@@ -610,16 +610,6 @@ arm_quadop_unone_unone_unone_none_pred_qualifiers[SIMD_MAX_BUILTIN_ARGS]
#define QUADOP_UNONE_UNONE_UNONE_NONE_PRED_QUALIFIERS \
(arm_quadop_unone_unone_unone_none_pred_qualifiers)
-static enum arm_type_qualifiers
-arm_ldrgu_qualifiers[SIMD_MAX_BUILTIN_ARGS]
- = { qualifier_unsigned, qualifier_pointer, qualifier_unsigned};
-#define LDRGU_QUALIFIERS (arm_ldrgu_qualifiers)
-
-static enum arm_type_qualifiers
-arm_ldrgs_qualifiers[SIMD_MAX_BUILTIN_ARGS]
- = { qualifier_none, qualifier_pointer, qualifier_unsigned};
-#define LDRGS_QUALIFIERS (arm_ldrgs_qualifiers)
-
static enum arm_type_qualifiers
arm_ldrgbs_qualifiers[SIMD_MAX_BUILTIN_ARGS]
= { qualifier_none, qualifier_unsigned, qualifier_immediate};
@@ -642,18 +632,6 @@ arm_ldrgbu_z_qualifiers[SIMD_MAX_BUILTIN_ARGS]
qualifier_predicate};
#define LDRGBU_Z_QUALIFIERS (arm_ldrgbu_z_qualifiers)
-static enum arm_type_qualifiers
-arm_ldrgs_z_qualifiers[SIMD_MAX_BUILTIN_ARGS]
- = { qualifier_none, qualifier_pointer, qualifier_unsigned,
- qualifier_predicate};
-#define LDRGS_Z_QUALIFIERS (arm_ldrgs_z_qualifiers)
-
-static enum arm_type_qualifiers
-arm_ldrgu_z_qualifiers[SIMD_MAX_BUILTIN_ARGS]
- = { qualifier_unsigned, qualifier_pointer, qualifier_unsigned,
- qualifier_predicate};
-#define LDRGU_Z_QUALIFIERS (arm_ldrgu_z_qualifiers)
-
static enum arm_type_qualifiers
arm_ldrgbwbxu_qualifiers[SIMD_MAX_BUILTIN_ARGS]
= { qualifier_unsigned, qualifier_unsigned, qualifier_immediate};
@@ -417,33 +417,56 @@ class vldrq_gather_impl : public load_extending
public:
using load_extending::load_extending;
+ CONSTEXPR vldrq_gather_impl (bool shifted,
+ type_suffix_index signed_memory_type,
+ type_suffix_index unsigned_memory_type,
+ type_suffix_index float_memory_type)
+ : load_extending (signed_memory_type, unsigned_memory_type, float_memory_type),
+ m_shifted (shifted)
+ {}
+
+ CONSTEXPR vldrq_gather_impl (bool shifted,
+ type_suffix_index signed_memory_type,
+ type_suffix_index unsigned_memory_type)
+ : load_extending (signed_memory_type, unsigned_memory_type, NUM_TYPE_SUFFIXES),
+ m_shifted (shifted)
+ {}
+
+ /* Shifted offset (true) or plain offset (false). */
+ bool m_shifted;
+
rtx expand (function_expander &e) const override
{
insn_code icode;
+ machine_mode memory_mode = e.memory_vector_mode ();
+ rtx_code extend = (e.type_suffix (0).unsigned_p
+ ? ZERO_EXTEND
+ : SIGN_EXTEND);
+
switch (e.pred)
{
case PRED_none:
- if (e.vector_mode (0) == e.memory_vector_mode ())
- /* Non-extending load case. */
- icode = code_for_mve_vldrq_gather_offset (e.vector_mode (0));
- else
- /* Extending load case. */
- icode = code_for_mve_vldrq_gather_offset_extend
- (e.memory_vector_mode (),
- e.type_suffix (0).unsigned_p
- ? ZERO_EXTEND
- : SIGN_EXTEND);
+ icode = (e.vector_mode (0) == memory_mode
+ /* Non-extending load case. */
+ ? (m_shifted
+ ? code_for_mve_vldrq_gather_shifted_offset (memory_mode)
+ : code_for_mve_vldrq_gather_offset (memory_mode))
+ /* Extending load case. */
+ : (m_shifted
+ ? code_for_mve_vldrq_gather_shifted_offset_extend_v4si (extend)
+ : code_for_mve_vldrq_gather_offset_extend (memory_mode,
+ extend)));
break;
case PRED_z:
- if (e.vector_mode (0) == e.memory_vector_mode ())
- icode = code_for_mve_vldrq_gather_offset_z (e.vector_mode (0));
- else
- icode = code_for_mve_vldrq_gather_offset_z_extend
- (e.memory_vector_mode (),
- e.type_suffix (0).unsigned_p
- ? ZERO_EXTEND
- : SIGN_EXTEND);
+ icode = (e.vector_mode (0) == memory_mode
+ ? (m_shifted
+ ? code_for_mve_vldrq_gather_shifted_offset_z (memory_mode)
+ : code_for_mve_vldrq_gather_offset_z (memory_mode))
+ : (m_shifted
+ ? code_for_mve_vldrq_gather_shifted_offset_z_extend_v4si (extend)
+ : code_for_mve_vldrq_gather_offset_z_extend (memory_mode,
+ extend)));
break;
default:
@@ -1251,12 +1274,15 @@ FUNCTION_WITH_M_N_NO_F (vhaddq, VHADDQ)
FUNCTION_WITH_M_N_NO_F (vhsubq, VHSUBQ)
FUNCTION (vld1q, vld1_impl,)
FUNCTION (vldrbq, vldrq_impl, (TYPE_SUFFIX_s8, TYPE_SUFFIX_u8))
-FUNCTION (vldrbq_gather, vldrq_gather_impl, (TYPE_SUFFIX_s8, TYPE_SUFFIX_u8))
-FUNCTION (vldrdq_gather, vldrq_gather_impl, (TYPE_SUFFIX_s64, TYPE_SUFFIX_u64, NUM_TYPE_SUFFIXES))
+FUNCTION (vldrbq_gather, vldrq_gather_impl, (false, TYPE_SUFFIX_s8, TYPE_SUFFIX_u8))
+FUNCTION (vldrdq_gather, vldrq_gather_impl, (false, TYPE_SUFFIX_s64, TYPE_SUFFIX_u64, NUM_TYPE_SUFFIXES))
+FUNCTION (vldrdq_gather_shifted, vldrq_gather_impl, (true, TYPE_SUFFIX_s64, TYPE_SUFFIX_u64, NUM_TYPE_SUFFIXES))
FUNCTION (vldrhq, vldrq_impl, (TYPE_SUFFIX_s16, TYPE_SUFFIX_u16, TYPE_SUFFIX_f16))
-FUNCTION (vldrhq_gather, vldrq_gather_impl, (TYPE_SUFFIX_s16, TYPE_SUFFIX_u16, TYPE_SUFFIX_f16))
+FUNCTION (vldrhq_gather, vldrq_gather_impl, (false, TYPE_SUFFIX_s16, TYPE_SUFFIX_u16, TYPE_SUFFIX_f16))
+FUNCTION (vldrhq_gather_shifted, vldrq_gather_impl, (true, TYPE_SUFFIX_s16, TYPE_SUFFIX_u16, TYPE_SUFFIX_f16))
FUNCTION (vldrwq, vldrq_impl, (TYPE_SUFFIX_s32, TYPE_SUFFIX_u32, TYPE_SUFFIX_f32))
-FUNCTION (vldrwq_gather, vldrq_gather_impl, (TYPE_SUFFIX_s32, TYPE_SUFFIX_u32, TYPE_SUFFIX_f32))
+FUNCTION (vldrwq_gather, vldrq_gather_impl, (false, TYPE_SUFFIX_s32, TYPE_SUFFIX_u32, TYPE_SUFFIX_f32))
+FUNCTION (vldrwq_gather_shifted, vldrq_gather_impl, (true, TYPE_SUFFIX_s32, TYPE_SUFFIX_u32, TYPE_SUFFIX_f32))
FUNCTION_PRED_P_S (vmaxavq, VMAXAVQ)
FUNCTION_WITHOUT_N_NO_U_F (vmaxaq, VMAXAQ)
FUNCTION_ONLY_F (vmaxnmaq, VMAXNMAQ)
@@ -62,10 +62,13 @@ DEF_MVE_FUNCTION (vld1q, load, all_integer, z_or_none)
DEF_MVE_FUNCTION (vldrbq, load_ext, all_integer, z_or_none)
DEF_MVE_FUNCTION (vldrbq_gather, load_ext_gather_offset, all_integer, z_or_none)
DEF_MVE_FUNCTION (vldrdq_gather, load_ext_gather_offset, integer_64, z_or_none)
+DEF_MVE_FUNCTION (vldrdq_gather_shifted, load_ext_gather_offset, integer_64, z_or_none)
DEF_MVE_FUNCTION (vldrhq, load_ext, integer_16_32, z_or_none)
DEF_MVE_FUNCTION (vldrhq_gather, load_ext_gather_offset, integer_16_32, z_or_none)
+DEF_MVE_FUNCTION (vldrhq_gather_shifted, load_ext_gather_offset, integer_16_32, z_or_none)
DEF_MVE_FUNCTION (vldrwq, load_ext, integer_32, z_or_none)
DEF_MVE_FUNCTION (vldrwq_gather, load_ext_gather_offset, integer_32, z_or_none)
+DEF_MVE_FUNCTION (vldrwq_gather_shifted, load_ext_gather_offset, integer_32, z_or_none)
DEF_MVE_FUNCTION (vmaxaq, binary_maxamina, all_signed, m_or_none)
DEF_MVE_FUNCTION (vmaxavq, binary_maxavminav, all_signed, p_or_none)
DEF_MVE_FUNCTION (vmaxq, binary, all_integer, mx_or_none)
@@ -231,8 +234,10 @@ DEF_MVE_FUNCTION (vfmsq, ternary, all_float, m_or_none)
DEF_MVE_FUNCTION (vld1q, load, all_float, z_or_none)
DEF_MVE_FUNCTION (vldrhq, load_ext, float_16, z_or_none)
DEF_MVE_FUNCTION (vldrhq_gather, load_ext_gather_offset, float_16, z_or_none)
+DEF_MVE_FUNCTION (vldrhq_gather_shifted, load_ext_gather_offset, float_16, z_or_none)
DEF_MVE_FUNCTION (vldrwq, load_ext, float_32, z_or_none)
DEF_MVE_FUNCTION (vldrwq_gather, load_ext_gather_offset, float_32, z_or_none)
+DEF_MVE_FUNCTION (vldrwq_gather_shifted, load_ext_gather_offset, float_32, z_or_none)
DEF_MVE_FUNCTION (vmaxnmaq, binary, all_float, m_or_none)
DEF_MVE_FUNCTION (vmaxnmavq, binary_maxvminv, all_float, p_or_none)
DEF_MVE_FUNCTION (vmaxnmq, binary, all_float, mx_or_none)
@@ -85,10 +85,13 @@ extern const function_base *const vld1q;
extern const function_base *const vldrbq;
extern const function_base *const vldrbq_gather;
extern const function_base *const vldrdq_gather;
+extern const function_base *const vldrdq_gather_shifted;
extern const function_base *const vldrhq;
extern const function_base *const vldrhq_gather;
+extern const function_base *const vldrhq_gather_shifted;
extern const function_base *const vldrwq;
extern const function_base *const vldrwq_gather;
+extern const function_base *const vldrwq_gather_shifted;
extern const function_base *const vmaxaq;
extern const function_base *const vmaxavq;
extern const function_base *const vmaxnmaq;
@@ -42,12 +42,6 @@
#ifndef __ARM_MVE_PRESERVE_USER_NAMESPACE
#define vst4q(__addr, __value) __arm_vst4q(__addr, __value)
-#define vldrhq_gather_shifted_offset(__base, __offset) __arm_vldrhq_gather_shifted_offset(__base, __offset)
-#define vldrhq_gather_shifted_offset_z(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z(__base, __offset, __p)
-#define vldrdq_gather_shifted_offset(__base, __offset) __arm_vldrdq_gather_shifted_offset(__base, __offset)
-#define vldrdq_gather_shifted_offset_z(__base, __offset, __p) __arm_vldrdq_gather_shifted_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 vuninitializedq(__v) __arm_vuninitializedq(__v)
#define vst2q(__addr, __value) __arm_vst2q(__addr, __value)
#define vld2q(__addr) __arm_vld2q(__addr)
@@ -69,32 +63,12 @@
#define vldrwq_gather_base_u32(__addr, __offset) __arm_vldrwq_gather_base_u32(__addr, __offset)
#define vldrwq_gather_base_z_u32(__addr, __offset, __p) __arm_vldrwq_gather_base_z_u32(__addr, __offset, __p)
#define vldrwq_gather_base_z_s32(__addr, __offset, __p) __arm_vldrwq_gather_base_z_s32(__addr, __offset, __p)
-#define vldrhq_gather_shifted_offset_s32(__base, __offset) __arm_vldrhq_gather_shifted_offset_s32(__base, __offset)
-#define vldrhq_gather_shifted_offset_s16(__base, __offset) __arm_vldrhq_gather_shifted_offset_s16(__base, __offset)
-#define vldrhq_gather_shifted_offset_u32(__base, __offset) __arm_vldrhq_gather_shifted_offset_u32(__base, __offset)
-#define vldrhq_gather_shifted_offset_u16(__base, __offset) __arm_vldrhq_gather_shifted_offset_u16(__base, __offset)
-#define vldrhq_gather_shifted_offset_z_s32(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_s32(__base, __offset, __p)
-#define vldrhq_gather_shifted_offset_z_s16(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_s16(__base, __offset, __p)
-#define vldrhq_gather_shifted_offset_z_u32(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_u32(__base, __offset, __p)
-#define vldrhq_gather_shifted_offset_z_u16(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_u16(__base, __offset, __p)
#define vldrdq_gather_base_s64(__addr, __offset) __arm_vldrdq_gather_base_s64(__addr, __offset)
#define vldrdq_gather_base_u64(__addr, __offset) __arm_vldrdq_gather_base_u64(__addr, __offset)
#define vldrdq_gather_base_z_s64(__addr, __offset, __p) __arm_vldrdq_gather_base_z_s64(__addr, __offset, __p)
#define vldrdq_gather_base_z_u64(__addr, __offset, __p) __arm_vldrdq_gather_base_z_u64(__addr, __offset, __p)
-#define vldrdq_gather_shifted_offset_s64(__base, __offset) __arm_vldrdq_gather_shifted_offset_s64(__base, __offset)
-#define vldrdq_gather_shifted_offset_u64(__base, __offset) __arm_vldrdq_gather_shifted_offset_u64(__base, __offset)
-#define vldrdq_gather_shifted_offset_z_s64(__base, __offset, __p) __arm_vldrdq_gather_shifted_offset_z_s64(__base, __offset, __p)
-#define vldrdq_gather_shifted_offset_z_u64(__base, __offset, __p) __arm_vldrdq_gather_shifted_offset_z_u64(__base, __offset, __p)
-#define vldrhq_gather_shifted_offset_f16(__base, __offset) __arm_vldrhq_gather_shifted_offset_f16(__base, __offset)
-#define vldrhq_gather_shifted_offset_z_f16(__base, __offset, __p) __arm_vldrhq_gather_shifted_offset_z_f16(__base, __offset, __p)
#define vldrwq_gather_base_f32(__addr, __offset) __arm_vldrwq_gather_base_f32(__addr, __offset)
#define vldrwq_gather_base_z_f32(__addr, __offset, __p) __arm_vldrwq_gather_base_z_f32(__addr, __offset, __p)
-#define vldrwq_gather_shifted_offset_f32(__base, __offset) __arm_vldrwq_gather_shifted_offset_f32(__base, __offset)
-#define vldrwq_gather_shifted_offset_s32(__base, __offset) __arm_vldrwq_gather_shifted_offset_s32(__base, __offset)
-#define vldrwq_gather_shifted_offset_u32(__base, __offset) __arm_vldrwq_gather_shifted_offset_u32(__base, __offset)
-#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 vuninitializedq_u8(void) __arm_vuninitializedq_u8(void)
#define vuninitializedq_u16(void) __arm_vuninitializedq_u16(void)
#define vuninitializedq_u32(void) __arm_vuninitializedq_u32(void)
@@ -278,62 +252,6 @@ __arm_vldrwq_gather_base_z_u32 (uint32x4_t __addr, const int __offset, mve_pred1
return __builtin_mve_vldrwq_gather_base_z_uv4si (__addr, __offset, __p);
}
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_s32 (int16_t const * __base, uint32x4_t __offset)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_sv4si ((__builtin_neon_hi *) __base, __offset);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_s16 (int16_t const * __base, uint16x8_t __offset)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_sv8hi ((__builtin_neon_hi *) __base, __offset);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_u32 (uint16_t const * __base, uint32x4_t __offset)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_uv4si ((__builtin_neon_hi *) __base, __offset);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_u16 (uint16_t const * __base, uint16x8_t __offset)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_uv8hi ((__builtin_neon_hi *) __base, __offset);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z_s32 (int16_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_z_sv4si ((__builtin_neon_hi *) __base, __offset, __p);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z_s16 (int16_t const * __base, uint16x8_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_z_sv8hi ((__builtin_neon_hi *) __base, __offset, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z_u32 (uint16_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_z_uv4si ((__builtin_neon_hi *) __base, __offset, __p);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z_u16 (uint16_t const * __base, uint16x8_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_z_uv8hi ((__builtin_neon_hi *) __base, __offset, __p);
-}
-
__extension__ extern __inline int64x2_t
__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
__arm_vldrdq_gather_base_s64 (uint64x2_t __addr, const int __offset)
@@ -362,62 +280,6 @@ __arm_vldrdq_gather_base_z_u64 (uint64x2_t __addr, const int __offset, mve_pred1
return __builtin_mve_vldrdq_gather_base_z_uv2di (__addr, __offset, __p);
}
-__extension__ extern __inline int64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset_s64 (int64_t const * __base, uint64x2_t __offset)
-{
- return __builtin_mve_vldrdq_gather_shifted_offset_sv2di ((__builtin_neon_di *) __base, __offset);
-}
-
-__extension__ extern __inline uint64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset_u64 (uint64_t const * __base, uint64x2_t __offset)
-{
- return __builtin_mve_vldrdq_gather_shifted_offset_uv2di ((__builtin_neon_di *) __base, __offset);
-}
-
-__extension__ extern __inline int64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset_z_s64 (int64_t const * __base, uint64x2_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrdq_gather_shifted_offset_z_sv2di ((__builtin_neon_di *) __base, __offset, __p);
-}
-
-__extension__ extern __inline uint64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset_z_u64 (uint64_t const * __base, uint64x2_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrdq_gather_shifted_offset_z_uv2di ((__builtin_neon_di *) __base, __offset, __p);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_s32 (int32_t const * __base, uint32x4_t __offset)
-{
- return __builtin_mve_vldrwq_gather_shifted_offset_sv4si ((__builtin_neon_si *) __base, __offset);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_u32 (uint32_t const * __base, uint32x4_t __offset)
-{
- return __builtin_mve_vldrwq_gather_shifted_offset_uv4si ((__builtin_neon_si *) __base, __offset);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_z_s32 (int32_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrwq_gather_shifted_offset_z_sv4si ((__builtin_neon_si *) __base, __offset, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_z_u32 (uint32_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrwq_gather_shifted_offset_z_uv4si ((__builtin_neon_si *) __base, __offset, __p);
-}
-
__extension__ extern __inline int64x2_t
__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
__arm_vldrdq_gather_base_wb_s64 (uint64x2_t * __addr, const int __offset)
@@ -928,20 +790,6 @@ __arm_vst4q_f32 (float32_t * __addr, float32x4x4_t __value)
__builtin_mve_vst4qv4sf (__addr, __rv.__o);
}
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_f16 (float16_t const * __base, uint16x8_t __offset)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_fv8hf ((__builtin_neon_hi *) __base, __offset);
-}
-
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z_f16 (float16_t const * __base, uint16x8_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrhq_gather_shifted_offset_z_fv8hf ((__builtin_neon_hi *) __base, __offset, __p);
-}
-
__extension__ extern __inline float32x4_t
__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
__arm_vldrwq_gather_base_f32 (uint32x4_t __addr, const int __offset)
@@ -956,20 +804,6 @@ __arm_vldrwq_gather_base_z_f32 (uint32x4_t __addr, const int __offset, mve_pred1
return __builtin_mve_vldrwq_gather_base_z_fv4sf (__addr, __offset, __p);
}
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_f32 (float32_t const * __base, uint32x4_t __offset)
-{
- return __builtin_mve_vldrwq_gather_shifted_offset_fv4sf ((__builtin_neon_si *) __base, __offset);
-}
-
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_z_f32 (float32_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __builtin_mve_vldrwq_gather_shifted_offset_z_fv4sf ((__builtin_neon_si *) __base, __offset, __p);
-}
-
__extension__ extern __inline float32x4_t
__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
__arm_vldrwq_gather_base_wb_f32 (uint32x4_t * __addr, const int __offset)
@@ -1122,118 +956,6 @@ __arm_vst4q (uint32_t * __addr, uint32x4x4_t __value)
__arm_vst4q_u32 (__addr, __value);
}
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset (int16_t const * __base, uint32x4_t __offset)
-{
- return __arm_vldrhq_gather_shifted_offset_s32 (__base, __offset);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset (int16_t const * __base, uint16x8_t __offset)
-{
- return __arm_vldrhq_gather_shifted_offset_s16 (__base, __offset);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset (uint16_t const * __base, uint32x4_t __offset)
-{
- return __arm_vldrhq_gather_shifted_offset_u32 (__base, __offset);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset (uint16_t const * __base, uint16x8_t __offset)
-{
- return __arm_vldrhq_gather_shifted_offset_u16 (__base, __offset);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z (int16_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrhq_gather_shifted_offset_z_s32 (__base, __offset, __p);
-}
-
-__extension__ extern __inline int16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z (int16_t const * __base, uint16x8_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrhq_gather_shifted_offset_z_s16 (__base, __offset, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z (uint16_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrhq_gather_shifted_offset_z_u32 (__base, __offset, __p);
-}
-
-__extension__ extern __inline uint16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z (uint16_t const * __base, uint16x8_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrhq_gather_shifted_offset_z_u16 (__base, __offset, __p);
-}
-
-__extension__ extern __inline int64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset (int64_t const * __base, uint64x2_t __offset)
-{
- return __arm_vldrdq_gather_shifted_offset_s64 (__base, __offset);
-}
-
-__extension__ extern __inline uint64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset (uint64_t const * __base, uint64x2_t __offset)
-{
- return __arm_vldrdq_gather_shifted_offset_u64 (__base, __offset);
-}
-
-__extension__ extern __inline int64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset_z (int64_t const * __base, uint64x2_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrdq_gather_shifted_offset_z_s64 (__base, __offset, __p);
-}
-
-__extension__ extern __inline uint64x2_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrdq_gather_shifted_offset_z (uint64_t const * __base, uint64x2_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrdq_gather_shifted_offset_z_u64 (__base, __offset, __p);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset (int32_t const * __base, uint32x4_t __offset)
-{
- return __arm_vldrwq_gather_shifted_offset_s32 (__base, __offset);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset (uint32_t const * __base, uint32x4_t __offset)
-{
- return __arm_vldrwq_gather_shifted_offset_u32 (__base, __offset);
-}
-
-__extension__ extern __inline int32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_z (int32_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrwq_gather_shifted_offset_z_s32 (__base, __offset, __p);
-}
-
-__extension__ extern __inline uint32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_z (uint32_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrwq_gather_shifted_offset_z_u32 (__base, __offset, __p);
-}
-
__extension__ extern __inline void
__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
__arm_vst2q (int8_t * __addr, int8x16x2_t __value)
@@ -1488,34 +1210,6 @@ __arm_vst4q (float32_t * __addr, float32x4x4_t __value)
__arm_vst4q_f32 (__addr, __value);
}
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset (float16_t const * __base, uint16x8_t __offset)
-{
- return __arm_vldrhq_gather_shifted_offset_f16 (__base, __offset);
-}
-
-__extension__ extern __inline float16x8_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrhq_gather_shifted_offset_z (float16_t const * __base, uint16x8_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrhq_gather_shifted_offset_z_f16 (__base, __offset, __p);
-}
-
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset (float32_t const * __base, uint32x4_t __offset)
-{
- return __arm_vldrwq_gather_shifted_offset_f32 (__base, __offset);
-}
-
-__extension__ extern __inline float32x4_t
-__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
-__arm_vldrwq_gather_shifted_offset_z (float32_t const * __base, uint32x4_t __offset, mve_pred16_t __p)
-{
- return __arm_vldrwq_gather_shifted_offset_z_f32 (__base, __offset, __p);
-}
-
__extension__ extern __inline float16x8x4_t
__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
__arm_vld4q (float16_t const * __addr)
@@ -1931,34 +1625,6 @@ extern void *__ARM_undef;
int (*)[__ARM_mve_type_float16_t_ptr]: __arm_vld4q_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *)), \
int (*)[__ARM_mve_type_float32_t_ptr]: __arm_vld4q_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *))))
-#define __arm_vldrhq_gather_shifted_offset(p0,p1) ({ __typeof(p1) __p1 = (p1); \
- _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t)), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t)), \
- int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t)));})
-
-#define __arm_vldrhq_gather_shifted_offset_z(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
- _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_z_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_z_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_z_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_z_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2), \
- int (*)[__ARM_mve_type_float16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_z_f16 (__ARM_mve_coerce_f16_ptr(p0, float16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2));})
-
-#define __arm_vldrwq_gather_shifted_offset(p0,p1) ( \
- _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
- int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vldrwq_gather_shifted_offset_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), p1), \
- int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vldrwq_gather_shifted_offset_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), p1), \
- int (*)[__ARM_mve_type_float32_t_ptr]: __arm_vldrwq_gather_shifted_offset_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), p1)))
-
-#define __arm_vldrwq_gather_shifted_offset_z(p0,p1,p2) ( \
- _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
- int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_s32 (__ARM_mve_coerce_s32_ptr(p0, int32_t *), p1, p2), \
- int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), p1, p2), \
- int (*)[__ARM_mve_type_float32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_f32 (__ARM_mve_coerce_f32_ptr(p0, float32_t *), p1, p2)))
-
#define __arm_vst2q(p0,p1) ({ __typeof(p1) __p1 = (p1); \
_Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16x2_t]: __arm_vst2q_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16x2_t)), \
@@ -2021,30 +1687,6 @@ extern void *__ARM_undef;
int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8x4_t]: __arm_vst4q_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8x4_t)), \
int (*)[__ARM_mve_type_uint32_t_ptr][__ARM_mve_type_uint32x4x4_t]: __arm_vst4q_u32 (__ARM_mve_coerce_u32_ptr(p0, uint32_t *), __ARM_mve_coerce(__p1, uint32x4x4_t)));})
-#define __arm_vldrhq_gather_shifted_offset(p0,p1) ({ __typeof(p1) __p1 = (p1); \
- _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t)), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t)), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t)));})
-
-#define __arm_vldrhq_gather_shifted_offset_z(p0,p1,p2) ({ __typeof(p1) __p1 = (p1); \
- _Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_z_s16 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
- int (*)[__ARM_mve_type_int16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_z_s32 (__ARM_mve_coerce_s16_ptr(p0, int16_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint16x8_t]: __arm_vldrhq_gather_shifted_offset_z_u16 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint16x8_t), p2), \
- int (*)[__ARM_mve_type_uint16_t_ptr][__ARM_mve_type_uint32x4_t]: __arm_vldrhq_gather_shifted_offset_z_u32 (__ARM_mve_coerce_u16_ptr(p0, uint16_t *), __ARM_mve_coerce(__p1, uint32x4_t), p2));})
-
-#define __arm_vldrwq_gather_shifted_offset(p0,p1) ({ __typeof(p0) __p0 = (p0); \
- _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \
- int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vldrwq_gather_shifted_offset_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1), \
- int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vldrwq_gather_shifted_offset_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1));})
-
-#define __arm_vldrwq_gather_shifted_offset_z(p0,p1,p2) ({ __typeof(p0) __p0 = (p0); \
- _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \
- int (*)[__ARM_mve_type_int32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_s32 (__ARM_mve_coerce_s32_ptr(__p0, int32_t *), p1, p2), \
- int (*)[__ARM_mve_type_uint32_t_ptr]: __arm_vldrwq_gather_shifted_offset_z_u32 (__ARM_mve_coerce_u32_ptr(__p0, uint32_t *), p1, p2));})
-
#define __arm_vst2q(p0,p1) ({ __typeof(p1) __p1 = (p1); \
_Generic( (int (*)[__ARM_mve_typeid(p0)][__ARM_mve_typeid(__p1)])0, \
int (*)[__ARM_mve_type_int8_t_ptr][__ARM_mve_type_int8x16x2_t]: __arm_vst2q_s8 (__ARM_mve_coerce_s8_ptr(p0, int8_t *), __ARM_mve_coerce(__p1, int8x16x2_t)), \
@@ -2108,14 +1750,6 @@ extern void *__ARM_undef;
#endif /* MVE Integer. */
-#define __arm_vldrdq_gather_shifted_offset(p0,p1) ( _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
- int (*)[__ARM_mve_type_int64_t_ptr]: __arm_vldrdq_gather_shifted_offset_s64 (__ARM_mve_coerce_s64_ptr(p0, int64_t *), p1), \
- int (*)[__ARM_mve_type_uint64_t_ptr]: __arm_vldrdq_gather_shifted_offset_u64 (__ARM_mve_coerce_u64_ptr(p0, uint64_t *), p1)))
-
-#define __arm_vldrdq_gather_shifted_offset_z(p0,p1,p2) ( _Generic( (int (*)[__ARM_mve_typeid(p0)])0, \
- int (*)[__ARM_mve_type_int64_t_ptr]: __arm_vldrdq_gather_shifted_offset_z_s64 (__ARM_mve_coerce_s64_ptr(p0, int64_t *), p1, p2), \
- int (*)[__ARM_mve_type_uint64_t_ptr]: __arm_vldrdq_gather_shifted_offset_z_u64 (__ARM_mve_coerce_u64_ptr(p0, uint64_t *), p1, p2)))
-
#endif /* __cplusplus */
#endif /* __ARM_FEATURE_MVE */
#endif /* _GCC_ARM_MVE_H. */
@@ -667,28 +667,12 @@ VAR1 (LDRGBS, vldrwq_gather_base_s, v4si)
VAR1 (LDRGBU, vldrwq_gather_base_u, v4si)
VAR1 (LDRGBS_Z, vldrwq_gather_base_z_s, v4si)
VAR1 (LDRGBU_Z, vldrwq_gather_base_z_u, v4si)
-VAR2 (LDRGU_Z, vldrhq_gather_shifted_offset_z_u, v8hi, v4si)
-VAR2 (LDRGU, vldrhq_gather_shifted_offset_u, v8hi, v4si)
-VAR2 (LDRGS_Z, vldrhq_gather_shifted_offset_z_s, v8hi, v4si)
-VAR2 (LDRGS, vldrhq_gather_shifted_offset_s, v8hi, v4si)
VAR1 (LDRGBS, vldrdq_gather_base_s, v2di)
VAR1 (LDRGBS, vldrwq_gather_base_f, v4sf)
VAR1 (LDRGBS_Z, vldrdq_gather_base_z_s, v2di)
VAR1 (LDRGBS_Z, vldrwq_gather_base_z_f, v4sf)
VAR1 (LDRGBU, vldrdq_gather_base_u, v2di)
VAR1 (LDRGBU_Z, vldrdq_gather_base_z_u, v2di)
-VAR1 (LDRGS, vldrdq_gather_shifted_offset_s, v2di)
-VAR1 (LDRGS, vldrhq_gather_shifted_offset_f, v8hf)
-VAR1 (LDRGS, vldrwq_gather_shifted_offset_f, v4sf)
-VAR1 (LDRGS, vldrwq_gather_shifted_offset_s, v4si)
-VAR1 (LDRGS_Z, vldrdq_gather_shifted_offset_z_s, v2di)
-VAR1 (LDRGS_Z, vldrhq_gather_shifted_offset_z_f, v8hf)
-VAR1 (LDRGS_Z, vldrwq_gather_shifted_offset_z_f, v4sf)
-VAR1 (LDRGS_Z, vldrwq_gather_shifted_offset_z_s, v4si)
-VAR1 (LDRGU, vldrdq_gather_shifted_offset_u, v2di)
-VAR1 (LDRGU, vldrwq_gather_shifted_offset_u, v4si)
-VAR1 (LDRGU_Z, vldrdq_gather_shifted_offset_z_u, v2di)
-VAR1 (LDRGU_Z, vldrwq_gather_shifted_offset_z_u, v4si)
VAR1 (LDRGBWBU_Z, vldrwq_gather_base_nowb_z_u, v4si)
VAR1 (LDRGBWBU_Z, vldrdq_gather_base_nowb_z_u, v2di)
VAR1 (LDRGBWBU, vldrwq_gather_base_nowb_u, v4si)
@@ -2539,11 +2539,7 @@ (define_int_attr supf [(VCVTQ_TO_F_S "s") (VCVTQ_TO_F_U "u") (VREV16Q_S "s")
(VMLALDAVAQ_P_S "s") (VMLALDAVAQ_P_U "u")
(VLDRWQGB_S "s")
(VLDRWQGB_U "u")
- (VLDRHQGSO_S "s") (VLDRHQGSO_U "u")
(VLDRDQGB_S "s") (VLDRDQGB_U "u")
- (VLDRDQGSO_S "s")
- (VLDRDQGSO_U "u")
- (VLDRWQGSO_S "s") (VLDRWQGSO_U "u")
(VSTRDQSB_S "s") (VSTRDQSB_U "u")
(VLDRWQGBWB_S "s") (VLDRWQGBWB_U "u") (VLDRDQGBWB_S "s")
(VLDRDQGBWB_U "u") (VADCQ_M_S "s")
@@ -2946,10 +2942,7 @@ (define_int_iterator VSHLLxQ_M_N [VSHLLBQ_M_N_U VSHLLBQ_M_N_S VSHLLTQ_M_N_U VSHL
(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 VLDRWGBQ [VLDRWQGB_S VLDRWQGB_U])
-(define_int_iterator VLDRHGSOQ [VLDRHQGSO_S VLDRHQGSO_U])
(define_int_iterator VLDRDGBQ [VLDRDQGB_S VLDRDQGB_U])
-(define_int_iterator VLDRDGSOQ [VLDRDQGSO_S VLDRDQGSO_U])
-(define_int_iterator VLDRWGSOQ [VLDRWQGSO_S VLDRWQGSO_U])
(define_int_iterator VLDRWGBWBQ [VLDRWQGBWB_S VLDRWQGBWB_U])
(define_int_iterator VLDRDGBWBQ [VLDRDQGBWB_S VLDRDQGBWB_U])
(define_int_iterator VxCIQ [VADCIQ_U VADCIQ_S VSBCIQ_U VSBCIQ_S])
@@ -3566,56 +3566,88 @@ (define_insn "mve_vldrwq_gather_base_z_<supf>v4si"
[(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_base_<supf>v4si"))
(set_attr "length" "8")])
+;; Gather loads with shifted offset
+;;
+;; [vldrhq_gather_shifted_offset_s vldrhq_gather_shifted_offset_u]
+;; [vldrhq_gather_shifted_offset_f]
+;; [vldrwq_gather_shifted_offset_s vldrwq_gather_shifted_offset_u]
+;; [vldrwq_gather_shifted_offset_f]
+;; [vldrdq_gather_shifted_offset_s vldrdq_gather_shifted_offset_u]
+;;
+(define_insn "@mve_vldrq_gather_shifted_offset_<mode>"
+ [(set (match_operand:MVE_VLD_ST_scatter_shifted 0 "s_register_operand" "=&w")
+ (unspec:MVE_VLD_ST_scatter_shifted
+ [(match_operand:SI 1 "register_operand" "r")
+ (match_operand:<MVE_scatter_offset> 2 "s_register_operand" "w")
+ (mem:BLK (scratch))]
+ VLDRGSOQ))
+ ]
+ "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+ || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+ "vldr<MVE_elem_ch>.<MVE_u_elem>\t%q0, [%1, %q2, uxtw #<MVE_scatter_shift>]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrq_gather_shifted_offset_<mode>"))
+ (set_attr "length" "4")])
+
+;; Extending gather loads with shifted offset
;;
;; [vldrhq_gather_shifted_offset_s vldrhq_gather_shifted_offset_u]
;;
-(define_insn "mve_vldrhq_gather_shifted_offset_<supf><mode>"
- [(set (match_operand:MVE_5 0 "s_register_operand" "=&w")
- (unspec:MVE_5 [(match_operand:<MVE_H_ELEM> 1 "memory_operand" "Us")
- (match_operand:MVE_5 2 "s_register_operand" "w")]
- VLDRHGSOQ))
+(define_insn "@mve_vldrq_gather_shifted_offset_extend_v4si<US>"
+ [(set (match_operand:V4SI 0 "s_register_operand" "=&w")
+ (SE:V4SI
+ (unspec:V4HI
+ [(match_operand:SI 1 "register_operand" "r")
+ (match_operand:V4SI 2 "s_register_operand" "w")
+ (mem:BLK (scratch))]
+ VLDRGSOQ_EXT)))
]
"TARGET_HAVE_MVE"
-{
- rtx ops[3];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- if (!strcmp ("<supf>","s") && <V_sz_elem> == 16)
- output_asm_insn ("vldrh.u16\t%q0, [%m1, %q2, uxtw #1]",ops);
- else
- output_asm_insn ("vldrh.<supf><V_sz_elem>\t%q0, [%m1, %q2, uxtw #1]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_gather_shifted_offset_<supf><mode>"))
+ "vldrh.<US>32\t%q0, [%1, %q2, uxtw #1]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrq_gather_shifted_offset_extend_v4si<US>"))
(set_attr "length" "4")])
+;; Predicated gather loads with shifted offset
;;
;; [vldrhq_gather_shifted_offset_z_s vldrhq_gather_shited_offset_z_u]
+;; [vldrhq_gather_shifted_offset_z_f]
+;; [vldrwq_gather_shifted_offset_z_s vldrwq_gather_shifted_offset_z_u]
+;; [vldrwq_gather_shifted_offset_z_f]
+;; [vldrdq_gather_shifted_offset_z_s vldrdq_gather_shifted_offset_z_u]
;;
-(define_insn "mve_vldrhq_gather_shifted_offset_z_<supf><mode>"
- [(set (match_operand:MVE_5 0 "s_register_operand" "=&w")
- (unspec:MVE_5 [(match_operand:<MVE_H_ELEM> 1 "memory_operand" "Us")
- (match_operand:MVE_5 2 "s_register_operand" "w")
- (match_operand:<MVE_VPRED> 3 "vpr_register_operand" "Up")
- ]VLDRHGSOQ))
+(define_insn "@mve_vldrq_gather_shifted_offset_z_<mode>"
+ [(set (match_operand:MVE_VLD_ST_scatter_shifted 0 "s_register_operand" "=&w")
+ (unspec:MVE_VLD_ST_scatter_shifted
+ [(match_operand:SI 1 "register_operand" "r")
+ (match_operand:<MVE_scatter_offset> 2 "s_register_operand" "w")
+ (match_operand:<MVE_VPRED> 3 "vpr_register_operand" "Up")
+ (mem:BLK (scratch))]
+ VLDRGSOQ_Z))
]
- "TARGET_HAVE_MVE"
-{
- rtx ops[4];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- ops[3] = operands[3];
- if (!strcmp ("<supf>","s") && <V_sz_elem> == 16)
- output_asm_insn ("vpst\n\tvldrht.u16\t%q0, [%m1, %q2, uxtw #1]",ops);
- else
- output_asm_insn ("vpst\n\tvldrht.<supf><V_sz_elem>\t%q0, [%m1, %q2, uxtw #1]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_gather_shifted_offset_<supf><mode>"))
+ "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
+ || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode))"
+ "vpst\n\tvldr<MVE_elem_ch>t.<MVE_u_elem>\t%q0, [%1, %q2, uxtw #<MVE_scatter_shift>]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrq_gather_shifted_offset_<mode>"))
(set_attr "length" "8")])
+;; Predicated extending gather loads with shifted offset
+;;
+;; [vldrhq_gather_shifted_offset_z_s vldrhq_gather_shifted_offset_z_u]
+;;
+(define_insn "@mve_vldrq_gather_shifted_offset_z_extend_v4si<US>"
+ [(set (match_operand:V4SI 0 "s_register_operand" "=&w")
+ (SE:V4SI
+ (unspec:V4HI
+ [(match_operand:SI 1 "register_operand" "r")
+ (match_operand:V4SI 2 "s_register_operand" "w")
+ (match_operand:V4BI 3 "vpr_register_operand" "Up")
+ (mem:BLK (scratch))]
+ VLDRGSOQ_EXT_Z)))
+ ]
+ "TARGET_HAVE_MVE"
+ "vpst\n\tvldrht.<US>32\t%q0, [%1, %q2, uxtw #1]"
+ [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrq_gather_shifted_offset_extend_v4si<US>"))
+ (set_attr "length" "4")])
+
;;
;; [vldrdq_gather_base_s vldrdq_gather_base_u]
;;
@@ -3659,93 +3691,6 @@ (define_insn "mve_vldrdq_gather_base_z_<supf>v2di"
[(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrdq_gather_base_<supf>v2di"))
(set_attr "length" "8")])
-;;
-;; [vldrdq_gather_shifted_offset_s vldrdq_gather_shifted_offset_u]
-;;
-(define_insn "mve_vldrdq_gather_shifted_offset_<supf>v2di"
- [(set (match_operand:V2DI 0 "s_register_operand" "=&w")
- (unspec:V2DI [(match_operand:V2DI 1 "memory_operand" "Us")
- (match_operand:V2DI 2 "s_register_operand" "w")]
- VLDRDGSOQ))
- ]
- "TARGET_HAVE_MVE"
-{
- rtx ops[3];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- output_asm_insn ("vldrd.u64\t%q0, [%m1, %q2, uxtw #3]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrdq_gather_shifted_offset_<supf>v2di"))
- (set_attr "length" "4")])
-
-;;
-;; [vldrdq_gather_shifted_offset_z_s vldrdq_gather_shifted_offset_z_u]
-;;
-(define_insn "mve_vldrdq_gather_shifted_offset_z_<supf>v2di"
- [(set (match_operand:V2DI 0 "s_register_operand" "=&w")
- (unspec:V2DI [(match_operand:V2DI 1 "memory_operand" "Us")
- (match_operand:V2DI 2 "s_register_operand" "w")
- (match_operand:V2QI 3 "vpr_register_operand" "Up")]
- VLDRDGSOQ))
- ]
- "TARGET_HAVE_MVE"
-{
- rtx ops[3];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- output_asm_insn ("vpst\n\tvldrdt.u64\t%q0, [%m1, %q2, uxtw #3]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrdq_gather_shifted_offset_<supf>v2di"))
- (set_attr "length" "8")])
-
-;;
-;; [vldrhq_gather_shifted_offset_f]
-;;
-(define_insn "mve_vldrhq_gather_shifted_offset_fv8hf"
- [(set (match_operand:V8HF 0 "s_register_operand" "=&w")
- (unspec:V8HF [(match_operand:V8HI 1 "memory_operand" "Us")
- (match_operand:V8HI 2 "s_register_operand" "w")]
- VLDRHQGSO_F))
- ]
- "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
- rtx ops[3];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- output_asm_insn ("vldrh.f16\t%q0, [%m1, %q2, uxtw #1]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_gather_shifted_offset_fv8hf"))
- (set_attr "length" "4")])
-
-;;
-;; [vldrhq_gather_shifted_offset_z_f]
-;;
-(define_insn "mve_vldrhq_gather_shifted_offset_z_fv8hf"
- [(set (match_operand:V8HF 0 "s_register_operand" "=&w")
- (unspec:V8HF [(match_operand:V8HI 1 "memory_operand" "Us")
- (match_operand:V8HI 2 "s_register_operand" "w")
- (match_operand:V8BI 3 "vpr_register_operand" "Up")]
- VLDRHQGSO_F))
- ]
- "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
- rtx ops[4];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- ops[3] = operands[3];
- output_asm_insn ("vpst\n\tvldrht.f16\t%q0, [%m1, %q2, uxtw #1]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrhq_gather_shifted_offset_fv8hf"))
- (set_attr "length" "8")])
-
;;
;; [vldrwq_gather_base_f]
;;
@@ -3789,94 +3734,6 @@ (define_insn "mve_vldrwq_gather_base_z_fv4sf"
[(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_base_fv4sf"))
(set_attr "length" "8")])
-;;
-;; [vldrwq_gather_shifted_offset_f]
-;;
-(define_insn "mve_vldrwq_gather_shifted_offset_fv4sf"
- [(set (match_operand:V4SF 0 "s_register_operand" "=&w")
- (unspec:V4SF [(match_operand:V4SI 1 "memory_operand" "Us")
- (match_operand:V4SI 2 "s_register_operand" "w")]
- VLDRWQGSO_F))
- ]
- "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
- rtx ops[3];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- output_asm_insn ("vldrw.u32\t%q0, [%m1, %q2, uxtw #2]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_shifted_offset_fv4sf"))
- (set_attr "length" "4")])
-
-;;
-;; [vldrwq_gather_shifted_offset_s vldrwq_gather_shifted_offset_u]
-;;
-(define_insn "mve_vldrwq_gather_shifted_offset_<supf>v4si"
- [(set (match_operand:V4SI 0 "s_register_operand" "=&w")
- (unspec:V4SI [(match_operand:V4SI 1 "memory_operand" "Us")
- (match_operand:V4SI 2 "s_register_operand" "w")]
- VLDRWGSOQ))
- ]
- "TARGET_HAVE_MVE"
-{
- rtx ops[3];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- output_asm_insn ("vldrw.u32\t%q0, [%m1, %q2, uxtw #2]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_shifted_offset_<supf>v4si"))
- (set_attr "length" "4")])
-
-;;
-;; [vldrwq_gather_shifted_offset_z_f]
-;;
-(define_insn "mve_vldrwq_gather_shifted_offset_z_fv4sf"
- [(set (match_operand:V4SF 0 "s_register_operand" "=&w")
- (unspec:V4SF [(match_operand:V4SI 1 "memory_operand" "Us")
- (match_operand:V4SI 2 "s_register_operand" "w")
- (match_operand:V4BI 3 "vpr_register_operand" "Up")]
- VLDRWQGSO_F))
- ]
- "TARGET_HAVE_MVE && TARGET_HAVE_MVE_FLOAT"
-{
- rtx ops[4];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- ops[3] = operands[3];
- output_asm_insn ("vpst\n\tvldrwt.u32\t%q0, [%m1, %q2, uxtw #2]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_shifted_offset_fv4sf"))
- (set_attr "length" "8")])
-
-;;
-;; [vldrwq_gather_shifted_offset_z_s vldrwq_gather_shifted_offset_z_u]
-;;
-(define_insn "mve_vldrwq_gather_shifted_offset_z_<supf>v4si"
- [(set (match_operand:V4SI 0 "s_register_operand" "=&w")
- (unspec:V4SI [(match_operand:V4SI 1 "memory_operand" "Us")
- (match_operand:V4SI 2 "s_register_operand" "w")
- (match_operand:V4BI 3 "vpr_register_operand" "Up")]
- VLDRWGSOQ))
- ]
- "TARGET_HAVE_MVE"
-{
- rtx ops[4];
- ops[0] = operands[0];
- ops[1] = operands[1];
- ops[2] = operands[2];
- ops[3] = operands[3];
- output_asm_insn ("vpst\n\tvldrwt.u32\t%q0, [%m1, %q2, uxtw #2]",ops);
- return "";
-}
- [(set (attr "mve_unpredicated_insn") (symbol_ref "CODE_FOR_mve_vldrwq_gather_shifted_offset_<supf>v4si"))
- (set_attr "length" "8")])
-
;; Vector scatter stores with shifted offset
;;
;; [vstrhq_scatter_shifted_offset_s vstrhq_scatter_shifted_offset_u]
@@ -1162,17 +1162,13 @@ (define_c_enum "unspec" [
VLDRGOQ_EXT_Z
VLDRWQGB_S
VLDRWQGB_U
- VLDRHQGSO_S
- VLDRHQGSO_U
+ VLDRGSOQ
+ VLDRGSOQ_Z
+ VLDRGSOQ_EXT
+ VLDRGSOQ_EXT_Z
VLDRDQGB_S
VLDRDQGB_U
- VLDRDQGSO_S
- VLDRDQGSO_U
- VLDRHQGSO_F
VLDRWQGB_F
- VLDRWQGSO_F
- VLDRWQGSO_S
- VLDRWQGSO_U
VSTRQ
VSTRQ_P
VSTRQ_TRUNC