avoid aligned alloc for infrequently written areas

Aligned allocations are wasteful and tend to fragment the heap.
They only benefit us when dealing with frequently _written_ data
that we don't want unnecessarily shared/bounced between threads.
So swap mog_cachealign() for xmalloc() wherever we have
infrequently-written data areas.

1 files changed, 1 insertions, 1 deletions

diff --git a/trywrite.c b/trywrite.c
index 2e0fca7..0c7a68b 100644
--- a/trywrite.c
+++ b/trywrite.c
@@ -12,7 +12,7 @@ struct mog_wbuf {
 
 static void * wbuf_new(size_t total, struct iovec *iov, int iovcnt)
 {
-        struct mog_wbuf *wbuf = mog_cachealign(sizeof(struct mog_wbuf) + total);
+        struct mog_wbuf *wbuf = xmalloc(sizeof(struct mog_wbuf) + total);
         void *dst = wbuf->buf;
         int i;