rewrite of f32 fma, and other optmization of simd kernel ops

This commit is contained in:
Youmu Liu
2025-11-10 20:40:19 +08:00
parent 1dee467220
commit 678a846e54
18 changed files with 302 additions and 156 deletions
@@ -26,6 +26,7 @@
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
using StgSharp.HighPerformance;
using System;
using System.Numerics;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
@@ -38,8 +39,9 @@ namespace StgSharp.Internal.Intrinsic
public delegate* unmanaged[Cdecl]<char*, int, int> city_hash_simplify;
public delegate* unmanaged[Cdecl]<void*, void*, void*, void> f32_add;
public delegate* unmanaged[Cdecl]<void*, ScalarPacket*, void> f32_fill;
public delegate* unmanaged[Cdecl]<void*, void*, void*, void> f32_fma;
public delegate* unmanaged[Cdecl]<void*, float, void*, void> f32_scalar_mul;
public delegate* unmanaged[Cdecl]<void*, void*, ScalarPacket*, void> f32_scalar_mul;
public delegate* unmanaged[Cdecl]<void*, void*, void*, void> f32_sub;
public delegate* unmanaged[Cdecl]<void*, void*, void> f32_transpose;
public delegate* unmanaged[Cdecl]<int, ulong> factorial_simd;
@@ -52,4 +54,4 @@ namespace StgSharp.Internal.Intrinsic
}
}
}
}