Squashed commit of the sb-vbs branch.

Includes the SD-VBS benchmarks modified to:
- Use libextra to loop as realtime jobs
- Preallocate memory before starting their main computation
- Accept input via stdin instead of via argc

Does not include the SD-VBS matlab code.

Fixes libextra execution in LITMUS^RT.

1 files changed, 188 insertions, 0 deletions

diff --git a/SD-VBS/benchmarks/svm/src/c/takeStep.c b/SD-VBS/benchmarks/svm/src/c/takeStep.c
new file mode 100644
index 0000000..ae550c2
--- /dev/null
+++ b/SD-VBS/benchmarks/svm/src/c/takeStep.c
@@ -0,0 +1,188 @@
+/********************************
+Author: Sravanthi Kota Venkata
+********************************/
+
+#include "svm.h"
+
+int takeStep(int i, int j, F2D* a, float C, F2D* e, F2D* Y, F2D* X, float eps, float* b, int N, int dim)
+{
+    int ret=1;
+    float s;
+    int m, n, k;
+    float Ei, Ej, gamma, L, H;
+    F2D *a_old;
+    float k11, k12, k22, eta;
+    F2D *temp, *temp1, *temp2;
+    float t, t1, t2;
+    float bnew, delta_b;
+    float c1, c2, Lobj, Hobj;
+    
+    if(i==j)
+        return 0;
+        
+    a_old = fDeepCopy(a);
+    
+    if( asubsref(a_old,i)>0 && asubsref(a_old,i)<C ) 
+        Ei = asubsref(e,i);
+    else
+        Ei = cal_learned_func(i, a, b, N, Y, X, dim) - asubsref(Y,i);
+
+    if( asubsref(a_old,j)>0 && asubsref(a_old,j)<C ) 
+        Ej = asubsref(e,j);
+    else
+        Ej = cal_learned_func(j, a, b, N, Y, X, dim) - asubsref(Y,j);
+
+    s = asubsref(Y,i) * asubsref(Y,j);
+
+    if( asubsref(Y,i) == asubsref(Y,j) )
+    {
+        gamma = asubsref(a_old,i) + asubsref(a_old,j);
+        if(gamma > C)
+        {
+            L = gamma - C;
+            H = C;
+        }
+        else
+        {
+            L = 0;
+            H = gamma;
+        }
+    }
+    else
+    {
+        gamma = asubsref(a_old,i) - asubsref(a_old,j);
+        if(gamma > 0)
+        {
+            L = 0;
+            H = C - gamma;
+        }
+        else
+        {
+            L = -gamma;
+            H = C;
+        }
+    }
+    
+
+    if(L==H)
+    {
+        fFreeHandle(a_old);
+        return 0;
+    }
+           
+    temp = fMallocHandle(1, X->width);
+    temp1 = fMallocHandle(1, X->width);
+
+    for(m=0; m<X->width; m++)
+    {
+        asubsref(temp,m) = subsref(X,i,m);
+        asubsref(temp1,m) = subsref(X,j,m);
+    }
+
+    k11 = polynomial(3, temp, temp, dim);
+    k12 = polynomial(3, temp, temp1, dim);
+    k22 = polynomial(3, temp1, temp1, dim);
+    eta = 2 * k12 - k11 - k22;
+
+    fFreeHandle(temp1);
+    fFreeHandle(temp);
+
+    if(eta<0)
+    {
+        asubsref(a,j) = asubsref(a_old,j) + asubsref(Y,j) * (Ej-Ei)/eta;
+        if( asubsref(a,j) < L)
+            asubsref(a,j) = L;
+        else if( asubsref(a,j) > H )
+            asubsref(a,j) = H;
+    }
+    else
+    {
+        c1 = eta/2;
+        c2 = asubsref(Y,j) * (Ei-Ej) - eta * asubsref(a_old,j);
+        Lobj = c1 * L * L + c2 * L;
+        Hobj = c1 * H * H + c2 * H;
+
+        if (Lobj > (Hobj+eps))
+            asubsref(a,j) = L;
+        else if (Lobj < (Hobj-eps))
+            asubsref(a,j) = H;
+        else
+            asubsref(a,j) = asubsref(a_old,j);
+    }
+    
+    if( fabsf( asubsref(a,j)- asubsref(a_old,j) ) < (eps* (asubsref(a,j) + asubsref(a_old,j) +eps)) )
+    {
+        fFreeHandle(a_old);
+        return 0;
+    }
+
+    asubsref(a,i) = asubsref(a_old,i) - s * ( asubsref(a,j) - asubsref(a_old,j) );
+
+    if( asubsref(a,i) < 0)
+    {
+        asubsref(a,j) = asubsref(a,j) + s * asubsref(a,i);
+        asubsref(a,i) = 0;
+    }
+    else if (asubsref(a,i) > C)
+    {
+        t = asubsref(a,i) - C;
+        asubsref(a,j) = asubsref(a,j) + s * t;
+        asubsref(a,i) = C;
+    }
+
+    /** Update threshold to react change in Lagrange multipliers **/
+
+    if( asubsref(a,i) > 0 && asubsref(a,i) < C )
+        bnew = arrayref(b,0) + Ei + asubsref(Y,i) * (asubsref(a,i) - asubsref(a_old,i)) * k11 + asubsref(Y,j) * (asubsref(a,j) - asubsref(a_old,j)) * k12;
+    else
+    {
+        if( asubsref(a,j) > 0 && asubsref(a,j) < C )
+            bnew = arrayref(b,0) + Ej + asubsref(Y,i) * (asubsref(a,i) - asubsref(a_old,i)) * k12 + asubsref(Y,j) * (asubsref(a,j) - asubsref(a_old,j)) * k22;
+        else
+        {
+            float b1, b2;
+            b1 = arrayref(b,0) + Ei + asubsref(Y,i) * (asubsref(a,i) - asubsref(a_old,i)) * k11 + asubsref(Y,j) * (asubsref(a,j) - asubsref(a_old,j)) * k12;
+            b2 = arrayref(b,0) + Ej + asubsref(Y,i) * (asubsref(a,i) - asubsref(a_old,i)) * k12 + asubsref(Y,j) * (asubsref(a,j) - asubsref(a_old,j)) * k22;
+            bnew = (b1 + b2) / 2;
+        }
+    }
+    delta_b = bnew - arrayref(b,0);
+    arrayref(b,0) = bnew;
+
+    /** Update error cache using new Lagrange multipliers 24ai **/
+
+    t1 = asubsref(Y,i) * (asubsref(a,i)-asubsref(a_old,i));
+    t2 = asubsref(Y,j) * (asubsref(a,j)-asubsref(a_old,j));
+
+    temp = fMallocHandle(1, X->width);
+    temp1 = fMallocHandle(1, X->width);
+    temp2 = fMallocHandle(1, X->width);
+
+    for (k=0; k<N; k++)
+    {
+        if (0 < asubsref(a_old,i) && asubsref(a_old,i) < C )
+        {
+
+            for(m=0; m<X->width; m++)
+            {
+                asubsref(temp,m) = subsref(X,i,m);
+                asubsref(temp1,m) = subsref(X,k,m);
+                asubsref(temp2,m) = subsref(X,j,m);
+            }
+
+            asubsref(e,k) = asubsref(e,k)+t1 * polynomial(3, temp, temp1, dim) + t2 * polynomial(3, temp2, temp1, dim) - delta_b;
+            asubsref(e,i) = 0;
+            asubsref(e,j) = 0;
+        }
+    }
+    
+    fFreeHandle(a_old);
+    fFreeHandle(temp);
+    fFreeHandle(temp1);
+    fFreeHandle(temp2);
+    ret = 1;
+    return ret;
+}
+
+
+