Scarfski/bx
1
0
Fork 0
mirror of https://github.com/bkaradzic/bx.git synced 2026-06-08 03:03:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed build. (#387)

Browse Source
This commit is contained in:
Branimir Karadžić
2026-05-03 12:56:46 -07:00
committed by GitHub
parent 22f7b051b9
commit 6ce72eac63
2 changed files with 28 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
include/bx/inline/simd128_neon.inl
View File

@@ -287,8 +287,13 @@ namespace bx
template<>
BX_SIMD_FORCE_INLINE simd128_neon_t simd128_splat(double _a)
{
#if BX_ARCH_64BIT
const float64x2_t tmp = vdupq_n_f64(_a);
return vreinterpretq_f32_f64(tmp);
#else
const float val = float(_a);
return vdupq_n_f32(val);
#endif // BX_ARCH_64BIT
}
template<>
@@ -392,6 +397,12 @@ namespace bx
return simd_f32_rcp_ni(_a);
}
template<>
BX_SIMD_FORCE_INLINE simd128_neon_t simd128_f32_rsqrt_est(simd128_neon_t _a)
{
return vrsqrteq_f32(_a);
}
template<>
BX_SIMD_FORCE_INLINE simd128_neon_t simd128_f32_sqrt(simd128_neon_t _a)
{
@@ -400,13 +411,7 @@ namespace bx
#else
const simd128_neon_t rsqrt = simd128_f32_rsqrt_est<simd128_neon_t>(_a);
return simd128_f32_mul<simd128_neon_t>(_a, rsqrt);
#endif
}
template<>
BX_SIMD_FORCE_INLINE simd128_neon_t simd128_f32_rsqrt_est(simd128_neon_t _a)
{
return vrsqrteq_f32(_a);
#endif // BX_ARCH_64BIT
}
template<>
@@ -842,19 +847,27 @@ namespace bx
template<>
BX_SIMD_FORCE_INLINE simd128_neon_t simd128_x8_shuffle(simd128_neon_t _a, simd128_neon_t _indices)
{
#if BX_ARCH_64BIT
const uint8x16_t a = vreinterpretq_u8_f32(_a);
const uint8x16_t indices = vreinterpretq_u8_f32(_indices);
const uint8x16_t result = vqtbl1q_u8(a, indices);
return vreinterpretq_f32_u8(result);
#else
return simd_x8_shuffle_ni(_a, _indices);
#endif // BX_ARCH_64BIT
}
template<>
BX_SIMD_FORCE_INLINE simd128_neon_t simd128_x8_shuffle(simd128_neon_t _a, simd128_neon_t _b, simd128_neon_t _indices)
{
#if BX_ARCH_64BIT
const uint8x16x2_t tbl = { { vreinterpretq_u8_f32(_a), vreinterpretq_u8_f32(_b) } };
const uint8x16_t indices = vreinterpretq_u8_f32(_indices);
const uint8x16_t result = vqtbl2q_u8(tbl, indices);
return vreinterpretq_f32_u8(result);
#else
return simd_x8_shuffle_ni(_a, _b, _indices);
#endif // BX_ARCH_64BIT
}
template<>

16
include/bx/inline/simd256_ref.inl
View File

@@ -258,7 +258,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_f32_madd(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _c)
inline BX_CONST_FUNC simd256_ref_t simd256_f32_madd(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _c)
{
#if BX_SIMD_LANGEXT
const simd256_f32_langext_t a = bitCast<simd256_f32_langext_t>(_a);
@@ -274,7 +274,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_f32_msub(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _c)
inline BX_CONST_FUNC simd256_ref_t simd256_f32_msub(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _c)
{
#if BX_SIMD_LANGEXT
const simd256_f32_langext_t a = bitCast<simd256_f32_langext_t>(_a);
@@ -290,7 +290,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_f32_nmsub(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _c)
inline BX_CONST_FUNC simd256_ref_t simd256_f32_nmsub(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _c)
{
#if BX_SIMD_LANGEXT
const simd256_f32_langext_t a = bitCast<simd256_f32_langext_t>(_a);
@@ -306,7 +306,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_f32_neg(simd256_ref_t _a)
inline BX_CONST_FUNC simd256_ref_t simd256_f32_neg(simd256_ref_t _a)
{
#if BX_SIMD_LANGEXT
const simd256_f32_langext_t a = bitCast<simd256_f32_langext_t>(_a);
@@ -351,7 +351,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_f32_lerp(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _s)
inline BX_CONST_FUNC simd256_ref_t simd256_f32_lerp(simd256_ref_t _a, simd256_ref_t _b, simd256_ref_t _s)
{
#if BX_SIMD_LANGEXT
const simd256_f32_langext_t a = bitCast<simd256_f32_langext_t>(_a);
@@ -450,7 +450,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_f32_cmpneq(simd256_ref_t _a, simd256_ref_t _b)
inline BX_CONST_FUNC simd256_ref_t simd256_f32_cmpneq(simd256_ref_t _a, simd256_ref_t _b)
{
#if BX_SIMD_LANGEXT
const simd256_f32_langext_t a = bitCast<simd256_f32_langext_t>(_a);
@@ -532,7 +532,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_i32_neg(simd256_ref_t _a)
inline BX_CONST_FUNC simd256_ref_t simd256_i32_neg(simd256_ref_t _a)
{
#if BX_SIMD_LANGEXT
const simd256_i32_langext_t a = bitCast<simd256_i32_langext_t>(_a);
@@ -813,7 +813,7 @@ namespace bx
}
template<>
inline BX_CONSTEXPR_FUNC simd256_ref_t simd256_andc(simd256_ref_t _a, simd256_ref_t _b)
inline BX_CONST_FUNC simd256_ref_t simd256_andc(simd256_ref_t _a, simd256_ref_t _b)
{
#if BX_SIMD_LANGEXT
const simd256_u32_langext_t a = bitCast<simd256_u32_langext_t>(_a);