Change rtc::VideoSinkWants to have target and a max pixel count

webrtc/media/base/videoadapter.cc

 /*
  *  Copyright (c) 2010 The WebRTC 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 "webrtc/media/base/videoadapter.h"
 
 #include <algorithm>
 #include <cmath>
 #include <cstdlib>
 #include <limits>
 
 #include "webrtc/base/arraysize.h"
 #include "webrtc/base/checks.h"
 #include "webrtc/base/logging.h"
 #include "webrtc/base/optional.h"
 #include "webrtc/media/base/mediaconstants.h"
 #include "webrtc/media/base/videocommon.h"
 
 namespace {
 struct Fraction {
   int numerator;
   int denominator;
+
+  // Determines number of output pixels if both width and height of an input of
+  // |input_pixels| pixels is scaled with the fraction numerator / denominator.
+  int scale_pixel_count(int input_pixels) {
+    return (numerator * numerator * input_pixels) / (denominator * denominator);
+  }
 };
 
 // Round |value_to_round| to a multiple of |multiple|. Prefer rounding upwards,
 // but never more than |max_value|.
 int roundUp(int value_to_round, int multiple, int max_value) {
   const int rounded_value =
       (value_to_round + multiple - 1) / multiple * multiple;
   return rounded_value <= max_value ? rounded_value
                                     : (max_value / multiple * multiple);
 }
 
-// Generates a scale factor that makes |input_num_pixels| smaller or
-// larger than |target_num_pixels|, depending on the value of |step_up|.
-Fraction FindScale(int input_num_pixels, int target_num_pixels, bool step_up) {
+// Generates a scale factor that makes |input_pixels| close to |target_pixels|,
+// but no higher than |max_pixels|.
+Fraction FindScale(int input_pixels, int target_pixels, int max_pixels) {
   // This function only makes sense for a positive target.
-  RTC_DCHECK_GT(target_num_pixels, 0);
+  RTC_DCHECK_GT(target_pixels, 0);
+  RTC_DCHECK_GT(max_pixels, 0);
+  RTC_DCHECK_GE(max_pixels, target_pixels);
+
+  // Don't scale up original.
+  if (target_pixels >= input_pixels)
+    return Fraction{1, 1};
+
+  Fraction current_scale = Fraction{1, 1};
   Fraction best_scale = Fraction{1, 1};
-  Fraction last_scale = Fraction{1, 1};
-  const float target_scale =
-      sqrt(target_num_pixels / static_cast<float>(input_num_pixels));
-  while (best_scale.numerator > (target_scale * best_scale.denominator)) {
-    last_scale = best_scale;
-    if (best_scale.numerator % 3 == 0 && best_scale.denominator % 2 == 0) {
-      // Multiply by 2/3
-      best_scale.numerator /= 3;
-      best_scale.denominator /= 2;
+  // The minimum (absolute) difference between the number of output pixels and
+  // the target pixel count.
+  int min_pixel_diff = input_pixels;
+  if (input_pixels < max_pixels) {
+    // Start condition for 1/1 case, if it is less than max.
+    min_pixel_diff = std::abs(input_pixels - target_pixels);
+  }
+
+  // Alternately scale down by 2/3 and 3/4. This results in fractions which are
+  // effectively scalable. For instance, starting at 1280x720 will result in
+  // the series (3/4) => 960x540, (1/2) => 640x360, (3/8) => 480x270,
+  // (1/4) => 320x180, (3/16) => 240x125, (1/8) => 160x90.
+  while (current_scale.scale_pixel_count(input_pixels) > target_pixels) {
+    if (current_scale.numerator % 3 == 0 &&
+        current_scale.denominator % 2 == 0) {
+      // Multiply by 2/3.
+      current_scale.numerator /= 3;
+      current_scale.denominator /= 2;
     } else {
-      // Multiply by 3/4
-      best_scale.numerator *= 3;
-      best_scale.denominator *= 4;
+      // Multiply by 3/4.
+      current_scale.numerator *= 3;
+      current_scale.denominator *= 4;
+    }
+
+    int output_pixels = current_scale.scale_pixel_count(input_pixels);
+    if (output_pixels <= max_pixels) {
+      int diff = std::abs(target_pixels - output_pixels);
+      if (diff < min_pixel_diff) {
+        min_pixel_diff = diff;
+        best_scale = current_scale;
+      }
     }
   }
-  if (step_up)
-    return last_scale;
+
   return best_scale;
 }
 }  // namespace
 
 namespace cricket {
 
 VideoAdapter::VideoAdapter(int required_resolution_alignment)
     : frames_in_(0),
       frames_out_(0),
       frames_scaled_(0),
       adaption_changes_(0),
       previous_width_(0),
       previous_height_(0),
       required_resolution_alignment_(required_resolution_alignment),
-      resolution_request_max_pixel_count_(std::numeric_limits<int>::max()),
-      step_up_(false) {}
+      resolution_request_target_pixel_count_(std::numeric_limits<int>::max()),
+      resolution_request_max_pixel_count_(std::numeric_limits<int>::max()) {}
 
 VideoAdapter::VideoAdapter() : VideoAdapter(1) {}
 
 VideoAdapter::~VideoAdapter() {}
 
 bool VideoAdapter::KeepFrame(int64_t in_timestamp_ns) {
   rtc::CritScope cs(&critical_section_);
   if (!requested_format_ || requested_format_->interval == 0)
     return true;
 
@@ skipping 28 matching lines @@
                                         int* cropped_height,
                                         int* out_width,
                                         int* out_height) {
   rtc::CritScope cs(&critical_section_);
   ++frames_in_;
 
   // The max output pixel count is the minimum of the requests from
   // OnOutputFormatRequest and OnResolutionRequest.
   int max_pixel_count = resolution_request_max_pixel_count_;
   if (requested_format_) {
-    // TODO(kthelgason): remove the - |step_up_| hack when we change how
-    // resolution is requested from VideoSourceProxy.
-    // This is required because we must not scale above the requested
-    // format so we subtract one when scaling up.
     max_pixel_count = std::min(
-        max_pixel_count, requested_format_->width * requested_format_->height -
-                             static_cast<int>(step_up_));

[kthelgason, 2017/02/06 12:13:20]

good riddance!

[sprang_webrtc, 2017/02/06 13:18:07]

:)

+        max_pixel_count, requested_format_->width * requested_format_->height);
   }
+  int target_pixel_count =
+      std::min(resolution_request_target_pixel_count_, max_pixel_count);
 
   // Drop the input frame if necessary.
   if (max_pixel_count <= 0 || !KeepFrame(in_timestamp_ns)) {
     // Show VAdapt log every 90 frames dropped. (3 seconds)
     if ((frames_in_ - frames_out_) % 90 == 0) {
       // TODO(fbarchard): Reduce to LS_VERBOSE when adapter info is not needed
       // in default calls.
       LOG(LS_INFO) << "VAdapt Drop Frame: scaled " << frames_scaled_
                    << " / out " << frames_out_
                    << " / in " << frames_in_
@@ skipping 21 matching lines @@
       std::swap(requested_format_->width, requested_format_->height);
     }
     const float requested_aspect =
         requested_format_->width /
         static_cast<float>(requested_format_->height);
     *cropped_width =
         std::min(in_width, static_cast<int>(in_height * requested_aspect));
     *cropped_height =
         std::min(in_height, static_cast<int>(in_width / requested_aspect));
   }
-  const Fraction scale =
-      FindScale(*cropped_width * *cropped_height, max_pixel_count, step_up_);
+  const Fraction scale = FindScale((*cropped_width) * (*cropped_height),
+                                   target_pixel_count, max_pixel_count);
   // Adjust cropping slightly to get even integer output size and a perfect
   // scale factor. Make sure the resulting dimensions are aligned correctly
   // to be nice to hardware encoders.
   *cropped_width =
       roundUp(*cropped_width,
               scale.denominator * required_resolution_alignment_, in_width);
   *cropped_height =
       roundUp(*cropped_height,
               scale.denominator * required_resolution_alignment_, in_height);
   RTC_DCHECK_EQ(0, *cropped_width % scale.denominator);
@@ skipping 26 matching lines @@
 
   return true;
 }
 
 void VideoAdapter::OnOutputFormatRequest(const VideoFormat& format) {
   rtc::CritScope cs(&critical_section_);
   requested_format_ = rtc::Optional<VideoFormat>(format);
   next_frame_timestamp_ns_ = rtc::Optional<int64_t>();
 }
 
-void VideoAdapter::OnResolutionRequest(
-    rtc::Optional<int> max_pixel_count,
-    rtc::Optional<int> max_pixel_count_step_up) {
+void VideoAdapter::OnResolutionRequest(rtc::Optional<int> target_pixel_count,
+                                       rtc::Optional<int> max_pixel_count) {
   rtc::CritScope cs(&critical_section_);
-  resolution_request_max_pixel_count_ = max_pixel_count.value_or(
-      max_pixel_count_step_up.value_or(std::numeric_limits<int>::max()));
-  step_up_ = static_cast<bool>(max_pixel_count_step_up);
+  resolution_request_max_pixel_count_ =
+      max_pixel_count.value_or(std::numeric_limits<int>::max());
+  resolution_request_target_pixel_count_ =
+      target_pixel_count.value_or(resolution_request_target_pixel_count_);
 }
 
 }  // namespace cricket