diff --git a/src/core/core.vcxproj b/src/core/core.vcxproj index 85cde4ff8..addb7a1a8 100644 --- a/src/core/core.vcxproj +++ b/src/core/core.vcxproj @@ -51,12 +51,6 @@ - - AdvancedVectorExtensions2 - %(AdditionalOptions) -mavx2 - true - NotUsing - diff --git a/src/core/core.vcxproj.filters b/src/core/core.vcxproj.filters index 5bd97eaf4..ec0c37da1 100644 --- a/src/core/core.vcxproj.filters +++ b/src/core/core.vcxproj.filters @@ -67,7 +67,6 @@ - diff --git a/src/core/gpu_sw_rasterizer.cpp b/src/core/gpu_sw_rasterizer.cpp index 371766ddb..b7dc3eabc 100644 --- a/src/core/gpu_sw_rasterizer.cpp +++ b/src/core/gpu_sw_rasterizer.cpp @@ -47,7 +47,7 @@ namespace GPU_SW_Rasterizer::SIMD { #define USE_VECTOR 1 #include "gpu_sw_rasterizer.inl" #undef USE_VECTOR -} +} // namespace GPU_SW_Rasterizer::SIMD #endif // Initialize with default implementation. @@ -78,10 +78,9 @@ void GPU_SW_Rasterizer::SelectImplementation() #if defined(CPU_ARCH_SSE) || defined(CPU_ARCH_NEON) const char* use_isa = std::getenv("SW_USE_ISA"); - // Default to scalar for now, until vector is finished. - use_isa = use_isa ? use_isa : "Scalar"; - -#if defined(CPU_ARCH_SSE) && defined(_MSC_VER) + // AVX2/256-bit path still has issues, and I need to make sure that it's not ODR'ing any shared + // symbols on top of the base symbols. +#if defined(CPU_ARCH_SSE) && defined(_MSC_VER) && 0 if (cpuinfo_has_x86_avx2() && (!use_isa || StringUtil::Strcasecmp(use_isa, "AVX2") == 0)) { SELECT_ALTERNATIVE_RASTERIZER(AVX2);