summaryrefslogtreecommitdiff
path: root/benchmarks
diff options
context:
space:
mode:
Diffstat (limited to 'benchmarks')
-rw-r--r--benchmarks/CUDA/template/template_kernel.cu8
1 files changed, 7 insertions, 1 deletions
diff --git a/benchmarks/CUDA/template/template_kernel.cu b/benchmarks/CUDA/template/template_kernel.cu
index 4ced538..14610d8 100644
--- a/benchmarks/CUDA/template/template_kernel.cu
+++ b/benchmarks/CUDA/template/template_kernel.cu
@@ -30,6 +30,12 @@
//! @param g_idata input data in global memory
//! @param g_odata output data in global memory
////////////////////////////////////////////////////////////////////////////////
+
+__device__ float foo( float bar )
+{
+ return 2.0f*bar;
+}
+
__global__ void
testKernel( float* g_idata, float* g_odata)
{
@@ -45,7 +51,7 @@ testKernel( float* g_idata, float* g_odata)
// read in input data from global memory
// use the bank checker macro to check for bank conflicts during host
// emulation
- SDATA(tid) = g_idata[tid];
+ SDATA(tid) = foo(g_idata[tid])/2.0;
__syncthreads();
printf("thread tid=%u reads %f from g_idata[]\n", tid, sdata[tid]);