target/arm: Optimize aarch64 rev16

It is much shorter to reverse all 4 half-words in parallel
than extract, reverse, and deposit each in turn.

Suggested-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

1 files changed, 6 insertions, 18 deletions

diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 3fa3902..5bb0f8e 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -4043,25 +4043,13 @@ static void handle_rev16(DisasContext *s, unsigned int sf,
     TCGv_i64 tcg_rd = cpu_reg(s, rd);
     TCGv_i64 tcg_tmp = tcg_temp_new_i64();
     TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
+    TCGv_i64 mask = tcg_const_i64(sf ? 0x00ff00ff00ff00ffull : 0x00ff00ff);
 
-    tcg_gen_andi_i64(tcg_tmp, tcg_rn, 0xffff);
-    tcg_gen_bswap16_i64(tcg_rd, tcg_tmp);
-
-    tcg_gen_shri_i64(tcg_tmp, tcg_rn, 16);
-    tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
-    tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
-    tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 16, 16);
-
-    if (sf) {
-        tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
-        tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
-        tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
-        tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 32, 16);
-
-        tcg_gen_shri_i64(tcg_tmp, tcg_rn, 48);
-        tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
-        tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 48, 16);
-    }
+    tcg_gen_shri_i64(tcg_tmp, tcg_rn, 8);
+    tcg_gen_and_i64(tcg_rd, tcg_rn, mask);
+    tcg_gen_and_i64(tcg_tmp, tcg_tmp, mask);
+    tcg_gen_shli_i64(tcg_rd, tcg_rd, 8);
+    tcg_gen_or_i64(tcg_rd, tcg_rd, tcg_tmp);
 
     tcg_temp_free_i64(tcg_tmp);
 }