From d2de1ca37b630e691d64401b4a4d74d862981dc9 Mon Sep 17 00:00:00 2001
From: Jingning Han <jingning@google.com>
Date: Wed, 24 Jul 2013 16:33:57 -0700
Subject: [PATCH] Merge vp9_dc_only_idct_add and vp9_short_idct4x4_1

They share the same functionality, so merging together.

Change-Id: I98a0386fcee052cb854f9ff90c283c1b844bcb79
---
 vp9/common/vp9_idct.c                 | 17 --------
 vp9/common/vp9_rtcd_defs.sh           |  3 --
 vp9/common/x86/vp9_idct_intrin_sse2.c | 58 ---------------------------
 vp9/decoder/vp9_idct_blk.c            |  2 +-
 4 files changed, 1 insertion(+), 79 deletions(-)

diff --git a/vp9/common/vp9_idct.c b/vp9/common/vp9_idct.c
index 3103be214..a95560a55 100644
--- a/vp9/common/vp9_idct.c
+++ b/vp9/common/vp9_idct.c
@@ -156,23 +156,6 @@ void vp9_short_idct4x4_1_add_c(int16_t *input, uint8_t *dest, int dest_stride) {
   }
 }
 
-void vp9_dc_only_idct_add_c(int input_dc, uint8_t *pred_ptr,
-                            uint8_t *dst_ptr, int pitch, int stride) {
-  int a1;
-  int r, c;
-  int16_t out = dct_const_round_shift(input_dc * cospi_16_64);
-  out = dct_const_round_shift(out * cospi_16_64);
-  a1 = ROUND_POWER_OF_TWO(out, 4);
-
-  for (r = 0; r < 4; r++) {
-    for (c = 0; c < 4; c++)
-      dst_ptr[c] = clip_pixel(a1 + pred_ptr[c]);
-
-    dst_ptr += stride;
-    pred_ptr += pitch;
-  }
-}
-
 static void idct8_1d(int16_t *input, int16_t *output) {
   int16_t step1[8], step2[8];
   int temp1, temp2;
diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
index 2a9e62d0f..f1789fb74 100644
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -337,9 +337,6 @@ prototype void vp9_idct4_1d "int16_t *input, int16_t *output"
 specialize vp9_idct4_1d sse2
 # dct and add
 
-prototype void vp9_dc_only_idct_add "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride"
-specialize vp9_dc_only_idct_add sse2 neon
-
 prototype void vp9_short_iwalsh4x4_1_add "int16_t *input, uint8_t *dest, int dest_stride"
 specialize vp9_short_iwalsh4x4_1_add
 
diff --git a/vp9/common/x86/vp9_idct_intrin_sse2.c b/vp9/common/x86/vp9_idct_intrin_sse2.c
index c380d4216..b4766df5b 100644
--- a/vp9/common/x86/vp9_idct_intrin_sse2.c
+++ b/vp9/common/x86/vp9_idct_intrin_sse2.c
@@ -15,64 +15,6 @@
 #include "vp9/common/vp9_common.h"
 #include "vp9/common/vp9_idct.h"
 
-// In order to improve performance, clip absolute diff values to [0, 255],
-// which allows to keep the additions/subtractions in 8 bits.
-void vp9_dc_only_idct_add_sse2(int input_dc, uint8_t *pred_ptr,
-                               uint8_t *dst_ptr, int pitch, int stride) {
-  int a1;
-  int16_t out;
-  uint8_t abs_diff;
-  __m128i p0, p1, p2, p3;
-  unsigned int extended_diff;
-  __m128i diff;
-
-  out = dct_const_round_shift(input_dc * cospi_16_64);
-  out = dct_const_round_shift(out * cospi_16_64);
-  a1 = ROUND_POWER_OF_TWO(out, 4);
-
-  // Read prediction data.
-  p0 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 0 * pitch));
-  p1 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 1 * pitch));
-  p2 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 2 * pitch));
-  p3 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 3 * pitch));
-
-  // Unpack prediction data, and store 4x4 array in 1 XMM register.
-  p0 = _mm_unpacklo_epi32(p0, p1);
-  p2 = _mm_unpacklo_epi32(p2, p3);
-  p0 = _mm_unpacklo_epi64(p0, p2);
-
-  // Clip dc value to [0, 255] range. Then, do addition or subtraction
-  // according to its sign.
-  if (a1 >= 0) {
-    abs_diff = (a1 > 255) ? 255 : a1;
-    extended_diff = abs_diff * 0x01010101u;
-    diff = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_diff), 0);
-
-    p1 = _mm_adds_epu8(p0, diff);
-  } else {
-    abs_diff = (a1 < -255) ? 255 : -a1;
-    extended_diff = abs_diff * 0x01010101u;
-    diff = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_diff), 0);
-
-    p1 = _mm_subs_epu8(p0, diff);
-  }
-
-  // Store results to dst.
-  *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
-  dst_ptr += stride;
-
-  p1 = _mm_srli_si128(p1, 4);
-  *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
-  dst_ptr += stride;
-
-  p1 = _mm_srli_si128(p1, 4);
-  *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
-  dst_ptr += stride;
-
-  p1 = _mm_srli_si128(p1, 4);
-  *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
-}
-
 void vp9_short_idct4x4_add_sse2(int16_t *input, uint8_t *dest, int stride) {
   const __m128i zero = _mm_setzero_si128();
   const __m128i eight = _mm_set1_epi16(8);
diff --git a/vp9/decoder/vp9_idct_blk.c b/vp9/decoder/vp9_idct_blk.c
index c52963c49..0217919da 100644
--- a/vp9/decoder/vp9_idct_blk.c
+++ b/vp9/decoder/vp9_idct_blk.c
@@ -66,7 +66,7 @@ void vp9_idct_add_c(int16_t *input, uint8_t *dest, int stride, int eob) {
     vp9_short_idct4x4_add(input, dest, stride);
     vpx_memset(input, 0, 32);
   } else {
-    vp9_dc_only_idct_add(input[0], dest, dest, stride, stride);
+    vp9_short_idct4x4_1_add(input, dest, stride);
     ((int *)input)[0] = 0;
   }
 }
-- 
2.40.0