动态并行允许从GPU端启动核函数,它要求显卡计算能力大于3.5.
动态并行Qt Creator的工程配置与不使用动态并行时有较大差别,以linux系统为例
QT += core greaterThan(QT_MAJOR_VERSION, 4): QT += widgets TARGET = test TEMPLATE = app DEFINES += QT_DEPRECATED_WARNINGS
INCLUDEPATH +=/usr/local/cuda-10.1/include SOURCES += main.cpp HEADERAS +=algorithm.h LIBS +=-L/usr/local/cuda-10.1/lib64 \ -lcublas -lcuda -lcudadevrt \ -lcudart -lcudart_static -lcufft \ -lcufftw -lcurand -lcusolver -lcusparse
OTHER_FILES +=./algorithm.cu CUDA_SOURCES+=./algorithm.cu CUDA_SDK ="/usr/local/cuda-10.1" CUDA_DIR ="/usr/local/cuda-10.1" QMAKE_LIBDIR += $$CUDA_DIR/lib64 SYSTEM_TYPE = 64 CUDA_ARCH = sm_60 NVCCFLAGS = --use_fast_math CUDA_INC = $$join(INCLUDEPATH,'" -I"','-I"','"') CUDA_LIBS= -L/us