diff --git a/vp9/decoder/vp9_decodeframe.c b/vp9/decoder/vp9_decodeframe.c
index bb5a5a22d039e5e14cbd932b60b9d39faa04c8d6..44d1e288335a88dfcd5b79da39c95c99029fe0ea 100644
--- a/vp9/decoder/vp9_decodeframe.c
+++ b/vp9/decoder/vp9_decodeframe.c
@@ -1338,7 +1338,8 @@ int vp9_decode_frame(VP9D_COMP *pbi,
   pbi->do_loopfilter_inline =
       (cm->log2_tile_rows | cm->log2_tile_cols) == 0 && cm->lf.filter_level;
   if (pbi->do_loopfilter_inline && pbi->lf_worker.data1 == NULL) {
-    CHECK_MEM_ERROR(cm, pbi->lf_worker.data1, vpx_malloc(sizeof(LFWorkerData)));
+    CHECK_MEM_ERROR(cm, pbi->lf_worker.data1,
+                    vpx_memalign(32, sizeof(LFWorkerData)));
     pbi->lf_worker.hook = (VP9WorkerHook)vp9_loop_filter_worker;
     if (pbi->oxcf.max_threads > 1 && !vp9_worker_reset(&pbi->lf_worker)) {
       vpx_internal_error(&cm->error, VPX_CODEC_ERROR,