[Qemu-devel] [PATCH v2 11/68] target/arm: Simplify op_smlawx for SMLAW*

[Richard Henderson]

By shifting the 16-bit input left by 16, we can align the desired
portion of the 48-bit product and use tcg_gen_muls2_i32.

Signed-off-by: Richard Henderson <address@hidden>
---
 target/arm/translate.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/target/arm/translate.c b/target/arm/translate.c
index 8557ef831f..9a2fb7d3aa 100644
--- a/target/arm/translate.c
+++ b/target/arm/translate.c
@@ -8213,7 +8213,6 @@ DO_SMLAX(SMLALTT, 2, 1, 1)
 static bool op_smlawx(DisasContext *s, arg_rrrr *a, bool add, bool mt)
 {
     TCGv_i32 t0, t1;
-    TCGv_i64 t64;
 
     if (!ENABLE_ARCH_5TE) {
         return false;
@@ -8221,16 +8220,17 @@ static bool op_smlawx(DisasContext *s, arg_rrrr *a, 
bool add, bool mt)
 
     t0 = load_reg(s, a->rn);
     t1 = load_reg(s, a->rm);
+    /*
+     * Since the nominal result is product<47:16>, shift the 16-bit
+     * input up by 16 bits, so that the result is at product<63:32>.
+     */
     if (mt) {
-        tcg_gen_sari_i32(t1, t1, 16);
+        tcg_gen_andi_i32(t1, t1, 0xffff0000);
     } else {
-        gen_sxth(t1);
+        tcg_gen_shli_i32(t1, t1, 16);
     }
-    t64 = gen_muls_i64_i32(t0, t1);
-    tcg_gen_shri_i64(t64, t64, 16);
-    t1 = tcg_temp_new_i32();
-    tcg_gen_extrl_i64_i32(t1, t64);
-    tcg_temp_free_i64(t64);
+    tcg_gen_muls2_i32(t0, t1, t0, t1);
+    tcg_temp_free_i32(t0);
     if (add) {
         t0 = load_reg(s, a->ra);
         gen_helper_add_setq(t1, cpu_env, t1, t0);
-- 
2.17.1