Fix valgrind errors in the NEON loop filters.

Like the ARMv6 code, these functions were accessing values below
 the stack pointer, which can be corrupted by signal delivery at
 any time.

diff --git a/vp8/common/arm/neon/mbloopfilterverticaledge_uv_neon.asm b/vp8/common/arm/neon/mbloopfilterverticaledge_uv_neon.asm
index 33cd55e1c95948026294eea262a378df220c887c..044b3a3a8fe326b64b84d685a64a38f26a373f43 100644 (file)
--- a/vp8/common/arm/neon/mbloopfilterverticaledge_uv_neon.asm
+++ b/vp8/common/arm/neon/mbloopfilterverticaledge_uv_neon.asm
@@ -67,9 +67,10 @@
 
     sub         sp, sp, #32
     vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    vst1.u8     {q3}, [sp]!
+    mov         r12, sp
+    vst1.u8     {q3}, [r12]!
+    vst1.u8     {q10}, [r12]!
     ldr         r12, _mbvlfuv_coeff_
-    vst1.u8     {q10}, [sp]!
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
@@ -165,8 +166,6 @@
 
     vld1.u8     {d5}, [r12]!                ;#27
 
-    sub         sp, sp, #32
-
     vqadd.s8    q6, q6, q13                 ; ps0 = vp8_signed_char_clamp(ps0 + Filter2)
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 

diff --git a/vp8/common/arm/neon/mbloopfilterverticaledge_y_neon.asm b/vp8/common/arm/neon/mbloopfilterverticaledge_y_neon.asm
index f51fd0bd4a01f36d8fde65c7374c9a65b1b56cb4..e0716625ce555b01137eb817b2337d2c989a6a63 100644 (file)
--- a/vp8/common/arm/neon/mbloopfilterverticaledge_y_neon.asm
+++ b/vp8/common/arm/neon/mbloopfilterverticaledge_y_neon.asm
@@ -63,11 +63,12 @@
     vtrn.8      q7, q8
     vtrn.8      q9, q10
 
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    vst1.u8     {q3}, [sp]!
     vld1.s8     {d4[], d5[]}, [r12]         ; thresh
+    vld1.s8     {d2[], d3[]}, [r3]          ; limit
+    mov         r12, sp
+    vst1.u8     {q3}, [r12]!
+    vst1.u8     {q10}, [r12]!
     ldr         r12, _mbvlfy_coeff_
-    vst1.u8     {q10}, [sp]!
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
@@ -150,7 +151,6 @@
 
     vld1.u8     {d6}, [r12]!                ;#18
     sub         r0, r0, r1, lsl #4
-    sub         sp, sp, #32
 
     add         r2, r0, r1