FindCUDA: Fix PTX selection with multiple architectures

This commit is contained in:
Bjoern Thiel
2017-02-22 09:11:34 +01:00
committed by Brad King
parent 6cbad490c2
commit 934eb497f2

View File

@@ -112,6 +112,7 @@ function(CUDA_SELECT_NVCC_ARCH_FLAGS out_variable)
list(REMOVE_DUPLICATES CUDA_ARCH_LIST)
foreach(arch_name ${CUDA_ARCH_LIST})
set(arch_bin)
set(arch_ptx)
set(add_ptx FALSE)
# Check to see if we are compiling PTX
if(arch_name MATCHES "(.*)\\+PTX$")