tcg: Use tcg_out_dupi_vec from temp_load

Having dupi pass though movi is confusing and arguably wrong.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

4 files changed, 46 insertions, 34 deletions

diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index ab199b143f..a2a588e3aa 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -1011,13 +1011,6 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
     case TCG_TYPE_I64:
         tcg_debug_assert(rd < 32);
         break;
-
-    case TCG_TYPE_V64:
-    case TCG_TYPE_V128:
-        tcg_debug_assert(rd >= 32);
-        tcg_out_dupi_vec(s, type, rd, value);
-        return;
-
     default:
         g_assert_not_reached();
     }
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 46e856f442..35554fd1e8 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -975,30 +975,32 @@ static void tcg_out_dupi_vec(TCGContext *s, TCGType type,
     }
 }
 
-static void tcg_out_movi(TCGContext *s, TCGType type,
-                         TCGReg ret, tcg_target_long arg)
+static void tcg_out_movi_vec(TCGContext *s, TCGType type,
+                             TCGReg ret, tcg_target_long arg)
 {
-    tcg_target_long diff;
-
-    switch (type) {
-    case TCG_TYPE_I32:
-#if TCG_TARGET_REG_BITS == 64
-    case TCG_TYPE_I64:
-#endif
-        if (ret < 16) {
-            break;
-        }
-        /* fallthru */
-    case TCG_TYPE_V64:
-    case TCG_TYPE_V128:
-    case TCG_TYPE_V256:
-        tcg_debug_assert(ret >= 16);
-        tcg_out_dupi_vec(s, type, ret, arg);
+    if (arg == 0) {
+        tcg_out_vex_modrm(s, OPC_PXOR, ret, ret, ret);
+        return;
+    }
+    if (arg == -1) {
+        tcg_out_vex_modrm(s, OPC_PCMPEQB, ret, ret, ret);
         return;
-    default:
-        g_assert_not_reached();
     }
 
+    int rexw = (type == TCG_TYPE_I32 ? 0 : P_REXW);
+    tcg_out_vex_modrm_pool(s, OPC_MOVD_VyEy + rexw, ret);
+    if (TCG_TARGET_REG_BITS == 64) {
+        new_pool_label(s, arg, R_386_PC32, s->code_ptr - 4, -4);
+    } else {
+        new_pool_label(s, arg, R_386_32, s->code_ptr - 4, 0);
+    }
+}
+
+static void tcg_out_movi_int(TCGContext *s, TCGType type,
+                             TCGReg ret, tcg_target_long arg)
+{
+    tcg_target_long diff;
+
     if (arg == 0) {
         tgen_arithr(s, ARITH_XOR, ret, ret);
         return;
@@ -1027,6 +1029,25 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
     tcg_out64(s, arg);
 }
 
+static void tcg_out_movi(TCGContext *s, TCGType type,
+                         TCGReg ret, tcg_target_long arg)
+{
+    switch (type) {
+    case TCG_TYPE_I32:
+#if TCG_TARGET_REG_BITS == 64
+    case TCG_TYPE_I64:
+#endif
+        if (ret < 16) {
+            tcg_out_movi_int(s, type, ret, arg);
+        } else {
+            tcg_out_movi_vec(s, type, ret, arg);
+        }
+        break;
+    default:
+        g_assert_not_reached();
+    }
+}
+
 static inline void tcg_out_pushi(TCGContext *s, tcg_target_long val)
 {
     if (val == (int8_t)val) {
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index 19a4a12f15..a3f1bd41cd 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -987,12 +987,6 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg ret,
         tcg_out_movi_int(s, type, ret, arg, false);
         break;
 
-    case TCG_TYPE_V64:
-    case TCG_TYPE_V128:
-        tcg_debug_assert(ret >= TCG_REG_V0);
-        tcg_out_dupi_vec(s, type, ret, arg);
-        break;
-
     default:
         g_assert_not_reached();
     }
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 472bf1755b..ded3c928e3 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -3387,7 +3387,11 @@ static void temp_load(TCGContext *s, TCGTemp *ts, TCGRegSet desired_regs,
     case TEMP_VAL_CONST:
         reg = tcg_reg_alloc(s, desired_regs, allocated_regs,
                             preferred_regs, ts->indirect_base);
-        tcg_out_movi(s, ts->type, reg, ts->val);
+        if (ts->type <= TCG_TYPE_I64) {
+            tcg_out_movi(s, ts->type, reg, ts->val);
+        } else {
+            tcg_out_dupi_vec(s, ts->type, reg, ts->val);
+        }
         ts->mem_coherent = 0;
         break;
     case TEMP_VAL_MEM: