diff --git a/libavcodec/x86/vp9mc.asm b/libavcodec/x86/vp9mc.asm
index 5393957..8109fa0 100644
--- a/libavcodec/x86/vp9mc.asm
+++ b/libavcodec/x86/vp9mc.asm
@@ -192,42 +192,49 @@ filter_sse2_h_fn avg

 %macro filter_h_fn 1
 %assign %%px mmsize/2
-cglobal vp9_%1_8tap_1d_h_ %+ %%px, 6, 6, 11, dst, dstride, src, sstride, h, filtery
-    mova        m6, [pw_256]
-    mova        m7, [filteryq+ 0]
+cglobal vp9_%1_8tap_1d_h_ %+ %%px, 6, 6, 10, dst, dstride, src, sstride, h, filtery
+    mova        m5, [pw_256]
+    mova        m6, [filteryq+ 0]
+    mova        m7, [filteryq+32]
 %if ARCH_X86_64 && mmsize > 8
-    mova        m8, [filteryq+32]
-    mova        m9, [filteryq+64]
-    mova       m10, [filteryq+96]
+    mova        m8, [filteryq+64]
+    mova        m9, [filteryq+96]
 %endif
 .loop:
-    movh        m0, [srcq-3]
-    movh        m1, [srcq-2]
-    movh        m2, [srcq-1]
-    movh        m3, [srcq+0]
-    movh        m4, [srcq+1]
-    movh        m5, [srcq+2]
-    punpcklbw   m0, m1
-    punpcklbw   m2, m3
-    movh        m1, [srcq+3]
-    movh        m3, [srcq+4]
+    movh        m1, [srcq-3]
+    movq        m0, [srcq+%%px-3]
+    punpcklbw   m1, m1
+%if mmsize == 8
+    punpckhbw   m3, m0, m0
+%endif
+    punpcklbw   m0, m0
     add       srcq, sstrideq
-    punpcklbw   m4, m5
-    punpcklbw   m1, m3
-    pmaddubsw   m0, m7
+%if mmsize == 8
+    mova        m4, m3
+    palignr     m3, m0, 1
+    palignr     m4, m0, 5
+%else
+    mova        m3, m0
+    mova        m4, m0
+    palignr     m3, m1, 9
+    palignr     m4, m1, 13
+%endif
+    mova        m2, m0
+    palignr     m0, m1, 1
+    palignr     m2, m1, 5
+    pmaddubsw   m0, m6
+    pmaddubsw   m2, m7
 %if ARCH_X86_64 && mmsize > 8
-    pmaddubsw   m2, m8
+    pmaddubsw   m3, m8
     pmaddubsw   m4, m9
-    pmaddubsw   m1, m10
 %else
-    pmaddubsw   m2, [filteryq+32]
-    pmaddubsw   m4, [filteryq+64]
-    pmaddubsw   m1, [filteryq+96]
+    pmaddubsw   m3, [filteryq+64]
+    pmaddubsw   m4, [filteryq+96]
 %endif
-    paddw       m0, m4
-    paddw       m2, m1
+    paddw       m0, m3
+    paddw       m2, m4
     paddsw      m0, m2
-    pmulhrsw    m0, m6
+    pmulhrsw    m0, m5
 %ifidn %1, avg
     movh        m1, [dstq]
 %endif
@@ -260,35 +267,36 @@ cglobal vp9_%1_8tap_1d_h_ %+ %%px, 6, 6, 14, dst, dstride, src, sstride, h, filt
     mova       m10, [filteryq+64]
     mova       m11, [filteryq+96]
 .loop:
-    movu        m0, [srcq-3]
-    movu        m1, [srcq-2]
-    movu        m2, [srcq-1]
-    movu        m3, [srcq+0]
-    movu        m4, [srcq+1]
-    movu        m5, [srcq+2]
-    movu        m6, [srcq+3]
-    movu        m7, [srcq+4]
+    movu       m12, [srcq-3]
+    movu        m1, [srcq+5]
+    punpckhbw   m0, m12, m12
+    punpckhbw   m1, m1
+    punpcklbw  m12, m12
     add       srcq, sstrideq
-    SBUTTERFLY  bw, 0, 1, 12
-    SBUTTERFLY  bw, 2, 3, 12
-    SBUTTERFLY  bw, 4, 5, 12
-    SBUTTERFLY  bw, 6, 7, 12
-    pmaddubsw   m0, m8
-    pmaddubsw   m1, m8
-    pmaddubsw   m2, m9
-    pmaddubsw   m3, m9
-    pmaddubsw   m4, m10
-    pmaddubsw   m5, m10
-    pmaddubsw   m6, m11
+    palignr     m7, m1, m0, 13
+    palignr     m6, m0, m12, 13
+    palignr     m5, m1, m0, 9
+    palignr     m4, m0, m12, 9
+    palignr     m3, m1, m0, 5
+    palignr     m2, m0, m12, 5
+    palignr     m1, m0, 1
+    palignr     m0, m12, 1
     pmaddubsw   m7, m11
-    paddw       m0, m4
-    paddw       m1, m5
-    paddw       m2, m6
+    pmaddubsw   m6, m11
+    pmaddubsw   m5, m10
+    pmaddubsw   m4, m10
+    pmaddubsw   m3, m9
+    pmaddubsw   m2, m9
+    pmaddubsw   m1, m8
+    pmaddubsw   m0, m8
     paddw       m3, m7
-    paddsw      m0, m2
+    paddw       m2, m6
+    paddw       m1, m5
+    paddw       m0, m4
     paddsw      m1, m3
-    pmulhrsw    m0, m13
+    paddsw      m0, m2
     pmulhrsw    m1, m13
+    pmulhrsw    m0, m13
     packuswb    m0, m1
 %ifidn %1, avg
     pavgb       m0, [dstq]