libvpx: Pull from upstream

source/libvpx/vp9/encoder/x86/vp9_highbd_sad4d_sse2.asm

 ;
 ;  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
 ;
 ;  Use of this source code is governed by a BSD-style license
 ;  that can be found in the LICENSE file in the root of the source
 ;  tree. An additional intellectual property rights grant can be found
 ;  in the file PATENTS.  All contributing project authors may
 ;  be found in the AUTHORS file in the root of the source tree.
 ;
 
@@ skipping 197 matching lines @@
   HIGH_PROCESS_32x2x4 %1, %2, %3, (%2 + 32), (%3 + 32)
   HIGH_PROCESS_32x2x4  0, %4, %5, (%4 + 32), (%5 + 32), %6
 %endmacro
 
 ; void vp9_highbd_sadNxNx4d_sse2(uint8_t *src,    int src_stride,
 ;                         uint8_t *ref[4], int ref_stride,
 ;                         unsigned int res[4]);
 ; where NxN = 64x64, 32x32, 16x16, 16x8, 8x16 or 8x8
 %macro HIGH_SADNXN4D 2
 %if UNIX64
-cglobal highbd_sad%1x%2x4d, 5, 9, 8, src, src_stride, ref1, ref_stride, \
-                              res, ref2, ref3, ref4, one
+cglobal highbd_sad%1x%2x4d, 5, 8, 8, src, src_stride, ref1, ref_stride, \
+                              res, ref2, ref3, ref4
 %else
-cglobal highbd_sad%1x%2x4d, 4, 8, 8, src, src_stride, ref1, ref_stride, \
-                              ref2, ref3, ref4, one
+cglobal highbd_sad%1x%2x4d, 4, 7, 8, src, src_stride, ref1, ref_stride, \
+                              ref2, ref3, ref4
 %endif
 
+; set m1
+  push                srcq
+  mov                 srcd, 0x00010001
+  movd                  m1, srcd
+  pshufd                m1, m1, 0x0
+  pop                 srcq
+
   movsxdifnidn src_strideq, src_strided
   movsxdifnidn ref_strideq, ref_strided
   mov                ref2q, [ref1q+gprsize*1]
   mov                ref3q, [ref1q+gprsize*2]
   mov                ref4q, [ref1q+gprsize*3]
   mov                ref1q, [ref1q+gprsize*0]
 
 ; convert byte pointers to short pointers
   shl                 srcq, 1
   shl                ref2q, 1
   shl                ref3q, 1
   shl                ref4q, 1
   shl                ref1q, 1
 
-  mov                 oned, 0x00010001
-  movd                  m1, oned
-  pshufd                m1, m1, 0x0
-
   HIGH_PROCESS_%1x2x4 1, 0, 0, src_strideq, ref_strideq, 1
 %rep (%2-4)/2
   HIGH_PROCESS_%1x2x4 0, 0, 0, src_strideq, ref_strideq, 1
 %endrep
   HIGH_PROCESS_%1x2x4 0, 0, 0, src_strideq, ref_strideq, 0
   ; N.B. HIGH_PROCESS outputs dwords (32 bits)
   ; so in high bit depth even the smallest width (4) needs 128bits i.e. XMM
   movhlps               m0, m4
   movhlps               m1, m5
   movhlps               m2, m6
@@ skipping 22 matching lines @@
 HIGH_SADNXN4D 32, 32
 HIGH_SADNXN4D 32, 16
 HIGH_SADNXN4D 16, 32
 HIGH_SADNXN4D 16, 16
 HIGH_SADNXN4D 16,  8
 HIGH_SADNXN4D  8, 16
 HIGH_SADNXN4D  8,  8
 HIGH_SADNXN4D  8,  4
 HIGH_SADNXN4D  4,  8
 HIGH_SADNXN4D  4,  4