3 files changed, 33 insertions, 12 deletions

diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index 27e9026..f56e59c 100644
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -171,6 +171,13 @@
   BUILTIN_VQN (TERNOP, raddhn2, 0, NONE)
   BUILTIN_VQN (TERNOP, rsubhn2, 0, NONE)
 
+  /* Implemented by aarch64_<us>xtl<mode>.  */
+  BUILTIN_VQN (UNOP, sxtl, 0, NONE)
+  BUILTIN_VQN (UNOPU, uxtl, 0, NONE)
+
+  /* Implemented by aarch64_xtn<mode>.  */
+  BUILTIN_VQN (UNOP, xtn, 0, NONE)
+
   BUILTIN_VSQN_HSDI (UNOPUS, sqmovun, 0, ALL)
   /* Implemented by aarch64_<sur>qmovn<mode>.  */
   BUILTIN_VSQN_HSDI (UNOP, sqmovn, 0, ALL)
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index 0827f0e..4b869de 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -7301,6 +7301,20 @@
   [(set_attr "type" "neon_shift_imm_long")]
 )
 
+(define_expand "aarch64_<su>xtl<mode>"
+  [(set (match_operand:VQN 0 "register_operand" "=w")
+	(ANY_EXTEND:VQN (match_operand:<VNARROWQ> 1 "register_operand" "w")))]
+  "TARGET_SIMD"
+  ""
+)
+
+(define_expand "aarch64_xtn<mode>"
+  [(set (match_operand:<VNARROWQ> 0 "register_operand" "=w")
+	(truncate:<VNARROWQ> (match_operand:VQN 1 "register_operand" "w")))]
+  "TARGET_SIMD"
+  ""
+)
+
 ;; Truncate a 128-bit integer vector to a 64-bit vector.
 (define_insn "trunc<mode><Vnarrowq>2"
   [(set (match_operand:<VNARROWQ> 0 "register_operand" "=w")
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 6095c0d..46331ae 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -8709,42 +8709,42 @@ __extension__ extern __inline int16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovl_s8 (int8x8_t __a)
 {
-  return __builtin_convertvector (__a, int16x8_t);
+  return __builtin_aarch64_sxtlv8hi (__a);
 }
 
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovl_s16 (int16x4_t __a)
 {
-  return __builtin_convertvector (__a, int32x4_t);
+  return __builtin_aarch64_sxtlv4si (__a);
 }
 
 __extension__ extern __inline int64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovl_s32 (int32x2_t __a)
 {
-  return __builtin_convertvector (__a, int64x2_t);
+  return __builtin_aarch64_sxtlv2di (__a);
 }
 
 __extension__ extern __inline uint16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovl_u8 (uint8x8_t __a)
 {
-  return __builtin_convertvector (__a, uint16x8_t);
+  return __builtin_aarch64_uxtlv8hi_uu (__a);
 }
 
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovl_u16 (uint16x4_t __a)
 {
-  return __builtin_convertvector (__a, uint32x4_t);
+  return __builtin_aarch64_uxtlv4si_uu (__a);
 }
 
 __extension__ extern __inline uint64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovl_u32 (uint32x2_t __a)
 {
-  return __builtin_convertvector (__a, uint64x2_t);
+  return __builtin_aarch64_uxtlv2di_uu (__a);
 }
 
 __extension__ extern __inline int8x16_t
@@ -8796,42 +8796,42 @@ __extension__ extern __inline int8x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovn_s16 (int16x8_t __a)
 {
-  return __builtin_convertvector (__a, int8x8_t);
+  return __builtin_aarch64_xtnv8hi (__a);
 }
 
 __extension__ extern __inline int16x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovn_s32 (int32x4_t __a)
 {
-  return __builtin_convertvector (__a, int16x4_t);
+  return __builtin_aarch64_xtnv4si (__a);
 }
 
 __extension__ extern __inline int32x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovn_s64 (int64x2_t __a)
 {
-  return __builtin_convertvector (__a, int32x2_t);
+  return __builtin_aarch64_xtnv2di (__a);
 }
 
 __extension__ extern __inline uint8x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovn_u16 (uint16x8_t __a)
 {
-  return __builtin_convertvector (__a, uint8x8_t);
+  return (uint8x8_t)__builtin_aarch64_xtnv8hi ((int16x8_t) __a);
 }
 
 __extension__ extern __inline uint16x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovn_u32 (uint32x4_t __a)
 {
-  return __builtin_convertvector (__a, uint16x4_t);
+  return (uint16x4_t) __builtin_aarch64_xtnv4si ((int32x4_t )__a);
 }
 
 __extension__ extern __inline uint32x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 vmovn_u64 (uint64x2_t __a)
 {
-  return __builtin_convertvector (__a, uint32x2_t);
+  return (uint32x2_t) __builtin_aarch64_xtnv2di ((int64x2_t) __a);
 }
 
 #define vmull_high_lane_s16(a, b, c)                                    \