EXPERIMENT2 in divmnu64.c to split out into two loops

diff --git a/openpower/sv/bitmanip/divmnu64.c b/openpower/sv/bitmanip/divmnu64.c
index 02bef0bb5061cf86f0c88b36f6e92fa63fcad5b4..192fa1674874d5629b056ba320db8f3a2278a362 100644 (file)
--- a/openpower/sv/bitmanip/divmnu64.c
+++ b/openpower/sv/bitmanip/divmnu64.c
@@ -111,10 +111,11 @@ again:
       for (i = 0; i < n; i++) {
          p = qhat*vn[i];
          t = un[i+j] - k - (p & 0xFFFFFFFFLL);
-         un[i+j] = t;
+         un[i+j] = (t & 0xffffffffLL);
          k = (p >> 32) - (t >> 32);
       }
-#else
+#endif
+#ifdef EXPERIMENT1
       for (i = 0; i < n; i++) {
          unsigned rhi;
          unsigned long long sum;
@@ -124,6 +125,22 @@ again:
          k = rhi;
          un[i+j] = sum & 0xffffffff;
       }
+#endif
+#define EXPERIMENT2
+#ifdef EXPERIMENT2
+      unsigned long phi[200]; // yes, not malloced, we know
+      unsigned long plo[200]; // yes, not malloced, we know
+      // double-width multiply with strange subtract only on bottom half
+      for (i = 0; i < n; i++) {
+        unsigned long long p = qhat*vn[i];
+        plo[i] = un[i+j] - (p&0xffffffffLL);
+        phi[i] = p >> 32;
+      }
+      for (i = 0; i < n; i++) {
+         t = plo[i] - k; // subtract previous carry
+         un[i+j] = (t & 0xffffffffLL);
+         k = phi[i] - (t >> 32); // take top-halves for new carry
+      }
 #endif
       t = un[j+n] - k;
       un[j+n] = t;