tcg: add ext{8,16,32}u_i{32,64} TCG ops

Currently zero extensions ops are implemented by a and op with a
constant. This is then catched in some backend, and replaced by
a zero extension instruction. While this works well on RISC
machines, this adds a useless register move on non-RISC machines.

Example on x86:
  ext16u_i32 r1, r2
is translated into
  mov    %eax,%ebx
  movzwl %bx, %ebx
while the optimized version should be:
  movzwl %ax, %ebx

This patch adds ext{8,16,32}u_i{32,64} TCG ops that can be
implemented in the backends to avoid emitting useless register
moves.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

2 files changed, 36 insertions, 3 deletions

diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index 7cb6934143..faf2e8b609 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -1189,16 +1189,22 @@ static inline void tcg_gen_ext16s_i32(TCGv_i32 ret, TCGv_i32 arg)
 #endif
 }
 
-/* These are currently just for convenience.
-   We assume a target will recognise these automatically .  */
 static inline void tcg_gen_ext8u_i32(TCGv_i32 ret, TCGv_i32 arg)
 {
+#ifdef TCG_TARGET_HAS_ext8u_i32
+    tcg_gen_op2_i32(INDEX_op_ext8u_i32, ret, arg);
+#else
     tcg_gen_andi_i32(ret, arg, 0xffu);
+#endif
 }
 
 static inline void tcg_gen_ext16u_i32(TCGv_i32 ret, TCGv_i32 arg)
 {
+#ifdef TCG_TARGET_HAS_ext16u_i32
+    tcg_gen_op2_i32(INDEX_op_ext16u_i32, ret, arg);
+#else
     tcg_gen_andi_i32(ret, arg, 0xffffu);
+#endif
 }
 
 /* Note: we assume the two high bytes are set to zero */
@@ -1358,17 +1364,29 @@ static inline void tcg_gen_ext32s_i64(TCGv_i64 ret, TCGv_i64 arg)
 
 static inline void tcg_gen_ext8u_i64(TCGv_i64 ret, TCGv_i64 arg)
 {
+#ifdef TCG_TARGET_HAS_ext8u_i64
+    tcg_gen_op2_i64(INDEX_op_ext8u_i64, ret, arg);
+#else
     tcg_gen_andi_i64(ret, arg, 0xffu);
+#endif
 }
 
 static inline void tcg_gen_ext16u_i64(TCGv_i64 ret, TCGv_i64 arg)
 {
+#ifdef TCG_TARGET_HAS_ext16u_i64
+    tcg_gen_op2_i64(INDEX_op_ext16u_i64, ret, arg);
+#else
     tcg_gen_andi_i64(ret, arg, 0xffffu);
+#endif
 }
 
 static inline void tcg_gen_ext32u_i64(TCGv_i64 ret, TCGv_i64 arg)
 {
+#ifdef TCG_TARGET_HAS_ext32u_i64
+    tcg_gen_op2_i64(INDEX_op_ext32u_i64, ret, arg);
+#else
     tcg_gen_andi_i64(ret, arg, 0xffffffffu);
+#endif
 }
 
 /* Note: we assume the target supports move between 32 and 64 bit
@@ -1382,7 +1400,7 @@ static inline void tcg_gen_trunc_i64_i32(TCGv_i32 ret, TCGv_i64 arg)
    registers */
 static inline void tcg_gen_extu_i32_i64(TCGv_i64 ret, TCGv_i32 arg)
 {
-    tcg_gen_andi_i64(ret, MAKE_TCGV_I64(GET_TCGV_I32(arg)), 0xffffffffu);
+    tcg_gen_ext32u_i64(ret, MAKE_TCGV_I64(GET_TCGV_I32(arg)));
 }
 
 /* Note: we assume the target supports move between 32 and 64 bit
diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h
index 3a095fce14..b7f3fd7e66 100644
--- a/tcg/tcg-opc.h
+++ b/tcg/tcg-opc.h
@@ -89,6 +89,12 @@ DEF2(ext8s_i32, 1, 1, 0, 0)
 #ifdef TCG_TARGET_HAS_ext16s_i32
 DEF2(ext16s_i32, 1, 1, 0, 0)
 #endif
+#ifdef TCG_TARGET_HAS_ext8u_i32
+DEF2(ext8u_i32, 1, 1, 0, 0)
+#endif
+#ifdef TCG_TARGET_HAS_ext16u_i32
+DEF2(ext16u_i32, 1, 1, 0, 0)
+#endif
 #ifdef TCG_TARGET_HAS_bswap16_i32
 DEF2(bswap16_i32, 1, 1, 0, 0)
 #endif
@@ -152,6 +158,15 @@ DEF2(ext16s_i64, 1, 1, 0, 0)
 #ifdef TCG_TARGET_HAS_ext32s_i64
 DEF2(ext32s_i64, 1, 1, 0, 0)
 #endif
+#ifdef TCG_TARGET_HAS_ext8u_i64
+DEF2(ext8u_i64, 1, 1, 0, 0)
+#endif
+#ifdef TCG_TARGET_HAS_ext16u_i64
+DEF2(ext16u_i64, 1, 1, 0, 0)
+#endif
+#ifdef TCG_TARGET_HAS_ext32u_i64
+DEF2(ext32u_i64, 1, 1, 0, 0)
+#endif
 #ifdef TCG_TARGET_HAS_bswap16_i64
 DEF2(bswap16_i64, 1, 1, 0, 0)
 #endif