mirror of
https://github.com/Kelsidavis/WoWee.git
synced 2026-04-04 20:23:51 +00:00
bbb560f93c
queryTimeMs and queryCallCount on WMORenderer and M2Renderer were plain mutable doubles/uint32s written by getFloorHeight (dispatched on async threads from CameraController) and read by the main thread. This is undefined behavior per C++ — thread sanitizer would flag it. Changed to std::atomic with relaxed ordering (adequate for diagnostics) and updated QueryTimer to use atomic fetch_add/compare_exchange.
70 lines
2.3 KiB
C++
70 lines
2.3 KiB
C++
#pragma once
|
|
|
|
#include <vulkan/vulkan.h>
|
|
#include <glm/glm.hpp>
|
|
#include <atomic>
|
|
#include <chrono>
|
|
|
|
namespace wowee {
|
|
namespace rendering {
|
|
|
|
// Must match the PerFrame UBO layout in all shaders (std140 alignment)
|
|
struct GPUPerFrameData {
|
|
glm::mat4 view;
|
|
glm::mat4 projection;
|
|
glm::mat4 lightSpaceMatrix;
|
|
glm::vec4 lightDir; // xyz = direction, w = unused
|
|
glm::vec4 lightColor; // xyz = color, w = unused
|
|
glm::vec4 ambientColor; // xyz = color, w = unused
|
|
glm::vec4 viewPos; // xyz = camera pos, w = unused
|
|
glm::vec4 fogColor; // xyz = color, w = unused
|
|
glm::vec4 fogParams; // x = fogStart, y = fogEnd, z = time, w = unused
|
|
glm::vec4 shadowParams; // x = enabled(0/1), y = strength, z = unused, w = unused
|
|
};
|
|
|
|
// Push constants for the model matrix (most common case)
|
|
struct GPUPushConstants {
|
|
glm::mat4 model;
|
|
};
|
|
|
|
// Push constants for shadow rendering passes
|
|
struct ShadowPush {
|
|
glm::mat4 lightSpaceMatrix;
|
|
glm::mat4 model;
|
|
};
|
|
|
|
// Uniform buffer for shadow rendering parameters (matches shader std140 layout)
|
|
struct ShadowParamsUBO {
|
|
int32_t useBones;
|
|
int32_t useTexture;
|
|
int32_t alphaTest;
|
|
int32_t foliageSway;
|
|
float windTime;
|
|
float foliageMotionDamp;
|
|
};
|
|
|
|
// Timer utility for performance profiling queries.
|
|
// Uses atomics because floor-height queries are dispatched on async threads
|
|
// from CameraController while the main thread may read the counters.
|
|
struct QueryTimer {
|
|
std::atomic<double>* totalMs = nullptr;
|
|
std::atomic<uint32_t>* callCount = nullptr;
|
|
std::chrono::steady_clock::time_point start = std::chrono::steady_clock::now();
|
|
QueryTimer(std::atomic<double>* total, std::atomic<uint32_t>* calls)
|
|
: totalMs(total), callCount(calls) {}
|
|
~QueryTimer() {
|
|
if (callCount) {
|
|
callCount->fetch_add(1, std::memory_order_relaxed);
|
|
}
|
|
if (totalMs) {
|
|
auto end = std::chrono::steady_clock::now();
|
|
double ms = std::chrono::duration<double, std::milli>(end - start).count();
|
|
// Relaxed is fine for diagnostics — exact ordering doesn't matter.
|
|
double old = totalMs->load(std::memory_order_relaxed);
|
|
while (!totalMs->compare_exchange_weak(old, old + ms, std::memory_order_relaxed)) {}
|
|
}
|
|
}
|
|
};
|
|
|
|
} // namespace rendering
|
|
} // namespace wowee
|