Little tweaks to make gcc compile this the faster way.sort-balanced-merge3

On master, gcc compiles this into a call to memcpy(), but on the branch,
it compiled it into a "rep movsq". Turns out that memcpy() is faster.
Move things around a little bit, so that we get the faster memcpy() version.

1 files changed, 7 insertions, 6 deletions

diff --git a/src/backend/utils/sort/logtape.c b/src/backend/utils/sort/logtape.c
index 891a663d31..789dc2dec1 100644
--- a/src/backend/utils/sort/logtape.c
+++ b/src/backend/utils/sort/logtape.c
@@ -625,7 +625,10 @@ LogicalTapeWrite(LogicalTape *lt, void *ptr, size_t size)
 	Assert(lt->buffer_size == BLCKSZ);
 	while (size > 0)
 	{
-		if (lt->pos >= TapeBlockPayloadSize)
+		int			bufleft;
+
+		bufleft = TapeBlockPayloadSize - lt->pos;
+		if (bufleft == 0)
 		{
 			/* Buffer full, dump it out */
 			long		nextBlockNumber;
@@ -646,18 +649,16 @@ LogicalTapeWrite(LogicalTape *lt, void *ptr, size_t size)
 			lt->curBlockNumber = nextBlockNumber;
 			lt->pos = 0;
 			lt->nbytes = 0;
+			bufleft = TapeBlockPayloadSize;
 		}
 
-		nthistime = TapeBlockPayloadSize - lt->pos;
-		if (nthistime > size)
-			nthistime = size;
+		nthistime = Min(bufleft, size);
 		Assert(nthistime > 0);
 
 		memcpy(lt->buffer + lt->pos, ptr, nthistime);
 
 		lt->pos += nthistime;
-		if (lt->nbytes < lt->pos)
-			lt->nbytes = lt->pos;
+		lt->nbytes = lt->pos;
 		ptr = (void *) ((char *) ptr + nthistime);
 		size -= nthistime;
 	}