Add SVE2 optimizations for SynetInnerProduct16bGemmNN

docs/2026.html

@@ -51,6 +51,7 @@ <h5>New features</h5>
  <li>SVE2 optimizations of class SynetConvolution32fNhwcDirect.</li>
  <li>SVE2 optimizations of class SynetInnerProduct32fGemm.</li>
  <li>SVE2 optimizations of class SynetInnerProduct32fProd.</li>
+ <li>SVE2 optimizations of class SynetInnerProduct16bGemmNN.</li>
 </ul>
 <h5>Renaming</h5>
 <ul>

prj/vs2022/Sve2.vcxproj

@@ -128,6 +128,8 @@
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetMergedConvolution8iInput.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetMergedConvolution8iOutput.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetGridSample2d32fBlZ.cpp" />
+    <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct16b.cpp" />
+    <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct16bGemmNN.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct32f.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct8i.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetNormalize.cpp" />

prj/vs2022/Sve2.vcxproj.filters

@@ -545,6 +545,12 @@
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetMergedConvolution8iOutput.cpp">
       <Filter>Sve2\Synet\MergedConvolution</Filter>
     </ClCompile>
+    <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct16b.cpp">
+      <Filter>Sve2\Synet\InnerProduct</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct16bGemmNN.cpp">
+      <Filter>Sve2\Synet\InnerProduct</Filter>
+    </ClCompile>
     <ClCompile Include="..\..\src\Simd\SimdSve2SynetInnerProduct32f.cpp">
       <Filter>Sve2\Synet\InnerProduct</Filter>
     </ClCompile>

src/Simd/SimdLib.cpp

@@ -6773,7 +6773,7 @@ SIMD_API void* SimdSynetInnerProduct16bInit(size_t M, size_t N, size_t K, SimdTe
     SIMD_EMPTY();
 #if defined(SIMD_SYNET_ENABLE)
     typedef void* (*SimdSynetInnerProduct16bInitPtr) (size_t M, size_t N, size_t K, SimdTensorDataType typeA, SimdTensorDataType typeB, SimdTensorDataType typeC, SimdBool transB, SimdBool constB, SimdBool bias, SimdConvolutionActivationType activation);
-    const static SimdSynetInnerProduct16bInitPtr simdSynetInnerProduct16bInit = SIMD_FUNC4(SynetInnerProduct16bInit, SIMD_AMXBF16_FUNC, SIMD_AVX512BW_FUNC, SIMD_AVX2_FUNC, SIMD_SSE41_FUNC);
+    const static SimdSynetInnerProduct16bInitPtr simdSynetInnerProduct16bInit = SIMD_FUNC5(SynetInnerProduct16bInit, SIMD_AMXBF16_FUNC, SIMD_AVX512BW_FUNC, SIMD_AVX2_FUNC, SIMD_SSE41_FUNC, SIMD_SVE2_FUNC);
 
     return simdSynetInnerProduct16bInit(M, N, K, typeA, typeB, typeC, transB, constB, bias, activation);
 #else

src/Simd/SimdSve2SynetInnerProduct16b.cpp

@@ -0,0 +1,42 @@
+/*
+* Simd Library (http://ermig1979.github.io/Simd).
+*
+* Copyright (c) 2011-2026 Yermalayeu Ihar.
+*
+* 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.
+*/
+#include "Simd/SimdSynetInnerProduct16b.h"
+
+namespace Simd
+{
+#if defined(SIMD_SVE2_ENABLE) && defined(SIMD_SYNET_ENABLE)      
+    namespace Sve2
+    {
+        void* SynetInnerProduct16bInit(size_t M, size_t N, size_t K, SimdTensorDataType typeA, SimdTensorDataType typeB, SimdTensorDataType typeC, SimdBool transB, SimdBool constB, SimdBool bias, SimdConvolutionActivationType activation)
+        {
+            InnerProductParam16b param(M, N, K, typeA, typeB, typeC, transB, constB, bias, activation);
+            if (!param.Valid())
+                return NULL;
+            if (Base::SynetInnerProduct16bGemmNN::Preferable(param))
+                return new Sve2::SynetInnerProduct16bGemmNN(param);
+            return Base::SynetInnerProduct16bInit(M, N, K, typeA, typeB, typeC, transB, constB, bias, activation);
+        }
+    }
+#endif
+}