libvpx: Pull from upstream

source/libvpx/vpx_dsp/x86/ssim_opt_x86_64.asm

 ;
 ;  Copyright (c) 2010 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 31 matching lines @@
         punpcklwd       %1,xmm0
         punpckhwd       xmm1,xmm0
         paddd           %1, xmm1
         SUM_ACROSS_Q    %1
 %endmacro
 ;void ssim_parms_sse2(
 ;    unsigned char *s,
 ;    int sp,
 ;    unsigned char *r,
 ;    int rp
-;    unsigned long *sum_s,
-;    unsigned long *sum_r,
-;    unsigned long *sum_sq_s,
-;    unsigned long *sum_sq_r,
-;    unsigned long *sum_sxr);
+;    uint32_t *sum_s,
+;    uint32_t *sum_r,
+;    uint32_t *sum_sq_s,
+;    uint32_t *sum_sq_r,
+;    uint32_t *sum_sxr);
 ;
 ; TODO: Use parm passing through structure, probably don't need the pxors
 ; ( calling app will initialize to 0 ) could easily fit everything in sse2
 ; without too much hastle, and can probably do better estimates with psadw
 ; or pavgb At this point this is just meant to be first pass for calculating
 ; all the parms needed for 16x16 ssim so we can play with dssim as distortion
 ; in mode selection code.
-global sym(vp8_ssim_parms_16x16_sse2) PRIVATE
-sym(vp8_ssim_parms_16x16_sse2):
+global sym(vpx_ssim_parms_16x16_sse2) PRIVATE
+sym(vpx_ssim_parms_16x16_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 9
     SAVE_XMM 15
     push        rsi
     push        rdi
     ; end prolog
 
     mov             rsi,        arg(0) ;s
     mov             rcx,        arg(1) ;sp
@@ skipping 56 matching lines @@
     RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
 
 ;void ssim_parms_sse2(
 ;    unsigned char *s,
 ;    int sp,
 ;    unsigned char *r,
 ;    int rp
-;    unsigned long *sum_s,
-;    unsigned long *sum_r,
-;    unsigned long *sum_sq_s,
-;    unsigned long *sum_sq_r,
-;    unsigned long *sum_sxr);
+;    uint32_t *sum_s,
+;    uint32_t *sum_r,
+;    uint32_t *sum_sq_s,
+;    uint32_t *sum_sq_r,
+;    uint32_t *sum_sxr);
 ;
 ; TODO: Use parm passing through structure, probably don't need the pxors
 ; ( calling app will initialize to 0 ) could easily fit everything in sse2
 ; without too much hastle, and can probably do better estimates with psadw
 ; or pavgb At this point this is just meant to be first pass for calculating
 ; all the parms needed for 16x16 ssim so we can play with dssim as distortion
 ; in mode selection code.
-global sym(vp8_ssim_parms_8x8_sse2) PRIVATE
-sym(vp8_ssim_parms_8x8_sse2):
+global sym(vpx_ssim_parms_8x8_sse2) PRIVATE
+sym(vpx_ssim_parms_8x8_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 9
     SAVE_XMM 15
     push        rsi
     push        rdi
     ; end prolog
 
     mov             rsi,        arg(0) ;s
     mov             rcx,        arg(1) ;sp
@@ skipping 41 matching lines @@
     mov             rdi,arg(8)
     movd            [rdi], xmm11;
 
     ; begin epilog
     pop         rdi
     pop         rsi
     RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret