Finish basic vector add.

.gitignore

@@ -1,3 +1,4 @@
 a.out
 .*.swp
 *~
+*-demo.c

01-sequential-soln.c

@@ -0,0 +1,46 @@
+#include "timing.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(int argc, char **argv)
+{
+  if (argc != 2)
+  {
+    // STRIP
+    fprintf(stderr, "need two arguments!\n");
+    abort();
+  }
+
+  const long n = atol(argv[1]);
+  double *x = (double *) malloc(sizeof(double) * n);
+  if (!x) { perror("alloc x"); abort(); }
+  double *y = (double *) malloc(sizeof(double) * n);
+  if (!y) { perror("alloc y"); abort(); }
+  double *z = (double *) malloc(sizeof(double) * n);
+  if (!z) { perror("alloc z"); abort(); }
+
+  for (int i = 0; i < n; ++i)
+  {
+    x[i] = i;
+    y[i] = 2*i;
+  }
+
+  struct timespec time1, time2;
+  clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time1);
+
+  const int ntrips = 2;
+  for (int trip = 0; trip < ntrips; ++trip)
+  {
+    for (int i = 0; i < n; ++i)
+    {
+      z[i] = x[i] + y[i];
+    }
+  }
+
+  clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time2);
+  double elapsed = timespec_diff_in_seconds(time1,time2)/ntrips;
+  printf("%f GB/s\n",
+      3*n*sizeof(double)/1e9/elapsed);
+  printf("%f GFlops/s\n",
+      n/1e9/elapsed);
+}

01-sequential-start.c

@@ -0,0 +1,33 @@
+#include "timing.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(int argc, char **argv)
+{
+  if (argc != 2)
+  {
+    // FIXME
+  }
+
+  const long n = atol(argv[1]);
+  double *x = (double *) malloc(sizeof(double) * n);
+  double *y = (double *) malloc(sizeof(double) * n);
+  double *z = (double *) malloc(sizeof(double) * n);
+
+  for (int i = 0; i < n; ++i)
+  {
+    x[i] = i;
+    y[i] = 2*i;
+  }
+
+  struct timespec time1, time2;
+  clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time1);
+
+  // FIXME
+
+  clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time2);
+  printf("%f seconds for %d entries\n",
+      timespec_diff_in_seconds(time1,time2),
+      n);
+}
+

02-flexible-soln.c

@@ -0,0 +1,46 @@
+#include "timing.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(int argc, char **argv)
+{
+  if (argc != 2)
+  {
+    // STRIP
+    fprintf(stderr, "need two arguments!\n");
+    abort();
+  }
+
+  const long n = atol(argv[1]);
+  double *x = (double *) malloc(sizeof(double) * n);
+  if (!x) { perror("alloc x"); abort(); }
+  double *y = (double *) malloc(sizeof(double) * n);
+  if (!y) { perror("alloc y"); abort(); }
+  double *z = (double *) malloc(sizeof(double) * n);
+  if (!z) { perror("alloc z"); abort(); }
+
+  for (int i = 0; i < n; ++i)
+  {
+    x[i] = i;
+    y[i] = 2*i;
+  }
+
+  struct timespec time1, time2;
+  clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time1);
+
+  const int ntrips = 2;
+  for (int trip = 0; trip < ntrips; ++trip)
+  {
+    for (int i = 0; i < n; ++i)
+    {
+      z[i] = x[i] + y[i];
+    }
+  }
+
+  clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time2);
+  double elapsed = timespec_diff_in_seconds(time1,time2)/ntrips;
+  printf("%f GB/s\n",
+      3*n*sizeof(double)/1e9/elapsed);
+  printf("%f GFlops/s\n",
+      n/1e9/elapsed);
+}

notes.txt

@@ -1,8 +1,23 @@
-000-timing-example.c:
+colorscheme evening
+
+00-timing-example.c:
 - basic timing
 - run into all errors
 - -O3
 - enable result printing
 - time again
 
-001-simple.c:
+01-simple.c:
+- simple timing
+- run into segfault
+  - debug it
+- what do the numbers mean?
+- timing repeatability
+- O3
+- set-governor performance/ondemand
+- run through a few decades
+  - why is it doing what it's doing?!
+  - Control flow overhead
+  - Cache
+
+02-flexible.c:

reset.sh

@@ -0,0 +1,5 @@
+#! /bin/sh
+
+for i in *-start.c; do
+  echo cp $i ${i%%-start.c}-demo.c
+done

set-governor

@@ -0,0 +1,5 @@
+#! /bin/sh
+
+for i in 0 1 2 3; do
+  cpufreq-set -g $1 -c 1
+done