Skip to content

Fix strict aliasing in SSE packed vector stores #356

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Chuck Walbourn (walbourn) merged 2 commits into from
Oct 16, 2026
Merged

Fix strict aliasing in SSE packed vector stores #356

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a8a1517
Fix SSE packed vector stores for GCC strict-aliasing optimizer
vera-rubin Oct 13, 2026
91c0991
Keep the _mm_store_ss pattern for MSVC in SSE packed vector stores
vera-rubin Oct 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
60 changes: 60 additions & 0 deletions Inc/DirectXPackedVector.inl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2040,7 +2040,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreColor(XMCOLOR* pDestinatio
// Mash to bytes
vInt = _mm_packus_epi16(vInt, vInt);
// Store the color
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->c), _mm_castsi128_ps(vInt));
#else
pDestination->c = static_cast<uint32_t>(_mm_cvtsi128_si32(vInt));
#endif
#endif
}

Expand All @@ -2050,7 +2054,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreHalf2(XMHALF2* pDestinatio
assert(pDestination);
#if defined(_XM_F16C_INTRINSICS_) && !defined(_XM_NO_INTRINSICS_)
__m128i V1 = _mm_cvtps_ph(V, _MM_FROUND_TO_NEAREST_INT);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(pDestination), _mm_castsi128_ps(V1));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(V1));
#endif
#else
pDestination->x = XMConvertFloatToHalf(XMVectorGetX(V));
pDestination->y = XMConvertFloatToHalf(XMVectorGetY(V));
Expand Down Expand Up @@ -2086,7 +2094,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreShortN2(XMSHORTN2* pDestin
vResult = _mm_mul_ps(vResult, g_ShortMax);
__m128i vResulti = _mm_cvtps_epi32(vResult);
vResulti = _mm_packs_epi32(vResulti, vResulti);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->x), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -2119,7 +2131,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreShort2(XMSHORT2* pDestinat
__m128i vInt = _mm_cvtps_epi32(vResult);
// Pack the ints into shorts
vInt = _mm_packs_epi32(vInt, vInt);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->x), _mm_castsi128_ps(vInt));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vInt));
#endif
#endif
}

Expand Down Expand Up @@ -2857,7 +2873,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreXDecN4(XMXDECN4* pDestinat
vResulti = _mm_or_si128(vResulti, _mm_castps_si128(vResult));
vResult = XM_PERMUTE_PS(vResult, _MM_SHUFFLE(0, 3, 2, 1));
vResulti = _mm_or_si128(vResulti, _mm_castps_si128(vResult));
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -2933,7 +2953,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreXDec4(XMXDEC4* pDestinatio
vResulti2 = _mm_add_epi32(vResulti2, vResulti2);
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3005,7 +3029,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreUDecN4(XMUDECN4* pDestinat
vResulti2 = _mm_add_epi32(vResulti2, vResulti2);
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3070,7 +3098,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreUDecN4_XR(XMUDECN4* pDesti
vResulti2 = _mm_add_epi32(vResulti2, vResulti2);
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3129,7 +3161,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreUDec4(XMUDEC4* pDestinatio
vResulti2 = _mm_add_epi32(vResulti2, vResulti2);
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3200,7 +3236,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreDecN4(XMDECN4* pDestinatio
vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1));
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3253,7 +3293,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreDec4(XMDEC4* pDestination,
vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1));
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3316,7 +3360,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreUByteN4(XMUBYTEN4* pDestin
vResulti2 = _mm_add_epi32(vResulti2, vResulti2);
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3366,7 +3414,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreUByte4(XMUBYTE4* pDestinat
vResulti2 = _mm_add_epi32(vResulti2, vResulti2);
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3416,7 +3468,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreByteN4(XMBYTEN4* pDestinat
vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1));
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down Expand Up @@ -3464,7 +3520,11 @@ _Use_decl_annotations_ inline void XM_CALLCONV XMStoreByte4(XMBYTE4* pDestinatio
vResulti2 = _mm_shuffle_epi32(vResulti, _MM_SHUFFLE(1, 1, 1, 1));
// i = x|y|z|w
vResulti = _mm_or_si128(vResulti, vResulti2);
#ifdef _MSC_VER
_mm_store_ss(reinterpret_cast<float*>(&pDestination->v), _mm_castsi128_ps(vResulti));
#else
pDestination->v = static_cast<uint32_t>(_mm_cvtsi128_si32(vResulti));
#endif
#endif
}

Expand Down
Loading