From: Richard Henderson <richard.henderson@linaro.org>
Date: Wed, 4 Aug 2021 05:26:30 +0000 (+0300)
Subject: tcg: Canonicalize alignment flags in MemOp
X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=76e366e728549b3324cc2dee6745d6a4f1af18e6;p=qmiga%2Fqemu.git

tcg: Canonicalize alignment flags in MemOp

Having observed e.g. al8+leq in dumps, canonicalize to al+leq.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---

diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index b1cfd36f29..61b492d89f 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -2765,7 +2765,12 @@ void tcg_gen_lookup_and_goto_ptr(void)
 static inline MemOp tcg_canonicalize_memop(MemOp op, bool is64, bool st)
 {
     /* Trigger the asserts within as early as possible.  */
-    (void)get_alignment_bits(op);
+    unsigned a_bits = get_alignment_bits(op);
+
+    /* Prefer MO_ALIGN+MO_XX over MO_ALIGN_XX+MO_XX */
+    if (a_bits == (op & MO_SIZE)) {
+        op = (op & ~MO_AMASK) | MO_ALIGN;
+    }
 
     switch (op & MO_SIZE) {
     case MO_8: