[ARM,1/2x] : MVE intrinsics with binary operands.

Message ID DBBPR08MB4775CA058897CA97C55A477B9B710@DBBPR08MB4775.eurprd08.prod.outlook.com
State New
Headers show
Series
  • [ARM,1/2x] : MVE intrinsics with binary operands.
Related show

Commit Message

Srinath Parvathaneni Nov. 14, 2019, 7:13 p.m.
Hello,

This patch supports following MVE ACLE intrinsics with binary operand.

vsubq_n_f16, vsubq_n_f32, vbrsrq_n_f16, vbrsrq_n_f32, vcvtq_n_f16_s16,
vcvtq_n_f32_s32, vcvtq_n_f16_u16, vcvtq_n_f32_u32, vcreateq_f16, vcreateq_f32.

Please refer to M-profile Vector Extension (MVE) intrinsics [1]  for more details.
[1] https://developer.arm.com/architectures/instruction-sets/simd-isas/helium/mve-intrinsics

In this patch new constraint "Rd" is added, which checks the constant is with in the range of 1 to 16.
Also a new predicate "mve_imm_16" is added, to check the the matching constraint Rd.

Regression tested on arm-none-eabi and found no regressions.

Ok for trunk?

Thanks,
Srinath.

gcc/ChangeLog:

2019-10-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
	    Mihail Ionescu  <mihail.ionescu@arm.com>
	    Srinath Parvathaneni  <srinath.parvathaneni@arm.com>

	* config/arm/arm-builtins.c (BINOP_NONE_NONE_NONE_QUALIFIERS): Define
	qualifier for binary operands.
	(BINOP_NONE_NONE_IMM_QUALIFIERS): Likewise.
	(BINOP_NONE_UNONE_IMM_QUALIFIERS): Likewise.
	(BINOP_NONE_UNONE_UNONE_QUALIFIERS): Likewise.
	* config/arm/arm_mve.h (vsubq_n_f16): Define macro.
	(vsubq_n_f32): Likewise.
	(vbrsrq_n_f16): Likewise.
	(vbrsrq_n_f32): Likewise.
	(vcvtq_n_f16_s16): Likewise.
	(vcvtq_n_f32_s32): Likewise.
	(vcvtq_n_f16_u16): Likewise.
	(vcvtq_n_f32_u32): Likewise.
	(vcreateq_f16): Likewise.
	(vcreateq_f32): Likewise.
	(__arm_vsubq_n_f16): Define intrinsic.
	(__arm_vsubq_n_f32): Likewise.
	(__arm_vbrsrq_n_f16): Likewise.
	(__arm_vbrsrq_n_f32): Likewise.
	(__arm_vcvtq_n_f16_s16): Likewise.
	(__arm_vcvtq_n_f32_s32): Likewise.
	(__arm_vcvtq_n_f16_u16): Likewise.
	(__arm_vcvtq_n_f32_u32): Likewise.
	(__arm_vcreateq_f16): Likewise.
	(__arm_vcreateq_f32): Likewise.
	(vsubq): Define polymorphic variant.
	(vbrsrq): Likewise.
	(vcvtq_n): Likewise.
	* config/arm/arm_mve_builtins.def (BINOP_NONE_NONE_NONE_QUALIFIERS): Use
	it.
        (BINOP_NONE_NONE_IMM_QUALIFIERS): Likewise.
        (BINOP_NONE_UNONE_IMM_QUALIFIERS): Likewise.
        (BINOP_NONE_UNONE_UNONE_QUALIFIERS): Likewise.
	* config/arm/constraints.md (Rd): Define constraint to check constant is
	in the range of 1 to 16.
	* config/arm/mve.md (mve_vsubq_n_f<mode>): Define RTL pattern.
	mve_vbrsrq_n_f<mode>: Likewise.
	mve_vcvtq_n_to_f_<supf><mode>: Likewise.
	mve_vcreateq_f<mode>: Likewise.
	* config/arm/predicates.md (mve_imm_16): Define predicate to check
	the matching constraint Rd.

gcc/testsuite/ChangeLog:

2019-10-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
	    Mihail Ionescu  <mihail.ionescu@arm.com>
	    Srinath Parvathaneni  <srinath.parvathaneni@arm.com>

	* gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c: New test.
	* gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vcreateq_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vcreateq_f32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vsubq_n_f16.c: Likewise.
	* gcc.target/arm/mve/intrinsics/vsubq_n_f32.c: Likewise.


###############     Attachment also inlined for ease of reply    ###############
diff --git a/gcc/config/arm/arm-builtins.c b/gcc/config/arm/arm-builtins.c
index cd82aa159089c288607e240de02a85dcbb134a14..c2dad057d1365914477c64d559aa1fd1c32bbf19 100644
--- a/gcc/config/arm/arm-builtins.c
+++ b/gcc/config/arm/arm-builtins.c
@@ -349,6 +349,30 @@ arm_unop_unone_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]
 #define UNOP_UNONE_IMM_QUALIFIERS \
   (arm_unop_unone_imm_qualifiers)
 
+static enum arm_type_qualifiers
+arm_binop_none_none_none_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_none, qualifier_none };
+#define BINOP_NONE_NONE_NONE_QUALIFIERS \
+  (arm_binop_none_none_none_qualifiers)
+
+static enum arm_type_qualifiers
+arm_binop_none_none_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_none, qualifier_immediate };
+#define BINOP_NONE_NONE_IMM_QUALIFIERS \
+  (arm_binop_none_none_imm_qualifiers)
+
+static enum arm_type_qualifiers
+arm_binop_none_unone_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_unsigned, qualifier_immediate };
+#define BINOP_NONE_UNONE_IMM_QUALIFIERS \
+  (arm_binop_none_unone_imm_qualifiers)
+
+static enum arm_type_qualifiers
+arm_binop_none_unone_unone_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_unsigned, qualifier_unsigned };
+#define BINOP_NONE_UNONE_UNONE_QUALIFIERS \
+  (arm_binop_none_unone_unone_qualifiers)
+
 /* End of Qualifier for MVE builtins.  */
 
    /* void ([T element type] *, T, immediate).  */
diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h
index c8d9b6471634725cea9bab3f9fa145810b506938..15b7ada025fe57b682f3873f0f275c43a30d1273 100644
--- a/gcc/config/arm/arm_mve.h
+++ b/gcc/config/arm/arm_mve.h
@@ -197,6 +197,16 @@ typedef struct { uint8x16_t val[4]; } uint8x16x4_t;
 #define vctp64q(__a) __arm_vctp64q(__a)
 #define vctp8q(__a) __arm_vctp8q(__a)
 #define vpnot(__a) __arm_vpnot(__a)
+#define vsubq_n_f16(__a, __b) __arm_vsubq_n_f16(__a, __b)
+#define vsubq_n_f32(__a, __b) __arm_vsubq_n_f32(__a, __b)
+#define vbrsrq_n_f16(__a, __b) __arm_vbrsrq_n_f16(__a, __b)
+#define vbrsrq_n_f32(__a, __b) __arm_vbrsrq_n_f32(__a, __b)
+#define vcvtq_n_f16_s16(__a,  __imm6) __arm_vcvtq_n_f16_s16(__a,  __imm6)
+#define vcvtq_n_f32_s32(__a,  __imm6) __arm_vcvtq_n_f32_s32(__a,  __imm6)
+#define vcvtq_n_f16_u16(__a,  __imm6) __arm_vcvtq_n_f16_u16(__a,  __imm6)
+#define vcvtq_n_f32_u32(__a,  __imm6) __arm_vcvtq_n_f32_u32(__a,  __imm6)
+#define vcreateq_f16(__a, __b) __arm_vcreateq_f16(__a, __b)
+#define vcreateq_f32(__a, __b) __arm_vcreateq_f32(__a, __b)
 #endif
 
 __extension__ extern __inline void
@@ -1085,6 +1095,76 @@ __arm_vcvtmq_s32_f32 (float32x4_t __a)
   return __builtin_mve_vcvtmq_sv4si (__a);
 }
 
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vsubq_n_f16 (float16x8_t __a, float16_t __b)
+{
+  return __builtin_mve_vsubq_n_fv8hf (__a, __b);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vsubq_n_f32 (float32x4_t __a, float32_t __b)
+{
+  return __builtin_mve_vsubq_n_fv4sf (__a, __b);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vbrsrq_n_f16 (float16x8_t __a, int32_t __b)
+{
+  return __builtin_mve_vbrsrq_n_fv8hf (__a, __b);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vbrsrq_n_f32 (float32x4_t __a, int32_t __b)
+{
+  return __builtin_mve_vbrsrq_n_fv4sf (__a, __b);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f16_s16 (int16x8_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_sv8hf (__a, __imm6);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f32_s32 (int32x4_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_sv4sf (__a, __imm6);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f16_u16 (uint16x8_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_uv8hf (__a, __imm6);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f32_u32 (uint32x4_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_uv4sf (__a, __imm6);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcreateq_f16 (uint64_t __a, uint64_t __b)
+{
+  return __builtin_mve_vcreateq_fv8hf (__a, __b);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcreateq_f32 (uint64_t __a, uint64_t __b)
+{
+  return __builtin_mve_vcreateq_fv4sf (__a, __b);
+}
+
 #endif
 
 enum {
@@ -1373,6 +1453,27 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint16x8_t]: __arm_vcvtq_f16_u16 (__ARM_mve_coerce(__p0, uint16x8_t)), \
   int (*)[__ARM_mve_type_uint32x4_t]: __arm_vcvtq_f32_u32 (__ARM_mve_coerce(__p0, uint32x4_t)));})
 
+#define vsubq(p0,p1) __arm_vsubq(p0,p1)
+#define __arm_vsubq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
+  __typeof(p1) __p1 = (p1); \
+  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
+  int (*)[__ARM_mve_type_float16x8_t][__ARM_mve_type_float16_t]: __arm_vsubq_n_f16 (__ARM_mve_coerce(__p0, float16x8_t), __ARM_mve_coerce(__p1, float16_t)), \
+  int (*)[__ARM_mve_type_float32x4_t][__ARM_mve_type_float32_t]: __arm_vsubq_n_f32 (__ARM_mve_coerce(__p0, float32x4_t), __ARM_mve_coerce(__p1, float32_t)));})
+
+#define vbrsrq(p0,p1) __arm_vbrsrq(p0,p1)
+#define __arm_vbrsrq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
+  _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \
+  int (*)[__ARM_mve_type_float16x8_t]: __arm_vbrsrq_n_f16 (__ARM_mve_coerce(__p0, float16x8_t), p1), \
+  int (*)[__ARM_mve_type_float32x4_t]: __arm_vbrsrq_n_f32 (__ARM_mve_coerce(__p0, float32x4_t), p1));})
+
+#define vcvtq_n(p0,p1) __arm_vcvtq_n(p0,p1)
+#define __arm_vcvtq_n(p0,p1) ({ __typeof(p0) __p0 = (p0); \
+  _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \
+  int (*)[__ARM_mve_type_int16x8_t]: __arm_vcvtq_n_f16_s16 (__ARM_mve_coerce(__p0, int16x8_t), p1), \
+  int (*)[__ARM_mve_type_int32x4_t]: __arm_vcvtq_n_f32_s32 (__ARM_mve_coerce(__p0, int32x4_t), p1), \
+  int (*)[__ARM_mve_type_uint16x8_t]: __arm_vcvtq_n_f16_u16 (__ARM_mve_coerce(__p0, uint16x8_t), p1), \
+  int (*)[__ARM_mve_type_uint32x4_t]: __arm_vcvtq_n_f32_u32 (__ARM_mve_coerce(__p0, uint32x4_t), p1));})
+
 #else /* MVE Interger.  */
 
 #define vst4q(p0,p1) __arm_vst4q(p0,p1)
diff --git a/gcc/config/arm/arm_mve_builtins.def b/gcc/config/arm/arm_mve_builtins.def
index 816b6dfca7fb221275212ca5f06fc6f679860a38..8d1e4fac3d75e87fbe334e64e1073cb1fef0d96d 100644
--- a/gcc/config/arm/arm_mve_builtins.def
+++ b/gcc/config/arm/arm_mve_builtins.def
@@ -6,7 +6,7 @@
 
     GCC is free software; you can redistribute it and/or modify it
     under the terms of the GNU General Public License as published
-    by the Free Software Foundation; either version 3, or (at your
+    by the Free Software Foundation; either version 3, or  (at your
     option) any later version.
 
     GCC is distributed in the hope that it will be useful, but WITHOUT
@@ -76,3 +76,8 @@ VAR1 (UNOP_UNONE_UNONE, vctp32q, hi)
 VAR1 (UNOP_UNONE_UNONE, vctp64q, hi)
 VAR1 (UNOP_UNONE_UNONE, vctp8q, hi)
 VAR1 (UNOP_UNONE_UNONE, vpnot, hi)
+VAR2 (BINOP_NONE_NONE_NONE, vsubq_n_f, v8hf, v4sf)
+VAR2 (BINOP_NONE_NONE_NONE, vbrsrq_n_f, v8hf, v4sf)
+VAR2 (BINOP_NONE_NONE_IMM, vcvtq_n_to_f_s, v8hf, v4sf)
+VAR2 (BINOP_NONE_UNONE_IMM, vcvtq_n_to_f_u, v8hf, v4sf)
+VAR2 (BINOP_NONE_UNONE_UNONE, vcreateq_f, v8hf, v4sf)
diff --git a/gcc/config/arm/constraints.md b/gcc/config/arm/constraints.md
index 4c193d20a48d5e9ed43bdca76468381fc17682da..86cd82334533acbe58be4a0c547cf38737d2bb84 100644
--- a/gcc/config/arm/constraints.md
+++ b/gcc/config/arm/constraints.md
@@ -35,7 +35,7 @@
 ;; in ARM/Thumb-2 state: Da, Db, Dc, Dd, Dn, DN, Dm, Dl, DL, Do, Dv, Dy, Di,
 ;;			 Dt, Dp, Dz, Tu
 ;; in Thumb-1 state: Pa, Pb, Pc, Pd, Pe
-;; in Thumb-2 state: Ha, Pj, PJ, Ps, Pt, Pu, Pv, Pw, Px, Py, Pz
+;; in Thumb-2 state: Ha, Pj, PJ, Ps, Pt, Pu, Pv, Pw, Px, Py, Pz, Rd
 ;; in all states: Pf, Pg, UM, U1
 
 ;; The following memory constraints have been used:
@@ -51,6 +51,11 @@
   "MVE EVEN registers @code{r0}, @code{r2}, @code{r4}, @code{r6}, @code{r8},
    @code{r10}, @code{r12}, @code{r14}")
 
+(define_constraint "Rd"
+  "@internal In Thumb-2 state a constant in range 1 to 16"
+  (and (match_code "const_int")
+       (match_test "TARGET_HAVE_MVE && ival >= 1 && ival <= 16")))
+
 (define_register_constraint "t" "TARGET_32BIT ? VFP_LO_REGS : NO_REGS"
  "The VFP registers @code{s0}-@code{s31}.")
 
diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md
index ee2263e04309e8274ec76ae4478afb7cfba59f8f..06701b024dabae85446bb60060a8b331e540cd6d 100644
--- a/gcc/config/arm/mve.md
+++ b/gcc/config/arm/mve.md
@@ -36,7 +36,9 @@
 			 VREV32Q_U VREV32Q_S VMOVLTQ_U VMOVLTQ_S VMOVLBQ_S
 			 VMOVLBQ_U VCVTQ_FROM_F_S VCVTQ_FROM_F_U VCVTPQ_S
 			 VCVTPQ_U VCVTNQ_S VCVTNQ_U VCVTMQ_S VCVTMQ_U
-			 VADDLVQ_U VCTP8Q VCTP16Q VCTP32Q VCTP64Q VPNOT])
+			 VADDLVQ_U VCTP8Q VCTP16Q VCTP32Q VCTP64Q VPNOT
+			 VCREATEQ_F VCVTQ_N_TO_F_S VCVTQ_N_TO_F_U VBRSRQ_N_F
+			 VSUBQ_N_F])
 
 (define_mode_attr MVE_CNVT [(V8HI "V8HF") (V4SI "V4SF")
 			    (V8HF "V8HI") (V4SF "V4SI")])
@@ -52,7 +54,8 @@
 		       (VCVTPQ_S "s") (VCVTPQ_U "u") (VCVTNQ_S "s")
 		       (VCVTNQ_U "u") (VCVTMQ_S "s") (VCVTMQ_U "u")
 		       (VCLZQ_U "u") (VCLZQ_S "s") (VREV32Q_U "u")
-		       (VREV32Q_S "s") (VADDLVQ_U "u") (VADDLVQ_S "s")])
+		       (VREV32Q_S "s") (VADDLVQ_U "u") (VADDLVQ_S "s")
+		       (VCVTQ_N_TO_F_S "s") (VCVTQ_N_TO_F_U "u")])
 
 (define_int_attr mode1 [(VCTP8Q "8") (VCTP16Q "16") (VCTP32Q "32")
 			(VCTP64Q "64")])
@@ -75,6 +78,7 @@
 (define_int_iterator VCVTMQ [VCVTMQ_S VCVTMQ_U])
 (define_int_iterator VADDLVQ [VADDLVQ_U VADDLVQ_S])
 (define_int_iterator VCTPQ [VCTP8Q VCTP16Q VCTP32Q VCTP64Q])
+(define_int_iterator VCVTQ_N_TO_F [VCVTQ_N_TO_F_S VCVTQ_N_TO_F_U])
 
 (define_insn "*mve_mov<mode>"
   [(set (match_operand:MVE_types 0 "s_register_operand" "=w,w,r,w,w,r,w")
@@ -680,3 +684,62 @@
   "vpnot"
   [(set_attr "type" "mve_move")
 ])
+
+;;
+;; [vsubq_n_f])
+;;
+(define_insn "mve_vsubq_n_f<mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:MVE_0 1 "s_register_operand" "w")
+		       (match_operand:<V_elem> 2 "s_register_operand" "r")]
+	 VSUBQ_N_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vsub.f%#<V_sz_elem>  %q0, %q1, %2"
+  [(set_attr "type" "mve_move")
+])
+
+;;
+;; [vbrsrq_n_f])
+;;
+(define_insn "mve_vbrsrq_n_f<mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:MVE_0 1 "s_register_operand" "w")
+		       (match_operand:SI 2 "s_register_operand" "r")]
+	 VBRSRQ_N_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vbrsr.%#<V_sz_elem>  %q0, %q1, %2"
+  [(set_attr "type" "mve_move")
+])
+
+;;
+;; [vcvtq_n_to_f_s, vcvtq_n_to_f_u])
+;;
+(define_insn "mve_vcvtq_n_to_f_<supf><mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:<MVE_CNVT> 1 "s_register_operand" "w")
+		       (match_operand:SI 2 "mve_imm_16" "Rd")]
+	 VCVTQ_N_TO_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vcvt.f%#<V_sz_elem>.<supf>%#<V_sz_elem>\t%q0, %q1, %2"
+  [(set_attr "type" "mve_move")
+])
+
+;; [vcreateq_f])
+;;
+(define_insn "mve_vcreateq_f<mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:DI 1 "s_register_operand" "r")
+		       (match_operand:DI 2 "s_register_operand" "r")]
+	 VCREATEQ_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vmov %q0[2], %q0[0], %Q2, %Q1\;vmov %q0[3], %q0[1], %R2, %R1"
+  [(set_attr "type" "mve_move")
+   (set_attr "length""8")])
diff --git a/gcc/config/arm/predicates.md b/gcc/config/arm/predicates.md
index 9d74165fe065b03c77918fe9e4611967799535f1..80a44bb32cbce1ee3f850a44b70a0e4ceed548be 100644
--- a/gcc/config/arm/predicates.md
+++ b/gcc/config/arm/predicates.md
@@ -31,6 +31,10 @@
 	      || REGNO_REG_CLASS (REGNO (op)) != NO_REGS));
 })
 
+;; True for immediates in the range of 1 to 16 for MVE.
+(define_predicate "mve_imm_16"
+  (match_test "satisfies_constraint_Rd (op)"))
+
 ; Predicate for stack protector guard's address in
 ; stack_protect_combined_set_insn and stack_protect_combined_test_insn patterns
 (define_predicate "guard_addr_operand"
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c
new file mode 100644
index 0000000000000000000000000000000000000000..90f52b50e6bd0dbccc40e9da7e2f8badb593727d
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (float16x8_t a, int32_t b)
+{
+  return vbrsrq_n_f16 (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.16"  }  } */
+
+float16x8_t
+foo1 (float16x8_t a, int32_t b)
+{
+  return vbrsrq (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c
new file mode 100644
index 0000000000000000000000000000000000000000..4ccd8dd313319ce1933e85f76955d1ea12952abd
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (float32x4_t a, int32_t b)
+{
+  return vbrsrq_n_f32 (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.32"  }  } */
+
+float32x4_t
+foo1 (float32x4_t a, int32_t b)
+{
+  return vbrsrq (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.32"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c
new file mode 100644
index 0000000000000000000000000000000000000000..a83203ad0de1e6bfc6b4b9d469748108a335a391
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c
@@ -0,0 +1,13 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (uint64_t a, uint64_t b)
+{
+  return vcreateq_f16 (a, b);
+}
+
+/* { dg-final { scan-assembler "vmov"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c
new file mode 100644
index 0000000000000000000000000000000000000000..279e922ed45165d0b104b236e4a2c8f782c4a59d
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c
@@ -0,0 +1,13 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (uint64_t a, uint64_t b)
+{
+  return vcreateq_f32 (a, b);
+}
+
+/* { dg-final { scan-assembler "vmov"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c
new file mode 100644
index 0000000000000000000000000000000000000000..440d4df99b295deb66b91e3996cbca8f35f7d12b
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (int16x8_t a)
+{
+  return vcvtq_n_f16_s16 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.s16"  }  } */
+
+float16x8_t
+foo1 (int16x8_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.s16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c
new file mode 100644
index 0000000000000000000000000000000000000000..4d648d653ef91436bb87d2e07e15ccee0844309e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (uint16x8_t a)
+{
+  return vcvtq_n_f16_u16 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.u16"  }  } */
+
+float16x8_t
+foo1 (uint16x8_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.u16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c
new file mode 100644
index 0000000000000000000000000000000000000000..836727a9b2d6a2621b34de5b7d4643ee1bebf47c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (int32x4_t a)
+{
+  return vcvtq_n_f32_s32 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.s32"  }  } */
+
+float32x4_t
+foo1 (int32x4_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.s32"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c
new file mode 100644
index 0000000000000000000000000000000000000000..fcd66a42fd49c78e1c11ac89393dc38cb06efbf5
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (uint32x4_t a)
+{
+  return vcvtq_n_f32_u32 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.u32"  }  } */
+
+float32x4_t
+foo1 (uint32x4_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.u32"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c
new file mode 100644
index 0000000000000000000000000000000000000000..e6ff11edf3933de7acd7bdffd33d8b79c88a7f48
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (float16x8_t a, float16_t b)
+{
+  return vsubq_n_f16 (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f16"  }  } */
+
+float16x8_t
+foo1 (float16x8_t a, float16_t b)
+{
+  return vsubq (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c
new file mode 100644
index 0000000000000000000000000000000000000000..1092fd0aa6abc2a0ed0320a17d6b563f0396fb58
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c
@@ -0,0 +1,21 @@
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (float32x4_t a, float32_t b)
+{
+  return vsubq_n_f32 (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f32"  }  } */
+
+float32x4_t
+foo1 (float32x4_t a, float32_t b)
+{
+  return vsubq (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f32"  }  } */

Comments

Kyrill Tkachov Dec. 19, 2019, 6:16 p.m. | #1
Hi Srinath,

On 11/14/19 7:13 PM, Srinath Parvathaneni wrote:
> Hello,

>

> This patch supports following MVE ACLE intrinsics with binary operand.

>

> vsubq_n_f16, vsubq_n_f32, vbrsrq_n_f16, vbrsrq_n_f32, vcvtq_n_f16_s16,

> vcvtq_n_f32_s32, vcvtq_n_f16_u16, vcvtq_n_f32_u32, vcreateq_f16, 

> vcreateq_f32.

>

> Please refer to M-profile Vector Extension (MVE) intrinsics [1]  for 

> more details.

> [1] 

> https://developer.arm.com/architectures/instruction-sets/simd-isas/helium/mve-intrinsics

>

> In this patch new constraint "Rd" is added, which checks the constant 

> is with in the range of 1 to 16.

> Also a new predicate "mve_imm_16" is added, to check the the matching 

> constraint Rd.

>

> Regression tested on arm-none-eabi and found no regressions.

>

> Ok for trunk?

>

> Thanks,

> Srinath.

>

> gcc/ChangeLog:

>

> 2019-10-21  Andre Vieira <andre.simoesdiasvieira@arm.com>

>             Mihail Ionescu  <mihail.ionescu@arm.com>

>             Srinath Parvathaneni <srinath.parvathaneni@arm.com>

>

>         * config/arm/arm-builtins.c (BINOP_NONE_NONE_NONE_QUALIFIERS): 

> Define

>         qualifier for binary operands.

>         (BINOP_NONE_NONE_IMM_QUALIFIERS): Likewise.

>         (BINOP_NONE_UNONE_IMM_QUALIFIERS): Likewise.

>         (BINOP_NONE_UNONE_UNONE_QUALIFIERS): Likewise.

>         * config/arm/arm_mve.h (vsubq_n_f16): Define macro.

>         (vsubq_n_f32): Likewise.

>         (vbrsrq_n_f16): Likewise.

>         (vbrsrq_n_f32): Likewise.

>         (vcvtq_n_f16_s16): Likewise.

>         (vcvtq_n_f32_s32): Likewise.

>         (vcvtq_n_f16_u16): Likewise.

>         (vcvtq_n_f32_u32): Likewise.

>         (vcreateq_f16): Likewise.

>         (vcreateq_f32): Likewise.

>         (__arm_vsubq_n_f16): Define intrinsic.

>         (__arm_vsubq_n_f32): Likewise.

>         (__arm_vbrsrq_n_f16): Likewise.

>         (__arm_vbrsrq_n_f32): Likewise.

>         (__arm_vcvtq_n_f16_s16): Likewise.

>         (__arm_vcvtq_n_f32_s32): Likewise.

>         (__arm_vcvtq_n_f16_u16): Likewise.

>         (__arm_vcvtq_n_f32_u32): Likewise.

>         (__arm_vcreateq_f16): Likewise.

>         (__arm_vcreateq_f32): Likewise.

>         (vsubq): Define polymorphic variant.

>         (vbrsrq): Likewise.

>         (vcvtq_n): Likewise.

>         * config/arm/arm_mve_builtins.def 

> (BINOP_NONE_NONE_NONE_QUALIFIERS): Use

>         it.

>         (BINOP_NONE_NONE_IMM_QUALIFIERS): Likewise.

>         (BINOP_NONE_UNONE_IMM_QUALIFIERS): Likewise.

>         (BINOP_NONE_UNONE_UNONE_QUALIFIERS): Likewise.

>         * config/arm/constraints.md (Rd): Define constraint to check 

> constant is

>         in the range of 1 to 16.

>         * config/arm/mve.md (mve_vsubq_n_f<mode>): Define RTL pattern.

>         mve_vbrsrq_n_f<mode>: Likewise.

>         mve_vcvtq_n_to_f_<supf><mode>: Likewise.

>         mve_vcreateq_f<mode>: Likewise.

>         * config/arm/predicates.md (mve_imm_16): Define predicate to check

>         the matching constraint Rd.

>

> gcc/testsuite/ChangeLog:

>

> 2019-10-21  Andre Vieira <andre.simoesdiasvieira@arm.com>

>             Mihail Ionescu  <mihail.ionescu@arm.com>

>             Srinath Parvathaneni <srinath.parvathaneni@arm.com>

>

>         * gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c: New test.

>         * gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vcreateq_f16.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vcreateq_f32.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vsubq_n_f16.c: Likewise.

>         * gcc.target/arm/mve/intrinsics/vsubq_n_f32.c: Likewise.

>

>

> ###############     Attachment also inlined for ease of reply    

> ###############

>

>

> diff --git a/gcc/config/arm/arm-builtins.c b/gcc/config/arm/arm-builtins.c

> index 

> cd82aa159089c288607e240de02a85dcbb134a14..c2dad057d1365914477c64d559aa1fd1c32bbf19 

> 100644

> --- a/gcc/config/arm/arm-builtins.c

> +++ b/gcc/config/arm/arm-builtins.c

> @@ -349,6 +349,30 @@ arm_unop_unone_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]

>  #define UNOP_UNONE_IMM_QUALIFIERS \

>    (arm_unop_unone_imm_qualifiers)

>

> +static enum arm_type_qualifiers

> +arm_binop_none_none_none_qualifiers[SIMD_MAX_BUILTIN_ARGS]

> +  = { qualifier_none, qualifier_none, qualifier_none };

> +#define BINOP_NONE_NONE_NONE_QUALIFIERS \

> +  (arm_binop_none_none_none_qualifiers)

> +

> +static enum arm_type_qualifiers

> +arm_binop_none_none_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]

> +  = { qualifier_none, qualifier_none, qualifier_immediate };

> +#define BINOP_NONE_NONE_IMM_QUALIFIERS \

> +  (arm_binop_none_none_imm_qualifiers)

> +

> +static enum arm_type_qualifiers

> +arm_binop_none_unone_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]

> +  = { qualifier_none, qualifier_unsigned, qualifier_immediate };

> +#define BINOP_NONE_UNONE_IMM_QUALIFIERS \

> +  (arm_binop_none_unone_imm_qualifiers)

> +

> +static enum arm_type_qualifiers

> +arm_binop_none_unone_unone_qualifiers[SIMD_MAX_BUILTIN_ARGS]

> +  = { qualifier_none, qualifier_unsigned, qualifier_unsigned };

> +#define BINOP_NONE_UNONE_UNONE_QUALIFIERS \

> +  (arm_binop_none_unone_unone_qualifiers)

> +

>  /* End of Qualifier for MVE builtins.  */

>

>     /* void ([T element type] *, T, immediate).  */

> diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h

> index 

> c8d9b6471634725cea9bab3f9fa145810b506938..15b7ada025fe57b682f3873f0f275c43a30d1273 

> 100644

> --- a/gcc/config/arm/arm_mve.h

> +++ b/gcc/config/arm/arm_mve.h

> @@ -197,6 +197,16 @@ typedef struct { uint8x16_t val[4]; } uint8x16x4_t;

>  #define vctp64q(__a) __arm_vctp64q(__a)

>  #define vctp8q(__a) __arm_vctp8q(__a)

>  #define vpnot(__a) __arm_vpnot(__a)

> +#define vsubq_n_f16(__a, __b) __arm_vsubq_n_f16(__a, __b)

> +#define vsubq_n_f32(__a, __b) __arm_vsubq_n_f32(__a, __b)

> +#define vbrsrq_n_f16(__a, __b) __arm_vbrsrq_n_f16(__a, __b)

> +#define vbrsrq_n_f32(__a, __b) __arm_vbrsrq_n_f32(__a, __b)

> +#define vcvtq_n_f16_s16(__a,  __imm6) __arm_vcvtq_n_f16_s16(__a,  __imm6)

> +#define vcvtq_n_f32_s32(__a,  __imm6) __arm_vcvtq_n_f32_s32(__a,  __imm6)

> +#define vcvtq_n_f16_u16(__a,  __imm6) __arm_vcvtq_n_f16_u16(__a,  __imm6)

> +#define vcvtq_n_f32_u32(__a,  __imm6) __arm_vcvtq_n_f32_u32(__a,  __imm6)

> +#define vcreateq_f16(__a, __b) __arm_vcreateq_f16(__a, __b)

> +#define vcreateq_f32(__a, __b) __arm_vcreateq_f32(__a, __b)

>  #endif

>

>  __extension__ extern __inline void

> @@ -1085,6 +1095,76 @@ __arm_vcvtmq_s32_f32 (float32x4_t __a)

>    return __builtin_mve_vcvtmq_sv4si (__a);

>  }

>

> +__extension__ extern __inline float16x8_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vsubq_n_f16 (float16x8_t __a, float16_t __b)

> +{

> +  return __builtin_mve_vsubq_n_fv8hf (__a, __b);

> +}

> +

> +__extension__ extern __inline float32x4_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vsubq_n_f32 (float32x4_t __a, float32_t __b)

> +{

> +  return __builtin_mve_vsubq_n_fv4sf (__a, __b);

> +}

> +

> +__extension__ extern __inline float16x8_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vbrsrq_n_f16 (float16x8_t __a, int32_t __b)

> +{

> +  return __builtin_mve_vbrsrq_n_fv8hf (__a, __b);

> +}

> +

> +__extension__ extern __inline float32x4_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vbrsrq_n_f32 (float32x4_t __a, int32_t __b)

> +{

> +  return __builtin_mve_vbrsrq_n_fv4sf (__a, __b);

> +}

> +

> +__extension__ extern __inline float16x8_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vcvtq_n_f16_s16 (int16x8_t __a, const int __imm6)

> +{

> +  return __builtin_mve_vcvtq_n_to_f_sv8hf (__a, __imm6);

> +}

> +

> +__extension__ extern __inline float32x4_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vcvtq_n_f32_s32 (int32x4_t __a, const int __imm6)

> +{

> +  return __builtin_mve_vcvtq_n_to_f_sv4sf (__a, __imm6);

> +}

> +

> +__extension__ extern __inline float16x8_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vcvtq_n_f16_u16 (uint16x8_t __a, const int __imm6)

> +{

> +  return __builtin_mve_vcvtq_n_to_f_uv8hf (__a, __imm6);

> +}

> +

> +__extension__ extern __inline float32x4_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vcvtq_n_f32_u32 (uint32x4_t __a, const int __imm6)

> +{

> +  return __builtin_mve_vcvtq_n_to_f_uv4sf (__a, __imm6);

> +}

> +

> +__extension__ extern __inline float16x8_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vcreateq_f16 (uint64_t __a, uint64_t __b)

> +{

> +  return __builtin_mve_vcreateq_fv8hf (__a, __b);

> +}

> +

> +__extension__ extern __inline float32x4_t

> +__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))

> +__arm_vcreateq_f32 (uint64_t __a, uint64_t __b)

> +{

> +  return __builtin_mve_vcreateq_fv4sf (__a, __b);

> +}

> +

>  #endif

>

>  enum {

> @@ -1373,6 +1453,27 @@ extern void *__ARM_undef;

>    int (*)[__ARM_mve_type_uint16x8_t]: __arm_vcvtq_f16_u16 

> (__ARM_mve_coerce(__p0, uint16x8_t)), \

>    int (*)[__ARM_mve_type_uint32x4_t]: __arm_vcvtq_f32_u32 

> (__ARM_mve_coerce(__p0, uint32x4_t)));})

>

> +#define vsubq(p0,p1) __arm_vsubq(p0,p1)

> +#define __arm_vsubq(p0,p1) ({ __typeof(p0) __p0 = (p0); \

> +  __typeof(p1) __p1 = (p1); \

> +  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \

> +  int (*)[__ARM_mve_type_float16x8_t][__ARM_mve_type_float16_t]: 

> __arm_vsubq_n_f16 (__ARM_mve_coerce(__p0, float16x8_t), 

> __ARM_mve_coerce(__p1, float16_t)), \

> +  int (*)[__ARM_mve_type_float32x4_t][__ARM_mve_type_float32_t]: 

> __arm_vsubq_n_f32 (__ARM_mve_coerce(__p0, float32x4_t), 

> __ARM_mve_coerce(__p1, float32_t)));})

> +

> +#define vbrsrq(p0,p1) __arm_vbrsrq(p0,p1)

> +#define __arm_vbrsrq(p0,p1) ({ __typeof(p0) __p0 = (p0); \

> +  _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \

> +  int (*)[__ARM_mve_type_float16x8_t]: __arm_vbrsrq_n_f16 

> (__ARM_mve_coerce(__p0, float16x8_t), p1), \

> +  int (*)[__ARM_mve_type_float32x4_t]: __arm_vbrsrq_n_f32 

> (__ARM_mve_coerce(__p0, float32x4_t), p1));})

> +

> +#define vcvtq_n(p0,p1) __arm_vcvtq_n(p0,p1)

> +#define __arm_vcvtq_n(p0,p1) ({ __typeof(p0) __p0 = (p0); \

> +  _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \

> +  int (*)[__ARM_mve_type_int16x8_t]: __arm_vcvtq_n_f16_s16 

> (__ARM_mve_coerce(__p0, int16x8_t), p1), \

> +  int (*)[__ARM_mve_type_int32x4_t]: __arm_vcvtq_n_f32_s32 

> (__ARM_mve_coerce(__p0, int32x4_t), p1), \

> +  int (*)[__ARM_mve_type_uint16x8_t]: __arm_vcvtq_n_f16_u16 

> (__ARM_mve_coerce(__p0, uint16x8_t), p1), \

> +  int (*)[__ARM_mve_type_uint32x4_t]: __arm_vcvtq_n_f32_u32 

> (__ARM_mve_coerce(__p0, uint32x4_t), p1));})

> +

>  #else /* MVE Interger.  */

>

>  #define vst4q(p0,p1) __arm_vst4q(p0,p1)

> diff --git a/gcc/config/arm/arm_mve_builtins.def 

> b/gcc/config/arm/arm_mve_builtins.def

> index 

> 816b6dfca7fb221275212ca5f06fc6f679860a38..8d1e4fac3d75e87fbe334e64e1073cb1fef0d96d 

> 100644

> --- a/gcc/config/arm/arm_mve_builtins.def

> +++ b/gcc/config/arm/arm_mve_builtins.def

> @@ -6,7 +6,7 @@

>

>      GCC is free software; you can redistribute it and/or modify it

>      under the terms of the GNU General Public License as published

> -    by the Free Software Foundation; either version 3, or (at your

> +    by the Free Software Foundation; either version 3, or (at your

>      option) any later version.

>

>      GCC is distributed in the hope that it will be useful, but WITHOUT

> @@ -76,3 +76,8 @@ VAR1 (UNOP_UNONE_UNONE, vctp32q, hi)

>  VAR1 (UNOP_UNONE_UNONE, vctp64q, hi)

>  VAR1 (UNOP_UNONE_UNONE, vctp8q, hi)

>  VAR1 (UNOP_UNONE_UNONE, vpnot, hi)

> +VAR2 (BINOP_NONE_NONE_NONE, vsubq_n_f, v8hf, v4sf)

> +VAR2 (BINOP_NONE_NONE_NONE, vbrsrq_n_f, v8hf, v4sf)

> +VAR2 (BINOP_NONE_NONE_IMM, vcvtq_n_to_f_s, v8hf, v4sf)

> +VAR2 (BINOP_NONE_UNONE_IMM, vcvtq_n_to_f_u, v8hf, v4sf)

> +VAR2 (BINOP_NONE_UNONE_UNONE, vcreateq_f, v8hf, v4sf)

> diff --git a/gcc/config/arm/constraints.md b/gcc/config/arm/constraints.md

> index 

> 4c193d20a48d5e9ed43bdca76468381fc17682da..86cd82334533acbe58be4a0c547cf38737d2bb84 

> 100644

> --- a/gcc/config/arm/constraints.md

> +++ b/gcc/config/arm/constraints.md

> @@ -35,7 +35,7 @@

>  ;; in ARM/Thumb-2 state: Da, Db, Dc, Dd, Dn, DN, Dm, Dl, DL, Do, Dv, 

> Dy, Di,

>  ;;                       Dt, Dp, Dz, Tu

>  ;; in Thumb-1 state: Pa, Pb, Pc, Pd, Pe

> -;; in Thumb-2 state: Ha, Pj, PJ, Ps, Pt, Pu, Pv, Pw, Px, Py, Pz

> +;; in Thumb-2 state: Ha, Pj, PJ, Ps, Pt, Pu, Pv, Pw, Px, Py, Pz, Rd

>  ;; in all states: Pf, Pg, UM, U1

>

>  ;; The following memory constraints have been used:

> @@ -51,6 +51,11 @@

>    "MVE EVEN registers @code{r0}, @code{r2}, @code{r4}, @code{r6}, 

> @code{r8},

>     @code{r10}, @code{r12}, @code{r14}")

>

> +(define_constraint "Rd"

> +  "@internal In Thumb-2 state a constant in range 1 to 16"

> +  (and (match_code "const_int")

> +       (match_test "TARGET_HAVE_MVE && ival >= 1 && ival <= 16")))

> +

>  (define_register_constraint "t" "TARGET_32BIT ? VFP_LO_REGS : NO_REGS"

>   "The VFP registers @code{s0}-@code{s31}.")

>

> diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md

> index 

> ee2263e04309e8274ec76ae4478afb7cfba59f8f..06701b024dabae85446bb60060a8b331e540cd6d 

> 100644

> --- a/gcc/config/arm/mve.md

> +++ b/gcc/config/arm/mve.md

> @@ -36,7 +36,9 @@

>                           VREV32Q_U VREV32Q_S VMOVLTQ_U VMOVLTQ_S 

> VMOVLBQ_S

>                           VMOVLBQ_U VCVTQ_FROM_F_S VCVTQ_FROM_F_U VCVTPQ_S

>                           VCVTPQ_U VCVTNQ_S VCVTNQ_U VCVTMQ_S VCVTMQ_U

> -                        VADDLVQ_U VCTP8Q VCTP16Q VCTP32Q VCTP64Q VPNOT])

> +                        VADDLVQ_U VCTP8Q VCTP16Q VCTP32Q VCTP64Q VPNOT

> +                        VCREATEQ_F VCVTQ_N_TO_F_S VCVTQ_N_TO_F_U 

> VBRSRQ_N_F

> +                        VSUBQ_N_F])

>

>  (define_mode_attr MVE_CNVT [(V8HI "V8HF") (V4SI "V4SF")

>                              (V8HF "V8HI") (V4SF "V4SI")])

> @@ -52,7 +54,8 @@

>                         (VCVTPQ_S "s") (VCVTPQ_U "u") (VCVTNQ_S "s")

>                         (VCVTNQ_U "u") (VCVTMQ_S "s") (VCVTMQ_U "u")

>                         (VCLZQ_U "u") (VCLZQ_S "s") (VREV32Q_U "u")

> -                      (VREV32Q_S "s") (VADDLVQ_U "u") (VADDLVQ_S "s")])

> +                      (VREV32Q_S "s") (VADDLVQ_U "u") (VADDLVQ_S "s")

> +                      (VCVTQ_N_TO_F_S "s") (VCVTQ_N_TO_F_U "u")])

>

>  (define_int_attr mode1 [(VCTP8Q "8") (VCTP16Q "16") (VCTP32Q "32")

>                          (VCTP64Q "64")])

> @@ -75,6 +78,7 @@

>  (define_int_iterator VCVTMQ [VCVTMQ_S VCVTMQ_U])

>  (define_int_iterator VADDLVQ [VADDLVQ_U VADDLVQ_S])

>  (define_int_iterator VCTPQ [VCTP8Q VCTP16Q VCTP32Q VCTP64Q])

> +(define_int_iterator VCVTQ_N_TO_F [VCVTQ_N_TO_F_S VCVTQ_N_TO_F_U])

>

>  (define_insn "*mve_mov<mode>"

>    [(set (match_operand:MVE_types 0 "s_register_operand" "=w,w,r,w,w,r,w")

> @@ -680,3 +684,62 @@

>    "vpnot"

>    [(set_attr "type" "mve_move")

>  ])

> +

> +;;

> +;; [vsubq_n_f])

> +;;

> +(define_insn "mve_vsubq_n_f<mode>"

> +  [

> +   (set (match_operand:MVE_0 0 "s_register_operand" "=w")

> +       (unspec:MVE_0 [(match_operand:MVE_0 1 "s_register_operand" "w")

> +                      (match_operand:<V_elem> 2 "s_register_operand" 

> "r")]

> +        VSUBQ_N_F))

> +  ]

> +  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"

> +  "vsub.f%#<V_sz_elem>  %q0, %q1, %2"



What's with the "f%#<V_sz_elem>" ?

I think just "f<V_sz_elem>" should work here, and in the rest of the patch.

Ok with that fixed.

Thanks,

Kyrill


> +  [(set_attr "type" "mve_move")

> +])

> +

> +;;

> +;; [vbrsrq_n_f])

> +;;

> +(define_insn "mve_vbrsrq_n_f<mode>"

> +  [

> +   (set (match_operand:MVE_0 0 "s_register_operand" "=w")

> +       (unspec:MVE_0 [(match_operand:MVE_0 1 "s_register_operand" "w")

> +                      (match_operand:SI 2 "s_register_operand" "r")]

> +        VBRSRQ_N_F))

> +  ]

> +  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"

> +  "vbrsr.%#<V_sz_elem>  %q0, %q1, %2"

> +  [(set_attr "type" "mve_move")

> +])

> +

> +;;

> +;; [vcvtq_n_to_f_s, vcvtq_n_to_f_u])

> +;;

> +(define_insn "mve_vcvtq_n_to_f_<supf><mode>"

> +  [

> +   (set (match_operand:MVE_0 0 "s_register_operand" "=w")

> +       (unspec:MVE_0 [(match_operand:<MVE_CNVT> 1 

> "s_register_operand" "w")

> +                      (match_operand:SI 2 "mve_imm_16" "Rd")]

> +        VCVTQ_N_TO_F))

> +  ]

> +  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"

> + "vcvt.f%#<V_sz_elem>.<supf>%#<V_sz_elem>\t%q0, %q1, %2"

> +  [(set_attr "type" "mve_move")

> +])

> +

> +;; [vcreateq_f])

> +;;

> +(define_insn "mve_vcreateq_f<mode>"

> +  [

> +   (set (match_operand:MVE_0 0 "s_register_operand" "=w")

> +       (unspec:MVE_0 [(match_operand:DI 1 "s_register_operand" "r")

> +                      (match_operand:DI 2 "s_register_operand" "r")]

> +        VCREATEQ_F))

> +  ]

> +  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"

> +  "vmov %q0[2], %q0[0], %Q2, %Q1\;vmov %q0[3], %q0[1], %R2, %R1"

> +  [(set_attr "type" "mve_move")

> +   (set_attr "length""8")])

> diff --git a/gcc/config/arm/predicates.md b/gcc/config/arm/predicates.md

> index 

> 9d74165fe065b03c77918fe9e4611967799535f1..80a44bb32cbce1ee3f850a44b70a0e4ceed548be 

> 100644

> --- a/gcc/config/arm/predicates.md

> +++ b/gcc/config/arm/predicates.md

> @@ -31,6 +31,10 @@

>                || REGNO_REG_CLASS (REGNO (op)) != NO_REGS));

>  })

>

> +;; True for immediates in the range of 1 to 16 for MVE.

> +(define_predicate "mve_imm_16"

> +  (match_test "satisfies_constraint_Rd (op)"))

> +

>  ; Predicate for stack protector guard's address in

>  ; stack_protect_combined_set_insn and 

> stack_protect_combined_test_insn patterns

>  (define_predicate "guard_addr_operand"

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..90f52b50e6bd0dbccc40e9da7e2f8badb593727d

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float16x8_t

> +foo (float16x8_t a, int32_t b)

> +{

> +  return vbrsrq_n_f16 (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vbrsr.16"  }  } */

> +

> +float16x8_t

> +foo1 (float16x8_t a, int32_t b)

> +{

> +  return vbrsrq (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vbrsr.16"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..4ccd8dd313319ce1933e85f76955d1ea12952abd

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float32x4_t

> +foo (float32x4_t a, int32_t b)

> +{

> +  return vbrsrq_n_f32 (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vbrsr.32"  }  } */

> +

> +float32x4_t

> +foo1 (float32x4_t a, int32_t b)

> +{

> +  return vbrsrq (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vbrsr.32"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..a83203ad0de1e6bfc6b4b9d469748108a335a391

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c

> @@ -0,0 +1,13 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float16x8_t

> +foo (uint64_t a, uint64_t b)

> +{

> +  return vcreateq_f16 (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vmov"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..279e922ed45165d0b104b236e4a2c8f782c4a59d

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c

> @@ -0,0 +1,13 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float32x4_t

> +foo (uint64_t a, uint64_t b)

> +{

> +  return vcreateq_f32 (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vmov"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..440d4df99b295deb66b91e3996cbca8f35f7d12b

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float16x8_t

> +foo (int16x8_t a)

> +{

> +  return vcvtq_n_f16_s16 (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f16.s16"  }  } */

> +

> +float16x8_t

> +foo1 (int16x8_t a)

> +{

> +  return vcvtq_n (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f16.s16"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..4d648d653ef91436bb87d2e07e15ccee0844309e

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float16x8_t

> +foo (uint16x8_t a)

> +{

> +  return vcvtq_n_f16_u16 (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f16.u16"  }  } */

> +

> +float16x8_t

> +foo1 (uint16x8_t a)

> +{

> +  return vcvtq_n (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f16.u16"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..836727a9b2d6a2621b34de5b7d4643ee1bebf47c

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float32x4_t

> +foo (int32x4_t a)

> +{

> +  return vcvtq_n_f32_s32 (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f32.s32"  }  } */

> +

> +float32x4_t

> +foo1 (int32x4_t a)

> +{

> +  return vcvtq_n (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f32.s32"  }  } */

> diff --git 

> a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..fcd66a42fd49c78e1c11ac89393dc38cb06efbf5

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float32x4_t

> +foo (uint32x4_t a)

> +{

> +  return vcvtq_n_f32_u32 (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f32.u32"  }  } */

> +

> +float32x4_t

> +foo1 (uint32x4_t a)

> +{

> +  return vcvtq_n (a, 1);

> +}

> +

> +/* { dg-final { scan-assembler "vcvt.f32.u32"  }  } */

> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..e6ff11edf3933de7acd7bdffd33d8b79c88a7f48

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float16x8_t

> +foo (float16x8_t a, float16_t b)

> +{

> +  return vsubq_n_f16 (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vsub.f16"  }  } */

> +

> +float16x8_t

> +foo1 (float16x8_t a, float16_t b)

> +{

> +  return vsubq (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vsub.f16"  }  } */

> diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c 

> b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c

> new file mode 100644

> index 

> 0000000000000000000000000000000000000000..1092fd0aa6abc2a0ed0320a17d6b563f0396fb58

> --- /dev/null

> +++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c

> @@ -0,0 +1,21 @@

> +/* { dg-do compile  } */

> +/* { dg-additional-options "-march=armv8.1-m.main+mve.fp 

> -mfloat-abi=hard -O2"  }  */

> +/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} 

> } */

> +

> +#include "arm_mve.h"

> +

> +float32x4_t

> +foo (float32x4_t a, float32_t b)

> +{

> +  return vsubq_n_f32 (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vsub.f32"  }  } */

> +

> +float32x4_t

> +foo1 (float32x4_t a, float32_t b)

> +{

> +  return vsubq (a, b);

> +}

> +

> +/* { dg-final { scan-assembler "vsub.f32"  }  } */

>

Patch

diff --git a/gcc/config/arm/arm-builtins.c b/gcc/config/arm/arm-builtins.c
index cd82aa159089c288607e240de02a85dcbb134a14..c2dad057d1365914477c64d559aa1fd1c32bbf19 100644
--- a/gcc/config/arm/arm-builtins.c
+++ b/gcc/config/arm/arm-builtins.c
@@ -349,6 +349,30 @@  arm_unop_unone_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]
 #define UNOP_UNONE_IMM_QUALIFIERS \
   (arm_unop_unone_imm_qualifiers)
 
+static enum arm_type_qualifiers
+arm_binop_none_none_none_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_none, qualifier_none };
+#define BINOP_NONE_NONE_NONE_QUALIFIERS \
+  (arm_binop_none_none_none_qualifiers)
+
+static enum arm_type_qualifiers
+arm_binop_none_none_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_none, qualifier_immediate };
+#define BINOP_NONE_NONE_IMM_QUALIFIERS \
+  (arm_binop_none_none_imm_qualifiers)
+
+static enum arm_type_qualifiers
+arm_binop_none_unone_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_unsigned, qualifier_immediate };
+#define BINOP_NONE_UNONE_IMM_QUALIFIERS \
+  (arm_binop_none_unone_imm_qualifiers)
+
+static enum arm_type_qualifiers
+arm_binop_none_unone_unone_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_none, qualifier_unsigned, qualifier_unsigned };
+#define BINOP_NONE_UNONE_UNONE_QUALIFIERS \
+  (arm_binop_none_unone_unone_qualifiers)
+
 /* End of Qualifier for MVE builtins.  */
 
    /* void ([T element type] *, T, immediate).  */
diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h
index c8d9b6471634725cea9bab3f9fa145810b506938..15b7ada025fe57b682f3873f0f275c43a30d1273 100644
--- a/gcc/config/arm/arm_mve.h
+++ b/gcc/config/arm/arm_mve.h
@@ -197,6 +197,16 @@  typedef struct { uint8x16_t val[4]; } uint8x16x4_t;
 #define vctp64q(__a) __arm_vctp64q(__a)
 #define vctp8q(__a) __arm_vctp8q(__a)
 #define vpnot(__a) __arm_vpnot(__a)
+#define vsubq_n_f16(__a, __b) __arm_vsubq_n_f16(__a, __b)
+#define vsubq_n_f32(__a, __b) __arm_vsubq_n_f32(__a, __b)
+#define vbrsrq_n_f16(__a, __b) __arm_vbrsrq_n_f16(__a, __b)
+#define vbrsrq_n_f32(__a, __b) __arm_vbrsrq_n_f32(__a, __b)
+#define vcvtq_n_f16_s16(__a,  __imm6) __arm_vcvtq_n_f16_s16(__a,  __imm6)
+#define vcvtq_n_f32_s32(__a,  __imm6) __arm_vcvtq_n_f32_s32(__a,  __imm6)
+#define vcvtq_n_f16_u16(__a,  __imm6) __arm_vcvtq_n_f16_u16(__a,  __imm6)
+#define vcvtq_n_f32_u32(__a,  __imm6) __arm_vcvtq_n_f32_u32(__a,  __imm6)
+#define vcreateq_f16(__a, __b) __arm_vcreateq_f16(__a, __b)
+#define vcreateq_f32(__a, __b) __arm_vcreateq_f32(__a, __b)
 #endif
 
 __extension__ extern __inline void
@@ -1085,6 +1095,76 @@  __arm_vcvtmq_s32_f32 (float32x4_t __a)
   return __builtin_mve_vcvtmq_sv4si (__a);
 }
 
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vsubq_n_f16 (float16x8_t __a, float16_t __b)
+{
+  return __builtin_mve_vsubq_n_fv8hf (__a, __b);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vsubq_n_f32 (float32x4_t __a, float32_t __b)
+{
+  return __builtin_mve_vsubq_n_fv4sf (__a, __b);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vbrsrq_n_f16 (float16x8_t __a, int32_t __b)
+{
+  return __builtin_mve_vbrsrq_n_fv8hf (__a, __b);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vbrsrq_n_f32 (float32x4_t __a, int32_t __b)
+{
+  return __builtin_mve_vbrsrq_n_fv4sf (__a, __b);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f16_s16 (int16x8_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_sv8hf (__a, __imm6);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f32_s32 (int32x4_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_sv4sf (__a, __imm6);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f16_u16 (uint16x8_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_uv8hf (__a, __imm6);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcvtq_n_f32_u32 (uint32x4_t __a, const int __imm6)
+{
+  return __builtin_mve_vcvtq_n_to_f_uv4sf (__a, __imm6);
+}
+
+__extension__ extern __inline float16x8_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcreateq_f16 (uint64_t __a, uint64_t __b)
+{
+  return __builtin_mve_vcreateq_fv8hf (__a, __b);
+}
+
+__extension__ extern __inline float32x4_t
+__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
+__arm_vcreateq_f32 (uint64_t __a, uint64_t __b)
+{
+  return __builtin_mve_vcreateq_fv4sf (__a, __b);
+}
+
 #endif
 
 enum {
@@ -1373,6 +1453,27 @@  extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint16x8_t]: __arm_vcvtq_f16_u16 (__ARM_mve_coerce(__p0, uint16x8_t)), \
   int (*)[__ARM_mve_type_uint32x4_t]: __arm_vcvtq_f32_u32 (__ARM_mve_coerce(__p0, uint32x4_t)));})
 
+#define vsubq(p0,p1) __arm_vsubq(p0,p1)
+#define __arm_vsubq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
+  __typeof(p1) __p1 = (p1); \
+  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
+  int (*)[__ARM_mve_type_float16x8_t][__ARM_mve_type_float16_t]: __arm_vsubq_n_f16 (__ARM_mve_coerce(__p0, float16x8_t), __ARM_mve_coerce(__p1, float16_t)), \
+  int (*)[__ARM_mve_type_float32x4_t][__ARM_mve_type_float32_t]: __arm_vsubq_n_f32 (__ARM_mve_coerce(__p0, float32x4_t), __ARM_mve_coerce(__p1, float32_t)));})
+
+#define vbrsrq(p0,p1) __arm_vbrsrq(p0,p1)
+#define __arm_vbrsrq(p0,p1) ({ __typeof(p0) __p0 = (p0); \
+  _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \
+  int (*)[__ARM_mve_type_float16x8_t]: __arm_vbrsrq_n_f16 (__ARM_mve_coerce(__p0, float16x8_t), p1), \
+  int (*)[__ARM_mve_type_float32x4_t]: __arm_vbrsrq_n_f32 (__ARM_mve_coerce(__p0, float32x4_t), p1));})
+
+#define vcvtq_n(p0,p1) __arm_vcvtq_n(p0,p1)
+#define __arm_vcvtq_n(p0,p1) ({ __typeof(p0) __p0 = (p0); \
+  _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \
+  int (*)[__ARM_mve_type_int16x8_t]: __arm_vcvtq_n_f16_s16 (__ARM_mve_coerce(__p0, int16x8_t), p1), \
+  int (*)[__ARM_mve_type_int32x4_t]: __arm_vcvtq_n_f32_s32 (__ARM_mve_coerce(__p0, int32x4_t), p1), \
+  int (*)[__ARM_mve_type_uint16x8_t]: __arm_vcvtq_n_f16_u16 (__ARM_mve_coerce(__p0, uint16x8_t), p1), \
+  int (*)[__ARM_mve_type_uint32x4_t]: __arm_vcvtq_n_f32_u32 (__ARM_mve_coerce(__p0, uint32x4_t), p1));})
+
 #else /* MVE Interger.  */
 
 #define vst4q(p0,p1) __arm_vst4q(p0,p1)
diff --git a/gcc/config/arm/arm_mve_builtins.def b/gcc/config/arm/arm_mve_builtins.def
index 816b6dfca7fb221275212ca5f06fc6f679860a38..8d1e4fac3d75e87fbe334e64e1073cb1fef0d96d 100644
--- a/gcc/config/arm/arm_mve_builtins.def
+++ b/gcc/config/arm/arm_mve_builtins.def
@@ -6,7 +6,7 @@ 
 
     GCC is free software; you can redistribute it and/or modify it
     under the terms of the GNU General Public License as published
-    by the Free Software Foundation; either version 3, or (at your
+    by the Free Software Foundation; either version 3, or  (at your
     option) any later version.
 
     GCC is distributed in the hope that it will be useful, but WITHOUT
@@ -76,3 +76,8 @@  VAR1 (UNOP_UNONE_UNONE, vctp32q, hi)
 VAR1 (UNOP_UNONE_UNONE, vctp64q, hi)
 VAR1 (UNOP_UNONE_UNONE, vctp8q, hi)
 VAR1 (UNOP_UNONE_UNONE, vpnot, hi)
+VAR2 (BINOP_NONE_NONE_NONE, vsubq_n_f, v8hf, v4sf)
+VAR2 (BINOP_NONE_NONE_NONE, vbrsrq_n_f, v8hf, v4sf)
+VAR2 (BINOP_NONE_NONE_IMM, vcvtq_n_to_f_s, v8hf, v4sf)
+VAR2 (BINOP_NONE_UNONE_IMM, vcvtq_n_to_f_u, v8hf, v4sf)
+VAR2 (BINOP_NONE_UNONE_UNONE, vcreateq_f, v8hf, v4sf)
diff --git a/gcc/config/arm/constraints.md b/gcc/config/arm/constraints.md
index 4c193d20a48d5e9ed43bdca76468381fc17682da..86cd82334533acbe58be4a0c547cf38737d2bb84 100644
--- a/gcc/config/arm/constraints.md
+++ b/gcc/config/arm/constraints.md
@@ -35,7 +35,7 @@ 
 ;; in ARM/Thumb-2 state: Da, Db, Dc, Dd, Dn, DN, Dm, Dl, DL, Do, Dv, Dy, Di,
 ;;			 Dt, Dp, Dz, Tu
 ;; in Thumb-1 state: Pa, Pb, Pc, Pd, Pe
-;; in Thumb-2 state: Ha, Pj, PJ, Ps, Pt, Pu, Pv, Pw, Px, Py, Pz
+;; in Thumb-2 state: Ha, Pj, PJ, Ps, Pt, Pu, Pv, Pw, Px, Py, Pz, Rd
 ;; in all states: Pf, Pg, UM, U1
 
 ;; The following memory constraints have been used:
@@ -51,6 +51,11 @@ 
   "MVE EVEN registers @code{r0}, @code{r2}, @code{r4}, @code{r6}, @code{r8},
    @code{r10}, @code{r12}, @code{r14}")
 
+(define_constraint "Rd"
+  "@internal In Thumb-2 state a constant in range 1 to 16"
+  (and (match_code "const_int")
+       (match_test "TARGET_HAVE_MVE && ival >= 1 && ival <= 16")))
+
 (define_register_constraint "t" "TARGET_32BIT ? VFP_LO_REGS : NO_REGS"
  "The VFP registers @code{s0}-@code{s31}.")
 
diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md
index ee2263e04309e8274ec76ae4478afb7cfba59f8f..06701b024dabae85446bb60060a8b331e540cd6d 100644
--- a/gcc/config/arm/mve.md
+++ b/gcc/config/arm/mve.md
@@ -36,7 +36,9 @@ 
 			 VREV32Q_U VREV32Q_S VMOVLTQ_U VMOVLTQ_S VMOVLBQ_S
 			 VMOVLBQ_U VCVTQ_FROM_F_S VCVTQ_FROM_F_U VCVTPQ_S
 			 VCVTPQ_U VCVTNQ_S VCVTNQ_U VCVTMQ_S VCVTMQ_U
-			 VADDLVQ_U VCTP8Q VCTP16Q VCTP32Q VCTP64Q VPNOT])
+			 VADDLVQ_U VCTP8Q VCTP16Q VCTP32Q VCTP64Q VPNOT
+			 VCREATEQ_F VCVTQ_N_TO_F_S VCVTQ_N_TO_F_U VBRSRQ_N_F
+			 VSUBQ_N_F])
 
 (define_mode_attr MVE_CNVT [(V8HI "V8HF") (V4SI "V4SF")
 			    (V8HF "V8HI") (V4SF "V4SI")])
@@ -52,7 +54,8 @@ 
 		       (VCVTPQ_S "s") (VCVTPQ_U "u") (VCVTNQ_S "s")
 		       (VCVTNQ_U "u") (VCVTMQ_S "s") (VCVTMQ_U "u")
 		       (VCLZQ_U "u") (VCLZQ_S "s") (VREV32Q_U "u")
-		       (VREV32Q_S "s") (VADDLVQ_U "u") (VADDLVQ_S "s")])
+		       (VREV32Q_S "s") (VADDLVQ_U "u") (VADDLVQ_S "s")
+		       (VCVTQ_N_TO_F_S "s") (VCVTQ_N_TO_F_U "u")])
 
 (define_int_attr mode1 [(VCTP8Q "8") (VCTP16Q "16") (VCTP32Q "32")
 			(VCTP64Q "64")])
@@ -75,6 +78,7 @@ 
 (define_int_iterator VCVTMQ [VCVTMQ_S VCVTMQ_U])
 (define_int_iterator VADDLVQ [VADDLVQ_U VADDLVQ_S])
 (define_int_iterator VCTPQ [VCTP8Q VCTP16Q VCTP32Q VCTP64Q])
+(define_int_iterator VCVTQ_N_TO_F [VCVTQ_N_TO_F_S VCVTQ_N_TO_F_U])
 
 (define_insn "*mve_mov<mode>"
   [(set (match_operand:MVE_types 0 "s_register_operand" "=w,w,r,w,w,r,w")
@@ -680,3 +684,62 @@ 
   "vpnot"
   [(set_attr "type" "mve_move")
 ])
+
+;;
+;; [vsubq_n_f])
+;;
+(define_insn "mve_vsubq_n_f<mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:MVE_0 1 "s_register_operand" "w")
+		       (match_operand:<V_elem> 2 "s_register_operand" "r")]
+	 VSUBQ_N_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vsub.f%#<V_sz_elem>  %q0, %q1, %2"
+  [(set_attr "type" "mve_move")
+])
+
+;;
+;; [vbrsrq_n_f])
+;;
+(define_insn "mve_vbrsrq_n_f<mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:MVE_0 1 "s_register_operand" "w")
+		       (match_operand:SI 2 "s_register_operand" "r")]
+	 VBRSRQ_N_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vbrsr.%#<V_sz_elem>  %q0, %q1, %2"
+  [(set_attr "type" "mve_move")
+])
+
+;;
+;; [vcvtq_n_to_f_s, vcvtq_n_to_f_u])
+;;
+(define_insn "mve_vcvtq_n_to_f_<supf><mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:<MVE_CNVT> 1 "s_register_operand" "w")
+		       (match_operand:SI 2 "mve_imm_16" "Rd")]
+	 VCVTQ_N_TO_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vcvt.f%#<V_sz_elem>.<supf>%#<V_sz_elem>\t%q0, %q1, %2"
+  [(set_attr "type" "mve_move")
+])
+
+;; [vcreateq_f])
+;;
+(define_insn "mve_vcreateq_f<mode>"
+  [
+   (set (match_operand:MVE_0 0 "s_register_operand" "=w")
+	(unspec:MVE_0 [(match_operand:DI 1 "s_register_operand" "r")
+		       (match_operand:DI 2 "s_register_operand" "r")]
+	 VCREATEQ_F))
+  ]
+  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
+  "vmov %q0[2], %q0[0], %Q2, %Q1\;vmov %q0[3], %q0[1], %R2, %R1"
+  [(set_attr "type" "mve_move")
+   (set_attr "length""8")])
diff --git a/gcc/config/arm/predicates.md b/gcc/config/arm/predicates.md
index 9d74165fe065b03c77918fe9e4611967799535f1..80a44bb32cbce1ee3f850a44b70a0e4ceed548be 100644
--- a/gcc/config/arm/predicates.md
+++ b/gcc/config/arm/predicates.md
@@ -31,6 +31,10 @@ 
 	      || REGNO_REG_CLASS (REGNO (op)) != NO_REGS));
 })
 
+;; True for immediates in the range of 1 to 16 for MVE.
+(define_predicate "mve_imm_16"
+  (match_test "satisfies_constraint_Rd (op)"))
+
 ; Predicate for stack protector guard's address in
 ; stack_protect_combined_set_insn and stack_protect_combined_test_insn patterns
 (define_predicate "guard_addr_operand"
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c
new file mode 100644
index 0000000000000000000000000000000000000000..90f52b50e6bd0dbccc40e9da7e2f8badb593727d
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f16.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (float16x8_t a, int32_t b)
+{
+  return vbrsrq_n_f16 (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.16"  }  } */
+
+float16x8_t
+foo1 (float16x8_t a, int32_t b)
+{
+  return vbrsrq (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c
new file mode 100644
index 0000000000000000000000000000000000000000..4ccd8dd313319ce1933e85f76955d1ea12952abd
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vbrsrq_n_f32.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (float32x4_t a, int32_t b)
+{
+  return vbrsrq_n_f32 (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.32"  }  } */
+
+float32x4_t
+foo1 (float32x4_t a, int32_t b)
+{
+  return vbrsrq (a, b);
+}
+
+/* { dg-final { scan-assembler "vbrsr.32"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c
new file mode 100644
index 0000000000000000000000000000000000000000..a83203ad0de1e6bfc6b4b9d469748108a335a391
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f16.c
@@ -0,0 +1,13 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (uint64_t a, uint64_t b)
+{
+  return vcreateq_f16 (a, b);
+}
+
+/* { dg-final { scan-assembler "vmov"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c
new file mode 100644
index 0000000000000000000000000000000000000000..279e922ed45165d0b104b236e4a2c8f782c4a59d
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcreateq_f32.c
@@ -0,0 +1,13 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (uint64_t a, uint64_t b)
+{
+  return vcreateq_f32 (a, b);
+}
+
+/* { dg-final { scan-assembler "vmov"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c
new file mode 100644
index 0000000000000000000000000000000000000000..440d4df99b295deb66b91e3996cbca8f35f7d12b
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_s16.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (int16x8_t a)
+{
+  return vcvtq_n_f16_s16 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.s16"  }  } */
+
+float16x8_t
+foo1 (int16x8_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.s16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c
new file mode 100644
index 0000000000000000000000000000000000000000..4d648d653ef91436bb87d2e07e15ccee0844309e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f16_u16.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (uint16x8_t a)
+{
+  return vcvtq_n_f16_u16 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.u16"  }  } */
+
+float16x8_t
+foo1 (uint16x8_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f16.u16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c
new file mode 100644
index 0000000000000000000000000000000000000000..836727a9b2d6a2621b34de5b7d4643ee1bebf47c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_s32.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (int32x4_t a)
+{
+  return vcvtq_n_f32_s32 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.s32"  }  } */
+
+float32x4_t
+foo1 (int32x4_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.s32"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c
new file mode 100644
index 0000000000000000000000000000000000000000..fcd66a42fd49c78e1c11ac89393dc38cb06efbf5
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vcvtq_n_f32_u32.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (uint32x4_t a)
+{
+  return vcvtq_n_f32_u32 (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.u32"  }  } */
+
+float32x4_t
+foo1 (uint32x4_t a)
+{
+  return vcvtq_n (a, 1);
+}
+
+/* { dg-final { scan-assembler "vcvt.f32.u32"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c
new file mode 100644
index 0000000000000000000000000000000000000000..e6ff11edf3933de7acd7bdffd33d8b79c88a7f48
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f16.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float16x8_t
+foo (float16x8_t a, float16_t b)
+{
+  return vsubq_n_f16 (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f16"  }  } */
+
+float16x8_t
+foo1 (float16x8_t a, float16_t b)
+{
+  return vsubq (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f16"  }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c
new file mode 100644
index 0000000000000000000000000000000000000000..1092fd0aa6abc2a0ed0320a17d6b563f0396fb58
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/vsubq_n_f32.c
@@ -0,0 +1,21 @@ 
+/* { dg-do compile  } */
+/* { dg-additional-options "-march=armv8.1-m.main+mve.fp -mfloat-abi=hard -O2"  }  */
+/* { dg-skip-if "Skip if not auto" {*-*-*} {"-mfpu=*"} {"-mfpu=auto"} } */
+
+#include "arm_mve.h"
+
+float32x4_t
+foo (float32x4_t a, float32_t b)
+{
+  return vsubq_n_f32 (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f32"  }  } */
+
+float32x4_t
+foo1 (float32x4_t a, float32_t b)
+{
+  return vsubq (a, b);
+}
+
+/* { dg-final { scan-assembler "vsub.f32"  }  } */