From: Jim Bankoski <jimbankoski@google.com>
Date: Mon, 19 May 2014 19:46:22 +0000 (-0700)
Subject: Fix valgrind read out of bounds error.
X-Git-Tag: v1.4.0~1544^2
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=bb2041d8de5e1b23e86bbe5e94484f03239e3123;p=libvpx

Fix valgrind read out of bounds error.

MMX variance code in vp8 was reading out of bounds..

TODO(JBB):  The best fix would involve removing duplicate library
functions between vp8 and vp9...

Change-Id: I5722853a6a58d3b55257ff385fa54c773bf98ded
---

diff --git a/vp8/common/x86/variance_impl_mmx.asm b/vp8/common/x86/variance_impl_mmx.asm
index d9120d0d4..7d5e6810b 100644
--- a/vp8/common/x86/variance_impl_mmx.asm
+++ b/vp8/common/x86/variance_impl_mmx.asm
@@ -342,8 +342,8 @@ sym(vp8_get4x4var_mmx):
         movsxd      rdx, dword ptr arg(3) ;[recon_stride]
 
         ; Row 1
-        movq        mm0, [rax]                  ; Copy eight bytes to mm0
-        movq        mm1, [rbx]                  ; Copy eight bytes to mm1
+        movd        mm0, [rax]                  ; Copy four bytes to mm0
+        movd        mm1, [rbx]                  ; Copy four bytes to mm1
         punpcklbw   mm0, mm6                    ; unpack to higher prrcision
         punpcklbw   mm1, mm6
         psubsw      mm0, mm1                    ; A-B (low order) to MM0
@@ -351,12 +351,12 @@ sym(vp8_get4x4var_mmx):
         pmaddwd     mm0, mm0                    ; square and accumulate
         add         rbx,rdx                     ; Inc pointer into ref data
         add         rax,rcx                     ; Inc pointer into the new data
-        movq        mm1, [rbx]                  ; Copy eight bytes to mm1
+        movd        mm1, [rbx]                  ; Copy four bytes to mm1
         paddd       mm7, mm0                    ; accumulate in mm7
 
 
         ; Row 2
-        movq        mm0, [rax]                  ; Copy eight bytes to mm0
+        movd        mm0, [rax]                  ; Copy four bytes to mm0
         punpcklbw   mm0, mm6                    ; unpack to higher prrcision
         punpcklbw   mm1, mm6
         psubsw      mm0, mm1                    ; A-B (low order) to MM0
@@ -365,12 +365,12 @@ sym(vp8_get4x4var_mmx):
         pmaddwd     mm0, mm0                    ; square and accumulate
         add         rbx,rdx                     ; Inc pointer into ref data
         add         rax,rcx                     ; Inc pointer into the new data
-        movq        mm1, [rbx]                  ; Copy eight bytes to mm1
+        movd        mm1, [rbx]                  ; Copy four bytes to mm1
         paddd       mm7, mm0                    ; accumulate in mm7
 
         ; Row 3
-        movq        mm0, [rax]                  ; Copy eight bytes to mm0
-        punpcklbw   mm0, mm6                    ; unpack to higher prrcision
+        movd        mm0, [rax]                  ; Copy four bytes to mm0
+        punpcklbw   mm0, mm6                    ; unpack to higher precision
         punpcklbw   mm1, mm6
         psubsw      mm0, mm1                    ; A-B (low order) to MM0
         paddw       mm5, mm0                    ; accumulate differences in mm5
@@ -378,11 +378,11 @@ sym(vp8_get4x4var_mmx):
         pmaddwd     mm0, mm0                    ; square and accumulate
         add         rbx,rdx                     ; Inc pointer into ref data
         add         rax,rcx                     ; Inc pointer into the new data
-        movq        mm1, [rbx]                  ; Copy eight bytes to mm1
+        movd        mm1, [rbx]                  ; Copy four bytes to mm1
         paddd       mm7, mm0                    ; accumulate in mm7
 
         ; Row 4
-        movq        mm0, [rax]                  ; Copy eight bytes to mm0
+        movd        mm0, [rax]                  ; Copy four bytes to mm0
 
         punpcklbw   mm0, mm6                    ; unpack to higher prrcision
         punpcklbw   mm1, mm6