C2H Compilation Error - Explicit Inclusion

You aren't going to be able to convert a call to malloc() into VHDL!

A general rule of thumb: don't rely on standard C libraries for anything you accelerate. Even if they were supported they would probably result in really big and slow hardware. The closest you can get is call malloc from outside the function and then pass in the pointer allocated by malloc to the accelerator.

Good points made dsl/BadOmen. I ended up rewriting the function so it only contains the math/recursive operators and no calls to external functions.

Well have scrubbed out all function references in the code I am accelerating. Now, getting a new error which occurs when Quartus tried to compile the C2H-generated VHDL file (C2H compiler runs successfully):

 
int DTW_calcVectorCost_mxn_C2H(int **x, int xsize, int **y, int ysize, int nSize, int **Dist, int **globdist, int **move, int **temp, int **warp)
 
{
 
/* DESCRIPTION */
/* Compute a distance matrix on 2 multi-parameter vectors
and perform dynamic time warping on the distance matrix */
 
int top, mid, bot, cheapest, total;
int I, X, Y, n, i, j, k, cost;
int params = nSize;
 
/*Compute distance matrix*/
for(i=0;i<xsize;i++) {
for(j=0;j<ysize;j++) {
total = 0;
for (k=0;k<params;k++) {
total = total + ((x - y) * (x - y));
}
Dist = total;
}
}
 
/*% for first frame, only possible match is at (0,0)*/
globdist = Dist;
for (j=1; j<xsize; j++)
globdist = VERY_BIG_C2H;
 
globdist = VERY_BIG_C2H;
globdist = globdist + Dist;
move = 2;
 
for(j=2;j<xsize;j++)
globdist = VERY_BIG_C2H;
 
for(i=2;i<ysize;i++) {
globdist = VERY_BIG_C2H;
globdist = globdist + Dist;
 
for(j=2;j<xsize;j++) {
top = globdist + Dist + Dist;
mid = globdist + Dist;
bot = globdist + Dist + Dist;
if( (top < mid) && (top < bot))
{
cheapest = top;
I = 1;
}
elseif (mid < bot)
{
cheapest = mid;
I = 2;
}
else {cheapest = bot;
I = 3;
}
 
/*if all costs are equal, pick middle path*/
if( ( top == mid) && (mid == bot))
I = 2;
 
globdist = cheapest;
move = I;
}
}
 
X = ysize-1; Y = xsize-1; n=0;
warp = X; warp = Y;
 
while (X > 0 && Y > 0) {
n=n+1;
 
if (n>ysize *2)
{
return VERY_BIG_C2H; // Return default cost (large) if failed.
// exit(1);
}
 
if (move  == 1 )
{
warp = X-1; warp = Y;
n=n+1;
X=X-2; Y = Y-1;
}
elseif (move  == 2)
{
X=X-1; Y = Y-1;
}
elseif (move  == 3 )
{
warp  = X;
warp  = Y-1;
n=n+1;
X=X-1; Y = Y-2;
}
else
{
return VERY_BIG_C2H; // Return default cost (large) if failed.
}
warp  =X;
warp  =Y;
 
}
 
/*flip warp*/
for (i=0;i<=n;i++) {
temp = warp;
temp = warp;
 
}
 
for (i=0;i<=n;i++) {
warp = temp;
warp = temp;
 
}
 
// Cast 'globdist' as int 'cost' (to be returned)
cost = globdist;
return cost;
}

Just some notes:

When you create a global variable used by the accelerated function C2H will create a master to access that value. Just like any other multi-core design, sharing data via global variables can be tricky. There are some corner cases of global variables that are documented that you should refer to in the C2H documentation.

Also make sure to read this doc, I tried to shovel as much information as possible into it: http://www.altera.com/literature/hb/nios2/edh_ed51005.pdf

Appreciate all your input BadOmen - have printed everything out from the PDF and have some good reading ahead of me!

I'm not familiar with VHDL (I switched to verilog long ago and have never looked back :)). By the sounds of it the accelerated HDL happens to contain a previously declared module/name called "IO". Just in case there are old files being left behind one thing you could try is deleting all the generated HDL files and the /db directory for your project, and then regenerate to see if that will compile.

Thanks BadOmen, yes will look into what the error is in the VHDL syntax.