John Platts

Here are lines 186-188 of simde/simde-features.h: ``` #if defined(SIMDE_X86_AVX_NATIVE) && !defined(SIMDE_X86_SSE4_1_NATIVE) #define SIMDE_X86_SSE4_2_NATIVE #endif ``` Here is how lines 186-188 of simde/simde-features.h should read: ``` #if defined(SIMDE_X86_AVX_NATIVE) && !defined(SIMDE_X86_SSE4_2_NATIVE) #define...

Including "simde/arm/neon.h" fails to compile on Visual C++ for an ARM64 target since some of the SIMD_CONSTIFY_2_, SIMDE_CONSTIFY_4_, SIMDE_CONSTIFY_8_, and SIMDE_CONSTIFY_16_ macros used in some of the header files in...

The simde_vaddvq_u8, simde_vaddlvq_u8, simde_vaddvq_s8, and simde_vaddlvq_s8 routines can be implemented on x86 platforms with SSE2 support using the _mm_sad_epu8 intrinsic. Here is how simde_vaddvq_u8 could be implemented on x86 platforms...

While the fix in simde/x86/avx512/permutex2var.h that was applied to commit https://github.com/simd-everywhere/simde/commit/88931164bdb8b628702079bfa177087a614cc975 does partially fix the issue with getting simde/x86/avx512/permutex2var.h to compile with Visual C++, I am still getting the following...

Add support for TwoTablesLookupBytes

2

Some SIMD instruction sets such as NEON, Altivec, and AVX-512VBMI have 2-table lookup instructions. TwoTablesLookupBytes is more efficient than doing two TableLookupBytes operations, computing a mask, and combining the results...

Add support for SaturatedAdd and SaturatedSub for int32_t, uint32_t, int64_t, and uint64_t vector types

3

Some SIMD instruction sets such as NEON, SVE, and RVV have instructions for saturated addition/subtraction of int32_t, uint32_t, int64_t, and uint64_t SIMD vectors. AltiVec also has instructions for saturated addition/subtraction...

Here is the current implementation of FirstN for 32-bit AVX-512 targets for the lane size == 1, vector size == 512 bits case: https://github.com/google/highway/blob/22e3d7276f4157d4a47586ba9fd91dd6303f441a/hwy/ops/x86_512-inl.h#L476-L480 Here is a better FirstN implementation...

Memory allocation fixes in flatlaf-natives/flatlaf-natives-windows

1

I noticed that the operator delete[] overload is missing (and should be added) to the flatlaf-natives/flatlaf-natives-windows/src/main/cpp/Runtime.cpp file as follows: ``` void __cdecl operator delete[]( void* pv, size_t cb ) {...

Here is the current implementation of NettyDataBuffer.retainedSlice: https://github.com/spring-projects/spring-framework/blob/a41f97bc2bd547dbb635f032413fe586968912f8/spring-core/src/main/java/org/springframework/core/io/buffer/NettyDataBuffer.java#L265-L269 The above version of NettyDataBuffer.retainedSlice will fail to release the slice returned by this.byteBuf.retainedSlice if this.byteBuf.retainedSlice succeeds but allocation of the new...

Updated F32 to BF16 demote ops and BF16FromF32

5

Updated BF16FromF32 implementation in hwy/base.h to round the F32 value to the nearest BF16 value (with ties to even) instead of simply truncating the F32 value to a BF16 to...