Added cache testing to valgrind example

valgrind/Readme.rst

@@ -1,3 +1,40 @@
-Run this line to see errors in this code. 
+Compile each program without optimization first.
+
+For simpleTest.cc, run this line to see errors in this code. 
+
+::
+
+  valgrind --track-origins=yes --leak-check=full ./simpleTest 300 300
+
+
+We also have a cache test line. Run this line to see the cache errors.
+
+::
+
+  valgrind --tool=cachegrind ./a.out 0 1000 100000
+
+There are two paths in this code. If the first input is 1, it runs a cache-sensitive version of the loop. 
+If it is 0, it runs a cache-insensitive version.
+
+FYI: on the Trieste lab machines, this is what cache looks like:
+
+::
+
+  guy ~>dmesg | grep cache
+  CPU: L1 I cache: 32K, L1 D cache: 32K
+  CPU: L2 cache: 6144K
+  CPU: L1 I cache: 32K, L1 D cache: 32K
+  CPU: L2 cache: 6144K
+
+You can run the same command to see cache on your linux machine. Another way to see the exact cache setup that 
+valgrind found is the following:
+
+::
+
+  cg_annotate --auto=yes cachegrind.out.21960
+
+Note that your cachegrind.out will have a different number. This command is also handy because it shows which functions caused cache
+misses.
+
+
 
-valgrind --track-origins=yes --leak-check=full ./a.out

valgrind/cacheTest.cc

@@ -7,23 +7,34 @@ Tests cache misses.
 
 int main(int argc, char **argv)
 {
-  if (argc < 2){
-    printf("Usage: cacheTest sizeI sizeJ\n");
+  if (argc < 4){
+    printf("Usage: cacheTest fast sizeI sizeJ\nIn first input, use 1 for fast code (cache-smart) and anything else for slow (cache-poor) execution.\n");
     return 1;
   }
-  int sI = atoi(argv[1]);
-  int sJ = atoi(argv[2]);
-  int *arr = new int[sI*sJ]; // double array.
+  int runFast = atoi(argv[1]);
+  long sI = atoi(argv[2]);
+  long sJ = atoi(argv[3]);
 
+  printf("Operating on matrix of size %d by %d\n",sI,sJ);
+
+  long *arr = new long[sI*sJ]; // double array.
+
+  // El mejor!
   //for (int i=0; i < sI*sJ; ++i) arr[i] = i;
 
-  for (int i=0; i < sI; ++i)
-    for (int j=0; j < sJ; ++j)
-       arr[(i * (sJ)) + j ] = i;
+  if (runFast == 1){
+
+    for (long i=0; i < sI; ++i)
+      for (long j=0; j < sJ; ++j)
+         arr[(i * (sJ)) + j ] = i;
 
-  for (int i=0; i < sI; ++i)
-    for (int j=0; j < sJ; ++j)
-       arr[(j * (sI)) + i ] = i;
+  }else{
+
+    for (long i=0; i < sI; ++i)
+      for (long j=0; j < sJ; ++j)
+         arr[(j * (sI)) + i ] = i;
+
+  }
 
   return 1;
 }