CMSIS-NN: Added optimized version of arm_avgpool_s8

Testing/GeneratedInclude/Pooling_decl.h

@@ -20,6 +20,8 @@ static const int REF8_S8_ID=15;
 
 // Output IDs
 static const int OUTPUT_S8_ID=0;
+static const int TEMP_S8_ID=1;
+static const int TEMPINPUT_S8_ID=2;
 
 // Test IDs
 static const int TEST_AVGPOOL_S8_1=1;

Testing/GeneratedInclude/TestDrive.h

@@ -44,8 +44,10 @@ __ALIGNED(8) const char testDesc[]={
 129,15,0,0,
 224,19,0,0,
 1,0,0,0,
-1,0,0,0,
+3,0,0,0,
 'O','u','t','p','u','t','\0',
+'T','e','m','p','\0',
+'T','e','m','p','I','n','p','u','t','\0',
 0,0,0,0,
 1,0,0,0,
 1,0,0,0,

Testing/Include/Pooling.h

@@ -11,7 +11,9 @@ class Pooling:public Client::Suite
             
             Client::Pattern<q7_t> input;
 
+            Client::LocalPattern<q7_t> tmpInput;
             Client::LocalPattern<q7_t> output;
+            Client::LocalPattern<q15_t> temp;
 
             // Reference patterns are not loaded when we are in dump mode
             Client::RefPattern<q7_t> ref;

Testing/Source/Pooling.cpp

@@ -9,7 +9,11 @@
     void Pooling::test_avgpool_s8()
     {
        const q7_t *inp = input.ptr();
+       q7_t *tmpin = tmpInput.ptr();
        q7_t *outp = output.ptr();
+       q15_t *tempp = temp.ptr();
+
+       memcpy(tmpin,inp,input.nbSamples());
 
        arm_avgpool_s8(
            DIM_IN_Y,
@@ -25,7 +29,8 @@
            ACT_MIN,
            ACT_MAX,
            IN_CHANNEL,
-           inp,
+           tmpin,
+           tempp,
            outp);
        
        ASSERT_EQ(ref,output);
@@ -194,8 +199,12 @@
 
 
        }
+       temp.create(this->DIM_OUT_X * this->IN_CHANNEL,Pooling::TEMP_S8_ID,mgr);
 
        output.create(ref.nbSamples(),Pooling::OUTPUT_S8_ID,mgr);
+       tmpInput.create(input.nbSamples(),Pooling::TEMPINPUT_S8_ID,mgr);
+
+
 
 
     }

Testing/TestDesc.txt

@@ -27,8 +27,10 @@ Input7.txt
 Ref7.txt
 Input8.txt
 Ref8.txt
-1
+3
 Output
+Temp
+TempInput
 0
 1 1
 n

Testing/desc.txt

@@ -627,6 +627,8 @@ group Root {
           Pattern REF8_S8_ID : Ref8.txt
 
           Output  OUTPUT_S8_ID : Output
+          Output  TEMP_S8_ID : Temp
+          Output  TEMPINPUT_S8_ID : TempInput
 
           Functions {
             arm_avgpool_s8 Test 1:test_avgpool_s8