<<<<<<< HEAD:README

/*NOTES(tony): This help file is prepared in a rash, please feel free to contact me with any difficulties you faced.*/
*******bin**********
To benchmark on your machine,
	1. go to bin/release run handshake.bat if you are run it first time on new machine.
	2. go to each *scheduler folder, run test.bat
To modify-run each level of scheduling, you need to do modify in src folder and then:
	1. you need to compile the KernelEngine into DLL,Copy the lib into CoProcessores TonyLib.(If you never done so or made modification on kernelEngine project). 
	3. Compile CoProcessor, copy the exe to bin/release folder described in previous section.
/*NOTES*/
	current code rely on AMD's opencl SDK on windows, you might need to recompile it with other vendor's Opencl sdk, i.e intel's or Nvidia's
*******src code description**************
Three level of scheduling are put into three folder correspondingly, 
from each source folder, you will find: 
	1. CoProcessor folder : which contains the main entry point for certain level scheduling program. 
	2. KernelEngine folder: code for the engine, compile into DLL.
/*SETTINGS*/
Total query numbers are now hard-code set to *30* in order of Q1 to Q6.
Total number of threads are now hard-code set to *4*.
I will update them into easy-benchmark version later.

/*NOTES*/
Do be aware that, if parallism is too high(say, 10 true concrrent threads), GPU memory may not enough to buffering requests,
which will lead program crash. This bug will be addressed later.