swiss ok, needs more benchmark

2026-03-06 18:22:30 +08:00
parent 83997f64d0
commit f6e897ed06
70 changed files with 1889 additions and 3371 deletions
@@ -0,0 +1,106 @@
+//-----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+// file="MatrixParallelThread.Fma"
+// Project: StgSharp
+// AuthorGroup: Nitload
+// Copyright (c) Nitload. All rights reserved.
+//     
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+//     
+// -----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+using StgSharp.Internal;
+using StgSharp.Mathematics;
+using StgSharp.Mathematics.Internal;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Runtime.CompilerServices;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace StgSharp.Mathematics.Numeric
+{
+    public partial class MatrixParallelThread
+    {
+
+        internal static unsafe void ComputeFmaPanelF32(
+                                    MatrixParallelTask* p
+        )
+        {
+            int f32 = MatrixElementType.F32.IntrinsicNode;
+            MatrixKernel* leftBuffer = (MatrixKernel*)p->Left;
+            MatrixKernel* rightBuffer = (MatrixKernel*)p->Right;
+            MatrixKernel* ansBuffer = (MatrixKernel*)p->Answer;
+            ref long offsetRef = ref Unsafe.As<int, long>(ref p->PrimTileOffset);
+            ref long lengthRef = ref Unsafe.As<int, long>(ref p->PrimCount);
+            int ansWidth = p->AnsPrimOffset;
+            int ansHeight = p->AnsSecOffset;
+            int commonK = p->LeftPrimOffset;
+
+            M128* enumeration = stackalloc M128[1];
+            enumeration->Member<int>(2) = commonK;
+            enumeration->Member<int>(3) = ansHeight;
+
+            // in this occasion size must be 1
+            for (long c = offsetRef; c < offsetRef + lengthRef; c++)
+            {
+                int i = (int)(c / ansHeight);
+                int j = (int)(c % ansHeight);
+                enumeration->Member<int>(0) = i;
+                enumeration->Member<int>(1) = j;
+                NumericalModule.GlobalContext.mat_sk[f32].kernel_tile_fma(leftBuffer, rightBuffer,
+                                                                         ansBuffer, enumeration);
+            }
+        }
+
+        internal static unsafe void ComputeFmaPanelF64(
+                                    MatrixParallelTask* p
+        )
+        {
+            int f64 = MatrixElementType.F64.IntrinsicNode;
+
+            // int size = 1;
+            MatrixKernel* leftBuffer = (MatrixKernel*)p->Left;
+            MatrixKernel* rightBuffer = (MatrixKernel*)p->Right;
+            MatrixKernel* ansBuffer = (MatrixKernel*)p->Answer;
+            ref long offsetRef = ref Unsafe.As<int, long>(ref p->PrimTileOffset);
+            ref long lengthRef = ref Unsafe.As<int, long>(ref p->PrimCount);
+            int ansWidth = p->AnsPrimOffset;
+            int ansHeight = p->AnsSecOffset;
+            int commonK = p->LeftPrimOffset;
+
+            M128* enumeration = stackalloc M128[1];
+            enumeration->Member<int>(2) = commonK;
+            enumeration->Member<int>(3) = ansHeight;
+
+            // in this occasion size must be 1
+            for (long c = offsetRef; c < offsetRef + lengthRef; c++)
+            {
+                int i = (int)(c / ansHeight);
+                int j = (int)(c % ansHeight);
+                enumeration->Member<int>(0) = i;
+                enumeration->Member<int>(1) = j;
+                NumericalModule.GlobalContext.mat_sk[f64].kernel_tile_fma(leftBuffer, rightBuffer,
+                                                                         ansBuffer, enumeration);
+            }
+        }
+
+    }
+}