diff --git a/pcsx2/GS/GSBlock.cpp b/pcsx2/GS/GSBlock.cpp index b3bd16aa9a..c4e735a6d6 100644 --- a/pcsx2/GS/GSBlock.cpp +++ b/pcsx2/GS/GSBlock.cpp @@ -16,6 +16,8 @@ #include "PrecompiledHeader.h" #include "GSBlock.h" +MULTI_ISA_UNSHARED_IMPL; + CONSTINIT const GSVector4i GSBlock::m_r16mask(0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15); CONSTINIT const GSVector4i GSBlock::m_r8mask(0, 4, 2, 6, 8, 12, 10, 14, 1, 5, 3, 7, 9, 13, 11, 15); CONSTINIT const GSVector4i GSBlock::m_r4mask(0, 8, 4, 12, 1, 9, 5, 13, 2, 10, 6, 14, 3, 11, 7, 15); diff --git a/pcsx2/GS/GSBlock.h b/pcsx2/GS/GSBlock.h index eb475184bd..d70d47e2b1 100644 --- a/pcsx2/GS/GSBlock.h +++ b/pcsx2/GS/GSBlock.h @@ -18,6 +18,9 @@ #include "GSRegs.h" #include "GSTables.h" #include "GSVector.h" +#include "MultiISA.h" + +MULTI_ISA_UNSHARED_START class GSBlock { @@ -1940,3 +1943,5 @@ public: // TODO: ReadAndExpandBlock4HH_16 }; + +MULTI_ISA_UNSHARED_END diff --git a/pcsx2/GS/GSLocalMemory.h b/pcsx2/GS/GSLocalMemory.h index a6e8e4fea5..c8d07dc82c 100644 --- a/pcsx2/GS/GSLocalMemory.h +++ b/pcsx2/GS/GSLocalMemory.h @@ -17,7 +17,6 @@ #include "GSTables.h" #include "GSVector.h" -#include "GSBlock.h" #include "GSClut.h" #include "MultiISA.h" #include diff --git a/pcsx2/GS/GSLocalMemoryMultiISA.cpp b/pcsx2/GS/GSLocalMemoryMultiISA.cpp index 5cba6b585f..67a0b3e2d0 100644 --- a/pcsx2/GS/GSLocalMemoryMultiISA.cpp +++ b/pcsx2/GS/GSLocalMemoryMultiISA.cpp @@ -16,6 +16,7 @@ #include "PrecompiledHeader.h" #include "GSLocalMemory.h" #include "GS.h" +#include "GSBlock.h" #include "GSExtra.h" class CURRENT_ISA::GSLocalMemoryFunctions diff --git a/pcsx2/GS/GSVector.cpp b/pcsx2/GS/GSVector.cpp index deb00ac09d..dab402d2b1 100644 --- a/pcsx2/GS/GSVector.cpp +++ b/pcsx2/GS/GSVector.cpp @@ -71,8 +71,6 @@ CONSTINIT const GSVector4 GSVector4::m_xc1e00000000fffff = cxpr64(0xc1e00000000f CONSTINIT const GSVector4 GSVector4::m_max = cxpr(FLT_MAX); CONSTINIT const GSVector4 GSVector4::m_min = cxpr(FLT_MIN); -#if _M_SSE >= 0x500 - CONSTINIT const GSVector8 GSVector8::m_half = cxpr(0.5f); CONSTINIT const GSVector8 GSVector8::m_one = cxpr(1.0f); CONSTINIT const GSVector8 GSVector8::m_x7fffffff = cxpr(0x7fffffff); @@ -83,9 +81,6 @@ CONSTINIT const GSVector8 GSVector8::m_xc1e00000000fffff = cxpr64(0xc1e00000000f CONSTINIT const GSVector8 GSVector8::m_max = cxpr(FLT_MAX); CONSTINIT const GSVector8 GSVector8::m_min = cxpr(FLT_MAX); -#endif - -#if _M_SSE >= 0x501 CONSTINIT const GSVector8i GSVector8i::m_xff[33] = { cxpr(0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000), @@ -159,7 +154,6 @@ CONSTINIT const GSVector8i GSVector8i::m_x0f[33] = cxpr(0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x000f0f0f), cxpr(0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f, 0x0f0f0f0f), }; -#endif GSVector4i GSVector4i::fit(int arx, int ary) const { diff --git a/pcsx2/GS/GSVector8.h b/pcsx2/GS/GSVector8.h index 1f60768f39..c0b87b3c1d 100644 --- a/pcsx2/GS/GSVector8.h +++ b/pcsx2/GS/GSVector8.h @@ -15,8 +15,6 @@ #include -#if _M_SSE >= 0x500 - class alignas(32) GSVector8 { struct cxpr_init_tag {}; @@ -53,7 +51,9 @@ public: u16 U16[16]; u32 U32[8]; u64 U64[4]; +#if _M_SSE >= 0x500 __m256 m; +#endif __m128 m0, m1; }; @@ -104,6 +104,8 @@ public: return GSVector8(cxpr_init, x, x, x, x); } +#if _M_SSE >= 0x500 + __forceinline GSVector8(float x0, float y0, float z0, float w0, float x1, float y1, float z1, float w1) { m = _mm256_set_ps(w1, z1, y1, x1, w0, z0, y0, x0); @@ -948,6 +950,5 @@ public: // TODO: v.(x0|y0|z0|w0|x1|y1|z1|w1) // broadcast element #endif -}; - #endif +}; diff --git a/pcsx2/GS/GSVector8i.h b/pcsx2/GS/GSVector8i.h index 3456ef889d..8bff2eef42 100644 --- a/pcsx2/GS/GSVector8i.h +++ b/pcsx2/GS/GSVector8i.h @@ -15,8 +15,6 @@ #include -#if _M_SSE >= 0x501 - class alignas(32) GSVector8i { static const GSVector8i m_xff[33]; @@ -45,7 +43,9 @@ public: u16 U16[16]; u32 U32[8]; u64 U64[4]; +#if _M_SSE >= 0x501 __m256i m; +#endif __m128i m0, m1; }; @@ -61,6 +61,8 @@ public: return GSVector8i(cxpr_init, x, x, x, x, x, x, x, x); } +#if _M_SSE >= 0x501 + __forceinline explicit GSVector8i(const GSVector8& v, bool truncate = true); __forceinline static GSVector8i cast(const GSVector8& v); @@ -1887,6 +1889,6 @@ public: __forceinline static GSVector8i xff(int n) { return m_xff[n]; } __forceinline static GSVector8i x0f(int n) { return m_x0f[n]; } +#endif }; -#endif diff --git a/tests/ctest/GS/swizzle_test_main.cpp b/tests/ctest/GS/swizzle_test_main.cpp index 9057aa49b4..fccd85334c 100644 --- a/tests/ctest/GS/swizzle_test_main.cpp +++ b/tests/ctest/GS/swizzle_test_main.cpp @@ -16,9 +16,12 @@ #include "PrecompiledHeader.h" #include "GSBlock.h" #include "GSClut.h" +#include "MultiISA.h" #include #include +using namespace CURRENT_ISA; + static void swizzle(const u8* table, u8* dst, const u8* src, int bpp, bool deswizzle) { int pxbytes = bpp / 8;