fix a strange bug of cuda 11

src/spconv/indice.cu

@@ -105,6 +105,7 @@ int create_conv_indice_pair_p2_cuda(
   auto kernelVolume = indiceNum.size(0);
   if (numActIn == 0)
     return 0;
+  bool failed = false;
   tv::dispatch_torch<int32_t>(indicesIn.scalar_type(), [&](auto IndexValue) {
     using Index = TV_DECLTYPE(IndexValue);
     using IndexGrid = int32_t;
@@ -131,7 +132,8 @@ int create_conv_indice_pair_p2_cuda(
         cudaFree(d_values);
         TV_CHECK_CUDA_ERR_V2("cudaFree failed");
         if (!res) {
-          return -1; // use -1 to tell outside use CPU implementation
+          failed = true;
+          return;
         }
         assignIndiceOutKernel<Index, NDim>
             <<<tv::cuda::getBlocks(numAct), tv::cuda::CUDA_NUM_THREADS, 0,
@@ -190,6 +192,9 @@ int create_conv_indice_pair_p2_cuda(
       }
     });
   });
+  if (failed){
+    return -1;
+  }
   return numAct;
 }
 
@@ -211,6 +216,8 @@ int create_submconv_indice_pair_cuda(
   auto kernelVolume = indiceNum.size(0);
   if (numActIn == 0)
     return 0;
+  bool failed = false;
+
   tv::dispatch_torch<int32_t>(indicesIn.scalar_type(), [&](auto IndexValue) {
     using Index = TV_DECLTYPE(IndexValue);
     using IndexGrid = int32_t;
@@ -245,7 +252,8 @@ int create_submconv_indice_pair_cuda(
         cudaFree(d_keyvalues);
         TV_CHECK_CUDA_ERR_V2("cudaFree failed");
         if (!res) {
-          return -1; // use -1 to tell outside use CPU implementation
+          failed = true;
+          return;
         }
         auto tableSize = table.get_table_size();
         auto tableData = table.data();
@@ -349,6 +357,10 @@ int create_submconv_indice_pair_cuda(
       }
     });
   });
+  if (failed){
+    return -1;
+  }
+
   return numActIn;
 }