From b5f99aabb0087fadfdc45256a3d9d055b62b299a Mon Sep 17 00:00:00 2001
From: Xuan Hien <hiennguyen9874@gmail.com>
Date: Wed, 30 Nov 2022 23:07:31 +0700
Subject: [PATCH] Fix

---
 .../efficientNMSCustomInference.cu               | 16 ++++++++--------
 .../efficientNMSLandmarkInference.cu             | 16 ++++++++--------
 plugin/roIAlignPlugin/roIAlignPlugin.h           |  2 +-
 3 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/plugin/efficientNMSCustomPlugin/efficientNMSCustomInference.cu b/plugin/efficientNMSCustomPlugin/efficientNMSCustomInference.cu
index 5952daa3..4798db2a 100644
--- a/plugin/efficientNMSCustomPlugin/efficientNMSCustomInference.cu
+++ b/plugin/efficientNMSCustomPlugin/efficientNMSCustomInference.cu
@@ -590,11 +590,11 @@ pluginStatus_t EfficientNMSCustomDispatch(EfficientNMSCustomParameters param, co
     void* nmsClassesOutput, void* nmsIndicesOutput, void* workspace, cudaStream_t stream)
 {
     // Clear Outputs (not all elements will get overwritten by the kernels, so safer to clear everything out)
-    CSC(cudaMemsetAsync(numDetectionsOutput, 0x00, param.batchSize * sizeof(int), stream));
-    CSC(cudaMemsetAsync(nmsScoresOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(T), stream));
-    CSC(cudaMemsetAsync(nmsBoxesOutput, 0x00, param.batchSize * param.numOutputBoxes * 4 * sizeof(T), stream));
-    CSC(cudaMemsetAsync(nmsClassesOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(int), stream));
-    CSC(cudaMemsetAsync(nmsIndicesOutput, 0xFF, param.batchSize * param.numOutputBoxes * sizeof(int), stream));
+    CSC(cudaMemsetAsync(numDetectionsOutput, 0x00, param.batchSize * sizeof(int), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsScoresOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(T), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsBoxesOutput, 0x00, param.batchSize * param.numOutputBoxes * 4 * sizeof(T), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsClassesOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(int), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsIndicesOutput, 0xFF, param.batchSize * param.numOutputBoxes * sizeof(int), stream), STATUS_FAILURE);
 
     // Empty Inputs
     if (param.numScoreElements < 1)
@@ -610,7 +610,7 @@ pluginStatus_t EfficientNMSCustomDispatch(EfficientNMSCustomParameters param, co
     int* topOffsetsEndData = topNumData + 2 * param.batchSize;
     int* outputIndexData = topNumData + 3 * param.batchSize;
     int* outputClassData = topNumData + 4 * param.batchSize;
-    CSC(cudaMemsetAsync(topNumData, 0x00, countersTotalSize * sizeof(int), stream));
+    CSC(cudaMemsetAsync(topNumData, 0x00, countersTotalSize * sizeof(int), stream), STATUS_FAILURE);
     cudaError_t status = cudaGetLastError();
     CSC(status, STATUS_FAILURE);
 
@@ -633,9 +633,9 @@ pluginStatus_t EfficientNMSCustomDispatch(EfficientNMSCustomParameters param, co
 
     // Device Specific Properties
     int device;
-    CSC(cudaGetDevice(&device));
+    CSC(cudaGetDevice(&device), STATUS_FAILURE);
     struct cudaDeviceProp properties;
-    CSC(cudaGetDeviceProperties(&properties, device));
+    CSC(cudaGetDeviceProperties(&properties, device), STATUS_FAILURE);
     if (properties.regsPerBlock >= 65536)
     {
         // Most Devices
diff --git a/plugin/efficientNMSLandmarkPlugin/efficientNMSLandmarkInference.cu b/plugin/efficientNMSLandmarkPlugin/efficientNMSLandmarkInference.cu
index 33628734..457d559b 100644
--- a/plugin/efficientNMSLandmarkPlugin/efficientNMSLandmarkInference.cu
+++ b/plugin/efficientNMSLandmarkPlugin/efficientNMSLandmarkInference.cu
@@ -598,11 +598,11 @@ pluginStatus_t EfficientNMSLandmarkDispatch(EfficientNMSLandmarkParameters param
     cudaStream_t stream)
 {
     // Clear Outputs (not all elements will get overwritten by the kernels, so safer to clear everything out)
-    CSC(cudaMemsetAsync(numDetectionsOutput, 0x00, param.batchSize * sizeof(int), stream));
-    CSC(cudaMemsetAsync(nmsScoresOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(T), stream));
-    CSC(cudaMemsetAsync(nmsBoxesOutput, 0x00, param.batchSize * param.numOutputBoxes * 4 * sizeof(T), stream));
-    CSC(cudaMemsetAsync(nmsClassesOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(int), stream));
-    CSC(cudaMemsetAsync(nmsLandmarksOutput, 0x00, param.batchSize * param.numOutputBoxes * 10 * sizeof(T), stream));
+    CSC(cudaMemsetAsync(numDetectionsOutput, 0x00, param.batchSize * sizeof(int), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsScoresOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(T), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsBoxesOutput, 0x00, param.batchSize * param.numOutputBoxes * 4 * sizeof(T), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsClassesOutput, 0x00, param.batchSize * param.numOutputBoxes * sizeof(int), stream), STATUS_FAILURE);
+    CSC(cudaMemsetAsync(nmsLandmarksOutput, 0x00, param.batchSize * param.numOutputBoxes * 10 * sizeof(T), stream), STATUS_FAILURE);
 
     // Empty Inputs
     if (param.numScoreElements < 1)
@@ -618,7 +618,7 @@ pluginStatus_t EfficientNMSLandmarkDispatch(EfficientNMSLandmarkParameters param
     int* topOffsetsEndData = topNumData + 2 * param.batchSize;
     int* outputIndexData = topNumData + 3 * param.batchSize;
     int* outputClassData = topNumData + 4 * param.batchSize;
-    CSC(cudaMemsetAsync(topNumData, 0x00, countersTotalSize * sizeof(int), stream));
+    CSC(cudaMemsetAsync(topNumData, 0x00, countersTotalSize * sizeof(int), stream), STATUS_FAILURE);
     cudaError_t status = cudaGetLastError();
     CSC(status, STATUS_FAILURE);
 
@@ -642,9 +642,9 @@ pluginStatus_t EfficientNMSLandmarkDispatch(EfficientNMSLandmarkParameters param
 
     // Device Specific Properties
     int device;
-    CSC(cudaGetDevice(&device));
+    CSC(cudaGetDevice(&device), STATUS_FAILURE);
     struct cudaDeviceProp properties;
-    CSC(cudaGetDeviceProperties(&properties, device));
+    CSC(cudaGetDeviceProperties(&properties, device), STATUS_FAILURE);
     if (properties.regsPerBlock >= 65536)
     {
         // Most Devices
diff --git a/plugin/roIAlignPlugin/roIAlignPlugin.h b/plugin/roIAlignPlugin/roIAlignPlugin.h
index 5b669e69..5d873e23 100644
--- a/plugin/roIAlignPlugin/roIAlignPlugin.h
+++ b/plugin/roIAlignPlugin/roIAlignPlugin.h
@@ -88,7 +88,7 @@ class RoIAlignDynamicPlugin : public IPluginV2DynamicExt
     bool mAligned;
 };
 
-class RoIAlignBasePluginCreator : public BaseCreator
+class RoIAlignBasePluginCreator : public nvinfer1::pluginInternal::BaseCreator
 {
 public:
     RoIAlignBasePluginCreator() noexcept;