+add AVX-512BW optimizations of class SynetConvolution16bNhwcSpecV1.

Author: ermig1979

Diff for: docs/2025.html

@@ -40,7 +40,7 @@ <h3 id="R149">May X, 2025 (version 6.1.149)</h3>
 <h4>Algorithms</h4>
 <h5>New features</h5>
 <ul>
- <li>Base implementation, SSE4.1, AVX2 optimizations of class SynetConvolution16bNhwcSpecV1.</li>
+ <li>Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of class SynetConvolution16bNhwcSpecV1.</li>
 </ul>
 <h5>Bug fixing</h5>
 <ul>

Diff for: prj/vs2019/Avx512bw.vcxproj

@@ -94,6 +94,7 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcDeptwise.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcGemm.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV0.cpp" />
+    <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV1.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution32f.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution32fDirectNchw.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution32fGemm.cpp" />

Diff for: prj/vs2019/Avx512bw.vcxproj.filters

@@ -412,6 +412,9 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV0.cpp">
       <Filter>Avx512bw</Filter>
     </ClCompile>
+    <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV1.cpp">
+      <Filter>Avx512bw</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <Filter Include="Avx512bw">

Diff for: prj/vs2022/Avx512bw.vcxproj

@@ -94,6 +94,7 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcDeptwise.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcGemm.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV0.cpp" />
+    <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV1.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution32f.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution32fDirectNchw.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution32fGemm.cpp" />

Diff for: prj/vs2022/Avx512bw.vcxproj.filters

@@ -412,6 +412,9 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV0.cpp">
       <Filter>Avx512bw</Filter>
     </ClCompile>
+    <ClCompile Include="..\..\src\Simd\SimdAvx512bwSynetConvolution16bNhwcSpecV1.cpp">
+      <Filter>Avx512bw</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <Filter Include="Avx512bw">

Diff for: src/Simd/SimdAvx512bwSynetConvolution16b.cpp

@@ -33,6 +33,8 @@ namespace Simd
             ConvParam param(batch, conv, compatibility);
             if (!param.Valid(SimdTensorData32f, SimdTensorData16b))
                 return NULL;
+            if (SynetConvolution16bNhwcSpecV1::Preferable(param))
+                return new Avx512bw::SynetConvolution16bNhwcSpecV1(param);
             if (SynetConvolution16bNhwcSpecV0::Preferable(param))
                 return new Avx512bw::SynetConvolution16bNhwcSpecV0(param);
             if (SynetConvolution16bNhwcGemm::Preferable(param))

Diff for: src/Simd/SimdAvx512bwSynetConvolution16bNhwcSpecV1.cpp

@@ -423,6 +423,13 @@ namespace Simd
             virtual String Ext() const { return "Avx512bw"; }
         };
 
+        class SynetConvolution16bNhwcSpecV1 : public Avx2::SynetConvolution16bNhwcSpecV1
+        {
+        public:
+            SynetConvolution16bNhwcSpecV1(const ConvParam& p);
+
+            virtual String Ext() const { return "Avx512bw"; }
+        };
 
         class SynetConvolution16bNhwcDepthwise : public Avx2::SynetConvolution16bNhwcDepthwise
         {