libvpx: Pull from upstream

source/libvpx/vp9/encoder/x86/vp9_quantize_ssse3_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.
 ;
 
 %include "third_party/x86inc/x86inc.asm"
 
 SECTION_RODATA
 pw_1: times 8 dw 1
 
 SECTION .text
 
 %macro QUANTIZE_FN 2
 cglobal quantize_%1, 0, %2, 15, coeff, ncoeff, skip, zbin, round, quant, \
-                                shift, qcoeff, dqcoeff, dequant, zbin_oq, \
+                                shift, qcoeff, dqcoeff, dequant, \
                                 eob, scan, iscan
   cmp                    dword skipm, 0
   jne .blank
 
   ; actual quantize loop - setup pointers, rounders, etc.
   movifnidn                   coeffq, coeffmp
   movifnidn                  ncoeffq, ncoeffmp
   mov                             r2, dequantmp
   movifnidn                    zbinq, zbinmp
   movifnidn                   roundq, roundmp
   movifnidn                   quantq, quantmp
-  movd                            m4, dword zbin_oqm       ; m4 = zbin_oq
   mova                            m0, [zbinq]              ; m0 = zbin
-  punpcklwd                       m4, m4
   mova                            m1, [roundq]             ; m1 = round
-  pshufd                          m4, m4, 0
   mova                            m2, [quantq]             ; m2 = quant
-  paddw                           m0, m4                   ; m0 = zbin + zbin_oq
 %ifidn %1, b_32x32
   pcmpeqw                         m5, m5
   psrlw                           m5, 15
   paddw                           m0, m5
   paddw                           m1, m5
   psrlw                           m0, 1                    ; m0 = (m0 + 1) / 2
   psrlw                           m1, 1                    ; m1 = (m1 + 1) / 2
 %endif
   mova                            m3, [r2q]                ; m3 = dequant
   psubw                           m0, [pw_1]
   mov                             r2, shiftmp
   mov                             r3, qcoeffmp
   mova                            m4, [r2]                 ; m4 = shift
   mov                             r4, dqcoeffmp
   mov                             r5, iscanmp
 %ifidn %1, b_32x32
   psllw                           m4, 1
 %endif
   pxor                            m5, m5                   ; m5 = dedicated zero
-  DEFINE_ARGS coeff, ncoeff, d1, qcoeff, dqcoeff, iscan, d2, d3, d4, d5, d6, eob
+  DEFINE_ARGS coeff, ncoeff, d1, qcoeff, dqcoeff, iscan, d2, d3, d4, d5, eob
   lea                         coeffq, [  coeffq+ncoeffq*2]
   lea                         iscanq, [  iscanq+ncoeffq*2]
   lea                        qcoeffq, [ qcoeffq+ncoeffq*2]
   lea                       dqcoeffq, [dqcoeffq+ncoeffq*2]
   neg                        ncoeffq
 
   ; get DC and first 15 AC coeffs
   mova                            m9, [  coeffq+ncoeffq*2+ 0] ; m9 = c[i]
   mova                           m10, [  coeffq+ncoeffq*2+16] ; m10 = c[i]
   pabsw                           m6, m9                   ; m6 = abs(m9)
@@ skipping 46 matching lines @@
   jz .accumulate_eob
 
 .ac_only_loop:
   mova                            m9, [  coeffq+ncoeffq*2+ 0] ; m9 = c[i]
   mova                           m10, [  coeffq+ncoeffq*2+16] ; m10 = c[i]
   pabsw                           m6, m9                   ; m6 = abs(m9)
   pabsw                          m11, m10                  ; m11 = abs(m10)
   pcmpgtw                         m7, m6, m0               ; m7 = c[i] >= zbin
   pcmpgtw                        m12, m11, m0              ; m12 = c[i] >= zbin
 %ifidn %1, b_32x32
-  pmovmskb                        r6, m7
-  pmovmskb                        r2, m12
+  pmovmskb                       r6d, m7
+  pmovmskb                       r2d, m12
   or                              r6, r2
   jz .skip_iter
 %endif
   paddsw                          m6, m1                   ; m6 += round
   paddsw                         m11, m1                   ; m11 += round
   pmulhw                         m14, m6, m2               ; m14 = m6*q>>16
   pmulhw                         m13, m11, m2              ; m13 = m11*q>>16
   paddw                          m14, m6                   ; m14 += m6
   paddw                          m13, m11                  ; m13 += m11
   pmulhw                         m14, m4                   ; m14 = m14*qsh>>16
@@ skipping 76 matching lines @@
   mov                    word [eobq], 0
   RET
 %endmacro
 
 INIT_XMM ssse3
 QUANTIZE_FN b, 7
 QUANTIZE_FN b_32x32, 7
 
 %macro QUANTIZE_FP 2
 cglobal quantize_%1, 0, %2, 15, coeff, ncoeff, skip, zbin, round, quant, \
-                                shift, qcoeff, dqcoeff, dequant, zbin_oq, \
+                                shift, qcoeff, dqcoeff, dequant, \
                                 eob, scan, iscan
   cmp                    dword skipm, 0
   jne .blank
 
   ; actual quantize loop - setup pointers, rounders, etc.
   movifnidn                   coeffq, coeffmp
   movifnidn                  ncoeffq, ncoeffmp
   mov                             r2, dequantmp
   movifnidn                    zbinq, zbinmp
   movifnidn                   roundq, roundmp
   movifnidn                   quantq, quantmp
   mova                            m1, [roundq]             ; m1 = round
   mova                            m2, [quantq]             ; m2 = quant
 %ifidn %1, fp_32x32
   pcmpeqw                         m5, m5
   psrlw                           m5, 15
   paddw                           m1, m5
   psrlw                           m1, 1                    ; m1 = (m1 + 1) / 2
 %endif
   mova                            m3, [r2q]                ; m3 = dequant
   mov                             r3, qcoeffmp
   mov                             r4, dqcoeffmp
   mov                             r5, iscanmp
 %ifidn %1, fp_32x32
   psllw                           m2, 1
 %endif
   pxor                            m5, m5                   ; m5 = dedicated zero
-  DEFINE_ARGS coeff, ncoeff, d1, qcoeff, dqcoeff, iscan, d2, d3, d4, d5, d6, eob
+  DEFINE_ARGS coeff, ncoeff, d1, qcoeff, dqcoeff, iscan, d2, d3, d4, d5, eob
   lea                         coeffq, [  coeffq+ncoeffq*2]
   lea                         iscanq, [  iscanq+ncoeffq*2]
   lea                        qcoeffq, [ qcoeffq+ncoeffq*2]
   lea                       dqcoeffq, [dqcoeffq+ncoeffq*2]
   neg                        ncoeffq
 
   ; get DC and first 15 AC coeffs
   mova                            m9, [  coeffq+ncoeffq*2+ 0] ; m9 = c[i]
   mova                           m10, [  coeffq+ncoeffq*2+16] ; m10 = c[i]
   pabsw                           m6, m9                   ; m6 = abs(m9)
@@ skipping 39 matching lines @@
   jz .accumulate_eob
 
 .ac_only_loop:
   mova                            m9, [  coeffq+ncoeffq*2+ 0] ; m9 = c[i]
   mova                           m10, [  coeffq+ncoeffq*2+16] ; m10 = c[i]
   pabsw                           m6, m9                   ; m6 = abs(m9)
   pabsw                          m11, m10                  ; m11 = abs(m10)
 %ifidn %1, fp_32x32
   pcmpgtw                         m7, m6,  m0
   pcmpgtw                        m12, m11, m0
-  pmovmskb                        r6, m7
-  pmovmskb                        r2, m12
+  pmovmskb                       r6d, m7
+  pmovmskb                       r2d, m12
 
   or                              r6, r2
   jz .skip_iter
 %endif
   pcmpeqw                         m7, m7
 
   paddsw                          m6, m1                   ; m6 += round
   paddsw                         m11, m1                   ; m11 += round
   pmulhw                         m14, m6, m2               ; m14 = m6*q>>16
   pmulhw                         m13, m11, m2              ; m13 = m11*q>>16
@@ skipping 70 matching lines @@
   mova        [qcoeffq+ncoeffq*2+16], m7
   add                        ncoeffq, mmsize
   jl .blank_loop
   mov                    word [eobq], 0
   RET
 %endmacro
 
 INIT_XMM ssse3
 QUANTIZE_FP fp, 7
 QUANTIZE_FP fp_32x32, 7