Audio support - Replace Audio QA samples

utilities/test_suite/HIP/Tensor_audio_hip.cpp

@@ -114,8 +114,8 @@ int main(int argc, char **argv)
     // set buffer sizes for src/dst
     if(testCase == MEL_FILTER_BANK)
     {
-        iBufferSize = (Rpp64u)MEL_FILTER_BANK_MAX_HEIGHT * (Rpp64u)srcDescPtr->w * (Rpp64u)srcDescPtr->c * (Rpp64u)srcDescPtr->n;
-        oBufferSize = (Rpp64u)MEL_FILTER_BANK_MAX_HEIGHT * (Rpp64u)dstDescPtr->w * (Rpp64u)dstDescPtr->c * (Rpp64u)dstDescPtr->n;
+        iBufferSize = (Rpp64u)AUDIO_MAX_HEIGHT * (Rpp64u)srcDescPtr->w * (Rpp64u)srcDescPtr->c * (Rpp64u)srcDescPtr->n;
+        oBufferSize = (Rpp64u)AUDIO_MAX_HEIGHT * (Rpp64u)dstDescPtr->w * (Rpp64u)dstDescPtr->c * (Rpp64u)dstDescPtr->n;
     }
     else
     {
@@ -124,12 +124,12 @@ int main(int argc, char **argv)
     }
 
     // compute maximum possible buffer size of resample
-    Rpp64u resampleMaxBufferSize = dstDescPtr->n * dstDescPtr->strides.nStride * 1.15;
+    Rpp64u resampleMaxBufferSize = dstDescPtr->n * dstDescPtr->strides.nStride * AUDIO_MAX_HEIGHT;
     if (testCase == RESAMPLE)
         oBufferSize = resampleMaxBufferSize;
 
     // compute maximum possible buffer size of spectrogram
-    Rpp64u spectrogramMaxBufferSize = 257 * 3754 * dstDescPtr->n;
+    Rpp64u spectrogramMaxBufferSize = AUDIO_MAX_HEIGHT * SPECTROGRAM_MAX_WIDTH * dstDescPtr->n;
     if (testCase == SPECTROGRAM)
         oBufferSize = spectrogramMaxBufferSize;
 
@@ -348,11 +348,11 @@ int main(int argc, char **argv)
                     Rpp32s numFilter = 80;
                     bool normalize = true;
                     srcDimsTensor[0] = 257;
-                    srcDimsTensor[1] = 225;
+                    srcDimsTensor[1] = 3170;
                     srcDimsTensor[2] = 257;
-                    srcDimsTensor[3] = 211;
+                    srcDimsTensor[3] = 552;
                     srcDimsTensor[4] = 257;
-                    srcDimsTensor[5] = 214;
+                    srcDimsTensor[5] = 1131;
 
                     init_mel_filter_bank(&inputf32, &outputf32, srcDescPtr, dstDescPtr, dstDims, offsetInBytes, numFilter, batchSize, srcDimsTensor, scriptPath, testType);
 
@@ -375,7 +375,7 @@ int main(int argc, char **argv)
             if (missingFuncFlag == 1)
             {
                 cout << "\nThe functionality " << func << " doesn't yet exist in RPP\n";
-                return -1;
+                return RPP_ERROR_NOT_IMPLEMENTED;
             }
 
             wallTime = endWallTime - startWallTime;

utilities/test_suite/HOST/Tensor_audio_host.cpp

@@ -121,15 +121,15 @@ int main(int argc, char **argv)
         descriptorPtr3D->offsetInBytes = 0;
         descriptorPtr3D->dataType = RpptDataType::F32;
         descriptorPtr3D->dims[0] = batchSize;
-        descriptorPtr3D->dims[1] = maxSrcWidth;
+        descriptorPtr3D->dims[1] = (maxSrcWidth / 8) * 8 + 8; // Ensure a consistent dimension order between generic and typed descriptors to prevent errors.
         descriptorPtr3D->strides[0] = descriptorPtr3D->dims[1];
     }
 
     // set buffer sizes for src/dst
     if(testCase == MEL_FILTER_BANK)
     {
-        iBufferSize = (Rpp64u)MEL_FILTER_BANK_MAX_HEIGHT * (Rpp64u)srcDescPtr->w * (Rpp64u)srcDescPtr->c * (Rpp64u)srcDescPtr->n;
-        oBufferSize = (Rpp64u)MEL_FILTER_BANK_MAX_HEIGHT * (Rpp64u)dstDescPtr->w * (Rpp64u)dstDescPtr->c * (Rpp64u)dstDescPtr->n;
+        iBufferSize = (Rpp64u)AUDIO_MAX_HEIGHT * (Rpp64u)srcDescPtr->w * (Rpp64u)srcDescPtr->c * (Rpp64u)srcDescPtr->n;
+        oBufferSize = (Rpp64u)AUDIO_MAX_HEIGHT * (Rpp64u)dstDescPtr->w * (Rpp64u)dstDescPtr->c * (Rpp64u)dstDescPtr->n;
     }
     else
     {
@@ -138,12 +138,12 @@ int main(int argc, char **argv)
     }
 
     // compute maximum possible buffer size of resample
-    Rpp64u resampleMaxBufferSize = dstDescPtr->n * dstDescPtr->strides.nStride * 1.15;
+    Rpp64u resampleMaxBufferSize = dstDescPtr->n * dstDescPtr->strides.nStride * RESAMPLE_BUFFER_SCALE_FACTOR;
     if (testCase == RESAMPLE)
         oBufferSize = resampleMaxBufferSize;
 
     // compute maximum possible buffer size of spectrogram
-    Rpp64u spectrogramMaxBufferSize = 257 * 3754 * dstDescPtr->n;
+    Rpp64u spectrogramMaxBufferSize = AUDIO_MAX_HEIGHT * SPECTROGRAM_MAX_WIDTH * dstDescPtr->n;
     if (testCase == SPECTROGRAM)
         oBufferSize = spectrogramMaxBufferSize;
 
@@ -323,8 +323,8 @@ int main(int argc, char **argv)
                     maxDstWidth = 0;
                     for(int i = 0, j = 0; i < batchSize; i++, j += 2)
                     {
-                        inRateTensor[i] = 16000;
-                        outRateTensor[i] = 16000 * 1.15f;
+                        inRateTensor[i] = 44100;
+                        outRateTensor[i] = 44100 * 1.15f;
                         Rpp32f scaleRatio = outRateTensor[i] / inRateTensor[i];
                         srcDimsTensor[j] = srcLengthTensor[i];
                         srcDimsTensor[j + 1] = channelsTensor[i];
@@ -364,7 +364,7 @@ int main(int argc, char **argv)
                     Rpp32s numFilter = 80;
                     bool normalize = true;
                     // (height, width) for each tensor in a batch for given QA inputs.
-                    Rpp32s srcDimsTensor[] = {257, 225, 257, 211, 257, 214};
+                    Rpp32s srcDimsTensor[] = {257, 3170, 257, 552, 257, 1131};
 
                     init_mel_filter_bank(&inputf32, &outputf32, srcDescPtr, dstDescPtr, dstDims, offsetInBytes, numFilter, batchSize, srcDimsTensor, scriptPath, testType);
 
@@ -384,7 +384,7 @@ int main(int argc, char **argv)
             if (missingFuncFlag == 1)
             {
                 cout << "\nThe functionality " << func << " doesn't yet exist in RPP\n";
-                return -1;
+                return RPP_ERROR_NOT_IMPLEMENTED;
             }
 
             wallTime = endWallTime - startWallTime;

utilities/test_suite/rpp_test_suite_audio.h

@@ -35,7 +35,9 @@ SOFTWARE.
 #include <sndfile.h>
 using namespace std;
 
-#define MEL_FILTER_BANK_MAX_HEIGHT 257 // Maximum height for mel filter bank set to 257 to ensure compatibility with test configuration
+#define AUDIO_MAX_HEIGHT             257  // Maximum height for mel filter bank and spectrogram set to 257 to ensure compatibility with test configuration, calculated as (nfft / 2) + 1 for a standard nfft of 512
+#define RESAMPLE_BUFFER_SCALE_FACTOR 1.16 // Scale factor to allocate a safe maximum buffer size for resampling, allowing for upsampling
+#define SPECTROGRAM_MAX_WIDTH        3170 // Maximum width for a spectrogram, pre-calculated based on the longest audio file in the test dataset
 
 std::map<int, string> audioAugmentationMap =
 {
@@ -63,9 +65,9 @@ enum Augmentation {
 // Golden outputs for Non Silent Region Detection
 std::map<string, std::vector<int>> NonSilentRegionReferenceOutputs =
 {
-    {"sample1", {0, 35840}},
-    {"sample2", {0, 33680}},
-    {"sample3", {0, 34160}}
+    {"sample1", {0, 507150}},
+    {"sample2", {0, 88200}},
+    {"sample3", {0, 180810}}
 };
 
 // Cutoff values for audio kernels listed for HOST backend followed by HIP