[AArch64] [3/4 Fix vtbx1]Implement bsl intrinsics using builtins

gcc/
	* config/aarch64/aarch64-builtins.c
	(aarch64_types_bsl_p_qualifiers): New.
	(aarch64_types_bsl_s_qualifiers): Likewise.
	(aarch64_types_bsl_u_qualifiers): Likewise.
	(TYPES_BSL_P): Likewise.
	(TYPES_BSL_S): Likewise.
	(TYPES_BSL_U): Likewise.
	(BUILTIN_VALLDIF): Likewise.
	(BUILTIN_VDQQH): Likewise.
	* config/aarch64/aarch64-simd-builtins.def (simd_bsl): New.
	* config/aarch64/aarch64-simd.md
	(aarch64_simd_bsl<mode>_internal): Handle more modes.
	(aarch64_simd_bsl<mode>): Likewise.
	* config/aarch64/arm_neon.h
	(vbsl<q>_<fpsu><8,16,32,64): Implement using builtins.
	* config/aarch64/iterators.md (VALLDIF): New.
	(Vbtype): Handle more modes.

From-SVN: r205385

6 files changed, 201 insertions, 263 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 7704433..52c507d 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,6 +1,26 @@
 2013-11-26  James Greenhalgh  <james.greenhalgh@arm.com>
 
 	* config/aarch64/aarch64-builtins.c
+	(aarch64_types_bsl_p_qualifiers): New.
+	(aarch64_types_bsl_s_qualifiers): Likewise.
+	(aarch64_types_bsl_u_qualifiers): Likewise.
+	(TYPES_BSL_P): Likewise.
+	(TYPES_BSL_S): Likewise.
+	(TYPES_BSL_U): Likewise.
+	(BUILTIN_VALLDIF): Likewise.
+	(BUILTIN_VDQQH): Likewise.
+	* config/aarch64/aarch64-simd-builtins.def (simd_bsl): New.
+	* config/aarch64/aarch64-simd.md
+	(aarch64_simd_bsl<mode>_internal): Handle more modes.
+	(aarch64_simd_bsl<mode>): Likewise.
+	* config/aarch64/arm_neon.h
+	(vbsl<q>_<fpsu><8,16,32,64): Implement using builtins.
+	* config/aarch64/iterators.md (VALLDIF): New.
+	(Vbtype): Handle more modes.
+
+2013-11-26  James Greenhalgh  <james.greenhalgh@arm.com>
+
+	* config/aarch64/aarch64-builtins.c
 	(aarch64_type_qualifiers): Add qualifier_poly.
 	(aarch64_build_scalar_type): Also build Poly types.
 	(aarch64_build_vector_type): Likewise.
diff --git a/gcc/config/aarch64/aarch64-builtins.c b/gcc/config/aarch64/aarch64-builtins.c
index cb80f9a..1bc3cc5 100644
--- a/gcc/config/aarch64/aarch64-builtins.c
+++ b/gcc/config/aarch64/aarch64-builtins.c
@@ -181,6 +181,22 @@ aarch64_types_load1_qualifiers[SIMD_MAX_BUILTIN_ARGS]
 #define TYPES_LOAD1 (aarch64_types_load1_qualifiers)
 #define TYPES_LOADSTRUCT (aarch64_types_load1_qualifiers)
 
+static enum aarch64_type_qualifiers
+aarch64_types_bsl_p_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_poly, qualifier_unsigned,
+      qualifier_poly, qualifier_poly };
+#define TYPES_BSL_P (aarch64_types_bsl_p_qualifiers)
+static enum aarch64_type_qualifiers
+aarch64_types_bsl_s_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_unsigned,
+      qualifier_none, qualifier_none };
+#define TYPES_BSL_S (aarch64_types_bsl_s_qualifiers)
+static enum aarch64_type_qualifiers
+aarch64_types_bsl_u_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_unsigned, qualifier_unsigned,
+      qualifier_unsigned, qualifier_unsigned };
+#define TYPES_BSL_U (aarch64_types_bsl_u_qualifiers)
+
 /* The first argument (return type) of a store should be void type,
    which we represent with qualifier_void.  Their first operand will be
    a DImode pointer to the location to store to, so we must use
@@ -255,6 +271,9 @@ aarch64_types_store1_qualifiers[SIMD_MAX_BUILTIN_ARGS]
 #define BUILTIN_VALLDI(T, N, MAP) \
   VAR11 (T, N, MAP, v8qi, v16qi, v4hi, v8hi, v2si, \
 	 v4si, v2di, v2sf, v4sf, v2df, di)
+#define BUILTIN_VALLDIF(T, N, MAP) \
+  VAR12 (T, N, MAP, v8qi, v16qi, v4hi, v8hi, v2si, \
+	 v4si, v2di, v2sf, v4sf, v2df, di, df)
 #define BUILTIN_VB(T, N, MAP) \
   VAR2 (T, N, MAP, v8qi, v16qi)
 #define BUILTIN_VD(T, N, MAP) \
@@ -279,6 +298,8 @@ aarch64_types_store1_qualifiers[SIMD_MAX_BUILTIN_ARGS]
   VAR6 (T, N, MAP, v8qi, v16qi, v4hi, v8hi, v2si, v4si)
 #define BUILTIN_VDQV(T, N, MAP) \
   VAR5 (T, N, MAP, v8qi, v16qi, v4hi, v8hi, v4si)
+#define BUILTIN_VDQQH(T, N, MAP) \
+  VAR4 (T, N, MAP, v8qi, v16qi, v4hi, v8hi)
 #define BUILTIN_VDQ_BHSI(T, N, MAP) \
   VAR6 (T, N, MAP, v8qi, v16qi, v4hi, v8hi, v2si, v4si)
 #define BUILTIN_VDQ_I(T, N, MAP) \
diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def
index c18b150..1dc3c1f 100644
--- a/gcc/config/aarch64/aarch64-simd-builtins.def
+++ b/gcc/config/aarch64/aarch64-simd-builtins.def
@@ -362,3 +362,8 @@
   /* Implemented by fma<mode>4.  */
   BUILTIN_VDQF (TERNOP, fma, 4)
 
+  /* Implemented by aarch64_simd_bsl<mode>.  */
+  BUILTIN_VDQQH (BSL_P, simd_bsl, 0)
+  BUILTIN_VSDQ_I_DI (BSL_U, simd_bsl, 0)
+  BUILTIN_VALLDIF (BSL_S, simd_bsl, 0)
+
diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md
index 5dcbc62..158b3dc 100644
--- a/gcc/config/aarch64/aarch64-simd.md
+++ b/gcc/config/aarch64/aarch64-simd.md
@@ -1662,15 +1662,15 @@
 ;;     bif op0, op1, mask
 
 (define_insn "aarch64_simd_bsl<mode>_internal"
-  [(set (match_operand:VALL 0 "register_operand"		"=w,w,w")
-	(ior:VALL
-	   (and:VALL
+  [(set (match_operand:VALLDIF 0 "register_operand"		"=w,w,w")
+	(ior:VALLDIF
+	   (and:VALLDIF
 	     (match_operand:<V_cmp_result> 1 "register_operand"	" 0,w,w")
-	     (match_operand:VALL 2 "register_operand"		" w,w,0"))
-	   (and:VALL
+	     (match_operand:VALLDIF 2 "register_operand"	" w,w,0"))
+	   (and:VALLDIF
 	     (not:<V_cmp_result>
 		(match_dup:<V_cmp_result> 1))
-	     (match_operand:VALL 3 "register_operand"		" w,0,w"))
+	     (match_operand:VALLDIF 3 "register_operand"	" w,0,w"))
 	))]
   "TARGET_SIMD"
   "@
@@ -1681,10 +1681,10 @@
 )
 
 (define_expand "aarch64_simd_bsl<mode>"
-  [(match_operand:VALL 0 "register_operand")
+  [(match_operand:VALLDIF 0 "register_operand")
    (match_operand:<V_cmp_result> 1 "register_operand")
-   (match_operand:VALL 2 "register_operand")
-   (match_operand:VALL 3 "register_operand")]
+   (match_operand:VALLDIF 2 "register_operand")
+   (match_operand:VALLDIF 3 "register_operand")]
  "TARGET_SIMD"
 {
   /* We can't alias operands together if they have different modes.  */
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index dc56170..6826ffb 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -4839,259 +4839,6 @@ vaddlvq_u32 (uint32x4_t a)
   return result;
 }
 
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vbsl_f32 (uint32x2_t a, float32x2_t b, float32x2_t c)
-{
-  float32x2_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
-vbsl_p8 (uint8x8_t a, poly8x8_t b, poly8x8_t c)
-{
-  poly8x8_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline poly16x4_t __attribute__ ((__always_inline__))
-vbsl_p16 (uint16x4_t a, poly16x4_t b, poly16x4_t c)
-{
-  poly16x4_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
-vbsl_s8 (uint8x8_t a, int8x8_t b, int8x8_t c)
-{
-  int8x8_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
-vbsl_s16 (uint16x4_t a, int16x4_t b, int16x4_t c)
-{
-  int16x4_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
-vbsl_s32 (uint32x2_t a, int32x2_t b, int32x2_t c)
-{
-  int32x2_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int64x1_t __attribute__ ((__always_inline__))
-vbsl_s64 (uint64x1_t a, int64x1_t b, int64x1_t c)
-{
-  int64x1_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__))
-vbsl_u8 (uint8x8_t a, uint8x8_t b, uint8x8_t c)
-{
-  uint8x8_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__))
-vbsl_u16 (uint16x4_t a, uint16x4_t b, uint16x4_t c)
-{
-  uint16x4_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__))
-vbsl_u32 (uint32x2_t a, uint32x2_t b, uint32x2_t c)
-{
-  uint32x2_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint64x1_t __attribute__ ((__always_inline__))
-vbsl_u64 (uint64x1_t a, uint64x1_t b, uint64x1_t c)
-{
-  uint64x1_t result;
-  __asm__ ("bsl %0.8b, %2.8b, %3.8b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
-vbslq_f32 (uint32x4_t a, float32x4_t b, float32x4_t c)
-{
-  float32x4_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
-vbslq_f64 (uint64x2_t a, float64x2_t b, float64x2_t c)
-{
-  float64x2_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline poly8x16_t __attribute__ ((__always_inline__))
-vbslq_p8 (uint8x16_t a, poly8x16_t b, poly8x16_t c)
-{
-  poly8x16_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline poly16x8_t __attribute__ ((__always_inline__))
-vbslq_p16 (uint16x8_t a, poly16x8_t b, poly16x8_t c)
-{
-  poly16x8_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int8x16_t __attribute__ ((__always_inline__))
-vbslq_s8 (uint8x16_t a, int8x16_t b, int8x16_t c)
-{
-  int8x16_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
-vbslq_s16 (uint16x8_t a, int16x8_t b, int16x8_t c)
-{
-  int16x8_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
-vbslq_s32 (uint32x4_t a, int32x4_t b, int32x4_t c)
-{
-  int32x4_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline int64x2_t __attribute__ ((__always_inline__))
-vbslq_s64 (uint64x2_t a, int64x2_t b, int64x2_t c)
-{
-  int64x2_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint8x16_t __attribute__ ((__always_inline__))
-vbslq_u8 (uint8x16_t a, uint8x16_t b, uint8x16_t c)
-{
-  uint8x16_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint16x8_t __attribute__ ((__always_inline__))
-vbslq_u16 (uint16x8_t a, uint16x8_t b, uint16x8_t c)
-{
-  uint16x8_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint32x4_t __attribute__ ((__always_inline__))
-vbslq_u32 (uint32x4_t a, uint32x4_t b, uint32x4_t c)
-{
-  uint32x4_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline uint64x2_t __attribute__ ((__always_inline__))
-vbslq_u64 (uint64x2_t a, uint64x2_t b, uint64x2_t c)
-{
-  uint64x2_t result;
-  __asm__ ("bsl %0.16b, %2.16b, %3.16b"
-           : "=w"(result)
-           : "0"(a), "w"(b), "w"(c)
-           : /* No clobbers */);
-  return result;
-}
-
 __extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
 vcls_s8 (int8x8_t a)
 {
@@ -15793,6 +15540,146 @@ vaddvq_f64 (float64x2_t __a)
   return vgetq_lane_f64 (__t, __LANE0 (2));
 }
 
+/* vbsl  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vbsl_f32 (uint32x2_t __a, float32x2_t __b, float32x2_t __c)
+{
+  return __builtin_aarch64_simd_bslv2sf_suss (__a, __b, __c);
+}
+
+__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
+vbsl_p8 (uint8x8_t __a, poly8x8_t __b, poly8x8_t __c)
+{
+  return __builtin_aarch64_simd_bslv8qi_pupp (__a, __b, __c);
+}
+
+__extension__ static __inline poly16x4_t __attribute__ ((__always_inline__))
+vbsl_p16 (uint16x4_t __a, poly16x4_t __b, poly16x4_t __c)
+{
+  return __builtin_aarch64_simd_bslv4hi_pupp (__a, __b, __c);
+}
+
+__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
+vbsl_s8 (uint8x8_t __a, int8x8_t __b, int8x8_t __c)
+{
+  return __builtin_aarch64_simd_bslv8qi_suss (__a, __b, __c);
+}
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vbsl_s16 (uint16x4_t __a, int16x4_t __b, int16x4_t __c)
+{
+  return __builtin_aarch64_simd_bslv4hi_suss (__a, __b, __c);
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vbsl_s32 (uint32x2_t __a, int32x2_t __b, int32x2_t __c)
+{
+  return __builtin_aarch64_simd_bslv2si_suss (__a, __b, __c);
+}
+
+__extension__ static __inline int64x1_t __attribute__ ((__always_inline__))
+vbsl_s64 (uint64x1_t __a, int64x1_t __b, int64x1_t __c)
+{
+  return __builtin_aarch64_simd_bsldi_suss (__a, __b, __c);
+}
+
+__extension__ static __inline uint8x8_t __attribute__ ((__always_inline__))
+vbsl_u8 (uint8x8_t __a, uint8x8_t __b, uint8x8_t __c)
+{
+  return __builtin_aarch64_simd_bslv8qi_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline uint16x4_t __attribute__ ((__always_inline__))
+vbsl_u16 (uint16x4_t __a, uint16x4_t __b, uint16x4_t __c)
+{
+  return __builtin_aarch64_simd_bslv4hi_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline uint32x2_t __attribute__ ((__always_inline__))
+vbsl_u32 (uint32x2_t __a, uint32x2_t __b, uint32x2_t __c)
+{
+  return __builtin_aarch64_simd_bslv2si_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline uint64x1_t __attribute__ ((__always_inline__))
+vbsl_u64 (uint64x1_t __a, uint64x1_t __b, uint64x1_t __c)
+{
+  return __builtin_aarch64_simd_bsldi_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vbslq_f32 (uint32x4_t __a, float32x4_t __b, float32x4_t __c)
+{
+  return __builtin_aarch64_simd_bslv4sf_suss (__a, __b, __c);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vbslq_f64 (uint64x2_t __a, float64x2_t __b, float64x2_t __c)
+{
+  return __builtin_aarch64_simd_bslv2df_suss (__a, __b, __c);
+}
+
+__extension__ static __inline poly8x16_t __attribute__ ((__always_inline__))
+vbslq_p8 (uint8x16_t __a, poly8x16_t __b, poly8x16_t __c)
+{
+  return __builtin_aarch64_simd_bslv16qi_pupp (__a, __b, __c);
+}
+
+__extension__ static __inline poly16x8_t __attribute__ ((__always_inline__))
+vbslq_p16 (uint16x8_t __a, poly16x8_t __b, poly16x8_t __c)
+{
+  return __builtin_aarch64_simd_bslv8hi_pupp (__a, __b, __c);
+}
+
+__extension__ static __inline int8x16_t __attribute__ ((__always_inline__))
+vbslq_s8 (uint8x16_t __a, int8x16_t __b, int8x16_t __c)
+{
+  return __builtin_aarch64_simd_bslv16qi_suss (__a, __b, __c);
+}
+
+__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
+vbslq_s16 (uint16x8_t __a, int16x8_t __b, int16x8_t __c)
+{
+  return __builtin_aarch64_simd_bslv8hi_suss (__a, __b, __c);
+}
+
+__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
+vbslq_s32 (uint32x4_t __a, int32x4_t __b, int32x4_t __c)
+{
+  return __builtin_aarch64_simd_bslv4si_suss (__a, __b, __c);
+}
+
+__extension__ static __inline int64x2_t __attribute__ ((__always_inline__))
+vbslq_s64 (uint64x2_t __a, int64x2_t __b, int64x2_t __c)
+{
+  return __builtin_aarch64_simd_bslv2di_suss (__a, __b, __c);
+}
+
+__extension__ static __inline uint8x16_t __attribute__ ((__always_inline__))
+vbslq_u8 (uint8x16_t __a, uint8x16_t __b, uint8x16_t __c)
+{
+  return __builtin_aarch64_simd_bslv16qi_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline uint16x8_t __attribute__ ((__always_inline__))
+vbslq_u16 (uint16x8_t __a, uint16x8_t __b, uint16x8_t __c)
+{
+  return __builtin_aarch64_simd_bslv8hi_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline uint32x4_t __attribute__ ((__always_inline__))
+vbslq_u32 (uint32x4_t __a, uint32x4_t __b, uint32x4_t __c)
+{
+  return __builtin_aarch64_simd_bslv4si_uuuu (__a, __b, __c);
+}
+
+__extension__ static __inline uint64x2_t __attribute__ ((__always_inline__))
+vbslq_u64 (uint64x2_t __a, uint64x2_t __b, uint64x2_t __c)
+{
+  return __builtin_aarch64_simd_bslv2di_uuuu (__a, __b, __c);
+}
+
 /* vcage  */
 
 __extension__ static __inline uint32_t __attribute__ ((__always_inline__))
diff --git a/gcc/config/aarch64/iterators.md b/gcc/config/aarch64/iterators.md
index fd7152c..43279ad 100644
--- a/gcc/config/aarch64/iterators.md
+++ b/gcc/config/aarch64/iterators.md
@@ -107,6 +107,10 @@
 ;; All vector modes and DI.
 (define_mode_iterator VALLDI [V8QI V16QI V4HI V8HI V2SI V4SI V2DI V2SF V4SF V2DF DI])
 
+;; All vector modes and DI and DF.
+(define_mode_iterator VALLDIF [V8QI V16QI V4HI V8HI V2SI V4SI
+			       V2DI V2SF V4SF V2DF DI DF])
+
 ;; Vector modes for Integer reduction across lanes.
 (define_mode_iterator VDQV [V8QI V16QI V4HI V8HI V4SI V2DI])
 
@@ -363,7 +367,8 @@
 			  (V4HI "8b") (V8HI  "16b")
 			  (V2SI "8b") (V4SI  "16b")
 			  (V2DI "16b") (V2SF  "8b")
-			  (V4SF "16b") (V2DF  "16b")])
+			  (V4SF "16b") (V2DF  "16b")
+			  (DI   "8b")  (DF    "8b")])
 
 ;; Define element mode for each vector mode.
 (define_mode_attr VEL [(V8QI "QI") (V16QI "QI")