From cwou@npac.syr.edu Mon Jul 14 20:56:11 1997
Date: Mon, 07 Jul 1997 14:37:28 -0400
From: Chao-Wei Ou <cwou@npac.syr.edu>
To: Paul Coddington <paulc@postoffice.npac.syr.edu>
Cc: cwou@dante.npac.syr.edu
Subject: Re: status 



On 7 Jul 1997 14:8:2, Paul Coddington wrote: 

> 
> How are things with the NT port? 

Still stuck at the same memory allocation problem on quad-cpu NT.
Currently, no solution is found. But I'll try to grab more help
from others to sove it.

> 
> Can you give me a definition of the new API for running the MC programs.

If you expect the answer according to the last discussion about running
varable time steps, this would need to reimplement the kernel of MC.
I need to reimplement the part to generate samples. This would be a
significant change to Miloje's code, or I should say a new MC code.
Before I implement this feature, I need you to verify the probablity
function. 

Basically, the API will be the following new data structure.

typedef struct mc_interface {
    int   NumberOfRates;             /* Total number of rates */
    char  ProxyRatesContainer[80];   /* Output proxy rate file */
    int   NumberOfCycles;            /* Number of MC cycle */
    int   RNGSeed;                   /* Random number seed */
    int   NumberOfMeasurementBlocks; /* Number of measurement blocks */
    int   SizeOfMeasurementBlock;    /* Number of trials in a block */
    int   NumberOfRelaxBlocks;       /* NUmber of relaxation blocks */
    REAL  *InitialValues;            /* Initial values for rates */
    int   NumberOfDiffTimeSteps;     /* Number of different time steps */
    TIMESTEPS *timestep; /* pointer of variable time steps with size
	                    NumberOfDiffTimeSteps */
} MC_INTERFACE;

typedef struct timesteps {
        int     unit;   /* 1|2|3|4 represent dayly|weekly|monthly|yearly */
        int     upper;  /* the upper bound for this time interval */
        int     lower;  /* the lower bound for this time interval */
        DATE    start_date;     /* Starting date of this time interval */
        DATE    end_date;       /* Ending date of this time interval */
        int     distribution;   /* choice of the distribution */
        float   variance;       /* associated variance */
        float   **correlation;  /* associated correlation matrix */
} TIMESTEPS;

typedef struct date {
        int     day;    /* the input day of the month */
        int     month;  /* the input month of the year */
        int     year;   /* the input year */
} DATE;


> Have you implemented a mechanism for returning the status (e.g. number of
> iterations)?
> 

The return status for number of loops is done. All I have done just adding
a print statement in the MC loops to indicate the current loop.