[How to set up]
1. Replace the original versions of prog/score_sim.m and prog/score_real.m with the new versions 
   attached. Note that these new versions invoke another matlab process with the 'matlab' command. 
   If your matlab executable has a different name, please change the 'matlabcmd' variable value in 
   these M files. Also make sure that your matlab path is included in the search path. 

2. Store SRMR_run.m in the 'SRMRtoolbox-ReverbChallenge' directory. 


[Important note]
Before using these M files, please make sure that you can run (at least) two matlab processes 
simultaneously on your machine since these M files generate a separate matlab process. 


[Brief description]
Here is a brief description of how this updated version avoids the garbage accumulation problem that
happens with the original evaluation tools in some matlab versions. Our SRMR function is provided 
in the p-code form. Although we are not absolutely sure, it appears that the memory space consumed 
by a p-code is not released until the matlab returns to the interactive mode. Neither the clear 
function nor garbage collection works appropriately during batch processing. 

To provide a quick (and less elegant) fix to this problem, whenever the SRMR p-code is invoked, this
updated version generates a new matlab process to execute the p-code and stores the result in a 
temporary file. This matlab process is terminated immediately, and therefore the memory space used 
by the p-code is release after the p-code is evaluated. By doing so, this updated version avoids 
the garbage from being accumulated. Please note that, this version takes a longer time to complete 
the entire evaluation because it launches matlab processes many times. 

