Catcrafts/Crafter.Graphics
2
0
Fork 0
Code Issues 1 Pull requests Projects Releases Packages Wiki Activity Actions

perf(window): ring buffer for CRAFTER_TIMING frame times (#44) #76

Merged
catbot merged 1 commit from claude/issue-44 into master 2026-06-16 17:25:58 +02:00
Conversation 0 Commits 1 Files changed 2 +26 -17
2 changed files with 26 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files
Showing only changes of commit 41d99890c1 - Show all commits

perf(window): ring buffer for CRAFTER_TIMING frame times (#44)

Prev Next 
Replace the std::vector<nanoseconds> frameTimes plus
erase(begin()) with a fixed-size std::array ring buffer (head +
count). The previous code memmoved ~99 elements left every frame once
the window filled. LogTiming computes order-independent sum/avg/min/max,
so the ring's write order is irrelevant to the reported stats.

Behind CRAFTER_TIMING only; shipping builds are unaffected.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
catbot 2026-06-16 15:25:22 +00:00
commit 41d99890c1

32
implementations/Crafter.Graphics-Window.cpp
View file

@ -657,11 +657,12 @@ void Window::Update() {
#ifdef CRAFTER_TIMING #ifdef CRAFTER_TIMING
frameEnd = std::chrono::high_resolution_clock::now(); frameEnd = std::chrono::high_resolution_clock::now();
frameTimes.push_back(totalUpdate+totalRender); // Ring buffer: overwrite the oldest entry once full. Order doesn't matter
// to LogTiming, so we never need to shift elements.
// Keep only the last 100 frame times frameTimes[frameTimesHead] = totalUpdate+totalRender;
if (frameTimes.size() > 100) { frameTimesHead = (frameTimesHead + 1) % frameTimeCapacity;
frameTimes.erase(frameTimes.begin()); if (frameTimesCount < frameTimeCapacity) {
++frameTimesCount;
} }
#endif #endif
lastFrameBegin = startTime; lastFrameBegin = startTime;
@ -843,21 +844,22 @@ void Window::LogTiming() {
std::cout << std::format("Total: {}", duration_cast<std::chrono::milliseconds>(totalUpdate+totalRender)) << std::endl; std::cout << std::format("Total: {}", duration_cast<std::chrono::milliseconds>(totalUpdate+totalRender)) << std::endl;
std::cout << std::format("Vblank: {}", duration_cast<std::chrono::milliseconds>(vblank)) << std::endl; std::cout << std::format("Vblank: {}", duration_cast<std::chrono::milliseconds>(vblank)) << std::endl;
// Add 100-frame average and min-max timing info // Add 100-frame average and min-max timing info. Only the first
if (!frameTimes.empty()) { // frameTimesCount entries of the ring buffer hold valid samples.
if (frameTimesCount != 0) {
// Calculate average // Calculate average
std::chrono::nanoseconds sum(0); std::chrono::nanoseconds sum(0);
for (const auto& frameTime : frameTimes) { for (std::size_t i = 0; i < frameTimesCount; ++i) {
sum += frameTime; sum += frameTimes[i];
} }
auto average = sum / frameTimes.size(); auto average = sum / frameTimesCount;
// Find min and max // Find min and max
auto min = frameTimes.front(); auto min = frameTimes[0];
auto max = frameTimes.front(); auto max = frameTimes[0];
for (const auto& frameTime : frameTimes) { for (std::size_t i = 0; i < frameTimesCount; ++i) {
if (frameTime < min) min = frameTime; if (frameTimes[i] < min) min = frameTimes[i];
if (frameTime > max) max = frameTime; if (frameTimes[i] > max) max = frameTimes[i];
} }
std::cout << std::format("Last 100 Frame Times - Avg: {}, Min: {}, Max: {}", std::cout << std::format("Last 100 Frame Times - Avg: {}, Min: {}, Max: {}",

9
interfaces/Crafter.Graphics-Window.cppm
View file

@ -164,7 +164,14 @@ export namespace Crafter {
std::chrono::nanoseconds vblank; std::chrono::nanoseconds vblank;
std::chrono::nanoseconds totalFrame; std::chrono::nanoseconds totalFrame;
std::chrono::time_point<std::chrono::high_resolution_clock> frameEnd; std::chrono::time_point<std::chrono::high_resolution_clock> frameEnd;
std::vector<std::chrono::nanoseconds> frameTimes; // Fixed-size ring buffer of the most recent frame times. LogTiming does
// order-independent sum/avg/min/max, so head position is irrelevant to
// the reported stats; this avoids the per-frame memmove a vector::erase
// at the front would incur once full.
static constexpr std::size_t frameTimeCapacity = 100;
std::array<std::chrono::nanoseconds, frameTimeCapacity> frameTimes{};
std::size_t frameTimesHead = 0;
std::size_t frameTimesCount = 0;
void LogTiming(); void LogTiming();
#endif #endif