arm: do not use aligned stores in mc_weight_w4_*neon

mc_weight_w4_*neon is also used for width 2 which does not guarantee
4-byte aligned destination. Fixes crashes caused by random memory
corruption.

diff --git a/common/arm/mc-a.S b/common/arm/mc-a.S
index cd57920cbaa021e54878355b622cdb9c6a18d5bf..2b7acdac8b187dcdc7974cf56a9e3cab87ab79b0 100644 (file)
--- a/common/arm/mc-a.S
+++ b/common/arm/mc-a.S
@@ -567,8 +567,8 @@ weight4_loop:
     vrshl.s16   q8,  q8,  q2
     vadd.s16    q8,  q8,  q1
     vqmovun.s16 d16, q8
-    vst1.32     {d16[0]}, [r0,:32], r1
-    vst1.32     {d16[1]}, [r0,:32], r1
+    vst1.32     {d16[0]}, [r0], r1
+    vst1.32     {d16[1]}, [r0], r1
     bgt         weight4_loop
     pop         {r4-r5,pc}
 endfunc
@@ -670,8 +670,8 @@ weight4_nodenom_loop:
     vmla.s16    d20, d16, d0
     vmla.s16    d21, d18, d0
     vqmovun.s16 d16, q10
-    vst1.32     {d16[0]}, [r0,:32], r1
-    vst1.32     {d16[1]}, [r0,:32], r1
+    vst1.32     {d16[0]}, [r0], r1
+    vst1.32     {d16[1]}, [r0], r1
     bgt         weight4_nodenom_loop
     pop         {r4-r5,pc}
 endfunc
@@ -734,8 +734,8 @@ weight4_\name\()_loop:
     vld1.32     {d16[]}, [r2], r3
     vld1.32     {d17[]}, [r2], r3
     \op         q8,  q8,  q1
-    vst1.32     {d16[0]}, [r0,:32], r1
-    vst1.32     {d17[0]}, [r0,:32], r1
+    vst1.32     {d16[0]}, [r0], r1
+    vst1.32     {d17[0]}, [r0], r1
     bgt         weight4_\name\()_loop
     pop         {pc}
 endfunc