Clean up canonical mt benchmarks and reorganize extra versions in /mt. All versions support support at least 1/2/4 threads.

1 files changed, 22 insertions, 0 deletions

diff --git a/mt/vvadd3.c b/mt/vvadd3.c
new file mode 100755
index 0000000..d18d2de
--- /dev/null
+++ b/mt/vvadd3.c
@@ -0,0 +1,22 @@
+#include "stdlib.h"
+#include "dataset.h"
+
+void __attribute__((noinline)) vvadd(int coreid, int ncores, size_t n, const data_t* x, const data_t* y, data_t* z)
+{
+  data_t* to = &z[coreid * (n / ncores)];
+  const data_t* from1 = &x[coreid * (n / ncores)];
+  const data_t* from2 = &y[coreid * (n / ncores)];
+  size_t count = n / ncores;
+  size_t c = (count + 7) / 8;
+  switch(count % 8) {
+    case 0: do {  *to++ = *from1++ + *from2++;
+    case 7:       *to++ = *from1++ + *from2++;
+    case 6:       *to++ = *from1++ + *from2++;
+    case 5:       *to++ = *from1++ + *from2++;
+    case 4:       *to++ = *from1++ + *from2++;
+    case 3:       *to++ = *from1++ + *from2++;
+    case 2:       *to++ = *from1++ + *from2++;
+    case 1:       *to++ = *from1++ + *from2++;
+  } while(--c > 0);
+  }
+}