testing cms hcal/ecal regression

• git clone <repo> and cd <repo>
• mkdir build and cd build
• cmake ../ -DEIGEN_HOME=<path to eigen root dir>
• cmake ../ -DDOpenCL_INCLUDE_DIR=<path to opencl root dir>
• cmake ../ -DDEBUG=ON to compile in debug mode>
• cmake ../ -DTURNOFF_CUDA=ON to disable cuda>
• cmake ../ -DTURNOFF_OPENCL=ON to disable opencl>
• if u r @vinavx2 use: cmake ../ -DEIGEN_HOME=/data/user/vkhriste/eigen/eigen -DOpenCL_INCLUDE_DIR=/usr/local/cuda-9.2/include

• generate pulses ./gen/gen_data -13
• run multifit cpu ./multifit_cpu/multifit_cpu ../data/<filename> <num_iterations> <num_channels_per_iteration>
• run multifit gpu ./multifit_gpu/multifit_gpu ../data/<filename> <num_iterations> <num_channels_per_iteration>

• compile with -g and optimizations
• run valgrind --tool=cachegrind <program> specifying the program to be profiled
• run KCachegrind with on the file produced by valgrind to visualize the results

• gen: test data generator.
• io: input data parser.
• legacy_multifit_cpu: old cms_sw hcal/ecal code.
• legacy_multifit_cpu: gpu porting of old cms_sw hcal/ecal code.
• multifit_cpu: cms_sw hcal/ecal code updated with the new version of nnls/fnnls.
• multifit_gpu: gpu porting of multifit_cpu
• nnls: CPU/GPU version of nnls/fnnls and inplace fnnls by Marco.
• test_nnls_cpu: tests for nnls and io.

• source /data/user/vkhriste/setup.sh
• proceed to build and run