arm: [MVE intrinsics] rework vstr?q_scatter_offset

This patch implements vstr?q_scatter_offset using the new MVE builtins
framework.

It uses a similar approach to a previous patch which grouped
truncating and non-truncating stores in two sets of patterns, rather
than having groups of patterns depending on the destination size.

We need to add the 'integer_64' types of suffixes in order to support
vstrdq_scatter_offset.

The patch introduces the MVE_VLD_ST_scatter iterator, similar to
MVE_VLD_ST but which also includes V2DI (again, for
vstrdq_scatter_offset).

The new MVE_scatter_offset mode attribute is used to map the
destination type to the offset type (both are usually equal, except
when the destination is floating-point).

We end up with four sets of patterns:
- vector scatter stores with offset (non-truncating)
- predicated vector scatter stores with offset (non-truncating)
- truncating vector scatter stores with offset
- predicated truncating vector scatter stores with offset

gcc/ChangeLog:

	* config/arm/arm-mve-builtins-base.cc (class vstrq_scatter_impl):
	New.
	(vstrbq_scatter, vstrhq_scatter, vstrwq_scatter, vstrdq_scatter):
	New.
	* config/arm/arm-mve-builtins-base.def (vstrbq_scatter)
	(vstrhq_scatter, vstrwq_scatter, vstrdq_scatter): New.
	* config/arm/arm-mve-builtins-base.h (vstrbq_scatter)
	(vstrhq_scatter, vstrwq_scatter, vstrdq_scatter): New.
	* config/arm/arm-mve-builtins.cc (integer_64): New.
	* config/arm/arm_mve.h (vstrbq_scatter_offset): Delete.
	(vstrbq_scatter_offset_p): Delete.
	(vstrhq_scatter_offset): Delete.
	(vstrhq_scatter_offset_p): Delete.
	(vstrdq_scatter_offset_p): Delete.
	(vstrdq_scatter_offset): Delete.
	(vstrwq_scatter_offset_p): Delete.
	(vstrwq_scatter_offset): Delete.
	(vstrbq_scatter_offset_s8): Delete.
	(vstrbq_scatter_offset_u8): Delete.
	(vstrbq_scatter_offset_u16): Delete.
	(vstrbq_scatter_offset_s16): Delete.
	(vstrbq_scatter_offset_u32): Delete.
	(vstrbq_scatter_offset_s32): Delete.
	(vstrbq_scatter_offset_p_s8): Delete.
	(vstrbq_scatter_offset_p_s32): Delete.
	(vstrbq_scatter_offset_p_s16): Delete.
	(vstrbq_scatter_offset_p_u8): Delete.
	(vstrbq_scatter_offset_p_u32): Delete.
	(vstrbq_scatter_offset_p_u16): Delete.
	(vstrhq_scatter_offset_s32): Delete.
	(vstrhq_scatter_offset_s16): Delete.
	(vstrhq_scatter_offset_u32): Delete.
	(vstrhq_scatter_offset_u16): Delete.
	(vstrhq_scatter_offset_p_s32): Delete.
	(vstrhq_scatter_offset_p_s16): Delete.
	(vstrhq_scatter_offset_p_u32): Delete.
	(vstrhq_scatter_offset_p_u16): Delete.
	(vstrdq_scatter_offset_p_s64): Delete.
	(vstrdq_scatter_offset_p_u64): Delete.
	(vstrdq_scatter_offset_s64): Delete.
	(vstrdq_scatter_offset_u64): Delete.
	(vstrhq_scatter_offset_f16): Delete.
	(vstrhq_scatter_offset_p_f16): Delete.
	(vstrwq_scatter_offset_f32): Delete.
	(vstrwq_scatter_offset_p_f32): Delete.
	(vstrwq_scatter_offset_p_s32): Delete.
	(vstrwq_scatter_offset_p_u32): Delete.
	(vstrwq_scatter_offset_s32): Delete.
	(vstrwq_scatter_offset_u32): Delete.
	(__arm_vstrbq_scatter_offset_s8): Delete.
	(__arm_vstrbq_scatter_offset_s32): Delete.
	(__arm_vstrbq_scatter_offset_s16): Delete.
	(__arm_vstrbq_scatter_offset_u8): Delete.
	(__arm_vstrbq_scatter_offset_u32): Delete.
	(__arm_vstrbq_scatter_offset_u16): Delete.
	(__arm_vstrbq_scatter_offset_p_s8): Delete.
	(__arm_vstrbq_scatter_offset_p_s32): Delete.
	(__arm_vstrbq_scatter_offset_p_s16): Delete.
	(__arm_vstrbq_scatter_offset_p_u8): Delete.
	(__arm_vstrbq_scatter_offset_p_u32): Delete.
	(__arm_vstrbq_scatter_offset_p_u16): Delete.
	(__arm_vstrhq_scatter_offset_s32): Delete.
	(__arm_vstrhq_scatter_offset_s16): Delete.
	(__arm_vstrhq_scatter_offset_u32): Delete.
	(__arm_vstrhq_scatter_offset_u16): Delete.
	(__arm_vstrhq_scatter_offset_p_s32): Delete.
	(__arm_vstrhq_scatter_offset_p_s16): Delete.
	(__arm_vstrhq_scatter_offset_p_u32): Delete.
	(__arm_vstrhq_scatter_offset_p_u16): Delete.
	(__arm_vstrdq_scatter_offset_p_s64): Delete.
	(__arm_vstrdq_scatter_offset_p_u64): Delete.
	(__arm_vstrdq_scatter_offset_s64): Delete.
	(__arm_vstrdq_scatter_offset_u64): Delete.
	(__arm_vstrwq_scatter_offset_p_s32): Delete.
	(__arm_vstrwq_scatter_offset_p_u32): Delete.
	(__arm_vstrwq_scatter_offset_s32): Delete.
	(__arm_vstrwq_scatter_offset_u32): Delete.
	(__arm_vstrhq_scatter_offset_f16): Delete.
	(__arm_vstrhq_scatter_offset_p_f16): Delete.
	(__arm_vstrwq_scatter_offset_f32): Delete.
	(__arm_vstrwq_scatter_offset_p_f32): Delete.
	(__arm_vstrbq_scatter_offset): Delete.
	(__arm_vstrbq_scatter_offset_p): Delete.
	(__arm_vstrhq_scatter_offset): Delete.
	(__arm_vstrhq_scatter_offset_p): Delete.
	(__arm_vstrdq_scatter_offset_p): Delete.
	(__arm_vstrdq_scatter_offset): Delete.
	(__arm_vstrwq_scatter_offset_p): Delete.
	(__arm_vstrwq_scatter_offset): Delete.
	* config/arm/arm_mve_builtins.def (vstrbq_scatter_offset_s)
	(vstrbq_scatter_offset_u, vstrbq_scatter_offset_p_s)
	(vstrbq_scatter_offset_p_u, vstrhq_scatter_offset_p_u)
	(vstrhq_scatter_offset_u, vstrhq_scatter_offset_p_s)
	(vstrhq_scatter_offset_s, vstrdq_scatter_offset_s)
	(vstrhq_scatter_offset_f, vstrwq_scatter_offset_f)
	(vstrwq_scatter_offset_s, vstrdq_scatter_offset_p_s)
	(vstrhq_scatter_offset_p_f, vstrwq_scatter_offset_p_f)
	(vstrwq_scatter_offset_p_s, vstrdq_scatter_offset_u)
	(vstrwq_scatter_offset_u, vstrdq_scatter_offset_p_u)
	(vstrwq_scatter_offset_p_u) Delete.
	* config/arm/iterators.md (MVE_VLD_ST_scatter): New.
	(MVE_scatter_offset): New.
	(MVE_elem_ch): Add entry for V2DI.
	(supf): Remove VSTRBQSO_S, VSTRBQSO_U, VSTRHQSO_S, VSTRHQSO_U,
	VSTRDQSO_S, VSTRDQSO_U, VSTRWQSO_U, VSTRWQSO_S.
	(VSTRBSOQ, VSTRHSOQ, VSTRDSOQ, VSTRWSOQ): Delete.
	* config/arm/mve.md (mve_vstrbq_scatter_offset_<supf><mode>):
	Delete.
	(mve_vstrbq_scatter_offset_<supf><mode>_insn): Delete.
	(mve_vstrbq_scatter_offset_p_<supf><mode>): Delete.
	(mve_vstrbq_scatter_offset_p_<supf><mode>_insn): Delete.
	(mve_vstrhq_scatter_offset_p_<supf><mode>): Delete.
	(mve_vstrhq_scatter_offset_p_<supf><mode>_insn): Delete.
	(mve_vstrhq_scatter_offset_<supf><mode>): Delete.
	(mve_vstrhq_scatter_offset_<supf><mode>_insn): Delete.
	(mve_vstrdq_scatter_offset_p_<supf>v2di): Delete.
	(mve_vstrdq_scatter_offset_p_<supf>v2di_insn): Delete.
	(mve_vstrdq_scatter_offset_<supf>v2di): Delete.
	(mve_vstrdq_scatter_offset_<supf>v2di_insn): Delete.
	(mve_vstrhq_scatter_offset_fv8hf): Delete.
	(mve_vstrhq_scatter_offset_fv8hf_insn): Delete.
	(mve_vstrhq_scatter_offset_p_fv8hf): Delete.
	(mve_vstrhq_scatter_offset_p_fv8hf_insn): Delete.
	(mve_vstrwq_scatter_offset_fv4sf): Delete.
	(mve_vstrwq_scatter_offset_fv4sf_insn): Delete.
	(mve_vstrwq_scatter_offset_p_fv4sf): Delete.
	(mve_vstrwq_scatter_offset_p_fv4sf_insn): Delete.
	(mve_vstrwq_scatter_offset_p_<supf>v4si): Delete.
	(mve_vstrwq_scatter_offset_p_<supf>v4si_insn): Delete.
	(mve_vstrwq_scatter_offset_<supf>v4si): Delete.
	(mve_vstrwq_scatter_offset_<supf>v4si_insn): Delete.
	(@mve_vstrq_scatter_offset_<mode>): New.
	(@mve_vstrq_scatter_offset_p_<mode>): New.
	(@mve_vstrq_truncate_scatter_offset_<mode>): New.
	(@mve_vstrq_truncate_scatter_offset_p_<mode>): New.
	* config/arm/unspecs.md (VSTRBQSO_S, VSTRBQSO_U, VSTRHQSO_S)
	(VSTRDQSO_S, VSTRDQSO_U, VSTRWQSO_S, VSTRWQSO_U, VSTRHQSO_F)
	(VSTRWQSO_F, VSTRHQSO_U): Delete.
	(VSTRQSO, VSTRQSO_P, VSTRQSO_TRUNC, VSTRQSO_TRUNC_P): New.

diff --git a/gcc/config/arm/arm-mve-builtins-base.cc b/gcc/config/arm/arm-mve-builtins-base.cc
index 2c8ff461c539f7b2ec725b8622c49244842edc52..855115c009fbb2b2cee548fb7fcf2f56d2f989e6 100644 (file)
--- a/gcc/config/arm/arm-mve-builtins-base.cc
+++ b/gcc/config/arm/arm-mve-builtins-base.cc
@@ -239,6 +239,40 @@ public:
   }
 };
 
+  /* Builds the vstrq_scatter*offset intrinsics.  */
+class vstrq_scatter_impl : public store_truncating
+{
+public:
+  using store_truncating::store_truncating;
+
+  rtx expand (function_expander &e) const override
+  {
+    insn_code icode;
+    machine_mode memory_mode = e.memory_vector_mode ();
+
+    switch (e.pred)
+      {
+      case PRED_none:
+       icode = (e.vector_mode (0) == memory_mode
+                /* Non-truncating store case.  */
+                ? code_for_mve_vstrq_scatter_offset (memory_mode)
+                /* Truncating store case.  */
+                : code_for_mve_vstrq_truncate_scatter_offset (memory_mode));
+       break;
+
+      case PRED_p:
+       icode = (e.vector_mode (0) == memory_mode
+                ? code_for_mve_vstrq_scatter_offset_p (memory_mode)
+                : code_for_mve_vstrq_truncate_scatter_offset_p (memory_mode));
+       break;
+
+      default:
+       gcc_unreachable ();
+      }
+    return e.use_exact_insn (icode);
+  }
+};
+
 /* Builds the vldrq* intrinsics.  */
 class vldrq_impl : public load_extending
 {
@@ -1207,8 +1241,12 @@ 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 (vstrbq_scatter, vstrq_scatter_impl, (QImode, opt_scalar_mode ()))
+FUNCTION (vstrdq_scatter, vstrq_scatter_impl, (DImode, opt_scalar_mode ()))
 FUNCTION (vstrhq, vstrq_impl, (HImode, HFmode))
+FUNCTION (vstrhq_scatter, vstrq_scatter_impl, (HImode, HFmode))
 FUNCTION (vstrwq, vstrq_impl, (SImode, SFmode))
+FUNCTION (vstrwq_scatter, vstrq_scatter_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 6166f1b38f48336bfe4cf08850bfd55512f7fe4d..30b576f01ed8066846859e6d81337b9df131e2f9 100644 (file)
--- a/gcc/config/arm/arm-mve-builtins-base.def
+++ b/gcc/config/arm/arm-mve-builtins-base.def
@@ -171,8 +171,12 @@ 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, p_or_none)
 DEF_MVE_FUNCTION (vstrbq, store, all_integer, p_or_none)
+DEF_MVE_FUNCTION (vstrbq_scatter, store_scatter_offset, all_integer, p_or_none)
 DEF_MVE_FUNCTION (vstrhq, store, integer_16_32, p_or_none)
+DEF_MVE_FUNCTION (vstrhq_scatter, store_scatter_offset, integer_16_32, p_or_none)
 DEF_MVE_FUNCTION (vstrwq, store, integer_32, p_or_none)
+DEF_MVE_FUNCTION (vstrwq_scatter, store_scatter_offset, integer_32, p_or_none)
+DEF_MVE_FUNCTION (vstrdq_scatter, store_scatter_offset, integer_64, 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
@@ -242,7 +246,9 @@ 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, p_or_none)
 DEF_MVE_FUNCTION (vstrhq, store, float_16, p_or_none)
+DEF_MVE_FUNCTION (vstrhq_scatter, store_scatter_offset, float_16, p_or_none)
 DEF_MVE_FUNCTION (vstrwq, store, float_32, p_or_none)
+DEF_MVE_FUNCTION (vstrwq_scatter, store_scatter_offset, 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 7c866d81c447bc61c6ba27b737405fc3328247e0..6ff3b149bc658da54f6b6dd33c7697bb38a21483 100644 (file)
--- a/gcc/config/arm/arm-mve-builtins-base.h
+++ b/gcc/config/arm/arm-mve-builtins-base.h
@@ -206,8 +206,12 @@ 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 vstrbq_scatter;
+extern const function_base *const vstrdq_scatter;
 extern const function_base *const vstrhq;
+extern const function_base *const vstrhq_scatter;
 extern const function_base *const vstrwq;
+extern const function_base *const vstrwq_scatter;
 extern const function_base *const vsubq;
 extern const function_base *const vuninitializedq;
 

diff --git a/gcc/config/arm/arm-mve-builtins.cc b/gcc/config/arm/arm-mve-builtins.cc
index 0d8def8c2a57e21b2cf1199e99f2192c21cb497f..8e6fff089c42e4c81f42dc93898d1e18af86a0ee 100644 (file)
--- a/gcc/config/arm/arm-mve-builtins.cc
+++ b/gcc/config/arm/arm-mve-builtins.cc
@@ -207,6 +207,10 @@ CONSTEXPR const type_suffix_info type_suffixes[NUM_TYPE_SUFFIXES + 1] = {
 #define TYPES_signed_32(S, D) \
   S (s32)
 
+/* s64 _u64.  */
+#define TYPES_integer_64(S, D) \
+  S (s64), S (u64)
+
 /* All the type combinations allowed by vcvtq.  */
 #define TYPES_cvt(S, D) \
   D (f16, s16), \
@@ -314,6 +318,7 @@ DEF_MVE_TYPES_ARRAY (integer_8);
 DEF_MVE_TYPES_ARRAY (integer_8_16);
 DEF_MVE_TYPES_ARRAY (integer_16_32);
 DEF_MVE_TYPES_ARRAY (integer_32);
+DEF_MVE_TYPES_ARRAY (integer_64);
 DEF_MVE_TYPES_ARRAY (poly_8_16);
 DEF_MVE_TYPES_ARRAY (signed_16_32);
 DEF_MVE_TYPES_ARRAY (signed_32);

diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h
index 21a2ae7353bf63449d7603875d76cbd6ab632c41..7946734651d7040a2bd91cf96dde500bd010a290 100644 (file)
--- a/gcc/config/arm/arm_mve.h
+++ b/gcc/config/arm/arm_mve.h
@@ -46,10 +46,8 @@
 
 #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 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_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)
 #define vldrhq_gather_offset(__base, __offset) __arm_vldrhq_gather_offset(__base, __offset)
@@ -64,18 +62,12 @@
 #define vldrwq_gather_offset_z(__base, __offset, __p) __arm_vldrwq_gather_offset_z(__base, __offset, __p)
 #define vldrwq_gather_shifted_offset(__base, __offset) __arm_vldrwq_gather_shifted_offset(__base, __offset)
 #define vldrwq_gather_shifted_offset_z(__base, __offset, __p) __arm_vldrwq_gather_shifted_offset_z(__base, __offset, __p)
-#define vstrhq_scatter_offset(__base, __offset, __value) __arm_vstrhq_scatter_offset(__base, __offset, __value)
-#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 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)
-#define vstrdq_scatter_offset(__base, __offset, __value) __arm_vstrdq_scatter_offset(__base, __offset, __value)
 #define vstrdq_scatter_shifted_offset_p(__base, __offset, __value, __p) __arm_vstrdq_scatter_shifted_offset_p(__base, __offset, __value, __p)
 #define vstrdq_scatter_shifted_offset(__base, __offset, __value) __arm_vstrdq_scatter_shifted_offset(__base, __offset, __value)
-#define vstrwq_scatter_offset_p(__base, __offset, __value, __p) __arm_vstrwq_scatter_offset_p(__base, __offset, __value, __p)
-#define vstrwq_scatter_offset(__base, __offset, __value) __arm_vstrwq_scatter_offset(__base, __offset, __value)
 #define vstrwq_scatter_shifted_offset_p(__base, __offset, __value, __p) __arm_vstrwq_scatter_shifted_offset_p(__base, __offset, __value, __p)
 #define vstrwq_scatter_shifted_offset(__base, __offset, __value) __arm_vstrwq_scatter_shifted_offset(__base, __offset, __value)
 #define vuninitializedq(__v) __arm_vuninitializedq(__v)
@@ -99,12 +91,6 @@
 #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_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_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_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)
@@ -115,12 +101,6 @@
 #define vldrbq_gather_offset_s32(__base, __offset) __arm_vldrbq_gather_offset_s32(__base, __offset)
 #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_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)
-#define vstrbq_scatter_offset_p_u8( __base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p_u8( __base, __offset, __value, __p)
-#define vstrbq_scatter_offset_p_u32( __base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p_u32( __base, __offset, __value, __p)
-#define vstrbq_scatter_offset_p_u16( __base, __offset, __value, __p) __arm_vstrbq_scatter_offset_p_u16( __base, __offset, __value, __p)
 #define vstrwq_scatter_base_p_s32(__addr,  __offset, __value, __p) __arm_vstrwq_scatter_base_p_s32(__addr,  __offset, __value, __p)
 #define vstrwq_scatter_base_p_u32(__addr,  __offset, __value, __p) __arm_vstrwq_scatter_base_p_u32(__addr,  __offset, __value, __p)
 #define vldrbq_gather_offset_z_s16(__base, __offset, __p) __arm_vldrbq_gather_offset_z_s16(__base, __offset, __p)
@@ -177,14 +157,6 @@
 #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_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)
-#define vstrhq_scatter_offset_u16( __base, __offset, __value) __arm_vstrhq_scatter_offset_u16( __base, __offset, __value)
-#define vstrhq_scatter_offset_p_s32( __base, __offset, __value, __p) __arm_vstrhq_scatter_offset_p_s32( __base, __offset, __value, __p)
-#define vstrhq_scatter_offset_p_s16( __base, __offset, __value, __p) __arm_vstrhq_scatter_offset_p_s16( __base, __offset, __value, __p)
-#define vstrhq_scatter_offset_p_u32( __base, __offset, __value, __p) __arm_vstrhq_scatter_offset_p_u32( __base, __offset, __value, __p)
-#define vstrhq_scatter_offset_p_u16( __base, __offset, __value, __p) __arm_vstrhq_scatter_offset_p_u16( __base, __offset, __value, __p)
 #define vstrhq_scatter_shifted_offset_s32( __base, __offset, __value) __arm_vstrhq_scatter_shifted_offset_s32( __base, __offset, __value)
 #define vstrhq_scatter_shifted_offset_s16( __base, __offset, __value) __arm_vstrhq_scatter_shifted_offset_s16( __base, __offset, __value)
 #define vstrhq_scatter_shifted_offset_u32( __base, __offset, __value) __arm_vstrhq_scatter_shifted_offset_u32( __base, __offset, __value)
@@ -197,26 +169,14 @@
 #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)
 #define vstrdq_scatter_base_u64(__addr, __offset, __value) __arm_vstrdq_scatter_base_u64(__addr, __offset, __value)
-#define vstrdq_scatter_offset_p_s64(__base, __offset, __value, __p) __arm_vstrdq_scatter_offset_p_s64(__base, __offset, __value, __p)
-#define vstrdq_scatter_offset_p_u64(__base, __offset, __value, __p) __arm_vstrdq_scatter_offset_p_u64(__base, __offset, __value, __p)
-#define vstrdq_scatter_offset_s64(__base, __offset, __value) __arm_vstrdq_scatter_offset_s64(__base, __offset, __value)
-#define vstrdq_scatter_offset_u64(__base, __offset, __value) __arm_vstrdq_scatter_offset_u64(__base, __offset, __value)
 #define vstrdq_scatter_shifted_offset_p_s64(__base, __offset, __value, __p) __arm_vstrdq_scatter_shifted_offset_p_s64(__base, __offset, __value, __p)
 #define vstrdq_scatter_shifted_offset_p_u64(__base, __offset, __value, __p) __arm_vstrdq_scatter_shifted_offset_p_u64(__base, __offset, __value, __p)
 #define vstrdq_scatter_shifted_offset_s64(__base, __offset, __value) __arm_vstrdq_scatter_shifted_offset_s64(__base, __offset, __value)
 #define vstrdq_scatter_shifted_offset_u64(__base, __offset, __value) __arm_vstrdq_scatter_shifted_offset_u64(__base, __offset, __value)
-#define vstrhq_scatter_offset_f16(__base, __offset, __value) __arm_vstrhq_scatter_offset_f16(__base, __offset, __value)
-#define vstrhq_scatter_offset_p_f16(__base, __offset, __value, __p) __arm_vstrhq_scatter_offset_p_f16(__base, __offset, __value, __p)
 #define vstrhq_scatter_shifted_offset_f16(__base, __offset, __value) __arm_vstrhq_scatter_shifted_offset_f16(__base, __offset, __value)
 #define vstrhq_scatter_shifted_offset_p_f16(__base, __offset, __value, __p) __arm_vstrhq_scatter_shifted_offset_p_f16(__base, __offset, __value, __p)
 #define vstrwq_scatter_base_f32(__addr, __offset, __value) __arm_vstrwq_scatter_base_f32(__addr, __offset, __value)
 #define vstrwq_scatter_base_p_f32(__addr, __offset, __value, __p) __arm_vstrwq_scatter_base_p_f32(__addr, __offset, __value, __p)
-#define vstrwq_scatter_offset_f32(__base, __offset, __value) __arm_vstrwq_scatter_offset_f32(__base, __offset, __value)
-#define vstrwq_scatter_offset_p_f32(__base, __offset, __value, __p) __arm_vstrwq_scatter_offset_p_f32(__base, __offset, __value, __p)
-#define vstrwq_scatter_offset_p_s32(__base, __offset, __value, __p) __arm_vstrwq_scatter_offset_p_s32(__base, __offset, __value, __p)
-#define vstrwq_scatter_offset_p_u32(__base, __offset, __value, __p) __arm_vstrwq_scatter_offset_p_u32(__base, __offset, __value, __p)
-#define vstrwq_scatter_offset_s32(__base, __offset, __value) __arm_vstrwq_scatter_offset_s32(__base, __offset, __value)
-#define vstrwq_scatter_offset_u32(__base, __offset, __value) __arm_vstrwq_scatter_offset_u32(__base, __offset, __value)
 #define vstrwq_scatter_shifted_offset_f32(__base, __offset, __value) __arm_vstrwq_scatter_shifted_offset_f32(__base, __offset, __value)
 #define vstrwq_scatter_shifted_offset_p_f32(__base, __offset, __value, __p) __arm_vstrwq_scatter_shifted_offset_p_f32(__base, __offset, __value, __p)
 #define vstrwq_scatter_shifted_offset_p_s32(__base, __offset, __value, __p) __arm_vstrwq_scatter_shifted_offset_p_s32(__base, __offset, __value, __p)
@@ -388,48 +348,6 @@ __arm_vpnot (mve_pred16_t __a)
   return __builtin_mve_vpnotv16bi (__a);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_s8 (int8_t * __base, uint8x16_t __offset, int8x16_t __value)
-{
-  __builtin_mve_vstrbq_scatter_offset_sv16qi ((__builtin_neon_qi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_s32 (int8_t * __base, uint32x4_t __offset, int32x4_t __value)
-{
-  __builtin_mve_vstrbq_scatter_offset_sv4si ((__builtin_neon_qi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_s16 (int8_t * __base, uint16x8_t __offset, int16x8_t __value)
-{
-  __builtin_mve_vstrbq_scatter_offset_sv8hi ((__builtin_neon_qi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_u8 (uint8_t * __base, uint8x16_t __offset, uint8x16_t __value)
-{
-  __builtin_mve_vstrbq_scatter_offset_uv16qi ((__builtin_neon_qi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_u32 (uint8_t * __base, uint32x4_t __offset, uint32x4_t __value)
-{
-  __builtin_mve_vstrbq_scatter_offset_uv4si ((__builtin_neon_qi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_u16 (uint8_t * __base, uint16x8_t __offset, uint16x8_t __value)
-{
-  __builtin_mve_vstrbq_scatter_offset_uv8hi ((__builtin_neon_qi *) __base, __offset, __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)
@@ -500,48 +418,6 @@ __arm_vldrwq_gather_base_u32 (uint32x4_t __addr, const int __offset)
   return __builtin_mve_vldrwq_gather_base_uv4si (__addr, __offset);
 }
 
-__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);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p_s32 (int8_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_scatter_offset_p_sv4si ((__builtin_neon_qi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p_s16 (int8_t * __base, uint16x8_t __offset, int16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_scatter_offset_p_sv8hi ((__builtin_neon_qi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p_u8 (uint8_t * __base, uint8x16_t __offset, uint8x16_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_scatter_offset_p_uv16qi ((__builtin_neon_qi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p_u32 (uint8_t * __base, uint32x4_t __offset, uint32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_scatter_offset_p_uv4si ((__builtin_neon_qi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p_u16 (uint8_t * __base, uint16x8_t __offset, uint16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrbq_scatter_offset_p_uv8hi ((__builtin_neon_qi *) __base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_base_p_s32 (uint32x4_t __addr, const int __offset, int32x4_t __value, mve_pred16_t __p)
@@ -865,62 +741,6 @@ __arm_vldrwq_gather_shifted_offset_z_u32 (uint32_t const * __base, uint32x4_t __
   return __builtin_mve_vldrwq_gather_shifted_offset_z_uv4si ((__builtin_neon_si *) __base, __offset, __p);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_s32 (int16_t * __base, uint32x4_t __offset, int32x4_t __value)
-{
-  __builtin_mve_vstrhq_scatter_offset_sv4si ((__builtin_neon_hi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_s16 (int16_t * __base, uint16x8_t __offset, int16x8_t __value)
-{
-  __builtin_mve_vstrhq_scatter_offset_sv8hi ((__builtin_neon_hi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_u32 (uint16_t * __base, uint32x4_t __offset, uint32x4_t __value)
-{
-  __builtin_mve_vstrhq_scatter_offset_uv4si ((__builtin_neon_hi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_u16 (uint16_t * __base, uint16x8_t __offset, uint16x8_t __value)
-{
-  __builtin_mve_vstrhq_scatter_offset_uv8hi ((__builtin_neon_hi *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p_s32 (int16_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_scatter_offset_p_sv4si ((__builtin_neon_hi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p_s16 (int16_t * __base, uint16x8_t __offset, int16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_scatter_offset_p_sv8hi ((__builtin_neon_hi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p_u32 (uint16_t * __base, uint32x4_t __offset, uint32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_scatter_offset_p_uv4si ((__builtin_neon_hi *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p_u16 (uint16_t * __base, uint16x8_t __offset, uint16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_scatter_offset_p_uv8hi ((__builtin_neon_hi *) __base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrhq_scatter_shifted_offset_s32 (int16_t * __base, uint32x4_t __offset, int32x4_t __value)
@@ -1005,34 +825,6 @@ __arm_vstrdq_scatter_base_u64 (uint64x2_t __addr, const int __offset, uint64x2_t
   __builtin_mve_vstrdq_scatter_base_uv2di (__addr, __offset, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset_p_s64 (int64_t * __base, uint64x2_t __offset, int64x2_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrdq_scatter_offset_p_sv2di ((__builtin_neon_di *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset_p_u64 (uint64_t * __base, uint64x2_t __offset, uint64x2_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrdq_scatter_offset_p_uv2di ((__builtin_neon_di *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset_s64 (int64_t * __base, uint64x2_t __offset, int64x2_t __value)
-{
-  __builtin_mve_vstrdq_scatter_offset_sv2di ((__builtin_neon_di *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset_u64 (uint64_t * __base, uint64x2_t __offset, uint64x2_t __value)
-{
-  __builtin_mve_vstrdq_scatter_offset_uv2di ((__builtin_neon_di *) __base, __offset, __value);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrdq_scatter_shifted_offset_p_s64 (int64_t * __base, uint64x2_t __offset, int64x2_t __value, mve_pred16_t __p)
@@ -1061,34 +853,6 @@ __arm_vstrdq_scatter_shifted_offset_u64 (uint64_t * __base, uint64x2_t __offset,
   __builtin_mve_vstrdq_scatter_shifted_offset_uv2di ((__builtin_neon_di *) __base, __offset, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_p_s32 (int32_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrwq_scatter_offset_p_sv4si ((__builtin_neon_si *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_p_u32 (uint32_t * __base, uint32x4_t __offset, uint32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrwq_scatter_offset_p_uv4si ((__builtin_neon_si *) __base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_s32 (int32_t * __base, uint32x4_t __offset, int32x4_t __value)
-{
-  __builtin_mve_vstrwq_scatter_offset_sv4si ((__builtin_neon_si *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_u32 (uint32_t * __base, uint32x4_t __offset, uint32x4_t __value)
-{
-  __builtin_mve_vstrwq_scatter_offset_uv4si ((__builtin_neon_si *) __base, __offset, __value);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_shifted_offset_p_s32 (int32_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
@@ -1753,20 +1517,6 @@ __arm_vldrwq_gather_shifted_offset_z_f32 (float32_t const * __base, uint32x4_t _
   return __builtin_mve_vldrwq_gather_shifted_offset_z_fv4sf ((__builtin_neon_si *) __base, __offset, __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);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p_f16 (float16_t * __base, uint16x8_t __offset, float16x8_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrhq_scatter_offset_p_fv8hf ((__builtin_neon_hi *) __base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrhq_scatter_shifted_offset_f16 (float16_t * __base, uint16x8_t __offset, float16x8_t __value)
@@ -1795,20 +1545,6 @@ __arm_vstrwq_scatter_base_p_f32 (uint32x4_t __addr, const int __offset, float32x
   __builtin_mve_vstrwq_scatter_base_p_fv4sf (__addr, __offset, __value, __p);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_f32 (float32_t * __base, uint32x4_t __offset, float32x4_t __value)
-{
-  __builtin_mve_vstrwq_scatter_offset_fv4sf ((__builtin_neon_si *) __base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_p_f32 (float32_t * __base, uint32x4_t __offset, float32x4_t __value, mve_pred16_t __p)
-{
-  __builtin_mve_vstrwq_scatter_offset_p_fv4sf ((__builtin_neon_si *) __base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_shifted_offset_f32 (float32_t * __base, uint32x4_t __offset, float32x4_t __value)
@@ -1989,48 +1725,6 @@ __arm_vst4q (uint32_t * __addr, uint32x4x4_t __value)
  __arm_vst4q_u32 (__addr, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset (int8_t * __base, uint8x16_t __offset, int8x16_t __value)
-{
- __arm_vstrbq_scatter_offset_s8 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset (int8_t * __base, uint32x4_t __offset, int32x4_t __value)
-{
- __arm_vstrbq_scatter_offset_s32 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset (int8_t * __base, uint16x8_t __offset, int16x8_t __value)
-{
- __arm_vstrbq_scatter_offset_s16 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset (uint8_t * __base, uint8x16_t __offset, uint8x16_t __value)
-{
- __arm_vstrbq_scatter_offset_u8 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset (uint8_t * __base, uint32x4_t __offset, uint32x4_t __value)
-{
- __arm_vstrbq_scatter_offset_u32 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset (uint8_t * __base, uint16x8_t __offset, uint16x8_t __value)
-{
- __arm_vstrbq_scatter_offset_u16 (__base, __offset, __value);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_base (uint32x4_t __addr, const int __offset, int32x4_t __value)
@@ -2087,48 +1781,6 @@ __arm_vldrbq_gather_offset (int8_t const * __base, uint32x4_t __offset)
  return __arm_vldrbq_gather_offset_s32 (__base, __offset);
 }
 
-__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);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p (int8_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_scatter_offset_p_s32 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p (int8_t * __base, uint16x8_t __offset, int16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_scatter_offset_p_s16 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p (uint8_t * __base, uint8x16_t __offset, uint8x16_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_scatter_offset_p_u8 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p (uint8_t * __base, uint32x4_t __offset, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_scatter_offset_p_u32 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrbq_scatter_offset_p (uint8_t * __base, uint16x8_t __offset, uint16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrbq_scatter_offset_p_u16 (__base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_base_p (uint32x4_t __addr, const int __offset, int32x4_t __value, mve_pred16_t __p)
@@ -2409,62 +2061,6 @@ __arm_vldrwq_gather_shifted_offset_z (uint32_t const * __base, uint32x4_t __offs
  return __arm_vldrwq_gather_shifted_offset_z_u32 (__base, __offset, __p);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset (int16_t * __base, uint32x4_t __offset, int32x4_t __value)
-{
- __arm_vstrhq_scatter_offset_s32 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset (int16_t * __base, uint16x8_t __offset, int16x8_t __value)
-{
- __arm_vstrhq_scatter_offset_s16 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset (uint16_t * __base, uint32x4_t __offset, uint32x4_t __value)
-{
- __arm_vstrhq_scatter_offset_u32 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset (uint16_t * __base, uint16x8_t __offset, uint16x8_t __value)
-{
- __arm_vstrhq_scatter_offset_u16 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p (int16_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_scatter_offset_p_s32 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p (int16_t * __base, uint16x8_t __offset, int16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_scatter_offset_p_s16 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p (uint16_t * __base, uint32x4_t __offset, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_scatter_offset_p_u32 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p (uint16_t * __base, uint16x8_t __offset, uint16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_scatter_offset_p_u16 (__base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrhq_scatter_shifted_offset (int16_t * __base, uint32x4_t __offset, int32x4_t __value)
@@ -2549,34 +2145,6 @@ __arm_vstrdq_scatter_base (uint64x2_t __addr, const int __offset, uint64x2_t __v
  __arm_vstrdq_scatter_base_u64 (__addr, __offset, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset_p (int64_t * __base, uint64x2_t __offset, int64x2_t __value, mve_pred16_t __p)
-{
- __arm_vstrdq_scatter_offset_p_s64 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset_p (uint64_t * __base, uint64x2_t __offset, uint64x2_t __value, mve_pred16_t __p)
-{
- __arm_vstrdq_scatter_offset_p_u64 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset (int64_t * __base, uint64x2_t __offset, int64x2_t __value)
-{
- __arm_vstrdq_scatter_offset_s64 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrdq_scatter_offset (uint64_t * __base, uint64x2_t __offset, uint64x2_t __value)
-{
- __arm_vstrdq_scatter_offset_u64 (__base, __offset, __value);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrdq_scatter_shifted_offset_p (int64_t * __base, uint64x2_t __offset, int64x2_t __value, mve_pred16_t __p)
@@ -2605,34 +2173,6 @@ __arm_vstrdq_scatter_shifted_offset (uint64_t * __base, uint64x2_t __offset, uin
  __arm_vstrdq_scatter_shifted_offset_u64 (__base, __offset, __value);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_p (int32_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrwq_scatter_offset_p_s32 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_p (uint32_t * __base, uint32x4_t __offset, uint32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrwq_scatter_offset_p_u32 (__base, __offset, __value, __p);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset (int32_t * __base, uint32x4_t __offset, int32x4_t __value)
-{
- __arm_vstrwq_scatter_offset_s32 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset (uint32_t * __base, uint32x4_t __offset, uint32x4_t __value)
-{
- __arm_vstrwq_scatter_offset_u32 (__base, __offset, __value);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_shifted_offset_p (int32_t * __base, uint32x4_t __offset, int32x4_t __value, mve_pred16_t __p)
@@ -3027,20 +2567,6 @@ __arm_vldrwq_gather_shifted_offset_z (float32_t const * __base, uint32x4_t __off
  return __arm_vldrwq_gather_shifted_offset_z_f32 (__base, __offset, __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);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrhq_scatter_offset_p (float16_t * __base, uint16x8_t __offset, float16x8_t __value, mve_pred16_t __p)
-{
- __arm_vstrhq_scatter_offset_p_f16 (__base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrhq_scatter_shifted_offset (float16_t * __base, uint16x8_t __offset, float16x8_t __value)
@@ -3069,20 +2595,6 @@ __arm_vstrwq_scatter_base_p (uint32x4_t __addr, const int __offset, float32x4_t
  __arm_vstrwq_scatter_base_p_f32 (__addr, __offset, __value, __p);
 }
 
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset (float32_t * __base, uint32x4_t __offset, float32x4_t __value)
-{
- __arm_vstrwq_scatter_offset_f32 (__base, __offset, __value);
-}
-
-__extension__ extern __inline void
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vstrwq_scatter_offset_p (float32_t * __base, uint32x4_t __offset, float32x4_t __value, mve_pred16_t __p)
-{
- __arm_vstrwq_scatter_offset_p_f32 (__base, __offset, __value, __p);
-}
-
 __extension__ extern __inline void
 __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
 __arm_vstrwq_scatter_shifted_offset (float32_t * __base, uint32x4_t __offset, float32x4_t __value)
@@ -3593,24 +3105,6 @@ 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_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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t), p3), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_p_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_offset_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_offset_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t), p3), \
-  int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_float16x8_t]: __arm_vstrhq_scatter_offset_p_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, float16x8_t), p3));})
-
-#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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t)), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_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_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_offset_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, float16x8_t)));})
-
 #define __arm_vstrhq_scatter_shifted_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, \
@@ -3629,24 +3123,6 @@ 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_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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t)), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_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_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_offset_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, float16x8_t)));})
-
-#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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t), p3), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_p_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_offset_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_offset_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t), p3), \
-  int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_float16x8_t]: __arm_vstrhq_scatter_offset_p_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, float16x8_t), p3));})
-
 #define __arm_vstrhq_scatter_shifted_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, \
@@ -3677,20 +3153,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_uint32x4_t]: __arm_vstrwq_scatter_base_p_u32(p0, p1, __ARM_mve_coerce(__p2, uint32x4_t), p3), \
   int (*)[__ARM_mve_type_float32x4_t]: __arm_vstrwq_scatter_base_p_f32(p0, p1, __ARM_mve_coerce(__p2, float32x4_t), p3));})
 
-#define __arm_vstrwq_scatter_offset(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_scatter_offset_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1, __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_scatter_offset_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1, __ARM_mve_coerce(__p2, uint32x4_t)), \
-  int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4_t]: __arm_vstrwq_scatter_offset_f32 (__ARM_mve_coerce_f32_ptr(__p0, float32_t *), p1, __ARM_mve_coerce(__p2, float32x4_t)));})
-
-#define __arm_vstrwq_scatter_offset_p(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_scatter_offset_p_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1, __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_scatter_offset_p_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1, __ARM_mve_coerce(__p2, uint32x4_t), p3), \
-  int (*)[__ARM_mve_type_float32_t_ptr][__ARM_mve_type_float32x4_t]: __arm_vstrwq_scatter_offset_p_f32 (__ARM_mve_coerce_f32_ptr(__p0, float32_t *), p1, __ARM_mve_coerce(__p2, float32x4_t), p3));})
-
 #define __arm_vstrwq_scatter_shifted_offset(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p2)])0, \
@@ -3868,21 +3330,6 @@ 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_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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t), p3), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_p_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_offset_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_offset_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t), p3));})
-
-#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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t)), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_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_offset_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t)));})
 
 #define __arm_vstrhq_scatter_shifted_offset_p(p0,p1,p2,p3) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
@@ -3910,22 +3357,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_int64x2_t]: __arm_vstrdq_scatter_base_s64 (p0, p1, __ARM_mve_coerce(__p2, int64x2_t)), \
   int (*)[__ARM_mve_type_uint64x2_t]: __arm_vstrdq_scatter_base_u64 (p0, p1, __ARM_mve_coerce(__p2, uint64x2_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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t)), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_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_offset_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t)));})
-
-#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, \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrhq_scatter_offset_p_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t), p3), \
-  int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrhq_scatter_offset_p_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrhq_scatter_offset_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t), p3), \
-  int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_offset_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t), p3));})
-
 #define __arm_vstrhq_scatter_shifted_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, \
@@ -3942,18 +3373,6 @@ 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_p_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t), p3), \
   int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrhq_scatter_shifted_offset_p_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t), p3));})
 
-#define __arm_vstrwq_scatter_offset(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_scatter_offset_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1, __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_scatter_offset_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1, __ARM_mve_coerce(__p2, uint32x4_t)));})
-
-#define __arm_vstrwq_scatter_offset_p(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int32_t_ptr][__ARM_mve_type_int32x4_t]: __arm_vstrwq_scatter_offset_p_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1, __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vstrwq_scatter_offset_p_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1, __ARM_mve_coerce(__p2, uint32x4_t), p3));})
-
 #define __arm_vstrwq_scatter_shifted_offset(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
   __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p2)])0, \
@@ -4074,40 +3493,6 @@ extern void *__ARM_undef;
   int (*)[__ARM_mve_type_int64x2_t]: __arm_vstrdq_scatter_base_p_s64 (p0, p1, __ARM_mve_coerce(__p2, int64x2_t), p3), \
   int (*)[__ARM_mve_type_uint64x2_t]: __arm_vstrdq_scatter_base_p_u64 (p0, p1, __ARM_mve_coerce(__p2, uint64x2_t), p3));})
 
-#define __arm_vstrbq_scatter_offset(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p1) __p1 = (p1); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_uint8x16_t][__ARM_mve_type_int8x16_t]: __arm_vstrbq_scatter_offset_s8 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, uint8x16_t), __ARM_mve_coerce(__p2, int8x16_t)), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrbq_scatter_offset_s16 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t)), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrbq_scatter_offset_s32 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t)), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint8x16_t][__ARM_mve_type_uint8x16_t]: __arm_vstrbq_scatter_offset_u8 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint8x16_t), __ARM_mve_coerce(__p2, uint8x16_t)), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrbq_scatter_offset_u16 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t)), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrbq_scatter_offset_u32 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t)));})
-
-#define __arm_vstrbq_scatter_offset_p(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p1) __p1 = (p1); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_uint8x16_t][__ARM_mve_type_int8x16_t]: __arm_vstrbq_scatter_offset_p_s8 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, uint8x16_t), __ARM_mve_coerce(__p2, int8x16_t), p3), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vstrbq_scatter_offset_p_s16 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, int16x8_t), p3), \
-  int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vstrbq_scatter_offset_p_s32 (__ARM_mve_coerce_s8_ptr(__p0, int8_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, int32x4_t), p3), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint8x16_t][__ARM_mve_type_uint8x16_t]: __arm_vstrbq_scatter_offset_p_u8 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint8x16_t), __ARM_mve_coerce(__p2, uint8x16_t), p3), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint16x8_t][__ARM_mve_type_uint16x8_t]: __arm_vstrbq_scatter_offset_p_u16 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint16x8_t), __ARM_mve_coerce(__p2, uint16x8_t), p3), \
-  int (*)[__ARM_mve_type_uint8_t_ptr][__ARM_mve_type_uint32x4_t][__ARM_mve_type_uint32x4_t]: __arm_vstrbq_scatter_offset_p_u32 (__ARM_mve_coerce_u8_ptr(__p0, uint8_t *), __ARM_mve_coerce(__p1, uint32x4_t), __ARM_mve_coerce(__p2, uint32x4_t), p3));})
-
-#define __arm_vstrdq_scatter_offset_p(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int64_t_ptr][__ARM_mve_type_int64x2_t]: __arm_vstrdq_scatter_offset_p_s64 (__ARM_mve_coerce_s64_ptr(__p0, int64_t *), p1, __ARM_mve_coerce(__p2, int64x2_t), p3), \
-  int (*)[__ARM_mve_type_uint64_t_ptr][__ARM_mve_type_uint64x2_t]: __arm_vstrdq_scatter_offset_p_u64 (__ARM_mve_coerce_u64_ptr(__p0, uint64_t *), p1, __ARM_mve_coerce(__p2, uint64x2_t), p3));})
-
-#define __arm_vstrdq_scatter_offset(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
-  __typeof(p2) __p2 = (p2); \
-  _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \
-  int (*)[__ARM_mve_type_int64_t_ptr][__ARM_mve_type_int64x2_t]: __arm_vstrdq_scatter_offset_s64 (__ARM_mve_coerce_s64_ptr(__p0, int64_t *), p1, __ARM_mve_coerce(__p2, int64x2_t)), \
-  int (*)[__ARM_mve_type_uint64_t_ptr][__ARM_mve_type_uint64x2_t]: __arm_vstrdq_scatter_offset_u64 (__ARM_mve_coerce_u64_ptr(__p0, uint64_t *), p1, __ARM_mve_coerce(__p2, uint64x2_t)));})
-
 #define __arm_vstrdq_scatter_shifted_offset_p(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \
   __typeof(p2) __p2 = (p2); \
   _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p2)])0, \

diff --git a/gcc/config/arm/arm_mve_builtins.def b/gcc/config/arm/arm_mve_builtins.def
index 5a0c76063396abb8203585e986f9282a6ba653f8..167d272916c6aad25ee306985d8a9881eb2577d6 100644 (file)
--- a/gcc/config/arm/arm_mve_builtins.def
+++ b/gcc/config/arm/arm_mve_builtins.def
@@ -663,16 +663,12 @@ 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 (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)
 VAR1 (LDRGBS, vldrwq_gather_base_s, v4si)
 VAR1 (LDRGBU, vldrwq_gather_base_u, 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)
 VAR1 (STRSBU_P, vstrwq_scatter_base_p_u, v4si)
 VAR1 (LDRGBS_Z, vldrwq_gather_base_z_s, v4si)
@@ -718,42 +714,26 @@ 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)
 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)
-VAR2 (STRSU, vstrhq_scatter_offset_u, v8hi, v4si)
 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 (STRSBS, vstrdq_scatter_base_s, v2di)
 VAR1 (STRSBS, vstrwq_scatter_base_f, v4sf)
 VAR1 (STRSBS_P, vstrdq_scatter_base_p_s, v2di)
 VAR1 (STRSBS_P, vstrwq_scatter_base_p_f, v4sf)
 VAR1 (STRSBU, vstrdq_scatter_base_u, v2di)
 VAR1 (STRSBU_P, vstrdq_scatter_base_p_u, v2di)
-VAR1 (STRSS, vstrdq_scatter_offset_s, v2di)
 VAR1 (STRSS, vstrdq_scatter_shifted_offset_s, v2di)
-VAR1 (STRSS, vstrhq_scatter_offset_f, v8hf)
 VAR1 (STRSS, vstrhq_scatter_shifted_offset_f, v8hf)
-VAR1 (STRSS, vstrwq_scatter_offset_f, v4sf)
-VAR1 (STRSS, vstrwq_scatter_offset_s, v4si)
 VAR1 (STRSS, vstrwq_scatter_shifted_offset_f, v4sf)
 VAR1 (STRSS, vstrwq_scatter_shifted_offset_s, v4si)
-VAR1 (STRSS_P, vstrdq_scatter_offset_p_s, v2di)
 VAR1 (STRSS_P, vstrdq_scatter_shifted_offset_p_s, v2di)
-VAR1 (STRSS_P, vstrhq_scatter_offset_p_f, v8hf)
 VAR1 (STRSS_P, vstrhq_scatter_shifted_offset_p_f, v8hf)
-VAR1 (STRSS_P, vstrwq_scatter_offset_p_f, v4sf)
-VAR1 (STRSS_P, vstrwq_scatter_offset_p_s, v4si)
 VAR1 (STRSS_P, vstrwq_scatter_shifted_offset_p_f, v4sf)
 VAR1 (STRSS_P, vstrwq_scatter_shifted_offset_p_s, v4si)
-VAR1 (STRSU, vstrdq_scatter_offset_u, v2di)
 VAR1 (STRSU, vstrdq_scatter_shifted_offset_u, v2di)
-VAR1 (STRSU, vstrwq_scatter_offset_u, v4si)
 VAR1 (STRSU, vstrwq_scatter_shifted_offset_u, v4si)
-VAR1 (STRSU_P, vstrdq_scatter_offset_p_u, v2di)
 VAR1 (STRSU_P, vstrdq_scatter_shifted_offset_p_u, v2di)
-VAR1 (STRSU_P, vstrwq_scatter_offset_p_u, v4si)
 VAR1 (STRSU_P, vstrwq_scatter_shifted_offset_p_u, v4si)
 VAR1 (STRSBWBU, vstrwq_scatter_base_wb_u, v4si)
 VAR1 (STRSBWBU, vstrdq_scatter_base_wb_u, v2di)

diff --git a/gcc/config/arm/iterators.md b/gcc/config/arm/iterators.md
index 22f8c1805657c3610c4cb8380c495bb8e2d742aa..f046225584a808b77042c6c208428e8f0532f94c 100644 (file)
--- a/gcc/config/arm/iterators.md
+++ b/gcc/config/arm/iterators.md
@@ -273,6 +273,7 @@
 (define_mode_iterator MVE_types [V16QI V8HI V4SI V2DI TI V8HF V4SF V2DF])
 (define_mode_iterator MVE_vecs [V16QI V8HI V4SI V2DI V8HF V4SF V2DF])
 (define_mode_iterator MVE_VLD_ST [V16QI V8HI V4SI V8HF V4SF])
+(define_mode_iterator MVE_VLD_ST_scatter [V16QI V8HI V4SI V8HF V4SF V2DI])
 (define_mode_iterator MVE_0 [V8HF V4SF])
 (define_mode_iterator MVE_1 [V16QI V8HI V4SI V2DI])
 (define_mode_iterator MVE_3 [V16QI V8HI])
@@ -290,6 +291,7 @@
 (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")])
+(define_mode_attr MVE_scatter_offset [(V16QI "V16QI") (V8HI "V8HI") (V4SI "V4SI") (V8HF "V8HI") (V4SF "V4SI") (V2DI "V2DI")])
 
 ;;----------------------------------------------------------------------------
 ;; Code iterators
@@ -1817,7 +1819,8 @@
 
 (define_mode_attr MVE_elem_ch [(V4QI "b") (V8QI "b") (V16QI "b")
                               (V4HI "h") (V8HI "h") (V8HF "h")
-                              (V4SI "w") (V4SF "w")])
+                              (V4SI "w") (V4SF "w")
+                              (V2DI "d")])
 
 (define_mode_attr VH_elem_ch [(V4HI "s") (V8HI  "s")
                              (V4HF "s") (V8HF  "s")
@@ -2521,8 +2524,7 @@
                       (VQRSHRNBQ_M_N_S "s") (VQRSHRNBQ_M_N_U "u")
                       (VMLALDAVAXQ_P_S "s")
                       (VMLALDAVAQ_P_S "s") (VMLALDAVAQ_P_U "u")
-                      (VSTRWQSB_S "s") (VSTRWQSB_U "u") (VSTRBQSO_S "s")
-                      (VSTRBQSO_U "u")
+                      (VSTRWQSB_S "s") (VSTRWQSB_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")
@@ -2530,11 +2532,10 @@
                       (VLDRDQGO_S "s") (VLDRDQGO_U "u") (VLDRDQGSO_S "s")
                       (VLDRDQGSO_U "u") (VLDRWQGO_S "s") (VLDRWQGO_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")
+                      (VSTRDQSB_S "s") (VSTRDQSB_U "u")
+                      (VSTRDQSSO_S "s") (VSTRDQSSO_U "u")
+                      (VSTRWQSSO_U "u")
                       (VSTRWQSSO_S "s") (VSTRWQSBWB_S "s") (VSTRWQSBWB_U "u")
                       (VLDRWQGBWB_S "s") (VLDRWQGBWB_U "u") (VLDRDQGBWB_S "s")
                       (VLDRDQGBWB_U "u") (VSTRDQSBWB_S "s") (VADCQ_M_S "s")
@@ -2937,7 +2938,6 @@
 (define_int_iterator VSHRNBQ_M_N [VSHRNBQ_M_N_S VSHRNBQ_M_N_U])
 (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 VLDRBGOQ [VLDRBQGO_S VLDRBQGO_U])
 (define_int_iterator VLDRWGBQ [VLDRWQGB_S VLDRWQGB_U])
 (define_int_iterator VLDRHGOQ [VLDRHQGO_S VLDRHQGO_U])
@@ -2947,12 +2947,9 @@
 (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 VSTRHSOQ [VSTRHQSO_S VSTRHQSO_U])
 (define_int_iterator VSTRHSSOQ [VSTRHQSSO_S VSTRHQSSO_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])
-(define_int_iterator VSTRWSOQ [VSTRWQSO_S VSTRWQSO_U])
 (define_int_iterator VSTRWSSOQ [VSTRWQSSO_S VSTRWQSSO_U])
 (define_int_iterator VSTRWSBWBQ [VSTRWQSBWB_S VSTRWQSBWB_U])
 (define_int_iterator VLDRWGBWBQ [VLDRWQGBWB_S VLDRWQGBWB_U])

diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md
index e54153edb5d929c1871a128fdef4a038b1e6d5b0..b2fb2d878c49a006d1cfdaa89198a348ab1c57d1 100644 (file)
--- a/gcc/config/arm/mve.md
+++ b/gcc/config/arm/mve.md
@@ -3311,36 +3311,93 @@
   (set_attr "type" "mve_move")
   (set_attr "length" "8")])
 
+;; Vector scatter stores with offset
 ;;
-;; [vstrbq_scatter_offset_s vstrbq_scatter_offset_u]
+;; [vstrbq_scatter_offset_s8,  vstrbq_scatter_offset_u8,
+;;  vstrhq_scatter_offset_s16, vstrhq_scatter_offset_u16,
+;;  vstrwq_scatter_offset_s32, vstrwq_scatter_offset_u32,
+;;  vstrdq_scatter_offset_s64, vstrdq_scatter_offset_u64,
+;;  vstrhq_scatter_offset_f16,
+;;  vstrwq_scatter_offset_f32]
 ;;
-(define_expand "mve_vstrbq_scatter_offset_<supf><mode>"
-  [(match_operand:<MVE_B_ELEM> 0 "mve_scatter_memory")
-   (match_operand:MVE_2 1 "s_register_operand")
-   (match_operand:MVE_2 2 "s_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRBSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrbq_scatter_offset_<supf><mode>_insn (ind, operands[1],
-                                                             operands[2]));
-  DONE;
-})
+(define_insn "@mve_vstrq_scatter_offset_<mode>"
+  [(set (mem:BLK (scratch))
+       (unspec:BLK
+         [(match_operand:SI 0 "register_operand" "r")
+          (match_operand:<MVE_scatter_offset> 1 "s_register_operand" "w")
+          (match_operand:MVE_VLD_ST_scatter 2 "s_register_operand" "w")]
+         VSTRQSO))]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+   || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+  "vstr<MVE_elem_ch>.<V_sz_elem>\t%q2, [%0, %q1]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrq_scatter_offset_<mode>"))
+  (set_attr "length" "4")])
 
-(define_insn "mve_vstrbq_scatter_offset_<supf><mode>_insn"
+;; Predicated vector scatter stores with offset
+;;
+;; [vstrbq_scatter_offset_p_s8, vstrbq_scatter_offset_p_u8,
+;; [vstrhq_scatter_offset_p_s16, vstrhq_scatter_offset_p_u16,
+;; [vstrwq_scatter_offset_p_s32, vstrwq_scatter_offset_p_u32,
+;; [vstrdq_scatter_offset_p_s64, vstrdq_scatter_offset_p_u64,
+;; [vstrhq_scatter_offset_p_f16,
+;; [vstrwq_scatter_offset_p_f32]
+;;
+(define_insn "@mve_vstrq_scatter_offset_p_<mode>"
   [(set (mem:BLK (scratch))
        (unspec:BLK
          [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:MVE_2 1 "s_register_operand" "w")
-          (match_operand:MVE_2 2 "s_register_operand" "w")]
-         VSTRBSOQ))]
-  "TARGET_HAVE_MVE"
-  "vstrb.<V_sz_elem>\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrbq_scatter_offset_<supf><mode>_insn"))
+          (match_operand:<MVE_scatter_offset> 1 "s_register_operand" "w")
+          (match_operand:MVE_VLD_ST_scatter 2 "s_register_operand" "w")
+          (match_operand:<MVE_VPRED> 3 "vpr_register_operand" "Up")]
+         VSTRQSO_P))]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+   || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+  "vpst\;vstr<MVE_elem_ch>t.<V_sz_elem>\t%q2, [%0, %q1]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrq_scatter_offset_<mode>"))
+  (set_attr "length" "8")])
+
+;; Truncating vector scatter stores with offset
+;;
+;; [vstrbq_scatter_offset_s16, vstrbq_scatter_offset_u16,
+;; [vstrbq_scatter_offset_s32, vstrbq_scatter_offset_u32,
+;; [vstrhq_scatter_offset_s32, vstrhq_scatter_offset_u32]
+;;
+(define_insn "@mve_vstrq_truncate_scatter_offset_<mode>"
+  [(set (mem:BLK (scratch))
+       (unspec:BLK
+         [(match_operand:SI 0 "register_operand" "r")
+          (match_operand:<MVE_wide_n_TYPE> 1 "s_register_operand" "w")
+          (truncate:MVE_w_narrow_TYPE
+            (match_operand:<MVE_wide_n_TYPE> 2 "s_register_operand" "w"))]
+         VSTRQSO_TRUNC))]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MVE_wide_n_TYPE>mode))"
+  "vstr<MVE_elem_ch>.<MVE_wide_n_sz_elem>\t%q2, [%0, %q1]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrq_truncate_scatter_offset_<mode>"))
   (set_attr "length" "4")])
 
+
+;; Predicated truncating vector scatter stores with offset
 ;;
+;; [vstrbq_scatter_offset_p_s16, vstrbq_scatter_offset_p_u16,
+;; [vstrbq_scatter_offset_p_s32, vstrbq_scatter_offset_p_u32,
+;; [vstrhq_scatter_offset_p_s32, vstrhq_scatter_offset_p_u32]
+;;
+(define_insn "@mve_vstrq_truncate_scatter_offset_p_<mode>"
+  [(set (mem:BLK (scratch))
+       (unspec:BLK
+         [(match_operand:SI 0 "register_operand" "r")
+          (match_operand:<MVE_wide_n_TYPE> 1 "s_register_operand" "w")
+          (truncate:MVE_w_narrow_TYPE
+            (match_operand:<MVE_wide_n_TYPE> 2 "s_register_operand" "w"))
+          (match_operand:<MVE_wide_n_VPRED> 3 "vpr_register_operand" "Up")]
+         VSTRQSO_TRUNC_P))]
+  "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MVE_wide_n_TYPE>mode))"
+  "vpst\;vstr<MVE_elem_ch>t.<MVE_wide_n_sz_elem>\t%q2, [%0, %q1]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrq_truncate_scatter_offset_<mode>"))
+  (set_attr "length" "8")])
+
+;
+;
 ;; [vstrwq_scatter_base_s vstrwq_scatter_base_u]
 ;;
 (define_insn "mve_vstrwq_scatter_base_<supf>v4si"
@@ -3407,40 +3464,6 @@
 }
  [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_base_<supf>v4si"))
   (set_attr "length" "4")])
-
-;;
-;; [vstrbq_scatter_offset_p_s vstrbq_scatter_offset_p_u]
-;;
-(define_expand "mve_vstrbq_scatter_offset_p_<supf><mode>"
-  [(match_operand:<MVE_B_ELEM>  0 "mve_scatter_memory")
-   (match_operand:MVE_2 1 "s_register_operand")
-   (match_operand:MVE_2 2 "s_register_operand")
-   (match_operand:<MVE_VPRED> 3 "vpr_register_operand" "Up")
-   (unspec:V4SI [(const_int 0)] VSTRBSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (
-    gen_mve_vstrbq_scatter_offset_p_<supf><mode>_insn (ind, operands[1],
-                                                      operands[2],
-                                                      operands[3]));
-  DONE;
-})
-
-(define_insn "mve_vstrbq_scatter_offset_p_<supf><mode>_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:MVE_2 1 "s_register_operand" "w")
-          (match_operand:MVE_2 2 "s_register_operand" "w")
-          (match_operand:<MVE_VPRED> 3 "vpr_register_operand" "Up")]
-         VSTRBSOQ))]
-  "TARGET_HAVE_MVE"
-  "vpst\;vstrbt.<V_sz_elem>\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrbq_scatter_offset_<supf><mode>_insn"))
-  (set_attr "length" "8")])
-
 ;;
 ;; [vstrwq_scatter_base_p_s vstrwq_scatter_base_p_u]
 ;;
@@ -4049,68 +4072,6 @@
  [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_shifted_offset_<supf>v4si"))
   (set_attr "length" "8")])
 
-;;
-;; [vstrhq_scatter_offset_p_s vstrhq_scatter_offset_p_u]
-;;
-(define_expand "mve_vstrhq_scatter_offset_p_<supf><mode>"
-  [(match_operand:<MVE_H_ELEM> 0 "mve_scatter_memory")
-   (match_operand:MVE_5 1 "s_register_operand")
-   (match_operand:MVE_5 2 "s_register_operand")
-   (match_operand:<MVE_VPRED> 3 "vpr_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRHSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (
-    gen_mve_vstrhq_scatter_offset_p_<supf><mode>_insn (ind, operands[1],
-                                                      operands[2],
-                                                      operands[3]));
-  DONE;
-})
-
-(define_insn "mve_vstrhq_scatter_offset_p_<supf><mode>_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:MVE_5 1 "s_register_operand" "w")
-          (match_operand:MVE_5 2 "s_register_operand" "w")
-          (match_operand:<MVE_VPRED> 3 "vpr_register_operand" "Up")]
-         VSTRHSOQ))]
-  "TARGET_HAVE_MVE"
-  "vpst\;vstrht.<V_sz_elem>\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_scatter_offset_<supf><mode>_insn"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrhq_scatter_offset_s vstrhq_scatter_offset_u]
-;;
-(define_expand "mve_vstrhq_scatter_offset_<supf><mode>"
-  [(match_operand:<MVE_H_ELEM> 0 "mve_scatter_memory")
-   (match_operand:MVE_5 1 "s_register_operand")
-   (match_operand:MVE_5 2 "s_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRHSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrhq_scatter_offset_<supf><mode>_insn (ind, operands[1],
-                                                             operands[2]));
-  DONE;
-})
-
-(define_insn "mve_vstrhq_scatter_offset_<supf><mode>_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:MVE_5 1 "s_register_operand" "w")
-          (match_operand:MVE_5 2 "s_register_operand" "w")]
-         VSTRHSOQ))]
-  "TARGET_HAVE_MVE"
-  "vstrh.<V_sz_elem>\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_scatter_offset_<supf><mode>_insn"))
-  (set_attr "length" "4")])
-
 ;;
 ;; [vstrhq_scatter_shifted_offset_p_s vstrhq_scatter_shifted_offset_p_u]
 ;;
@@ -4221,67 +4182,6 @@
  [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrdq_scatter_base_<supf>v2di"))
   (set_attr "length" "4")])
 
-;;
-;; [vstrdq_scatter_offset_p_s vstrdq_scatter_offset_p_u]
-;;
-(define_expand "mve_vstrdq_scatter_offset_p_<supf>v2di"
-  [(match_operand:V2DI 0 "mve_scatter_memory")
-   (match_operand:V2DI 1 "s_register_operand")
-   (match_operand:V2DI 2 "s_register_operand")
-   (match_operand:V2QI 3 "vpr_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRDSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrdq_scatter_offset_p_<supf>v2di_insn (ind, operands[1],
-                                                             operands[2],
-                                                             operands[3]));
-  DONE;
-})
-
-(define_insn "mve_vstrdq_scatter_offset_p_<supf>v2di_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V2DI 1 "s_register_operand" "w")
-          (match_operand:V2DI 2 "s_register_operand" "w")
-          (match_operand:V2QI 3 "vpr_register_operand" "Up")]
-         VSTRDSOQ))]
-  "TARGET_HAVE_MVE"
-  "vpst\;vstrdt.64\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrdq_scatter_offset_<supf>v2di_insn"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrdq_scatter_offset_s vstrdq_scatter_offset_u]
-;;
-(define_expand "mve_vstrdq_scatter_offset_<supf>v2di"
-  [(match_operand:V2DI 0 "mve_scatter_memory")
-   (match_operand:V2DI 1 "s_register_operand")
-   (match_operand:V2DI 2 "s_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRDSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrdq_scatter_offset_<supf>v2di_insn (ind, operands[1],
-                                                           operands[2]));
-  DONE;
-})
-
-(define_insn "mve_vstrdq_scatter_offset_<supf>v2di_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V2DI 1 "s_register_operand" "w")
-          (match_operand:V2DI 2 "s_register_operand" "w")]
-         VSTRDSOQ))]
-  "TARGET_HAVE_MVE"
-  "vstrd.64\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrdq_scatter_offset_<supf>v2di_insn"))
-  (set_attr "length" "4")])
-
 ;;
 ;; [vstrdq_scatter_shifted_offset_p_s vstrdq_scatter_shifted_offset_p_u]
 ;;
@@ -4345,67 +4245,6 @@
  [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrdq_scatter_shifted_offset_<supf>v2di_insn"))
   (set_attr "length" "4")])
 
-;;
-;; [vstrhq_scatter_offset_f]
-;;
-(define_expand "mve_vstrhq_scatter_offset_fv8hf"
-  [(match_operand:V8HI 0 "mve_scatter_memory")
-   (match_operand:V8HI 1 "s_register_operand")
-   (match_operand:V8HF 2 "s_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRHQSO_F)]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrhq_scatter_offset_fv8hf_insn (ind, operands[1],
-                                                      operands[2]));
-  DONE;
-})
-
-(define_insn "mve_vstrhq_scatter_offset_fv8hf_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V8HI 1 "s_register_operand" "w")
-          (match_operand:V8HF 2 "s_register_operand" "w")]
-         VSTRHQSO_F))]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-  "vstrh.16\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_scatter_offset_fv8hf_insn"))
-  (set_attr "length" "4")])
-
-;;
-;; [vstrhq_scatter_offset_p_f]
-;;
-(define_expand "mve_vstrhq_scatter_offset_p_fv8hf"
-  [(match_operand:V8HI 0 "mve_scatter_memory")
-   (match_operand:V8HI 1 "s_register_operand")
-   (match_operand:V8HF 2 "s_register_operand")
-   (match_operand:V8BI 3 "vpr_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRHQSO_F)]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrhq_scatter_offset_p_fv8hf_insn (ind, operands[1],
-                                                        operands[2],
-                                                        operands[3]));
-  DONE;
-})
-
-(define_insn "mve_vstrhq_scatter_offset_p_fv8hf_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V8HI 1 "s_register_operand" "w")
-          (match_operand:V8HF 2 "s_register_operand" "w")
-          (match_operand:V8BI 3 "vpr_register_operand" "Up")]
-         VSTRHQSO_F))]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-  "vpst\;vstrht.16\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrhq_scatter_offset_fv8hf_insn"))
-  (set_attr "length" "8")])
-
 ;;
 ;; [vstrhq_scatter_shifted_offset_f]
 ;;
@@ -4515,128 +4354,6 @@
  [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_scatter_base_fv4sf"))
   (set_attr "length" "8")])
 
-;;
-;; [vstrwq_scatter_offset_f]
-;;
-(define_expand "mve_vstrwq_scatter_offset_fv4sf"
-  [(match_operand:V4SI 0 "mve_scatter_memory")
-   (match_operand:V4SI 1 "s_register_operand")
-   (match_operand:V4SF 2 "s_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRWQSO_F)]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrwq_scatter_offset_fv4sf_insn (ind, operands[1],
-                                                      operands[2]));
-  DONE;
-})
-
-(define_insn "mve_vstrwq_scatter_offset_fv4sf_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V4SI 1 "s_register_operand" "w")
-          (match_operand:V4SF 2 "s_register_operand" "w")]
-         VSTRWQSO_F))]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-  "vstrw.32\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_scatter_offset_fv4sf_insn"))
-  (set_attr "length" "4")])
-
-;;
-;; [vstrwq_scatter_offset_p_f]
-;;
-(define_expand "mve_vstrwq_scatter_offset_p_fv4sf"
-  [(match_operand:V4SI 0 "mve_scatter_memory")
-   (match_operand:V4SI 1 "s_register_operand")
-   (match_operand:V4SF 2 "s_register_operand")
-   (match_operand:V4BI 3 "vpr_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRWQSO_F)]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrwq_scatter_offset_p_fv4sf_insn (ind, operands[1],
-                                                        operands[2],
-                                                        operands[3]));
-  DONE;
-})
-
-(define_insn "mve_vstrwq_scatter_offset_p_fv4sf_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V4SI 1 "s_register_operand" "w")
-          (match_operand:V4SF 2 "s_register_operand" "w")
-          (match_operand:V4BI 3 "vpr_register_operand" "Up")]
-         VSTRWQSO_F))]
-  "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-  "vpst\;vstrwt.32\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_scatter_offset_fv4sf_insn"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrwq_scatter_offset_s vstrwq_scatter_offset_u]
-;;
-(define_expand "mve_vstrwq_scatter_offset_p_<supf>v4si"
-  [(match_operand:V4SI 0 "mve_scatter_memory")
-   (match_operand:V4SI 1 "s_register_operand")
-   (match_operand:V4SI 2 "s_register_operand")
-   (match_operand:V4BI 3 "vpr_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRWSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrwq_scatter_offset_p_<supf>v4si_insn (ind, operands[1],
-                                                             operands[2],
-                                                             operands[3]));
-  DONE;
-})
-
-(define_insn "mve_vstrwq_scatter_offset_p_<supf>v4si_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V4SI 1 "s_register_operand" "w")
-          (match_operand:V4SI 2 "s_register_operand" "w")
-          (match_operand:V4BI 3 "vpr_register_operand" "Up")]
-         VSTRWSOQ))]
-  "TARGET_HAVE_MVE"
-  "vpst\;vstrwt.32\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_scatter_offset_<supf>v4si_insn"))
-  (set_attr "length" "8")])
-
-;;
-;; [vstrwq_scatter_offset_s vstrwq_scatter_offset_u]
-;;
-(define_expand "mve_vstrwq_scatter_offset_<supf>v4si"
-  [(match_operand:V4SI 0 "mve_scatter_memory")
-   (match_operand:V4SI 1 "s_register_operand")
-   (match_operand:V4SI 2 "s_register_operand")
-   (unspec:V4SI [(const_int 0)] VSTRWSOQ)]
-  "TARGET_HAVE_MVE"
-{
-  rtx ind = XEXP (operands[0], 0);
-  gcc_assert (REG_P (ind));
-  emit_insn (gen_mve_vstrwq_scatter_offset_<supf>v4si_insn (ind, operands[1],
-                                                           operands[2]));
-  DONE;
-})
-
-(define_insn "mve_vstrwq_scatter_offset_<supf>v4si_insn"
-  [(set (mem:BLK (scratch))
-       (unspec:BLK
-         [(match_operand:SI 0 "register_operand" "r")
-          (match_operand:V4SI 1 "s_register_operand" "w")
-          (match_operand:V4SI 2 "s_register_operand" "w")]
-         VSTRWSOQ))]
-  "TARGET_HAVE_MVE"
-  "vstrw.32\t%q2, [%0, %q1]"
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vstrwq_scatter_offset_<supf>v4si_insn"))
-  (set_attr "length" "4")])
-
 ;;
 ;; [vstrwq_scatter_shifted_offset_f]
 ;;

diff --git a/gcc/config/arm/unspecs.md b/gcc/config/arm/unspecs.md
index 01963d54cd4dbe719622d10e5ce69ee5c02b789f..5315db580f3d8190392e1bccc1f658943353cff8 100644 (file)
--- a/gcc/config/arm/unspecs.md
+++ b/gcc/config/arm/unspecs.md
@@ -1148,8 +1148,10 @@
   VSUBQ_M_F
   VSTRWQSB_S
   VSTRWQSB_U
-  VSTRBQSO_S
-  VSTRBQSO_U
+  VSTRQSO
+  VSTRQSO_P
+  VSTRQSO_TRUNC
+  VSTRQSO_TRUNC_P
   VLDRQ
   VLDRQ_Z
   VLDRQ_EXT
@@ -1181,21 +1183,14 @@
   VSTRQ_P
   VSTRQ_TRUNC
   VSTRQ_TRUNC_P
-  VSTRHQSO_S
   VSTRDQSB_S
   VSTRDQSB_U
-  VSTRDQSO_S
-  VSTRDQSO_U
   VSTRDQSSO_S
   VSTRDQSSO_U
-  VSTRWQSO_S
-  VSTRWQSO_U
   VSTRWQSSO_S
   VSTRWQSSO_U
-  VSTRHQSO_F
   VSTRHQSSO_F
   VSTRWQSB_F
-  VSTRWQSO_F
   VSTRWQSSO_F
   VDDUPQ
   VDDUPQ_M
@@ -1236,7 +1231,6 @@
   VST2Q
   VSHLCQ_M_U
   VSHLCQ_M_S
-  VSTRHQSO_U
   VSTRHQSSO_S
   VSTRHQSSO_U
   VSTRHQ_S