New rule for size of array part

Array part needs 1/3 of its elements filled, instead of 1/2.
Array entries use ~1/3 the memory of hash entries, so this new rule
still ensures that array parts do not use more memory than keeping
the values in the hash, while allowing more uses of the array part,
which is more efficient than the hash.

1 files changed, 20 insertions, 7 deletions

diff --git a/ltable.c b/ltable.c
index 3451445c..923f3eaa 100644
--- a/ltable.c
+++ b/ltable.c
@@ -471,12 +471,23 @@ typedef struct {
   unsigned nums[MAXABITS + 1];
 } Counters;
 
+
+/*
+** Check whether it is worth to use 'na' array entries instead of 'nh'
+** hash nodes. (A hash node uses ~3 times more memory than an array
+** entry: Two values plus 'next' versus one value.) Evaluate with size_t
+** to avoid overflows.
+*/
+#define arrayXhash(na,nh)       (cast_sizet(na) <= cast_sizet(nh) * 3)
+
 /*
 ** Compute the optimal size for the array part of table 't'.
+** This size maximizes the number of elements going to the array part
+** while satisfying the condition 'arrayXhash' with the use of memory if
+** all those elements went to the hash part.
 ** 'ct->na' enters with the total number of array indices in the table
 ** and leaves with the number of keys that will go to the array part;
-** return the optimal size.  (The condition 'twotoi > 0' in the for loop
-** stops the loop if 'twotoi' overflows.)
+** return the optimal size for the array part.
 */
 static unsigned computesizes (Counters *ct) {
   int i;
@@ -484,17 +495,19 @@ static unsigned computesizes (Counters *ct) {
   unsigned int a = 0;  /* number of elements smaller than 2^i */
   unsigned int na = 0;  /* number of elements to go to array part */
   unsigned int optimal = 0;  /* optimal size for array part */
-  /* loop while keys can fill more than half of total size */
+  /* traverse slices while 'twotoi' does not overflow and total of array
+     indices still can satisfy 'arrayXhash' against the array size */
   for (i = 0, twotoi = 1;
-       twotoi > 0 && ct->na > twotoi / 2;
+       twotoi > 0 && arrayXhash(twotoi, ct->na);
        i++, twotoi *= 2) {
-    a += ct->nums[i];
-    if (a > twotoi/2) {  /* more than half elements present? */
+    unsigned nums = ct->nums[i];
+    a += nums;
+    if (nums > 0 &&  /* grows array only if it gets more elements... */
+        arrayXhash(twotoi, a)) {  /* ...while using "less memory" */
       optimal = twotoi;  /* optimal size (till now) */
       na = a;  /* all elements up to 'optimal' will go to array part */
     }
   }
-  lua_assert((optimal == 0 || optimal / 2 < na) && na <= optimal);
   ct->na = na;
   return optimal;
 }