# README
#     for learnOpenCL
#     Copyright (C) 2015 Steve Novakov
#
#     This program is free software; you can redistribute it and/or modify
#     it under the terms of the GNU General Public License as published by
#     the Free Software Foundation; either version 2 of the License, or
#     (at your option) any later version.
#
#     This program is distributed in the hope that it will be useful,
#     but WITHOUT ANY WARRANTY; without even the implied warranty of
#     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#     GNU General Public License for more details.
#
#     You should have received a copy of the GNU General Public License along
#     with this program; if not, write to the Free Software Foundation, Inc.,
#     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
#
# NOTE: MAKE SURE ALL INDENTS IN MAKEFILE ARE **TABS** AND NOT SPACES
#

#
# COMPILATION
#

Before attempting compilation, please make sure that:
	- a working OpenCL environment is installed. Installing the relevant hardware drivers and either the NVIDIA CUDA SDK or the AMD APP SDK will satisfy this requirement.
  - that cl.hpp has been copied over from the khronos site into /usr/local/cuda/include. The following commands should do it
    $ cd /usr/local/cuda/include/CL
    $ sudo wget https://www.khronos.org/registry/cl/api/1.1/cl.hpp
    (or whatever your supported version of OpenCL is )
  - that the system variable LD_LIBRARY_PATH has an entry for the path to libOpenCL.so.1
  - that the system variables C_INCLUDE_PATH and CPLUS_INCLUDE_PATH have entries for the path to cl.h/cl.hpp (for nvidia it's /usr/local/cuda/include).

You can compile non debug builds with just
$ make
and debug builds with
$ make debug
run in the root repository directory.

Once compiled you can check that the correct linkage has occured by running

$ ldd program
  linux-vdso.so.1 =>  (0x00007ffe72ab4000)
  libOpenCL.so.1 => /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 (0x00007f5a13d13000)
  libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f5a13a0f000)
  libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f5a13708000)
  libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f5a134f2000)
  libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f5a132d4000)
  libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f5a12f0e000)
  libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f5a12d0a000)
  /lib64/ld-linux-x86-64.so.2 (0x00007f5a13f42000)

for example.

#
# Execution
#

Execute the program with the default settings (data size 100MB, chunk suze 10MB)
by simply running

$ ./program

You can specify total data size and chunk size (or average chunk size, depending
on branch) with the following command line arguments

$ ./program -datasize=2000 -chunksize=10 -gpus=0,1

where, in that example, the total input/output arrays are 2GB in size, with a
chunk size of 10MB (do not recommend this, will just take long for no reason,
recommend chunk sizes ~1/10-1/5 of the data size for a quick but representative
set of data). Make sure that the gpu indeces listed in -gpus are valid (i.e.
that ocl_devices.at(x_i) exists for every x in -gpus=x0,..,xN).

If -gpus is empty, or not passed, all gpus available in the opencl context
will be used.

#
# OpenCL Profiling
#

Just run the profiling.sh script included in this repo
with any command line arguments you wish to pass to the program
$ ./profiling.sh ./program -datasize=2000 -chunksize=10 -gpus=0,1

to view it then run

$ nvvp cuda_profile.log

the logging configuration file and the sed file are included in this repo.