[FFmpeg-cvslog] swscale/x86/input: prevent RGB32_TO_UV_FN from reading into the padding

[Michael Niedermayer]

ffmpeg | branch: master | Michael Niedermayer <michaelni at gmx.at> | Sun Jun 22 02:32:30 2014 +0200| [a1a76b209b42cdd9f8f16bfce7509883d636b940] | committer: Michael Niedermayer

swscale/x86/input: prevent RGB32_TO_UV_FN from reading into the padding

This avoids reading uninitialized bytes and confusing valgrind

Signed-off-by: Michael Niedermayer <michaelni at gmx.at>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a1a76b209b42cdd9f8f16bfce7509883d636b940
---

 libswscale/x86/input.asm |   26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/libswscale/x86/input.asm b/libswscale/x86/input.asm
index 0c4f30e..1f0f1ef 100644
--- a/libswscale/x86/input.asm
+++ b/libswscale/x86/input.asm
@@ -406,6 +406,7 @@ cglobal %2%3%4%5 %+ ToUV, 7, 7, %1, dstU, dstV, u1, src, u2, w, table
     mov            wq, r5m
 %endif
     add            wq, wq
+    sub            wq, mmsize - 1
     add         dstUq, wq
     add         dstVq, wq
     lea          srcq, [srcq+wq*2]
@@ -449,6 +450,31 @@ cglobal %2%3%4%5 %+ ToUV, 7, 7, %1, dstU, dstV, u1, src, u2, w, table
 %endif ; mmsize == 8/16
     add            wq, mmsize
     jl .loop
+    sub            wq, mmsize - 1
+    jz .end
+    add            srcq , 2*mmsize - 2
+    add            dstUq, mmsize - 1
+    add            dstVq, mmsize - 1
+.loop2:
+    movd           m0, [srcq+wq*2]        ; (byte) { Bx, Gx, Rx, xx }[0-3]
+    DEINTB          1,  0,  5,  4,  7     ; (word) { Gx, xx (m0/m4) or Bx, Rx (m1/m5) }[0-3]/[4-7]
+    pmaddwd        m3, m1, coeffV1        ; (dword) { Bx*BV + Rx*RV }[0-3]
+    pmaddwd        m2, m0, coeffV2        ; (dword) { Gx*GV }[0-3]
+    pmaddwd        m1, coeffU1            ; (dword) { Bx*BU + Rx*RU }[0-3]
+    pmaddwd        m0, coeffU2            ; (dword) { Gx*GU }[0-3]
+    paddd          m3, m6                 ; += rgb_UVrnd
+    paddd          m1, m6                 ; += rgb_UVrnd
+    paddd          m2, m3                 ; (dword) { V[0-3] }
+    paddd          m0, m1                 ; (dword) { U[0-3] }
+    psrad          m0, 9
+    psrad          m2, 9
+    packssdw       m0, m0                 ; (word) { U[0-7] }
+    packssdw       m2, m2                 ; (word) { V[0-7] }
+    movd   [dstUq+wq], m0
+    movd   [dstVq+wq], m2
+    add            wq, 2
+    jl .loop2
+.end:
     REP_RET
 %endif ; ARCH_X86_64 && %0 == 3
 %endmacro