libvpx: Pull from upstream

source/libvpx/vp9/encoder/x86/vp9_dct_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.
  */
 
 #include <assert.h>
 #include <emmintrin.h>  // SSE2
+
+#include "./vp9_rtcd.h"
 #include "vp9/common/vp9_idct.h"  // for cospi constants
 #include "vp9/encoder/vp9_dct.h"
 #include "vp9/encoder/x86/vp9_dct_sse2.h"
 #include "vpx_ports/mem.h"
 
 void vp9_fdct4x4_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
   __m128i in0, in1;
   __m128i tmp;
   const __m128i zero = _mm_setzero_si128();
   in0  = _mm_loadl_epi64((const __m128i *)(input +  0 * stride));
@@ skipping 66 matching lines @@
   res[0] = _mm_unpacklo_epi32(tr0_0, tr0_1);
   res[2] = _mm_unpackhi_epi32(tr0_0, tr0_1);
 
   // 00 10 20 30 01 11 21 31
   // 02 12 22 32 03 13 23 33
   // only use the first 4 16-bit integers
   res[1] = _mm_unpackhi_epi64(res[0], res[0]);
   res[3] = _mm_unpackhi_epi64(res[2], res[2]);
 }
 
-void fdct4_sse2(__m128i *in) {
+static void fdct4_sse2(__m128i *in) {
   const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t)cospi_16_64);
   const __m128i k__cospi_p16_m16 = pair_set_epi16(cospi_16_64, -cospi_16_64);
   const __m128i k__cospi_p08_p24 = pair_set_epi16(cospi_8_64, cospi_24_64);
   const __m128i k__cospi_p24_m08 = pair_set_epi16(cospi_24_64, -cospi_8_64);
   const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
 
   __m128i u[4], v[4];
   u[0]=_mm_unpacklo_epi16(in[0], in[1]);
   u[1]=_mm_unpacklo_epi16(in[3], in[2]);
 
@@ skipping 12 matching lines @@
   u[0] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
   u[1] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
   u[2] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
   u[3] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
 
   in[0] = _mm_packs_epi32(u[0], u[1]);
   in[1] = _mm_packs_epi32(u[2], u[3]);
   transpose_4x4(in);
 }
 
-void fadst4_sse2(__m128i *in) {
+static void fadst4_sse2(__m128i *in) {
   const __m128i k__sinpi_p01_p02 = pair_set_epi16(sinpi_1_9, sinpi_2_9);
   const __m128i k__sinpi_p04_m01 = pair_set_epi16(sinpi_4_9, -sinpi_1_9);
   const __m128i k__sinpi_p03_p04 = pair_set_epi16(sinpi_3_9, sinpi_4_9);
   const __m128i k__sinpi_m03_p02 = pair_set_epi16(-sinpi_3_9, sinpi_2_9);
   const __m128i k__sinpi_p03_p03 = _mm_set1_epi16((int16_t)sinpi_3_9);
   const __m128i kZero = _mm_set1_epi16(0);
   const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
   __m128i u[8], v[8];
   __m128i in7 = _mm_add_epi16(in[0], in[1]);
 
@@ skipping 681 matching lines @@
   // 00 10 20 30 40 50 60 70
   // 01 11 21 31 41 51 61 71
   // 02 12 22 32 42 52 62 72
   // 03 13 23 33 43 53 63 73
   // 04 14 24 34 44 54 64 74
   // 05 15 25 35 45 55 65 75
   // 06 16 26 36 46 56 66 76
   // 07 17 27 37 47 57 67 77
 }
 
-void fdct8_sse2(__m128i *in) {
+static void fdct8_sse2(__m128i *in) {
   // constants
   const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t)cospi_16_64);
   const __m128i k__cospi_p16_m16 = pair_set_epi16(cospi_16_64, -cospi_16_64);
   const __m128i k__cospi_p24_p08 = pair_set_epi16(cospi_24_64, cospi_8_64);
   const __m128i k__cospi_m08_p24 = pair_set_epi16(-cospi_8_64, cospi_24_64);
   const __m128i k__cospi_p28_p04 = pair_set_epi16(cospi_28_64, cospi_4_64);
   const __m128i k__cospi_m04_p28 = pair_set_epi16(-cospi_4_64, cospi_28_64);
   const __m128i k__cospi_p12_p20 = pair_set_epi16(cospi_12_64, cospi_20_64);
   const __m128i k__cospi_m20_p12 = pair_set_epi16(-cospi_20_64, cospi_12_64);
   const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
@@ skipping 119 matching lines @@
 
   in[1] = _mm_packs_epi32(v0, v1);
   in[3] = _mm_packs_epi32(v4, v5);
   in[5] = _mm_packs_epi32(v2, v3);
   in[7] = _mm_packs_epi32(v6, v7);
 
   // transpose
   array_transpose_8x8(in, in);
 }
 
-void fadst8_sse2(__m128i *in) {
+static void fadst8_sse2(__m128i *in) {
   // Constants
   const __m128i k__cospi_p02_p30 = pair_set_epi16(cospi_2_64, cospi_30_64);
   const __m128i k__cospi_p30_m02 = pair_set_epi16(cospi_30_64, -cospi_2_64);
   const __m128i k__cospi_p10_p22 = pair_set_epi16(cospi_10_64, cospi_22_64);
   const __m128i k__cospi_p22_m10 = pair_set_epi16(cospi_22_64, -cospi_10_64);
   const __m128i k__cospi_p18_p14 = pair_set_epi16(cospi_18_64, cospi_14_64);
   const __m128i k__cospi_p14_m18 = pair_set_epi16(cospi_14_64, -cospi_18_64);
   const __m128i k__cospi_p26_p06 = pair_set_epi16(cospi_26_64, cospi_6_64);
   const __m128i k__cospi_p06_m26 = pair_set_epi16(cospi_6_64, -cospi_26_64);
   const __m128i k__cospi_p08_p24 = pair_set_epi16(cospi_8_64, cospi_24_64);
@@ skipping 361 matching lines @@
 }
 
 static INLINE void right_shift_16x16(__m128i *res0, __m128i *res1) {
   // perform rounding operations
   right_shift_8x8(res0, 2);
   right_shift_8x8(res0 + 8, 2);
   right_shift_8x8(res1, 2);
   right_shift_8x8(res1 + 8, 2);
 }
 
-void fdct16_8col(__m128i *in) {
+static void fdct16_8col(__m128i *in) {
   // perform 16x16 1-D DCT for 8 columns
   __m128i i[8], s[8], p[8], t[8], u[16], v[16];
   const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t)cospi_16_64);
   const __m128i k__cospi_p16_m16 = pair_set_epi16(cospi_16_64, -cospi_16_64);
   const __m128i k__cospi_m16_p16 = pair_set_epi16(-cospi_16_64, cospi_16_64);
   const __m128i k__cospi_p24_p08 = pair_set_epi16(cospi_24_64, cospi_8_64);
   const __m128i k__cospi_p08_m24 = pair_set_epi16(cospi_8_64, -cospi_24_64);
   const __m128i k__cospi_m08_p24 = pair_set_epi16(-cospi_8_64, cospi_24_64);
   const __m128i k__cospi_p28_p04 = pair_set_epi16(cospi_28_64, cospi_4_64);
   const __m128i k__cospi_m04_p28 = pair_set_epi16(-cospi_4_64, cospi_28_64);
@@ skipping 301 matching lines @@
   in[1]  = _mm_packs_epi32(v[0], v[1]);
   in[9]  = _mm_packs_epi32(v[2], v[3]);
   in[5]  = _mm_packs_epi32(v[4], v[5]);
   in[13] = _mm_packs_epi32(v[6], v[7]);
   in[3]  = _mm_packs_epi32(v[8], v[9]);
   in[11] = _mm_packs_epi32(v[10], v[11]);
   in[7]  = _mm_packs_epi32(v[12], v[13]);
   in[15] = _mm_packs_epi32(v[14], v[15]);
 }
 
-void fadst16_8col(__m128i *in) {
+static void fadst16_8col(__m128i *in) {
   // perform 16x16 1-D ADST for 8 columns
   __m128i s[16], x[16], u[32], v[32];
   const __m128i k__cospi_p01_p31 = pair_set_epi16(cospi_1_64, cospi_31_64);
   const __m128i k__cospi_p31_m01 = pair_set_epi16(cospi_31_64, -cospi_1_64);
   const __m128i k__cospi_p05_p27 = pair_set_epi16(cospi_5_64, cospi_27_64);
   const __m128i k__cospi_p27_m05 = pair_set_epi16(cospi_27_64, -cospi_5_64);
   const __m128i k__cospi_p09_p23 = pair_set_epi16(cospi_9_64, cospi_23_64);
   const __m128i k__cospi_p23_m09 = pair_set_epi16(cospi_23_64, -cospi_9_64);
   const __m128i k__cospi_p13_p19 = pair_set_epi16(cospi_13_64, cospi_19_64);
   const __m128i k__cospi_p19_m13 = pair_set_epi16(cospi_19_64, -cospi_13_64);
@@ skipping 449 matching lines @@
   in[8] = _mm_packs_epi32(v[2], v[3]);
   in[9] = _mm_packs_epi32(v[10], v[11]);
   in[10] = _mm_packs_epi32(v[14], v[15]);
   in[11] = _mm_packs_epi32(v[6], v[7]);
   in[12] = s[5];
   in[13] = _mm_sub_epi16(kZero, s[13]);
   in[14] = s[9];
   in[15] = _mm_sub_epi16(kZero, s[1]);
 }
 
-void fdct16_sse2(__m128i *in0, __m128i *in1) {
+static void fdct16_sse2(__m128i *in0, __m128i *in1) {
   fdct16_8col(in0);
   fdct16_8col(in1);
   array_transpose_16x16(in0, in1);
 }
 
-void fadst16_sse2(__m128i *in0, __m128i *in1) {
+static void fadst16_sse2(__m128i *in0, __m128i *in1) {
   fadst16_8col(in0);
   fadst16_8col(in1);
   array_transpose_16x16(in0, in1);
 }
 
 void vp9_fht16x16_sse2(const int16_t *input, tran_low_t *output,
                        int stride, int tx_type) {
   __m128i in0[16], in1[16];
 
   switch (tx_type) {
@@ skipping 162 matching lines @@
     for (i = 0; i < 8; ++i) {
       for (j = 0; j < 8; ++j)
         temp_in[j] = out[j + i * 8];
       ht.rows(temp_in, temp_out);
       for (j = 0; j < 8; ++j)
         output[j + i * 8] = (temp_out[j] + (temp_out[j] < 0)) >> 1;
     }
   }
 }
 
-void vp9_highbd_fht16x16_sse2(int16_t *input, tran_low_t *output,
+void vp9_highbd_fht16x16_sse2(const int16_t *input, tran_low_t *output,
                               int stride, int tx_type) {
   if (tx_type == DCT_DCT) {
     vp9_highbd_fdct16x16_sse2(input, output, stride);
   } else {
     tran_low_t out[256];
     tran_low_t *outptr = &out[0];
     int i, j;
     tran_low_t temp_in[16], temp_out[16];
     const transform_2d ht = FHT_16[tx_type];
 
@@ skipping 13 matching lines @@
       ht.rows(temp_in, temp_out);
       for (j = 0; j < 16; ++j)
         output[j + i * 16] = temp_out[j];
     }
   }
 }
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
 /*
  * The DCTnxn functions are defined using the macros below. The main code for
- * them is in separate files (vp9/encoder/x86/vp9_dct_impl_sse2.c &
- * vp9/encoder/x86/vp9_dct32x32_sse2.c) which are used by both the 8 bit code
+ * them is in separate files (vp9/encoder/x86/vp9_dct_sse2_impl.h &
+ * vp9/encoder/x86/vp9_dct32x32_sse2_impl.h) which are used by both the 8 bit code
  * and the high bit depth code.
  */
 
 #define DCT_HIGH_BIT_DEPTH 0
 
 #define FDCT4x4_2D vp9_fdct4x4_sse2
 #define FDCT8x8_2D vp9_fdct8x8_sse2
 #define FDCT16x16_2D vp9_fdct16x16_sse2
-#include "vp9/encoder/x86/vp9_dct_impl_sse2.c"
+#include "vp9/encoder/x86/vp9_dct_sse2_impl.h"
 #undef  FDCT4x4_2D
 #undef  FDCT8x8_2D
 #undef  FDCT16x16_2D
 
 #define FDCT32x32_2D vp9_fdct32x32_rd_sse2
 #define FDCT32x32_HIGH_PRECISION 0
-#include "vp9/encoder/x86/vp9_dct32x32_sse2.c"
+#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h"
 #undef  FDCT32x32_2D
 #undef  FDCT32x32_HIGH_PRECISION
 
 #define FDCT32x32_2D vp9_fdct32x32_sse2
 #define FDCT32x32_HIGH_PRECISION 1
-#include "vp9/encoder/x86/vp9_dct32x32_sse2.c" // NOLINT
+#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h" // NOLINT
 #undef  FDCT32x32_2D
 #undef  FDCT32x32_HIGH_PRECISION
 
 #undef  DCT_HIGH_BIT_DEPTH
 
 
 #if CONFIG_VP9_HIGHBITDEPTH
 
 #define DCT_HIGH_BIT_DEPTH 1
 
 #define FDCT4x4_2D vp9_highbd_fdct4x4_sse2
 #define FDCT8x8_2D vp9_highbd_fdct8x8_sse2
 #define FDCT16x16_2D vp9_highbd_fdct16x16_sse2
-#include "vp9/encoder/x86/vp9_dct_impl_sse2.c" // NOLINT
+#include "vp9/encoder/x86/vp9_dct_sse2_impl.h" // NOLINT
 #undef  FDCT4x4_2D
 #undef  FDCT8x8_2D
 #undef  FDCT16x16_2D
 
 #define FDCT32x32_2D vp9_highbd_fdct32x32_rd_sse2
 #define FDCT32x32_HIGH_PRECISION 0
-#include "vp9/encoder/x86/vp9_dct32x32_sse2.c" // NOLINT
+#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h" // NOLINT
 #undef  FDCT32x32_2D
 #undef  FDCT32x32_HIGH_PRECISION
 
 #define FDCT32x32_2D vp9_highbd_fdct32x32_sse2
 #define FDCT32x32_HIGH_PRECISION 1
-#include "vp9/encoder/x86/vp9_dct32x32_sse2.c" // NOLINT
+#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h" // NOLINT
 #undef  FDCT32x32_2D
 #undef  FDCT32x32_HIGH_PRECISION
 
 #undef  DCT_HIGH_BIT_DEPTH
 
 #endif  // CONFIG_VP9_HIGHBITDEPTH