From: Daniel Arndt <arndtd@ornl.gov>
Date: Thu, 25 Apr 2019 22:10:05 +0000 (-0600)
Subject: Use integers in cudaMemset
X-Git-Tag: v9.1.0-rc1~169^2
X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=1fb975f018c4fb380eb30cf54b79b5513fa9e6c5;p=dealii.git

Use integers in cudaMemset
---

diff --git a/include/deal.II/lac/la_parallel_vector.templates.h b/include/deal.II/lac/la_parallel_vector.templates.h
index b2e890c30e..22bfea5531 100644
--- a/include/deal.II/lac/la_parallel_vector.templates.h
+++ b/include/deal.II/lac/la_parallel_vector.templates.h
@@ -349,7 +349,7 @@ namespace LinearAlgebra
           Number *    result_device;
           cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
           AssertCuda(error_code);
-          error_code = cudaMemset(result_device, Number(), sizeof(Number));
+          error_code = cudaMemset(result_device, 0, sizeof(Number));
 
           const int n_blocks =
             1 + (size - 1) / (::dealii::CUDAWrappers::chunk_size *
diff --git a/include/deal.II/lac/vector_operations_internal.h b/include/deal.II/lac/vector_operations_internal.h
index 542c31499f..47bef5ac61 100644
--- a/include/deal.II/lac/vector_operations_internal.h
+++ b/include/deal.II/lac/vector_operations_internal.h
@@ -2363,7 +2363,7 @@ namespace internal
         cudaError_t error_code =
           cudaMalloc(&result_device, size * sizeof(Number));
         AssertCuda(error_code);
-        error_code = cudaMemset(result_device, Number(), sizeof(Number));
+        error_code = cudaMemset(result_device, 0, sizeof(Number));
 
         const int n_blocks = 1 + (size - 1) / (chunk_size * block_size);
         ::dealii::LinearAlgebra::CUDAWrappers::kernel::double_vector_reduction<
@@ -2414,7 +2414,7 @@ namespace internal
         Number *    result_device;
         cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
         AssertCuda(error_code);
-        error_code = cudaMemset(result_device, Number(), sizeof(Number));
+        error_code = cudaMemset(result_device, 0, sizeof(Number));
 
         const int n_blocks = 1 + (size - 1) / (chunk_size * block_size);
         ::dealii::LinearAlgebra::CUDAWrappers::kernel::reduction<
@@ -2451,7 +2451,7 @@ namespace internal
         Number *    result_device;
         cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
         AssertCuda(error_code);
-        error_code = cudaMemset(result_device, Number(), sizeof(Number));
+        error_code = cudaMemset(result_device, 0, sizeof(Number));
 
         const int n_blocks = 1 + (size - 1) / (chunk_size * block_size);
         ::dealii::LinearAlgebra::CUDAWrappers::kernel::reduction<
@@ -2501,7 +2501,7 @@ namespace internal
         Number *    res_d;
         cudaError_t error_code = cudaMalloc(&res_d, sizeof(Number));
         AssertCuda(error_code);
-        error_code = cudaMemset(res_d, 0., sizeof(Number));
+        error_code = cudaMemset(res_d, 0, sizeof(Number));
         AssertCuda(error_code);
 
         const int n_blocks = 1 + (size - 1) / (chunk_size * block_size);
diff --git a/source/lac/cuda_vector.cu b/source/lac/cuda_vector.cu
index 0787cb679c..a0adad9280 100644
--- a/source/lac/cuda_vector.cu
+++ b/source/lac/cuda_vector.cu
@@ -313,7 +313,7 @@ namespace LinearAlgebra
       cudaError_t error_code =
         cudaMalloc(&result_device, n_elements * sizeof(Number));
       AssertCuda(error_code);
-      error_code = cudaMemset(result_device, Number(), sizeof(Number));
+      error_code = cudaMemset(result_device, 0, sizeof(Number));
 
       const int n_blocks = 1 + (n_elements - 1) / (chunk_size * block_size);
       kernel::double_vector_reduction<Number, kernel::DotProduct<Number>>
@@ -540,7 +540,7 @@ namespace LinearAlgebra
       Number *    result_device;
       cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
       AssertCuda(error_code);
-      error_code = cudaMemset(result_device, Number(), sizeof(Number));
+      error_code = cudaMemset(result_device, 0, sizeof(Number));
 
       const int n_blocks = 1 + (n_elements - 1) / (chunk_size * block_size);
       kernel::reduction<Number, kernel::ElemSum<Number>>
@@ -571,7 +571,7 @@ namespace LinearAlgebra
       Number *    result_device;
       cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
       AssertCuda(error_code);
-      error_code = cudaMemset(result_device, Number(), sizeof(Number));
+      error_code = cudaMemset(result_device, 0, sizeof(Number));
 
       const int n_blocks = 1 + (n_elements - 1) / (chunk_size * block_size);
       kernel::reduction<Number, kernel::L1Norm<Number>>
@@ -619,7 +619,7 @@ namespace LinearAlgebra
       Number *    result_device;
       cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
       AssertCuda(error_code);
-      error_code = cudaMemset(result_device, Number(), sizeof(Number));
+      error_code = cudaMemset(result_device, 0, sizeof(Number));
 
       const int n_blocks = 1 + (n_elements - 1) / (chunk_size * block_size);
       kernel::reduction<Number, kernel::LInfty<Number>>
@@ -667,7 +667,7 @@ namespace LinearAlgebra
       Number *    result_device;
       cudaError_t error_code = cudaMalloc(&result_device, sizeof(Number));
       AssertCuda(error_code);
-      error_code = cudaMemset(result_device, 0., sizeof(Number));
+      error_code = cudaMemset(result_device, 0, sizeof(Number));
       AssertCuda(error_code);
 
       const int n_blocks = 1 + (n_elements - 1) / (chunk_size * block_size);