libvpx: Pull from upstream

source/libvpx/vp9/common/x86/vp9_idct_intrin_sse2.c

 /*
  *  Copyright (c) 2012 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 3555 matching lines @@
       RECON_AND_STORE(dest, in[31]);
 
       dest += 8 - (stride * 32);
     }
   }
 
 void vp9_idct32x32_1024_add_sse2(const int16_t *input, uint8_t *dest,
                                  int stride) {
   const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
   const __m128i final_rounding = _mm_set1_epi16(1<<5);
+  const __m128i zero = _mm_setzero_si128();
 
   // idct constants for each stage
   const __m128i stg1_0 = pair_set_epi16(cospi_31_64, -cospi_1_64);
   const __m128i stg1_1 = pair_set_epi16(cospi_1_64, cospi_31_64);
   const __m128i stg1_2 = pair_set_epi16(cospi_15_64, -cospi_17_64);
   const __m128i stg1_3 = pair_set_epi16(cospi_17_64, cospi_15_64);
   const __m128i stg1_4 = pair_set_epi16(cospi_23_64, -cospi_9_64);
   const __m128i stg1_5 = pair_set_epi16(cospi_9_64, cospi_23_64);
   const __m128i stg1_6 = pair_set_epi16(cospi_7_64, -cospi_25_64);
   const __m128i stg1_7 = pair_set_epi16(cospi_25_64, cospi_7_64);
@@ skipping 42 matching lines @@
           stp1_16, stp1_17, stp1_18, stp1_19, stp1_20, stp1_21, stp1_22,
           stp1_23, stp1_24, stp1_25, stp1_26, stp1_27, stp1_28, stp1_29,
           stp1_30, stp1_31;
   __m128i stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7,
           stp2_8, stp2_9, stp2_10, stp2_11, stp2_12, stp2_13, stp2_14, stp2_15,
           stp2_16, stp2_17, stp2_18, stp2_19, stp2_20, stp2_21, stp2_22,
           stp2_23, stp2_24, stp2_25, stp2_26, stp2_27, stp2_28, stp2_29,
           stp2_30, stp2_31;
   __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   int i, j, i32;
-  int zero_flag[2];
 
   for (i = 0; i < 4; i++) {
     i32 = (i << 5);
       // First 1-D idct
       // Load input data.
       LOAD_DQCOEFF(in[0], input);
       LOAD_DQCOEFF(in[8], input);
       LOAD_DQCOEFF(in[16], input);
       LOAD_DQCOEFF(in[24], input);
       LOAD_DQCOEFF(in[1], input);
@@ skipping 54 matching lines @@
       zero_idx[7] = _mm_or_si128(zero_idx[14], zero_idx[15]);
 
       zero_idx[8] = _mm_or_si128(zero_idx[0], zero_idx[1]);
       zero_idx[9] = _mm_or_si128(zero_idx[2], zero_idx[3]);
       zero_idx[10] = _mm_or_si128(zero_idx[4], zero_idx[5]);
       zero_idx[11] = _mm_or_si128(zero_idx[6], zero_idx[7]);
       zero_idx[12] = _mm_or_si128(zero_idx[8], zero_idx[9]);
       zero_idx[13] = _mm_or_si128(zero_idx[10], zero_idx[11]);
       zero_idx[14] = _mm_or_si128(zero_idx[12], zero_idx[13]);
 
-      zero_idx[0] = _mm_unpackhi_epi64(zero_idx[14], zero_idx[14]);
-      zero_idx[1] = _mm_or_si128(zero_idx[0], zero_idx[14]);
-      zero_idx[2] = _mm_srli_epi64(zero_idx[1], 32);
-      zero_flag[0] = _mm_cvtsi128_si32(zero_idx[1]);
-      zero_flag[1] = _mm_cvtsi128_si32(zero_idx[2]);
-
-      if (!zero_flag[0] && !zero_flag[1]) {
+      if (_mm_movemask_epi8(_mm_cmpeq_epi32(zero_idx[14], zero)) == 0xFFFF) {
         col[i32 + 0] = _mm_setzero_si128();
         col[i32 + 1] = _mm_setzero_si128();
         col[i32 + 2] = _mm_setzero_si128();
         col[i32 + 3] = _mm_setzero_si128();
         col[i32 + 4] = _mm_setzero_si128();
         col[i32 + 5] = _mm_setzero_si128();
         col[i32 + 6] = _mm_setzero_si128();
         col[i32 + 7] = _mm_setzero_si128();
         col[i32 + 8] = _mm_setzero_si128();
         col[i32 + 9] = _mm_setzero_si128();
@@ skipping 58 matching lines @@
       col[i32 + 24] = _mm_sub_epi16(stp1_7, stp1_24);
       col[i32 + 25] = _mm_sub_epi16(stp1_6, stp1_25);
       col[i32 + 26] = _mm_sub_epi16(stp1_5, stp1_26);
       col[i32 + 27] = _mm_sub_epi16(stp1_4, stp1_27);
       col[i32 + 28] = _mm_sub_epi16(stp1_3, stp1_28);
       col[i32 + 29] = _mm_sub_epi16(stp1_2, stp1_29);
       col[i32 + 30] = _mm_sub_epi16(stp1_1, stp1_30);
       col[i32 + 31] = _mm_sub_epi16(stp1_0, stp1_31);
     }
   for (i = 0; i < 4; i++) {
-      const __m128i zero = _mm_setzero_si128();
       // Second 1-D idct
       j = i << 3;
 
       // Transpose 32x8 block to 8x32 block
       array_transpose_8x8(col+j, in);
       array_transpose_8x8(col+j+32, in+8);
       array_transpose_8x8(col+j+64, in+16);
       array_transpose_8x8(col+j+96, in+24);
 
       IDCT32
@@ skipping 176 matching lines @@
     RECON_AND_STORE(dest, dc_value);
     RECON_AND_STORE(dest, dc_value);
     RECON_AND_STORE(dest, dc_value);
     RECON_AND_STORE(dest, dc_value);
     RECON_AND_STORE(dest, dc_value);
     RECON_AND_STORE(dest, dc_value);
     RECON_AND_STORE(dest, dc_value);
     dest += 8 - (stride * 32);
   }
 }