Q6: step 3

Q6-cache/info.txt

@@ -12,6 +12,7 @@
 
 Βήμα 1: Λόγος διάμεσων χρόνων: 20.265193309
 
+Βήμα 2: Λόγος διάμεσων χρόνων: 6.719256593 (s=8)
 Βήμα 2: Λόγος διάμεσων χρόνων: 24.61113257 (s=1024)
 
-Βήμα 3: Λόγος διάμεσων χρόνων: xxxx
+Βήμα 3: Λόγος διάμεσων χρόνων: 4.667896091 (s=8)

Q6-cache/src/main.c

@@ -49,6 +49,7 @@ mMult_ft multSelect (char* order) {
    else if (! strcmp ((const char*)order, "jki"))  return matrixMult_jki;
    else if (! strcmp ((const char*)order, "kij"))  return matrixMult_kij;
    else if (! strcmp ((const char*)order, "kji"))  return matrixMult_kji;
+   else if (! strcmp ((const char*)order, "ikj8")) return matrixMult_ikj8;
    else                                            return matrixMult_ijk;
 }
 

Q6-cache/src/matmul.c

@@ -273,6 +273,44 @@ void matrixMult_kji_block(float * const C, float const * const A, float const *
                      C[ sub2ind(I+i,J+j,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+j,n) ];
 }
 
+/*!
+ * Square Matrix multiplication in unrolling block of 8 - ikj
+ * \param   C  pointer to output matrix
+ * \param   A  pointer to input matrix A
+ * \param   B  pointer to input matrix B
+ * \param   n  Size of matrices (both sizes)
+ * \param   s  The block size
+ * \return     none
+ */
+void matrixMult_ikj8(float * const C, float const * const A, float const * const B, int const n) {
+   for (int I =0; I<n; I +=8)
+      for (int K =0; K<n; K +=8)
+         for (int J = 0; J<n; J +=8)
+            for (int i =0; i<8; ++i)
+               for (int k =0; k<8; ++k) {
+                  if ((k+K) == 0) {
+                     C[ sub2ind(I+i,J+0,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+0,n) ];
+                     C[ sub2ind(I+i,J+1,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+1,n) ];
+                     C[ sub2ind(I+i,J+2,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+2,n) ];
+                     C[ sub2ind(I+i,J+3,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+3,n) ];
+                     C[ sub2ind(I+i,J+4,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+4,n) ];
+                     C[ sub2ind(I+i,J+5,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+5,n) ];
+                     C[ sub2ind(I+i,J+6,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+6,n) ];
+                     C[ sub2ind(I+i,J+7,n) ] = A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+7,n) ];
+                  }
+                  else {
+                     C[ sub2ind(I+i,J+0,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+0,n) ];
+                     C[ sub2ind(I+i,J+1,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+1,n) ];
+                     C[ sub2ind(I+i,J+2,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+2,n) ];
+                     C[ sub2ind(I+i,J+3,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+3,n) ];
+                     C[ sub2ind(I+i,J+4,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+4,n) ];
+                     C[ sub2ind(I+i,J+5,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+5,n) ];
+                     C[ sub2ind(I+i,J+6,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+6,n) ];
+                     C[ sub2ind(I+i,J+7,n) ] += A[ sub2ind(I+i,K+k,n) ] * B[ sub2ind(K+k,J+7,n) ];
+                  }
+               }
+}
+
 /*!
  * Initialize matrix with random indices and return the matrix pointer.
  *

Q6-cache/src/matmul.h

@@ -33,6 +33,7 @@ void matrixMult_jki_block(float * const C, float const * const A, float const *
 void matrixMult_kij_block(float * const C, float const * const A, float const * const B, int const n, int const s);
 void matrixMult_kji_block(float * const C, float const * const A, float const * const B, int const n, int const s);
 
+void matrixMult_ikj8(float * const C, float const * const A, float const * const B, int const n);
 
 float* matrixInit(int const n);