aarch64: Use memcpy to copy vector tables in vst3[q] intrinsics
Use __builtin_memcpy to copy vector structures instead of building a new opaque structure one vector at a time in each of the vst3[q] Neon intrinsics in arm_neon.h. This simplifies the header file and also improves code generation - superfluous move instructions were emitted for every register extraction/set in this additional structure. Add new code generation tests to verify that superfluous move instructions are no longer generated for the vst3q intrinsics. gcc/ChangeLog: 2021-07-21 Jonathan Wright <jonathan.wright@arm.com> * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy instead of constructing __builtin_aarch64_simd_ci one vector at a time. (vst3_u64): Likewise. (vst3_f64): Likewise. (vst3_s8): Likewise. (vst3_p8): Likewise. (vst3_s16): Likewise. (vst3_p16): Likewise. (vst3_s32): Likewise. (vst3_u8): Likewise. (vst3_u16): Likewise. (vst3_u32): Likewise. (vst3_f16): Likewise. (vst3_f32): Likewise. (vst3_p64): Likewise. (vst3q_s8): Likewise. (vst3q_p8): Likewise. (vst3q_s16): Likewise. (vst3q_p16): Likewise. (vst3q_s32): Likewise. (vst3q_s64): Likewise. (vst3q_u8): Likewise. (vst3q_u16): Likewise. (vst3q_u32): Likewise. (vst3q_u64): Likewise. (vst3q_f16): Likewise. (vst3q_f32): Likewise. (vst3q_f64): Likewise. (vst3q_p64): Likewise. gcc/testsuite/ChangeLog: * gcc.target/aarch64/vector_structure_intrinsics.c: Add new tests.
This commit is contained in:
Jonathan Wright
parent
e8de7edde6
commit
95509ee2c1
|
|
@ -27543,9 +27543,7 @@ vst3_s64 (int64_t * __a, int64x1x3_t __val)
|
|||
__temp.val[0] = vcombine_s64 (__val.val[0], vcreate_s64 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[1] = vcombine_s64 (__val.val[1], vcreate_s64 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[2] = vcombine_s64 (__val.val[2], vcreate_s64 (__AARCH64_INT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3di ((__builtin_aarch64_simd_di *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27558,9 +27556,7 @@ vst3_u64 (uint64_t * __a, uint64x1x3_t __val)
|
|||
__temp.val[0] = vcombine_u64 (__val.val[0], vcreate_u64 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_u64 (__val.val[1], vcreate_u64 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_u64 (__val.val[2], vcreate_u64 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3di ((__builtin_aarch64_simd_di *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27573,9 +27569,7 @@ vst3_f64 (float64_t * __a, float64x1x3_t __val)
|
|||
__temp.val[0] = vcombine_f64 (__val.val[0], vcreate_f64 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_f64 (__val.val[1], vcreate_f64 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_f64 (__val.val[2], vcreate_f64 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv2df (__o, (float64x2_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2df (__o, (float64x2_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2df (__o, (float64x2_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3df ((__builtin_aarch64_simd_df *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27588,9 +27582,7 @@ vst3_s8 (int8_t * __a, int8x8x3_t __val)
|
|||
__temp.val[0] = vcombine_s8 (__val.val[0], vcreate_s8 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[1] = vcombine_s8 (__val.val[1], vcreate_s8 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[2] = vcombine_s8 (__val.val[2], vcreate_s8 (__AARCH64_INT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v8qi ((__builtin_aarch64_simd_qi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27603,9 +27595,7 @@ vst3_p8 (poly8_t * __a, poly8x8x3_t __val)
|
|||
__temp.val[0] = vcombine_p8 (__val.val[0], vcreate_p8 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_p8 (__val.val[1], vcreate_p8 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_p8 (__val.val[2], vcreate_p8 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v8qi ((__builtin_aarch64_simd_qi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27618,9 +27608,7 @@ vst3_s16 (int16_t * __a, int16x4x3_t __val)
|
|||
__temp.val[0] = vcombine_s16 (__val.val[0], vcreate_s16 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[1] = vcombine_s16 (__val.val[1], vcreate_s16 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[2] = vcombine_s16 (__val.val[2], vcreate_s16 (__AARCH64_INT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v4hi ((__builtin_aarch64_simd_hi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27633,9 +27621,7 @@ vst3_p16 (poly16_t * __a, poly16x4x3_t __val)
|
|||
__temp.val[0] = vcombine_p16 (__val.val[0], vcreate_p16 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_p16 (__val.val[1], vcreate_p16 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_p16 (__val.val[2], vcreate_p16 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v4hi ((__builtin_aarch64_simd_hi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27648,9 +27634,7 @@ vst3_s32 (int32_t * __a, int32x2x3_t __val)
|
|||
__temp.val[0] = vcombine_s32 (__val.val[0], vcreate_s32 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[1] = vcombine_s32 (__val.val[1], vcreate_s32 (__AARCH64_INT64_C (0)));
|
||||
__temp.val[2] = vcombine_s32 (__val.val[2], vcreate_s32 (__AARCH64_INT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v2si ((__builtin_aarch64_simd_si *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27663,9 +27647,7 @@ vst3_u8 (uint8_t * __a, uint8x8x3_t __val)
|
|||
__temp.val[0] = vcombine_u8 (__val.val[0], vcreate_u8 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_u8 (__val.val[1], vcreate_u8 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_u8 (__val.val[2], vcreate_u8 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v8qi ((__builtin_aarch64_simd_qi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27678,9 +27660,7 @@ vst3_u16 (uint16_t * __a, uint16x4x3_t __val)
|
|||
__temp.val[0] = vcombine_u16 (__val.val[0], vcreate_u16 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_u16 (__val.val[1], vcreate_u16 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_u16 (__val.val[2], vcreate_u16 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v4hi ((__builtin_aarch64_simd_hi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27693,9 +27673,7 @@ vst3_u32 (uint32_t * __a, uint32x2x3_t __val)
|
|||
__temp.val[0] = vcombine_u32 (__val.val[0], vcreate_u32 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_u32 (__val.val[1], vcreate_u32 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_u32 (__val.val[2], vcreate_u32 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v2si ((__builtin_aarch64_simd_si *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27708,9 +27686,7 @@ vst3_f16 (float16_t * __a, float16x4x3_t __val)
|
|||
__temp.val[0] = vcombine_f16 (__val.val[0], vcreate_f16 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_f16 (__val.val[1], vcreate_f16 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_f16 (__val.val[2], vcreate_f16 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv8hf (__o, (float16x8_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hf (__o, (float16x8_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hf (__o, (float16x8_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v4hf ((__builtin_aarch64_simd_hf *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27723,9 +27699,7 @@ vst3_f32 (float32_t * __a, float32x2x3_t __val)
|
|||
__temp.val[0] = vcombine_f32 (__val.val[0], vcreate_f32 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_f32 (__val.val[1], vcreate_f32 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_f32 (__val.val[2], vcreate_f32 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv4sf (__o, (float32x4_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv4sf (__o, (float32x4_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv4sf (__o, (float32x4_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3v2sf ((__builtin_aarch64_simd_sf *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27738,12 +27712,7 @@ vst3_p64 (poly64_t * __a, poly64x1x3_t __val)
|
|||
__temp.val[0] = vcombine_p64 (__val.val[0], vcreate_p64 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[1] = vcombine_p64 (__val.val[1], vcreate_p64 (__AARCH64_UINT64_C (0)));
|
||||
__temp.val[2] = vcombine_p64 (__val.val[2], vcreate_p64 (__AARCH64_UINT64_C (0)));
|
||||
__o = __builtin_aarch64_set_qregciv2di_ssps (__o,
|
||||
(poly64x2_t) __temp.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2di_ssps (__o,
|
||||
(poly64x2_t) __temp.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2di_ssps (__o,
|
||||
(poly64x2_t) __temp.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__temp, sizeof (__temp));
|
||||
__builtin_aarch64_st3di ((__builtin_aarch64_simd_di *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27752,9 +27721,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_s8 (int8_t * __a, int8x16x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v16qi ((__builtin_aarch64_simd_qi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27763,9 +27730,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_p8 (poly8_t * __a, poly8x16x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v16qi ((__builtin_aarch64_simd_qi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27774,9 +27739,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_s16 (int16_t * __a, int16x8x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v8hi ((__builtin_aarch64_simd_hi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27785,9 +27748,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_p16 (poly16_t * __a, poly16x8x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v8hi ((__builtin_aarch64_simd_hi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27796,9 +27757,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_s32 (int32_t * __a, int32x4x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v4si ((__builtin_aarch64_simd_si *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27807,9 +27766,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_s64 (int64_t * __a, int64x2x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v2di ((__builtin_aarch64_simd_di *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27818,9 +27775,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_u8 (uint8_t * __a, uint8x16x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv16qi (__o, (int8x16_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v16qi ((__builtin_aarch64_simd_qi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27829,9 +27784,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_u16 (uint16_t * __a, uint16x8x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hi (__o, (int16x8_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v8hi ((__builtin_aarch64_simd_hi *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27840,9 +27793,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_u32 (uint32_t * __a, uint32x4x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv4si (__o, (int32x4_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v4si ((__builtin_aarch64_simd_si *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27851,9 +27802,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_u64 (uint64_t * __a, uint64x2x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2di (__o, (int64x2_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v2di ((__builtin_aarch64_simd_di *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27862,9 +27811,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_f16 (float16_t * __a, float16x8x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv8hf (__o, (float16x8_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv8hf (__o, (float16x8_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv8hf (__o, (float16x8_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v8hf ((__builtin_aarch64_simd_hf *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27873,9 +27820,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_f32 (float32_t * __a, float32x4x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv4sf (__o, (float32x4_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv4sf (__o, (float32x4_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv4sf (__o, (float32x4_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v4sf ((__builtin_aarch64_simd_sf *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27884,9 +27829,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_f64 (float64_t * __a, float64x2x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv2df (__o, (float64x2_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2df (__o, (float64x2_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2df (__o, (float64x2_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v2df ((__builtin_aarch64_simd_df *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
@ -27895,12 +27838,7 @@ __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
|
|||
vst3q_p64 (poly64_t * __a, poly64x2x3_t __val)
|
||||
{
|
||||
__builtin_aarch64_simd_ci __o;
|
||||
__o = __builtin_aarch64_set_qregciv2di_ssps (__o,
|
||||
(poly64x2_t) __val.val[0], 0);
|
||||
__o = __builtin_aarch64_set_qregciv2di_ssps (__o,
|
||||
(poly64x2_t) __val.val[1], 1);
|
||||
__o = __builtin_aarch64_set_qregciv2di_ssps (__o,
|
||||
(poly64x2_t) __val.val[2], 2);
|
||||
__builtin_memcpy (&__o, &__val, sizeof (__val));
|
||||
__builtin_aarch64_st3v2di ((__builtin_aarch64_simd_di *) __a, __o);
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -103,8 +103,30 @@ TEST_STX (vst4q, uint64x2x4_t, uint64_t*, u64);
|
|||
TEST_STX (vst4q, float64x2x4_t, float64_t*, f64);
|
||||
TEST_STX (vst4q, poly64x2x4_t, poly64_t*, p64);
|
||||
|
||||
#define TEST_ST3(name, tbltype, ptrtype, ts) \
|
||||
void test_ ## name ## _ ## ts (ptrtype a, int8x8_t dummy, tbltype b) \
|
||||
{ \
|
||||
name ## _ ## ts (a, b); \
|
||||
}
|
||||
|
||||
TEST_ST3 (vst3q, int8x16x3_t, int8_t*, s8);
|
||||
TEST_ST3 (vst3q, uint8x16x3_t, uint8_t*, u8);
|
||||
TEST_ST3 (vst3q, poly8x16x3_t, poly8_t*, p8);
|
||||
TEST_ST3 (vst3q, int16x8x3_t, int16_t*, s16);
|
||||
TEST_ST3 (vst3q, uint16x8x3_t, uint16_t*, u16);
|
||||
TEST_ST3 (vst3q, poly16x8x3_t, poly16_t*, p16);
|
||||
TEST_ST3 (vst3q, float16x8x3_t, float16_t*, f16);
|
||||
TEST_ST3 (vst3q, int32x4x3_t, int32_t*, s32);
|
||||
TEST_ST3 (vst3q, uint32x4x3_t, uint32_t*, u32);
|
||||
TEST_ST3 (vst3q, float32x4x3_t, float32_t*, f32);
|
||||
TEST_ST3 (vst3q, int64x2x3_t, int64_t*, s64);
|
||||
TEST_ST3 (vst3q, uint64x2x3_t, uint64_t*, u64);
|
||||
TEST_ST3 (vst3q, float64x2x3_t, float64_t*, f64);
|
||||
TEST_ST3 (vst3q, poly64x2x3_t, poly64_t*, p64);
|
||||
|
||||
/* { dg-final { scan-assembler-not "mov\\t" } } */
|
||||
|
||||
/* { dg-final { scan-assembler-times "tbl\\t" 18} } */
|
||||
/* { dg-final { scan-assembler-times "tbx\\t" 18} } */
|
||||
/* { dg-final { scan-assembler-times "st4\\t" 14} } */
|
||||
/* { dg-final { scan-assembler-times "st3\\t" 14} } */
|
||||
|
|
|
|||
Loading…
Reference in New Issue