+add AVX2 optimizations of function SynetSoftmax16b (part 1).

Author: ermig1979

docs/2026.html

@@ -47,7 +47,7 @@ <h5>New features</h5>
  <li>Parameter 'params' to function SimdSynetInnerProduct16bSetParams.</li>
  <li>Base implementation of class SynetGatherElements.</li>
  <li>Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function SynetNormalizeLayerForward16bV2.</li>
- <li>Base implementation, SSE4.1 optimizations of function SynetSoftmax16b.</li>
+ <li>Base implementation, SSE4.1, AVX2 optimizations of function SynetSoftmax16b.</li>
 </ul>
 <h5>Improving</h5>
 <ul>

prj/vs2022/Avx2.vcxproj

@@ -152,7 +152,8 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetQuantizeLinear.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetScale.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetScale16b.cpp" />
-    <ClCompile Include="..\..\src\Simd\SimdAvx2SynetSoftmax.cpp" />
+    <ClCompile Include="..\..\src\Simd\SimdAvx2SynetSoftmax16b.cpp" />
+    <ClCompile Include="..\..\src\Simd\SimdAvx2SynetSoftmax32f.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetUnaryOperation.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx2Texture.cpp" />
     <ClCompile Include="..\..\src\Simd\SimdAvx2Transform.cpp" />

prj/vs2022/Avx2.vcxproj.filters

@@ -406,9 +406,6 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetScale16b.cpp">
       <Filter>Avx2\Synet\Other</Filter>
     </ClCompile>
-    <ClCompile Include="..\..\src\Simd\SimdAvx2SynetSoftmax.cpp">
-      <Filter>Avx2\Synet\Other</Filter>
-    </ClCompile>
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetUnaryOperation.cpp">
       <Filter>Avx2\Synet\Other</Filter>
     </ClCompile>
@@ -472,6 +469,12 @@
     <ClCompile Include="..\..\src\Simd\SimdAvx2SynetNormalize32f.cpp">
       <Filter>Avx2\Synet\Normalize</Filter>
     </ClCompile>
+    <ClCompile Include="..\..\src\Simd\SimdAvx2SynetSoftmax16b.cpp">
+      <Filter>Avx2\Synet\Softmax</Filter>
+    </ClCompile>
+    <ClCompile Include="..\..\src\Simd\SimdAvx2SynetSoftmax32f.cpp">
+      <Filter>Avx2\Synet\Softmax</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <Filter Include="Avx2">
@@ -537,6 +540,9 @@
     <Filter Include="Avx2\Synet\Normalize">
       <UniqueIdentifier>{73ef30bd-9802-44ff-a136-fe77b4e92c11}</UniqueIdentifier>
     </Filter>
+    <Filter Include="Avx2\Synet\Softmax">
+      <UniqueIdentifier>{86bdc095-1674-4e46-b8a5-8da827051b9a}</UniqueIdentifier>
+    </Filter>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="..\..\src\Simd\SimdAvx2.h">

src/Simd/SimdAvx2.h

@@ -569,6 +569,8 @@ namespace Simd
 
         void SynetSoftmax32f(const float * src, size_t outer, size_t size, size_t inner, float * dst);
 
+        void SynetSoftmax16b(const uint16_t* src, size_t outer, size_t count, size_t inner, uint16_t* dst);
+
         void SynetSoftplus32f(const float* src, size_t size, const float* beta, const float* threshold, float* dst);
 
         void SynetShuffleLayerForward(const float* src0, const float* src1, size_t channels0, size_t channels1, size_t spatial, float* dst0, float* dst1, SimdTensorFormatType format, int type);