arm: [MVE intrinsics] Rework MVE vld/vst intrinsics

Implement the mve vld and vst intrinsics using the MVE builtins framework.

The main part of the patch is to reimplement to vstr/vldr patterns
such that we now have much fewer of them:
- non-truncating stores
- predicated non-truncating stores
- truncating stores
- predicated truncating stores
- non-extending loads
- predicated non-extending loads
- extending loads
- predicated extending loads

This enables us to update the implementation of vld1/vst1 and use the
new vldr/vstr builtins.

The patch also adds support for the predicated vld1/vst1 versions.

gcc.target/arm/pr112337.c needs an update, to call the intrinsic
instead of the builtin, which this patch deletes.

2024-09-11  Alfie Richards  <Alfie.Richards@arm.com>
	    Christophe Lyon  <christophe.lyon@arm.com>

	gcc/

	* config/arm/arm-mve-builtins-base.cc (vld1q_impl): Add support
	for predicated version.
	(vst1q_impl): Likewise.
	(vstrq_impl): New class.
	(vldrq_impl): New class.
	(vldrbq): New.
	(vldrhq): New.
	(vldrwq): New.
	(vstrbq): New.
	(vstrhq): New.
	(vstrwq): New.
	* config/arm/arm-mve-builtins-base.def (vld1q): Add predicated
	version.
	(vldrbq): New.
	(vldrhq): New.
	(vldrwq): New.
	(vst1q): Add predicated version.
	(vstrbq): New.
	(vstrhq): New.
	(vstrwq): New.
	(vrev32q): Update types to float_16.
	* config/arm/arm-mve-builtins-base.h (vldrbq): New.
	(vldrhq): New.
	(vldrwq): New.
	(vstrbq): New.
	(vstrhq): New.
	(vstrwq): New.
	* config/arm/arm-mve-builtins-functions.h (memory_vector_mode):
	Remove conversion of floating point vectors to integer.
	* config/arm/arm-mve-builtins.cc (TYPES_float16): Change to...
	(TYPES_float_16): ...this.
	(TYPES_float_32): New.
	(float16): Change to...
	(float_16): ...this.
	(float_32): New.
	(preds_z_or_none): New.
	(function_resolver::check_gp_argument): Add support for _z
	predicate.
	* config/arm/arm_mve.h (vstrbq): Remove.
	(vstrbq_p): Likewise.
	(vstrhq): Likewise.
	(vstrhq_p): Likewise.
	(vstrwq): Likewise.
	(vstrwq_p): Likewise.
	(vst1q_p): Likewise.
	(vld1q_z): Likewise.
	(vldrbq_s8): Likewise.
	(vldrbq_u8): Likewise.
	(vldrbq_s16): Likewise.
	(vldrbq_u16): Likewise.
	(vldrbq_s32): Likewise.
	(vldrbq_u32): Likewise.
	(vstrbq_s8): Likewise.
	(vstrbq_s32): Likewise.
	(vstrbq_s16): Likewise.
	(vstrbq_u8): Likewise.
	(vstrbq_u32): Likewise.
	(vstrbq_u16): Likewise.
	(vstrbq_p_s8): Likewise.
	(vstrbq_p_s32): Likewise.
	(vstrbq_p_s16): Likewise.
	(vstrbq_p_u8): Likewise.
	(vstrbq_p_u32): Likewise.
	(vstrbq_p_u16): Likewise.
	(vldrbq_z_s16): Likewise.
	(vldrbq_z_u8): Likewise.
	(vldrbq_z_s8): Likewise.
	(vldrbq_z_s32): Likewise.
	(vldrbq_z_u16): Likewise.
	(vldrbq_z_u32): Likewise.
	(vldrhq_s32): Likewise.
	(vldrhq_s16): Likewise.
	(vldrhq_u32): Likewise.
	(vldrhq_u16): Likewise.
	(vldrhq_z_s32): Likewise.
	(vldrhq_z_s16): Likewise.
	(vldrhq_z_u32): Likewise.
	(vldrhq_z_u16): Likewise.
	(vldrwq_s32): Likewise.
	(vldrwq_u32): Likewise.
	(vldrwq_z_s32): Likewise.
	(vldrwq_z_u32): Likewise.
	(vldrhq_f16): Likewise.
	(vldrhq_z_f16): Likewise.
	(vldrwq_f32): Likewise.
	(vldrwq_z_f32): Likewise.
	(vstrhq_f16): Likewise.
	(vstrhq_s32): Likewise.
	(vstrhq_s16): Likewise.
	(vstrhq_u32): Likewise.
	(vstrhq_u16): Likewise.
	(vstrhq_p_f16): Likewise.
	(vstrhq_p_s32): Likewise.
	(vstrhq_p_s16): Likewise.
	(vstrhq_p_u32): Likewise.
	(vstrhq_p_u16): Likewise.
	(vstrwq_f32): Likewise.
	(vstrwq_s32): Likewise.
	(vstrwq_u32): Likewise.
	(vstrwq_p_f32): Likewise.
	(vstrwq_p_s32): Likewise.
	(vstrwq_p_u32): Likewise.
	(vst1q_p_u8): Likewise.
	(vst1q_p_s8): Likewise.
	(vld1q_z_u8): Likewise.
	(vld1q_z_s8): Likewise.
	(vst1q_p_u16): Likewise.
	(vst1q_p_s16): Likewise.
	(vld1q_z_u16): Likewise.
	(vld1q_z_s16): Likewise.
	(vst1q_p_u32): Likewise.
	(vst1q_p_s32): Likewise.
	(vld1q_z_u32): Likewise.
	(vld1q_z_s32): Likewise.
	(vld1q_z_f16): Likewise.
	(vst1q_p_f16): Likewise.
	(vld1q_z_f32): Likewise.
	(vst1q_p_f32): Likewise.
	(__arm_vstrbq_s8): Likewise.
	(__arm_vstrbq_s32): Likewise.
	(__arm_vstrbq_s16): Likewise.
	(__arm_vstrbq_u8): Likewise.
	(__arm_vstrbq_u32): Likewise.
	(__arm_vstrbq_u16): Likewise.
	(__arm_vldrbq_s8): Likewise.
	(__arm_vldrbq_u8): Likewise.
	(__arm_vldrbq_s16): Likewise.
	(__arm_vldrbq_u16): Likewise.
	(__arm_vldrbq_s32): Likewise.
	(__arm_vldrbq_u32): Likewise.
	(__arm_vstrbq_p_s8): Likewise.
	(__arm_vstrbq_p_s32): Likewise.
	(__arm_vstrbq_p_s16): Likewise.
	(__arm_vstrbq_p_u8): Likewise.
	(__arm_vstrbq_p_u32): Likewise.
	(__arm_vstrbq_p_u16): Likewise.
	(__arm_vldrbq_z_s8): Likewise.
	(__arm_vldrbq_z_s32): Likewise.
	(__arm_vldrbq_z_s16): Likewise.
	(__arm_vldrbq_z_u8): Likewise.
	(__arm_vldrbq_z_u32): Likewise.
	(__arm_vldrbq_z_u16): Likewise.
	(__arm_vldrhq_s32): Likewise.
	(__arm_vldrhq_s16): Likewise.
	(__arm_vldrhq_u32): Likewise.
	(__arm_vldrhq_u16): Likewise.
	(__arm_vldrhq_z_s32): Likewise.
	(__arm_vldrhq_z_s16): Likewise.
	(__arm_vldrhq_z_u32): Likewise.
	(__arm_vldrhq_z_u16): Likewise.
	(__arm_vldrwq_s32): Likewise.
	(__arm_vldrwq_u32): Likewise.
	(__arm_vldrwq_z_s32): Likewise.
	(__arm_vldrwq_z_u32): Likewise.
	(__arm_vstrhq_s32): Likewise.
	(__arm_vstrhq_s16): Likewise.
	(__arm_vstrhq_u32): Likewise.
	(__arm_vstrhq_u16): Likewise.
	(__arm_vstrhq_p_s32): Likewise.
	(__arm_vstrhq_p_s16): Likewise.
	(__arm_vstrhq_p_u32): Likewise.
	(__arm_vstrhq_p_u16): Likewise.
	(__arm_vstrwq_s32): Likewise.
	(__arm_vstrwq_u32): Likewise.
	(__arm_vstrwq_p_s32): Likewise.
	(__arm_vstrwq_p_u32): Likewise.
	(__arm_vst1q_p_u8): Likewise.
	(__arm_vst1q_p_s8): Likewise.
	(__arm_vld1q_z_u8): Likewise.
	(__arm_vld1q_z_s8): Likewise.
	(__arm_vst1q_p_u16): Likewise.
	(__arm_vst1q_p_s16): Likewise.
	(__arm_vld1q_z_u16): Likewise.
	(__arm_vld1q_z_s16): Likewise.
	(__arm_vst1q_p_u32): Likewise.
	(__arm_vst1q_p_s32): Likewise.
	(__arm_vld1q_z_u32): Likewise.
	(__arm_vld1q_z_s32): Likewise.
	(__arm_vldrwq_f32): Likewise.
	(__arm_vldrwq_z_f32): Likewise.
	(__arm_vldrhq_z_f16): Likewise.
	(__arm_vldrhq_f16): Likewise.
	(__arm_vstrwq_p_f32): Likewise.
	(__arm_vstrwq_f32): Likewise.
	(__arm_vstrhq_f16): Likewise.
	(__arm_vstrhq_p_f16): Likewise.
	(__arm_vld1q_z_f16): Likewise.
	(__arm_vst1q_p_f16): Likewise.
	(__arm_vld1q_z_f32): Likewise.
	(__arm_vst2q_f32): Likewise.
	(__arm_vst1q_p_f32): Likewise.
	(__arm_vstrbq): Likewise.
	(__arm_vstrbq_p): Likewise.
	(__arm_vstrhq): Likewise.
	(__arm_vstrhq_p): Likewise.
	(__arm_vstrwq): Likewise.
	(__arm_vstrwq_p): Likewise.
	(__arm_vst1q_p): Likewise.
	(__arm_vld1q_z): Likewise.
	* config/arm/arm_mve_builtins.def:
	(vstrbq_s): Delete.
	(vstrbq_u): Likewise.
	(vldrbq_s): Likewise.
	(vldrbq_u): Likewise.
	(vstrbq_p_s): Likewise.
	(vstrbq_p_u): Likewise.
	(vldrbq_z_s): Likewise.
	(vldrbq_z_u): Likewise.
	(vld1q_u): Likewise.
	(vld1q_s): Likewise.
	(vldrhq_z_u): Likewise.
	(vldrhq_u): Likewise.
	(vldrhq_z_s): Likewise.
	(vldrhq_s): Likewise.
	(vld1q_f): Likewise.
	(vldrhq_f): Likewise.
	(vldrhq_z_f): Likewise.
	(vldrwq_f): Likewise.
	(vldrwq_s): Likewise.
	(vldrwq_u): Likewise.
	(vldrwq_z_f): Likewise.
	(vldrwq_z_s): Likewise.
	(vldrwq_z_u): Likewise.
	(vst1q_u): Likewise.
	(vst1q_s): Likewise.
	(vstrhq_p_u): Likewise.
	(vstrhq_u): Likewise.
	(vstrhq_p_s): Likewise.
	(vstrhq_s): Likewise.
	(vst1q_f): Likewise.
	(vstrhq_f): Likewise.
	(vstrhq_p_f): Likewise.
	(vstrwq_f): Likewise.
	(vstrwq_s): Likewise.
	(vstrwq_u): Likewise.
	(vstrwq_p_f): Likewise.
	(vstrwq_p_s): Likewise.
	(vstrwq_p_u): Likewise.
	* config/arm/iterators.md (MVE_w_narrow_TYPE): New iterator.
	(MVE_w_narrow_type): New iterator.
	(MVE_wide_n_TYPE): New attribute.
	(MVE_wide_n_type): New attribute.
	(MVE_wide_n_sz_elem): New attribute.
	(MVE_wide_n_VPRED): New attribute.
	(MVE_elem_ch): New attribute.
	(supf): Remove VSTRBQ_S, VSTRBQ_U, VLDRBQ_S, VLDRBQ_U, VLD1Q_S,
	VLD1Q_U, VLDRHQ_S, VLDRHQ_U, VLDRWQ_S, VLDRWQ_U, VST1Q_S, VST1Q_U,
	VSTRHQ_S, VSTRHQ_U, VSTRWQ_S, VSTRWQ_U.
	(VSTRBQ, VLDRBQ, VLD1Q, VLDRHQ, VLDRWQ, VST1Q, VSTRHQ, VSTRWQ):
	Delete.
	* config/arm/mve.md (mve_vstrbq_<supf><mode>): Remove.
	(mve_vldrbq_<supf><mode>): Likewise.
	(mve_vstrbq_p_<supf><mode>): Likewise.
	(mve_vldrbq_z_<supf><mode>): Likewise.
	(mve_vldrhq_fv8hf): Likewise.
	(mve_vldrhq_<supf><mode>): Likewise.
	(mve_vldrhq_z_fv8hf): Likewise.
	(mve_vldrhq_z_<supf><mode>): Likewise.
	(mve_vldrwq_fv4sf): Likewise.
	(mve_vldrwq_<supf>v4si): Likewise.
	(mve_vldrwq_z_fv4sf): Likewise.
	(mve_vldrwq_z_<supf>v4si): Likewise.
	(@mve_vld1q_f<mode>): Likewise.
	(@mve_vld1q_<supf><mode>): Likewise.
	(mve_vstrhq_fv8hf): Likewise.
	(mve_vstrhq_p_fv8hf): Likewise.
	(mve_vstrhq_p_<supf><mode>): Likewise.
	(mve_vstrhq_<supf><mode>): Likewise.
	(mve_vstrwq_fv4sf): Likewise.
	(mve_vstrwq_p_fv4sf): Likewise.
	(mve_vstrwq_p_<supf>v4si): Likewise.
	(mve_vstrwq_<supf>v4si): Likewise.
	(@mve_vst1q_f<mode>): Likewise.
	(@mve_vst1q_<supf><mode>): Likewise.
	(@mve_vstrq_<mode>): New.
	(@mve_vstrq_p_<mode>): New.
	(@mve_vstrq_truncate_<mode>): New.
	(@mve_vstrq_p_truncate_<mode>): New.
	(@mve_vldrq_<mode>): New.
	(@mve_vldrq_z_<mode>): New.
	(@mve_vldrq_extend_<mode><US>): New.
	(@mve_vldrq_z_extend_<mode><US>): New.
	* config/arm/unspecs.md:
	(VSTRBQ_S): Remove.
	(VSTRBQ_U): Likewise.
	(VLDRBQ_S): Likewise.
	(VLDRBQ_U): Likewise.
	(VLD1Q_F): Likewise.
	(VLD1Q_S): Likewise.
	(VLD1Q_U): Likewise.
	(VLDRHQ_F): Likewise.
	(VLDRHQ_U): Likewise.
	(VLDRHQ_S): Likewise.
	(VLDRWQ_F): Likewise.
	(VLDRWQ_S): Likewise.
	(VLDRWQ_U): Likewise.
	(VSTRHQ_F): Likewise.
	(VST1Q_S): Likewise.
	(VST1Q_U): Likewise.
	(VSTRHQ_U): Likewise.
	(VSTRWQ_S): Likewise.
	(VSTRWQ_U): Likewise.
	(VSTRWQ_F): Likewise.
	(VST1Q_F): Likewise.
	(VLDRQ): New.
	(VLDRQ_Z): Likewise.
	(VLDRQ_EXT): Likewise.
	(VLDRQ_EXT_Z): Likewise.
	(VSTRQ): Likewise.
	(VSTRQ_P): Likewise.
	(VSTRQ_TRUNC): Likewise.
	(VSTRQ_TRUNC_P): Likewise.

	gcc/testsuite/
	* gcc.target/arm/pr112337.c: Call intrinsic instead of builtin.

11 files changed, 369 insertions, 1579 deletions

diff --git a/gcc/config/arm/arm-mve-builtins-base.cc b/gcc/config/arm/arm-mve-builtins-base.cc
index cadd413..2c8ff46 100644
--- a/gcc/config/arm/arm-mve-builtins-base.cc
+++ b/gcc/config/arm/arm-mve-builtins-base.cc
@@ -150,16 +150,18 @@ public:
   expand (function_expander &e) const override
   {
     insn_code icode;
-    if (e.type_suffix (0).float_p)
-      icode = code_for_mve_vld1q_f(e.vector_mode (0));
-    else
+    switch (e.pred)
       {
-	if (e.type_suffix (0).unsigned_p)
-	  icode = code_for_mve_vld1q(VLD1Q_U,
-				     e.vector_mode (0));
-	else
-	  icode = code_for_mve_vld1q(VLD1Q_S,
-				     e.vector_mode (0));
+      case PRED_none:
+	icode = code_for_mve_vldrq (e.vector_mode (0));
+	break;
+
+      case PRED_z:
+	icode = code_for_mve_vldrq_z (e.vector_mode (0));
+	break;
+
+      default:
+	gcc_unreachable ();
       }
     return e.use_contiguous_load_insn (icode);
   }
@@ -178,21 +180,113 @@ public:
   expand (function_expander &e) const override
   {
     insn_code icode;
-    if (e.type_suffix (0).float_p)
-      icode = code_for_mve_vst1q_f(e.vector_mode (0));
-    else
+    switch (e.pred)
       {
-	if (e.type_suffix (0).unsigned_p)
-	  icode = code_for_mve_vst1q(VST1Q_U,
-				     e.vector_mode (0));
+      case PRED_none:
+	icode = code_for_mve_vstrq (e.vector_mode (0));
+	break;
+
+      case PRED_p:
+	icode = code_for_mve_vstrq_p (e.vector_mode (0));
+	break;
+
+      default:
+	gcc_unreachable ();
+      }
+    return e.use_contiguous_store_insn (icode);
+  }
+};
+
+/* Builds the vstrq* intrinsics.  */
+class vstrq_impl : public store_truncating
+{
+public:
+  using store_truncating::store_truncating;
+
+  unsigned int call_properties (const function_instance &) const override
+  {
+    return CP_WRITE_MEMORY;
+  }
+
+  rtx expand (function_expander &e) const override
+  {
+    insn_code icode;
+    switch (e.pred)
+      {
+      case PRED_none:
+	if (e.vector_mode (0) == e.memory_vector_mode ())
+	  /* Non-truncating store case.  */
+	  icode = code_for_mve_vstrq (e.vector_mode (0));
+	else
+	  /* Truncating store case.
+	     (there is only one possible truncation for each memory mode so only
+	     one mode argument is needed).  */
+	  icode = code_for_mve_vstrq_truncate (e.memory_vector_mode ());
+	break;
+
+      case PRED_p:
+	if (e.vector_mode (0) == e.memory_vector_mode ())
+	  icode = code_for_mve_vstrq_p (e.vector_mode (0));
 	else
-	  icode = code_for_mve_vst1q(VST1Q_S,
-				     e.vector_mode (0));
+	  icode = code_for_mve_vstrq_p_truncate (e.memory_vector_mode ());
+	break;
+
+      default:
+	gcc_unreachable ();
       }
+
     return e.use_contiguous_store_insn (icode);
   }
 };
 
+/* Builds the vldrq* intrinsics.  */
+class vldrq_impl : public load_extending
+{
+public:
+  using load_extending::load_extending;
+
+  unsigned int call_properties (const function_instance &) const override
+  {
+    return CP_READ_MEMORY;
+  }
+
+  rtx expand (function_expander &e) const override
+  {
+    insn_code icode;
+    switch (e.pred)
+      {
+      case PRED_none:
+	if (e.vector_mode (0) == e.memory_vector_mode ())
+	  /* Non-extending load case.  */
+	  icode = code_for_mve_vldrq (e.vector_mode (0));
+	else
+	  /* Extending load case.
+	     (there is only one extension for each memory mode so only one type
+	     argument is needed).  */
+	  icode = code_for_mve_vldrq_extend (e.memory_vector_mode (),
+					     e.type_suffix (0).unsigned_p
+					     ? ZERO_EXTEND
+					     : SIGN_EXTEND);
+	break;
+
+      case PRED_z:
+	if (e.vector_mode (0) == e.memory_vector_mode ())
+	  icode = code_for_mve_vldrq_z (e.vector_mode (0));
+	else
+	  icode = code_for_mve_vldrq_z_extend (e.memory_vector_mode (),
+					       e.type_suffix (0).unsigned_p
+					       ? ZERO_EXTEND
+					       : SIGN_EXTEND);
+	break;
+
+      default:
+	gcc_unreachable ();
+      }
+
+    return e.use_contiguous_load_insn (icode);
+  }
+};
+
   /* Implements vctp8q, vctp16q, vctp32q and vctp64q intrinsics.  */
 class vctpq_impl : public function_base
 {
@@ -989,6 +1083,9 @@ FUNCTION (vfmsq, unspec_mve_function_exact_insn, (-1, -1, VFMSQ_F, -1, -1, -1, -
 FUNCTION_WITH_M_N_NO_F (vhaddq, VHADDQ)
 FUNCTION_WITH_M_N_NO_F (vhsubq, VHSUBQ)
 FUNCTION (vld1q, vld1_impl,)
+FUNCTION (vldrbq, vldrq_impl, (TYPE_SUFFIX_s8, TYPE_SUFFIX_u8))
+FUNCTION (vldrhq, vldrq_impl, (TYPE_SUFFIX_s16, TYPE_SUFFIX_u16, TYPE_SUFFIX_f16))
+FUNCTION (vldrwq, vldrq_impl, (TYPE_SUFFIX_s32, TYPE_SUFFIX_u32, TYPE_SUFFIX_f32))
 FUNCTION_PRED_P_S (vmaxavq, VMAXAVQ)
 FUNCTION_WITHOUT_N_NO_U_F (vmaxaq, VMAXAQ)
 FUNCTION_ONLY_F (vmaxnmaq, VMAXNMAQ)
@@ -1109,6 +1206,9 @@ FUNCTION_ONLY_N_NO_F (vshrq, VSHRQ)
 FUNCTION_ONLY_N_NO_F (vsliq, VSLIQ)
 FUNCTION_ONLY_N_NO_F (vsriq, VSRIQ)
 FUNCTION (vst1q, vst1_impl,)
+FUNCTION (vstrbq, vstrq_impl, (QImode, opt_scalar_mode ()))
+FUNCTION (vstrhq, vstrq_impl, (HImode, HFmode))
+FUNCTION (vstrwq, vstrq_impl, (SImode, SFmode))
 FUNCTION_WITH_RTX_M_N (vsubq, MINUS, VSUBQ)
 FUNCTION (vuninitializedq, vuninitializedq_impl,)
 
diff --git a/gcc/config/arm/arm-mve-builtins-base.def b/gcc/config/arm/arm-mve-builtins-base.def
index b8a8cf2..6166f1b 100644
--- a/gcc/config/arm/arm-mve-builtins-base.def
+++ b/gcc/config/arm/arm-mve-builtins-base.def
@@ -58,7 +58,10 @@ DEF_MVE_FUNCTION (vhcaddq_rot90, binary, all_signed, mx_or_none)
 DEF_MVE_FUNCTION (vhsubq, binary_opt_n, all_integer, mx_or_none)
 DEF_MVE_FUNCTION (vidupq, viddup, all_unsigned, mx_or_none)
 DEF_MVE_FUNCTION (viwdupq, vidwdup, all_unsigned, mx_or_none)
-DEF_MVE_FUNCTION (vld1q, load, all_integer, none)
+DEF_MVE_FUNCTION (vld1q, load, all_integer, z_or_none)
+DEF_MVE_FUNCTION (vldrbq, load_ext, all_integer, z_or_none)
+DEF_MVE_FUNCTION (vldrhq, load_ext, integer_16_32, z_or_none)
+DEF_MVE_FUNCTION (vldrwq, load_ext, integer_32, z_or_none)
 DEF_MVE_FUNCTION (vmaxaq, binary_maxamina, all_signed, m_or_none)
 DEF_MVE_FUNCTION (vmaxavq, binary_maxavminav, all_signed, p_or_none)
 DEF_MVE_FUNCTION (vmaxq, binary, all_integer, mx_or_none)
@@ -166,7 +169,10 @@ DEF_MVE_FUNCTION (vshrntq, binary_rshift_narrow, integer_16_32, m_or_none)
 DEF_MVE_FUNCTION (vshrq, binary_rshift, all_integer, mx_or_none)
 DEF_MVE_FUNCTION (vsliq, ternary_lshift, all_integer, m_or_none)
 DEF_MVE_FUNCTION (vsriq, ternary_rshift, all_integer, m_or_none)
-DEF_MVE_FUNCTION (vst1q, store, all_integer, none)
+DEF_MVE_FUNCTION (vst1q, store, all_integer, p_or_none)
+DEF_MVE_FUNCTION (vstrbq, store, all_integer, p_or_none)
+DEF_MVE_FUNCTION (vstrhq, store, integer_16_32, p_or_none)
+DEF_MVE_FUNCTION (vstrwq, store, integer_32, p_or_none)
 DEF_MVE_FUNCTION (vsubq, binary_opt_n, all_integer, mx_or_none)
 DEF_MVE_FUNCTION (vuninitializedq, inherent, all_integer_with_64, none)
 #undef REQUIRES_FLOAT
@@ -209,7 +215,9 @@ DEF_MVE_FUNCTION (veorq, binary, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vfmaq, ternary_opt_n, all_float, m_or_none)
 DEF_MVE_FUNCTION (vfmasq, ternary_n, all_float, m_or_none)
 DEF_MVE_FUNCTION (vfmsq, ternary, all_float, m_or_none)
-DEF_MVE_FUNCTION (vld1q, load, all_float, none)
+DEF_MVE_FUNCTION (vld1q, load, all_float, z_or_none)
+DEF_MVE_FUNCTION (vldrhq, load_ext, float_16, z_or_none)
+DEF_MVE_FUNCTION (vldrwq, load_ext, float_32, z_or_none)
 DEF_MVE_FUNCTION (vmaxnmaq, binary, all_float, m_or_none)
 DEF_MVE_FUNCTION (vmaxnmavq, binary_maxvminv, all_float, p_or_none)
 DEF_MVE_FUNCTION (vmaxnmq, binary, all_float, mx_or_none)
@@ -224,7 +232,7 @@ DEF_MVE_FUNCTION (vornq, binary_orrq, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vorrq, binary_orrq, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vpselq, vpsel, all_float, none)
 DEF_MVE_FUNCTION (vreinterpretq, unary_convert, reinterpret_float, none)
-DEF_MVE_FUNCTION (vrev32q, unary, float16, mx_or_none)
+DEF_MVE_FUNCTION (vrev32q, unary, float_16, mx_or_none)
 DEF_MVE_FUNCTION (vrev64q, unary, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vrndaq, unary, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vrndmq, unary, all_float, mx_or_none)
@@ -232,7 +240,9 @@ DEF_MVE_FUNCTION (vrndnq, unary, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vrndpq, unary, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vrndq, unary, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vrndxq, unary, all_float, mx_or_none)
-DEF_MVE_FUNCTION (vst1q, store, all_float, none)
+DEF_MVE_FUNCTION (vst1q, store, all_float, p_or_none)
+DEF_MVE_FUNCTION (vstrhq, store, float_16, p_or_none)
+DEF_MVE_FUNCTION (vstrwq, store, float_32, p_or_none)
 DEF_MVE_FUNCTION (vsubq, binary_opt_n, all_float, mx_or_none)
 DEF_MVE_FUNCTION (vuninitializedq, inherent, all_float, none)
 #undef REQUIRES_FLOAT
diff --git a/gcc/config/arm/arm-mve-builtins-base.h b/gcc/config/arm/arm-mve-builtins-base.h
index da630d4..7c866d8 100644
--- a/gcc/config/arm/arm-mve-builtins-base.h
+++ b/gcc/config/arm/arm-mve-builtins-base.h
@@ -82,6 +82,9 @@ extern const function_base *const vhsubq;
 extern const function_base *const vidupq;
 extern const function_base *const viwdupq;
 extern const function_base *const vld1q;
+extern const function_base *const vldrbq;
+extern const function_base *const vldrhq;
+extern const function_base *const vldrwq;
 extern const function_base *const vmaxaq;
 extern const function_base *const vmaxavq;
 extern const function_base *const vmaxnmaq;
@@ -202,6 +205,9 @@ extern const function_base *const vshrq;
 extern const function_base *const vsliq;
 extern const function_base *const vsriq;
 extern const function_base *const vst1q;
+extern const function_base *const vstrbq;
+extern const function_base *const vstrhq;
+extern const function_base *const vstrwq;
 extern const function_base *const vsubq;
 extern const function_base *const vuninitializedq;
 
diff --git a/gcc/config/arm/arm-mve-builtins-functions.h b/gcc/config/arm/arm-mve-builtins-functions.h
index 48b9e79..0ade215 100644
--- a/gcc/config/arm/arm-mve-builtins-functions.h
+++ b/gcc/config/arm/arm-mve-builtins-functions.h
@@ -958,19 +958,6 @@ public:
   memory_vector_mode (const function_instance &fi) const override
   {
     machine_mode mode = fi.vector_mode (0);
-    /* Vectors of floating-point are managed in memory as vectors of
-       integers.  */
-    switch (mode)
-      {
-      case E_V4SFmode:
-	mode = E_V4SImode;
-	break;
-      case E_V8HFmode:
-	mode = E_V8HImode;
-	break;
-      default:
-	break;
-      }
 
     if (m_vectors_per_tuple != 1)
       mode = targetm.array_mode (mode, m_vectors_per_tuple).require ();
diff --git a/gcc/config/arm/arm-mve-builtins.cc b/gcc/config/arm/arm-mve-builtins.cc
index 2bde822..af19086 100644
--- a/gcc/config/arm/arm-mve-builtins.cc
+++ b/gcc/config/arm/arm-mve-builtins.cc
@@ -150,8 +150,10 @@ CONSTEXPR const type_suffix_info type_suffixes[NUM_TYPE_SUFFIXES + 1] = {
    class ("b", "f", etc.) and a numerical bit count.  */
 
 /* _f16.  */
-#define TYPES_float16(S, D) \
-  S (f16)
+#define TYPES_float_16(S, D) S (f16)
+
+/* _f32.  */
+#define TYPES_float_32(S, D) S (f32)
 
 /* _f16 _f32.  */
 #define TYPES_all_float(S, D) \
@@ -304,7 +306,8 @@ static const type_suffix_pair types_none[] = {
 
 DEF_MVE_TYPES_ARRAY (all_integer);
 DEF_MVE_TYPES_ARRAY (all_integer_with_64);
-DEF_MVE_TYPES_ARRAY (float16);
+DEF_MVE_TYPES_ARRAY (float_16);
+DEF_MVE_TYPES_ARRAY (float_32);
 DEF_MVE_TYPES_ARRAY (all_float);
 DEF_MVE_TYPES_ARRAY (all_signed);
 DEF_MVE_TYPES_ARRAY (all_unsigned);
@@ -343,6 +346,11 @@ static const predication_index preds_p_or_none[] = {
   PRED_p, PRED_none, NUM_PREDS
 };
 
+/* Used by functions that have the z predicated form, in addition to
+   an unpredicated form.  */
+static const predication_index preds_z_or_none[]
+  = {PRED_z, PRED_none, NUM_PREDS};
+
 /* A list of all MVE ACLE functions.  */
 static CONSTEXPR const function_group_info function_groups[] = {
 #define DEF_MVE_FUNCTION(NAME, SHAPE, TYPES, PREDS)			\
@@ -1661,6 +1669,7 @@ function_resolver::check_gp_argument (unsigned int nops,
 
 	case PRED_p:
 	case PRED_x:
+	case PRED_z:
 	  /* Add final predicate.  */
 	  nargs = nops + 1;
 	  break;
diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h
index 34f024b..8ffdbc7 100644
--- a/gcc/config/arm/arm_mve.h
+++ b/gcc/config/arm/arm_mve.h
@@ -43,10 +43,8 @@
 #ifndef __ARM_MVE_PRESERVE_USER_NAMESPACE
 #define vst4q(__addr, __value) __arm_vst4q(__addr, __value)
 #define vstrbq_scatter_offset(__base, __offset, __value) __arm_vstrbq_scatter_offset(__base, __offset, __value)
-#define vstrbq(__addr, __value) __arm_vstrbq(__addr, __value)
 #define vstrwq_scatter_base(__addr, __offset, __value) __arm_vstrwq_scatter_base(__addr, __offset, __value)
 #define vldrbq_gather_offset(__base, __offset) __arm_vldrbq_gather_offset(__base, __offset)
-#define vstrbq_p(__addr, __value, __p) __arm_vstrbq_p(__addr, __value, __p)
 #define vstrbq_scatter_offset_p(__base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p(__base, __offset, __value, __p)
 #define vstrwq_scatter_base_p(__addr, __offset, __value, __p) __arm_vstrwq_scatter_base_p(__addr, __offset, __value, __p)
 #define vldrbq_gather_offset_z(__base, __offset, __p) __arm_vldrbq_gather_offset_z(__base, __offset, __p)
@@ -66,10 +64,6 @@
 #define vstrhq_scatter_offset_p(__base, __offset, __value, __p) __arm_vstrhq_scatter_offset_p(__base, __offset, __value, __p)
 #define vstrhq_scatter_shifted_offset(__base, __offset, __value) __arm_vstrhq_scatter_shifted_offset(__base, __offset, __value)
 #define vstrhq_scatter_shifted_offset_p(__base, __offset, __value, __p) __arm_vstrhq_scatter_shifted_offset_p(__base, __offset, __value, __p)
-#define vstrhq(__addr, __value) __arm_vstrhq(__addr, __value)
-#define vstrhq_p(__addr, __value, __p) __arm_vstrhq_p(__addr, __value, __p)
-#define vstrwq(__addr, __value) __arm_vstrwq(__addr, __value)
-#define vstrwq_p(__addr, __value, __p) __arm_vstrwq_p(__addr, __value, __p)
 #define vstrdq_scatter_base_p(__addr, __offset, __value, __p) __arm_vstrdq_scatter_base_p(__addr, __offset, __value, __p)
 #define vstrdq_scatter_base(__addr, __offset, __value) __arm_vstrdq_scatter_base(__addr, __offset, __value)
 #define vstrdq_scatter_offset_p(__base, __offset, __value, __p) __arm_vstrdq_scatter_offset_p(__base, __offset, __value, __p)
@@ -85,9 +79,7 @@
 #define vstrdq_scatter_base_wb_p(__addr, __offset, __value, __p) __arm_vstrdq_scatter_base_wb_p(__addr, __offset, __value, __p)
 #define vstrwq_scatter_base_wb_p(__addr, __offset, __value, __p) __arm_vstrwq_scatter_base_wb_p(__addr, __offset, __value, __p)
 #define vstrwq_scatter_base_wb(__addr, __offset, __value) __arm_vstrwq_scatter_base_wb(__addr, __offset, __value)
-#define vst1q_p(__addr, __value, __p) __arm_vst1q_p(__addr, __value, __p)
 #define vst2q(__addr, __value) __arm_vst2q(__addr, __value)
-#define vld1q_z(__base, __p) __arm_vld1q_z(__base, __p)
 #define vld2q(__addr) __arm_vld2q(__addr)
 #define vld4q(__addr) __arm_vld4q(__addr)
 #define vsetq_lane(__a, __b, __idx) __arm_vsetq_lane(__a, __b, __idx)
@@ -103,40 +95,22 @@
 #define vst4q_f16( __addr, __value) __arm_vst4q_f16( __addr, __value)
 #define vst4q_f32( __addr, __value) __arm_vst4q_f32( __addr, __value)
 #define vpnot(__a) __arm_vpnot(__a)
-#define vstrbq_s8( __addr, __value) __arm_vstrbq_s8( __addr, __value)
-#define vstrbq_u8( __addr, __value) __arm_vstrbq_u8( __addr, __value)
-#define vstrbq_u16( __addr, __value) __arm_vstrbq_u16( __addr, __value)
 #define vstrbq_scatter_offset_s8( __base, __offset, __value) __arm_vstrbq_scatter_offset_s8( __base, __offset, __value)
 #define vstrbq_scatter_offset_u8( __base, __offset, __value) __arm_vstrbq_scatter_offset_u8( __base, __offset, __value)
 #define vstrbq_scatter_offset_u16( __base, __offset, __value) __arm_vstrbq_scatter_offset_u16( __base, __offset, __value)
-#define vstrbq_s16( __addr, __value) __arm_vstrbq_s16( __addr, __value)
-#define vstrbq_u32( __addr, __value) __arm_vstrbq_u32( __addr, __value)
 #define vstrbq_scatter_offset_s16( __base, __offset, __value) __arm_vstrbq_scatter_offset_s16( __base, __offset, __value)
 #define vstrbq_scatter_offset_u32( __base, __offset, __value) __arm_vstrbq_scatter_offset_u32( __base, __offset, __value)
-#define vstrbq_s32( __addr, __value) __arm_vstrbq_s32( __addr, __value)
 #define vstrbq_scatter_offset_s32( __base, __offset, __value) __arm_vstrbq_scatter_offset_s32( __base, __offset, __value)
 #define vstrwq_scatter_base_s32(__addr,  __offset, __value) __arm_vstrwq_scatter_base_s32(__addr,  __offset, __value)
 #define vstrwq_scatter_base_u32(__addr,  __offset, __value) __arm_vstrwq_scatter_base_u32(__addr,  __offset, __value)
 #define vldrbq_gather_offset_u8(__base, __offset) __arm_vldrbq_gather_offset_u8(__base, __offset)
 #define vldrbq_gather_offset_s8(__base, __offset) __arm_vldrbq_gather_offset_s8(__base, __offset)
-#define vldrbq_s8(__base) __arm_vldrbq_s8(__base)
-#define vldrbq_u8(__base) __arm_vldrbq_u8(__base)
 #define vldrbq_gather_offset_u16(__base, __offset) __arm_vldrbq_gather_offset_u16(__base, __offset)
 #define vldrbq_gather_offset_s16(__base, __offset) __arm_vldrbq_gather_offset_s16(__base, __offset)
-#define vldrbq_s16(__base) __arm_vldrbq_s16(__base)
-#define vldrbq_u16(__base) __arm_vldrbq_u16(__base)
 #define vldrbq_gather_offset_u32(__base, __offset) __arm_vldrbq_gather_offset_u32(__base, __offset)
 #define vldrbq_gather_offset_s32(__base, __offset) __arm_vldrbq_gather_offset_s32(__base, __offset)
-#define vldrbq_s32(__base) __arm_vldrbq_s32(__base)
-#define vldrbq_u32(__base) __arm_vldrbq_u32(__base)
 #define vldrwq_gather_base_s32(__addr,  __offset) __arm_vldrwq_gather_base_s32(__addr,  __offset)
 #define vldrwq_gather_base_u32(__addr,  __offset) __arm_vldrwq_gather_base_u32(__addr,  __offset)
-#define vstrbq_p_s8( __addr, __value, __p) __arm_vstrbq_p_s8( __addr, __value, __p)
-#define vstrbq_p_s32( __addr, __value, __p) __arm_vstrbq_p_s32( __addr, __value, __p)
-#define vstrbq_p_s16( __addr, __value, __p) __arm_vstrbq_p_s16( __addr, __value, __p)
-#define vstrbq_p_u8( __addr, __value, __p) __arm_vstrbq_p_u8( __addr, __value, __p)
-#define vstrbq_p_u32( __addr, __value, __p) __arm_vstrbq_p_u32( __addr, __value, __p)
-#define vstrbq_p_u16( __addr, __value, __p) __arm_vstrbq_p_u16( __addr, __value, __p)
 #define vstrbq_scatter_offset_p_s8( __base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p_s8( __base, __offset, __value, __p)
 #define vstrbq_scatter_offset_p_s32( __base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p_s32( __base, __offset, __value, __p)
 #define vstrbq_scatter_offset_p_s16( __base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p_s16( __base, __offset, __value, __p)
@@ -151,12 +125,6 @@
 #define vldrbq_gather_offset_z_u16(__base, __offset, __p) __arm_vldrbq_gather_offset_z_u16(__base, __offset, __p)
 #define vldrbq_gather_offset_z_u32(__base, __offset, __p) __arm_vldrbq_gather_offset_z_u32(__base, __offset, __p)
 #define vldrbq_gather_offset_z_s8(__base, __offset, __p) __arm_vldrbq_gather_offset_z_s8(__base, __offset, __p)
-#define vldrbq_z_s16(__base, __p) __arm_vldrbq_z_s16(__base, __p)
-#define vldrbq_z_u8(__base, __p) __arm_vldrbq_z_u8(__base, __p)
-#define vldrbq_z_s8(__base, __p) __arm_vldrbq_z_s8(__base, __p)
-#define vldrbq_z_s32(__base, __p) __arm_vldrbq_z_s32(__base, __p)
-#define vldrbq_z_u16(__base, __p) __arm_vldrbq_z_u16(__base, __p)
-#define vldrbq_z_u32(__base, __p) __arm_vldrbq_z_u32(__base, __p)
 #define vldrwq_gather_base_z_u32(__addr,  __offset, __p) __arm_vldrwq_gather_base_z_u32(__addr,  __offset, __p)
 #define vldrwq_gather_base_z_s32(__addr,  __offset, __p) __arm_vldrwq_gather_base_z_s32(__addr,  __offset, __p)
 #define vldrhq_gather_offset_s32(__base, __offset) __arm_vldrhq_gather_offset_s32(__base, __offset)
@@ -175,22 +143,6 @@
 #define vldrhq_gather_shifted_offset_z_s16(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_s16(__base, __offset, __p)
 #define vldrhq_gather_shifted_offset_z_u32(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_u32(__base, __offset, __p)
 #define vldrhq_gather_shifted_offset_z_u16(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_u16(__base, __offset, __p)
-#define vldrhq_s32(__base) __arm_vldrhq_s32(__base)
-#define vldrhq_s16(__base) __arm_vldrhq_s16(__base)
-#define vldrhq_u32(__base) __arm_vldrhq_u32(__base)
-#define vldrhq_u16(__base) __arm_vldrhq_u16(__base)
-#define vldrhq_z_s32(__base, __p) __arm_vldrhq_z_s32(__base, __p)
-#define vldrhq_z_s16(__base, __p) __arm_vldrhq_z_s16(__base, __p)
-#define vldrhq_z_u32(__base, __p) __arm_vldrhq_z_u32(__base, __p)
-#define vldrhq_z_u16(__base, __p) __arm_vldrhq_z_u16(__base, __p)
-#define vldrwq_s32(__base) __arm_vldrwq_s32(__base)
-#define vldrwq_u32(__base) __arm_vldrwq_u32(__base)
-#define vldrwq_z_s32(__base, __p) __arm_vldrwq_z_s32(__base, __p)
-#define vldrwq_z_u32(__base, __p) __arm_vldrwq_z_u32(__base, __p)
-#define vldrhq_f16(__base) __arm_vldrhq_f16(__base)
-#define vldrhq_z_f16(__base, __p) __arm_vldrhq_z_f16(__base, __p)
-#define vldrwq_f32(__base) __arm_vldrwq_f32(__base)
-#define vldrwq_z_f32(__base, __p) __arm_vldrwq_z_f32(__base, __p)
 #define vldrdq_gather_base_s64(__addr,  __offset) __arm_vldrdq_gather_base_s64(__addr,  __offset)
 #define vldrdq_gather_base_u64(__addr,  __offset) __arm_vldrdq_gather_base_u64(__addr,  __offset)
 #define vldrdq_gather_base_z_s64(__addr,  __offset, __p) __arm_vldrdq_gather_base_z_s64(__addr,  __offset, __p)
@@ -221,7 +173,6 @@
 #define vldrwq_gather_shifted_offset_z_f32(__base, __offset, __p) __arm_vldrwq_gather_shifted_offset_z_f32(__base, __offset, __p)
 #define vldrwq_gather_shifted_offset_z_s32(__base, __offset, __p) __arm_vldrwq_gather_shifted_offset_z_s32(__base, __offset, __p)
 #define vldrwq_gather_shifted_offset_z_u32(__base, __offset, __p) __arm_vldrwq_gather_shifted_offset_z_u32(__base, __offset, __p)
-#define vstrhq_f16(__addr, __value) __arm_vstrhq_f16(__addr, __value)
 #define vstrhq_scatter_offset_s32( __base, __offset, __value) __arm_vstrhq_scatter_offset_s32( __base, __offset, __value)
 #define vstrhq_scatter_offset_s16( __base, __offset, __value) __arm_vstrhq_scatter_offset_s16( __base, __offset, __value)
 #define vstrhq_scatter_offset_u32( __base, __offset, __value) __arm_vstrhq_scatter_offset_u32( __base, __offset, __value)
@@ -238,21 +189,6 @@
 #define vstrhq_scatter_shifted_offset_p_s16( __base, __offset, __value, __p) __arm_vstrhq_scatter_shifted_offset_p_s16( __base, __offset, __value, __p)
 #define vstrhq_scatter_shifted_offset_p_u32( __base, __offset, __value, __p) __arm_vstrhq_scatter_shifted_offset_p_u32( __base, __offset, __value, __p)
 #define vstrhq_scatter_shifted_offset_p_u16( __base, __offset, __value, __p) __arm_vstrhq_scatter_shifted_offset_p_u16( __base, __offset, __value, __p)
-#define vstrhq_s32(__addr, __value) __arm_vstrhq_s32(__addr, __value)
-#define vstrhq_s16(__addr, __value) __arm_vstrhq_s16(__addr, __value)
-#define vstrhq_u32(__addr, __value) __arm_vstrhq_u32(__addr, __value)
-#define vstrhq_u16(__addr, __value) __arm_vstrhq_u16(__addr, __value)
-#define vstrhq_p_f16(__addr, __value, __p) __arm_vstrhq_p_f16(__addr, __value, __p)
-#define vstrhq_p_s32(__addr, __value, __p) __arm_vstrhq_p_s32(__addr, __value, __p)
-#define vstrhq_p_s16(__addr, __value, __p) __arm_vstrhq_p_s16(__addr, __value, __p)
-#define vstrhq_p_u32(__addr, __value, __p) __arm_vstrhq_p_u32(__addr, __value, __p)
-#define vstrhq_p_u16(__addr, __value, __p) __arm_vstrhq_p_u16(__addr, __value, __p)
-#define vstrwq_f32(__addr, __value) __arm_vstrwq_f32(__addr, __value)
-#define vstrwq_s32(__addr, __value) __arm_vstrwq_s32(__addr, __value)
-#define vstrwq_u32(__addr, __value) __arm_vstrwq_u32(__addr, __value)
-#define vstrwq_p_f32(__addr, __value, __p) __arm_vstrwq_p_f32(__addr, __value, __p)
-#define vstrwq_p_s32(__addr, __value, __p) __arm_vstrwq_p_s32(__addr, __value, __p)
-#define vstrwq_p_u32(__addr, __value, __p) __arm_vstrwq_p_u32(__addr, __value, __p)
 #define vstrdq_scatter_base_p_s64(__addr, __offset, __value, __p) __arm_vstrdq_scatter_base_p_s64(__addr, __offset, __value, __p)
 #define vstrdq_scatter_base_p_u64(__addr, __offset, __value, __p) __arm_vstrdq_scatter_base_p_u64(__addr, __offset, __value, __p)
 #define vstrdq_scatter_base_s64(__addr, __offset, __value) __arm_vstrdq_scatter_base_s64(__addr, __offset, __value)
@@ -313,46 +249,30 @@
 #define vstrwq_scatter_base_wb_s32(__addr, __offset, __value) __arm_vstrwq_scatter_base_wb_s32(__addr, __offset, __value)
 #define vstrwq_scatter_base_wb_u32(__addr, __offset, __value) __arm_vstrwq_scatter_base_wb_u32(__addr, __offset, __value)
 #define vstrwq_scatter_base_wb_f32(__addr, __offset, __value) __arm_vstrwq_scatter_base_wb_f32(__addr, __offset, __value)
-#define vst1q_p_u8(__addr, __value, __p) __arm_vst1q_p_u8(__addr, __value, __p)
-#define vst1q_p_s8(__addr, __value, __p) __arm_vst1q_p_s8(__addr, __value, __p)
 #define vst2q_s8(__addr, __value) __arm_vst2q_s8(__addr, __value)
 #define vst2q_u8(__addr, __value) __arm_vst2q_u8(__addr, __value)
-#define vld1q_z_u8(__base, __p) __arm_vld1q_z_u8(__base, __p)
-#define vld1q_z_s8(__base, __p) __arm_vld1q_z_s8(__base, __p)
 #define vld2q_s8(__addr) __arm_vld2q_s8(__addr)
 #define vld2q_u8(__addr) __arm_vld2q_u8(__addr)
 #define vld4q_s8(__addr) __arm_vld4q_s8(__addr)
 #define vld4q_u8(__addr) __arm_vld4q_u8(__addr)
-#define vst1q_p_u16(__addr, __value, __p) __arm_vst1q_p_u16(__addr, __value, __p)
-#define vst1q_p_s16(__addr, __value, __p) __arm_vst1q_p_s16(__addr, __value, __p)
 #define vst2q_s16(__addr, __value) __arm_vst2q_s16(__addr, __value)
 #define vst2q_u16(__addr, __value) __arm_vst2q_u16(__addr, __value)
-#define vld1q_z_u16(__base, __p) __arm_vld1q_z_u16(__base, __p)
-#define vld1q_z_s16(__base, __p) __arm_vld1q_z_s16(__base, __p)
 #define vld2q_s16(__addr) __arm_vld2q_s16(__addr)
 #define vld2q_u16(__addr) __arm_vld2q_u16(__addr)
 #define vld4q_s16(__addr) __arm_vld4q_s16(__addr)
 #define vld4q_u16(__addr) __arm_vld4q_u16(__addr)
-#define vst1q_p_u32(__addr, __value, __p) __arm_vst1q_p_u32(__addr, __value, __p)
-#define vst1q_p_s32(__addr, __value, __p) __arm_vst1q_p_s32(__addr, __value, __p)
 #define vst2q_s32(__addr, __value) __arm_vst2q_s32(__addr, __value)
 #define vst2q_u32(__addr, __value) __arm_vst2q_u32(__addr, __value)
-#define vld1q_z_u32(__base, __p) __arm_vld1q_z_u32(__base, __p)
-#define vld1q_z_s32(__base, __p) __arm_vld1q_z_s32(__base, __p)
 #define vld2q_s32(__addr) __arm_vld2q_s32(__addr)
 #define vld2q_u32(__addr) __arm_vld2q_u32(__addr)
 #define vld4q_s32(__addr) __arm_vld4q_s32(__addr)
 #define vld4q_u32(__addr) __arm_vld4q_u32(__addr)
 #define vld4q_f16(__addr) __arm_vld4q_f16(__addr)
 #define vld2q_f16(__addr) __arm_vld2q_f16(__addr)
-#define vld1q_z_f16(__base, __p) __arm_vld1q_z_f16(__base, __p)
 #define vst2q_f16(__addr, __value) __arm_vst2q_f16(__addr, __value)
-#define vst1q_p_f16(__addr, __value, __p) __arm_vst1q_p_f16(__addr, __value, __p)
 #define vld4q_f32(__addr) __arm_vld4q_f32(__addr)
 #define vld2q_f32(__addr) __arm_vld2q_f32(__addr)
-#define vld1q_z_f32(__base, __p) __arm_vld1q_z_f32(__base, __p)
 #define vst2q_f32(__addr, __value) __arm_vst2q_f32(__addr, __value)
-#define vst1q_p_f32(__addr, __value, __p) __arm_vst1q_p_f32(__addr, __value, __p)
 #define vsetq_lane_f16(__a, __b,  __idx) __arm_vsetq_lane_f16(__a, __b,  __idx)
 #define vsetq_lane_f32(__a, __b,  __idx) __arm_vsetq_lane_f32(__a, __b,  __idx)
 #define vsetq_lane_s16(__a, __b,  __idx) __arm_vsetq_lane_s16(__a, __b,  __idx)
@@ -508,48 +428,6 @@ __arm_vstrbq_scatter_offset_u16 (uint8_t * __base, uint16x8_t __offset, uint16x8
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_s8 (int8_t * __addr, int8x16_t __value)
-{
-  __builtin_mve_vstrbq_sv16qi ((__builtin_neon_qi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_s32 (int8_t * __addr, int32x4_t __value)
-{
-  __builtin_mve_vstrbq_sv4si ((__builtin_neon_qi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_s16 (int8_t * __addr, int16x8_t __value)
-{
-  __builtin_mve_vstrbq_sv8hi ((__builtin_neon_qi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_u8 (uint8_t * __addr, uint8x16_t __value)
-{
-  __builtin_mve_vstrbq_uv16qi ((__builtin_neon_qi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_u32 (uint8_t * __addr, uint32x4_t __value)
-{
-  __builtin_mve_vstrbq_uv4si ((__builtin_neon_qi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_u16 (uint8_t * __addr, uint16x8_t __value)
-{
-  __builtin_mve_vstrbq_uv8hi ((__builtin_neon_qi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_base_s32 (uint32x4_t __addr, const int __offset, int32x4_t __value)
 {
   __builtin_mve_vstrwq_scatter_base_sv4si (__addr, __offset, __value);
@@ -576,20 +454,6 @@ __arm_vldrbq_gather_offset_s8 (int8_t const * __base, uint8x16_t __offset)
   return __builtin_mve_vldrbq_gather_offset_sv16qi ((__builtin_neon_qi *) __base, __offset);
 }
 
-__extension__ extern __inline int8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_s8 (int8_t const * __base)
-{
-  return __builtin_mve_vldrbq_sv16qi ((__builtin_neon_qi *) __base);
-}
-
-__extension__ extern __inline uint8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_u8 (uint8_t const * __base)
-{
-  return __builtin_mve_vldrbq_uv16qi ((__builtin_neon_qi *) __base);
-}
-
 __extension__ extern __inline uint16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vldrbq_gather_offset_u16 (uint8_t const * __base, uint16x8_t __offset)
@@ -604,20 +468,6 @@ __arm_vldrbq_gather_offset_s16 (int8_t const * __base, uint16x8_t __offset)
   return __builtin_mve_vldrbq_gather_offset_sv8hi ((__builtin_neon_qi *) __base, __offset);
 }
 
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_s16 (int8_t const * __base)
-{
-  return __builtin_mve_vldrbq_sv8hi ((__builtin_neon_qi *) __base);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_u16 (uint8_t const * __base)
-{
-  return __builtin_mve_vldrbq_uv8hi ((__builtin_neon_qi *) __base);
-}
-
 __extension__ extern __inline uint32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vldrbq_gather_offset_u32 (uint8_t const * __base, uint32x4_t __offset)
@@ -634,20 +484,6 @@ __arm_vldrbq_gather_offset_s32 (int8_t const * __base, uint32x4_t __offset)
 
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_s32 (int8_t const * __base)
-{
-  return __builtin_mve_vldrbq_sv4si ((__builtin_neon_qi *) __base);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_u32 (uint8_t const * __base)
-{
-  return __builtin_mve_vldrbq_uv4si ((__builtin_neon_qi *) __base);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vldrwq_gather_base_s32 (uint32x4_t __addr, const int __offset)
 {
   return __builtin_mve_vldrwq_gather_base_sv4si (__addr, __offset);
@@ -662,48 +498,6 @@ __arm_vldrwq_gather_base_u32 (uint32x4_t __addr, const int __offset)
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p_s8 (int8_t * __addr, int8x16_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_p_sv16qi ((__builtin_neon_qi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p_s32 (int8_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_p_sv4si ((__builtin_neon_qi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p_s16 (int8_t * __addr, int16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_p_sv8hi ((__builtin_neon_qi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p_u8 (uint8_t * __addr, uint8x16_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_p_uv16qi ((__builtin_neon_qi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p_u32 (uint8_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_p_uv4si ((__builtin_neon_qi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p_u16 (uint8_t * __addr, uint16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_p_uv8hi ((__builtin_neon_qi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrbq_scatter_offset_p_s8 (int8_t * __base, uint8x16_t __offset, int8x16_t __value, mve_pred16_t __p)
 {
   __builtin_mve_vstrbq_scatter_offset_p_sv16qi ((__builtin_neon_qi *) __base, __offset, __value, __p);
@@ -800,48 +594,6 @@ __arm_vldrbq_gather_offset_z_u16 (uint8_t const * __base, uint16x8_t __offset, m
   return __builtin_mve_vldrbq_gather_offset_z_uv8hi ((__builtin_neon_qi *) __base, __offset, __p);
 }
 
-__extension__ extern __inline int8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_z_s8 (int8_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrbq_z_sv16qi ((__builtin_neon_qi *) __base, __p);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_z_s32 (int8_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrbq_z_sv4si ((__builtin_neon_qi *) __base, __p);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_z_s16 (int8_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrbq_z_sv8hi ((__builtin_neon_qi *) __base, __p);
-}
-
-__extension__ extern __inline uint8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_z_u8 (uint8_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrbq_z_uv16qi ((__builtin_neon_qi *) __base, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_z_u32 (uint8_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrbq_z_uv4si ((__builtin_neon_qi *) __base, __p);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrbq_z_u16 (uint8_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrbq_z_uv8hi ((__builtin_neon_qi *) __base, __p);
-}
-
 __extension__ extern __inline int32x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vldrwq_gather_base_z_s32 (uint32x4_t __addr, const int __offset, mve_pred16_t __p)
@@ -968,91 +720,6 @@ __arm_vldrhq_gather_shifted_offset_z_u16 (uint16_t const * __base, uint16x8_t __
   return __builtin_mve_vldrhq_gather_shifted_offset_z_uv8hi ((__builtin_neon_hi *) __base, __offset, __p);
 }
 
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_s32 (int16_t const * __base)
-{
-  return __builtin_mve_vldrhq_sv4si ((__builtin_neon_hi *) __base);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_s16 (int16_t const * __base)
-{
-  return __builtin_mve_vldrhq_sv8hi ((__builtin_neon_hi *) __base);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_u32 (uint16_t const * __base)
-{
-  return __builtin_mve_vldrhq_uv4si ((__builtin_neon_hi *) __base);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_u16 (uint16_t const * __base)
-{
-  return __builtin_mve_vldrhq_uv8hi ((__builtin_neon_hi *) __base);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_z_s32 (int16_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrhq_z_sv4si ((__builtin_neon_hi *) __base, __p);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_z_s16 (int16_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrhq_z_sv8hi ((__builtin_neon_hi *) __base, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_z_u32 (uint16_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrhq_z_uv4si ((__builtin_neon_hi *) __base, __p);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_z_u16 (uint16_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrhq_z_uv8hi ((__builtin_neon_hi *) __base, __p);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_s32 (int32_t const * __base)
-{
-  return __builtin_mve_vldrwq_sv4si ((__builtin_neon_si *) __base);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_u32 (uint32_t const * __base)
-{
-  return __builtin_mve_vldrwq_uv4si ((__builtin_neon_si *) __base);
-}
-
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_z_s32 (int32_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrwq_z_sv4si ((__builtin_neon_si *) __base, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_z_u32 (uint32_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrwq_z_uv4si ((__builtin_neon_si *) __base, __p);
-}
-
 __extension__ extern __inline int64x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vldrdq_gather_base_s64 (uint64x2_t __addr, const int __offset)
@@ -1308,90 +975,6 @@ __arm_vstrhq_scatter_shifted_offset_p_u16 (uint16_t * __base, uint16x8_t __offse
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_s32 (int16_t * __addr, int32x4_t __value)
-{
-  __builtin_mve_vstrhq_sv4si ((__builtin_neon_hi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_s16 (int16_t * __addr, int16x8_t __value)
-{
-  __builtin_mve_vstrhq_sv8hi ((__builtin_neon_hi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_u32 (uint16_t * __addr, uint32x4_t __value)
-{
-  __builtin_mve_vstrhq_uv4si ((__builtin_neon_hi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_u16 (uint16_t * __addr, uint16x8_t __value)
-{
-  __builtin_mve_vstrhq_uv8hi ((__builtin_neon_hi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p_s32 (int16_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_p_sv4si ((__builtin_neon_hi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p_s16 (int16_t * __addr, int16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_p_sv8hi ((__builtin_neon_hi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p_u32 (uint16_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_p_uv4si ((__builtin_neon_hi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p_u16 (uint16_t * __addr, uint16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_p_uv8hi ((__builtin_neon_hi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_s32 (int32_t * __addr, int32x4_t __value)
-{
-  __builtin_mve_vstrwq_sv4si ((__builtin_neon_si *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_u32 (uint32_t * __addr, uint32x4_t __value)
-{
-  __builtin_mve_vstrwq_uv4si ((__builtin_neon_si *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_p_s32 (int32_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrwq_p_sv4si ((__builtin_neon_si *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_p_u32 (uint32_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrwq_p_uv4si ((__builtin_neon_si *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrdq_scatter_base_p_s64 (uint64x2_t __addr, const int __offset, int64x2_t __value, mve_pred16_t __p)
 {
   __builtin_mve_vstrdq_scatter_base_p_sv2di (__addr, __offset, __value, __p);
@@ -1668,20 +1251,6 @@ __arm_vstrwq_scatter_base_wb_u32 (uint32x4_t * __addr, const int __offset, uint3
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_u8 (uint8_t * __addr, uint8x16_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrbq_p_u8 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_s8 (int8_t * __addr, int8x16_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrbq_p_s8 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q_s8 (int8_t * __addr, int8x16x2_t __value)
 {
   union { int8x16x2_t __i; __builtin_neon_oi __o; } __rv;
@@ -1698,20 +1267,6 @@ __arm_vst2q_u8 (uint8_t * __addr, uint8x16x2_t __value)
   __builtin_mve_vst2qv16qi ((__builtin_neon_qi *) __addr, __rv.__o);
 }
 
-__extension__ extern __inline uint8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_u8 (uint8_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrbq_z_u8 ( __base, __p);
-}
-
-__extension__ extern __inline int8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_s8 (int8_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrbq_z_s8 ( __base, __p);
-}
-
 __extension__ extern __inline int8x16x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld2q_s8 (int8_t const * __addr)
@@ -1750,20 +1305,6 @@ __arm_vld4q_u8 (uint8_t const * __addr)
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_u16 (uint16_t * __addr, uint16x8_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrhq_p_u16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_s16 (int16_t * __addr, int16x8_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrhq_p_s16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q_s16 (int16_t * __addr, int16x8x2_t __value)
 {
   union { int16x8x2_t __i; __builtin_neon_oi __o; } __rv;
@@ -1780,20 +1321,6 @@ __arm_vst2q_u16 (uint16_t * __addr, uint16x8x2_t __value)
   __builtin_mve_vst2qv8hi ((__builtin_neon_hi *) __addr, __rv.__o);
 }
 
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_u16 (uint16_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrhq_z_u16 ( __base, __p);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_s16 (int16_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrhq_z_s16 ( __base, __p);
-}
-
 __extension__ extern __inline int16x8x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld2q_s16 (int16_t const * __addr)
@@ -1832,20 +1359,6 @@ __arm_vld4q_u16 (uint16_t const * __addr)
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_u32 (uint32_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrwq_p_u32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_s32 (int32_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrwq_p_s32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q_s32 (int32_t * __addr, int32x4x2_t __value)
 {
   union { int32x4x2_t __i; __builtin_neon_oi __o; } __rv;
@@ -1862,20 +1375,6 @@ __arm_vst2q_u32 (uint32_t * __addr, uint32x4x2_t __value)
   __builtin_mve_vst2qv4si ((__builtin_neon_si *) __addr, __rv.__o);
 }
 
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_u32 (uint32_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrwq_z_u32 ( __base, __p);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_s32 (int32_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrwq_z_s32 ( __base, __p);
-}
-
 __extension__ extern __inline int32x4x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld2q_s32 (int32_t const * __addr)
@@ -2180,34 +1679,6 @@ __arm_vst4q_f32 (float32_t * __addr, float32x4x4_t __value)
   __builtin_mve_vst4qv4sf (__addr, __rv.__o);
 }
 
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_f32 (float32_t const * __base)
-{
-  return __builtin_mve_vldrwq_fv4sf((__builtin_neon_si *) __base);
-}
-
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_z_f32 (float32_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrwq_z_fv4sf((__builtin_neon_si *) __base, __p);
-}
-
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_z_f16 (float16_t const * __base, mve_pred16_t __p)
-{
-  return __builtin_mve_vldrhq_z_fv8hf((__builtin_neon_hi *) __base, __p);
-}
-
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_f16 (float16_t const * __base)
-{
-  return __builtin_mve_vldrhq_fv8hf((__builtin_neon_hi *) __base);
-}
-
 __extension__ extern __inline float16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vldrhq_gather_offset_f16 (float16_t const * __base, uint16x8_t __offset)
@@ -2280,34 +1751,6 @@ __arm_vldrwq_gather_shifted_offset_z_f32 (float32_t const * __base, uint32x4_t _
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_p_f32 (float32_t * __addr, float32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrwq_p_fv4sf ((__builtin_neon_si *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_f32 (float32_t * __addr, float32x4_t __value)
-{
-  __builtin_mve_vstrwq_fv4sf ((__builtin_neon_si *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_f16 (float16_t * __addr, float16x8_t __value)
-{
-  __builtin_mve_vstrhq_fv8hf ((__builtin_neon_hi *) __addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p_f16 (float16_t * __addr, float16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_p_fv8hf ((__builtin_neon_hi *) __addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrhq_scatter_offset_f16 (float16_t * __base, uint16x8_t __offset, float16x8_t __value)
 {
   __builtin_mve_vstrhq_scatter_offset_fv8hf ((__builtin_neon_hi *) __base, __offset, __value);
@@ -2428,13 +1871,6 @@ __arm_vld2q_f16 (float16_t const * __addr)
   return __rv.__i;
 }
 
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_f16 (float16_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrhq_z_f16 (__base, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q_f16 (float16_t * __addr, float16x8x2_t __value)
@@ -2444,13 +1880,6 @@ __arm_vst2q_f16 (float16_t * __addr, float16x8x2_t __value)
   __builtin_mve_vst2qv8hf (__addr, __rv.__o);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_f16 (float16_t * __addr, float16x8_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrhq_p_f16 (__addr, __value, __p);
-}
-
 __extension__ extern __inline float32x4x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld4q_f32 (float32_t const * __addr)
@@ -2469,13 +1898,6 @@ __arm_vld2q_f32 (float32_t const * __addr)
   return __rv.__i;
 }
 
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z_f32 (float32_t const *__base, mve_pred16_t __p)
-{
-  return __arm_vldrwq_z_f32 (__base, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q_f32 (float32_t * __addr, float32x4x2_t __value)
@@ -2485,13 +1907,6 @@ __arm_vst2q_f32 (float32_t * __addr, float32x4x2_t __value)
   __builtin_mve_vst2qv4sf (__addr, __rv.__o);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p_f32 (float32_t * __addr, float32x4_t __value, mve_pred16_t __p)
-{
-  return __arm_vstrwq_p_f32 (__addr, __value, __p);
-}
-
 __extension__ extern __inline float16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vsetq_lane_f16 (float16_t __a, float16x8_t __b, const int __idx)
@@ -2614,48 +2029,6 @@ __arm_vstrbq_scatter_offset (uint8_t * __base, uint16x8_t __offset, uint16x8_t _
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq (int8_t * __addr, int8x16_t __value)
-{
- __arm_vstrbq_s8 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq (int8_t * __addr, int32x4_t __value)
-{
- __arm_vstrbq_s32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq (int8_t * __addr, int16x8_t __value)
-{
- __arm_vstrbq_s16 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq (uint8_t * __addr, uint8x16_t __value)
-{
- __arm_vstrbq_u8 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq (uint8_t * __addr, uint32x4_t __value)
-{
- __arm_vstrbq_u32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq (uint8_t * __addr, uint16x8_t __value)
-{
- __arm_vstrbq_u16 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_base (uint32x4_t __addr, const int __offset, int32x4_t __value)
 {
  __arm_vstrwq_scatter_base_s32 (__addr, __offset, __value);
@@ -2712,48 +2085,6 @@ __arm_vldrbq_gather_offset (int8_t const * __base, uint32x4_t __offset)
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p (int8_t * __addr, int8x16_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_p_s8 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p (int8_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_p_s32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p (int8_t * __addr, int16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_p_s16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p (uint8_t * __addr, uint8x16_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_p_u8 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p (uint8_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_p_u32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_p (uint8_t * __addr, uint16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_p_u16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrbq_scatter_offset_p (int8_t * __base, uint8x16_t __offset, int8x16_t __value, mve_pred16_t __p)
 {
  __arm_vstrbq_scatter_offset_p_s8 (__base, __offset, __value, __p);
@@ -3188,90 +2519,6 @@ __arm_vstrhq_scatter_shifted_offset_p (uint16_t * __base, uint16x8_t __offset, u
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq (int16_t * __addr, int32x4_t __value)
-{
- __arm_vstrhq_s32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq (int16_t * __addr, int16x8_t __value)
-{
- __arm_vstrhq_s16 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq (uint16_t * __addr, uint32x4_t __value)
-{
- __arm_vstrhq_u32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq (uint16_t * __addr, uint16x8_t __value)
-{
- __arm_vstrhq_u16 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p (int16_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_p_s32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p (int16_t * __addr, int16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_p_s16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p (uint16_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_p_u32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p (uint16_t * __addr, uint16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_p_u16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq (int32_t * __addr, int32x4_t __value)
-{
- __arm_vstrwq_s32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq (uint32_t * __addr, uint32x4_t __value)
-{
- __arm_vstrwq_u32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_p (int32_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrwq_p_s32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_p (uint32_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrwq_p_u32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrdq_scatter_base_p (uint64x2_t __addr, const int __offset, int64x2_t __value, mve_pred16_t __p)
 {
  __arm_vstrdq_scatter_base_p_s64 (__addr, __offset, __value, __p);
@@ -3468,20 +2715,6 @@ __arm_vstrwq_scatter_base_wb (uint32x4_t * __addr, const int __offset, uint32x4_
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (uint8_t * __addr, uint8x16_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_u8 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (int8_t * __addr, int8x16_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_s8 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q (int8_t * __addr, int8x16x2_t __value)
 {
  __arm_vst2q_s8 (__addr, __value);
@@ -3494,20 +2727,6 @@ __arm_vst2q (uint8_t * __addr, uint8x16x2_t __value)
  __arm_vst2q_u8 (__addr, __value);
 }
 
-__extension__ extern __inline uint8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (uint8_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_u8 (__base, __p);
-}
-
-__extension__ extern __inline int8x16_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (int8_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_s8 (__base, __p);
-}
-
 __extension__ extern __inline int8x16x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld2q (int8_t const * __addr)
@@ -3538,20 +2757,6 @@ __arm_vld4q (uint8_t const * __addr)
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (uint16_t * __addr, uint16x8_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_u16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (int16_t * __addr, int16x8_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_s16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q (int16_t * __addr, int16x8x2_t __value)
 {
  __arm_vst2q_s16 (__addr, __value);
@@ -3564,20 +2769,6 @@ __arm_vst2q (uint16_t * __addr, uint16x8x2_t __value)
  __arm_vst2q_u16 (__addr, __value);
 }
 
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (uint16_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_u16 (__base, __p);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (int16_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_s16 (__base, __p);
-}
-
 __extension__ extern __inline int16x8x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld2q (int16_t const * __addr)
@@ -3608,20 +2799,6 @@ __arm_vld4q (uint16_t const * __addr)
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (uint32_t * __addr, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_u32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (int32_t * __addr, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_s32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q (int32_t * __addr, int32x4x2_t __value)
 {
  __arm_vst2q_s32 (__addr, __value);
@@ -3634,20 +2811,6 @@ __arm_vst2q (uint32_t * __addr, uint32x4x2_t __value)
  __arm_vst2q_u32 (__addr, __value);
 }
 
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (uint32_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_u32 (__base, __p);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (int32_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_s32 (__base, __p);
-}
-
 __extension__ extern __inline int32x4x2_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld2q (int32_t const * __addr)
@@ -3862,34 +3025,6 @@ __arm_vldrwq_gather_shifted_offset_z (float32_t const * __base, uint32x4_t __off
 
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_p (float32_t * __addr, float32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrwq_p_f32 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq (float32_t * __addr, float32x4_t __value)
-{
- __arm_vstrwq_f32 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq (float16_t * __addr, float16x8_t __value)
-{
- __arm_vstrhq_f16 (__addr, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_p (float16_t * __addr, float16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_p_f16 (__addr, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrhq_scatter_offset (float16_t * __base, uint16x8_t __offset, float16x8_t __value)
 {
  __arm_vstrhq_scatter_offset_f16 (__base, __offset, __value);
@@ -3986,13 +3121,6 @@ __arm_vld2q (float16_t const * __addr)
  return __arm_vld2q_f16 (__addr);
 }
 
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (float16_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_f16 (__base, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q (float16_t * __addr, float16x8x2_t __value)
@@ -4000,13 +3128,6 @@ __arm_vst2q (float16_t * __addr, float16x8x2_t __value)
  __arm_vst2q_f16 (__addr, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (float16_t * __addr, float16x8_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_f16 (__addr, __value, __p);
-}
-
 __extension__ extern __inline float32x4x4_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vld4q (float32_t const * __addr)
@@ -4021,13 +3142,6 @@ __arm_vld2q (float32_t const * __addr)
  return __arm_vld2q_f32 (__addr);
 }
 
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vld1q_z (float32_t const *__base, mve_pred16_t __p)
-{
- return __arm_vld1q_z_f32 (__base, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vst2q (float32_t * __addr, float32x4x2_t __value)
@@ -4035,13 +3149,6 @@ __arm_vst2q (float32_t * __addr, float32x4x2_t __value)
  __arm_vst2q_f32 (__addr, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vst1q_p (float32_t * __addr, float32x4_t __value, mve_pred16_t __p)
-{
- __arm_vst1q_p_f32 (__addr, __value, __p);
-}
-
 __extension__ extern __inline float16x8_t
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vsetq_lane (float16_t __a, float16x8_t __b, const int __idx)
@@ -4393,17 +3500,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_float16x8x4_t]: __arm_vst4q_f16 (__ARM_mve_coerce_f16_ptr(__p0, float16_t *), __ARM_mve_coerce(__p1, float16x8x4_t)), \
   int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4x4_t]: __arm_vst4q_f32 (__ARM_mve_coerce_f32_ptr(__p0, float32_t *), __ARM_mve_coerce(__p1, float32x4x4_t)));})
 
-#define __arm_vld1q_z(p0,p1) ( \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr]: __arm_vld1q_z_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), p1), \
-  int (*)[__ARM_mve_type_int16_t_ptr]: __arm_vld1q_z_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), p1), \
-  int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vld1q_z_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), p1), \
-  int (*)[__ARM_mve_type_uint8_t_ptr]: __arm_vld1q_z_u8 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), p1), \
-  int (*)[__ARM_mve_type_uint16_t_ptr]: __arm_vld1q_z_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), p1), \
-  int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vld1q_z_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), p1), \
-  int (*)[__ARM_mve_type_float16_t_ptr]: __arm_vld1q_z_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), p1), \
-  int (*)[__ARM_mve_type_float32_t_ptr]: __arm_vld1q_z_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), p1)))
-
 #define __arm_vld2q(p0) ( \
   _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
   int (*)[__ARM_mve_type_int8_t_ptr]: __arm_vld2q_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *)), \
@@ -4482,17 +3578,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), p1, p2), \
   int (*)[__ARM_mve_type_float32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), p1, p2)))
 
-#define __arm_vst1q_p(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16_t]: __arm_vst1q_p_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vst1q_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vst1q_p_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint8x16_t]: __arm_vst1q_p_u8 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint8x16_t), p2), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vst1q_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vst1q_p_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2), \
-  int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_float16x8_t]: __arm_vst1q_p_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, float16x8_t), p2), \
-  int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4_t]: __arm_vst1q_p_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), __ARM_mve_coerce(__p1, float32x4_t), p2));})
-
 #define __arm_vst2q(p0,p1) ({ __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
   int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16x2_t]: __arm_vst2q_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16x2_t)), \
@@ -4504,22 +3589,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_float16x8x2_t]: __arm_vst2q_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, float16x8x2_t)), \
   int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4x2_t]: __arm_vst2q_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), __ARM_mve_coerce(__p1, float32x4x2_t)));})
 
-#define __arm_vstrhq(p0,p1) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vstrhq_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrhq_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t)), \
-  int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_float16x8_t]: __arm_vstrhq_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, float16x8_t)));})
-
-#define __arm_vstrhq_p(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vstrhq_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrhq_p_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2), \
-  int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_float16x8_t]: __arm_vstrhq_p_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, float16x8_t), p2));})
-
 #define __arm_vstrhq_scatter_offset_p(p0,p1,p2,p3) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)][__ARM_mve_typeid(__p2)])0, \
@@ -4556,18 +3625,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_shifted_offset_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t)), \
   int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_float16x8_t]: __arm_vstrhq_scatter_shifted_offset_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, float16x8_t)));})
 
-#define __arm_vstrwq_p(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_p_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_p_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2), \
-  int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4_t]: __arm_vstrwq_p_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), __ARM_mve_coerce(__p1, float32x4_t), p2));})
-
-#define __arm_vstrwq(p0,p1) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4_t)), \
-  int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4_t]: __arm_vstrwq_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), __ARM_mve_coerce(__p1, float32x4_t)));})
-
 #define __arm_vstrhq_scatter_offset(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)][__ARM_mve_typeid(__p2)])0, \
@@ -4798,15 +3855,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1, p2), \
   int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1, p2));})
 
-#define __arm_vst1q_p(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16_t]: __arm_vst1q_p_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vst1q_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vst1q_p_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint8x16_t]: __arm_vst1q_p_u8 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint8x16_t), p2), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vst1q_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vst1q_p_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
-
 #define __arm_vst2q(p0,p1) ({ __typeof(p1) __p1 = (p1); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
   int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16x2_t]: __arm_vst2q_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16x2_t)), \
@@ -4816,20 +3864,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8x2_t]: __arm_vst2q_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8x2_t)), \
   int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4x2_t]: __arm_vst2q_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4x2_t)));})
 
-#define __arm_vstrhq(p0,p1) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vstrhq_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrhq_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t)));})
-
-#define __arm_vstrhq_p(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vstrhq_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrhq_p_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
-
 #define __arm_vstrhq_scatter_offset_p(p0,p1,p2,p3) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)][__ARM_mve_typeid(__p2)])0, \
@@ -4862,17 +3896,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_shifted_offset_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t)), \
   int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_shifted_offset_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t)));})
 
-
-#define __arm_vstrwq(p0,p1) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4_t)));})
-
-#define __arm_vstrwq_p(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_p_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_p_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
-
 #define __arm_vstrdq_scatter_base_p(p0,p1,p2,p3) ({ __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(__p2)])0, \
   int (*)[__ARM_mve_type_int64x2_t]: __arm_vstrdq_scatter_base_p_s64 (p0, p1, __ARM_mve_coerce(__p2, int64x2_t), p3), \
@@ -4950,14 +3973,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint32x4_t]: __arm_vuninitializedq_u32 (), \
   int (*)[__ARM_mve_type_uint64x2_t]: __arm_vuninitializedq_u64 ());})
 
-#define __arm_vld1q_z(p0,p1) ( _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr]: __arm_vld1q_z_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), p1), \
-  int (*)[__ARM_mve_type_int16_t_ptr]: __arm_vld1q_z_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), p1), \
-  int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vld1q_z_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), p1), \
-  int (*)[__ARM_mve_type_uint8_t_ptr]: __arm_vld1q_z_u8 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), p1), \
-  int (*)[__ARM_mve_type_uint16_t_ptr]: __arm_vld1q_z_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), p1), \
-  int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vld1q_z_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), p1)))
-
 #define __arm_vld2q(p0) ( _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
   int (*)[__ARM_mve_type_int8_t_ptr]: __arm_vld2q_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *)), \
   int (*)[__ARM_mve_type_int16_t_ptr]: __arm_vld2q_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *)), \
@@ -5045,25 +4060,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrbq_gather_offset_u16(__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
   int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrbq_gather_offset_u32(__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint32x4_t)));})
 
-#define __arm_vstrbq(p0,p1) ({ __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16_t]: __arm_vstrbq_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16_t)), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vstrbq_s16 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int16x8_t)), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrbq_s32 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint8x16_t]: __arm_vstrbq_u8 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint8x16_t)), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vstrbq_u16 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrbq_u32 (__ARM_mve_coerce_u8_ptr(p0, uint8_t *), __ARM_mve_coerce(__p1, uint32x4_t)));})
-
-#define __arm_vstrbq_p(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p1) __p1 = (p1); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16_t]: __arm_vstrbq_p_s8 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, int8x16_t), p2), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int16x8_t]: __arm_vstrbq_p_s16 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, int16x8_t), p2), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrbq_p_s32 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, int32x4_t), p2), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint8x16_t]: __arm_vstrbq_p_u8 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint8x16_t), p2), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vstrbq_p_u16 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrbq_p_u32 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
-
 #define __arm_vstrdq_scatter_base(p0,p1,p2) ({ __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(__p2)])0, \
   int (*)[__ARM_mve_type_int64x2_t]: __arm_vstrdq_scatter_base_s64 (p0, p1, __ARM_mve_coerce(__p2, int64x2_t)), \
diff --git a/gcc/config/arm/arm_mve_builtins.def b/gcc/config/arm/arm_mve_builtins.def
index 0a9d321..5a0c760 100644
--- a/gcc/config/arm/arm_mve_builtins.def
+++ b/gcc/config/arm/arm_mve_builtins.def
@@ -663,20 +663,14 @@ VAR2 (QUADOP_NONE_NONE_NONE_NONE_PRED, vandq_m_f, v8hf, v4sf)
 VAR2 (QUADOP_NONE_NONE_NONE_NONE_PRED, vaddq_m_n_f, v8hf, v4sf)
 VAR2 (QUADOP_NONE_NONE_NONE_NONE_PRED, vaddq_m_f, v8hf, v4sf)
 VAR2 (QUADOP_NONE_NONE_NONE_NONE_PRED, vabdq_m_f, v8hf, v4sf)
-VAR3 (STRS, vstrbq_s, v16qi, v8hi, v4si)
-VAR3 (STRU, vstrbq_u, v16qi, v8hi, v4si)
 VAR3 (STRSS, vstrbq_scatter_offset_s, v16qi, v8hi, v4si)
 VAR3 (STRSU, vstrbq_scatter_offset_u, v16qi, v8hi, v4si)
 VAR1 (STRSBS, vstrwq_scatter_base_s, v4si)
 VAR1 (STRSBU, vstrwq_scatter_base_u, v4si)
 VAR3 (LDRGU, vldrbq_gather_offset_u, v16qi, v8hi, v4si)
 VAR3 (LDRGS, vldrbq_gather_offset_s, v16qi, v8hi, v4si)
-VAR3 (LDRS, vldrbq_s, v16qi, v8hi, v4si)
-VAR3 (LDRU, vldrbq_u, v16qi, v8hi, v4si)
 VAR1 (LDRGBS, vldrwq_gather_base_s, v4si)
 VAR1 (LDRGBU, vldrwq_gather_base_u, v4si)
-VAR3 (STRS_P, vstrbq_p_s, v16qi, v8hi, v4si)
-VAR3 (STRU_P, vstrbq_p_u, v16qi, v8hi, v4si)
 VAR3 (STRSS_P, vstrbq_scatter_offset_p_s, v16qi, v8hi, v4si)
 VAR3 (STRSU_P, vstrbq_scatter_offset_p_u, v16qi, v8hi, v4si)
 VAR1 (STRSBS_P, vstrwq_scatter_base_p_s, v4si)
@@ -685,15 +679,6 @@ VAR1 (LDRGBS_Z, vldrwq_gather_base_z_s, v4si)
 VAR1 (LDRGBU_Z, vldrwq_gather_base_z_u, v4si)
 VAR3 (LDRGS_Z, vldrbq_gather_offset_z_s, v16qi, v8hi, v4si)
 VAR3 (LDRGU_Z, vldrbq_gather_offset_z_u, v16qi, v8hi, v4si)
-VAR3 (LDRS_Z, vldrbq_z_s, v16qi, v8hi, v4si)
-VAR3 (LDRU_Z, vldrbq_z_u, v16qi, v8hi, v4si)
-VAR3 (LDRU, vld1q_u, v16qi, v8hi, v4si)
-VAR3 (LDRS, vld1q_s, v16qi, v8hi, v4si)
-VAR2 (LDRU_Z, vldrhq_z_u, v8hi, v4si)
-VAR2 (LDRU, vldrhq_u, v8hi, v4si)
-VAR2 (LDRS_Z, vldrhq_z_s, v8hi, v4si)
-VAR2 (LDRS, vldrhq_s, v8hi, v4si)
-VAR2 (LDRS, vld1q_f, v8hf, v4sf)
 VAR2 (LDRGU_Z, vldrhq_gather_shifted_offset_z_u, v8hi, v4si)
 VAR2 (LDRGU_Z, vldrhq_gather_offset_z_u, v8hi, v4si)
 VAR2 (LDRGU, vldrhq_gather_shifted_offset_u, v8hi, v4si)
@@ -702,14 +687,6 @@ VAR2 (LDRGS_Z, vldrhq_gather_shifted_offset_z_s, v8hi, v4si)
 VAR2 (LDRGS_Z, vldrhq_gather_offset_z_s, v8hi, v4si)
 VAR2 (LDRGS, vldrhq_gather_shifted_offset_s, v8hi, v4si)
 VAR2 (LDRGS, vldrhq_gather_offset_s, v8hi, v4si)
-VAR1 (LDRS, vldrhq_f, v8hf)
-VAR1 (LDRS_Z, vldrhq_z_f, v8hf)
-VAR1 (LDRS, vldrwq_f, v4sf)
-VAR1 (LDRS, vldrwq_s, v4si)
-VAR1 (LDRU, vldrwq_u, v4si)
-VAR1 (LDRS_Z, vldrwq_z_f, v4sf)
-VAR1 (LDRS_Z, vldrwq_z_s, v4si)
-VAR1 (LDRU_Z, vldrwq_z_u, v4si)
 VAR1 (LDRGBS, vldrdq_gather_base_s, v2di)
 VAR1 (LDRGBS, vldrwq_gather_base_f, v4sf)
 VAR1 (LDRGBS_Z, vldrdq_gather_base_z_s, v2di)
@@ -740,13 +717,6 @@ VAR1 (LDRGU_Z, vldrdq_gather_offset_z_u, v2di)
 VAR1 (LDRGU_Z, vldrdq_gather_shifted_offset_z_u, v2di)
 VAR1 (LDRGU_Z, vldrwq_gather_offset_z_u, v4si)
 VAR1 (LDRGU_Z, vldrwq_gather_shifted_offset_z_u, v4si)
-VAR3 (STRU, vst1q_u, v16qi, v8hi, v4si)
-VAR3 (STRS, vst1q_s, v16qi, v8hi, v4si)
-VAR2 (STRU_P, vstrhq_p_u, v8hi, v4si)
-VAR2 (STRU, vstrhq_u, v8hi, v4si)
-VAR2 (STRS_P, vstrhq_p_s, v8hi, v4si)
-VAR2 (STRS, vstrhq_s, v8hi, v4si)
-VAR2 (STRS, vst1q_f, v8hf, v4sf)
 VAR2 (STRSU_P, vstrhq_scatter_shifted_offset_p_u, v8hi, v4si)
 VAR2 (STRSU_P, vstrhq_scatter_offset_p_u, v8hi, v4si)
 VAR2 (STRSU, vstrhq_scatter_shifted_offset_u, v8hi, v4si)
@@ -755,14 +725,6 @@ VAR2 (STRSS_P, vstrhq_scatter_shifted_offset_p_s, v8hi, v4si)
 VAR2 (STRSS_P, vstrhq_scatter_offset_p_s, v8hi, v4si)
 VAR2 (STRSS, vstrhq_scatter_shifted_offset_s, v8hi, v4si)
 VAR2 (STRSS, vstrhq_scatter_offset_s, v8hi, v4si)
-VAR1 (STRS, vstrhq_f, v8hf)
-VAR1 (STRS_P, vstrhq_p_f, v8hf)
-VAR1 (STRS, vstrwq_f, v4sf)
-VAR1 (STRS, vstrwq_s, v4si)
-VAR1 (STRU, vstrwq_u, v4si)
-VAR1 (STRS_P, vstrwq_p_f, v4sf)
-VAR1 (STRS_P, vstrwq_p_s, v4si)
-VAR1 (STRU_P, vstrwq_p_u, v4si)
 VAR1 (STRSBS, vstrdq_scatter_base_s, v2di)
 VAR1 (STRSBS, vstrwq_scatter_base_f, v4sf)
 VAR1 (STRSBS_P, vstrdq_scatter_base_p_s, v2di)
diff --git a/gcc/config/arm/iterators.md b/gcc/config/arm/iterators.md
index 59e112b..22f8c18 100644
--- a/gcc/config/arm/iterators.md
+++ b/gcc/config/arm/iterators.md
@@ -283,6 +283,14 @@
 (define_mode_iterator MVE_V8HF [V8HF])
 (define_mode_iterator MVE_V16QI [V16QI])
 
+;; Types for MVE truncating stores and widening loads
+(define_mode_iterator MVE_w_narrow_TYPE [V8QI V4QI V4HI])
+(define_mode_attr MVE_w_narrow_type [(V8QI "v8qi") (V4QI "v4qi") (V4HI "v4hi")])
+(define_mode_attr MVE_wide_n_TYPE [(V8QI "V8HI") (V4QI "V4SI") (V4HI "V4SI")])
+(define_mode_attr MVE_wide_n_type [(V8QI "v8hi") (V4QI "v4si") (V4HI "v4si")])
+(define_mode_attr MVE_wide_n_sz_elem [(V8QI "16") (V4QI "32") (V4HI "32")])
+(define_mode_attr MVE_wide_n_VPRED [(V8QI "V8BI") (V4QI "V4BI") (V4HI "V4BI")])
+
 ;;----------------------------------------------------------------------------
 ;; Code iterators
 ;;----------------------------------------------------------------------------
@@ -1807,6 +1815,10 @@
 			     (V2SF "s") (V4SF  "s")
 			     (V2SF "s") (V4SF  "s")])
 
+(define_mode_attr MVE_elem_ch [(V4QI "b") (V8QI "b") (V16QI "b")
+			       (V4HI "h") (V8HI "h") (V8HF "h")
+			       (V4SI "w") (V4SF "w")])
+
 (define_mode_attr VH_elem_ch [(V4HI "s") (V8HI  "s")
 			      (V4HF "s") (V8HF  "s")
 			      (HF "s")])
@@ -2510,19 +2522,16 @@
 		       (VMLALDAVAXQ_P_S "s")
 		       (VMLALDAVAQ_P_S "s") (VMLALDAVAQ_P_U "u")
 		       (VSTRWQSB_S "s") (VSTRWQSB_U "u") (VSTRBQSO_S "s")
-		       (VSTRBQSO_U "u") (VSTRBQ_S "s") (VSTRBQ_U "u")
-		       (VLDRBQGO_S "s") (VLDRBQGO_U "u") (VLDRBQ_S "s")
-		       (VLDRBQ_U "u") (VLDRWQGB_S "s") (VLDRWQGB_U "u")
-		       (VLD1Q_S "s") (VLD1Q_U "u") (VLDRHQGO_S "s")
+		       (VSTRBQSO_U "u")
+		       (VLDRBQGO_S "s") (VLDRBQGO_U "u") (VLDRWQGB_S "s")
+		       (VLDRWQGB_U "u") (VLDRHQGO_S "s")
 		       (VLDRHQGO_U "u") (VLDRHQGSO_S "s") (VLDRHQGSO_U "u")
-		       (VLDRHQ_S "s") (VLDRHQ_U "u") (VLDRWQ_S "s")
-		       (VLDRWQ_U "u") (VLDRDQGB_S "s") (VLDRDQGB_U "u")
+		       (VLDRDQGB_S "s") (VLDRDQGB_U "u")
 		       (VLDRDQGO_S "s") (VLDRDQGO_U "u") (VLDRDQGSO_S "s")
 		       (VLDRDQGSO_U "u") (VLDRWQGO_S "s") (VLDRWQGO_U "u")
-		       (VLDRWQGSO_S "s") (VLDRWQGSO_U "u") (VST1Q_S "s")
-		       (VST1Q_U "u") (VSTRHQSO_S "s") (VSTRHQSO_U "u")
-		       (VSTRHQSSO_S "s") (VSTRHQSSO_U "u") (VSTRHQ_S "s")
-		       (VSTRHQ_U "u") (VSTRWQ_S "s") (VSTRWQ_U "u")
+		       (VLDRWQGSO_S "s") (VLDRWQGSO_U "u")
+		       (VSTRHQSO_S "s") (VSTRHQSO_U "u")
+		       (VSTRHQSSO_S "s") (VSTRHQSSO_U "u")
 		       (VSTRDQSB_S "s") (VSTRDQSB_U "u") (VSTRDQSO_S "s")
 		       (VSTRDQSO_U "u") (VSTRDQSSO_S "s") (VSTRDQSSO_U "u")
 		       (VSTRWQSO_U "u") (VSTRWQSO_S "s") (VSTRWQSSO_U "u")
@@ -2929,25 +2938,17 @@
 (define_int_iterator VSHRNTQ_M_N [VSHRNTQ_M_N_S VSHRNTQ_M_N_U])
 (define_int_iterator VSTRWSBQ [VSTRWQSB_S VSTRWQSB_U])
 (define_int_iterator VSTRBSOQ [VSTRBQSO_S VSTRBQSO_U])
-(define_int_iterator VSTRBQ [VSTRBQ_S VSTRBQ_U])
 (define_int_iterator VLDRBGOQ [VLDRBQGO_S VLDRBQGO_U])
-(define_int_iterator VLDRBQ [VLDRBQ_S VLDRBQ_U])
 (define_int_iterator VLDRWGBQ [VLDRWQGB_S VLDRWQGB_U])
-(define_int_iterator VLD1Q [VLD1Q_S VLD1Q_U])
 (define_int_iterator VLDRHGOQ [VLDRHQGO_S VLDRHQGO_U])
 (define_int_iterator VLDRHGSOQ [VLDRHQGSO_S VLDRHQGSO_U])
-(define_int_iterator VLDRHQ [VLDRHQ_S VLDRHQ_U])
-(define_int_iterator VLDRWQ [VLDRWQ_S VLDRWQ_U])
 (define_int_iterator VLDRDGBQ [VLDRDQGB_S VLDRDQGB_U])
 (define_int_iterator VLDRDGOQ [VLDRDQGO_S VLDRDQGO_U])
 (define_int_iterator VLDRDGSOQ [VLDRDQGSO_S VLDRDQGSO_U])
 (define_int_iterator VLDRWGOQ [VLDRWQGO_S VLDRWQGO_U])
 (define_int_iterator VLDRWGSOQ [VLDRWQGSO_S VLDRWQGSO_U])
-(define_int_iterator VST1Q [VST1Q_S VST1Q_U])
 (define_int_iterator VSTRHSOQ [VSTRHQSO_S VSTRHQSO_U])
 (define_int_iterator VSTRHSSOQ [VSTRHQSSO_S VSTRHQSSO_U])
-(define_int_iterator VSTRHQ [VSTRHQ_S VSTRHQ_U])
-(define_int_iterator VSTRWQ [VSTRWQ_S VSTRWQ_U])
 (define_int_iterator VSTRDSBQ [VSTRDQSB_S VSTRDQSB_U])
 (define_int_iterator VSTRDSOQ [VSTRDQSO_S VSTRDQSO_U])
 (define_int_iterator VSTRDSSOQ [VSTRDQSSO_S VSTRDQSSO_U])
diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md
index 8c69670..e54153e 100644
--- a/gcc/config/arm/mve.md
+++ b/gcc/config/arm/mve.md
@@ -3116,26 +3116,201 @@
   (set_attr "type" "mve_move")
    (set_attr "length""8")])
 
-;;
-;; [vstrbq_s vstrbq_u]
-;;
-(define_insn "mve_vstrbq_<supf><mode>"
-  [(set (match_operand:<MVE_B_ELEM> 0 "mve_memory_operand" "=Ux")
-	(unspec:<MVE_B_ELEM> [(match_operand:MVE_2 1 "s_register_operand" "w")]
-	 VSTRBQ))
+;; Vector stores
+;; [vstrbq_s8, vstrhq_s16, vstrwq_s32,
+;;  vstrbq_u8, vstrhq_u16, vstrwq_u32,
+;;  vst1q ]
+(define_insn "@mve_vstrq_<mode>"
+  [(set (match_operand:MVE_VLD_ST 0 "mve_memory_operand" "=Ux")
+	(unspec:MVE_VLD_ST
+	  [(match_operand:MVE_VLD_ST 1 "s_register_operand" "w")]
+	  VSTRQ))
+  ]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+   || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[1]);
+  ops[1] = gen_rtx_REG (TImode, regno);
+  ops[0]  = operands[0];
+  output_asm_insn ("vstr<MVE_elem_ch>.<V_sz_elem>\t%q1, %E0",ops);
+  return "";
+}
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrq_<mode>"))
+  (set_attr "length" "4")])
+
+;; Predicated vector stores
+;; [vstrbq_p_s8, vstrhq_p_s16, vstrwq_p_s32,
+;;  vstrbq_p_u8, vstrhq_p_u16, vstrwq_p_u32,
+;;  vst1q_p ]
+(define_insn "@mve_vstrq_p_<mode>"
+  [(set (match_operand:MVE_VLD_ST 0 "mve_memory_operand" "=Ux")
+	(unspec:MVE_VLD_ST [
+	   (match_operand:MVE_VLD_ST 1 "s_register_operand" "w")
+	   (match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")
+	   (match_dup 0)
+	] VSTRQ_P))
   ]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+   || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[1]);
+  ops[1] = gen_rtx_REG (TImode, regno);
+  ops[0]  = operands[0];
+  output_asm_insn ("vpst\;vstr<MVE_elem_ch>t.<V_sz_elem>\t%q1, %E0",ops);
+  return "";
+}
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrq_<mode>"))
+  (set_attr "type" "mve_move")
+  (set_attr "length" "8")])
+
+;; Truncating vector stores
+;; [vstrbq_s16, vstrbq_s32, vstrhq_s32,
+;;  vstrbq_u16, vstrbq_u32, vstrhq_u32]
+(define_insn "@mve_vstrq_truncate_<mode>"
+  [(set (match_operand:MVE_w_narrow_TYPE 0 "mve_memory_operand" "=Ux")
+	(unspec:MVE_w_narrow_TYPE
+	  [(truncate:MVE_w_narrow_TYPE
+	    (match_operand:<MVE_wide_n_TYPE> 1 "s_register_operand" "w"))]
+	  VSTRQ_TRUNC
+	))]
   "TARGET_HAVE_MVE"
 {
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn("vstrb.<V_sz_elem>\t%q1, %E0",ops);
-   return "";
+  rtx ops[2];
+  int regno = REGNO (operands[1]);
+  ops[1] = gen_rtx_REG (TImode, regno);
+  ops[0]  = operands[0];
+  output_asm_insn ("vstr<MVE_elem_ch>.<MVE_wide_n_sz_elem>\t%q1, %E0",ops);
+  return "";
+}
+  [(set (attr "mve_unpredicated_insn")
+	(symbol_ref "CODE_FOR_mve_vstrq_truncate_<mode>"))
+   (set_attr "length" "4")])
+
+;; Predicated truncating vector stores
+;; [vstrbq_p_s16, vstrbq_p_s32, vstrhq_p_s32,
+;;  vstrbq_p_u16, vstrbq_p_u32, vstrhq_p_u32]
+(define_insn "@mve_vstrq_p_truncate_<mode>"
+  [(set (match_operand:MVE_w_narrow_TYPE 0 "mve_memory_operand" "=Ux")
+	(unspec:MVE_w_narrow_TYPE [
+	  (truncate:MVE_w_narrow_TYPE
+	    (match_operand:<MVE_wide_n_TYPE> 1 "s_register_operand" "w"))
+	  (match_operand:<MVE_wide_n_VPRED> 2 "vpr_register_operand" "Up")
+	  (match_dup 0)
+	] VSTRQ_TRUNC_P))]
+  "TARGET_HAVE_MVE"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[1]);
+  ops[1] = gen_rtx_REG (TImode, regno);
+  ops[0]  = operands[0];
+  output_asm_insn (
+    "vpst\;vstr<MVE_elem_ch>t.<MVE_wide_n_sz_elem>\t%q1, %E0",
+    ops
+  );
+  return "";
 }
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrbq_<supf><mode>"))
+ [(set (attr "mve_unpredicated_insn")
+       (symbol_ref "CODE_FOR_mve_vstrq_truncate_<mode>"))
+  (set_attr "type" "mve_move")
+  (set_attr "length" "8")])
+
+;; Vector Loads
+;; [vldrbq_s8, vldrhq_s16, vldrwq_s32,
+;;  vldrbq_u8, vldrhq_u16, vldrwq_u32,
+;;  vld1q ]
+(define_insn "@mve_vldrq_<mode>"
+  [(set (match_operand:MVE_VLD_ST 0 "s_register_operand" "=w")
+	(unspec:MVE_VLD_ST
+	  [(match_operand:MVE_VLD_ST 1 "mve_memory_operand" "Ux")]
+	  VLDRQ))]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+   || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[0]);
+  ops[0] = gen_rtx_REG (TImode, regno);
+  ops[1]  = operands[1];
+  output_asm_insn ("vldr<MVE_elem_ch>.<V_sz_elem>\t%q0, %E1",ops);
+  return "";
+ }
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrq_<mode>"))
   (set_attr "length" "4")])
 
+;; Predicated vector loads
+;; [vldrbq_z_s8, vldrhq_z_s16, vldrwq_z_s32,
+;;  vldrbq_z_u8, vldrhq_z_u16, vldrwq_z_u32,
+;;  vld1q_z ]
+(define_insn "@mve_vldrq_z_<mode>"
+  [(set (match_operand:MVE_VLD_ST 0 "s_register_operand" "=w")
+	(unspec:MVE_VLD_ST [
+	   (match_operand:MVE_VLD_ST 1 "mve_memory_operand" "Ux")
+	   (match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")
+	] VLDRQ_Z))]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+   || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[0]);
+  ops[0] = gen_rtx_REG (TImode, regno);
+  ops[1]  = operands[1];
+  output_asm_insn ("vpst\;vldr<MVE_elem_ch>t.<V_sz_elem>\t%q0, %E1",ops);
+  return "";
+}
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrq_<mode>"))
+  (set_attr "type" "mve_move")
+  (set_attr "length" "8")])
+
+;; Extending vector loads
+;; [vldrbq_s16, vldrbq_s32, vldrhq_s32,
+;;  vldrbq_u16, vldrbq_u32, vldrhq_u32]
+(define_insn "@mve_vldrq_extend_<mode><US>"
+  [(set (match_operand:<MVE_wide_n_TYPE> 0 "s_register_operand" "=w")
+	(unspec:<MVE_wide_n_TYPE>
+	  [(SE:<MVE_wide_n_TYPE>
+	    (match_operand:MVE_w_narrow_TYPE 1 "mve_memory_operand" "Ux"))]
+	  VLDRQ_EXT))]
+  "TARGET_HAVE_MVE"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[0]);
+  ops[0] = gen_rtx_REG (TImode, regno);
+  ops[1]  = operands[1];
+  output_asm_insn ("vldr<MVE_elem_ch>.<US><MVE_wide_n_sz_elem>\t%q0, %E1",ops);
+  return "";
+}
+ [(set (attr "mve_unpredicated_insn")
+       (symbol_ref "CODE_FOR_mve_vldrq_extend_<mode><US>"))
+  (set_attr "length" "4")])
+
+;; Predicated extending vector loads
+;; [vldrbq_z_s16, vldrbq_z_s32, vldrhq_z_s32,
+;;  vldrbq_z_u16, vldrbq_z_u32, vldrhq_z_u32]
+(define_insn "@mve_vldrq_z_extend_<mode><US>"
+  [(set (match_operand:<MVE_wide_n_TYPE> 0 "s_register_operand" "=w")
+	  (unspec:<MVE_wide_n_TYPE> [
+	      (SE:<MVE_wide_n_TYPE>
+		(match_operand:MVE_w_narrow_TYPE 1 "mve_memory_operand" "Ux"))
+	      (match_operand:<MVE_wide_n_VPRED> 2 "vpr_register_operand" "Up")
+	  ] VLDRQ_EXT_Z))]
+  "TARGET_HAVE_MVE"
+{
+  rtx ops[2];
+  int regno = REGNO (operands[0]);
+  ops[0] = gen_rtx_REG (TImode, regno);
+  ops[1]  = operands[1];
+  output_asm_insn (
+    "vpst\;vldr<MVE_elem_ch>t.<US><MVE_wide_n_sz_elem>\t%q0, %E1",
+    ops
+  );
+  return "";
+}
+ [(set (attr "mve_unpredicated_insn")
+       (symbol_ref "CODE_FOR_mve_vldrq_extend_<mode><US>"))
+  (set_attr "type" "mve_move")
+  (set_attr "length" "8")])
+
 ;;
 ;; [vstrbq_scatter_offset_s vstrbq_scatter_offset_u]
 ;;
@@ -3213,29 +3388,6 @@
   (set_attr "length" "4")])
 
 ;;
-;; [vldrbq_s vldrbq_u]
-;;
-(define_insn "mve_vldrbq_<supf><mode>"
-  [(set (match_operand:MVE_2 0 "s_register_operand" "=w")
-	(unspec:MVE_2 [(match_operand:<MVE_B_ELEM> 1 "mve_memory_operand" "Ux")]
-	 VLDRBQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   if (<V_sz_elem> == 8)
-     output_asm_insn ("vldrb.<V_sz_elem>\t%q0, %E1",ops);
-   else
-     output_asm_insn ("vldrb.<supf><V_sz_elem>\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrbq_<supf><mode>"))
-  (set_attr "length" "4")])
-
-;;
 ;; [vldrwq_gather_base_s vldrwq_gather_base_u]
 ;;
 (define_insn "mve_vldrwq_gather_base_<supf>v4si"
@@ -3313,25 +3465,6 @@
  [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_scatter_base_<supf>v4si"))
   (set_attr "length" "8")])
 
-(define_insn "mve_vstrbq_p_<supf><mode>"
-  [(set (match_operand:<MVE_B_ELEM> 0 "mve_memory_operand" "=Ux")
-	(unspec:<MVE_B_ELEM>
-	 [(match_operand:MVE_2 1 "s_register_operand" "w")
-	  (match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")
-	  (match_dup 0)]
-	 VSTRBQ))]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vpst\;vstrbt.<V_sz_elem>\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrbq_<supf><mode>"))
-  (set_attr "length" "8")])
-
 ;;
 ;; [vldrbq_gather_offset_z_s vldrbq_gather_offset_z_u]
 ;;
@@ -3359,30 +3492,6 @@
   (set_attr "length" "8")])
 
 ;;
-;; [vldrbq_z_s vldrbq_z_u]
-;;
-(define_insn "mve_vldrbq_z_<supf><mode>"
-  [(set (match_operand:MVE_2 0 "s_register_operand" "=w")
-	(unspec:MVE_2 [(match_operand:<MVE_B_ELEM> 1 "mve_memory_operand" "Ux")
-		       (match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")]
-	 VLDRBQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   if (<V_sz_elem> == 8)
-     output_asm_insn ("vpst\;vldrbt.<V_sz_elem>\t%q0, %E1",ops);
-   else
-     output_asm_insn ("vpst\;vldrbt.<supf><V_sz_elem>\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrbq_<supf><mode>"))
-  (set_attr "length" "8")])
-
-;;
 ;; [vldrwq_gather_base_z_s vldrwq_gather_base_z_u]
 ;;
 (define_insn "mve_vldrwq_gather_base_z_<supf>v4si"
@@ -3405,26 +3514,6 @@
   (set_attr "length" "8")])
 
 ;;
-;; [vldrhq_f]
-;;
-(define_insn "mve_vldrhq_fv8hf"
-  [(set (match_operand:V8HF 0 "s_register_operand" "=w")
-	(unspec:V8HF [(match_operand:V8HI 1 "mve_memory_operand" "Ux")]
-	 VLDRHQ_F))
-  ]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   output_asm_insn ("vldrh.16\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_fv8hf"))
-  (set_attr "length" "4")])
-
-;;
 ;; [vldrhq_gather_offset_s vldrhq_gather_offset_u]
 ;;
 (define_insn "mve_vldrhq_gather_offset_<supf><mode>"
@@ -3525,176 +3614,6 @@
   (set_attr "length" "8")])
 
 ;;
-;; [vldrhq_s, vldrhq_u]
-;;
-(define_insn "mve_vldrhq_<supf><mode>"
-  [(set (match_operand:MVE_5 0 "s_register_operand" "=w")
-	(unspec:MVE_5 [(match_operand:<MVE_H_ELEM> 1 "mve_memory_operand" "Ux")]
-	 VLDRHQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   if (<V_sz_elem> == 16)
-     output_asm_insn ("vldrh.16\t%q0, %E1",ops);
-   else
-     output_asm_insn ("vldrh.<supf><V_sz_elem>\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_<supf><mode>"))
-  (set_attr "length" "4")])
-
-;;
-;; [vldrhq_z_f]
-;;
-(define_insn "mve_vldrhq_z_fv8hf"
-  [(set (match_operand:V8HF 0 "s_register_operand" "=w")
-	(unspec:V8HF [(match_operand:V8HI 1 "mve_memory_operand" "Ux")
-	(match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")]
-	 VLDRHQ_F))
-  ]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   output_asm_insn ("vpst\;vldrht.16\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_fv8hf"))
-  (set_attr "length" "8")])
-
-;;
-;; [vldrhq_z_s vldrhq_z_u]
-;;
-(define_insn "mve_vldrhq_z_<supf><mode>"
-  [(set (match_operand:MVE_5 0 "s_register_operand" "=w")
-	(unspec:MVE_5 [(match_operand:<MVE_H_ELEM> 1 "mve_memory_operand" "Ux")
-	(match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")]
-	 VLDRHQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   if (<V_sz_elem> == 16)
-     output_asm_insn ("vpst\;vldrht.16\t%q0, %E1",ops);
-   else
-     output_asm_insn ("vpst\;vldrht.<supf><V_sz_elem>\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_<supf><mode>"))
-  (set_attr "length" "8")])
-
-;;
-;; [vldrwq_f]
-;;
-(define_insn "mve_vldrwq_fv4sf"
-  [(set (match_operand:V4SF 0 "s_register_operand" "=w")
-	(unspec:V4SF [(match_operand:V4SI 1 "mve_memory_operand" "Ux")]
-	 VLDRWQ_F))
-  ]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   output_asm_insn ("vldrw.32\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_fv4sf"))
-  (set_attr "length" "4")])
-
-;;
-;; [vldrwq_s vldrwq_u]
-;;
-(define_insn "mve_vldrwq_<supf>v4si"
-  [(set (match_operand:V4SI 0 "s_register_operand" "=w")
-	(unspec:V4SI [(match_operand:V4SI 1 "mve_memory_operand" "Ux")]
-	 VLDRWQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   output_asm_insn ("vldrw.32\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_<supf>v4si"))
-  (set_attr "length" "4")])
-
-;;
-;; [vldrwq_z_f]
-;;
-(define_insn "mve_vldrwq_z_fv4sf"
-  [(set (match_operand:V4SF 0 "s_register_operand" "=w")
-	(unspec:V4SF [(match_operand:V4SI 1 "mve_memory_operand" "Ux")
-	(match_operand:V4BI 2 "vpr_register_operand" "Up")]
-	 VLDRWQ_F))
-  ]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   output_asm_insn ("vpst\;vldrwt.32\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_fv4sf"))
-  (set_attr "length" "8")])
-
-;;
-;; [vldrwq_z_s vldrwq_z_u]
-;;
-(define_insn "mve_vldrwq_z_<supf>v4si"
-  [(set (match_operand:V4SI 0 "s_register_operand" "=w")
-	(unspec:V4SI [(match_operand:V4SI 1 "mve_memory_operand" "Ux")
-	(match_operand:V4BI 2 "vpr_register_operand" "Up")]
-	 VLDRWQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[0]);
-   ops[0] = gen_rtx_REG (TImode, regno);
-   ops[1]  = operands[1];
-   output_asm_insn ("vpst\;vldrwt.32\t%q0, %E1",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_<supf>v4si"))
-  (set_attr "length" "8")])
-
-(define_expand "@mve_vld1q_f<mode>"
-  [(match_operand:MVE_0 0 "s_register_operand")
-   (unspec:MVE_0 [(match_operand:<MVE_CNVT> 1 "mve_memory_operand")] VLD1Q_F)
-  ]
-  "TARGET_HAVE_MVE || TARGET_HAVE_MVE_FLOAT"
-{
-  emit_insn (gen_mve_vldr<V_sz_elem1>q_f<mode>(operands[0],operands[1]));
-  DONE;
-})
-
-(define_expand "@mve_vld1q_<supf><mode>"
-  [(match_operand:MVE_2 0 "s_register_operand")
-   (unspec:MVE_2 [(match_operand:MVE_2 1 "mve_memory_operand")] VLD1Q)
-  ]
-  "TARGET_HAVE_MVE"
-{
-  emit_insn (gen_mve_vldr<V_sz_elem1>q_<supf><mode>(operands[0],operands[1]));
-  DONE;
-})
-
-;;
 ;; [vldrdq_gather_base_s vldrdq_gather_base_u]
 ;;
 (define_insn "mve_vldrdq_gather_base_<supf>v2di"
@@ -4131,71 +4050,6 @@
   (set_attr "length" "8")])
 
 ;;
-;; [vstrhq_f]
-;;
-(define_insn "mve_vstrhq_fv8hf"
-  [(set (match_operand:V8HI 0 "mve_memory_operand" "=Ux")
-	(unspec:V8HI [(match_operand:V8HF 1 "s_register_operand" "w")]
-	 VSTRHQ_F))
-  ]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vstrh.16\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_fv8hf"))
-  (set_attr "length" "4")])
-
-;;
-;; [vstrhq_p_f]
-;;
-(define_insn "mve_vstrhq_p_fv8hf"
-  [(set (match_operand:V8HI 0 "mve_memory_operand" "=Ux")
-	(unspec:V8HI
-	 [(match_operand:V8HF 1 "s_register_operand" "w")
-	  (match_operand:V8BI 2 "vpr_register_operand" "Up")
-	  (match_dup 0)]
-	 VSTRHQ_F))]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vpst\;vstrht.16\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_fv8hf"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrhq_p_s vstrhq_p_u]
-;;
-(define_insn "mve_vstrhq_p_<supf><mode>"
-  [(set (match_operand:<MVE_H_ELEM> 0 "mve_memory_operand" "=Ux")
-	(unspec:<MVE_H_ELEM>
-	 [(match_operand:MVE_5 1 "s_register_operand" "w")
-	  (match_operand:<MVE_VPRED> 2 "vpr_register_operand" "Up")
-	  (match_dup 0)]
-	 VSTRHQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vpst\;vstrht.<V_sz_elem>\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_<supf><mode>"))
-  (set_attr "length" "8")])
-
-;;
 ;; [vstrhq_scatter_offset_p_s vstrhq_scatter_offset_p_u]
 ;;
 (define_expand "mve_vstrhq_scatter_offset_p_<supf><mode>"
@@ -4321,130 +4175,6 @@
   (set_attr "length" "4")])
 
 ;;
-;; [vstrhq_s, vstrhq_u]
-;;
-(define_insn "mve_vstrhq_<supf><mode>"
-  [(set (match_operand:<MVE_H_ELEM> 0 "mve_memory_operand" "=Ux")
-	(unspec:<MVE_H_ELEM> [(match_operand:MVE_5 1 "s_register_operand" "w")]
-	 VSTRHQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vstrh.<V_sz_elem>\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_<supf><mode>"))
-  (set_attr "length" "4")])
-
-;;
-;; [vstrwq_f]
-;;
-(define_insn "mve_vstrwq_fv4sf"
-  [(set (match_operand:V4SI 0 "mve_memory_operand" "=Ux")
-	(unspec:V4SI [(match_operand:V4SF 1 "s_register_operand" "w")]
-	 VSTRWQ_F))
-  ]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vstrw.32\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_fv4sf"))
-  (set_attr "length" "4")])
-
-;;
-;; [vstrwq_p_f]
-;;
-(define_insn "mve_vstrwq_p_fv4sf"
-  [(set (match_operand:V4SI 0 "mve_memory_operand" "=Ux")
-	(unspec:V4SI
-	 [(match_operand:V4SF 1 "s_register_operand" "w")
-	  (match_operand:V4BI 2 "vpr_register_operand" "Up")
-	  (match_dup 0)]
-	 VSTRWQ_F))]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vpst\;vstrwt.32\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_fv4sf"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrwq_p_s vstrwq_p_u]
-;;
-(define_insn "mve_vstrwq_p_<supf>v4si"
-  [(set (match_operand:V4SI 0 "mve_memory_operand" "=Ux")
-	(unspec:V4SI
-	 [(match_operand:V4SI 1 "s_register_operand" "w")
-	  (match_operand:V4BI 2 "vpr_register_operand" "Up")
-	  (match_dup 0)]
-	 VSTRWQ))]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vpst\;vstrwt.32\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_<supf>v4si"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrwq_s vstrwq_u]
-;;
-(define_insn "mve_vstrwq_<supf>v4si"
-  [(set (match_operand:V4SI 0 "mve_memory_operand" "=Ux")
-	(unspec:V4SI [(match_operand:V4SI 1 "s_register_operand" "w")]
-	 VSTRWQ))
-  ]
-  "TARGET_HAVE_MVE"
-{
-   rtx ops[2];
-   int regno = REGNO (operands[1]);
-   ops[1] = gen_rtx_REG (TImode, regno);
-   ops[0]  = operands[0];
-   output_asm_insn ("vstrw.32\t%q1, %E0",ops);
-   return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_<supf>v4si"))
-  (set_attr "length" "4")])
-
-(define_expand "@mve_vst1q_f<mode>"
-  [(match_operand:<MVE_CNVT> 0 "mve_memory_operand")
-   (unspec:<MVE_CNVT> [(match_operand:MVE_0 1 "s_register_operand")] VST1Q_F)
-  ]
-  "TARGET_HAVE_MVE || TARGET_HAVE_MVE_FLOAT"
-{
-  emit_insn (gen_mve_vstr<V_sz_elem1>q_f<mode>(operands[0],operands[1]));
-  DONE;
-})
-
-(define_expand "@mve_vst1q_<supf><mode>"
-  [(match_operand:MVE_2 0 "mve_memory_operand")
-   (unspec:MVE_2 [(match_operand:MVE_2 1 "s_register_operand")] VST1Q)
-  ]
-  "TARGET_HAVE_MVE"
-{
-  emit_insn (gen_mve_vstr<V_sz_elem1>q_<supf><mode>(operands[0],operands[1]));
-  DONE;
-})
-
-;;
 ;; [vstrdq_scatter_base_p_s vstrdq_scatter_base_p_u]
 ;;
 (define_insn "mve_vstrdq_scatter_base_p_<supf>v2di"
diff --git a/gcc/config/arm/unspecs.md b/gcc/config/arm/unspecs.md
index f5f4d15..01963d5 100644
--- a/gcc/config/arm/unspecs.md
+++ b/gcc/config/arm/unspecs.md
@@ -1150,27 +1150,18 @@
   VSTRWQSB_U
   VSTRBQSO_S
   VSTRBQSO_U
-  VSTRBQ_S
-  VSTRBQ_U
+  VLDRQ
+  VLDRQ_Z
+  VLDRQ_EXT
+  VLDRQ_EXT_Z
   VLDRBQGO_S
   VLDRBQGO_U
-  VLDRBQ_S
-  VLDRBQ_U
   VLDRWQGB_S
   VLDRWQGB_U
-  VLD1Q_F
-  VLD1Q_S
-  VLD1Q_U
-  VLDRHQ_F
   VLDRHQGO_S
   VLDRHQGO_U
   VLDRHQGSO_S
   VLDRHQGSO_U
-  VLDRHQ_S
-  VLDRHQ_U
-  VLDRWQ_F
-  VLDRWQ_S
-  VLDRWQ_U
   VLDRDQGB_S
   VLDRDQGB_U
   VLDRDQGO_S
@@ -1186,15 +1177,11 @@
   VLDRWQGSO_F
   VLDRWQGSO_S
   VLDRWQGSO_U
-  VSTRHQ_F
-  VST1Q_S
-  VST1Q_U
+  VSTRQ
+  VSTRQ_P
+  VSTRQ_TRUNC
+  VSTRQ_TRUNC_P
   VSTRHQSO_S
-  VSTRHQ_U
-  VSTRWQ_S
-  VSTRWQ_U
-  VSTRWQ_F
-  VST1Q_F
   VSTRDQSB_S
   VSTRDQSB_U
   VSTRDQSO_S
diff --git a/gcc/testsuite/gcc.target/arm/pr112337.c b/gcc/testsuite/gcc.target/arm/pr112337.c
index 10b7881..599229c 100644
--- a/gcc/testsuite/gcc.target/arm/pr112337.c
+++ b/gcc/testsuite/gcc.target/arm/pr112337.c
@@ -4,7 +4,9 @@
 /* { dg-add-options arm_v8_1m_mve } */
 
 #pragma GCC arm "arm_mve_types.h"
-int32x4_t h(void *p) { return __builtin_mve_vldrwq_sv4si(p); }
+#pragma GCC arm "arm_mve.h" false
+
+int32x4_t h(void *p) { return vldrwq_s32(p); }
 void g(int32x4_t);
 void f(int, int, int, short, int *p) {
   int *bias = p;