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

1 files changed, 32 insertions, 0 deletions

diff --git a/mt/ds_matmul.c b/mt/ds_matmul.c
new file mode 100644
index 0000000..3b48267
--- /dev/null
+++ b/mt/ds_matmul.c
@@ -0,0 +1,32 @@
+#include "stdlib.h"
+
+#include "util.h"
+
+#include "dataset.h"
+void __attribute__((noinline)) matmul(const int coreid, const int ncores, const int lda,  const data_t A[], const data_t B[], data_t C[] )
+{
+int i,j,k,a,b,b1,a1,a2,a3,c,c1,c2,c3,b2,b3;
+        for (j=coreid*4; j<lda; j+=4*ncores){
+                a=j*lda;
+                a1=(j+1)*lda;
+                a2=(j+2)*lda;
+                a3=(j+3)*lda;
+                for (k=0;k<lda; k+=2)
+                {
+                        b = k*lda;
+                        b1 = (k+1)*lda;
+                        for (i=0;i<lda;i++){
+                                c = B[b+i];
+                                c1 = B[b1+i];
+                                C[i+a]+=A[a+k]*c;
+                                C[i+a1]+=A[a1+k]*c;
+                                C[i+a2]+=A[a2+k]*c;
+                                C[i+a3]+=A[a3+k]*c;
+                                C[i+a]+=A[a+k+1]*c1;
+                                C[i+a1]+=A[a1+k+1]*c1;
+                                C[i+a2]+=A[a2+k+1]*c1;
+                                C[i+a3]+=A[a3+k+1]*c1;
+}
+}
+}
+}