diff --git a/libavfilter/vf_gblur.c b/libavfilter/vf_gblur.c
index 1957d79e0f..9b3e168b1d 100644
--- a/libavfilter/vf_gblur.c
+++ b/libavfilter/vf_gblur.c
@@ -236,7 +236,7 @@ static int config_input(AVFilterLink *inlink)
 
     s->nb_planes = av_pix_fmt_count_planes(inlink->format);
 
-    s->buffer = av_malloc_array(inlink->w, inlink->h * sizeof(*s->buffer));
+    s->buffer = av_malloc_array(FFALIGN(inlink->w, 16), FFALIGN(inlink->h, 16) * sizeof(*s->buffer));
     if (!s->buffer)
         return AVERROR(ENOMEM);
 
diff --git a/libavfilter/x86/vf_gblur.asm b/libavfilter/x86/vf_gblur.asm
index 762c953c85..a25b1659f5 100644
--- a/libavfilter/x86/vf_gblur.asm
+++ b/libavfilter/x86/vf_gblur.asm
@@ -100,7 +100,7 @@ cglobal horiz_slice, 4, 9, 9, ptr, width, height, steps, nu, bscale, x, y, step,
 
         add widthq, remainq
         cmp xq, widthq
-        je .end_scalar
+        jge .end_scalar
 
         .loop_scalar:
             ; ptr[x] += nu * ptr[x-1]
@@ -148,7 +148,7 @@ cglobal horiz_slice, 4, 9, 9, ptr, width, height, steps, nu, bscale, x, y, step,
             jg .loop_x_back
 
         cmp xq, 0
-        je .end_scalar_back
+        jle .end_scalar_back
 
         .loop_scalar_back:
             ; ptr[x-1] += nu * ptr[x]