[x264-devel] Fix 10L in r1912

[Jason Garrett-Glaser]

x264 | branch: master | Jason Garrett-Glaser <jason at x264.com> | Fri Feb 18 17:50:42 2011 -0800| [5fd3dce0c72a40722df6a9bddf599980846f6fe8] | committer: Jason Garrett-Glaser

Fix 10L in r1912
SSSE3 code got used in MMX/SSE2 and vice versa (in hpel).

> http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=5fd3dce0c72a40722df6a9bddf599980846f6fe8
---

 common/x86/mc-a2.asm |   16 ++++++++--------
 1 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/common/x86/mc-a2.asm b/common/x86/mc-a2.asm
index 6fdde60..d4ea8b2 100644
--- a/common/x86/mc-a2.asm
+++ b/common/x86/mc-a2.asm
@@ -323,12 +323,18 @@ cglobal hpel_filter_v_%1, 5,6,%2
     lea r2, [r2+r4*2]
     neg r4
 %if %3
-    mova m0, [filt_mul15]
-%else
     pxor m0, m0
+%else
+    mova m0, [filt_mul15]
 %endif
 .loop:
 %if %3
+    LOAD_ADD_2 m1, m4, [r1     ], [r5+r3*2], m6, m7            ; a0 / a1
+    LOAD_ADD_2 m2, m5, [r1+r3  ], [r5+r3  ], m6, m7            ; b0 / b1
+    LOAD_ADD   m3,     [r1+r3*2], [r5     ], m7                ; c0
+    LOAD_ADD   m6,     [r1+r3*2+mmsize/2], [r5+mmsize/2], m7   ; c1
+    FILT_V2 m1, m2, m3, m4, m5, m6
+%else
     mova m1, [r1]
     mova m4, [r1+r3]
     mova m2, [r5+r3*2]
@@ -348,12 +354,6 @@ cglobal hpel_filter_v_%1, 5,6,%2
     paddw  m4, m5
     paddw  m1, m3
     paddw  m4, m6
-%else
-    LOAD_ADD_2 m1, m4, [r1     ], [r5+r3*2], m6, m7            ; a0 / a1
-    LOAD_ADD_2 m2, m5, [r1+r3  ], [r5+r3  ], m6, m7            ; b0 / b1
-    LOAD_ADD   m3,     [r1+r3*2], [r5     ], m7                ; c0
-    LOAD_ADD   m6,     [r1+r3*2+mmsize/2], [r5+mmsize/2], m7   ; c1
-    FILT_V2 m1, m2, m3, m4, m5, m6
 %endif
     mova      m7, [pw_16]
     mova      [r2+r4*2], m1