNOW) {
-        if (!(s->cpuid_ext2_features & CPUID_EXT2_3DNOW)) {
-            goto illegal_op;
-        }
-    }
     /* simple MMX/SSE operation */
     if (s->flags & HF_TS_MASK) {
         gen_exception(s, EXCP07_PREX, pc_start - s->cs_base);
@@ -3397,15 +3355,6 @@ static void gen_sse(CPUX86State *env, DisasContext *s, 
int b,
         && (b != 0x38 && b != 0x3a)) {
         goto unknown_op;
     }
-    if (b == 0x0e) {
-        if (!(s->cpuid_ext2_features & CPUID_EXT2_3DNOW)) {
-            /* If we were fully decoding this we might use illegal_op.  */
-            goto unknown_op;
-        }
-        /* femms */
-        gen_helper_emms(cpu_env);
-        return;
-    }
     if (b == 0x77) {
         /* emms */
         gen_helper_emms(cpu_env);
@@ -4563,18 +4512,6 @@ static void gen_sse(CPUX86State *env, DisasContext *s, 
int b,
                 rm = (modrm & 7);
                 op2_offset = offsetof(CPUX86State,fpregs[rm].mmx);
             }
-            if (sse_op_flags & SSE_OPF_3DNOW) {
-                /* 3DNow! data insns */
-                val = x86_ldub_code(env, s);
-                SSEFunc_0_epp op_3dnow = sse_op_table5[val];
-                if (!op_3dnow) {
-                    goto unknown_op;
-                }
-                tcg_gen_addi_ptr(s->ptr0, cpu_env, op1_offset);
-                tcg_gen_addi_ptr(s->ptr1, cpu_env, op2_offset);
-                op_3dnow(cpu_env, s->ptr0, s->ptr1);
-                return;
-            }
         }
 
 
@@ -4696,7 +4633,7 @@ static target_ulong disas_insn(DisasContext *s, CPUState 
*cpu)
 #endif
         if (use_new &&
             (b == 0x138 || b == 0x13a ||
-             (b >= 0x110 && b <= 0x117) ||
+             (b >= 0x10e && b <= 0x117) ||
              (b >= 0x128 && b <= 0x12f) ||
              (b >= 0x150 && b <= 0x17f) ||
              b == 0x1c2 || (b >= 0x1c4 && b <= 0x1c6) ||
@@ -8533,10 +8470,6 @@ static target_ulong disas_insn(DisasContext *s, CPUState 
*cpu)
 
         set_cc_op(s, CC_OP_POPCNT);
         break;
-    case 0x10e ... 0x10f:
-        /* 3DNow! instructions, ignore prefixes */
-        s->prefix &= ~(PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA);
-        /* fall through */
     case 0x110 ... 0x117:
     case 0x128 ... 0x12f:
     case 0x138 ... 0x13a:
-- 
2.37.2