bilinear scale use fast 2 pass unless image is beyond row buffer size.

BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/965020

git-svn-id: http://libyuv.googlecode.com/svn/trunk@484 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 483
+Version: 484
 License: BSD
 License File: LICENSE
 

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 483
+#define LIBYUV_VERSION 484
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/scale.cc

@@ -2799,7 +2799,7 @@ void ScalePlaneBilinear(int src_width, int src_height,
                         const uint8* src_ptr, uint8* dst_ptr) {
   assert(dst_width > 0);
   assert(dst_height > 0);
-  if (!IS_ALIGNED(src_width, 8) || (src_width > kMaxInputWidth)) {
+  if (src_width > kMaxInputWidth) {
     ScalePlaneBilinearSimple(src_width, src_height, dst_width, dst_height,
                              src_stride, dst_stride, src_ptr, dst_ptr);
 

unit_test/compare_test.cc

@@ -30,23 +30,6 @@ static uint32 ReferenceHashDjb2(const uint8* src, uint64 count, uint32 seed) {
   return hash;
 }
 
-TEST_F(libyuvTest, BenchmakDjb2_C) {
-  const int kMaxTest = benchmark_width_ * benchmark_height_;
-  align_buffer_16(src_a, kMaxTest)
-  for (int i = 0; i < kMaxTest; ++i) {
-    src_a[i] = i;
-  }
-  uint32 h2 = ReferenceHashDjb2(src_a, kMaxTest, 5381);
-  uint32 h1;
-  MaskCpuFlags(0);
-  for (int i = 0; i < benchmark_iterations_; ++i) {
-    h1 = HashDjb2(src_a, kMaxTest, 5381);
-  }
-  MaskCpuFlags(-1);
-  EXPECT_EQ(h1, h2);
-  free_aligned_buffer_16(src_a)
-}
-
 TEST_F(libyuvTest, BenchmakDjb2_OPT) {
   const int kMaxTest = benchmark_width_ * benchmark_height_;
   align_buffer_16(src_a, kMaxTest)
@@ -78,39 +61,6 @@ TEST_F(libyuvTest, BenchmakDjb2_Unaligned_OPT) {
   free_aligned_buffer_16(src_a)
 }
 
-TEST_F(libyuvTest, BenchmarkSumSquareError_C) {
-  const int kMaxWidth = 4096 * 3;
-  align_buffer_16(src_a, kMaxWidth)
-  align_buffer_16(src_b, kMaxWidth)
-  memset(src_a, 0, kMaxWidth);
-  memset(src_b, 0, kMaxWidth);
-
-  MaskCpuFlags(0);
-  memcpy(src_a, "test0123test4567", 16);
-  memcpy(src_b, "tick0123tock4567", 16);
-  uint64 h1 = ComputeSumSquareError(src_a, src_b, 16);
-  EXPECT_EQ(790u, h1);
-
-  for (int i = 0; i < kMaxWidth; ++i) {
-    src_a[i] = i;
-    src_b[i] = i;
-  }
-  memset(src_a, 0, kMaxWidth);
-  memset(src_b, 0, kMaxWidth);
-
-  int count = benchmark_iterations_ *
-      (benchmark_width_ * benchmark_height_ + kMaxWidth - 1) / kMaxWidth;
-  for (int i = 0; i < count; ++i) {
-    h1 = ComputeSumSquareError(src_a, src_b, kMaxWidth);
-  }
-
-  MaskCpuFlags(-1);
-  EXPECT_EQ(0, h1);
-
-  free_aligned_buffer_16(src_a)
-  free_aligned_buffer_16(src_b)
-}
-
 TEST_F(libyuvTest, BenchmarkSumSquareError_OPT) {
   const int kMaxWidth = 4096 * 3;
   align_buffer_16(src_a, kMaxWidth)
@@ -184,33 +134,6 @@ TEST_F(libyuvTest, SumSquareError) {
   free_aligned_buffer_16(src_b)
 }
 
-TEST_F(libyuvTest, BenchmarkPsnr_C) {
-  align_buffer_16(src_a, benchmark_width_ * benchmark_height_)
-  align_buffer_16(src_b, benchmark_width_ * benchmark_height_)
-  for (int i = 0; i < benchmark_width_ * benchmark_height_; ++i) {
-    src_a[i] = i;
-    src_b[i] = i;
-  }
-
-  MaskCpuFlags(0);
-
-  double c_time = get_time();
-  for (int i = 0; i < benchmark_iterations_; ++i)
-    CalcFramePsnr(src_a, benchmark_width_,
-                  src_b, benchmark_width_,
-                  benchmark_width_, benchmark_height_);
-
-  c_time = (get_time() - c_time) / benchmark_iterations_;
-  printf("BenchmarkPsnr_C - %8.2f us c\n", c_time * 1e6);
-
-  MaskCpuFlags(-1);
-
-  EXPECT_EQ(0, 0);
-
-  free_aligned_buffer_16(src_a)
-  free_aligned_buffer_16(src_b)
-}
-
 TEST_F(libyuvTest, BenchmarkPsnr_OPT) {
   align_buffer_16(src_a, benchmark_width_ * benchmark_height_)
   align_buffer_16(src_b, benchmark_width_ * benchmark_height_)
@@ -313,33 +236,6 @@ TEST_F(libyuvTest, Psnr) {
   free_aligned_buffer_16(src_b)
 }
 
-TEST_F(libyuvTest, BenchmarkSsim_C) {
-  align_buffer_16(src_a, benchmark_width_ * benchmark_height_)
-  align_buffer_16(src_b, benchmark_width_ * benchmark_height_)
-  for (int i = 0; i < benchmark_width_ * benchmark_height_; ++i) {
-    src_a[i] = i;
-    src_b[i] = i;
-  }
-
-  MaskCpuFlags(0);
-
-  double c_time = get_time();
-  for (int i = 0; i < benchmark_iterations_; ++i)
-    CalcFrameSsim(src_a, benchmark_width_,
-                  src_b, benchmark_width_,
-                  benchmark_width_, benchmark_height_);
-
-  c_time = (get_time() - c_time) / benchmark_iterations_;
-  printf("BenchmarkSsim_C - %8.2f us c\n", c_time * 1e6);
-
-  MaskCpuFlags(-1);
-
-  EXPECT_EQ(0, 0);  // Pass if we get this far.
-
-  free_aligned_buffer_16(src_a)
-  free_aligned_buffer_16(src_b)
-}
-
 TEST_F(libyuvTest, BenchmarkSsim_OPT) {
   align_buffer_16(src_a, benchmark_width_ * benchmark_height_)
   align_buffer_16(src_b, benchmark_width_ * benchmark_height_)