Q6: step 1

Q6-cache/Makefile

@@ -86,8 +86,8 @@ include $(wildcard $(DEP))
 $(BUILD_DIR)/$(TARGET): $(OBJ)
 	@mkdir -p $(@D)
 	@echo Linking to target: $(TARGET)
-	@echo $(CXX) $(LDFLAGS) -o $(@D)/$(TARGET) '$$(OBJ)'
-	@$(CXX) $(LDFLAGS) -o $(@D)/$(TARGET) $(OBJ)
+#@echo $(CXX) $(LDFLAGS) -o $(@D)/$(TARGET) '$$(OBJ)'
+	$(CXX) -o $(@D)/$(TARGET) $(OBJ) $(LDFLAGS) 
 	@echo
 	@echo Print size information
 	@$(CSIZE) $(@D)/$(TARGET)

Q6-cache/info.txt

@@ -1,9 +1,17 @@
 Επεξεργαστής: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
 Συχνότητα επεξεργαστή: 800 - 2600 MHz
-Μέγεθος μνήμης L1: 192 KiB / 192 KiB (L1d/L1i)
+Μέγεθος μνήμης L1: 192 KiB / 192 KiB
 Μέγεθος μνήμης L2: 1.5 MiB
 Μέγεθος μνήμης L3: 12 MiB
-Μέγεθος μνήμης RAM: 31893MiB
+Μέγεθος μνήμης RAM: 31893 MiB
 Έκδοση gcc: 9.2.1 20191008 (Ubuntu 9.2.1-9ubuntu2)
 Έκδοση Λειτουργικού Συστήματος: Ubuntu 19.10 x86_64
 
+Διάμεσος χρόνος εκτέλεσης: 50984.906000 [msec]
+Τυπική απόκλιση: 221.533052 [msec]
+
+Βήμα 1: Λόγος διάμεσων χρόνων: 20.265193309
+
+Βήμα 2: Λόγος διάμεσων χρόνων: xxxx
+
+Βήμα 3: Λόγος διάμεσων χρόνων: xxxx

Q6-cache/src/main.c

@@ -0,0 +1,110 @@
+/*!
+  \file   main.c
+  \brief  Matrix multiplication implementation.
+
+  \author Nikos Pitsianis
+  \author Dimitris Floros
+  \author Christos Choutouridis 8997 <cchoutou@ece.auth.gr>
+  \date   2020-05-05
+*/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#include <sys/time.h>
+#include <assert.h>
+
+#include "matmul.h"
+
+extern double sqrt (double);
+
+int cmpfunc (const void * a, const void * b) {
+   double v =*(double*)a - *(double*)b;
+   return (v < 0) ? -1 : (v > 0) ? 1 : 0;
+}
+
+double median (double* t, size_t n) {
+   qsort ((void*)t, n, sizeof(t[0]), cmpfunc);
+   return (n % 2) ? t[n/2] : (t[n/2] + t[n/2 -1]) /2;
+}
+
+double std_deviation (double* t, size_t n) {
+   double av =0;
+   for (size_t i=0 ; i<n ; ++i) {
+      av += t[i];
+   }
+   av /= n;
+   double s =0;
+   for (size_t i=0 ; i<n ; ++i) {
+      s += (t[i]-av)*(t[i]-av);
+   }
+   return sqrt (s/n);
+}
+
+mMult_ft multSelect (char* order) {
+   if (! strcmp ((const char*)order, "ijk"))       return matrixMult_ijk;
+   else if (! strcmp ((const char*)order, "ikj"))  return matrixMult_ikj;
+   else if (! strcmp ((const char*)order, "jik"))  return matrixMult_jik;
+   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                                            return matrixMult_ijk;
+}
+
+
+
+/*!
+ * A unit testing like main function to profile our code
+ */
+int main(int argc, char **argv) {
+  struct timeval start, end;     /* time structs */
+  double time[MAX_ITER] = {0.0}; /* execution time array in ms */
+  float *A, *B, *C;              /* matrix declarations */
+
+  /* read matrix size (or use default) */
+  if (argc != 3){
+    fprintf( stderr,
+          "Usage:\n"
+          "%s n order, where \n"
+          "   n: is the matrix size.\n"
+          "   order: the loop order ex: ijk , jik, ...\n",
+             argv[0]);
+    exit(1);
+  }
+  int n = atoi( argv[1] );
+  mMult_ft  mMult =  multSelect(argv[2]);
+
+  /* initialize matrices */
+  A = matrixInit( n );
+  B = matrixInit( n );
+  C = (float *) malloc( n*n*sizeof(float) );
+
+  /* compute matrix multiplication */
+  for (int it = 0; it < MAX_ITER; it++) {
+    gettimeofday(&start, NULL);
+    mMult( C, A, B, n );
+    gettimeofday(&end, NULL);
+
+    time[it] =  (end.tv_sec  - start.tv_sec)  * 1000.0 +    /* sec to ms */
+                (end.tv_usec - start.tv_usec) / 1000.0;     /* us to ms */
+
+    printf("Iter: %d Time: %f ms\n", it, time[it]);
+  }
+
+  /* we need to use the result -- verify it */
+  for (int i = 0; i < n; i++) {             /* rows */
+    for (int j = 0; j < n; j++) {           /* cols */
+      float gold = 0;
+      for (int k = 0; k < n; k++) {         /* accumulate products */
+        gold += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      }
+      assert( (gold - C[sub2ind(i,j,n)]) < 1e-3 );
+    }
+  }
+
+  // statistical data
+  printf ("Median: %f [msec]\n", median (time, MAX_ITER));
+  printf ("Std.Dev: %f [msec]\n", std_deviation(time, MAX_ITER));
+}
+

Q6-cache/src/matmul.c

@@ -7,48 +7,139 @@
   \author Christos Choutouridis 8997 <cchoutou@ece.auth.gr>
   \date   2020-05-05
 */
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <math.h>
-#include <sys/time.h>
-#include <assert.h>
-
-#define MAX_ITER 10
-#define sub2ind(i,j,n) (j) + (i)*(n)
+#include "matmul.h"
 
 /*!
- * Square Matrix multiplication
+ * Square Matrix multiplication - ijk
  * \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)
  * \return 	   none
- *
- * \note
- * 		This version executes row major order ijk
  */
-void matrixMult(float * const C, float const * const A, float const * const B, int const n) {
+void matrixMult_ijk(float * const C, float const * const A, float const * const B, int const n) {
+   for (int i = 0; i < n; ++i)
+      for (int j = 0; j < n; ++j) {
+         int k =0;
+         C[ sub2ind(i,j,n) ] = A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+         for (k = 1; k < n; ++k)
+            C[ sub2ind(i,j,n) ] += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      }
+}
 
-   for (int i = 0; i < n; i++) {          /* rows */
-      for (int j = 0; j < n; j++) {       /* cols */
-         C[ sub2ind(i,j,n) ] = 0;         /* initialize output value */
-         for (int k = 0; k < n; k++) {    /* accumulate products */
-            C[ sub2ind(i,j,n) ] +=
-            A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+/*!
+ * Square Matrix multiplication - 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)
+ * \return     none
+ */
+void matrixMult_ikj(float * const C, float const * const A, float const * const B, int const n) {
+   for (int i = 0; i < n; ++i)
+      for (int k = 0; k < n; ++k) {
+         if (!k) {
+            for (int j = 0; j < n; ++j)
+               C[ sub2ind(i,j,n) ] = A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+         } else {
+            for (int j = 0; j < n; ++j)
+               C[ sub2ind(i,j,n) ] += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
          }
       }
+}
+
+/*!
+ * Square Matrix multiplication - jik
+ * \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)
+ * \return     none
+ */
+void matrixMult_jik(float * const C, float const * const A, float const * const B, int const n) {
+   for (int j = 0; j < n; j++)
+      for (int i = 0; i < n; i++) {
+         int k =0;
+         C[ sub2ind(i,j,n) ] = A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+         for (k = 1; k < n; k++)
+            C[ sub2ind(i,j,n) ] += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      }
+}
+
+/*!
+ * Square Matrix multiplication - jki
+ * \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)
+ * \return     none
+ */
+void matrixMult_jki(float * const C, float const * const A, float const * const B, int const n) {
+   for (int j = 0; j < n; ++j)
+      for (int k = 0; k < n; ++k) {
+         if (!k) {
+            for (int i = 0; i < n; ++i)
+               C[ sub2ind(i,j,n) ] = A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+         } else {
+            for (int i = 0; i < n; ++i)
+               C[ sub2ind(i,j,n) ] += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+         }
+      }
+}
+
+/*!
+ * Square Matrix multiplication - kij
+ * \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)
+ * \return     none
+ */
+void matrixMult_kij(float * const C, float const * const A, float const * const B, int const n) {
+   for (int k = 0; k < n; ++k) {
+      if (!k) {
+         for (int i = 0; i < n; ++i)
+            for (int j = 0; j < n; ++j)
+               C[ sub2ind(i,j,n) ] = A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      } else {
+         for (int i = 0; i < n; ++i)
+            for (int j = 0; j < n; ++j)
+               C[ sub2ind(i,j,n) ] += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      }
    }
+}
 
+/*!
+ * Square Matrix multiplication - kji
+ * \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)
+ * \return     none
+ * xxx
+ */
+void matrixMult_kji(float * const C, float const * const A, float const * const B, int const n) {
+   for (int k = 0; k < n; ++k) {
+      if (!k) {
+         for (int j = 0; j < n; ++j)
+            for (int i = 0; i < n; ++i)
+               C[ sub2ind(i,j,n) ] = A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      } else {
+         for (int j = 0; j < n; ++j)
+            for (int i = 0; i < n; ++i)
+               C[ sub2ind(i,j,n) ] += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
+      }
+   }
 }
 
+
 /*!
  * Initialize matrix with random indices and return the matrix pointer.
  *
  * \param   n  The size of the matrix (both of them)
  * \return     Pointer to allocated and initialized matrix
  */
-float * matrixInit(int const n) {
+float* matrixInit(int const n) {
 
    float *M = (float *) malloc( n*n*sizeof(float) );
 
@@ -59,67 +150,5 @@ float * matrixInit(int const n) {
    return M;
 }
 
-int cmpfunc (const void * a, const void * b) {
-   double v =*(double*)a - *(double*)b;
-   return (v < 0) ? -1 : (v > 0) ? 1 : 0;
-}
 
-/*!
- * A unit testing like main function to profile our code
- */
-int main(int argc, char **argv)
-{
-
-  struct timeval start, end;     /* time structs */
-  double time[MAX_ITER] = {0.0}; /* execution time array in ms */
-  float *A, *B, *C;              /* matrix declarations */
-  int n;                         /* matrix size */
-
-  /* read matrix size (or use default) */
-  if (argc != 2){
-    fprintf( stderr, "Usage:\n %s n\n   where n is the matrix size.\n",
-             argv[0]);
-    exit(1);
-  }
-  n = atoi( argv[1] );
-  
-  /* initialize matrices */
-  A = matrixInit( n );
-  B = matrixInit( n );
-  C = (float *) malloc( n*n*sizeof(float) );
-  
-  /* compute matrix multiplication */
-  for (int it = 0; it < MAX_ITER; it++) {
-    gettimeofday(&start, NULL);
-    matrixMult( C, A, B, n );
-    gettimeofday(&end, NULL);
-
-    time[it] =  (end.tv_sec  - start.tv_sec)  * 1000.0 +    /* sec to ms */
-                (end.tv_usec - start.tv_usec) / 1000.0;     /* us to ms */
-               
-    printf("Iter: %d Time: %f ms\n", it, time[it]);
-  }
-
-  /* we need to use the result -- verify it */
-  for (int i = 0; i < n; i++) {             /* rows */
-    for (int j = 0; j < n; j++) {           /* cols */
-
-      float gold = 0;
-
-      for (int k = 0; k < n; k++) {         /* accumulate products */
-        gold += A[ sub2ind(i,k,n) ] * B[ sub2ind(k,j,n) ];
-      }
-
-      assert( (gold - C[sub2ind(i,j,n)]) < 1e-3 );
-      
-    }
-  }
-
-  // median calculation
-  qsort ((void*)time, MAX_ITER, sizeof(time[0]), cmpfunc);
-  printf("Median: %f [msec]\n", (MAX_ITER % 2) ?
-        time[MAX_ITER/2] :
-        (time[MAX_ITER/2] + time[MAX_ITER/2 -1]) /2
-  );
-}
 

Q6-cache/src/matmul.h

@@ -0,0 +1,30 @@
+/*!
+  \file   matmul.h
+  \brief  Matrix multiplication implementation.
+
+  \author Nikos Pitsianis
+  \author Dimitris Floros
+  \author Christos Choutouridis 8997 <cchoutou@ece.auth.gr>
+  \date   2020-05-05
+*/
+#ifndef SRC_MATMUL_H_
+#define SRC_MATMUL_H_
+
+#include <stdlib.h>
+
+#define MAX_ITER 10
+#define sub2ind(i,j,n) (j) + (i)*(n)
+
+//! Function pointer type to matrix multiplication back-end.
+typedef void (*mMult_ft)(float * const C, float const * const A, float const * const B, int const n);
+
+void matrixMult_ijk(float * const C, float const * const A, float const * const B, int const n);
+void matrixMult_ikj(float * const C, float const * const A, float const * const B, int const n);
+void matrixMult_jik(float * const C, float const * const A, float const * const B, int const n);
+void matrixMult_jki(float * const C, float const * const A, float const * const B, int const n);
+void matrixMult_kij(float * const C, float const * const A, float const * const B, int const n);
+void matrixMult_kji(float * const C, float const * const A, float const * const B, int const n);
+
+float* matrixInit(int const n);
+
+#endif /* SRC_MATMUL_H_ */