diff options
Diffstat (limited to 'drivers')
64 files changed, 2837 insertions, 952 deletions
diff --git a/drivers/SCsub b/drivers/SCsub index e0bfa138f5..03ad70649b 100644 --- a/drivers/SCsub +++ b/drivers/SCsub @@ -14,7 +14,8 @@ SConscript("windows/SCsub") # Sounds drivers SConscript("alsa/SCsub") -SConscript("coreaudio/SCsub") +if env["platform"] == "ios" or env["platform"] == "macos": + SConscript("coreaudio/SCsub") SConscript("pulseaudio/SCsub") if env["platform"] == "windows": SConscript("wasapi/SCsub") diff --git a/drivers/alsa/audio_driver_alsa.cpp b/drivers/alsa/audio_driver_alsa.cpp index 0c9635339b..6ee22b6eb5 100644 --- a/drivers/alsa/audio_driver_alsa.cpp +++ b/drivers/alsa/audio_driver_alsa.cpp @@ -80,7 +80,7 @@ Error AudioDriverALSA::init_output_device() { status = snd_pcm_open(&pcm_handle, "default", SND_PCM_STREAM_PLAYBACK, SND_PCM_NONBLOCK); } else { String device = output_device_name; - int pos = device.find(";"); + int pos = device.find_char(';'); if (pos != -1) { device = device.substr(0, pos); } diff --git a/drivers/coreaudio/SCsub b/drivers/coreaudio/SCsub index 69d667c57b..83ac27f4b6 100644 --- a/drivers/coreaudio/SCsub +++ b/drivers/coreaudio/SCsub @@ -4,4 +4,4 @@ from misc.utility.scons_hints import * Import("env") # Driver source files -env.add_source_files(env.drivers_sources, "*.cpp") +env.add_source_files(env.drivers_sources, "*.mm") diff --git a/drivers/coreaudio/audio_driver_coreaudio.h b/drivers/coreaudio/audio_driver_coreaudio.h index 67ff3f3efc..54e02514ec 100644 --- a/drivers/coreaudio/audio_driver_coreaudio.h +++ b/drivers/coreaudio/audio_driver_coreaudio.h @@ -38,6 +38,8 @@ #import <AudioUnit/AudioUnit.h> #ifdef MACOS_ENABLED #import <CoreAudio/AudioHardware.h> +#else +#import <AVFoundation/AVFoundation.h> #endif class AudioDriverCoreAudio : public AudioDriver { @@ -51,9 +53,11 @@ class AudioDriverCoreAudio : public AudioDriver { String input_device_name = "Default"; int mix_rate = 0; + int capture_mix_rate = 0; unsigned int channels = 2; unsigned int capture_channels = 2; unsigned int buffer_frames = 0; + unsigned int capture_buffer_frames = 0; Vector<int32_t> samples_in; Vector<int16_t> input_buf; @@ -89,11 +93,12 @@ class AudioDriverCoreAudio : public AudioDriver { public: virtual const char *get_name() const override { return "CoreAudio"; - }; + } virtual Error init() override; virtual void start() override; virtual int get_mix_rate() const override; + virtual int get_input_mix_rate() const override; virtual SpeakerMode get_speaker_mode() const override; virtual void lock() override; diff --git a/drivers/coreaudio/audio_driver_coreaudio.cpp b/drivers/coreaudio/audio_driver_coreaudio.mm index fd0adb1fd1..4e6ff2edb3 100644 --- a/drivers/coreaudio/audio_driver_coreaudio.cpp +++ b/drivers/coreaudio/audio_driver_coreaudio.mm @@ -1,5 +1,5 @@ /**************************************************************************/ -/* audio_driver_coreaudio.cpp */ +/* audio_driver_coreaudio.mm */ /**************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ @@ -121,7 +121,24 @@ Error AudioDriverCoreAudio::init() { break; } - mix_rate = _get_configured_mix_rate(); +#ifdef MACOS_ENABLED + AudioDeviceID device_id; + UInt32 dev_id_size = sizeof(AudioDeviceID); + + AudioObjectPropertyAddress property_dev_id = { kAudioHardwarePropertyDefaultOutputDevice, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMaster }; + result = AudioObjectGetPropertyData(kAudioObjectSystemObject, &property_dev_id, 0, nullptr, &dev_id_size, &device_id); + ERR_FAIL_COND_V(result != noErr, FAILED); + + double hw_mix_rate; + UInt32 hw_mix_rate_size = sizeof(hw_mix_rate); + + AudioObjectPropertyAddress property_sr = { kAudioDevicePropertyNominalSampleRate, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMain }; + result = AudioObjectGetPropertyData(device_id, &property_sr, 0, nullptr, &hw_mix_rate_size, &hw_mix_rate); + ERR_FAIL_COND_V(result != noErr, FAILED); +#else + double hw_mix_rate = [AVAudioSession sharedInstance].sampleRate; +#endif + mix_rate = hw_mix_rate; memset(&strdesc, 0, sizeof(strdesc)); strdesc.mFormatID = kAudioFormatLinearPCM; @@ -147,10 +164,10 @@ Error AudioDriverCoreAudio::init() { unsigned int buffer_size = buffer_frames * channels; samples_in.resize(buffer_size); - input_buf.resize(buffer_size); print_verbose("CoreAudio: detected " + itos(channels) + " channels"); - print_verbose("CoreAudio: audio buffer frames: " + itos(buffer_frames) + " calculated latency: " + itos(buffer_frames * 1000 / mix_rate) + "ms"); + print_verbose("CoreAudio: output sampling rate: " + itos(mix_rate) + " Hz"); + print_verbose("CoreAudio: output audio buffer frames: " + itos(buffer_frames) + " calculated latency: " + itos(buffer_frames * 1000 / mix_rate) + "ms"); AURenderCallbackStruct callback; memset(&callback, 0, sizeof(AURenderCallbackStruct)); @@ -250,7 +267,7 @@ OSStatus AudioDriverCoreAudio::input_callback(void *inRefCon, } void AudioDriverCoreAudio::start() { - if (!active) { + if (!active && audio_unit != nullptr) { OSStatus result = AudioOutputUnitStart(audio_unit); if (result != noErr) { ERR_PRINT("AudioOutputUnitStart failed, code: " + itos(result)); @@ -275,6 +292,10 @@ int AudioDriverCoreAudio::get_mix_rate() const { return mix_rate; } +int AudioDriverCoreAudio::get_input_mix_rate() const { + return capture_mix_rate; +} + AudioDriver::SpeakerMode AudioDriverCoreAudio::get_speaker_mode() const { return get_speaker_mode_by_total_channels(channels); } @@ -378,14 +399,14 @@ Error AudioDriverCoreAudio::init_input_device() { UInt32 size; #ifdef MACOS_ENABLED - AudioDeviceID deviceId; + AudioDeviceID device_id; size = sizeof(AudioDeviceID); AudioObjectPropertyAddress property = { kAudioHardwarePropertyDefaultInputDevice, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMain }; - result = AudioObjectGetPropertyData(kAudioObjectSystemObject, &property, 0, nullptr, &size, &deviceId); + result = AudioObjectGetPropertyData(kAudioObjectSystemObject, &property, 0, nullptr, &size, &device_id); ERR_FAIL_COND_V(result != noErr, FAILED); - result = AudioUnitSetProperty(input_unit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, 0, &deviceId, sizeof(AudioDeviceID)); + result = AudioUnitSetProperty(input_unit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, 0, &device_id, sizeof(AudioDeviceID)); ERR_FAIL_COND_V(result != noErr, FAILED); #endif @@ -410,13 +431,23 @@ Error AudioDriverCoreAudio::init_input_device() { break; } - mix_rate = _get_configured_mix_rate(); +#ifdef MACOS_ENABLED + double hw_mix_rate; + UInt32 hw_mix_rate_size = sizeof(hw_mix_rate); + + AudioObjectPropertyAddress property_sr = { kAudioDevicePropertyNominalSampleRate, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMain }; + result = AudioObjectGetPropertyData(device_id, &property_sr, 0, nullptr, &hw_mix_rate_size, &hw_mix_rate); + ERR_FAIL_COND_V(result != noErr, FAILED); +#else + double hw_mix_rate = [AVAudioSession sharedInstance].sampleRate; +#endif + capture_mix_rate = hw_mix_rate; memset(&strdesc, 0, sizeof(strdesc)); strdesc.mFormatID = kAudioFormatLinearPCM; strdesc.mFormatFlags = kLinearPCMFormatFlagIsSignedInteger | kLinearPCMFormatFlagIsPacked; strdesc.mChannelsPerFrame = capture_channels; - strdesc.mSampleRate = mix_rate; + strdesc.mSampleRate = capture_mix_rate; strdesc.mFramesPerPacket = 1; strdesc.mBitsPerChannel = 16; strdesc.mBytesPerFrame = strdesc.mBitsPerChannel * strdesc.mChannelsPerFrame / 8; @@ -425,6 +456,13 @@ Error AudioDriverCoreAudio::init_input_device() { result = AudioUnitSetProperty(input_unit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Output, kInputBus, &strdesc, sizeof(strdesc)); ERR_FAIL_COND_V(result != noErr, FAILED); + int latency = Engine::get_singleton()->get_audio_output_latency(); + // Sample rate is independent of channels (ref: https://stackoverflow.com/questions/11048825/audio-sample-frequency-rely-on-channels) + capture_buffer_frames = closest_power_of_2(latency * capture_mix_rate / 1000); + + unsigned int buffer_size = capture_buffer_frames * capture_channels; + input_buf.resize(buffer_size); + AURenderCallbackStruct callback; memset(&callback, 0, sizeof(AURenderCallbackStruct)); callback.inputProc = &AudioDriverCoreAudio::input_callback; @@ -435,6 +473,9 @@ Error AudioDriverCoreAudio::init_input_device() { result = AudioUnitInitialize(input_unit); ERR_FAIL_COND_V(result != noErr, FAILED); + print_verbose("CoreAudio: input sampling rate: " + itos(capture_mix_rate) + " Hz"); + print_verbose("CoreAudio: input audio buffer frames: " + itos(capture_buffer_frames) + " calculated latency: " + itos(capture_buffer_frames * 1000 / capture_mix_rate) + "ms"); + return OK; } @@ -477,7 +518,7 @@ void AudioDriverCoreAudio::finish_input_device() { } Error AudioDriverCoreAudio::input_start() { - input_buffer_init(buffer_frames); + input_buffer_init(capture_buffer_frames); OSStatus result = AudioOutputUnitStart(input_unit); if (result != noErr) { @@ -561,7 +602,7 @@ PackedStringArray AudioDriverCoreAudio::_get_device_list(bool input) { } void AudioDriverCoreAudio::_set_device(const String &output_device, bool input) { - AudioDeviceID deviceId; + AudioDeviceID device_id; bool found = false; if (output_device != "Default") { AudioObjectPropertyAddress prop; @@ -608,7 +649,7 @@ void AudioDriverCoreAudio::_set_device(const String &output_device, bool input) if (CFStringGetCString(cfname, buffer, maxSize, kCFStringEncodingUTF8)) { String name = String::utf8(buffer) + " (" + itos(audioDevices[i]) + ")"; if (name == output_device) { - deviceId = audioDevices[i]; + device_id = audioDevices[i]; found = true; } } @@ -626,14 +667,14 @@ void AudioDriverCoreAudio::_set_device(const String &output_device, bool input) UInt32 elem = input ? kAudioHardwarePropertyDefaultInputDevice : kAudioHardwarePropertyDefaultOutputDevice; AudioObjectPropertyAddress property = { elem, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMain }; - OSStatus result = AudioObjectGetPropertyData(kAudioObjectSystemObject, &property, 0, nullptr, &size, &deviceId); + OSStatus result = AudioObjectGetPropertyData(kAudioObjectSystemObject, &property, 0, nullptr, &size, &device_id); ERR_FAIL_COND(result != noErr); found = true; } if (found) { - OSStatus result = AudioUnitSetProperty(input ? input_unit : audio_unit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, 0, &deviceId, sizeof(AudioDeviceID)); + OSStatus result = AudioUnitSetProperty(input ? input_unit : audio_unit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, 0, &device_id, sizeof(AudioDeviceID)); ERR_FAIL_COND(result != noErr); if (input) { diff --git a/drivers/d3d12/rendering_device_driver_d3d12.cpp b/drivers/d3d12/rendering_device_driver_d3d12.cpp index 479afbba93..0b1b0651c8 100644 --- a/drivers/d3d12/rendering_device_driver_d3d12.cpp +++ b/drivers/d3d12/rendering_device_driver_d3d12.cpp @@ -2160,6 +2160,7 @@ RDD::FenceID RenderingDeviceDriverD3D12::fence_create() { Error RenderingDeviceDriverD3D12::fence_wait(FenceID p_fence) { FenceInfo *fence = (FenceInfo *)(p_fence.id); + fence->d3d_fence->SetEventOnCompletion(fence->fence_value, fence->event_handle); DWORD res = WaitForSingleObjectEx(fence->event_handle, INFINITE, FALSE); #ifdef PIX_ENABLED PIXNotifyWakeFromFenceSignal(fence->event_handle); @@ -2254,7 +2255,6 @@ Error RenderingDeviceDriverD3D12::command_queue_execute_and_present(CommandQueue FenceInfo *fence = (FenceInfo *)(p_cmd_fence.id); fence->fence_value++; command_queue->d3d_queue->Signal(fence->d3d_fence.Get(), fence->fence_value); - fence->d3d_fence->SetEventOnCompletion(fence->fence_value, fence->event_handle); } } @@ -2461,8 +2461,6 @@ Error RenderingDeviceDriverD3D12::swap_chain_resize(CommandQueueID p_cmd_queue, break; } - print_verbose("Using swap chain flags: " + itos(creation_flags) + ", sync interval: " + itos(sync_interval) + ", present flags: " + itos(present_flags)); - if (swap_chain->d3d_swap_chain != nullptr && creation_flags != swap_chain->creation_flags) { // The swap chain must be recreated if the creation flags are different. _swap_chain_release(swap_chain); @@ -6179,6 +6177,8 @@ uint64_t RenderingDeviceDriverD3D12::api_trait_get(ApiTrait p_trait) { return false; case API_TRAIT_USE_GENERAL_IN_COPY_QUEUES: return true; + case API_TRAIT_BUFFERS_REQUIRE_TRANSITIONS: + return !barrier_capabilities.enhanced_barriers_supported; default: return RenderingDeviceDriver::api_trait_get(p_trait); } @@ -6585,7 +6585,7 @@ static Error create_command_signature(ID3D12Device *device, D3D12_INDIRECT_ARGUM HRESULT res = device->CreateCommandSignature(&cs_desc, nullptr, IID_PPV_ARGS(r_cmd_sig->GetAddressOf())); ERR_FAIL_COND_V_MSG(!SUCCEEDED(res), ERR_CANT_CREATE, "CreateCommandSignature failed with error " + vformat("0x%08ux", (uint64_t)res) + "."); return OK; -}; +} Error RenderingDeviceDriverD3D12::_initialize_frames(uint32_t p_frame_count) { Error err; diff --git a/drivers/egl/egl_manager.cpp b/drivers/egl/egl_manager.cpp index 603dfadd4b..8ca0aa5943 100644 --- a/drivers/egl/egl_manager.cpp +++ b/drivers/egl/egl_manager.cpp @@ -414,6 +414,30 @@ EGLContext EGLManager::get_context(DisplayServer::WindowID p_window_id) { return display.egl_context; } +EGLDisplay EGLManager::get_display(DisplayServer::WindowID p_window_id) { + GLWindow &glwindow = windows[p_window_id]; + + if (!glwindow.initialized) { + return EGL_NO_CONTEXT; + } + + GLDisplay &display = displays[glwindow.gldisplay_id]; + + return display.egl_display; +} + +EGLConfig EGLManager::get_config(DisplayServer::WindowID p_window_id) { + GLWindow &glwindow = windows[p_window_id]; + + if (!glwindow.initialized) { + return nullptr; + } + + GLDisplay &display = displays[glwindow.gldisplay_id]; + + return display.egl_config; +} + Error EGLManager::initialize(void *p_native_display) { #if defined(GLAD_ENABLED) && !defined(EGL_STATIC) // Loading EGL with a new display gets us just the bare minimum API. We'll then diff --git a/drivers/egl/egl_manager.h b/drivers/egl/egl_manager.h index f1b3dc99b7..2e1ae6ec53 100644 --- a/drivers/egl/egl_manager.h +++ b/drivers/egl/egl_manager.h @@ -113,6 +113,8 @@ public: bool is_using_vsync() const; EGLContext get_context(DisplayServer::WindowID p_window_id); + EGLDisplay get_display(DisplayServer::WindowID p_window_id); + EGLConfig get_config(DisplayServer::WindowID p_window_id); Error initialize(void *p_native_display = nullptr); diff --git a/drivers/gles3/effects/feed_effects.cpp b/drivers/gles3/effects/feed_effects.cpp new file mode 100644 index 0000000000..8ca88da662 --- /dev/null +++ b/drivers/gles3/effects/feed_effects.cpp @@ -0,0 +1,128 @@ +/**************************************************************************/ +/* feed_effects.cpp */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ + +#ifdef GLES3_ENABLED + +#include "feed_effects.h" + +#ifdef ANDROID_ENABLED +#include <GLES3/gl3ext.h> +#endif + +#define GL_PROGRAM_POINT_SIZE 0x8642 + +using namespace GLES3; + +FeedEffects *FeedEffects::singleton = nullptr; + +FeedEffects *FeedEffects::get_singleton() { + return singleton; +} + +FeedEffects::FeedEffects() { + singleton = this; + + feed.shader.initialize(); + feed.shader_version = feed.shader.version_create(); + feed.shader.version_bind_shader(feed.shader_version, FeedShaderGLES3::MODE_DEFAULT); + + { // Screen Triangle. + glGenBuffers(1, &screen_triangle); + glBindBuffer(GL_ARRAY_BUFFER, screen_triangle); + + const float qv[6] = { + -1.0f, + -1.0f, + 3.0f, + -1.0f, + -1.0f, + 3.0f, + }; + + glBufferData(GL_ARRAY_BUFFER, sizeof(float) * 6, qv, GL_STATIC_DRAW); + glBindBuffer(GL_ARRAY_BUFFER, 0); //unbind + + glGenVertexArrays(1, &screen_triangle_array); + glBindVertexArray(screen_triangle_array); + glBindBuffer(GL_ARRAY_BUFFER, screen_triangle); + glVertexAttribPointer(RS::ARRAY_VERTEX, 2, GL_FLOAT, GL_FALSE, sizeof(float) * 2, nullptr); + glEnableVertexAttribArray(RS::ARRAY_VERTEX); + glBindVertexArray(0); + glBindBuffer(GL_ARRAY_BUFFER, 0); //unbind + } +} + +FeedEffects::~FeedEffects() { + singleton = nullptr; + glDeleteBuffers(1, &screen_triangle); + glDeleteVertexArrays(1, &screen_triangle_array); + feed.shader.version_free(feed.shader_version); +} + +Transform3D transform3D_from_mat4(const float *p_mat4) { + Transform3D res; + + res.basis.rows[0][0] = p_mat4[0]; + res.basis.rows[1][0] = p_mat4[1]; + res.basis.rows[2][0] = p_mat4[2]; + // p_mat4[3] = 0; + res.basis.rows[0][1] = p_mat4[4]; + res.basis.rows[1][1] = p_mat4[5]; + res.basis.rows[2][1] = p_mat4[6]; + // p_mat4[7] = 0; + res.basis.rows[0][2] = p_mat4[8]; + res.basis.rows[1][2] = p_mat4[9]; + res.basis.rows[2][2] = p_mat4[10]; + // p_mat4[11] = 0; + res.origin.x = p_mat4[12]; + res.origin.y = p_mat4[13]; + res.origin.z = p_mat4[14]; + // p_mat4[15] = 1; + + return res; +} + +void FeedEffects::draw() { + bool success = feed.shader.version_bind_shader(feed.shader_version, FeedShaderGLES3::MODE_DEFAULT, FeedShaderGLES3::USE_EXTERNAL_SAMPLER); + if (!success) { + OS::get_singleton()->print("Godot : FeedShaderGLES3 Could not bind version_bind_shader USE_EXTERNAL_SAMPLER"); + return; + } + + draw_screen_triangle(); +} + +void FeedEffects::draw_screen_triangle() { + glBindVertexArray(screen_triangle_array); + glDrawArrays(GL_TRIANGLES, 0, 3); + glBindVertexArray(0); +} + +#endif // GLES3_ENABLED diff --git a/drivers/gles3/effects/feed_effects.h b/drivers/gles3/effects/feed_effects.h new file mode 100644 index 0000000000..5856a3e04b --- /dev/null +++ b/drivers/gles3/effects/feed_effects.h @@ -0,0 +1,68 @@ +/**************************************************************************/ +/* feed_effects.h */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ + +#ifndef FEED_EFFECTS_GLES3_H +#define FEED_EFFECTS_GLES3_H + +#ifdef GLES3_ENABLED + +#include "drivers/gles3/shaders/feed.glsl.gen.h" + +namespace GLES3 { + +class FeedEffects { +private: + struct Feed { + FeedShaderGLES3 shader; + RID shader_version; + } feed; + + static FeedEffects *singleton; + + GLuint screen_triangle = 0; + GLuint screen_triangle_array = 0; + +public: + static FeedEffects *get_singleton(); + + FeedEffects(); + ~FeedEffects(); + + void draw(); + +private: + void draw_screen_triangle(); +}; + +} // namespace GLES3 + +#endif // GLES3_ENABLED + +#endif // FEED_EFFECTS_GLES3_H diff --git a/drivers/gles3/rasterizer_canvas_gles3.cpp b/drivers/gles3/rasterizer_canvas_gles3.cpp index 3c959f0143..5fd90744a4 100644 --- a/drivers/gles3/rasterizer_canvas_gles3.cpp +++ b/drivers/gles3/rasterizer_canvas_gles3.cpp @@ -687,6 +687,8 @@ void RasterizerCanvasGLES3::_render_items(RID p_to_render_target, int p_item_cou state.current_tex = RID(); + const uint64_t base_specialization = GLES3::Config::get_singleton()->float_texture_supported ? 0 : CanvasShaderGLES3::USE_RGBA_SHADOWS; + for (uint32_t i = 0; i <= state.current_batch_index; i++) { // Skipping when there is no instances. if (state.canvas_instance_batches[i].instance_count == 0) { @@ -705,10 +707,9 @@ void RasterizerCanvasGLES3::_render_items(RID p_to_render_target, int p_item_cou } GLES3::CanvasMaterialData *material_data = state.canvas_instance_batches[i].material_data; - CanvasShaderGLES3::ShaderVariant variant = state.canvas_instance_batches[i].shader_variant; - uint64_t specialization = 0; - specialization |= uint64_t(state.canvas_instance_batches[i].lights_disabled); - specialization |= uint64_t(!GLES3::Config::get_singleton()->float_texture_supported) << 1; + CanvasShaderGLES3::ShaderVariant variant = CanvasShaderGLES3::MODE_DEFAULT; + uint64_t specialization = state.canvas_instance_batches[i].specialization; + specialization |= base_specialization; RID shader_version = data.canvas_shader_default_version; if (material_data) { @@ -719,11 +720,14 @@ void RasterizerCanvasGLES3::_render_items(RID p_to_render_target, int p_item_cou } } - bool success = GLES3::MaterialStorage::get_singleton()->shaders.canvas_shader.version_bind_shader(shader_version, variant, specialization); + bool success = material_storage->shaders.canvas_shader.version_bind_shader(shader_version, variant, specialization); if (!success) { continue; } + // Bind per-batch uniforms. + material_storage->shaders.canvas_shader.version_set_uniform(CanvasShaderGLES3::BATCH_FLAGS, state.canvas_instance_batches[i].flags, shader_version, variant, specialization); + GLES3::CanvasShaderData::BlendMode blend_mode = state.canvas_instance_batches[i].blend_mode; Color blend_color = state.canvas_instance_batches[i].blend_color; @@ -810,6 +814,7 @@ void RasterizerCanvasGLES3::_render_items(RID p_to_render_target, int p_item_cou void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_render_target, const Transform2D &p_canvas_transform_inverse, Item *¤t_clip, GLES3::CanvasShaderData::BlendMode p_blend_mode, Light *p_lights, uint32_t &r_index, bool &r_batch_broken, bool &r_sdf_used, const Point2 &p_repeat_offset) { RenderingServer::CanvasItemTextureFilter texture_filter = p_item->texture_filter == RS::CANVAS_ITEM_TEXTURE_FILTER_DEFAULT ? state.default_filter : p_item->texture_filter; + const uint64_t specialization_command_mask = ~(CanvasShaderGLES3::USE_NINEPATCH | CanvasShaderGLES3::USE_PRIMITIVE | CanvasShaderGLES3::USE_ATTRIBUTES | CanvasShaderGLES3::USE_INSTANCING); if (texture_filter != state.canvas_instance_batches[state.current_batch_index].filter) { _new_batch(r_batch_broken); @@ -845,6 +850,7 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend uint32_t lights[4] = { 0, 0, 0, 0 }; uint16_t light_count = 0; + uint16_t shadow_mask = 0; { Light *light = p_lights; @@ -854,6 +860,10 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend uint32_t light_index = light->render_index_cache; lights[light_count >> 2] |= light_index << ((light_count & 3) * 8); + if (p_item->light_mask & light->item_shadow_mask) { + shadow_mask |= 1 << light_count; + } + light_count++; if (light_count == data.max_lights_per_item - 1) { @@ -863,14 +873,15 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend light = light->next_ptr; } - base_flags |= light_count << FLAGS_LIGHT_COUNT_SHIFT; + base_flags |= light_count << INSTANCE_FLAGS_LIGHT_COUNT_SHIFT; + base_flags |= shadow_mask << INSTANCE_FLAGS_SHADOW_MASKED_SHIFT; } bool lights_disabled = light_count == 0 && !state.using_directional_lights; - if (lights_disabled != state.canvas_instance_batches[state.current_batch_index].lights_disabled) { + if (lights_disabled != bool(state.canvas_instance_batches[state.current_batch_index].specialization & CanvasShaderGLES3::DISABLE_LIGHTING)) { _new_batch(r_batch_broken); - state.canvas_instance_batches[state.current_batch_index].lights_disabled = lights_disabled; + state.canvas_instance_batches[state.current_batch_index].specialization ^= CanvasShaderGLES3::DISABLE_LIGHTING; } const Item::Command *c = p_item->commands; @@ -904,7 +915,7 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend state.instance_data_array[r_index].lights[2] = lights[2]; state.instance_data_array[r_index].lights[3] = lights[3]; - state.instance_data_array[r_index].flags = base_flags | (state.instance_data_array[r_index == 0 ? 0 : r_index - 1].flags & (FLAGS_DEFAULT_NORMAL_MAP_USED | FLAGS_DEFAULT_SPECULAR_MAP_USED)); // Reset on each command for safety, keep canvastexture binding config. + state.instance_data_array[r_index].flags = base_flags; Color blend_color = base_color; GLES3::CanvasShaderData::BlendMode blend_mode = p_blend_mode; @@ -936,7 +947,8 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend state.canvas_instance_batches[state.current_batch_index].tex = rect->texture; state.canvas_instance_batches[state.current_batch_index].command_type = Item::Command::TYPE_RECT; state.canvas_instance_batches[state.current_batch_index].command = c; - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_QUAD; + state.canvas_instance_batches[state.current_batch_index].specialization &= specialization_command_mask; + state.canvas_instance_batches[state.current_batch_index].flags = 0; } _prepare_canvas_texture(rect->texture, state.canvas_instance_batches[state.current_batch_index].filter, state.canvas_instance_batches[state.current_batch_index].repeat, r_index, texpixel_size); @@ -959,20 +971,18 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend if (rect->flags & CANVAS_RECT_FLIP_H) { src_rect.size.x *= -1; - state.instance_data_array[r_index].flags |= FLAGS_FLIP_H; } if (rect->flags & CANVAS_RECT_FLIP_V) { src_rect.size.y *= -1; - state.instance_data_array[r_index].flags |= FLAGS_FLIP_V; } if (rect->flags & CANVAS_RECT_TRANSPOSE) { - state.instance_data_array[r_index].flags |= FLAGS_TRANSPOSE_RECT; + state.instance_data_array[r_index].flags |= INSTANCE_FLAGS_TRANSPOSE_RECT; } if (rect->flags & CANVAS_RECT_CLIP_UV) { - state.instance_data_array[r_index].flags |= FLAGS_CLIP_RECT_UV; + state.instance_data_array[r_index].flags |= INSTANCE_FLAGS_CLIP_RECT_UV; } } else { @@ -991,13 +1001,13 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend } if (rect->flags & CANVAS_RECT_MSDF) { - state.instance_data_array[r_index].flags |= FLAGS_USE_MSDF; + state.instance_data_array[r_index].flags |= INSTANCE_FLAGS_USE_MSDF; state.instance_data_array[r_index].msdf[0] = rect->px_range; // Pixel range. state.instance_data_array[r_index].msdf[1] = rect->outline; // Outline size. state.instance_data_array[r_index].msdf[2] = 0.f; // Reserved. state.instance_data_array[r_index].msdf[3] = 0.f; // Reserved. } else if (rect->flags & CANVAS_RECT_LCD) { - state.instance_data_array[r_index].flags |= FLAGS_USE_LCD; + state.instance_data_array[r_index].flags |= INSTANCE_FLAGS_USE_LCD; } state.instance_data_array[r_index].modulation[0] = rect->modulate.r * base_color.r; @@ -1026,7 +1036,9 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend state.canvas_instance_batches[state.current_batch_index].tex = np->texture; state.canvas_instance_batches[state.current_batch_index].command_type = Item::Command::TYPE_NINEPATCH; state.canvas_instance_batches[state.current_batch_index].command = c; - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_NINEPATCH; + state.canvas_instance_batches[state.current_batch_index].specialization &= specialization_command_mask; + state.canvas_instance_batches[state.current_batch_index].specialization |= CanvasShaderGLES3::USE_NINEPATCH; + state.canvas_instance_batches[state.current_batch_index].flags = 0; } _prepare_canvas_texture(np->texture, state.canvas_instance_batches[state.current_batch_index].filter, state.canvas_instance_batches[state.current_batch_index].repeat, r_index, texpixel_size); @@ -1064,11 +1076,11 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend state.instance_data_array[r_index].dst_rect[2] = dst_rect.size.width; state.instance_data_array[r_index].dst_rect[3] = dst_rect.size.height; - state.instance_data_array[r_index].flags |= int(np->axis_x) << FLAGS_NINEPATCH_H_MODE_SHIFT; - state.instance_data_array[r_index].flags |= int(np->axis_y) << FLAGS_NINEPATCH_V_MODE_SHIFT; + state.instance_data_array[r_index].flags |= int(np->axis_x) << INSTANCE_FLAGS_NINEPATCH_H_MODE_SHIFT; + state.instance_data_array[r_index].flags |= int(np->axis_y) << INSTANCE_FLAGS_NINEPATCH_V_MODE_SHIFT; if (np->draw_center) { - state.instance_data_array[r_index].flags |= FLAGS_NINEPACH_DRAW_CENTER; + state.instance_data_array[r_index].flags |= INSTANCE_FLAGS_NINEPACH_DRAW_CENTER; } state.instance_data_array[r_index].ninepatch_margins[0] = np->margin[SIDE_LEFT]; @@ -1092,7 +1104,9 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend state.canvas_instance_batches[state.current_batch_index].tex = polygon->texture; state.canvas_instance_batches[state.current_batch_index].command_type = Item::Command::TYPE_POLYGON; state.canvas_instance_batches[state.current_batch_index].command = c; - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_ATTRIBUTES; + state.canvas_instance_batches[state.current_batch_index].specialization &= specialization_command_mask; + state.canvas_instance_batches[state.current_batch_index].specialization |= CanvasShaderGLES3::USE_ATTRIBUTES; + state.canvas_instance_batches[state.current_batch_index].flags = 0; _prepare_canvas_texture(polygon->texture, state.canvas_instance_batches[state.current_batch_index].filter, state.canvas_instance_batches[state.current_batch_index].repeat, r_index, texpixel_size); @@ -1119,7 +1133,9 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend state.canvas_instance_batches[state.current_batch_index].primitive_points = primitive->point_count; state.canvas_instance_batches[state.current_batch_index].command_type = Item::Command::TYPE_PRIMITIVE; state.canvas_instance_batches[state.current_batch_index].command = c; - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_PRIMITIVE; + state.canvas_instance_batches[state.current_batch_index].specialization &= specialization_command_mask; + state.canvas_instance_batches[state.current_batch_index].specialization |= CanvasShaderGLES3::USE_PRIMITIVE; + state.canvas_instance_batches[state.current_batch_index].flags = 0; } _prepare_canvas_texture(state.canvas_instance_batches[state.current_batch_index].tex, state.canvas_instance_batches[state.current_batch_index].filter, state.canvas_instance_batches[state.current_batch_index].repeat, r_index, texpixel_size); @@ -1160,11 +1176,13 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend case Item::Command::TYPE_MESH: case Item::Command::TYPE_MULTIMESH: case Item::Command::TYPE_PARTICLES: { - // Mesh's can't be batched, so always create a new batch + // Meshes can't be batched, so always create a new batch. _new_batch(r_batch_broken); Color modulate(1, 1, 1, 1); - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_ATTRIBUTES; + state.canvas_instance_batches[state.current_batch_index].specialization &= specialization_command_mask; + state.canvas_instance_batches[state.current_batch_index].specialization |= CanvasShaderGLES3::USE_ATTRIBUTES; + state.canvas_instance_batches[state.current_batch_index].flags = 0; if (c->type == Item::Command::TYPE_MESH) { const Item::CommandMesh *m = static_cast<const Item::CommandMesh *>(c); state.canvas_instance_batches[state.current_batch_index].tex = m->texture; @@ -1174,13 +1192,13 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend } else if (c->type == Item::Command::TYPE_MULTIMESH) { const Item::CommandMultiMesh *mm = static_cast<const Item::CommandMultiMesh *>(c); state.canvas_instance_batches[state.current_batch_index].tex = mm->texture; - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_INSTANCED; + state.canvas_instance_batches[state.current_batch_index].specialization |= CanvasShaderGLES3::USE_INSTANCING; if (GLES3::MeshStorage::get_singleton()->multimesh_uses_colors(mm->multimesh)) { - state.instance_data_array[r_index].flags |= FLAGS_INSTANCING_HAS_COLORS; + state.canvas_instance_batches[state.current_batch_index].flags |= BATCH_FLAGS_INSTANCING_HAS_COLORS; } if (GLES3::MeshStorage::get_singleton()->multimesh_uses_custom_data(mm->multimesh)) { - state.instance_data_array[r_index].flags |= FLAGS_INSTANCING_HAS_CUSTOM_DATA; + state.canvas_instance_batches[state.current_batch_index].flags |= BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA; } } else if (c->type == Item::Command::TYPE_PARTICLES) { GLES3::ParticlesStorage *particles_storage = GLES3::ParticlesStorage::get_singleton(); @@ -1189,9 +1207,9 @@ void RasterizerCanvasGLES3::_record_item_commands(const Item *p_item, RID p_rend const Item::CommandParticles *pt = static_cast<const Item::CommandParticles *>(c); RID particles = pt->particles; state.canvas_instance_batches[state.current_batch_index].tex = pt->texture; - state.canvas_instance_batches[state.current_batch_index].shader_variant = CanvasShaderGLES3::MODE_INSTANCED; - state.instance_data_array[r_index].flags |= FLAGS_INSTANCING_HAS_COLORS; - state.instance_data_array[r_index].flags |= FLAGS_INSTANCING_HAS_CUSTOM_DATA; + state.canvas_instance_batches[state.current_batch_index].specialization |= CanvasShaderGLES3::USE_INSTANCING; + state.canvas_instance_batches[state.current_batch_index].flags |= BATCH_FLAGS_INSTANCING_HAS_COLORS; + state.canvas_instance_batches[state.current_batch_index].flags |= BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA; if (particles_storage->particles_has_collision(particles) && texture_storage->render_target_is_sdf_enabled(p_render_target)) { // Pass collision information. @@ -1654,28 +1672,39 @@ void RasterizerCanvasGLES3::light_update_shadow(RID p_rid, int p_shadow_index, c return; } - for (int i = 0; i < 4; i++) { - glViewport((state.shadow_texture_size / 4) * i, p_shadow_index * 2, (state.shadow_texture_size / 4), 2); + Projection projection; + { + real_t fov = 90; + real_t nearp = p_near; + real_t farp = p_far; + real_t aspect = 1.0; - Projection projection; - { - real_t fov = 90; - real_t nearp = p_near; - real_t farp = p_far; - real_t aspect = 1.0; + real_t ymax = nearp * Math::tan(Math::deg_to_rad(fov * 0.5)); + real_t ymin = -ymax; + real_t xmin = ymin * aspect; + real_t xmax = ymax * aspect; - real_t ymax = nearp * Math::tan(Math::deg_to_rad(fov * 0.5)); - real_t ymin = -ymax; - real_t xmin = ymin * aspect; - real_t xmax = ymax * aspect; + projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp); + } - projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp); - } + // Precomputed: + // Vector3 cam_target = Basis::from_euler(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); + // projection = projection * Projection(Transform3D().looking_at(cam_targets[i], Vector3(0, 0, -1)).affine_inverse()); + const Projection projections[4] = { + projection * Projection(Vector4(0, 0, -1, 0), Vector4(1, 0, 0, 0), Vector4(0, -1, 0, 0), Vector4(0, 0, 0, 1)), + + projection * Projection(Vector4(-1, 0, 0, 0), Vector4(0, 0, -1, 0), Vector4(0, -1, 0, 0), Vector4(0, 0, 0, 1)), + + projection * Projection(Vector4(0, 0, 1, 0), Vector4(-1, 0, 0, 0), Vector4(0, -1, 0, 0), Vector4(0, 0, 0, 1)), - Vector3 cam_target = Basis::from_euler(Vector3(0, 0, Math_TAU * ((i + 3) / 4.0))).xform(Vector3(0, 1, 0)); + projection * Projection(Vector4(1, 0, 0, 0), Vector4(0, 0, 1, 0), Vector4(0, -1, 0, 0), Vector4(0, 0, 0, 1)) + + }; + + for (int i = 0; i < 4; i++) { + glViewport((state.shadow_texture_size / 4) * i, p_shadow_index * 2, (state.shadow_texture_size / 4), 2); - projection = projection * Projection(Transform3D().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse()); - shadow_render.shader.version_set_uniform(CanvasOcclusionShaderGLES3::PROJECTION, projection, shadow_render.shader_version, variant); + shadow_render.shader.version_set_uniform(CanvasOcclusionShaderGLES3::PROJECTION, projections[i], shadow_render.shader_version, variant); static const Vector2 directions[4] = { Vector2(1, 0), Vector2(0, 1), Vector2(-1, 0), Vector2(0, -1) }; shadow_render.shader.version_set_uniform(CanvasOcclusionShaderGLES3::DIRECTION, directions[i].x, directions[i].y, shadow_render.shader_version, variant); @@ -2347,15 +2376,15 @@ void RasterizerCanvasGLES3::_prepare_canvas_texture(RID p_texture, RS::CanvasIte GLES3::Texture *normal_map = texture_storage->get_texture(ct->normal_map); if (ct->specular_color.a < 0.999) { - state.instance_data_array[r_index].flags |= FLAGS_DEFAULT_SPECULAR_MAP_USED; + state.canvas_instance_batches[state.current_batch_index].flags |= BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED; } else { - state.instance_data_array[r_index].flags &= ~FLAGS_DEFAULT_SPECULAR_MAP_USED; + state.canvas_instance_batches[state.current_batch_index].flags &= ~BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED; } if (normal_map) { - state.instance_data_array[r_index].flags |= FLAGS_DEFAULT_NORMAL_MAP_USED; + state.canvas_instance_batches[state.current_batch_index].flags |= BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED; } else { - state.instance_data_array[r_index].flags &= ~FLAGS_DEFAULT_NORMAL_MAP_USED; + state.canvas_instance_batches[state.current_batch_index].flags &= ~BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED; } state.instance_data_array[r_index].specular_shininess = uint32_t(CLAMP(ct->specular_color.a * 255.0, 0, 255)) << 24; diff --git a/drivers/gles3/rasterizer_canvas_gles3.h b/drivers/gles3/rasterizer_canvas_gles3.h index a82e2713e0..e099fd0cc0 100644 --- a/drivers/gles3/rasterizer_canvas_gles3.h +++ b/drivers/gles3/rasterizer_canvas_gles3.h @@ -54,29 +54,27 @@ class RasterizerCanvasGLES3 : public RendererCanvasRender { _FORCE_INLINE_ void _update_transform_to_mat4(const Transform3D &p_transform, float *p_mat4); enum { + INSTANCE_FLAGS_LIGHT_COUNT_SHIFT = 0, // 4 bits for light count. - FLAGS_INSTANCING_MASK = 0x7F, - FLAGS_INSTANCING_HAS_COLORS = (1 << 7), - FLAGS_INSTANCING_HAS_CUSTOM_DATA = (1 << 8), + INSTANCE_FLAGS_CLIP_RECT_UV = (1 << 4), + INSTANCE_FLAGS_TRANSPOSE_RECT = (1 << 5), + INSTANCE_FLAGS_USE_MSDF = (1 << 6), + INSTANCE_FLAGS_USE_LCD = (1 << 7), - FLAGS_CLIP_RECT_UV = (1 << 9), - FLAGS_TRANSPOSE_RECT = (1 << 10), + INSTANCE_FLAGS_NINEPACH_DRAW_CENTER = (1 << 8), + INSTANCE_FLAGS_NINEPATCH_H_MODE_SHIFT = 9, + INSTANCE_FLAGS_NINEPATCH_V_MODE_SHIFT = 11, - FLAGS_NINEPACH_DRAW_CENTER = (1 << 12), - - FLAGS_USE_SKELETON = (1 << 15), - FLAGS_NINEPATCH_H_MODE_SHIFT = 16, - FLAGS_NINEPATCH_V_MODE_SHIFT = 18, - FLAGS_LIGHT_COUNT_SHIFT = 20, - - FLAGS_DEFAULT_NORMAL_MAP_USED = (1 << 26), - FLAGS_DEFAULT_SPECULAR_MAP_USED = (1 << 27), + INSTANCE_FLAGS_SHADOW_MASKED_SHIFT = 13, // 16 bits. + }; - FLAGS_USE_MSDF = (1 << 28), - FLAGS_USE_LCD = (1 << 29), + enum { + BATCH_FLAGS_INSTANCING_MASK = 0x7F, + BATCH_FLAGS_INSTANCING_HAS_COLORS = (1 << 7), + BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA = (1 << 8), - FLAGS_FLIP_H = (1 << 30), - FLAGS_FLIP_V = (1 << 31), + BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED = (1 << 9), + BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED = (1 << 10), }; enum { @@ -273,14 +271,14 @@ public: RID material; GLES3::CanvasMaterialData *material_data = nullptr; - CanvasShaderGLES3::ShaderVariant shader_variant = CanvasShaderGLES3::MODE_QUAD; uint64_t vertex_input_mask = RS::ARRAY_FORMAT_VERTEX | RS::ARRAY_FORMAT_COLOR | RS::ARRAY_FORMAT_TEX_UV; + uint64_t specialization = 0; const Item::Command *command = nullptr; Item::Command::Type command_type = Item::Command::TYPE_ANIMATION_SLICE; // Can default to any type that doesn't form a batch. uint32_t primitive_points = 0; - bool lights_disabled = false; + uint32_t flags = 0; }; // DataBuffer contains our per-frame data. I.e. the resources that are updated each frame. diff --git a/drivers/gles3/rasterizer_gles3.cpp b/drivers/gles3/rasterizer_gles3.cpp index 6e508c6ebf..0fda42979f 100644 --- a/drivers/gles3/rasterizer_gles3.cpp +++ b/drivers/gles3/rasterizer_gles3.cpp @@ -35,6 +35,7 @@ #include "core/config/project_settings.h" #include "core/io/dir_access.h" +#include "core/io/image.h" #include "core/os/os.h" #include "storage/texture_storage.h" @@ -217,6 +218,7 @@ void RasterizerGLES3::finalize() { memdelete(glow); memdelete(cubemap_filter); memdelete(copy_effects); + memdelete(feed_effects); memdelete(light_storage); memdelete(particles_storage); memdelete(mesh_storage); @@ -365,6 +367,7 @@ RasterizerGLES3::RasterizerGLES3() { cubemap_filter = memnew(GLES3::CubemapFilter); glow = memnew(GLES3::Glow); post_effects = memnew(GLES3::PostEffects); + feed_effects = memnew(GLES3::FeedEffects); gi = memnew(GLES3::GI); fog = memnew(GLES3::Fog); canvas = memnew(RasterizerCanvasGLES3()); diff --git a/drivers/gles3/rasterizer_gles3.h b/drivers/gles3/rasterizer_gles3.h index 6765d8b4d5..abda2a5e06 100644 --- a/drivers/gles3/rasterizer_gles3.h +++ b/drivers/gles3/rasterizer_gles3.h @@ -35,6 +35,7 @@ #include "effects/copy_effects.h" #include "effects/cubemap_filter.h" +#include "effects/feed_effects.h" #include "effects/glow.h" #include "effects/post_effects.h" #include "environment/fog.h" @@ -78,6 +79,7 @@ protected: GLES3::CubemapFilter *cubemap_filter = nullptr; GLES3::Glow *glow = nullptr; GLES3::PostEffects *post_effects = nullptr; + GLES3::FeedEffects *feed_effects = nullptr; RasterizerCanvasGLES3 *canvas = nullptr; RasterizerSceneGLES3 *scene = nullptr; static RasterizerGLES3 *singleton; diff --git a/drivers/gles3/rasterizer_scene_gles3.cpp b/drivers/gles3/rasterizer_scene_gles3.cpp index a73f14c796..dfc7d02ac0 100644 --- a/drivers/gles3/rasterizer_scene_gles3.cpp +++ b/drivers/gles3/rasterizer_scene_gles3.cpp @@ -31,6 +31,7 @@ #include "rasterizer_scene_gles3.h" #include "drivers/gles3/effects/copy_effects.h" +#include "drivers/gles3/effects/feed_effects.h" #include "rasterizer_gles3.h" #include "storage/config.h" #include "storage/mesh_storage.h" @@ -39,6 +40,8 @@ #include "core/config/project_settings.h" #include "core/templates/sort_array.h" +#include "servers/camera/camera_feed.h" +#include "servers/camera_server.h" #include "servers/rendering/rendering_server_default.h" #include "servers/rendering/rendering_server_globals.h" @@ -235,7 +238,7 @@ void RasterizerSceneGLES3::_geometry_instance_add_surface_with_material(Geometry GLES3::SceneMaterialData *material_shadow = nullptr; void *surface_shadow = nullptr; - if (!p_material->shader_data->uses_particle_trails && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_prepass_alpha && !p_material->shader_data->uses_alpha_clip && !p_material->shader_data->uses_world_coordinates) { + if (!p_material->shader_data->uses_particle_trails && !p_material->shader_data->writes_modelview_or_projection && !p_material->shader_data->uses_vertex && !p_material->shader_data->uses_discard && !p_material->shader_data->uses_depth_prepass_alpha && !p_material->shader_data->uses_alpha_clip && !p_material->shader_data->uses_world_coordinates && !p_material->shader_data->wireframe) { flags |= GeometryInstanceSurface::FLAG_USES_SHARED_SHADOW_MATERIAL; material_shadow = static_cast<GLES3::SceneMaterialData *>(GLES3::MaterialStorage::get_singleton()->material_get_data(scene_globals.default_material, RS::SHADER_SPATIAL)); @@ -2259,6 +2262,7 @@ void RasterizerSceneGLES3::render_scene(const Ref<RenderSceneBuffers> &p_render_ render_data.inv_cam_transform = render_data.cam_transform.affine_inverse(); render_data.cam_projection = p_camera_data->main_projection; render_data.cam_orthogonal = p_camera_data->is_orthogonal; + render_data.cam_frustum = p_camera_data->is_frustum; render_data.camera_visible_layers = p_camera_data->visible_layers; render_data.main_cam_transform = p_camera_data->main_transform; @@ -2382,7 +2386,9 @@ void RasterizerSceneGLES3::render_scene(const Ref<RenderSceneBuffers> &p_render_ bool draw_sky = false; bool draw_sky_fog_only = false; bool keep_color = false; + bool draw_feed = false; float sky_energy_multiplier = 1.0; + int camera_feed_id = -1; if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { clear_color = Color(0, 0, 0, 1); //in overdraw mode, BG should always be black @@ -2427,6 +2433,8 @@ void RasterizerSceneGLES3::render_scene(const Ref<RenderSceneBuffers> &p_render_ keep_color = true; } break; case RS::ENV_BG_CAMERA_FEED: { + camera_feed_id = environment_get_camera_feed_id(render_data.environment); + draw_feed = true; } break; default: { } @@ -2538,7 +2546,7 @@ void RasterizerSceneGLES3::render_scene(const Ref<RenderSceneBuffers> &p_render_ glClear(GL_DEPTH_BUFFER_BIT); } - if (!keep_color) { + if (!keep_color && !draw_feed) { clear_color.a = render_data.transparent_bg ? 0.0f : 1.0f; glClearBufferfv(GL_COLOR, 0, clear_color.components); } else if (fbo != rt->fbo) { @@ -2578,6 +2586,29 @@ void RasterizerSceneGLES3::render_scene(const Ref<RenderSceneBuffers> &p_render_ spec_constant_base_flags |= SceneShaderGLES3::APPLY_TONEMAPPING; } } + + if (draw_feed && camera_feed_id > -1) { + RENDER_TIMESTAMP("Render Camera feed"); + + scene_state.enable_gl_depth_draw(false); + scene_state.enable_gl_depth_test(false); + scene_state.enable_gl_blend(false); + scene_state.set_gl_cull_mode(GLES3::SceneShaderData::CULL_BACK); + + Ref<CameraFeed> feed = CameraServer::get_singleton()->get_feed_by_id(camera_feed_id); + + if (feed.is_valid()) { + RID camera_YCBCR = feed->get_texture(CameraServer::FEED_YCBCR_IMAGE); + GLES3::TextureStorage::get_singleton()->texture_bind(camera_YCBCR, 0); + + GLES3::FeedEffects *feed_effects = GLES3::FeedEffects::get_singleton(); + feed_effects->draw(); + } + scene_state.enable_gl_depth_draw(true); + scene_state.enable_gl_depth_test(true); + scene_state.enable_gl_blend(true); + } + // Render Opaque Objects. RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, spec_constant_base_flags, use_wireframe); @@ -2592,14 +2623,18 @@ void RasterizerSceneGLES3::render_scene(const Ref<RenderSceneBuffers> &p_render_ scene_state.enable_gl_blend(false); scene_state.set_gl_cull_mode(GLES3::SceneShaderData::CULL_BACK); + Transform3D transform = render_data.cam_transform; Projection projection = render_data.cam_projection; if (is_reflection_probe) { Projection correction; correction.columns[1][1] = -1.0; projection = correction * render_data.cam_projection; + } else if (render_data.cam_frustum) { + // Sky is drawn upside down, the frustum offset doesn't know the image is upside down so needs a flip. + projection[2].y = -projection[2].y; } - _draw_sky(render_data.environment, projection, render_data.cam_transform, sky_energy_multiplier, render_data.luminance_multiplier, p_camera_data->view_count > 1, flip_y, apply_color_adjustments_in_post); + _draw_sky(render_data.environment, projection, transform, sky_energy_multiplier, render_data.luminance_multiplier, p_camera_data->view_count > 1, flip_y, apply_color_adjustments_in_post); } if (rt && (scene_state.used_screen_texture || scene_state.used_depth_texture)) { @@ -3122,7 +3157,7 @@ void RasterizerSceneGLES3::_render_list_template(RenderListParameters *p_params, } bool use_wireframe = false; - if (p_params->force_wireframe) { + if (p_params->force_wireframe || shader->wireframe) { GLuint wireframe_index_array_gl = mesh_storage->mesh_surface_get_index_buffer_wireframe(mesh_surface); if (wireframe_index_array_gl) { index_array_gl = wireframe_index_array_gl; diff --git a/drivers/gles3/rasterizer_scene_gles3.h b/drivers/gles3/rasterizer_scene_gles3.h index 06371b2b7f..4f088a0e7d 100644 --- a/drivers/gles3/rasterizer_scene_gles3.h +++ b/drivers/gles3/rasterizer_scene_gles3.h @@ -103,6 +103,7 @@ struct RenderDataGLES3 { Transform3D inv_cam_transform; Projection cam_projection; bool cam_orthogonal = false; + bool cam_frustum = false; uint32_t camera_visible_layers = 0xFFFFFFFF; // For billboards to cast correct shadows. diff --git a/drivers/gles3/shader_gles3.cpp b/drivers/gles3/shader_gles3.cpp index b73debf04a..83f74865f6 100644 --- a/drivers/gles3/shader_gles3.cpp +++ b/drivers/gles3/shader_gles3.cpp @@ -168,6 +168,10 @@ void ShaderGLES3::_build_variant_code(StringBuilder &builder, uint32_t p_variant builder.append("#version 300 es\n"); } + if (GLES3::Config::get_singleton()->polyfill_half2float) { + builder.append("#define USE_HALF2FLOAT\n"); + } + for (int i = 0; i < specialization_count; i++) { if (p_specialization & (uint64_t(1) << uint64_t(i))) { builder.append("#define " + String(specializations[i].name) + "\n"); diff --git a/drivers/gles3/shaders/SCsub b/drivers/gles3/shaders/SCsub index df2c515035..0207ba12b7 100644 --- a/drivers/gles3/shaders/SCsub +++ b/drivers/gles3/shaders/SCsub @@ -17,6 +17,7 @@ if "GLES3_GLSL" in env["BUILDERS"]: # as we have a few, not yet, converted files we name the ones we want to include: env.GLES3_GLSL("canvas.glsl") + env.GLES3_GLSL("feed.glsl") env.GLES3_GLSL("scene.glsl") env.GLES3_GLSL("sky.glsl") env.GLES3_GLSL("canvas_occlusion.glsl") diff --git a/drivers/gles3/shaders/canvas.glsl b/drivers/gles3/shaders/canvas.glsl index 5e7fb3b338..3857aa8841 100644 --- a/drivers/gles3/shaders/canvas.glsl +++ b/drivers/gles3/shaders/canvas.glsl @@ -1,17 +1,16 @@ /* clang-format off */ #[modes] -mode_quad = -mode_ninepatch = #define USE_NINEPATCH -mode_primitive = #define USE_PRIMITIVE -mode_attributes = #define USE_ATTRIBUTES -mode_instanced = #define USE_ATTRIBUTES \n#define USE_INSTANCING +mode_default = #[specializations] DISABLE_LIGHTING = true USE_RGBA_SHADOWS = false -SINGLE_INSTANCE = false +USE_NINEPATCH = false +USE_PRIMITIVE = false +USE_ATTRIBUTES = false +USE_INSTANCING = false #[vertex] @@ -111,6 +110,9 @@ layout(std140) uniform MaterialUniforms{ //ubo:4 }; #endif + +uniform mediump uint batch_flags; + /* clang-format on */ #include "canvas_uniforms_inc.glsl" @@ -180,13 +182,13 @@ void main() { vec2 uv = uv_attrib; #ifdef USE_INSTANCING - if (bool(read_draw_data_flags & FLAGS_INSTANCING_HAS_COLORS)) { + if (bool(batch_flags & BATCH_FLAGS_INSTANCING_HAS_COLORS)) { vec4 instance_color; instance_color.xy = unpackHalf2x16(uint(instance_color_custom_data.x)); instance_color.zw = unpackHalf2x16(uint(instance_color_custom_data.y)); color *= instance_color; } - if (bool(read_draw_data_flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + if (bool(batch_flags & BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { instance_custom.xy = unpackHalf2x16(instance_color_custom_data.z); instance_custom.zw = unpackHalf2x16(instance_color_custom_data.w); } @@ -219,7 +221,7 @@ void main() { else if (vertex_id == 5) vertex_base = vec2(1.0, 1.0); - vec2 uv = read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw) * ((read_draw_data_flags & FLAGS_TRANSPOSE_RECT) != uint(0) ? vertex_base.yx : vertex_base.xy); + vec2 uv = read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw) * ((read_draw_data_flags & INSTANCE_FLAGS_TRANSPOSE_RECT) != uint(0) ? vertex_base.yx : vertex_base.xy); vec4 color = read_draw_data_modulation; vec2 vertex = read_draw_data_dst_rect.xy + abs(read_draw_data_dst_rect.zw) * mix(vertex_base, vec2(1.0, 1.0) - vertex_base, lessThan(read_draw_data_src_rect.zw, vec2(0.0, 0.0))); @@ -337,6 +339,8 @@ uniform sampler2D specular_texture; //texunit:-7 uniform sampler2D color_texture; //texunit:0 +uniform mediump uint batch_flags; + layout(location = 0) out vec4 frag_color; /* clang-format off */ @@ -520,7 +524,7 @@ float map_ninepatch_axis(float pixel, float draw_size, float tex_pixel_size, flo } else if (pixel >= draw_size - margin_end) { return (tex_size - (draw_size - pixel)) * tex_pixel_size; } else { - if (!bool(read_draw_data_flags & FLAGS_NINEPACH_DRAW_CENTER)) { + if (!bool(read_draw_data_flags & INSTANCE_FLAGS_NINEPATCH_DRAW_CENTER)) { draw_center--; } @@ -568,8 +572,8 @@ void main() { int draw_center = 2; uv = vec2( - map_ninepatch_axis(pixel_size_interp.x, abs(read_draw_data_dst_rect_z), read_draw_data_color_texture_pixel_size.x, read_draw_data_ninepatch_margins.x, read_draw_data_ninepatch_margins.z, int(read_draw_data_flags >> FLAGS_NINEPATCH_H_MODE_SHIFT) & 0x3, draw_center), - map_ninepatch_axis(pixel_size_interp.y, abs(read_draw_data_dst_rect_w), read_draw_data_color_texture_pixel_size.y, read_draw_data_ninepatch_margins.y, read_draw_data_ninepatch_margins.w, int(read_draw_data_flags >> FLAGS_NINEPATCH_V_MODE_SHIFT) & 0x3, draw_center)); + map_ninepatch_axis(pixel_size_interp.x, abs(read_draw_data_dst_rect_z), read_draw_data_color_texture_pixel_size.x, read_draw_data_ninepatch_margins.x, read_draw_data_ninepatch_margins.z, int(read_draw_data_flags >> INSTANCE_FLAGS_NINEPATCH_H_MODE_SHIFT) & 0x3, draw_center), + map_ninepatch_axis(pixel_size_interp.y, abs(read_draw_data_dst_rect_w), read_draw_data_color_texture_pixel_size.y, read_draw_data_ninepatch_margins.y, read_draw_data_ninepatch_margins.w, int(read_draw_data_flags >> INSTANCE_FLAGS_NINEPATCH_V_MODE_SHIFT) & 0x3, draw_center)); if (draw_center == 0) { color.a = 0.0; @@ -578,7 +582,7 @@ void main() { uv = uv * read_draw_data_src_rect.zw + read_draw_data_src_rect.xy; //apply region if needed #endif - if (bool(read_draw_data_flags & FLAGS_CLIP_RECT_UV)) { + if (bool(read_draw_data_flags & INSTANCE_FLAGS_CLIP_RECT_UV)) { vec2 half_texpixel = read_draw_data_color_texture_pixel_size * 0.5; uv = clamp(uv, read_draw_data_src_rect.xy + half_texpixel, read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw) - half_texpixel); } @@ -586,7 +590,7 @@ void main() { #endif #ifndef USE_PRIMITIVE - if (bool(read_draw_data_flags & FLAGS_USE_MSDF)) { + if (bool(read_draw_data_flags & INSTANCE_FLAGS_USE_MSDF)) { float px_range = read_draw_data_ninepatch_margins.x; float outline_thickness = read_draw_data_ninepatch_margins.y; @@ -604,7 +608,7 @@ void main() { float a = clamp(d * px_size + 0.5, 0.0, 1.0); color.a = a * color.a; } - } else if (bool(read_draw_data_flags & FLAGS_USE_LCD)) { + } else if (bool(read_draw_data_flags & INSTANCE_FLAGS_USE_LCD)) { vec4 lcd_sample = texture(color_texture, uv); if (lcd_sample.a == 1.0) { color.rgb = lcd_sample.rgb * color.a; @@ -618,7 +622,7 @@ void main() { color *= texture(color_texture, uv); } - uint light_count = (read_draw_data_flags >> uint(FLAGS_LIGHT_COUNT_SHIFT)) & uint(0xF); //max 16 lights + uint light_count = read_draw_data_flags & uint(0xF); // Max 16 lights. bool using_light = light_count > 0u || directional_light_count > 0u; vec3 normal; @@ -629,17 +633,16 @@ void main() { bool normal_used = false; #endif - if (normal_used || (using_light && bool(read_draw_data_flags & FLAGS_DEFAULT_NORMAL_MAP_USED))) { + if (normal_used || (using_light && bool(batch_flags & BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED))) { normal.xy = texture(normal_texture, uv).xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); - if (bool(read_draw_data_flags & FLAGS_TRANSPOSE_RECT)) { + +#if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE) + if (bool(read_draw_data_flags & INSTANCE_FLAGS_TRANSPOSE_RECT)) { normal.xy = normal.yx; } - if (bool(read_draw_data_flags & FLAGS_FLIP_H)) { - normal.x = -normal.x; - } - if (bool(read_draw_data_flags & FLAGS_FLIP_V)) { - normal.y = -normal.y; - } + normal.xy *= sign(read_draw_data_src_rect.zw); +#endif + normal.z = sqrt(max(0.0, 1.0 - dot(normal.xy, normal.xy))); normal_used = true; } else { @@ -655,7 +658,7 @@ void main() { bool specular_shininess_used = false; #endif - if (specular_shininess_used || (using_light && normal_used && bool(read_draw_data_flags & FLAGS_DEFAULT_SPECULAR_MAP_USED))) { + if (specular_shininess_used || (using_light && normal_used && bool(batch_flags & BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED))) { specular_shininess = texture(specular_texture, uv); specular_shininess *= godot_unpackUnorm4x8(read_draw_data_specular_shininess); specular_shininess_used = true; @@ -728,7 +731,7 @@ void main() { } #endif - if (bool(light_array[light_base].flags & LIGHT_FLAGS_HAS_SHADOW)) { + if (bool(light_array[light_base].flags & LIGHT_FLAGS_HAS_SHADOW) && bool(read_draw_data_flags & uint(INSTANCE_FLAGS_SHADOW_MASKED << i))) { vec2 shadow_pos = (vec4(shadow_vertex, 0.0, 1.0) * mat4(light_array[light_base].shadow_matrix[0], light_array[light_base].shadow_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations. vec4 shadow_uv = vec4(shadow_pos.x, light_array[light_base].shadow_y_ofs, shadow_pos.y * light_array[light_base].shadow_zfar_inv, 1.0); @@ -803,7 +806,7 @@ void main() { } #endif - if (bool(light_array[light_base].flags & LIGHT_FLAGS_HAS_SHADOW)) { + if (bool(light_array[light_base].flags & LIGHT_FLAGS_HAS_SHADOW) && bool(read_draw_data_flags & uint(INSTANCE_FLAGS_SHADOW_MASKED << i))) { vec2 shadow_pos = (vec4(shadow_vertex, 0.0, 1.0) * mat4(light_array[light_base].shadow_matrix[0], light_array[light_base].shadow_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations. vec2 pos_norm = normalize(shadow_pos); diff --git a/drivers/gles3/shaders/canvas_uniforms_inc.glsl b/drivers/gles3/shaders/canvas_uniforms_inc.glsl index f6ad2b730a..94cbbdd8a7 100644 --- a/drivers/gles3/shaders/canvas_uniforms_inc.glsl +++ b/drivers/gles3/shaders/canvas_uniforms_inc.glsl @@ -1,33 +1,32 @@ - #define MAX_LIGHTS_PER_ITEM uint(16) #define M_PI 3.14159265359 #define SDF_MAX_LENGTH 16384.0 -//1 means enabled, 2+ means trails in use -#define FLAGS_INSTANCING_MASK uint(0x7F) -#define FLAGS_INSTANCING_HAS_COLORS uint(1 << 7) -#define FLAGS_INSTANCING_HAS_CUSTOM_DATA uint(1 << 8) - -#define FLAGS_CLIP_RECT_UV uint(1 << 9) -#define FLAGS_TRANSPOSE_RECT uint(1 << 10) -// (1 << 11) is for FLAGS_CONVERT_ATTRIBUTES_TO_LINEAR in RD backends, unused here. -#define FLAGS_NINEPACH_DRAW_CENTER uint(1 << 12) +#define INSTANCE_FLAGS_LIGHT_COUNT_SHIFT 0 // 4 bits. -#define FLAGS_NINEPATCH_H_MODE_SHIFT 16 -#define FLAGS_NINEPATCH_V_MODE_SHIFT 18 +#define INSTANCE_FLAGS_CLIP_RECT_UV uint(1 << 4) +#define INSTANCE_FLAGS_TRANSPOSE_RECT uint(1 << 5) +#define INSTANCE_FLAGS_USE_MSDF uint(1 << 6) +#define INSTANCE_FLAGS_USE_LCD uint(1 << 7) -#define FLAGS_LIGHT_COUNT_SHIFT 20 +#define INSTANCE_FLAGS_NINEPATCH_DRAW_CENTER uint(1 << 8) +#define INSTANCE_FLAGS_NINEPATCH_H_MODE_SHIFT 9 +#define INSTANCE_FLAGS_NINEPATCH_V_MODE_SHIFT 11 -#define FLAGS_DEFAULT_NORMAL_MAP_USED uint(1 << 26) -#define FLAGS_DEFAULT_SPECULAR_MAP_USED uint(1 << 27) +#define INSTANCE_FLAGS_SHADOW_MASKED_SHIFT 13u // 16 bits. +#define INSTANCE_FLAGS_SHADOW_MASKED uint(1 << INSTANCE_FLAGS_SHADOW_MASKED_SHIFT) -#define FLAGS_USE_MSDF uint(1 << 28) -#define FLAGS_USE_LCD uint(1 << 29) +// 1 means enabled, 2+ means trails in use +#define BATCH_FLAGS_INSTANCING_MASK uint(0x7F) +#define BATCH_FLAGS_INSTANCING_HAS_COLORS_SHIFT 7 +#define BATCH_FLAGS_INSTANCING_HAS_COLORS uint(1 << BATCH_FLAGS_INSTANCING_HAS_COLORS_SHIFT) +#define BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA_SHIFT 8 +#define BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA uint(1 << BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA_SHIFT) -#define FLAGS_FLIP_H uint(1 << 30) -#define FLAGS_FLIP_V uint(1 << 31) +#define BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED uint(1 << 9) +#define BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED uint(1 << 10) layout(std140) uniform GlobalShaderUniformData { //ubo:1 vec4 global_shader_uniforms[MAX_GLOBAL_SHADER_UNIFORMS]; diff --git a/drivers/gles3/shaders/feed.glsl b/drivers/gles3/shaders/feed.glsl new file mode 100644 index 0000000000..9d89fc699d --- /dev/null +++ b/drivers/gles3/shaders/feed.glsl @@ -0,0 +1,39 @@ +/* clang-format off */ +#[modes] + +mode_default = + +#[specializations] + +USE_EXTERNAL_SAMPLER = false + +#[vertex] + +layout(location = 0) in vec2 vertex_attrib; + +out vec2 uv_interp; + + +void main() { + uv_interp = vertex_attrib * 0.5 + 0.5; + gl_Position = vec4(vertex_attrib, 1.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +layout(location = 0) out vec4 frag_color; +in vec2 uv_interp; + +/* clang-format on */ +#ifdef USE_EXTERNAL_SAMPLER +uniform samplerExternalOES sourceFeed; // texunit:0 +#else +uniform sampler2D sourceFeed; // texunit:0 +#endif + +void main() { + vec4 color = texture(sourceFeed, uv_interp); + + frag_color = color; +} diff --git a/drivers/gles3/shaders/scene.glsl b/drivers/gles3/shaders/scene.glsl index fcfbeddb9e..2022c8ee43 100644 --- a/drivers/gles3/shaders/scene.glsl +++ b/drivers/gles3/shaders/scene.glsl @@ -1259,7 +1259,7 @@ float SchlickFresnel(float u) { return m2 * m2 * m; // pow(m,5) } -void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, bool is_directional, float attenuation, vec3 f0, float roughness, float metallic, float specular_amount, vec3 albedo, inout float alpha, +void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, bool is_directional, float attenuation, vec3 f0, float roughness, float metallic, float specular_amount, vec3 albedo, inout float alpha, vec2 screen_uv, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, #endif @@ -1423,7 +1423,7 @@ float get_omni_spot_attenuation(float distance, float inv_range, float decay) { } #if !defined(DISABLE_LIGHT_OMNI) || defined(ADDITIVE_OMNI) -void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f0, float roughness, float metallic, float shadow, vec3 albedo, inout float alpha, +void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f0, float roughness, float metallic, float shadow, vec3 albedo, inout float alpha, vec2 screen_uv, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, #endif @@ -1450,7 +1450,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f omni_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, false, omni_attenuation, f0, roughness, metallic, omni_lights[idx].specular_amount, albedo, alpha, + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, false, omni_attenuation, f0, roughness, metallic, omni_lights[idx].specular_amount, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1469,7 +1469,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f #endif // !DISABLE_LIGHT_OMNI #if !defined(DISABLE_LIGHT_SPOT) || defined(ADDITIVE_SPOT) -void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f0, float roughness, float metallic, float shadow, vec3 albedo, inout float alpha, +void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f0, float roughness, float metallic, float shadow, vec3 albedo, inout float alpha, vec2 screen_uv, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, #endif @@ -1506,7 +1506,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 f spot_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, false, spot_attenuation, f0, roughness, metallic, spot_lights[idx].specular_amount, albedo, alpha, + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, false, spot_attenuation, f0, roughness, metallic, spot_lights[idx].specular_amount, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -2114,7 +2114,7 @@ void main() { continue; } #endif - light_compute(normal, normalize(directional_lights[i].direction), normalize(view), directional_lights[i].size, directional_lights[i].color * directional_lights[i].energy, true, 1.0, f0, roughness, metallic, 1.0, albedo, alpha, + light_compute(normal, normalize(directional_lights[i].direction), normalize(view), directional_lights[i].size, directional_lights[i].color * directional_lights[i].energy, true, 1.0, f0, roughness, metallic, 1.0, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -2143,7 +2143,7 @@ void main() { continue; } #endif - light_process_omni(omni_light_indices[i], vertex, view, normal, f0, roughness, metallic, 1.0, albedo, alpha, + light_process_omni(omni_light_indices[i], vertex, view, normal, f0, roughness, metallic, 1.0, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -2171,7 +2171,7 @@ void main() { continue; } #endif - light_process_spot(spot_light_indices[i], vertex, view, normal, f0, roughness, metallic, 1.0, albedo, alpha, + light_process_spot(spot_light_indices[i], vertex, view, normal, f0, roughness, metallic, 1.0, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -2389,7 +2389,7 @@ void main() { #endif // SHADOWS_DISABLED #ifndef USE_VERTEX_LIGHTING - light_compute(normal, normalize(directional_lights[directional_shadow_index].direction), normalize(view), directional_lights[directional_shadow_index].size, directional_lights[directional_shadow_index].color * directional_lights[directional_shadow_index].energy, true, directional_shadow, f0, roughness, metallic, 1.0, albedo, alpha, + light_compute(normal, normalize(directional_lights[directional_shadow_index].direction), normalize(view), directional_lights[directional_shadow_index].size, directional_lights[directional_shadow_index].color * directional_lights[directional_shadow_index].energy, true, directional_shadow, f0, roughness, metallic, 1.0, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -2421,7 +2421,7 @@ void main() { #endif // SHADOWS_DISABLED #ifndef USE_VERTEX_LIGHTING - light_process_omni(omni_light_index, vertex, view, normal, f0, roughness, metallic, omni_shadow, albedo, alpha, + light_process_omni(omni_light_index, vertex, view, normal, f0, roughness, metallic, omni_shadow, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -2451,7 +2451,7 @@ void main() { #endif // SHADOWS_DISABLED #ifndef USE_VERTEX_LIGHTING - light_process_spot(spot_light_index, vertex, view, normal, f0, roughness, metallic, spot_shadow, albedo, alpha, + light_process_spot(spot_light_index, vertex, view, normal, f0, roughness, metallic, spot_shadow, albedo, alpha, screen_uv, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif diff --git a/drivers/gles3/shaders/sky.glsl b/drivers/gles3/shaders/sky.glsl index 186b630bc8..043023aee0 100644 --- a/drivers/gles3/shaders/sky.glsl +++ b/drivers/gles3/shaders/sky.glsl @@ -2,17 +2,15 @@ #[modes] mode_background = -mode_half_res = #define USE_HALF_RES_PASS -mode_quarter_res = #define USE_QUARTER_RES_PASS mode_cubemap = #define USE_CUBEMAP_PASS -mode_cubemap_half_res = #define USE_CUBEMAP_PASS \n#define USE_HALF_RES_PASS -mode_cubemap_quarter_res = #define USE_CUBEMAP_PASS \n#define USE_QUARTER_RES_PASS #[specializations] USE_MULTIVIEW = false USE_INVERTED_Y = true APPLY_TONEMAPPING = true +USE_QUARTER_RES_PASS = false +USE_HALF_RES_PASS = false #[vertex] diff --git a/drivers/gles3/shaders/stdlib_inc.glsl b/drivers/gles3/shaders/stdlib_inc.glsl index f88c218506..004db9e5d0 100644 --- a/drivers/gles3/shaders/stdlib_inc.glsl +++ b/drivers/gles3/shaders/stdlib_inc.glsl @@ -1,25 +1,28 @@ - // Compatibility renames. These are exposed with the "godot_" prefix // to work around two distinct Adreno bugs: // 1. Some Adreno devices expose ES310 functions in ES300 shaders. // Internally, we must use the "godot_" prefix, but user shaders // will be mapped automatically. // 2. Adreno 3XX devices have poor implementations of the other packing -// functions, so we just use our own everywhere to keep it simple. +// functions, so we just use our own there to keep it simple. +#ifdef USE_HALF2FLOAT // Floating point pack/unpack functions are part of the GLSL ES 300 specification used by web and mobile. +// It appears to be safe to expose these on mobile, but when running through ANGLE this appears to break. uint float2half(uint f) { - uint b = f + uint(0x00001000); - uint e = (b & uint(0x7F800000)) >> 23; - uint m = b & uint(0x007FFFFF); - return (b & uint(0x80000000)) >> uint(16) | uint(e > uint(112)) * ((((e - uint(112)) << uint(10)) & uint(0x7C00)) | m >> uint(13)) | (uint(e < uint(113)) & uint(e > uint(101))) * ((((uint(0x007FF000) + m) >> (uint(125) - e)) + uint(1)) >> uint(1)) | uint(e > uint(143)) * uint(0x7FFF); + uint e = f & uint(0x7f800000); + if (e <= uint(0x38000000)) { + return uint(0); + } else { + return ((f >> uint(16)) & uint(0x8000)) | + (((e - uint(0x38000000)) >> uint(13)) & uint(0x7c00)) | + ((f >> uint(13)) & uint(0x03ff)); + } } uint half2float(uint h) { - uint e = (h & uint(0x7C00)) >> uint(10); - uint m = (h & uint(0x03FF)) << uint(13); - uint v = m >> uint(23); - return (h & uint(0x8000)) << uint(16) | uint(e != uint(0)) * ((e + uint(112)) << uint(23) | m) | (uint(e == uint(0)) & uint(m != uint(0))) * ((v - uint(37)) << uint(23) | ((m << (uint(150) - v)) & uint(0x007FE000))); + uint h_e = h & uint(0x7c00); + return ((h & uint(0x8000)) << uint(16)) | uint((h_e >> uint(10)) != uint(0)) * (((h_e + uint(0x1c000)) << uint(13)) | ((h & uint(0x03ff)) << uint(13))); } uint godot_packHalf2x16(vec2 v) { @@ -50,6 +53,17 @@ vec2 godot_unpackSnorm2x16(uint p) { return clamp((v - 32767.0) * vec2(0.00003051851), vec2(-1.0), vec2(1.0)); } +#define packHalf2x16 godot_packHalf2x16 +#define unpackHalf2x16 godot_unpackHalf2x16 +#define packUnorm2x16 godot_packUnorm2x16 +#define unpackUnorm2x16 godot_unpackUnorm2x16 +#define packSnorm2x16 godot_packSnorm2x16 +#define unpackSnorm2x16 godot_unpackSnorm2x16 + +#endif // USE_HALF2FLOAT + +// Always expose these as they are ES310 functions and not available in ES300 or GLSL 330. + uint godot_packUnorm4x8(vec4 v) { uvec4 uv = uvec4(round(clamp(v, vec4(0.0), vec4(1.0)) * 255.0)); return uv.x | (uv.y << uint(8)) | (uv.z << uint(16)) | (uv.w << uint(24)); @@ -73,9 +87,3 @@ vec4 godot_unpackSnorm4x8(uint p) { #define unpackUnorm4x8 godot_unpackUnorm4x8 #define packSnorm4x8 godot_packSnorm4x8 #define unpackSnorm4x8 godot_unpackSnorm4x8 -#define packHalf2x16 godot_packHalf2x16 -#define unpackHalf2x16 godot_unpackHalf2x16 -#define packUnorm2x16 godot_packUnorm2x16 -#define unpackUnorm2x16 godot_unpackUnorm2x16 -#define packSnorm2x16 godot_packSnorm2x16 -#define unpackSnorm2x16 godot_unpackSnorm2x16 diff --git a/drivers/gles3/storage/config.cpp b/drivers/gles3/storage/config.cpp index 209b7dd7d2..07dd5f58c1 100644 --- a/drivers/gles3/storage/config.cpp +++ b/drivers/gles3/storage/config.cpp @@ -231,6 +231,13 @@ Config::Config() { } else if (rendering_device_name == "PowerVR Rogue GE8320") { disable_transform_feedback_shader_cache = true; } + + if (OS::get_singleton()->get_current_rendering_driver_name() == "opengl3_angle") { + polyfill_half2float = false; + } +#ifdef WEB_ENABLED + polyfill_half2float = false; +#endif } Config::~Config() { diff --git a/drivers/gles3/storage/config.h b/drivers/gles3/storage/config.h index d60f295d66..51f6714320 100644 --- a/drivers/gles3/storage/config.h +++ b/drivers/gles3/storage/config.h @@ -95,13 +95,16 @@ public: bool multiview_supported = false; bool external_texture_supported = false; - // Adreno 3XX compatibility - bool disable_particles_workaround = false; // set to 'true' to disable 'GPUParticles' + // Adreno 3XX compatibility. + bool disable_particles_workaround = false; // Set to 'true' to disable 'GPUParticles'. bool flip_xy_workaround = false; - // PowerVR GE 8320 workaround + // PowerVR GE 8320 workaround. bool disable_transform_feedback_shader_cache = false; + // ANGLE shader workaround. + bool polyfill_half2float = true; + #ifdef ANDROID_ENABLED PFNGLFRAMEBUFFERTEXTUREMULTIVIEWOVRPROC eglFramebufferTextureMultiviewOVR = nullptr; PFNGLTEXSTORAGE3DMULTISAMPLEPROC eglTexStorage3DMultisample = nullptr; @@ -110,7 +113,7 @@ public: PFNEGLIMAGETARGETTEXTURE2DOESPROC eglEGLImageTargetTexture2DOES = nullptr; #endif - static Config *get_singleton() { return singleton; }; + static Config *get_singleton() { return singleton; } Config(); ~Config(); diff --git a/drivers/gles3/storage/light_storage.cpp b/drivers/gles3/storage/light_storage.cpp index 9b976c2206..9b81430d45 100644 --- a/drivers/gles3/storage/light_storage.cpp +++ b/drivers/gles3/storage/light_storage.cpp @@ -213,6 +213,23 @@ void LightStorage::light_set_cull_mask(RID p_light, uint32_t p_mask) { light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); } +void LightStorage::light_set_shadow_caster_mask(RID p_light, uint32_t p_caster_mask) { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_NULL(light); + + light->shadow_caster_mask = p_caster_mask; + + light->version++; + light->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_LIGHT); +} + +uint32_t LightStorage::light_get_shadow_caster_mask(RID p_light) const { + Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_NULL_V(light, 0); + + return light->shadow_caster_mask; +} + void LightStorage::light_set_distance_fade(RID p_light, bool p_enabled, float p_begin, float p_shadow, float p_length) { Light *light = light_owner.get_or_null(p_light); ERR_FAIL_NULL(light); diff --git a/drivers/gles3/storage/light_storage.h b/drivers/gles3/storage/light_storage.h index ed00dd235f..0695102640 100644 --- a/drivers/gles3/storage/light_storage.h +++ b/drivers/gles3/storage/light_storage.h @@ -59,6 +59,7 @@ struct Light { RS::LightBakeMode bake_mode = RS::LIGHT_BAKE_DYNAMIC; uint32_t max_sdfgi_cascade = 2; uint32_t cull_mask = 0xFFFFFFFF; + uint32_t shadow_caster_mask = 0xFFFFFFFF; bool distance_fade = false; real_t distance_fade_begin = 40.0; real_t distance_fade_shadow = 50.0; @@ -305,8 +306,8 @@ public: /* Light API */ - Light *get_light(RID p_rid) { return light_owner.get_or_null(p_rid); }; - bool owns_light(RID p_rid) { return light_owner.owns(p_rid); }; + Light *get_light(RID p_rid) { return light_owner.get_or_null(p_rid); } + bool owns_light(RID p_rid) { return light_owner.owns(p_rid); } void _light_initialize(RID p_rid, RS::LightType p_type); @@ -327,6 +328,8 @@ public: virtual void light_set_cull_mask(RID p_light, uint32_t p_mask) override; virtual void light_set_distance_fade(RID p_light, bool p_enabled, float p_begin, float p_shadow, float p_length) override; virtual void light_set_reverse_cull_face_mode(RID p_light, bool p_enabled) override; + virtual void light_set_shadow_caster_mask(RID p_light, uint32_t p_caster_mask) override; + virtual uint32_t light_get_shadow_caster_mask(RID p_light) const override; virtual void light_set_bake_mode(RID p_light, RS::LightBakeMode p_bake_mode) override; virtual void light_set_max_sdfgi_cascade(RID p_light, uint32_t p_cascade) override {} @@ -431,8 +434,8 @@ public: /* LIGHT INSTANCE API */ - LightInstance *get_light_instance(RID p_rid) { return light_instance_owner.get_or_null(p_rid); }; - bool owns_light_instance(RID p_rid) { return light_instance_owner.owns(p_rid); }; + LightInstance *get_light_instance(RID p_rid) { return light_instance_owner.get_or_null(p_rid); } + bool owns_light_instance(RID p_rid) { return light_instance_owner.owns(p_rid); } virtual RID light_instance_create(RID p_light) override; virtual void light_instance_free(RID p_light_instance) override; @@ -630,8 +633,8 @@ public: /* PROBE API */ - ReflectionProbe *get_reflection_probe(RID p_rid) { return reflection_probe_owner.get_or_null(p_rid); }; - bool owns_reflection_probe(RID p_rid) { return reflection_probe_owner.owns(p_rid); }; + ReflectionProbe *get_reflection_probe(RID p_rid) { return reflection_probe_owner.get_or_null(p_rid); } + bool owns_reflection_probe(RID p_rid) { return reflection_probe_owner.owns(p_rid); } virtual RID reflection_probe_allocate() override; virtual void reflection_probe_initialize(RID p_rid) override; @@ -712,8 +715,8 @@ public: /* LIGHTMAP CAPTURE */ - Lightmap *get_lightmap(RID p_rid) { return lightmap_owner.get_or_null(p_rid); }; - bool owns_lightmap(RID p_rid) { return lightmap_owner.owns(p_rid); }; + Lightmap *get_lightmap(RID p_rid) { return lightmap_owner.get_or_null(p_rid); } + bool owns_lightmap(RID p_rid) { return lightmap_owner.owns(p_rid); } virtual RID lightmap_allocate() override; virtual void lightmap_initialize(RID p_rid) override; @@ -736,15 +739,15 @@ public: /* LIGHTMAP INSTANCE */ - LightmapInstance *get_lightmap_instance(RID p_rid) { return lightmap_instance_owner.get_or_null(p_rid); }; - bool owns_lightmap_instance(RID p_rid) { return lightmap_instance_owner.owns(p_rid); }; + LightmapInstance *get_lightmap_instance(RID p_rid) { return lightmap_instance_owner.get_or_null(p_rid); } + bool owns_lightmap_instance(RID p_rid) { return lightmap_instance_owner.owns(p_rid); } virtual RID lightmap_instance_create(RID p_lightmap) override; virtual void lightmap_instance_free(RID p_lightmap) override; virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) override; /* SHADOW ATLAS API */ - bool owns_shadow_atlas(RID p_rid) { return shadow_atlas_owner.owns(p_rid); }; + bool owns_shadow_atlas(RID p_rid) { return shadow_atlas_owner.owns(p_rid); } virtual RID shadow_atlas_create() override; virtual void shadow_atlas_free(RID p_atlas) override; diff --git a/drivers/gles3/storage/material_storage.cpp b/drivers/gles3/storage/material_storage.cpp index 684f179492..04cbf7f2cd 100644 --- a/drivers/gles3/storage/material_storage.cpp +++ b/drivers/gles3/storage/material_storage.cpp @@ -1237,6 +1237,8 @@ MaterialStorage::MaterialStorage() { actions.renames["PI"] = _MKSTR(Math_PI); actions.renames["TAU"] = _MKSTR(Math_TAU); actions.renames["E"] = _MKSTR(Math_E); + actions.renames["OUTPUT_IS_SRGB"] = "SHADER_IS_SRGB"; + actions.renames["CLIP_SPACE_FAR"] = "SHADER_SPACE_FAR"; actions.renames["VIEWPORT_SIZE"] = "scene_data.viewport_size"; actions.renames["FRAGCOORD"] = "gl_FragCoord"; @@ -1276,8 +1278,6 @@ MaterialStorage::MaterialStorage() { actions.renames["CUSTOM1"] = "custom1_attrib"; actions.renames["CUSTOM2"] = "custom2_attrib"; actions.renames["CUSTOM3"] = "custom3_attrib"; - actions.renames["OUTPUT_IS_SRGB"] = "SHADER_IS_SRGB"; - actions.renames["CLIP_SPACE_FAR"] = "SHADER_SPACE_FAR"; actions.renames["LIGHT_VERTEX"] = "light_vertex"; actions.renames["NODE_POSITION_WORLD"] = "model_matrix[3].xyz"; diff --git a/drivers/gles3/storage/material_storage.h b/drivers/gles3/storage/material_storage.h index 392ebcc570..6c21abc276 100644 --- a/drivers/gles3/storage/material_storage.h +++ b/drivers/gles3/storage/material_storage.h @@ -576,8 +576,8 @@ public: /* SHADER API */ - Shader *get_shader(RID p_rid) { return shader_owner.get_or_null(p_rid); }; - bool owns_shader(RID p_rid) { return shader_owner.owns(p_rid); }; + Shader *get_shader(RID p_rid) { return shader_owner.get_or_null(p_rid); } + bool owns_shader(RID p_rid) { return shader_owner.owns(p_rid); } void _shader_make_dirty(Shader *p_shader); @@ -598,8 +598,8 @@ public: /* MATERIAL API */ - Material *get_material(RID p_rid) { return material_owner.get_or_null(p_rid); }; - bool owns_material(RID p_rid) { return material_owner.owns(p_rid); }; + Material *get_material(RID p_rid) { return material_owner.get_or_null(p_rid); } + bool owns_material(RID p_rid) { return material_owner.owns(p_rid); } void _material_queue_update(Material *material, bool p_uniform, bool p_texture); void _update_queued_materials(); diff --git a/drivers/gles3/storage/mesh_storage.cpp b/drivers/gles3/storage/mesh_storage.cpp index 73d95d75ba..630bfdf1b7 100644 --- a/drivers/gles3/storage/mesh_storage.cpp +++ b/drivers/gles3/storage/mesh_storage.cpp @@ -1974,6 +1974,10 @@ void MeshStorage::_multimesh_set_buffer(RID p_multimesh, const Vector<float> &p_ } } +RID MeshStorage::_multimesh_get_buffer_rd_rid(RID p_multimesh) const { + ERR_FAIL_V_MSG(RID(), "GLES3 does not contain a Rid for the multimesh buffer."); +} + Vector<float> MeshStorage::_multimesh_get_buffer(RID p_multimesh) const { MultiMesh *multimesh = multimesh_owner.get_or_null(p_multimesh); ERR_FAIL_NULL_V(multimesh, Vector<float>()); diff --git a/drivers/gles3/storage/mesh_storage.h b/drivers/gles3/storage/mesh_storage.h index 8615b89a30..5e3a56366e 100644 --- a/drivers/gles3/storage/mesh_storage.h +++ b/drivers/gles3/storage/mesh_storage.h @@ -280,8 +280,8 @@ public: /* MESH API */ - Mesh *get_mesh(RID p_rid) { return mesh_owner.get_or_null(p_rid); }; - bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); }; + Mesh *get_mesh(RID p_rid) { return mesh_owner.get_or_null(p_rid); } + bool owns_mesh(RID p_rid) { return mesh_owner.owns(p_rid); } virtual RID mesh_allocate() override; virtual void mesh_initialize(RID p_rid) override; @@ -443,8 +443,8 @@ public: /* MESH INSTANCE API */ - MeshInstance *get_mesh_instance(RID p_rid) { return mesh_instance_owner.get_or_null(p_rid); }; - bool owns_mesh_instance(RID p_rid) { return mesh_instance_owner.owns(p_rid); }; + MeshInstance *get_mesh_instance(RID p_rid) { return mesh_instance_owner.get_or_null(p_rid); } + bool owns_mesh_instance(RID p_rid) { return mesh_instance_owner.owns(p_rid); } virtual RID mesh_instance_create(RID p_base) override; virtual void mesh_instance_free(RID p_rid) override; @@ -492,8 +492,8 @@ public: /* MULTIMESH API */ - MultiMesh *get_multimesh(RID p_rid) { return multimesh_owner.get_or_null(p_rid); }; - bool owns_multimesh(RID p_rid) { return multimesh_owner.owns(p_rid); }; + MultiMesh *get_multimesh(RID p_rid) { return multimesh_owner.get_or_null(p_rid); } + bool owns_multimesh(RID p_rid) { return multimesh_owner.owns(p_rid); } virtual RID _multimesh_allocate() override; virtual void _multimesh_initialize(RID p_rid) override; @@ -517,6 +517,7 @@ public: virtual Color _multimesh_instance_get_color(RID p_multimesh, int p_index) const override; virtual Color _multimesh_instance_get_custom_data(RID p_multimesh, int p_index) const override; virtual void _multimesh_set_buffer(RID p_multimesh, const Vector<float> &p_buffer) override; + virtual RID _multimesh_get_buffer_rd_rid(RID p_multimesh) const override; virtual Vector<float> _multimesh_get_buffer(RID p_multimesh) const override; virtual void _multimesh_set_visible_instances(RID p_multimesh, int p_visible) override; @@ -571,8 +572,8 @@ public: /* SKELETON API */ - Skeleton *get_skeleton(RID p_rid) { return skeleton_owner.get_or_null(p_rid); }; - bool owns_skeleton(RID p_rid) { return skeleton_owner.owns(p_rid); }; + Skeleton *get_skeleton(RID p_rid) { return skeleton_owner.get_or_null(p_rid); } + bool owns_skeleton(RID p_rid) { return skeleton_owner.owns(p_rid); } virtual RID skeleton_allocate() override; virtual void skeleton_initialize(RID p_rid) override; diff --git a/drivers/gles3/storage/particles_storage.h b/drivers/gles3/storage/particles_storage.h index 086f5f7936..0552a5324f 100644 --- a/drivers/gles3/storage/particles_storage.h +++ b/drivers/gles3/storage/particles_storage.h @@ -396,7 +396,7 @@ public: _FORCE_INLINE_ bool particles_has_collision(RID p_particles) { Particles *particles = particles_owner.get_or_null(p_particles); - ERR_FAIL_NULL_V(particles, 0); + ERR_FAIL_NULL_V(particles, false); return particles->has_collision_cache; } diff --git a/drivers/gles3/storage/texture_storage.cpp b/drivers/gles3/storage/texture_storage.cpp index d7b4d6911d..14bbe635a4 100644 --- a/drivers/gles3/storage/texture_storage.cpp +++ b/drivers/gles3/storage/texture_storage.cpp @@ -230,6 +230,32 @@ TextureStorage::TextureStorage() { sdf_shader.shader_version = sdf_shader.shader.version_create(); } + // Initialize texture placeholder data for the `texture_*_placeholder_initialize()` methods. + + constexpr int placeholder_size = 4; + texture_2d_placeholder = Image::create_empty(placeholder_size, placeholder_size, false, Image::FORMAT_RGBA8); + // Draw a magenta/black checkerboard pattern. + for (int i = 0; i < placeholder_size * placeholder_size; i++) { + const int x = i % placeholder_size; + const int y = i / placeholder_size; + texture_2d_placeholder->set_pixel(x, y, (x + y) % 2 == 0 ? Color(1, 0, 1) : Color(0, 0, 0)); + } + + texture_2d_array_placeholder.push_back(texture_2d_placeholder); + + for (int i = 0; i < 6; i++) { + cubemap_placeholder.push_back(texture_2d_placeholder); + } + + Ref<Image> texture_2d_placeholder_rotated; + texture_2d_placeholder_rotated.instantiate(); + texture_2d_placeholder_rotated->copy_from(texture_2d_placeholder); + texture_2d_placeholder_rotated->rotate_90(CLOCKWISE); + for (int i = 0; i < 4; i++) { + // Alternate checkerboard pattern on odd layers (by using a copy that is rotated 90 degrees). + texture_3d_placeholder.push_back(i % 2 == 0 ? texture_2d_placeholder : texture_2d_placeholder_rotated); + } + #ifdef GL_API_ENABLED if (RasterizerGLES3::is_gles_over_gl()) { glEnable(GL_PROGRAM_POINT_SIZE); @@ -1014,46 +1040,19 @@ void TextureStorage::texture_proxy_update(RID p_texture, RID p_proxy_to) { } void TextureStorage::texture_2d_placeholder_initialize(RID p_texture) { - //this could be better optimized to reuse an existing image , done this way - //for now to get it working - Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); - image->fill(Color(1, 0, 1, 1)); - - texture_2d_initialize(p_texture, image); + texture_2d_initialize(p_texture, texture_2d_placeholder); } -void TextureStorage::texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type) { - //this could be better optimized to reuse an existing image , done this way - //for now to get it working - Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); - image->fill(Color(1, 0, 1, 1)); - - Vector<Ref<Image>> images; +void TextureStorage::texture_2d_layered_placeholder_initialize(RID p_texture, RS::TextureLayeredType p_layered_type) { if (p_layered_type == RS::TEXTURE_LAYERED_2D_ARRAY) { - images.push_back(image); + texture_2d_layered_initialize(p_texture, texture_2d_array_placeholder, p_layered_type); } else { - //cube - for (int i = 0; i < 6; i++) { - images.push_back(image); - } + texture_2d_layered_initialize(p_texture, cubemap_placeholder, p_layered_type); } - - texture_2d_layered_initialize(p_texture, images, p_layered_type); } void TextureStorage::texture_3d_placeholder_initialize(RID p_texture) { - //this could be better optimized to reuse an existing image , done this way - //for now to get it working - Ref<Image> image = Image::create_empty(4, 4, false, Image::FORMAT_RGBA8); - image->fill(Color(1, 0, 1, 1)); - - Vector<Ref<Image>> images; - //cube - for (int i = 0; i < 4; i++) { - images.push_back(image); - } - - texture_3d_initialize(p_texture, Image::FORMAT_RGBA8, 4, 4, 4, false, images); + texture_3d_initialize(p_texture, Image::FORMAT_RGBA8, 4, 4, 4, false, texture_3d_placeholder); } Ref<Image> TextureStorage::texture_2d_get(RID p_texture) const { @@ -2453,7 +2452,7 @@ Point2i TextureStorage::render_target_get_position(RID p_render_target) const { ERR_FAIL_NULL_V(rt, Point2i()); return rt->position; -}; +} void TextureStorage::render_target_set_size(RID p_render_target, int p_width, int p_height, uint32_t p_view_count) { RenderTarget *rt = render_target_owner.get_or_null(p_render_target); diff --git a/drivers/gles3/storage/texture_storage.h b/drivers/gles3/storage/texture_storage.h index 3786c8c690..916662dcd6 100644 --- a/drivers/gles3/storage/texture_storage.h +++ b/drivers/gles3/storage/texture_storage.h @@ -36,6 +36,7 @@ #include "platform_gl.h" #include "config.h" +#include "core/io/image.h" #include "core/os/os.h" #include "core/templates/rid_owner.h" #include "servers/rendering/renderer_compositor.h" @@ -477,8 +478,8 @@ public: /* Canvas Texture API */ - CanvasTexture *get_canvas_texture(RID p_rid) { return canvas_texture_owner.get_or_null(p_rid); }; - bool owns_canvas_texture(RID p_rid) { return canvas_texture_owner.owns(p_rid); }; + CanvasTexture *get_canvas_texture(RID p_rid) { return canvas_texture_owner.get_or_null(p_rid); } + bool owns_canvas_texture(RID p_rid) { return canvas_texture_owner.owns(p_rid); } virtual RID canvas_texture_allocate() override; virtual void canvas_texture_initialize(RID p_rid) override; @@ -498,8 +499,8 @@ public: return texture_owner.get_or_null(texture->proxy_to); } return texture; - }; - bool owns_texture(RID p_rid) { return texture_owner.owns(p_rid); }; + } + bool owns_texture(RID p_rid) { return texture_owner.owns(p_rid); } void texture_2d_initialize_from_texture(RID p_texture, Texture &p_tex) { texture_owner.initialize_rid(p_texture, p_tex); @@ -521,6 +522,11 @@ public: virtual void texture_external_update(RID p_texture, int p_width, int p_height, uint64_t p_external_buffer) override; virtual void texture_proxy_update(RID p_proxy, RID p_base) override; + Ref<Image> texture_2d_placeholder; + Vector<Ref<Image>> texture_2d_array_placeholder; + Vector<Ref<Image>> cubemap_placeholder; + Vector<Ref<Image>> texture_3d_placeholder; + //these two APIs can be used together or in combination with the others. virtual void texture_2d_placeholder_initialize(RID p_texture) override; virtual void texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type) override; @@ -612,8 +618,8 @@ public: static GLuint system_fbo; - RenderTarget *get_render_target(RID p_rid) { return render_target_owner.get_or_null(p_rid); }; - bool owns_render_target(RID p_rid) { return render_target_owner.owns(p_rid); }; + RenderTarget *get_render_target(RID p_rid) { return render_target_owner.get_or_null(p_rid); } + bool owns_render_target(RID p_rid) { return render_target_owner.owns(p_rid); } void check_backbuffer(RenderTarget *rt, const bool uses_screen_texture, const bool uses_depth_texture); diff --git a/drivers/gles3/storage/utilities.h b/drivers/gles3/storage/utilities.h index 7c3b08717e..88ac802018 100644 --- a/drivers/gles3/storage/utilities.h +++ b/drivers/gles3/storage/utilities.h @@ -165,8 +165,8 @@ public: /* VISIBILITY NOTIFIER */ - VisibilityNotifier *get_visibility_notifier(RID p_rid) { return visibility_notifier_owner.get_or_null(p_rid); }; - bool owns_visibility_notifier(RID p_rid) const { return visibility_notifier_owner.owns(p_rid); }; + VisibilityNotifier *get_visibility_notifier(RID p_rid) { return visibility_notifier_owner.get_or_null(p_rid); } + bool owns_visibility_notifier(RID p_rid) const { return visibility_notifier_owner.owns(p_rid); } virtual RID visibility_notifier_allocate() override; virtual void visibility_notifier_initialize(RID p_notifier) override; diff --git a/drivers/metal/metal_objects.h b/drivers/metal/metal_objects.h index 030b353ee8..5eda826d2a 100644 --- a/drivers/metal/metal_objects.h +++ b/drivers/metal/metal_objects.h @@ -82,6 +82,9 @@ MTL_CLASS(Texture) } //namespace MTL +/// Metal buffer index for the view mask when rendering multi-view. +const uint32_t VIEW_MASK_BUFFER_INDEX = 24; + enum ShaderStageUsage : uint32_t { None = 0, Vertex = RDD::SHADER_STAGE_VERTEX_BIT, @@ -96,6 +99,22 @@ _FORCE_INLINE_ ShaderStageUsage &operator|=(ShaderStageUsage &p_a, int p_b) { return p_a; } +enum StageResourceUsage : uint32_t { + VertexRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_VERTEX * 2), + VertexWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_VERTEX * 2), + FragmentRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_FRAGMENT * 2), + FragmentWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_FRAGMENT * 2), + TesselationControlRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_TESSELATION_CONTROL * 2), + TesselationControlWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_TESSELATION_CONTROL * 2), + TesselationEvaluationRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_TESSELATION_EVALUATION * 2), + TesselationEvaluationWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_TESSELATION_EVALUATION * 2), + ComputeRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_COMPUTE * 2), + ComputeWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_COMPUTE * 2), +}; + +typedef LocalVector<__unsafe_unretained id<MTLResource>> ResourceVector; +typedef HashMap<StageResourceUsage, ResourceVector> ResourceUsageMap; + enum class MDCommandBufferStateType { None, Render, @@ -126,6 +145,12 @@ struct ClearAttKey { const static uint32_t STENCIL_INDEX = DEPTH_INDEX + 1; const static uint32_t ATTACHMENT_COUNT = STENCIL_INDEX + 1; + enum Flags : uint16_t { + CLEAR_FLAGS_NONE = 0, + CLEAR_FLAGS_LAYERED = 1 << 0, + }; + + Flags flags = CLEAR_FLAGS_NONE; uint16_t sample_count = 0; uint16_t pixel_formats[ATTACHMENT_COUNT] = { 0 }; @@ -134,19 +159,22 @@ struct ClearAttKey { _FORCE_INLINE_ void set_stencil_format(MTLPixelFormat p_fmt) { pixel_formats[STENCIL_INDEX] = p_fmt; } _FORCE_INLINE_ MTLPixelFormat depth_format() const { return (MTLPixelFormat)pixel_formats[DEPTH_INDEX]; } _FORCE_INLINE_ MTLPixelFormat stencil_format() const { return (MTLPixelFormat)pixel_formats[STENCIL_INDEX]; } + _FORCE_INLINE_ void enable_layered_rendering() { flags::set(flags, CLEAR_FLAGS_LAYERED); } _FORCE_INLINE_ bool is_enabled(uint32_t p_idx) const { return pixel_formats[p_idx] != 0; } _FORCE_INLINE_ bool is_depth_enabled() const { return pixel_formats[DEPTH_INDEX] != 0; } _FORCE_INLINE_ bool is_stencil_enabled() const { return pixel_formats[STENCIL_INDEX] != 0; } + _FORCE_INLINE_ bool is_layered_rendering_enabled() const { return flags::any(flags, CLEAR_FLAGS_LAYERED); } _FORCE_INLINE_ bool operator==(const ClearAttKey &p_rhs) const { return memcmp(this, &p_rhs, sizeof(ClearAttKey)) == 0; } uint32_t hash() const { - uint32_t h = hash_murmur3_one_32(sample_count); + uint32_t h = hash_murmur3_one_32(flags); + h = hash_murmur3_one_32(sample_count, h); h = hash_murmur3_buffer(pixel_formats, ATTACHMENT_COUNT * sizeof(pixel_formats[0]), h); - return h; + return hash_fmix32(h); } }; @@ -190,6 +218,97 @@ public: ~MDResourceCache() = default; }; +enum class MDAttachmentType : uint8_t { + None = 0, + Color = 1 << 0, + Depth = 1 << 1, + Stencil = 1 << 2, +}; + +_FORCE_INLINE_ MDAttachmentType &operator|=(MDAttachmentType &p_a, MDAttachmentType p_b) { + flags::set(p_a, p_b); + return p_a; +} + +_FORCE_INLINE_ bool operator&(MDAttachmentType p_a, MDAttachmentType p_b) { + return uint8_t(p_a) & uint8_t(p_b); +} + +struct MDSubpass { + uint32_t subpass_index = 0; + uint32_t view_count = 0; + LocalVector<RDD::AttachmentReference> input_references; + LocalVector<RDD::AttachmentReference> color_references; + RDD::AttachmentReference depth_stencil_reference; + LocalVector<RDD::AttachmentReference> resolve_references; + + MTLFmtCaps getRequiredFmtCapsForAttachmentAt(uint32_t p_index) const; +}; + +struct API_AVAILABLE(macos(11.0), ios(14.0)) MDAttachment { +private: + uint32_t index = 0; + uint32_t firstUseSubpassIndex = 0; + uint32_t lastUseSubpassIndex = 0; + +public: + MTLPixelFormat format = MTLPixelFormatInvalid; + MDAttachmentType type = MDAttachmentType::None; + MTLLoadAction loadAction = MTLLoadActionDontCare; + MTLStoreAction storeAction = MTLStoreActionDontCare; + MTLLoadAction stencilLoadAction = MTLLoadActionDontCare; + MTLStoreAction stencilStoreAction = MTLStoreActionDontCare; + uint32_t samples = 1; + + /*! + * @brief Returns true if this attachment is first used in the given subpass. + * @param p_subpass + * @return + */ + _FORCE_INLINE_ bool isFirstUseOf(MDSubpass const &p_subpass) const { + return p_subpass.subpass_index == firstUseSubpassIndex; + } + + /*! + * @brief Returns true if this attachment is last used in the given subpass. + * @param p_subpass + * @return + */ + _FORCE_INLINE_ bool isLastUseOf(MDSubpass const &p_subpass) const { + return p_subpass.subpass_index == lastUseSubpassIndex; + } + + void linkToSubpass(MDRenderPass const &p_pass); + + MTLStoreAction getMTLStoreAction(MDSubpass const &p_subpass, + bool p_is_rendering_entire_area, + bool p_has_resolve, + bool p_can_resolve, + bool p_is_stencil) const; + bool configureDescriptor(MTLRenderPassAttachmentDescriptor *p_desc, + PixelFormats &p_pf, + MDSubpass const &p_subpass, + id<MTLTexture> p_attachment, + bool p_is_rendering_entire_area, + bool p_has_resolve, + bool p_can_resolve, + bool p_is_stencil) const; + /** Returns whether this attachment should be cleared in the subpass. */ + bool shouldClear(MDSubpass const &p_subpass, bool p_is_stencil) const; +}; + +class API_AVAILABLE(macos(11.0), ios(14.0)) MDRenderPass { +public: + Vector<MDAttachment> attachments; + Vector<MDSubpass> subpasses; + + uint32_t get_sample_count() const { + return attachments.is_empty() ? 1 : attachments[0].samples; + } + + MDRenderPass(Vector<MDAttachment> &p_attachments, Vector<MDSubpass> &p_subpasses); +}; + class API_AVAILABLE(macos(11.0), ios(14.0)) MDCommandBuffer { private: RenderingDeviceDriverMetal *device_driver = nullptr; @@ -204,8 +323,8 @@ private: void _render_set_dirty_state(); void _render_bind_uniform_sets(); - static void _populate_vertices(simd::float4 *p_vertices, Size2i p_fb_size, VectorView<Rect2i> p_rects); - static uint32_t _populate_vertices(simd::float4 *p_vertices, uint32_t p_index, Rect2i const &p_rect, Size2i p_fb_size); + void _populate_vertices(simd::float4 *p_vertices, Size2i p_fb_size, VectorView<Rect2i> p_rects); + uint32_t _populate_vertices(simd::float4 *p_vertices, uint32_t p_index, Rect2i const &p_rect, Size2i p_fb_size); void _end_render_pass(); void _render_clear_render_area(); @@ -230,6 +349,7 @@ public: uint32_t index_offset = 0; LocalVector<id<MTLBuffer> __unsafe_unretained> vertex_buffers; LocalVector<NSUInteger> vertex_offsets; + ResourceUsageMap resource_usage; // clang-format off enum DirtyFlag: uint8_t { DIRTY_NONE = 0b0000'0000, @@ -251,26 +371,12 @@ public: // Bit mask of the uniform sets that are dirty, to prevent redundant binding. uint64_t uniform_set_mask = 0; - _FORCE_INLINE_ void reset() { - pass = nil; - frameBuffer = nil; - pipeline = nil; - current_subpass = UINT32_MAX; - render_area = {}; - is_rendering_entire_area = false; - desc = nil; - encoder = nil; - index_buffer = nil; - index_type = MTLIndexTypeUInt16; - dirty = DIRTY_NONE; - uniform_sets.clear(); - uniform_set_mask = 0; - clear_values.clear(); - viewports.clear(); - scissors.clear(); - blend_constants.reset(); - vertex_buffers.clear(); - vertex_offsets.clear(); + _FORCE_INLINE_ void reset(); + void end_encoding(); + + _ALWAYS_INLINE_ const MDSubpass &get_subpass() const { + DEV_ASSERT(pass != nullptr); + return pass->subpasses[current_subpass]; } _FORCE_INLINE_ void mark_viewport_dirty() { @@ -356,13 +462,20 @@ public: } render; // State specific for a compute pass. - struct { + struct ComputeState { MDComputePipeline *pipeline = nullptr; id<MTLComputeCommandEncoder> encoder = nil; + ResourceUsageMap resource_usage; _FORCE_INLINE_ void reset() { pipeline = nil; encoder = nil; + // Keep the keys, as they are likely to be used again. + for (KeyValue<StageResourceUsage, LocalVector<__unsafe_unretained id<MTLResource>>> &kv : resource_usage) { + kv.value.clear(); + } } + + void end_encoding(); } compute; // State specific to a blit pass. @@ -619,6 +732,7 @@ public: uint32_t size = 0; } frag; } push_constants; + bool needs_view_mask_buffer = false; MDLibrary *vert; MDLibrary *frag; @@ -629,20 +743,10 @@ public: void encode_push_constant_data(VectorView<uint32_t> p_data, MDCommandBuffer *p_cb) final; - MDRenderShader(CharString p_name, Vector<UniformSet> p_sets, MDLibrary *p_vert, MDLibrary *p_frag); -}; - -enum StageResourceUsage : uint32_t { - VertexRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_VERTEX * 2), - VertexWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_VERTEX * 2), - FragmentRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_FRAGMENT * 2), - FragmentWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_FRAGMENT * 2), - TesselationControlRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_TESSELATION_CONTROL * 2), - TesselationControlWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_TESSELATION_CONTROL * 2), - TesselationEvaluationRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_TESSELATION_EVALUATION * 2), - TesselationEvaluationWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_TESSELATION_EVALUATION * 2), - ComputeRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_COMPUTE * 2), - ComputeWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_COMPUTE * 2), + MDRenderShader(CharString p_name, + bool p_needs_view_mask_buffer, + Vector<UniformSet> p_sets, + MDLibrary *p_vert, MDLibrary *p_frag); }; _FORCE_INLINE_ StageResourceUsage &operator|=(StageResourceUsage &p_a, uint32_t p_b) { @@ -667,7 +771,13 @@ struct HashMapComparatorDefault<RDD::ShaderID> { struct BoundUniformSet { id<MTLBuffer> buffer; - HashMap<id<MTLResource>, StageResourceUsage> bound_resources; + ResourceUsageMap usage_to_resources; + + /// Perform a 2-way merge each key of `ResourceVector` resources from this set into the + /// destination set. + /// + /// Assumes the vectors of resources are sorted. + void merge_into(ResourceUsageMap &p_dst) const; }; class API_AVAILABLE(macos(11.0), ios(14.0)) MDUniformSet { @@ -679,96 +789,6 @@ public: BoundUniformSet &boundUniformSetForShader(MDShader *p_shader, id<MTLDevice> p_device); }; -enum class MDAttachmentType : uint8_t { - None = 0, - Color = 1 << 0, - Depth = 1 << 1, - Stencil = 1 << 2, -}; - -_FORCE_INLINE_ MDAttachmentType &operator|=(MDAttachmentType &p_a, MDAttachmentType p_b) { - flags::set(p_a, p_b); - return p_a; -} - -_FORCE_INLINE_ bool operator&(MDAttachmentType p_a, MDAttachmentType p_b) { - return uint8_t(p_a) & uint8_t(p_b); -} - -struct MDSubpass { - uint32_t subpass_index = 0; - LocalVector<RDD::AttachmentReference> input_references; - LocalVector<RDD::AttachmentReference> color_references; - RDD::AttachmentReference depth_stencil_reference; - LocalVector<RDD::AttachmentReference> resolve_references; - - MTLFmtCaps getRequiredFmtCapsForAttachmentAt(uint32_t p_index) const; -}; - -struct API_AVAILABLE(macos(11.0), ios(14.0)) MDAttachment { -private: - uint32_t index = 0; - uint32_t firstUseSubpassIndex = 0; - uint32_t lastUseSubpassIndex = 0; - -public: - MTLPixelFormat format = MTLPixelFormatInvalid; - MDAttachmentType type = MDAttachmentType::None; - MTLLoadAction loadAction = MTLLoadActionDontCare; - MTLStoreAction storeAction = MTLStoreActionDontCare; - MTLLoadAction stencilLoadAction = MTLLoadActionDontCare; - MTLStoreAction stencilStoreAction = MTLStoreActionDontCare; - uint32_t samples = 1; - - /*! - * @brief Returns true if this attachment is first used in the given subpass. - * @param p_subpass - * @return - */ - _FORCE_INLINE_ bool isFirstUseOf(MDSubpass const &p_subpass) const { - return p_subpass.subpass_index == firstUseSubpassIndex; - } - - /*! - * @brief Returns true if this attachment is last used in the given subpass. - * @param p_subpass - * @return - */ - _FORCE_INLINE_ bool isLastUseOf(MDSubpass const &p_subpass) const { - return p_subpass.subpass_index == lastUseSubpassIndex; - } - - void linkToSubpass(MDRenderPass const &p_pass); - - MTLStoreAction getMTLStoreAction(MDSubpass const &p_subpass, - bool p_is_rendering_entire_area, - bool p_has_resolve, - bool p_can_resolve, - bool p_is_stencil) const; - bool configureDescriptor(MTLRenderPassAttachmentDescriptor *p_desc, - PixelFormats &p_pf, - MDSubpass const &p_subpass, - id<MTLTexture> p_attachment, - bool p_is_rendering_entire_area, - bool p_has_resolve, - bool p_can_resolve, - bool p_is_stencil) const; - /** Returns whether this attachment should be cleared in the subpass. */ - bool shouldClear(MDSubpass const &p_subpass, bool p_is_stencil) const; -}; - -class API_AVAILABLE(macos(11.0), ios(14.0)) MDRenderPass { -public: - Vector<MDAttachment> attachments; - Vector<MDSubpass> subpasses; - - uint32_t get_sample_count() const { - return attachments.is_empty() ? 1 : attachments[0].samples; - } - - MDRenderPass(Vector<MDAttachment> &p_attachments, Vector<MDSubpass> &p_subpasses); -}; - class API_AVAILABLE(macos(11.0), ios(14.0)) MDPipeline { public: MDPipelineType type; @@ -818,7 +838,7 @@ public: if (!enabled) return; [p_enc setStencilFrontReferenceValue:front_reference backReferenceValue:back_reference]; - }; + } } stencil; struct { @@ -832,7 +852,7 @@ public: //if (!enabled) // return; [p_enc setBlendColorRed:r green:g blue:b alpha:a]; - }; + } } blend; _FORCE_INLINE_ void apply(id<MTLRenderCommandEncoder> __unsafe_unretained p_enc) const { @@ -869,13 +889,39 @@ public: }; class API_AVAILABLE(macos(11.0), ios(14.0)) MDFrameBuffer { -public: Vector<MTL::Texture> textures; + +public: Size2i size; MDFrameBuffer(Vector<MTL::Texture> p_textures, Size2i p_size) : textures(p_textures), size(p_size) {} MDFrameBuffer() {} + /// Returns the texture at the given index. + _ALWAYS_INLINE_ MTL::Texture get_texture(uint32_t p_idx) const { + return textures[p_idx]; + } + + /// Returns true if the texture at the given index is not nil. + _ALWAYS_INLINE_ bool has_texture(uint32_t p_idx) const { + return textures[p_idx] != nil; + } + + /// Set the texture at the given index. + _ALWAYS_INLINE_ void set_texture(uint32_t p_idx, MTL::Texture p_texture) { + textures.write[p_idx] = p_texture; + } + + /// Unset or nil the texture at the given index. + _ALWAYS_INLINE_ void unset_texture(uint32_t p_idx) { + textures.write[p_idx] = nil; + } + + /// Resizes buffers to the specified size. + _ALWAYS_INLINE_ void set_texture_count(uint32_t p_size) { + textures.resize(p_size); + } + virtual ~MDFrameBuffer() = default; }; diff --git a/drivers/metal/metal_objects.mm b/drivers/metal/metal_objects.mm index 596728212a..11ab209d60 100644 --- a/drivers/metal/metal_objects.mm +++ b/drivers/metal/metal_objects.mm @@ -58,7 +58,7 @@ void MDCommandBuffer::begin() { DEV_ASSERT(commandBuffer == nil); - commandBuffer = queue.commandBuffer; + commandBuffer = queue.commandBufferWithUnretainedReferences; } void MDCommandBuffer::end() { @@ -96,6 +96,9 @@ void MDCommandBuffer::bind_pipeline(RDD::PipelineID p_pipeline) { MDRenderPipeline *rp = (MDRenderPipeline *)p; if (render.encoder == nil) { + // This error would happen if the render pass failed. + ERR_FAIL_NULL_MSG(render.desc, "Render pass descriptor is null."); + // This condition occurs when there are no attachments when calling render_next_subpass() // and is due to the SUPPORTS_FRAGMENT_SHADER_WITH_ONLY_SIDE_EFFECTS flag. render.desc.defaultRasterSampleCount = static_cast<NSUInteger>(rp->sample_count); @@ -223,8 +226,9 @@ void MDCommandBuffer::render_bind_uniform_set(RDD::UniformSetID p_uniform_set, R void MDCommandBuffer::render_clear_attachments(VectorView<RDD::AttachmentClear> p_attachment_clears, VectorView<Rect2i> p_rects) { DEV_ASSERT(type == MDCommandBufferStateType::Render); - uint32_t vertex_count = p_rects.size() * 6; + const MDSubpass &subpass = render.get_subpass(); + uint32_t vertex_count = p_rects.size() * 6 * subpass.view_count; simd::float4 vertices[vertex_count]; simd::float4 clear_colors[ClearAttKey::ATTACHMENT_COUNT]; @@ -235,6 +239,9 @@ void MDCommandBuffer::render_clear_attachments(VectorView<RDD::AttachmentClear> ClearAttKey key; key.sample_count = render.pass->get_sample_count(); + if (subpass.view_count > 1) { + key.enable_layered_rendering(); + } float depth_value = 0; uint32_t stencil_value = 0; @@ -245,7 +252,7 @@ void MDCommandBuffer::render_clear_attachments(VectorView<RDD::AttachmentClear> if (attClear.aspect.has_flag(RDD::TEXTURE_ASPECT_COLOR_BIT)) { attachment_index = attClear.color_attachment; } else { - attachment_index = render.pass->subpasses[render.current_subpass].depth_stencil_reference.attachment; + attachment_index = subpass.depth_stencil_reference.attachment; } MDAttachment const &mda = render.pass->attachments[attachment_index]; @@ -310,6 +317,13 @@ void MDCommandBuffer::render_clear_attachments(VectorView<RDD::AttachmentClear> void MDCommandBuffer::_render_set_dirty_state() { _render_bind_uniform_sets(); + MDSubpass const &subpass = render.get_subpass(); + if (subpass.view_count > 1) { + uint32_t view_range[2] = { 0, subpass.view_count }; + [render.encoder setVertexBytes:view_range length:sizeof(view_range) atIndex:VIEW_MASK_BUFFER_INDEX]; + [render.encoder setFragmentBytes:view_range length:sizeof(view_range) atIndex:VIEW_MASK_BUFFER_INDEX]; + } + if (render.dirty.has_flag(RenderState::DIRTY_PIPELINE)) { [render.encoder setRenderPipelineState:render.pipeline->state]; } @@ -390,6 +404,38 @@ void MDCommandBuffer::render_set_blend_constants(const Color &p_constants) { } } +void BoundUniformSet::merge_into(ResourceUsageMap &p_dst) const { + for (KeyValue<StageResourceUsage, ResourceVector> const &keyval : usage_to_resources) { + ResourceVector *resources = p_dst.getptr(keyval.key); + if (resources == nullptr) { + resources = &p_dst.insert(keyval.key, ResourceVector())->value; + } + // Reserve space for the new resources, assuming they are all added. + resources->reserve(resources->size() + keyval.value.size()); + + uint32_t i = 0, j = 0; + __unsafe_unretained id<MTLResource> *resources_ptr = resources->ptr(); + const __unsafe_unretained id<MTLResource> *keyval_ptr = keyval.value.ptr(); + // 2-way merge. + while (i < resources->size() && j < keyval.value.size()) { + if (resources_ptr[i] < keyval_ptr[j]) { + i++; + } else if (resources_ptr[i] > keyval_ptr[j]) { + resources->insert(i, keyval_ptr[j]); + i++; + j++; + } else { + i++; + j++; + } + } + // Append the remaining resources. + for (; j < keyval.value.size(); j++) { + resources->push_back(keyval_ptr[j]); + } + } +} + void MDCommandBuffer::_render_bind_uniform_sets() { DEV_ASSERT(type == MDCommandBufferStateType::Render); if (!render.dirty.has_flag(RenderState::DIRTY_UNIFORMS)) { @@ -408,7 +454,7 @@ void MDCommandBuffer::_render_bind_uniform_sets() { // Find the index of the next set bit. int index = __builtin_ctzll(set_uniforms); // Clear the set bit. - set_uniforms &= ~(1ULL << index); + set_uniforms &= (set_uniforms - 1); MDUniformSet *set = render.uniform_sets[index]; if (set == nullptr || set->index >= (uint32_t)shader->sets.size()) { continue; @@ -416,17 +462,7 @@ void MDCommandBuffer::_render_bind_uniform_sets() { UniformSet const &set_info = shader->sets[set->index]; BoundUniformSet &bus = set->boundUniformSetForShader(shader, device); - - for (KeyValue<id<MTLResource>, StageResourceUsage> const &keyval : bus.bound_resources) { - MTLResourceUsage usage = resource_usage_for_stage(keyval.value, RDD::ShaderStage::SHADER_STAGE_VERTEX); - if (usage != 0) { - [enc useResource:keyval.key usage:usage stages:MTLRenderStageVertex]; - } - usage = resource_usage_for_stage(keyval.value, RDD::ShaderStage::SHADER_STAGE_FRAGMENT); - if (usage != 0) { - [enc useResource:keyval.key usage:usage stages:MTLRenderStageFragment]; - } - } + bus.merge_into(render.resource_usage); // Set the buffer for the vertex stage. { @@ -470,36 +506,40 @@ uint32_t MDCommandBuffer::_populate_vertices(simd::float4 *p_vertices, uint32_t simd::float4 vtx; uint32_t idx = p_index; - vtx.z = 0.0; - vtx.w = (float)1; + uint32_t endLayer = render.get_subpass().view_count; - // Top left vertex - First triangle. - vtx.y = topPos; - vtx.x = leftPos; - p_vertices[idx++] = vtx; + for (uint32_t layer = 0; layer < endLayer; layer++) { + vtx.z = 0.0; + vtx.w = (float)layer; - // Bottom left vertex. - vtx.y = bottomPos; - vtx.x = leftPos; - p_vertices[idx++] = vtx; + // Top left vertex - First triangle. + vtx.y = topPos; + vtx.x = leftPos; + p_vertices[idx++] = vtx; - // Bottom right vertex. - vtx.y = bottomPos; - vtx.x = rightPos; - p_vertices[idx++] = vtx; + // Bottom left vertex. + vtx.y = bottomPos; + vtx.x = leftPos; + p_vertices[idx++] = vtx; - // Bottom right vertex - Second triangle. - p_vertices[idx++] = vtx; + // Bottom right vertex. + vtx.y = bottomPos; + vtx.x = rightPos; + p_vertices[idx++] = vtx; - // Top right vertex. - vtx.y = topPos; - vtx.x = rightPos; - p_vertices[idx++] = vtx; + // Bottom right vertex - Second triangle. + p_vertices[idx++] = vtx; - // Top left vertex. - vtx.y = topPos; - vtx.x = leftPos; - p_vertices[idx++] = vtx; + // Top right vertex. + vtx.y = topPos; + vtx.x = rightPos; + p_vertices[idx++] = vtx; + + // Top left vertex. + vtx.y = topPos; + vtx.x = leftPos; + p_vertices[idx++] = vtx; + } return idx; } @@ -526,8 +566,7 @@ void MDCommandBuffer::render_begin_pass(RDD::RenderPassID p_render_pass, RDD::Fr void MDCommandBuffer::_end_render_pass() { MDFrameBuffer const &fb_info = *render.frameBuffer; - MDRenderPass const &pass_info = *render.pass; - MDSubpass const &subpass = pass_info.subpasses[render.current_subpass]; + MDSubpass const &subpass = render.get_subpass(); PixelFormats &pf = device_driver->get_pixel_formats(); @@ -535,23 +574,22 @@ void MDCommandBuffer::_end_render_pass() { uint32_t color_index = subpass.color_references[i].attachment; uint32_t resolve_index = subpass.resolve_references[i].attachment; DEV_ASSERT((color_index == RDD::AttachmentReference::UNUSED) == (resolve_index == RDD::AttachmentReference::UNUSED)); - if (color_index == RDD::AttachmentReference::UNUSED || !fb_info.textures[color_index]) { + if (color_index == RDD::AttachmentReference::UNUSED || !fb_info.has_texture(color_index)) { continue; } - id<MTLTexture> resolve_tex = fb_info.textures[resolve_index]; + id<MTLTexture> resolve_tex = fb_info.get_texture(resolve_index); CRASH_COND_MSG(!flags::all(pf.getCapabilities(resolve_tex.pixelFormat), kMTLFmtCapsResolve), "not implemented: unresolvable texture types"); // see: https://github.com/KhronosGroup/MoltenVK/blob/d20d13fe2735adb845636a81522df1b9d89c0fba/MoltenVK/MoltenVK/GPUObjects/MVKRenderPass.mm#L407 } - [render.encoder endEncoding]; - render.encoder = nil; + render.end_encoding(); } void MDCommandBuffer::_render_clear_render_area() { MDRenderPass const &pass = *render.pass; - MDSubpass const &subpass = pass.subpasses[render.current_subpass]; + MDSubpass const &subpass = render.get_subpass(); // First determine attachments that should be cleared. LocalVector<RDD::AttachmentClear> clears; @@ -598,9 +636,14 @@ void MDCommandBuffer::render_next_subpass() { MDFrameBuffer const &fb = *render.frameBuffer; MDRenderPass const &pass = *render.pass; - MDSubpass const &subpass = pass.subpasses[render.current_subpass]; + MDSubpass const &subpass = render.get_subpass(); MTLRenderPassDescriptor *desc = MTLRenderPassDescriptor.renderPassDescriptor; + + if (subpass.view_count > 1) { + desc.renderTargetArrayLength = subpass.view_count; + } + PixelFormats &pf = device_driver->get_pixel_formats(); uint32_t attachmentCount = 0; @@ -617,7 +660,7 @@ void MDCommandBuffer::render_next_subpass() { bool has_resolve = resolveIdx != RDD::AttachmentReference::UNUSED; bool can_resolve = true; if (resolveIdx != RDD::AttachmentReference::UNUSED) { - id<MTLTexture> resolve_tex = fb.textures[resolveIdx]; + id<MTLTexture> resolve_tex = fb.get_texture(resolveIdx); can_resolve = flags::all(pf.getCapabilities(resolve_tex.pixelFormat), kMTLFmtCapsResolve); if (can_resolve) { ca.resolveTexture = resolve_tex; @@ -628,7 +671,9 @@ void MDCommandBuffer::render_next_subpass() { MDAttachment const &attachment = pass.attachments[idx]; - id<MTLTexture> tex = fb.textures[idx]; + id<MTLTexture> tex = fb.get_texture(idx); + ERR_FAIL_NULL_MSG(tex, "Frame buffer color texture is null."); + if ((attachment.type & MDAttachmentType::Color)) { if (attachment.configureDescriptor(ca, pf, subpass, tex, render.is_rendering_entire_area, has_resolve, can_resolve, false)) { Color clearColor = render.clear_values[idx].color; @@ -641,7 +686,8 @@ void MDCommandBuffer::render_next_subpass() { attachmentCount += 1; uint32_t idx = subpass.depth_stencil_reference.attachment; MDAttachment const &attachment = pass.attachments[idx]; - id<MTLTexture> tex = fb.textures[idx]; + id<MTLTexture> tex = fb.get_texture(idx); + ERR_FAIL_NULL_MSG(tex, "Frame buffer depth / stencil texture is null."); if (attachment.type & MDAttachmentType::Depth) { MTLRenderPassDepthAttachmentDescriptor *da = desc.depthAttachment; if (attachment.configureDescriptor(da, pf, subpass, tex, render.is_rendering_entire_area, false, false, false)) { @@ -681,8 +727,15 @@ void MDCommandBuffer::render_draw(uint32_t p_vertex_count, uint32_t p_base_vertex, uint32_t p_first_instance) { DEV_ASSERT(type == MDCommandBufferStateType::Render); + ERR_FAIL_NULL_MSG(render.pipeline, "No pipeline set for render command buffer."); + _render_set_dirty_state(); + MDSubpass const &subpass = render.get_subpass(); + if (subpass.view_count > 1) { + p_instance_count *= subpass.view_count; + } + DEV_ASSERT(render.dirty == 0); id<MTLRenderCommandEncoder> enc = render.encoder; @@ -730,8 +783,15 @@ void MDCommandBuffer::render_draw_indexed(uint32_t p_index_count, int32_t p_vertex_offset, uint32_t p_first_instance) { DEV_ASSERT(type == MDCommandBufferStateType::Render); + ERR_FAIL_NULL_MSG(render.pipeline, "No pipeline set for render command buffer."); + _render_set_dirty_state(); + MDSubpass const &subpass = render.get_subpass(); + if (subpass.view_count > 1) { + p_instance_count *= subpass.view_count; + } + id<MTLRenderCommandEncoder> enc = render.encoder; uint32_t index_offset = render.index_offset; @@ -749,6 +809,8 @@ void MDCommandBuffer::render_draw_indexed(uint32_t p_index_count, void MDCommandBuffer::render_draw_indexed_indirect(RDD::BufferID p_indirect_buffer, uint64_t p_offset, uint32_t p_draw_count, uint32_t p_stride) { DEV_ASSERT(type == MDCommandBufferStateType::Render); + ERR_FAIL_NULL_MSG(render.pipeline, "No pipeline set for render command buffer."); + _render_set_dirty_state(); id<MTLRenderCommandEncoder> enc = render.encoder; @@ -773,6 +835,8 @@ void MDCommandBuffer::render_draw_indexed_indirect_count(RDD::BufferID p_indirec void MDCommandBuffer::render_draw_indirect(RDD::BufferID p_indirect_buffer, uint64_t p_offset, uint32_t p_draw_count, uint32_t p_stride) { DEV_ASSERT(type == MDCommandBufferStateType::Render); + ERR_FAIL_NULL_MSG(render.pipeline, "No pipeline set for render command buffer."); + _render_set_dirty_state(); id<MTLRenderCommandEncoder> enc = render.encoder; @@ -795,11 +859,90 @@ void MDCommandBuffer::render_draw_indirect_count(RDD::BufferID p_indirect_buffer void MDCommandBuffer::render_end_pass() { DEV_ASSERT(type == MDCommandBufferStateType::Render); - [render.encoder endEncoding]; + render.end_encoding(); render.reset(); type = MDCommandBufferStateType::None; } +#pragma mark - RenderState + +void MDCommandBuffer::RenderState::reset() { + pass = nil; + frameBuffer = nil; + pipeline = nil; + current_subpass = UINT32_MAX; + render_area = {}; + is_rendering_entire_area = false; + desc = nil; + encoder = nil; + index_buffer = nil; + index_type = MTLIndexTypeUInt16; + dirty = DIRTY_NONE; + uniform_sets.clear(); + uniform_set_mask = 0; + clear_values.clear(); + viewports.clear(); + scissors.clear(); + blend_constants.reset(); + vertex_buffers.clear(); + vertex_offsets.clear(); + // Keep the keys, as they are likely to be used again. + for (KeyValue<StageResourceUsage, LocalVector<__unsafe_unretained id<MTLResource>>> &kv : resource_usage) { + kv.value.clear(); + } +} + +void MDCommandBuffer::RenderState::end_encoding() { + if (encoder == nil) { + return; + } + + // Bind all resources. + for (KeyValue<StageResourceUsage, ResourceVector> const &keyval : resource_usage) { + if (keyval.value.is_empty()) { + continue; + } + + MTLResourceUsage vert_usage = resource_usage_for_stage(keyval.key, RDD::ShaderStage::SHADER_STAGE_VERTEX); + MTLResourceUsage frag_usage = resource_usage_for_stage(keyval.key, RDD::ShaderStage::SHADER_STAGE_FRAGMENT); + if (vert_usage == frag_usage) { + [encoder useResources:keyval.value.ptr() count:keyval.value.size() usage:vert_usage stages:MTLRenderStageVertex | MTLRenderStageFragment]; + } else { + if (vert_usage != 0) { + [encoder useResources:keyval.value.ptr() count:keyval.value.size() usage:vert_usage stages:MTLRenderStageVertex]; + } + if (frag_usage != 0) { + [encoder useResources:keyval.value.ptr() count:keyval.value.size() usage:frag_usage stages:MTLRenderStageFragment]; + } + } + } + + [encoder endEncoding]; + encoder = nil; +} + +#pragma mark - ComputeState + +void MDCommandBuffer::ComputeState::end_encoding() { + if (encoder == nil) { + return; + } + + // Bind all resources. + for (KeyValue<StageResourceUsage, ResourceVector> const &keyval : resource_usage) { + if (keyval.value.is_empty()) { + continue; + } + MTLResourceUsage usage = resource_usage_for_stage(keyval.key, RDD::ShaderStage::SHADER_STAGE_COMPUTE); + if (usage != 0) { + [encoder useResources:keyval.value.ptr() count:keyval.value.size() usage:usage]; + } + } + + [encoder endEncoding]; + encoder = nil; +} + #pragma mark - Compute void MDCommandBuffer::compute_bind_uniform_set(RDD::UniformSetID p_uniform_set, RDD::ShaderID p_shader, uint32_t p_set_index) { @@ -813,13 +956,7 @@ void MDCommandBuffer::compute_bind_uniform_set(RDD::UniformSetID p_uniform_set, MDUniformSet *set = (MDUniformSet *)(p_uniform_set.id); BoundUniformSet &bus = set->boundUniformSetForShader(shader, device); - - for (KeyValue<id<MTLResource>, StageResourceUsage> &keyval : bus.bound_resources) { - MTLResourceUsage usage = resource_usage_for_stage(keyval.value, RDD::ShaderStage::SHADER_STAGE_COMPUTE); - if (usage != 0) { - [enc useResource:keyval.key usage:usage]; - } - } + bus.merge_into(compute.resource_usage); uint32_t const *offset = set_info.offsets.getptr(RDD::SHADER_STAGE_COMPUTE); if (offset) { @@ -848,7 +985,7 @@ void MDCommandBuffer::compute_dispatch_indirect(RDD::BufferID p_indirect_buffer, void MDCommandBuffer::_end_compute_dispatch() { DEV_ASSERT(type == MDCommandBufferStateType::Compute); - [compute.encoder endEncoding]; + compute.end_encoding(); compute.reset(); type = MDCommandBufferStateType::None; } @@ -879,8 +1016,11 @@ void MDComputeShader::encode_push_constant_data(VectorView<uint32_t> p_data, MDC [enc setBytes:ptr length:length atIndex:push_constants.binding]; } -MDRenderShader::MDRenderShader(CharString p_name, Vector<UniformSet> p_sets, MDLibrary *_Nonnull p_vert, MDLibrary *_Nonnull p_frag) : - MDShader(p_name, p_sets), vert(p_vert), frag(p_frag) { +MDRenderShader::MDRenderShader(CharString p_name, + bool p_needs_view_mask_buffer, + Vector<UniformSet> p_sets, + MDLibrary *_Nonnull p_vert, MDLibrary *_Nonnull p_frag) : + MDShader(p_name, p_sets), needs_view_mask_buffer(p_needs_view_mask_buffer), vert(p_vert), frag(p_frag) { } void MDRenderShader::encode_push_constant_data(VectorView<uint32_t> p_data, MDCommandBuffer *p_cb) { @@ -1052,7 +1192,20 @@ BoundUniformSet &MDUniformSet::boundUniformSetForShader(MDShader *p_shader, id<M } } - BoundUniformSet bs = { .buffer = enc_buffer, .bound_resources = bound_resources }; + SearchArray<__unsafe_unretained id<MTLResource>> search; + ResourceUsageMap usage_to_resources; + for (KeyValue<id<MTLResource>, StageResourceUsage> const &keyval : bound_resources) { + ResourceVector *resources = usage_to_resources.getptr(keyval.value); + if (resources == nullptr) { + resources = &usage_to_resources.insert(keyval.value, ResourceVector())->value; + } + int64_t pos = search.bisect(resources->ptr(), resources->size(), keyval.key, true); + if (pos == resources->size() || (*resources)[pos] != keyval.key) { + resources->insert(pos, keyval.key); + } + } + + BoundUniformSet bs = { .buffer = enc_buffer, .usage_to_resources = usage_to_resources }; bound_uniforms.insert(p_shader, bs); return bound_uniforms.get(p_shader); } @@ -1202,7 +1355,7 @@ typedef struct { typedef struct { float4 v_position [[position]]; - uint layer; + uint layer%s; } VaryingsPos; vertex VaryingsPos vertClear(AttributesPos attributes [[stage_in]], constant ClearColorsIn& ccIn [[buffer(0)]]) { @@ -1211,8 +1364,7 @@ vertex VaryingsPos vertClear(AttributesPos attributes [[stage_in]], constant Cle varyings.layer = uint(attributes.a_position.w); return varyings; } -)", - ClearAttKey::DEPTH_INDEX]; +)", p_key.is_layered_rendering_enabled() ? " [[render_target_array_index]]" : "", ClearAttKey::DEPTH_INDEX]; return new_func(msl, @"vertClear", nil); } @@ -1502,7 +1654,7 @@ void ShaderCacheEntry::notify_free() const { self->_library = library; self->_error = error; if (error) { - ERR_PRINT(String(U"Error compiling shader %s: %s").format(entry->name.get_data(), error.localizedDescription.UTF8String)); + ERR_PRINT(vformat(U"Error compiling shader %s: %s", entry->name.get_data(), error.localizedDescription.UTF8String)); } { diff --git a/drivers/metal/rendering_context_driver_metal.h b/drivers/metal/rendering_context_driver_metal.h index 7e0b09186d..1fdc255f93 100644 --- a/drivers/metal/rendering_context_driver_metal.h +++ b/drivers/metal/rendering_context_driver_metal.h @@ -107,6 +107,7 @@ public: uint32_t height = 0; DisplayServer::VSyncMode vsync_mode = DisplayServer::VSYNC_ENABLED; bool needs_resize = false; + double present_minimum_duration = 0.0; Surface( #ifdef __OBJC__ @@ -123,6 +124,7 @@ public: virtual Error resize(uint32_t p_desired_framebuffer_count) = 0; virtual RDD::FramebufferID acquire_next_frame_buffer() = 0; virtual void present(MDCommandBuffer *p_cmd_buffer) = 0; + void set_max_fps(int p_max_fps) { present_minimum_duration = p_max_fps ? 1.0 / p_max_fps : 0.0; } }; #ifdef __OBJC__ diff --git a/drivers/metal/rendering_context_driver_metal.mm b/drivers/metal/rendering_context_driver_metal.mm index b97b586352..199ec25d79 100644 --- a/drivers/metal/rendering_context_driver_metal.mm +++ b/drivers/metal/rendering_context_driver_metal.mm @@ -134,7 +134,7 @@ public: frame_buffers.resize(p_desired_framebuffer_count); for (uint32_t i = 0; i < p_desired_framebuffer_count; i++) { // Reserve space for the drawable texture. - frame_buffers[i].textures.resize(1); + frame_buffers[i].set_texture_count(1); } return OK; @@ -154,7 +154,7 @@ public: id<CAMetalDrawable> drawable = layer.nextDrawable; ERR_FAIL_NULL_V_MSG(drawable, RDD::FramebufferID(), "no drawable available"); drawables[rear] = drawable; - frame_buffer.textures.write[0] = drawable.texture; + frame_buffer.set_texture(0, drawable.texture); return RDD::FramebufferID(&frame_buffer); } @@ -165,14 +165,14 @@ public: } // Release texture and drawable. - frame_buffers[front].textures.write[0] = nil; + frame_buffers[front].unset_texture(0); id<MTLDrawable> drawable = drawables[front]; drawables[front] = nil; count--; front = (front + 1) % frame_buffers.size(); - [p_cmd_buffer->get_command_buffer() presentDrawable:drawable]; + [p_cmd_buffer->get_command_buffer() presentDrawable:drawable afterMinimumDuration:present_minimum_duration]; } }; diff --git a/drivers/metal/rendering_device_driver_metal.h b/drivers/metal/rendering_device_driver_metal.h index 7c23624e43..09ab7601e9 100644 --- a/drivers/metal/rendering_device_driver_metal.h +++ b/drivers/metal/rendering_device_driver_metal.h @@ -220,6 +220,7 @@ public: virtual FramebufferID swap_chain_acquire_framebuffer(CommandQueueID p_cmd_queue, SwapChainID p_swap_chain, bool &r_resize_required) override final; virtual RenderPassID swap_chain_get_render_pass(SwapChainID p_swap_chain) override final; virtual DataFormat swap_chain_get_format(SwapChainID p_swap_chain) override final; + virtual void swap_chain_set_max_fps(SwapChainID p_swap_chain, int p_max_fps) override final; virtual void swap_chain_free(SwapChainID p_swap_chain) override final; #pragma mark - Frame Buffer @@ -239,7 +240,13 @@ private: friend struct PushConstantData; private: - Error _reflect_spirv16(VectorView<ShaderStageSPIRVData> p_spirv, ShaderReflection &r_reflection); + /// Contains additional metadata about the shader. + struct ShaderMeta { + /// Indicates whether the shader uses multiview. + bool has_multiview = false; + }; + + Error _reflect_spirv16(VectorView<ShaderStageSPIRVData> p_spirv, ShaderReflection &r_reflection, ShaderMeta &r_shader_meta); public: virtual String shader_get_binary_cache_key() override final; @@ -410,7 +417,7 @@ public: virtual uint64_t api_trait_get(ApiTrait p_trait) override final; virtual bool has_feature(Features p_feature) override final; virtual const MultiviewCapabilities &get_multiview_capabilities() override final; - virtual String get_api_name() const override final { return "Metal"; }; + virtual String get_api_name() const override final { return "Metal"; } virtual String get_api_version() const override final; virtual String get_pipeline_cache_uuid() const override final; virtual const Capabilities &get_capabilities() const override final; diff --git a/drivers/metal/rendering_device_driver_metal.mm b/drivers/metal/rendering_device_driver_metal.mm index a4a408356a..784c9d5ae8 100644 --- a/drivers/metal/rendering_device_driver_metal.mm +++ b/drivers/metal/rendering_device_driver_metal.mm @@ -982,6 +982,12 @@ RDD::DataFormat RenderingDeviceDriverMetal::swap_chain_get_format(SwapChainID p_ return swap_chain->data_format; } +void RenderingDeviceDriverMetal::swap_chain_set_max_fps(SwapChainID p_swap_chain, int p_max_fps) { + SwapChain *swap_chain = (SwapChain *)(p_swap_chain.id); + RenderingContextDriverMetal::Surface *metal_surface = (RenderingContextDriverMetal::Surface *)(swap_chain->surface); + metal_surface->set_max_fps(p_max_fps); +} + void RenderingDeviceDriverMetal::swap_chain_free(SwapChainID p_swap_chain) { SwapChain *swap_chain = (SwapChain *)(p_swap_chain.id); _swap_chain_release(swap_chain); @@ -1026,7 +1032,7 @@ void RenderingDeviceDriverMetal::framebuffer_free(FramebufferID p_framebuffer) { #pragma mark - Shader -const uint32_t SHADER_BINARY_VERSION = 1; +const uint32_t SHADER_BINARY_VERSION = 3; // region Serialization @@ -1330,23 +1336,32 @@ struct ComputeSize { struct ShaderStageData { RD::ShaderStage stage = RD::ShaderStage::SHADER_STAGE_MAX; + uint32_t is_position_invariant = UINT32_MAX; + uint32_t supports_fast_math = UINT32_MAX; CharString entry_point_name; CharString source; size_t serialize_size() const { int comp_size = Compression::get_max_compressed_buffer_size(source.length(), Compression::MODE_ZSTD); return sizeof(uint32_t) // Stage. - + sizeof(uint32_t) /* entry_point_name.utf8().length */ + entry_point_name.length() + sizeof(uint32_t) /* uncompressed size */ + sizeof(uint32_t) /* compressed size */ + comp_size; + + sizeof(uint32_t) // is_position_invariant + + sizeof(uint32_t) // supports_fast_math + + sizeof(uint32_t) /* entry_point_name.utf8().length */ + + entry_point_name.length() + sizeof(uint32_t) /* uncompressed size */ + sizeof(uint32_t) /* compressed size */ + comp_size; } void serialize(BufWriter &p_writer) const { p_writer.write((uint32_t)stage); + p_writer.write(is_position_invariant); + p_writer.write(supports_fast_math); p_writer.write(entry_point_name); p_writer.write_compressed(source); } void deserialize(BufReader &p_reader) { p_reader.read((uint32_t &)stage); + p_reader.read(is_position_invariant); + p_reader.read(supports_fast_math); p_reader.read(entry_point_name); p_reader.read_compressed(source); } @@ -1503,6 +1518,7 @@ struct API_AVAILABLE(macos(11.0), ios(14.0)) ShaderBinaryData { uint32_t fragment_output_mask = UINT32_MAX; uint32_t spirv_specialization_constants_ids_mask = UINT32_MAX; uint32_t is_compute = UINT32_MAX; + uint32_t needs_view_mask_buffer = UINT32_MAX; ComputeSize compute_local_size; PushConstantData push_constant; LocalVector<ShaderStageData> stages; @@ -1523,6 +1539,7 @@ struct API_AVAILABLE(macos(11.0), ios(14.0)) ShaderBinaryData { size += sizeof(uint32_t); // fragment_output_mask size += sizeof(uint32_t); // spirv_specialization_constants_ids_mask size += sizeof(uint32_t); // is_compute + size += sizeof(uint32_t); // needs_view_mask_buffer size += compute_local_size.serialize_size(); // compute_local_size size += push_constant.serialize_size(); // push_constant size += sizeof(uint32_t); // stages.size() @@ -1547,6 +1564,7 @@ struct API_AVAILABLE(macos(11.0), ios(14.0)) ShaderBinaryData { p_writer.write(fragment_output_mask); p_writer.write(spirv_specialization_constants_ids_mask); p_writer.write(is_compute); + p_writer.write(needs_view_mask_buffer); p_writer.write(compute_local_size); p_writer.write(push_constant); p_writer.write(VectorView(stages)); @@ -1561,6 +1579,7 @@ struct API_AVAILABLE(macos(11.0), ios(14.0)) ShaderBinaryData { p_reader.read(fragment_output_mask); p_reader.read(spirv_specialization_constants_ids_mask); p_reader.read(is_compute); + p_reader.read(needs_view_mask_buffer); p_reader.read(compute_local_size); p_reader.read(push_constant); p_reader.read(stages); @@ -1572,14 +1591,16 @@ struct API_AVAILABLE(macos(11.0), ios(14.0)) ShaderBinaryData { // endregion String RenderingDeviceDriverMetal::shader_get_binary_cache_key() { - return "Metal-SV" + uitos(SHADER_BINARY_VERSION); + static const String cache_key = "Metal-SV" + uitos(SHADER_BINARY_VERSION); + return cache_key; } -Error RenderingDeviceDriverMetal::_reflect_spirv16(VectorView<ShaderStageSPIRVData> p_spirv, ShaderReflection &r_reflection) { +Error RenderingDeviceDriverMetal::_reflect_spirv16(VectorView<ShaderStageSPIRVData> p_spirv, ShaderReflection &r_reflection, ShaderMeta &r_shader_meta) { using namespace spirv_cross; using spirv_cross::Resource; r_reflection = {}; + r_shader_meta = {}; for (uint32_t i = 0; i < p_spirv.size(); i++) { ShaderStageSPIRVData const &v = p_spirv[i]; @@ -1811,6 +1832,20 @@ Error RenderingDeviceDriverMetal::_reflect_spirv16(VectorView<ShaderStageSPIRVDa } } + for (const BuiltInResource &res : resources.builtin_inputs) { + if (res.builtin == spv::BuiltInViewIndex || res.builtin == spv::BuiltInViewportIndex) { + r_shader_meta.has_multiview = true; + } + } + + if (!r_shader_meta.has_multiview) { + for (const BuiltInResource &res : resources.builtin_outputs) { + if (res.builtin == spv::BuiltInViewIndex || res.builtin == spv::BuiltInViewportIndex) { + r_shader_meta.has_multiview = true; + } + } + } + // Specialization constants. for (SpecializationConstant const &constant : compiler.get_specialization_constants()) { int32_t existing = -1; @@ -1874,7 +1909,8 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec using spirv_cross::Resource; ShaderReflection spirv_data; - ERR_FAIL_COND_V(_reflect_spirv16(p_spirv, spirv_data), Result()); + ShaderMeta shader_meta; + ERR_FAIL_COND_V(_reflect_spirv16(p_spirv, spirv_data, shader_meta), Result()); ShaderBinaryData bin_data{}; if (!p_shader_name.is_empty()) { @@ -1893,6 +1929,7 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec bin_data.is_compute = spirv_data.is_compute; bin_data.push_constant.size = spirv_data.push_constant_size; bin_data.push_constant.stages = (ShaderStageUsage)(uint8_t)spirv_data.push_constant_stages; + bin_data.needs_view_mask_buffer = shader_meta.has_multiview ? 1 : 0; for (uint32_t i = 0; i < spirv_data.uniform_sets.size(); i++) { const ::Vector<ShaderUniform> &spirv_set = spirv_data.uniform_sets[i]; @@ -1947,6 +1984,11 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec msl_options.pad_fragment_output_components = true; msl_options.r32ui_alignment_constant_id = R32UI_ALIGNMENT_CONSTANT_ID; msl_options.agx_manual_cube_grad_fixup = true; + if (shader_meta.has_multiview) { + msl_options.multiview = true; + msl_options.multiview_layered_rendering = true; + msl_options.view_mask_buffer_index = VIEW_MASK_BUFFER_INDEX; + } CompilerGLSL::Options options{}; options.vertex.flip_vert_y = true; @@ -1978,7 +2020,8 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec ERR_FAIL_COND_V_MSG(compiler.get_entry_points_and_stages().size() != 1, Result(), "Expected a single entry point and stage."); - EntryPoint &entry_point_stage = compiler.get_entry_points_and_stages().front(); + SmallVector<EntryPoint> entry_pts_stages = compiler.get_entry_points_and_stages(); + EntryPoint &entry_point_stage = entry_pts_stages.front(); SPIREntryPoint &entry_point = compiler.get_entry_point(entry_point_stage.name, entry_point_stage.execution_model); // Process specialization constants. @@ -2060,6 +2103,10 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec case BT::Sampler: { primary.dataType = MTLDataTypeSampler; + primary.arrayLength = 1; + for (uint32_t const &a : a_type.array) { + primary.arrayLength *= a; + } } break; default: { @@ -2067,7 +2114,7 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec } break; } - // Find array length. + // Find array length of image. if (basetype == BT::Image || basetype == BT::SampledImage) { primary.arrayLength = 1; for (uint32_t const &a : a_type.array) { @@ -2256,6 +2303,8 @@ Vector<uint8_t> RenderingDeviceDriverMetal::shader_compile_binary_from_spirv(Vec ShaderStageData stage_data; stage_data.stage = v.shader_stage; + stage_data.is_position_invariant = compiler.is_position_invariant(); + stage_data.supports_fast_math = !entry_point.flags.get(spv::ExecutionModeSignedZeroInfNanPreserve); stage_data.entry_point_name = entry_point.name.c_str(); stage_data.source = source.c_str(); bin_data.stages.push_back(stage_data); @@ -2328,7 +2377,8 @@ RDD::ShaderID RenderingDeviceDriverMetal::shader_create_from_bytecode(const Vect ShaderCacheEntry *cd = memnew(ShaderCacheEntry(*this, key)); cd->name = binary_data.shader_name; cd->stage = shader_data.stage; - + options.preserveInvariance = shader_data.is_position_invariant; + options.fastMathEnabled = YES; MDLibrary *library = [MDLibrary newLibraryWithCacheEntry:cd device:device source:source @@ -2444,7 +2494,7 @@ RDD::ShaderID RenderingDeviceDriverMetal::shader_create_from_bytecode(const Vect #endif shader = cs; } else { - MDRenderShader *rs = new MDRenderShader(binary_data.shader_name, uniform_sets, libraries[ShaderStage::SHADER_STAGE_VERTEX], libraries[ShaderStage::SHADER_STAGE_FRAGMENT]); + MDRenderShader *rs = new MDRenderShader(binary_data.shader_name, (bool)binary_data.needs_view_mask_buffer, uniform_sets, libraries[ShaderStage::SHADER_STAGE_VERTEX], libraries[ShaderStage::SHADER_STAGE_FRAGMENT]); uint32_t *vert_binding = binary_data.push_constant.msl_binding.getptr(SHADER_STAGE_VERTEX); if (vert_binding) { @@ -2952,6 +3002,7 @@ RDD::RenderPassID RenderingDeviceDriverMetal::render_pass_create(VectorView<Atta for (uint32_t i = 0; i < subpass_count; i++) { MDSubpass &subpass = subpasses.write[i]; subpass.subpass_index = i; + subpass.view_count = p_view_count; subpass.input_references = p_subpasses[i].input_references; subpass.color_references = p_subpasses[i].color_references; subpass.depth_stencil_reference = p_subpasses[i].depth_stencil_reference; @@ -3671,8 +3722,7 @@ void RenderingDeviceDriverMetal::set_object_name(ObjectType p_type, ID p_driver_ uint64_t RenderingDeviceDriverMetal::get_resource_native_handle(DriverResource p_type, ID p_driver_id) { switch (p_type) { case DRIVER_RESOURCE_LOGICAL_DEVICE: { - uintptr_t devicePtr = (uintptr_t)(__bridge void *)device; - return (uint64_t)devicePtr; + return (uint64_t)(uintptr_t)(__bridge void *)device; } case DRIVER_RESOURCE_PHYSICAL_DEVICE: { return 0; @@ -3681,7 +3731,7 @@ uint64_t RenderingDeviceDriverMetal::get_resource_native_handle(DriverResource p return 0; } case DRIVER_RESOURCE_COMMAND_QUEUE: { - return 0; + return (uint64_t)(uintptr_t)(__bridge void *)device_queue; } case DRIVER_RESOURCE_QUEUE_FAMILY: { return 0; @@ -3698,15 +3748,20 @@ uint64_t RenderingDeviceDriverMetal::get_resource_native_handle(DriverResource p case DRIVER_RESOURCE_SAMPLER: { return p_driver_id.id; } - case DRIVER_RESOURCE_UNIFORM_SET: + case DRIVER_RESOURCE_UNIFORM_SET: { return 0; + } case DRIVER_RESOURCE_BUFFER: { return p_driver_id.id; } - case DRIVER_RESOURCE_COMPUTE_PIPELINE: - return 0; - case DRIVER_RESOURCE_RENDER_PIPELINE: - return 0; + case DRIVER_RESOURCE_COMPUTE_PIPELINE: { + MDComputePipeline *pipeline = (MDComputePipeline *)(p_driver_id.id); + return (uint64_t)(uintptr_t)(__bridge void *)pipeline->state; + } + case DRIVER_RESOURCE_RENDER_PIPELINE: { + MDRenderPipeline *pipeline = (MDRenderPipeline *)(p_driver_id.id); + return (uint64_t)(uintptr_t)(__bridge void *)pipeline->state; + } default: { return 0; } @@ -3838,7 +3893,7 @@ uint64_t RenderingDeviceDriverMetal::api_trait_get(ApiTrait p_trait) { bool RenderingDeviceDriverMetal::has_feature(Features p_feature) { switch (p_feature) { case SUPPORTS_MULTIVIEW: - return false; + return multiview_capabilities.is_supported; case SUPPORTS_FSR_HALF_FLOAT: return true; case SUPPORTS_ATTACHMENT_VRS: @@ -3947,6 +4002,18 @@ Error RenderingDeviceDriverMetal::initialize(uint32_t p_device_index, uint32_t p metal_device_properties = memnew(MetalDeviceProperties(device)); pixel_formats = memnew(PixelFormats(device)); + if (metal_device_properties->features.layeredRendering) { + multiview_capabilities.is_supported = true; + multiview_capabilities.max_view_count = metal_device_properties->limits.maxViewports; + // NOTE: I'm not sure what the limit is as I don't see it referenced anywhere + multiview_capabilities.max_instance_count = UINT32_MAX; + + print_verbose("- Metal multiview supported:"); + print_verbose(" max view count: " + itos(multiview_capabilities.max_view_count)); + print_verbose(" max instances: " + itos(multiview_capabilities.max_instance_count)); + } else { + print_verbose("- Metal multiview not supported"); + } // Check required features and abort if any of them is missing. if (!metal_device_properties->features.imageCubeArray) { diff --git a/drivers/png/SCsub b/drivers/png/SCsub index fce37257b1..4268a66475 100644 --- a/drivers/png/SCsub +++ b/drivers/png/SCsub @@ -46,18 +46,18 @@ if env["builtin_libpng"]: if "S_compiler" in env: env_neon["CC"] = env["S_compiler"] neon_sources = [] - neon_sources.append(env_neon.Object(thirdparty_dir + "/arm/arm_init.c")) - neon_sources.append(env_neon.Object(thirdparty_dir + "/arm/filter_neon_intrinsics.c")) - neon_sources.append(env_neon.Object(thirdparty_dir + "/arm/filter_neon.S")) - neon_sources.append(env_neon.Object(thirdparty_dir + "/arm/palette_neon_intrinsics.c")) + neon_sources.append(env_neon.Object(thirdparty_dir + "arm/arm_init.c")) + neon_sources.append(env_neon.Object(thirdparty_dir + "arm/filter_neon_intrinsics.c")) + neon_sources.append(env_neon.Object(thirdparty_dir + "arm/filter_neon.S")) + neon_sources.append(env_neon.Object(thirdparty_dir + "arm/palette_neon_intrinsics.c")) thirdparty_obj += neon_sources elif env["arch"].startswith("x86"): env_thirdparty.Append(CPPDEFINES=["PNG_INTEL_SSE"]) - env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "/intel/intel_init.c") - env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "/intel/filter_sse2_intrinsics.c") + env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "intel/intel_init.c") + env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "intel/filter_sse2_intrinsics.c") elif env["arch"] == "ppc64": - env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "/powerpc/powerpc_init.c") - env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "/powerpc/filter_vsx_intrinsics.c") + env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "powerpc/powerpc_init.c") + env_thirdparty.add_source_files(thirdparty_obj, thirdparty_dir + "powerpc/filter_vsx_intrinsics.c") env.drivers_sources += thirdparty_obj diff --git a/drivers/pulseaudio/audio_driver_pulseaudio.h b/drivers/pulseaudio/audio_driver_pulseaudio.h index 585e882059..6456467cb9 100644 --- a/drivers/pulseaudio/audio_driver_pulseaudio.h +++ b/drivers/pulseaudio/audio_driver_pulseaudio.h @@ -100,7 +100,7 @@ class AudioDriverPulseAudio : public AudioDriver { public: virtual const char *get_name() const override { return "PulseAudio"; - }; + } virtual Error init() override; virtual void start() override; diff --git a/drivers/unix/file_access_unix.cpp b/drivers/unix/file_access_unix.cpp index 3d584341ed..4b92e5f8a1 100644 --- a/drivers/unix/file_access_unix.cpp +++ b/drivers/unix/file_access_unix.cpp @@ -41,6 +41,11 @@ #include <sys/types.h> #include <unistd.h> +#if defined(TOOLS_ENABLED) +#include <limits.h> +#include <stdlib.h> +#endif + void FileAccessUnix::check_errors() const { ERR_FAIL_NULL_MSG(f, "File must be opened before use."); @@ -87,6 +92,22 @@ Error FileAccessUnix::open_internal(const String &p_path, int p_mode_flags) { } } +#if defined(TOOLS_ENABLED) + if (p_mode_flags & READ) { + String real_path = get_real_path(); + if (real_path != path) { + // Don't warn on symlinks, since they can be used to simply share addons on multiple projects. + if (real_path.to_lower() == path.to_lower()) { + // The File system is case insensitive, but other platforms can be sensitive to it + // To ease cross-platform development, we issue a warning if users try to access + // a file using the wrong case (which *works* on Windows and macOS, but won't on other + // platforms). + WARN_PRINT(vformat("Case mismatch opening requested file '%s', stored as '%s' in the filesystem. This file will not open when exported to other case-sensitive platforms.", path, real_path)); + } + } + } +#endif + if (is_backup_save_enabled() && (p_mode_flags == WRITE)) { save_path = path; // Create a temporary file in the same directory as the target file. @@ -97,7 +118,7 @@ Error FileAccessUnix::open_internal(const String &p_path, int p_mode_flags) { last_error = ERR_FILE_CANT_OPEN; return last_error; } - fchmod(fd, 0666); + fchmod(fd, 0644); path = String::utf8(cs.ptr()); f = fdopen(fd, mode_string); @@ -173,6 +194,26 @@ String FileAccessUnix::get_path_absolute() const { return path; } +#if defined(TOOLS_ENABLED) +String FileAccessUnix::get_real_path() const { + char *resolved_path = ::realpath(path.utf8().get_data(), nullptr); + + if (!resolved_path) { + return path; + } + + String result; + Error parse_ok = result.parse_utf8(resolved_path); + ::free(resolved_path); + + if (parse_ok != OK) { + return path; + } + + return result.simplify_path(); +} +#endif + void FileAccessUnix::seek(uint64_t p_position) { ERR_FAIL_NULL_MSG(f, "File must be opened before use."); diff --git a/drivers/unix/file_access_unix.h b/drivers/unix/file_access_unix.h index 76f629f7c2..7caf8a14d7 100644 --- a/drivers/unix/file_access_unix.h +++ b/drivers/unix/file_access_unix.h @@ -51,6 +51,10 @@ class FileAccessUnix : public FileAccess { void _close(); +#if defined(TOOLS_ENABLED) + String get_real_path() const; // Returns the resolved real path for the current open file. +#endif + public: static CloseNotificationFunc close_notification_func; diff --git a/drivers/unix/file_access_unix_pipe.cpp b/drivers/unix/file_access_unix_pipe.cpp index 0a78429dec..fd60bec9d0 100644 --- a/drivers/unix/file_access_unix_pipe.cpp +++ b/drivers/unix/file_access_unix_pipe.cpp @@ -70,7 +70,7 @@ Error FileAccessUnixPipe::open_internal(const String &p_path, int p_mode_flags) struct stat st = {}; int err = stat(path.utf8().get_data(), &st); if (err) { - if (mkfifo(path.utf8().get_data(), 0666) != 0) { + if (mkfifo(path.utf8().get_data(), 0600) != 0) { last_error = ERR_FILE_CANT_OPEN; return last_error; } diff --git a/drivers/unix/ip_unix.cpp b/drivers/unix/ip_unix.cpp index 5987c1675d..5a1a26e874 100644 --- a/drivers/unix/ip_unix.cpp +++ b/drivers/unix/ip_unix.cpp @@ -28,22 +28,9 @@ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /**************************************************************************/ -#include "ip_unix.h" - -#if defined(UNIX_ENABLED) || defined(WINDOWS_ENABLED) - -#ifdef WINDOWS_ENABLED - -#define WIN32_LEAN_AND_MEAN -#include <windows.h> -#include <winsock2.h> -#include <ws2tcpip.h> - -#include <iphlpapi.h> - -#include <stdio.h> +#if defined(UNIX_ENABLED) -#else // UNIX +#include "ip_unix.h" #include <netdb.h> @@ -67,8 +54,6 @@ #include <net/if.h> // Order is important on OpenBSD, leave as last. -#endif // UNIX - #include <string.h> static IPAddress _sockaddr2ip(struct sockaddr *p_addr) { @@ -108,7 +93,7 @@ void IPUnix::_resolve_hostname(List<IPAddress> &r_addresses, const String &p_hos } if (result == nullptr || result->ai_addr == nullptr) { - print_verbose("Invalid response from getaddrinfo"); + print_verbose("Invalid response from getaddrinfo."); if (result) { freeaddrinfo(result); } @@ -132,56 +117,6 @@ void IPUnix::_resolve_hostname(List<IPAddress> &r_addresses, const String &p_hos freeaddrinfo(result); } -#if defined(WINDOWS_ENABLED) - -void IPUnix::get_local_interfaces(HashMap<String, Interface_Info> *r_interfaces) const { - ULONG buf_size = 1024; - IP_ADAPTER_ADDRESSES *addrs; - - while (true) { - addrs = (IP_ADAPTER_ADDRESSES *)memalloc(buf_size); - int err = GetAdaptersAddresses(AF_UNSPEC, GAA_FLAG_SKIP_ANYCAST | GAA_FLAG_SKIP_MULTICAST | GAA_FLAG_SKIP_DNS_SERVER | GAA_FLAG_SKIP_FRIENDLY_NAME, - nullptr, addrs, &buf_size); - if (err == NO_ERROR) { - break; - } - memfree(addrs); - if (err == ERROR_BUFFER_OVERFLOW) { - continue; // will go back and alloc the right size - } - - ERR_FAIL_MSG("Call to GetAdaptersAddresses failed with error " + itos(err) + "."); - } - - IP_ADAPTER_ADDRESSES *adapter = addrs; - - while (adapter != nullptr) { - Interface_Info info; - info.name = adapter->AdapterName; - info.name_friendly = adapter->FriendlyName; - info.index = String::num_uint64(adapter->IfIndex); - - IP_ADAPTER_UNICAST_ADDRESS *address = adapter->FirstUnicastAddress; - while (address != nullptr) { - int family = address->Address.lpSockaddr->sa_family; - if (family != AF_INET && family != AF_INET6) { - continue; - } - info.ip_addresses.push_front(_sockaddr2ip(address->Address.lpSockaddr)); - address = address->Next; - } - adapter = adapter->Next; - // Only add interface if it has at least one IP - if (info.ip_addresses.size() > 0) { - r_interfaces->insert(info.name, info); - } - } - - memfree(addrs); -} - -#else // UNIX - void IPUnix::get_local_interfaces(HashMap<String, Interface_Info> *r_interfaces) const { struct ifaddrs *ifAddrStruct = nullptr; struct ifaddrs *ifa = nullptr; @@ -219,8 +154,6 @@ void IPUnix::get_local_interfaces(HashMap<String, Interface_Info> *r_interfaces) } } -#endif // UNIX - void IPUnix::make_default() { _create = _create_unix; } @@ -232,4 +165,4 @@ IP *IPUnix::_create_unix() { IPUnix::IPUnix() { } -#endif // UNIX_ENABLED || WINDOWS_ENABLED +#endif // UNIX_ENABLED diff --git a/drivers/unix/ip_unix.h b/drivers/unix/ip_unix.h index 274b7c561e..7e496629ef 100644 --- a/drivers/unix/ip_unix.h +++ b/drivers/unix/ip_unix.h @@ -31,9 +31,9 @@ #ifndef IP_UNIX_H #define IP_UNIX_H -#include "core/io/ip.h" +#if defined(UNIX_ENABLED) -#if defined(UNIX_ENABLED) || defined(WINDOWS_ENABLED) +#include "core/io/ip.h" class IPUnix : public IP { GDCLASS(IPUnix, IP); @@ -49,6 +49,6 @@ public: IPUnix(); }; -#endif +#endif // UNIX_ENABLED #endif // IP_UNIX_H diff --git a/drivers/unix/net_socket_posix.cpp b/drivers/unix/net_socket_unix.cpp index 5caa33100e..3eaf1b2885 100644 --- a/drivers/unix/net_socket_posix.cpp +++ b/drivers/unix/net_socket_unix.cpp @@ -1,5 +1,5 @@ /**************************************************************************/ -/* net_socket_posix.cpp */ +/* net_socket_unix.cpp */ /**************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ @@ -28,13 +28,11 @@ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /**************************************************************************/ -#include "net_socket_posix.h" - // Some proprietary Unix-derived platforms don't expose Unix sockets // so this allows skipping this file to reimplement this API differently. -#ifndef UNIX_SOCKET_UNAVAILABLE +#if defined(UNIX_ENABLED) && !defined(UNIX_SOCKET_UNAVAILABLE) -#if defined(UNIX_ENABLED) +#include "net_socket_unix.h" #include <errno.h> #include <fcntl.h> @@ -62,44 +60,11 @@ #define IPV6_DROP_MEMBERSHIP IPV6_LEAVE_GROUP #endif -// Some custom defines to minimize ifdefs -#define SOCK_EMPTY -1 -#define SOCK_BUF(x) x -#define SOCK_CBUF(x) x -#define SOCK_IOCTL ioctl -#define SOCK_FIONREAD_LEN_TYPE int -#define SOCK_CLOSE ::close -#define SOCK_CONNECT(p_sock, p_addr, p_addr_len) ::connect(p_sock, p_addr, p_addr_len) - -/* Windows */ -#elif defined(WINDOWS_ENABLED) -#include <winsock2.h> -#include <ws2tcpip.h> - -#include <mswsock.h> -// Some custom defines to minimize ifdefs -#define SOCK_EMPTY INVALID_SOCKET -#define SOCK_BUF(x) (char *)(x) -#define SOCK_CBUF(x) (const char *)(x) -#define SOCK_IOCTL ioctlsocket -#define SOCK_FIONREAD_LEN_TYPE unsigned long -#define SOCK_CLOSE closesocket -// connect is broken on windows under certain conditions, reasons unknown: -// See https://github.com/godotengine/webrtc-native/issues/6 -#define SOCK_CONNECT(p_sock, p_addr, p_addr_len) ::WSAConnect(p_sock, p_addr, p_addr_len, nullptr, nullptr, nullptr, nullptr) - -// Workaround missing flag in MinGW -#if defined(__MINGW32__) && !defined(SIO_UDP_NETRESET) -#define SIO_UDP_NETRESET _WSAIOW(IOC_VENDOR, 15) -#endif - -#endif // UNIX_ENABLED - -size_t NetSocketPosix::_set_addr_storage(struct sockaddr_storage *p_addr, const IPAddress &p_ip, uint16_t p_port, IP::Type p_ip_type) { +size_t NetSocketUnix::_set_addr_storage(struct sockaddr_storage *p_addr, const IPAddress &p_ip, uint16_t p_port, IP::Type p_ip_type) { memset(p_addr, 0, sizeof(struct sockaddr_storage)); - if (p_ip_type == IP::TYPE_IPV6 || p_ip_type == IP::TYPE_ANY) { // IPv6 socket + if (p_ip_type == IP::TYPE_IPV6 || p_ip_type == IP::TYPE_ANY) { // IPv6 socket. - // IPv6 only socket with IPv4 address + // IPv6 only socket with IPv4 address. ERR_FAIL_COND_V(!p_ip.is_wildcard() && p_ip_type == IP::TYPE_IPV6 && p_ip.is_ipv4(), 0); struct sockaddr_in6 *addr6 = (struct sockaddr_in6 *)p_addr; @@ -111,14 +76,14 @@ size_t NetSocketPosix::_set_addr_storage(struct sockaddr_storage *p_addr, const addr6->sin6_addr = in6addr_any; } return sizeof(sockaddr_in6); - } else { // IPv4 socket + } else { // IPv4 socket. - // IPv4 socket with IPv6 address + // IPv4 socket with IPv6 address. ERR_FAIL_COND_V(!p_ip.is_wildcard() && !p_ip.is_ipv4(), 0); struct sockaddr_in *addr4 = (struct sockaddr_in *)p_addr; addr4->sin_family = AF_INET; - addr4->sin_port = htons(p_port); // short, network byte order + addr4->sin_port = htons(p_port); // Short, network byte order. if (p_ip.is_valid()) { memcpy(&addr4->sin_addr.s_addr, p_ip.get_ipv4(), 4); @@ -130,7 +95,7 @@ size_t NetSocketPosix::_set_addr_storage(struct sockaddr_storage *p_addr, const } } -void NetSocketPosix::_set_ip_port(struct sockaddr_storage *p_addr, IPAddress *r_ip, uint16_t *r_port) { +void NetSocketUnix::_set_ip_port(struct sockaddr_storage *p_addr, IPAddress *r_ip, uint16_t *r_port) { if (p_addr->ss_family == AF_INET) { struct sockaddr_in *addr4 = (struct sockaddr_in *)p_addr; if (r_ip) { @@ -150,34 +115,21 @@ void NetSocketPosix::_set_ip_port(struct sockaddr_storage *p_addr, IPAddress *r_ } } -NetSocket *NetSocketPosix::_create_func() { - return memnew(NetSocketPosix); +NetSocket *NetSocketUnix::_create_func() { + return memnew(NetSocketUnix); } -void NetSocketPosix::make_default() { -#if defined(WINDOWS_ENABLED) - if (_create == nullptr) { - WSADATA data; - WSAStartup(MAKEWORD(2, 2), &data); - } -#endif +void NetSocketUnix::make_default() { _create = _create_func; } -void NetSocketPosix::cleanup() { -#if defined(WINDOWS_ENABLED) - if (_create != nullptr) { - WSACleanup(); - } - _create = nullptr; -#endif +void NetSocketUnix::cleanup() { } -NetSocketPosix::NetSocketPosix() : - _sock(SOCK_EMPTY) { +NetSocketUnix::NetSocketUnix() { } -NetSocketPosix::~NetSocketPosix() { +NetSocketUnix::~NetSocketUnix() { close(); } @@ -188,30 +140,7 @@ NetSocketPosix::~NetSocketPosix() { #pragma GCC diagnostic ignored "-Wlogical-op" #endif -NetSocketPosix::NetError NetSocketPosix::_get_socket_error() const { -#if defined(WINDOWS_ENABLED) - int err = WSAGetLastError(); - if (err == WSAEISCONN) { - return ERR_NET_IS_CONNECTED; - } - if (err == WSAEINPROGRESS || err == WSAEALREADY) { - return ERR_NET_IN_PROGRESS; - } - if (err == WSAEWOULDBLOCK) { - return ERR_NET_WOULD_BLOCK; - } - if (err == WSAEADDRINUSE || err == WSAEADDRNOTAVAIL) { - return ERR_NET_ADDRESS_INVALID_OR_UNAVAILABLE; - } - if (err == WSAEACCES) { - return ERR_NET_UNAUTHORIZED; - } - if (err == WSAEMSGSIZE || err == WSAENOBUFS) { - return ERR_NET_BUFFER_TOO_SMALL; - } - print_verbose("Socket error: " + itos(err)); - return ERR_NET_OTHER; -#else +NetSocketUnix::NetError NetSocketUnix::_get_socket_error() const { if (errno == EISCONN) { return ERR_NET_IS_CONNECTED; } @@ -230,16 +159,15 @@ NetSocketPosix::NetError NetSocketPosix::_get_socket_error() const { if (errno == ENOBUFS) { return ERR_NET_BUFFER_TOO_SMALL; } - print_verbose("Socket error: " + itos(errno)); + print_verbose("Socket error: " + itos(errno) + "."); return ERR_NET_OTHER; -#endif } #if defined(__GNUC__) && !defined(__clang__) #pragma GCC diagnostic pop #endif -bool NetSocketPosix::_can_use_ip(const IPAddress &p_ip, const bool p_for_bind) const { +bool NetSocketUnix::_can_use_ip(const IPAddress &p_ip, const bool p_for_bind) const { if (p_for_bind && !(p_ip.is_valid() || p_ip.is_wildcard())) { return false; } else if (!p_for_bind && !p_ip.is_valid()) { @@ -250,11 +178,11 @@ bool NetSocketPosix::_can_use_ip(const IPAddress &p_ip, const bool p_for_bind) c return !(_ip_type != IP::TYPE_ANY && !p_ip.is_wildcard() && _ip_type != type); } -_FORCE_INLINE_ Error NetSocketPosix::_change_multicast_group(IPAddress p_ip, String p_if_name, bool p_add) { +_FORCE_INLINE_ Error NetSocketUnix::_change_multicast_group(IPAddress p_ip, String p_if_name, bool p_add) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); ERR_FAIL_COND_V(!_can_use_ip(p_ip, false), ERR_INVALID_PARAMETER); - // Need to force level and af_family to IP(v4) when using dual stacking and provided multicast group is IPv4 + // Need to force level and af_family to IP(v4) when using dual stacking and provided multicast group is IPv4. IP::Type type = _ip_type == IP::TYPE_ANY && p_ip.is_ipv4() ? IP::TYPE_IPV4 : _ip_type; // This needs to be the proper level for the multicast group, no matter if the socket is dual stacking. int level = type == IP::TYPE_IPV4 ? IPPROTO_IP : IPPROTO_IPV6; @@ -277,7 +205,7 @@ _FORCE_INLINE_ Error NetSocketPosix::_change_multicast_group(IPAddress p_ip, Str for (const IPAddress &F : c.ip_addresses) { if (!F.is_ipv4()) { - continue; // Wrong IP type + continue; // Wrong IP type. } if_ip = F; break; @@ -304,7 +232,7 @@ _FORCE_INLINE_ Error NetSocketPosix::_change_multicast_group(IPAddress p_ip, Str return OK; } -void NetSocketPosix::_set_socket(SOCKET_TYPE p_sock, IP::Type p_ip_type, bool p_is_stream) { +void NetSocketUnix::_set_socket(int p_sock, IP::Type p_ip_type, bool p_is_stream) { _sock = p_sock; _ip_type = p_ip_type; _is_stream = p_is_stream; @@ -312,15 +240,13 @@ void NetSocketPosix::_set_socket(SOCKET_TYPE p_sock, IP::Type p_ip_type, bool p_ _set_close_exec_enabled(true); } -void NetSocketPosix::_set_close_exec_enabled(bool p_enabled) { -#ifndef WINDOWS_ENABLED +void NetSocketUnix::_set_close_exec_enabled(bool p_enabled) { // Enable close on exec to avoid sharing with subprocesses. Off by default on Windows. int opts = fcntl(_sock, F_GETFD); fcntl(_sock, F_SETFD, opts | FD_CLOEXEC); -#endif } -Error NetSocketPosix::open(Type p_sock_type, IP::Type &ip_type) { +Error NetSocketUnix::open(Type p_sock_type, IP::Type &ip_type) { ERR_FAIL_COND_V(is_open(), ERR_ALREADY_IN_USE); ERR_FAIL_COND_V(ip_type > IP::TYPE_ANY || ip_type < IP::TYPE_NONE, ERR_INVALID_PARAMETER); @@ -336,7 +262,7 @@ Error NetSocketPosix::open(Type p_sock_type, IP::Type &ip_type) { int type = p_sock_type == TYPE_TCP ? SOCK_STREAM : SOCK_DGRAM; _sock = socket(family, type, protocol); - if (_sock == SOCK_EMPTY && ip_type == IP::TYPE_ANY) { + if (_sock == -1 && ip_type == IP::TYPE_ANY) { // Careful here, changing the referenced parameter so the caller knows that we are using an IPv4 socket // in place of a dual stack one, and further calls to _set_sock_addr will work as expected. ip_type = IP::TYPE_IPV4; @@ -344,11 +270,11 @@ Error NetSocketPosix::open(Type p_sock_type, IP::Type &ip_type) { _sock = socket(family, type, protocol); } - ERR_FAIL_COND_V(_sock == SOCK_EMPTY, FAILED); + ERR_FAIL_COND_V(_sock == -1, FAILED); _ip_type = ip_type; if (family == AF_INET6) { - // Select IPv4 over IPv6 mapping + // Select IPv4 over IPv6 mapping. set_ipv6_only_enabled(ip_type != IP::TYPE_ANY); } @@ -363,41 +289,27 @@ Error NetSocketPosix::open(Type p_sock_type, IP::Type &ip_type) { // Disable descriptor sharing with subprocesses. _set_close_exec_enabled(true); -#if defined(WINDOWS_ENABLED) - if (!_is_stream) { - // Disable windows feature/bug reporting WSAECONNRESET/WSAENETRESET when - // recv/recvfrom and an ICMP reply was received from a previous send/sendto. - unsigned long disable = 0; - if (ioctlsocket(_sock, SIO_UDP_CONNRESET, &disable) == SOCKET_ERROR) { - print_verbose("Unable to turn off UDP WSAECONNRESET behavior on Windows"); - } - if (ioctlsocket(_sock, SIO_UDP_NETRESET, &disable) == SOCKET_ERROR) { - // This feature seems not to be supported on wine. - print_verbose("Unable to turn off UDP WSAENETRESET behavior on Windows"); - } - } -#endif #if defined(SO_NOSIGPIPE) - // Disable SIGPIPE (should only be relevant to stream sockets, but seems to affect UDP too on iOS) + // Disable SIGPIPE (should only be relevant to stream sockets, but seems to affect UDP too on iOS). int par = 1; - if (setsockopt(_sock, SOL_SOCKET, SO_NOSIGPIPE, SOCK_CBUF(&par), sizeof(int)) != 0) { - print_verbose("Unable to turn off SIGPIPE on socket"); + if (setsockopt(_sock, SOL_SOCKET, SO_NOSIGPIPE, &par, sizeof(int)) != 0) { + print_verbose("Unable to turn off SIGPIPE on socket."); } #endif return OK; } -void NetSocketPosix::close() { - if (_sock != SOCK_EMPTY) { - SOCK_CLOSE(_sock); +void NetSocketUnix::close() { + if (_sock != -1) { + ::close(_sock); } - _sock = SOCK_EMPTY; + _sock = -1; _ip_type = IP::TYPE_NONE; _is_stream = false; } -Error NetSocketPosix::bind(IPAddress p_addr, uint16_t p_port) { +Error NetSocketUnix::bind(IPAddress p_addr, uint16_t p_port) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); ERR_FAIL_COND_V(!_can_use_ip(p_addr, true), ERR_INVALID_PARAMETER); @@ -406,7 +318,7 @@ Error NetSocketPosix::bind(IPAddress p_addr, uint16_t p_port) { if (::bind(_sock, (struct sockaddr *)&addr, addr_size) != 0) { NetError err = _get_socket_error(); - print_verbose("Failed to bind socket. Error: " + itos(err)); + print_verbose("Failed to bind socket. Error: " + itos(err) + "."); close(); return ERR_UNAVAILABLE; } @@ -414,7 +326,7 @@ Error NetSocketPosix::bind(IPAddress p_addr, uint16_t p_port) { return OK; } -Error NetSocketPosix::listen(int p_max_pending) { +Error NetSocketUnix::listen(int p_max_pending) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); if (::listen(_sock, p_max_pending) != 0) { @@ -427,26 +339,26 @@ Error NetSocketPosix::listen(int p_max_pending) { return OK; } -Error NetSocketPosix::connect_to_host(IPAddress p_host, uint16_t p_port) { +Error NetSocketUnix::connect_to_host(IPAddress p_host, uint16_t p_port) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); ERR_FAIL_COND_V(!_can_use_ip(p_host, false), ERR_INVALID_PARAMETER); struct sockaddr_storage addr; size_t addr_size = _set_addr_storage(&addr, p_host, p_port, _ip_type); - if (SOCK_CONNECT(_sock, (struct sockaddr *)&addr, addr_size) != 0) { + if (::connect(_sock, (struct sockaddr *)&addr, addr_size) != 0) { NetError err = _get_socket_error(); switch (err) { - // We are already connected + // We are already connected. case ERR_NET_IS_CONNECTED: return OK; - // Still waiting to connect, try again in a while + // Still waiting to connect, try again in a while. case ERR_NET_WOULD_BLOCK: case ERR_NET_IN_PROGRESS: return ERR_BUSY; default: - print_verbose("Connection to remote host failed!"); + print_verbose("Connection to remote host failed."); close(); return FAILED; } @@ -455,66 +367,9 @@ Error NetSocketPosix::connect_to_host(IPAddress p_host, uint16_t p_port) { return OK; } -Error NetSocketPosix::poll(PollType p_type, int p_timeout) const { +Error NetSocketUnix::poll(PollType p_type, int p_timeout) const { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); -#if defined(WINDOWS_ENABLED) - bool ready = false; - fd_set rd, wr, ex; - fd_set *rdp = nullptr; - fd_set *wrp = nullptr; - FD_ZERO(&rd); - FD_ZERO(&wr); - FD_ZERO(&ex); - FD_SET(_sock, &ex); - struct timeval timeout = { p_timeout / 1000, (p_timeout % 1000) * 1000 }; - // For blocking operation, pass nullptr timeout pointer to select. - struct timeval *tp = nullptr; - if (p_timeout >= 0) { - // If timeout is non-negative, we want to specify the timeout instead. - tp = &timeout; - } - - switch (p_type) { - case POLL_TYPE_IN: - FD_SET(_sock, &rd); - rdp = &rd; - break; - case POLL_TYPE_OUT: - FD_SET(_sock, &wr); - wrp = ≀ - break; - case POLL_TYPE_IN_OUT: - FD_SET(_sock, &rd); - FD_SET(_sock, &wr); - rdp = &rd; - wrp = ≀ - } - int ret = select(1, rdp, wrp, &ex, tp); - - if (ret == SOCKET_ERROR) { - return FAILED; - } - - if (ret == 0) { - return ERR_BUSY; - } - - if (FD_ISSET(_sock, &ex)) { - _get_socket_error(); - print_verbose("Exception when polling socket."); - return FAILED; - } - - if (rdp && FD_ISSET(_sock, rdp)) { - ready = true; - } - if (wrp && FD_ISSET(_sock, wrp)) { - ready = true; - } - - return ready ? OK : ERR_BUSY; -#else struct pollfd pfd; pfd.fd = _sock; pfd.events = POLLIN; @@ -544,13 +399,12 @@ Error NetSocketPosix::poll(PollType p_type, int p_timeout) const { } return OK; -#endif } -Error NetSocketPosix::recv(uint8_t *p_buffer, int p_len, int &r_read) { +Error NetSocketUnix::recv(uint8_t *p_buffer, int p_len, int &r_read) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); - r_read = ::recv(_sock, SOCK_BUF(p_buffer), p_len, 0); + r_read = ::recv(_sock, p_buffer, p_len, 0); if (r_read < 0) { NetError err = _get_socket_error(); @@ -568,14 +422,14 @@ Error NetSocketPosix::recv(uint8_t *p_buffer, int p_len, int &r_read) { return OK; } -Error NetSocketPosix::recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddress &r_ip, uint16_t &r_port, bool p_peek) { +Error NetSocketUnix::recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddress &r_ip, uint16_t &r_port, bool p_peek) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); struct sockaddr_storage from; socklen_t len = sizeof(struct sockaddr_storage); memset(&from, 0, len); - r_read = ::recvfrom(_sock, SOCK_BUF(p_buffer), p_len, p_peek ? MSG_PEEK : 0, (struct sockaddr *)&from, &len); + r_read = ::recvfrom(_sock, p_buffer, p_len, p_peek ? MSG_PEEK : 0, (struct sockaddr *)&from, &len); if (r_read < 0) { NetError err = _get_socket_error(); @@ -606,7 +460,7 @@ Error NetSocketPosix::recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddr return OK; } -Error NetSocketPosix::send(const uint8_t *p_buffer, int p_len, int &r_sent) { +Error NetSocketUnix::send(const uint8_t *p_buffer, int p_len, int &r_sent) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); int flags = 0; @@ -615,7 +469,7 @@ Error NetSocketPosix::send(const uint8_t *p_buffer, int p_len, int &r_sent) { flags = MSG_NOSIGNAL; } #endif - r_sent = ::send(_sock, SOCK_CBUF(p_buffer), p_len, flags); + r_sent = ::send(_sock, p_buffer, p_len, flags); if (r_sent < 0) { NetError err = _get_socket_error(); @@ -632,12 +486,12 @@ Error NetSocketPosix::send(const uint8_t *p_buffer, int p_len, int &r_sent) { return OK; } -Error NetSocketPosix::sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IPAddress p_ip, uint16_t p_port) { +Error NetSocketUnix::sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IPAddress p_ip, uint16_t p_port) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); struct sockaddr_storage addr; size_t addr_size = _set_addr_storage(&addr, p_ip, p_port, _ip_type); - r_sent = ::sendto(_sock, SOCK_CBUF(p_buffer), p_len, 0, (struct sockaddr *)&addr, addr_size); + r_sent = ::sendto(_sock, p_buffer, p_len, 0, (struct sockaddr *)&addr, addr_size); if (r_sent < 0) { NetError err = _get_socket_error(); @@ -654,7 +508,7 @@ Error NetSocketPosix::sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IP return OK; } -Error NetSocketPosix::set_broadcasting_enabled(bool p_enabled) { +Error NetSocketUnix::set_broadcasting_enabled(bool p_enabled) { ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); // IPv6 has no broadcast support. if (_ip_type == IP::TYPE_IPV6) { @@ -662,90 +516,68 @@ Error NetSocketPosix::set_broadcasting_enabled(bool p_enabled) { } int par = p_enabled ? 1 : 0; - if (setsockopt(_sock, SOL_SOCKET, SO_BROADCAST, SOCK_CBUF(&par), sizeof(int)) != 0) { - WARN_PRINT("Unable to change broadcast setting"); + if (setsockopt(_sock, SOL_SOCKET, SO_BROADCAST, &par, sizeof(int)) != 0) { + WARN_PRINT("Unable to change broadcast setting."); return FAILED; } return OK; } -void NetSocketPosix::set_blocking_enabled(bool p_enabled) { +void NetSocketUnix::set_blocking_enabled(bool p_enabled) { ERR_FAIL_COND(!is_open()); int ret = 0; -#if defined(WINDOWS_ENABLED) - unsigned long par = p_enabled ? 0 : 1; - ret = SOCK_IOCTL(_sock, FIONBIO, &par); -#else int opts = fcntl(_sock, F_GETFL); if (p_enabled) { ret = fcntl(_sock, F_SETFL, opts & ~O_NONBLOCK); } else { ret = fcntl(_sock, F_SETFL, opts | O_NONBLOCK); } -#endif if (ret != 0) { - WARN_PRINT("Unable to change non-block mode"); + WARN_PRINT("Unable to change non-block mode."); } } -void NetSocketPosix::set_ipv6_only_enabled(bool p_enabled) { +void NetSocketUnix::set_ipv6_only_enabled(bool p_enabled) { ERR_FAIL_COND(!is_open()); // This option is only available in IPv6 sockets. ERR_FAIL_COND(_ip_type == IP::TYPE_IPV4); int par = p_enabled ? 1 : 0; - if (setsockopt(_sock, IPPROTO_IPV6, IPV6_V6ONLY, SOCK_CBUF(&par), sizeof(int)) != 0) { - WARN_PRINT("Unable to change IPv4 address mapping over IPv6 option"); + if (setsockopt(_sock, IPPROTO_IPV6, IPV6_V6ONLY, &par, sizeof(int)) != 0) { + WARN_PRINT("Unable to change IPv4 address mapping over IPv6 option."); } } -void NetSocketPosix::set_tcp_no_delay_enabled(bool p_enabled) { +void NetSocketUnix::set_tcp_no_delay_enabled(bool p_enabled) { ERR_FAIL_COND(!is_open()); - ERR_FAIL_COND(!_is_stream); // Not TCP + ERR_FAIL_COND(!_is_stream); // Not TCP. int par = p_enabled ? 1 : 0; - if (setsockopt(_sock, IPPROTO_TCP, TCP_NODELAY, SOCK_CBUF(&par), sizeof(int)) < 0) { - ERR_PRINT("Unable to set TCP no delay option"); + if (setsockopt(_sock, IPPROTO_TCP, TCP_NODELAY, &par, sizeof(int)) < 0) { + WARN_PRINT("Unable to set TCP no delay option."); } } -void NetSocketPosix::set_reuse_address_enabled(bool p_enabled) { +void NetSocketUnix::set_reuse_address_enabled(bool p_enabled) { ERR_FAIL_COND(!is_open()); -// On Windows, enabling SO_REUSEADDR actually would also enable reuse port, very bad on TCP. Denying... -// Windows does not have this option, SO_REUSEADDR in this magical world means SO_REUSEPORT -#ifndef WINDOWS_ENABLED - int par = p_enabled ? 1 : 0; - if (setsockopt(_sock, SOL_SOCKET, SO_REUSEADDR, SOCK_CBUF(&par), sizeof(int)) < 0) { - WARN_PRINT("Unable to set socket REUSEADDR option!"); - } -#endif -} - -void NetSocketPosix::set_reuse_port_enabled(bool p_enabled) { - ERR_FAIL_COND(!is_open()); - -// See comment above... -#ifdef WINDOWS_ENABLED -#define SO_REUSEPORT SO_REUSEADDR -#endif int par = p_enabled ? 1 : 0; - if (setsockopt(_sock, SOL_SOCKET, SO_REUSEPORT, SOCK_CBUF(&par), sizeof(int)) < 0) { - WARN_PRINT("Unable to set socket REUSEPORT option!"); + if (setsockopt(_sock, SOL_SOCKET, SO_REUSEADDR, &par, sizeof(int)) < 0) { + WARN_PRINT("Unable to set socket REUSEADDR option."); } } -bool NetSocketPosix::is_open() const { - return _sock != SOCK_EMPTY; +bool NetSocketUnix::is_open() const { + return _sock != -1; } -int NetSocketPosix::get_available_bytes() const { +int NetSocketUnix::get_available_bytes() const { ERR_FAIL_COND_V(!is_open(), -1); - SOCK_FIONREAD_LEN_TYPE len; - int ret = SOCK_IOCTL(_sock, FIONREAD, &len); + int len; + int ret = ioctl(_sock, FIONREAD, &len); if (ret == -1) { _get_socket_error(); print_verbose("Error when checking available bytes on socket."); @@ -754,7 +586,7 @@ int NetSocketPosix::get_available_bytes() const { return len; } -Error NetSocketPosix::get_socket_address(IPAddress *r_ip, uint16_t *r_port) const { +Error NetSocketUnix::get_socket_address(IPAddress *r_ip, uint16_t *r_port) const { ERR_FAIL_COND_V(!is_open(), FAILED); struct sockaddr_storage saddr; @@ -768,14 +600,14 @@ Error NetSocketPosix::get_socket_address(IPAddress *r_ip, uint16_t *r_port) cons return OK; } -Ref<NetSocket> NetSocketPosix::accept(IPAddress &r_ip, uint16_t &r_port) { +Ref<NetSocket> NetSocketUnix::accept(IPAddress &r_ip, uint16_t &r_port) { Ref<NetSocket> out; ERR_FAIL_COND_V(!is_open(), out); struct sockaddr_storage their_addr; socklen_t size = sizeof(their_addr); - SOCKET_TYPE fd = ::accept(_sock, (struct sockaddr *)&their_addr, &size); - if (fd == SOCK_EMPTY) { + int fd = ::accept(_sock, (struct sockaddr *)&their_addr, &size); + if (fd == -1) { _get_socket_error(); print_verbose("Error when accepting socket connection."); return out; @@ -783,18 +615,18 @@ Ref<NetSocket> NetSocketPosix::accept(IPAddress &r_ip, uint16_t &r_port) { _set_ip_port(&their_addr, &r_ip, &r_port); - NetSocketPosix *ns = memnew(NetSocketPosix); + NetSocketUnix *ns = memnew(NetSocketUnix); ns->_set_socket(fd, _ip_type, _is_stream); ns->set_blocking_enabled(false); return Ref<NetSocket>(ns); } -Error NetSocketPosix::join_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) { +Error NetSocketUnix::join_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) { return _change_multicast_group(p_multi_address, p_if_name, true); } -Error NetSocketPosix::leave_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) { +Error NetSocketUnix::leave_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) { return _change_multicast_group(p_multi_address, p_if_name, false); } -#endif // UNIX_SOCKET_UNAVAILABLE +#endif // UNIX_ENABLED && !UNIX_SOCKET_UNAVAILABLE diff --git a/drivers/unix/net_socket_posix.h b/drivers/unix/net_socket_unix.h index aa59ff36ee..08c45f2ac3 100644 --- a/drivers/unix/net_socket_posix.h +++ b/drivers/unix/net_socket_unix.h @@ -1,5 +1,5 @@ /**************************************************************************/ -/* net_socket_posix.h */ +/* net_socket_unix.h */ /**************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ @@ -28,25 +28,18 @@ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /**************************************************************************/ -#ifndef NET_SOCKET_POSIX_H -#define NET_SOCKET_POSIX_H +#ifndef NET_SOCKET_UNIX_H +#define NET_SOCKET_UNIX_H -#include "core/io/net_socket.h" +#if defined(UNIX_ENABLED) && !defined(UNIX_SOCKET_UNAVAILABLE) -#if defined(WINDOWS_ENABLED) -#include <winsock2.h> -#include <ws2tcpip.h> -#define SOCKET_TYPE SOCKET +#include "core/io/net_socket.h" -#else #include <sys/socket.h> -#define SOCKET_TYPE int -#endif - -class NetSocketPosix : public NetSocket { +class NetSocketUnix : public NetSocket { private: - SOCKET_TYPE _sock; // NOLINT - the default value is defined in the .cpp + int _sock = -1; IP::Type _ip_type = IP::TYPE_NONE; bool _is_stream = false; @@ -61,7 +54,7 @@ private: }; NetError _get_socket_error() const; - void _set_socket(SOCKET_TYPE p_sock, IP::Type p_ip_type, bool p_is_stream); + void _set_socket(int p_sock, IP::Type p_ip_type, bool p_is_stream); _FORCE_INLINE_ Error _change_multicast_group(IPAddress p_ip, String p_if_name, bool p_add); _FORCE_INLINE_ void _set_close_exec_enabled(bool p_enabled); @@ -76,33 +69,34 @@ public: static void _set_ip_port(struct sockaddr_storage *p_addr, IPAddress *r_ip, uint16_t *r_port); static size_t _set_addr_storage(struct sockaddr_storage *p_addr, const IPAddress &p_ip, uint16_t p_port, IP::Type p_ip_type); - virtual Error open(Type p_sock_type, IP::Type &ip_type); - virtual void close(); - virtual Error bind(IPAddress p_addr, uint16_t p_port); - virtual Error listen(int p_max_pending); - virtual Error connect_to_host(IPAddress p_host, uint16_t p_port); - virtual Error poll(PollType p_type, int timeout) const; - virtual Error recv(uint8_t *p_buffer, int p_len, int &r_read); - virtual Error recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddress &r_ip, uint16_t &r_port, bool p_peek = false); - virtual Error send(const uint8_t *p_buffer, int p_len, int &r_sent); - virtual Error sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IPAddress p_ip, uint16_t p_port); - virtual Ref<NetSocket> accept(IPAddress &r_ip, uint16_t &r_port); - - virtual bool is_open() const; - virtual int get_available_bytes() const; - virtual Error get_socket_address(IPAddress *r_ip, uint16_t *r_port) const; - - virtual Error set_broadcasting_enabled(bool p_enabled); - virtual void set_blocking_enabled(bool p_enabled); - virtual void set_ipv6_only_enabled(bool p_enabled); - virtual void set_tcp_no_delay_enabled(bool p_enabled); - virtual void set_reuse_address_enabled(bool p_enabled); - virtual void set_reuse_port_enabled(bool p_enabled); - virtual Error join_multicast_group(const IPAddress &p_multi_address, const String &p_if_name); - virtual Error leave_multicast_group(const IPAddress &p_multi_address, const String &p_if_name); - - NetSocketPosix(); - ~NetSocketPosix(); + virtual Error open(Type p_sock_type, IP::Type &ip_type) override; + virtual void close() override; + virtual Error bind(IPAddress p_addr, uint16_t p_port) override; + virtual Error listen(int p_max_pending) override; + virtual Error connect_to_host(IPAddress p_host, uint16_t p_port) override; + virtual Error poll(PollType p_type, int timeout) const override; + virtual Error recv(uint8_t *p_buffer, int p_len, int &r_read) override; + virtual Error recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddress &r_ip, uint16_t &r_port, bool p_peek = false) override; + virtual Error send(const uint8_t *p_buffer, int p_len, int &r_sent) override; + virtual Error sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IPAddress p_ip, uint16_t p_port) override; + virtual Ref<NetSocket> accept(IPAddress &r_ip, uint16_t &r_port) override; + + virtual bool is_open() const override; + virtual int get_available_bytes() const override; + virtual Error get_socket_address(IPAddress *r_ip, uint16_t *r_port) const override; + + virtual Error set_broadcasting_enabled(bool p_enabled) override; + virtual void set_blocking_enabled(bool p_enabled) override; + virtual void set_ipv6_only_enabled(bool p_enabled) override; + virtual void set_tcp_no_delay_enabled(bool p_enabled) override; + virtual void set_reuse_address_enabled(bool p_enabled) override; + virtual Error join_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) override; + virtual Error leave_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) override; + + NetSocketUnix(); + ~NetSocketUnix() override; }; -#endif // NET_SOCKET_POSIX_H +#endif // UNIX_ENABLED && !UNIX_SOCKET_UNAVAILABLE + +#endif // NET_SOCKET_UNIX_H diff --git a/drivers/unix/os_unix.cpp b/drivers/unix/os_unix.cpp index 8a9b130068..ffc270cd36 100644 --- a/drivers/unix/os_unix.cpp +++ b/drivers/unix/os_unix.cpp @@ -38,7 +38,7 @@ #include "drivers/unix/dir_access_unix.h" #include "drivers/unix/file_access_unix.h" #include "drivers/unix/file_access_unix_pipe.h" -#include "drivers/unix/net_socket_posix.h" +#include "drivers/unix/net_socket_unix.h" #include "drivers/unix/thread_posix.h" #include "servers/rendering_server.h" @@ -77,6 +77,7 @@ #include <stdlib.h> #include <string.h> #include <sys/resource.h> +#include <sys/stat.h> #include <sys/time.h> #include <sys/wait.h> #include <time.h> @@ -166,7 +167,9 @@ void OS_Unix::initialize_core() { DirAccess::make_default<DirAccessUnix>(DirAccess::ACCESS_USERDATA); DirAccess::make_default<DirAccessUnix>(DirAccess::ACCESS_FILESYSTEM); - NetSocketPosix::make_default(); +#ifndef UNIX_SOCKET_UNAVAILABLE + NetSocketUnix::make_default(); +#endif IPUnix::make_default(); process_map = memnew((HashMap<ProcessID, ProcessInfo>)); @@ -175,16 +178,96 @@ void OS_Unix::initialize_core() { void OS_Unix::finalize_core() { memdelete(process_map); - NetSocketPosix::cleanup(); +#ifndef UNIX_SOCKET_UNAVAILABLE + NetSocketUnix::cleanup(); +#endif } Vector<String> OS_Unix::get_video_adapter_driver_info() const { return Vector<String>(); } -String OS_Unix::get_stdin_string() { - char buff[1024]; - return String::utf8(fgets(buff, 1024, stdin)); +String OS_Unix::get_stdin_string(int64_t p_buffer_size) { + Vector<uint8_t> data; + data.resize(p_buffer_size); + if (fgets((char *)data.ptrw(), data.size(), stdin)) { + return String::utf8((char *)data.ptr()); + } + return String(); +} + +PackedByteArray OS_Unix::get_stdin_buffer(int64_t p_buffer_size) { + Vector<uint8_t> data; + data.resize(p_buffer_size); + size_t sz = fread((void *)data.ptrw(), 1, data.size(), stdin); + if (sz > 0) { + data.resize(sz); + return data; + } + return PackedByteArray(); +} + +OS_Unix::StdHandleType OS_Unix::get_stdin_type() const { + int h = fileno(stdin); + if (h == -1) { + return STD_HANDLE_INVALID; + } + + if (isatty(h)) { + return STD_HANDLE_CONSOLE; + } + struct stat statbuf; + if (fstat(h, &statbuf) < 0) { + return STD_HANDLE_UNKNOWN; + } + if (S_ISFIFO(statbuf.st_mode)) { + return STD_HANDLE_PIPE; + } else if (S_ISREG(statbuf.st_mode) || S_ISLNK(statbuf.st_mode)) { + return STD_HANDLE_FILE; + } + return STD_HANDLE_UNKNOWN; +} + +OS_Unix::StdHandleType OS_Unix::get_stdout_type() const { + int h = fileno(stdout); + if (h == -1) { + return STD_HANDLE_INVALID; + } + + if (isatty(h)) { + return STD_HANDLE_CONSOLE; + } + struct stat statbuf; + if (fstat(h, &statbuf) < 0) { + return STD_HANDLE_UNKNOWN; + } + if (S_ISFIFO(statbuf.st_mode)) { + return STD_HANDLE_PIPE; + } else if (S_ISREG(statbuf.st_mode) || S_ISLNK(statbuf.st_mode)) { + return STD_HANDLE_FILE; + } + return STD_HANDLE_UNKNOWN; +} + +OS_Unix::StdHandleType OS_Unix::get_stderr_type() const { + int h = fileno(stderr); + if (h == -1) { + return STD_HANDLE_INVALID; + } + + if (isatty(h)) { + return STD_HANDLE_CONSOLE; + } + struct stat statbuf; + if (fstat(h, &statbuf) < 0) { + return STD_HANDLE_UNKNOWN; + } + if (S_ISFIFO(statbuf.st_mode)) { + return STD_HANDLE_PIPE; + } else if (S_ISREG(statbuf.st_mode) || S_ISLNK(statbuf.st_mode)) { + return STD_HANDLE_FILE; + } + return STD_HANDLE_UNKNOWN; } Error OS_Unix::get_entropy(uint8_t *r_buffer, int p_bytes) { @@ -777,7 +860,7 @@ String OS_Unix::get_locale() const { } String locale = get_environment("LANG"); - int tp = locale.find("."); + int tp = locale.find_char('.'); if (tp != -1) { locale = locale.substr(0, tp); } diff --git a/drivers/unix/os_unix.h b/drivers/unix/os_unix.h index 3add5df055..2c7920c142 100644 --- a/drivers/unix/os_unix.h +++ b/drivers/unix/os_unix.h @@ -58,7 +58,11 @@ public: virtual Vector<String> get_video_adapter_driver_info() const override; - virtual String get_stdin_string() override; + virtual String get_stdin_string(int64_t p_buffer_size = 1024) override; + virtual PackedByteArray get_stdin_buffer(int64_t p_buffer_size = 1024) override; + virtual StdHandleType get_stdin_type() const override; + virtual StdHandleType get_stdout_type() const override; + virtual StdHandleType get_stderr_type() const override; virtual Error get_entropy(uint8_t *r_buffer, int p_bytes) override; diff --git a/drivers/vulkan/rendering_device_driver_vulkan.cpp b/drivers/vulkan/rendering_device_driver_vulkan.cpp index 154095552b..b6e5ed0287 100644 --- a/drivers/vulkan/rendering_device_driver_vulkan.cpp +++ b/drivers/vulkan/rendering_device_driver_vulkan.cpp @@ -35,6 +35,16 @@ #include "thirdparty/misc/smolv.h" #include "vulkan_hooks.h" +#if defined(ANDROID_ENABLED) +#include "platform/android/java_godot_wrapper.h" +#include "platform/android/os_android.h" +#include "platform/android/thread_jandroid.h" +#endif + +#if defined(SWAPPY_FRAME_PACING_ENABLED) +#include "thirdparty/swappy-frame-pacing/swappyVk.h" +#endif + #define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) #define PRINT_NATIVE_COMMANDS 0 @@ -43,6 +53,10 @@ /**** GENERIC ****/ /*****************/ +#if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) +static const uint32_t BREADCRUMB_BUFFER_ENTRIES = 512u; +#endif + static const VkFormat RD_TO_VK_FORMAT[RDD::DATA_FORMAT_MAX] = { VK_FORMAT_R4G4_UNORM_PACK8, VK_FORMAT_R4G4B4A4_UNORM_PACK16, @@ -534,6 +548,37 @@ Error RenderingDeviceDriverVulkan::_initialize_device_extensions() { err = vkEnumerateDeviceExtensionProperties(physical_device, nullptr, &device_extension_count, device_extensions.ptr()); ERR_FAIL_COND_V(err != VK_SUCCESS, ERR_CANT_CREATE); +#if defined(SWAPPY_FRAME_PACING_ENABLED) + if (swappy_frame_pacer_enable) { + char **swappy_required_extensions; + uint32_t swappy_required_extensions_count = 0; + // Determine number of extensions required by Swappy frame pacer. + SwappyVk_determineDeviceExtensions(physical_device, device_extension_count, device_extensions.ptr(), &swappy_required_extensions_count, nullptr); + + if (swappy_required_extensions_count < device_extension_count) { + // Determine the actual extensions. + swappy_required_extensions = (char **)malloc(swappy_required_extensions_count * sizeof(char *)); + char *pRequiredExtensionsData = (char *)malloc(swappy_required_extensions_count * (VK_MAX_EXTENSION_NAME_SIZE + 1)); + for (uint32_t i = 0; i < swappy_required_extensions_count; i++) { + swappy_required_extensions[i] = &pRequiredExtensionsData[i * (VK_MAX_EXTENSION_NAME_SIZE + 1)]; + } + SwappyVk_determineDeviceExtensions(physical_device, device_extension_count, + device_extensions.ptr(), &swappy_required_extensions_count, swappy_required_extensions); + + // Enable extensions requested by Swappy. + for (uint32_t i = 0; i < swappy_required_extensions_count; i++) { + CharString extension_name(swappy_required_extensions[i]); + if (requested_device_extensions.has(extension_name)) { + enabled_device_extension_names.insert(extension_name); + } + } + + free(pRequiredExtensionsData); + free(swappy_required_extensions); + } + } +#endif + #ifdef DEV_ENABLED for (uint32_t i = 0; i < device_extension_count; i++) { print_verbose(String("VULKAN: Found device extension ") + String::utf8(device_extensions[i].extensionName)); @@ -1370,7 +1415,22 @@ Error RenderingDeviceDriverVulkan::initialize(uint32_t p_device_index, uint32_t ERR_FAIL_COND_V(err != OK, err); max_descriptor_sets_per_pool = GLOBAL_GET("rendering/rendering_device/vulkan/max_descriptors_per_pool"); - breadcrumb_buffer = buffer_create(sizeof(uint32_t), BufferUsageBits::BUFFER_USAGE_TRANSFER_TO_BIT, MemoryAllocationType::MEMORY_ALLOCATION_TYPE_CPU); + +#if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) + breadcrumb_buffer = buffer_create(2u * sizeof(uint32_t) * BREADCRUMB_BUFFER_ENTRIES, BufferUsageBits::BUFFER_USAGE_TRANSFER_TO_BIT, MemoryAllocationType::MEMORY_ALLOCATION_TYPE_CPU); +#endif + +#if defined(SWAPPY_FRAME_PACING_ENABLED) + swappy_frame_pacer_enable = GLOBAL_GET("display/window/frame_pacing/android/enable_frame_pacing"); + swappy_mode = GLOBAL_GET("display/window/frame_pacing/android/swappy_mode"); + + if (VulkanHooks::get_singleton() != nullptr) { + // Hooks control device creation & possibly presentation + // (e.g. OpenXR) thus it's too risky to use Swappy. + swappy_frame_pacer_enable = false; + OS::get_singleton()->print("VulkanHooks detected (e.g. OpenXR): Force-disabling Swappy Frame Pacing.\n"); + } +#endif return OK; } @@ -1459,7 +1519,7 @@ RDD::BufferID RenderingDeviceDriverVulkan::buffer_create(uint64_t p_size, BitFie ERR_FAIL_COND_V_MSG(err, BufferID(), "Can't create buffer of size: " + itos(p_size) + ", error " + itos(err) + "."); err = vmaAllocateMemoryForBuffer(allocator, vk_buffer, &alloc_create_info, &allocation, &alloc_info); ERR_FAIL_COND_V_MSG(err, BufferID(), "Can't allocate memory for buffer of size: " + itos(p_size) + ", error " + itos(err) + "."); - err = vmaBindBufferMemory2(allocator, allocation, 0, vk_buffer, NULL); + err = vmaBindBufferMemory2(allocator, allocation, 0, vk_buffer, nullptr); ERR_FAIL_COND_V_MSG(err, BufferID(), "Can't bind memory to buffer of size: " + itos(p_size) + ", error " + itos(err) + "."); // Bookkeep. @@ -1685,7 +1745,7 @@ RDD::TextureID RenderingDeviceDriverVulkan::texture_create(const TextureFormat & ERR_FAIL_COND_V_MSG(err, TextureID(), "vkCreateImage failed with error " + itos(err) + "."); err = vmaAllocateMemoryForImage(allocator, vk_image, &alloc_create_info, &allocation, &alloc_info); ERR_FAIL_COND_V_MSG(err, TextureID(), "Can't allocate memory for image, error: " + itos(err) + "."); - err = vmaBindImageMemory2(allocator, allocation, 0, vk_image, NULL); + err = vmaBindImageMemory2(allocator, allocation, 0, vk_image, nullptr); ERR_FAIL_COND_V_MSG(err, TextureID(), "Can't bind memory to image, error: " + itos(err) + "."); // Create view. @@ -2332,7 +2392,9 @@ RDD::CommandQueueFamilyID RenderingDeviceDriverVulkan::command_queue_family_get( } } - ERR_FAIL_COND_V_MSG(picked_family_index >= queue_family_properties.size(), CommandQueueFamilyID(), "A queue family with the requested bits could not be found."); + if (picked_family_index >= queue_family_properties.size()) { + return CommandQueueFamilyID(); + } // Since 0 is a valid index and we use 0 as the error case, we make the index start from 1 instead. return CommandQueueFamilyID(picked_family_index + 1); @@ -2357,6 +2419,14 @@ RDD::CommandQueueID RenderingDeviceDriverVulkan::command_queue_create(CommandQue ERR_FAIL_COND_V_MSG(picked_queue_index >= queue_family.size(), CommandQueueID(), "A queue in the picked family could not be found."); +#if defined(SWAPPY_FRAME_PACING_ENABLED) + if (swappy_frame_pacer_enable) { + VkQueue selected_queue; + vkGetDeviceQueue(vk_device, family_index, picked_queue_index, &selected_queue); + SwappyVk_setQueueFamilyIndex(vk_device, selected_queue, family_index); + } +#endif + // Create the virtual queue. CommandQueue *command_queue = memnew(CommandQueue); command_queue->queue_family = family_index; @@ -2502,7 +2572,16 @@ Error RenderingDeviceDriverVulkan::command_queue_execute_and_present(CommandQueu present_info.pResults = results.ptr(); device_queue.submit_mutex.lock(); +#if defined(SWAPPY_FRAME_PACING_ENABLED) + if (swappy_frame_pacer_enable) { + err = SwappyVk_queuePresent(device_queue.queue, &present_info); + } else { + err = device_functions.QueuePresentKHR(device_queue.queue, &present_info); + } +#else err = device_functions.QueuePresentKHR(device_queue.queue, &present_info); +#endif + device_queue.submit_mutex.unlock(); // Set the index to an invalid value. If any of the swap chains returned out of date, indicate it should be resized the next time it's acquired. @@ -2534,7 +2613,10 @@ Error RenderingDeviceDriverVulkan::command_queue_execute_and_present(CommandQueu // it'll lead to very low performance in Android by entering an endless loop where it'll always resize the swap chain // every frame. - ERR_FAIL_COND_V(err != VK_SUCCESS && err != VK_SUBOPTIMAL_KHR, FAILED); + ERR_FAIL_COND_V_MSG( + err != VK_SUCCESS && err != VK_SUBOPTIMAL_KHR, + FAILED, + "QueuePresentKHR failed with error: " + get_vulkan_result(err)); } return OK; @@ -2637,11 +2719,13 @@ bool RenderingDeviceDriverVulkan::command_buffer_begin(CommandBufferID p_cmd_buf bool RenderingDeviceDriverVulkan::command_buffer_begin_secondary(CommandBufferID p_cmd_buffer, RenderPassID p_render_pass, uint32_t p_subpass, FramebufferID p_framebuffer) { // Reset is implicit (VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT). + Framebuffer *framebuffer = (Framebuffer *)(p_framebuffer.id); + VkCommandBufferInheritanceInfo inheritance_info = {}; inheritance_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_INHERITANCE_INFO; inheritance_info.renderPass = (VkRenderPass)p_render_pass.id; inheritance_info.subpass = p_subpass; - inheritance_info.framebuffer = (VkFramebuffer)p_framebuffer.id; + inheritance_info.framebuffer = framebuffer->vk_framebuffer; VkCommandBufferBeginInfo cmd_buf_begin_info = {}; cmd_buf_begin_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; @@ -2682,6 +2766,14 @@ void RenderingDeviceDriverVulkan::_swap_chain_release(SwapChain *swap_chain) { swap_chain->framebuffers.clear(); if (swap_chain->vk_swapchain != VK_NULL_HANDLE) { +#if defined(SWAPPY_FRAME_PACING_ENABLED) + if (swappy_frame_pacer_enable) { + // Swappy has a bug where the ANativeWindow will be leaked if we call + // SwappyVk_destroySwapchain, so we must release it by hand. + SwappyVk_setWindow(vk_device, swap_chain->vk_swapchain, nullptr); + SwappyVk_destroySwapchain(vk_device, swap_chain->vk_swapchain); + } +#endif device_functions.DestroySwapchainKHR(vk_device, swap_chain->vk_swapchain, VKC::get_allocation_callbacks(VK_OBJECT_TYPE_SWAPCHAIN_KHR)); swap_chain->vk_swapchain = VK_NULL_HANDLE; } @@ -2798,9 +2890,30 @@ Error RenderingDeviceDriverVulkan::swap_chain_resize(CommandQueueID p_cmd_queue, VkResult err = functions.GetPhysicalDeviceSurfaceCapabilitiesKHR(physical_device, surface->vk_surface, &surface_capabilities); ERR_FAIL_COND_V(err != VK_SUCCESS, ERR_CANT_CREATE); + // No swapchain yet, this is the first time we're creating it. + if (!swap_chain->vk_swapchain) { + if (surface_capabilities.currentExtent.width == 0xFFFFFFFF) { + // The current extent is currently undefined, so the current surface width and height will be clamped to the surface's capabilities. + // We make sure to overwrite surface_capabilities.currentExtent.width so that the same check further below + // does not set extent.width = CLAMP( surface->width, ... ) on the first run of this function, because + // that'd be potentially unswapped. + surface_capabilities.currentExtent.width = CLAMP(surface->width, surface_capabilities.minImageExtent.width, surface_capabilities.maxImageExtent.width); + surface_capabilities.currentExtent.height = CLAMP(surface->height, surface_capabilities.minImageExtent.height, surface_capabilities.maxImageExtent.height); + } + + // We must SWAP() only once otherwise we'll keep ping-ponging between + // the right and wrong resolutions after multiple calls to swap_chain_resize(). + if (surface_capabilities.currentTransform & VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR || + surface_capabilities.currentTransform & VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR) { + // Swap to get identity width and height. + SWAP(surface_capabilities.currentExtent.width, surface_capabilities.currentExtent.height); + } + } + VkExtent2D extent; if (surface_capabilities.currentExtent.width == 0xFFFFFFFF) { // The current extent is currently undefined, so the current surface width and height will be clamped to the surface's capabilities. + // We can only be here on the second call to swap_chain_resize(), by which time surface->width & surface->height should already be swapped if needed. extent.width = CLAMP(surface->width, surface_capabilities.minImageExtent.width, surface_capabilities.maxImageExtent.width); extent.height = CLAMP(surface->height, surface_capabilities.minImageExtent.height, surface_capabilities.maxImageExtent.height); } else { @@ -2848,9 +2961,7 @@ Error RenderingDeviceDriverVulkan::swap_chain_resize(CommandQueueID p_cmd_queue, } bool present_mode_available = present_modes.has(present_mode); - if (present_mode_available) { - print_verbose("Using present mode: " + present_mode_name); - } else { + if (!present_mode_available) { // Present mode is not available, fall back to FIFO which is guaranteed to be supported. WARN_PRINT(vformat("The requested V-Sync mode %s is not available. Falling back to V-Sync mode Enabled.", present_mode_name)); surface->vsync_mode = DisplayServer::VSYNC_ENABLED; @@ -2864,15 +2975,8 @@ Error RenderingDeviceDriverVulkan::swap_chain_resize(CommandQueueID p_cmd_queue, desired_swapchain_images = MIN(desired_swapchain_images, surface_capabilities.maxImageCount); } - // Prefer identity transform if it's supported, use the current transform otherwise. - // This behavior is intended as Godot does not supported native rotation in platforms that use these bits. // Refer to the comment in command_queue_present() for more details. - VkSurfaceTransformFlagBitsKHR surface_transform_bits; - if (surface_capabilities.supportedTransforms & VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR) { - surface_transform_bits = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR; - } else { - surface_transform_bits = surface_capabilities.currentTransform; - } + VkSurfaceTransformFlagBitsKHR surface_transform_bits = surface_capabilities.currentTransform; VkCompositeAlphaFlagBitsKHR composite_alpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR; if (OS::get_singleton()->is_layered_allowed() || !(surface_capabilities.supportedCompositeAlpha & composite_alpha)) { @@ -2904,12 +3008,60 @@ Error RenderingDeviceDriverVulkan::swap_chain_resize(CommandQueueID p_cmd_queue, swap_create_info.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; swap_create_info.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; swap_create_info.preTransform = surface_transform_bits; + switch (swap_create_info.preTransform) { + case VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR: + swap_chain->pre_transform_rotation_degrees = 0; + break; + case VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR: + swap_chain->pre_transform_rotation_degrees = 90; + break; + case VK_SURFACE_TRANSFORM_ROTATE_180_BIT_KHR: + swap_chain->pre_transform_rotation_degrees = 180; + break; + case VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR: + swap_chain->pre_transform_rotation_degrees = 270; + break; + default: + WARN_PRINT("Unexpected swap_create_info.preTransform = " + itos(swap_create_info.preTransform) + "."); + swap_chain->pre_transform_rotation_degrees = 0; + break; + } swap_create_info.compositeAlpha = composite_alpha; swap_create_info.presentMode = present_mode; swap_create_info.clipped = true; err = device_functions.CreateSwapchainKHR(vk_device, &swap_create_info, VKC::get_allocation_callbacks(VK_OBJECT_TYPE_SWAPCHAIN_KHR), &swap_chain->vk_swapchain); ERR_FAIL_COND_V(err != VK_SUCCESS, ERR_CANT_CREATE); +#if defined(SWAPPY_FRAME_PACING_ENABLED) + if (swappy_frame_pacer_enable) { + SwappyVk_initAndGetRefreshCycleDuration(get_jni_env(), static_cast<OS_Android *>(OS::get_singleton())->get_godot_java()->get_activity(), physical_device, + vk_device, swap_chain->vk_swapchain, &swap_chain->refresh_duration); + SwappyVk_setWindow(vk_device, swap_chain->vk_swapchain, static_cast<OS_Android *>(OS::get_singleton())->get_native_window()); + SwappyVk_setSwapIntervalNS(vk_device, swap_chain->vk_swapchain, swap_chain->refresh_duration); + + enum SwappyModes { + PIPELINE_FORCED_ON, + AUTO_FPS_PIPELINE_FORCED_ON, + AUTO_FPS_AUTO_PIPELINE, + }; + + switch (swappy_mode) { + case PIPELINE_FORCED_ON: + SwappyVk_setAutoSwapInterval(true); + SwappyVk_setAutoPipelineMode(true); + break; + case AUTO_FPS_PIPELINE_FORCED_ON: + SwappyVk_setAutoSwapInterval(true); + SwappyVk_setAutoPipelineMode(false); + break; + case AUTO_FPS_AUTO_PIPELINE: + SwappyVk_setAutoSwapInterval(false); + SwappyVk_setAutoPipelineMode(false); + break; + } + } +#endif + uint32_t image_count = 0; err = device_functions.GetSwapchainImagesKHR(vk_device, swap_chain->vk_swapchain, &image_count, nullptr); ERR_FAIL_COND_V(err != VK_SUCCESS, ERR_CANT_CREATE); @@ -2951,12 +3103,16 @@ Error RenderingDeviceDriverVulkan::swap_chain_resize(CommandQueueID p_cmd_queue, fb_create_info.height = surface->height; fb_create_info.layers = 1; - VkFramebuffer framebuffer; + VkFramebuffer vk_framebuffer; for (uint32_t i = 0; i < image_count; i++) { fb_create_info.pAttachments = &swap_chain->image_views[i]; - err = vkCreateFramebuffer(vk_device, &fb_create_info, VKC::get_allocation_callbacks(VK_OBJECT_TYPE_FRAMEBUFFER), &framebuffer); + err = vkCreateFramebuffer(vk_device, &fb_create_info, VKC::get_allocation_callbacks(VK_OBJECT_TYPE_FRAMEBUFFER), &vk_framebuffer); ERR_FAIL_COND_V(err != VK_SUCCESS, ERR_CANT_CREATE); + Framebuffer *framebuffer = memnew(Framebuffer); + framebuffer->vk_framebuffer = vk_framebuffer; + framebuffer->swap_chain_image = swap_chain->images[i]; + framebuffer->swap_chain_image_subresource_range = view_create_info.subresourceRange; swap_chain->framebuffers.push_back(RDD::FramebufferID(framebuffer)); } @@ -3025,7 +3181,10 @@ RDD::FramebufferID RenderingDeviceDriverVulkan::swap_chain_acquire_framebuffer(C command_queue->pending_semaphores_for_fence.push_back(semaphore_index); // Return the corresponding framebuffer to the new current image. - return swap_chain->framebuffers[swap_chain->image_index]; + FramebufferID framebuffer_id = swap_chain->framebuffers[swap_chain->image_index]; + Framebuffer *framebuffer = (Framebuffer *)(framebuffer_id.id); + framebuffer->swap_chain_acquired = true; + return framebuffer_id; } RDD::RenderPassID RenderingDeviceDriverVulkan::swap_chain_get_render_pass(SwapChainID p_swap_chain) { @@ -3035,6 +3194,13 @@ RDD::RenderPassID RenderingDeviceDriverVulkan::swap_chain_get_render_pass(SwapCh return swap_chain->render_pass; } +int RenderingDeviceDriverVulkan::swap_chain_get_pre_rotation_degrees(SwapChainID p_swap_chain) { + DEV_ASSERT(p_swap_chain.id != 0); + + SwapChain *swap_chain = (SwapChain *)(p_swap_chain.id); + return swap_chain->pre_transform_rotation_degrees; +} + RDD::DataFormat RenderingDeviceDriverVulkan::swap_chain_get_format(SwapChainID p_swap_chain) { DEV_ASSERT(p_swap_chain.id != 0); @@ -3050,6 +3216,22 @@ RDD::DataFormat RenderingDeviceDriverVulkan::swap_chain_get_format(SwapChainID p } } +void RenderingDeviceDriverVulkan::swap_chain_set_max_fps(SwapChainID p_swap_chain, int p_max_fps) { + DEV_ASSERT(p_swap_chain.id != 0); + +#ifdef SWAPPY_FRAME_PACING_ENABLED + if (!swappy_frame_pacer_enable) { + return; + } + + SwapChain *swap_chain = (SwapChain *)(p_swap_chain.id); + if (swap_chain->vk_swapchain != VK_NULL_HANDLE) { + const uint64_t max_time = p_max_fps > 0 ? uint64_t((1000.0 * 1000.0 * 1000.0) / p_max_fps) : 0; + SwappyVk_setSwapIntervalNS(vk_device, swap_chain->vk_swapchain, MAX(swap_chain->refresh_duration, max_time)); + } +#endif +} + void RenderingDeviceDriverVulkan::swap_chain_free(SwapChainID p_swap_chain) { DEV_ASSERT(p_swap_chain.id != 0); @@ -3094,11 +3276,15 @@ RDD::FramebufferID RenderingDeviceDriverVulkan::framebuffer_create(RenderPassID } #endif - return FramebufferID(vk_framebuffer); + Framebuffer *framebuffer = memnew(Framebuffer); + framebuffer->vk_framebuffer = vk_framebuffer; + return FramebufferID(framebuffer); } void RenderingDeviceDriverVulkan::framebuffer_free(FramebufferID p_framebuffer) { - vkDestroyFramebuffer(vk_device, (VkFramebuffer)p_framebuffer.id, VKC::get_allocation_callbacks(VK_OBJECT_TYPE_FRAMEBUFFER)); + Framebuffer *framebuffer = (Framebuffer *)(p_framebuffer.id); + vkDestroyFramebuffer(vk_device, framebuffer->vk_framebuffer, VKC::get_allocation_callbacks(VK_OBJECT_TYPE_FRAMEBUFFER)); + memdelete(framebuffer); } /****************/ @@ -3865,7 +4051,7 @@ RDD::UniformSetID RenderingDeviceDriverVulkan::uniform_set_create(VectorView<Bou } // Need a descriptor pool. - DescriptorSetPools::Iterator pool_sets_it = {}; + DescriptorSetPools::Iterator pool_sets_it; VkDescriptorPool vk_pool = _descriptor_set_pool_find_or_create(pool_key, &pool_sets_it); DEV_ASSERT(vk_pool); pool_sets_it->value[vk_pool]++; @@ -4316,10 +4502,25 @@ void RenderingDeviceDriverVulkan::render_pass_free(RenderPassID p_render_pass) { static_assert(ARRAYS_COMPATIBLE_FIELDWISE(RDD::RenderPassClearValue, VkClearValue)); void RenderingDeviceDriverVulkan::command_begin_render_pass(CommandBufferID p_cmd_buffer, RenderPassID p_render_pass, FramebufferID p_framebuffer, CommandBufferType p_cmd_buffer_type, const Rect2i &p_rect, VectorView<RenderPassClearValue> p_clear_values) { + Framebuffer *framebuffer = (Framebuffer *)(p_framebuffer.id); + if (framebuffer->swap_chain_acquired) { + // Insert a barrier to wait for the acquisition of the framebuffer before the render pass begins. + VkImageMemoryBarrier image_barrier = {}; + image_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; + image_barrier.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; + image_barrier.newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + image_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + image_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + image_barrier.image = framebuffer->swap_chain_image; + image_barrier.subresourceRange = framebuffer->swap_chain_image_subresource_range; + vkCmdPipelineBarrier((VkCommandBuffer)p_cmd_buffer.id, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_barrier); + framebuffer->swap_chain_acquired = false; + } + VkRenderPassBeginInfo render_pass_begin = {}; render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; render_pass_begin.renderPass = (VkRenderPass)p_render_pass.id; - render_pass_begin.framebuffer = (VkFramebuffer)p_framebuffer.id; + render_pass_begin.framebuffer = framebuffer->vk_framebuffer; render_pass_begin.renderArea.offset.x = p_rect.position.x; render_pass_begin.renderArea.offset.y = p_rect.position.y; @@ -4976,10 +5177,65 @@ void RenderingDeviceDriverVulkan::command_end_label(CommandBufferID p_cmd_buffer /**** DEBUG *****/ /****************/ void RenderingDeviceDriverVulkan::command_insert_breadcrumb(CommandBufferID p_cmd_buffer, uint32_t p_data) { +#if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) if (p_data == BreadcrumbMarker::NONE) { return; } - vkCmdFillBuffer((VkCommandBuffer)p_cmd_buffer.id, ((BufferInfo *)breadcrumb_buffer.id)->vk_buffer, 0, sizeof(uint32_t), p_data); + + if (Engine::get_singleton()->is_accurate_breadcrumbs_enabled()) { + // Force a full barrier so commands are not executed in parallel. + // This will mean that the last breadcrumb to see was actually the + // last (group of) command to be executed (hence, the one causing the crash). + VkMemoryBarrier memoryBarrier; + memoryBarrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER; + memoryBarrier.pNext = nullptr; + memoryBarrier.srcAccessMask = VK_ACCESS_INDIRECT_COMMAND_READ_BIT | + VK_ACCESS_INDEX_READ_BIT | + VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | + VK_ACCESS_UNIFORM_READ_BIT | + VK_ACCESS_INPUT_ATTACHMENT_READ_BIT | + VK_ACCESS_SHADER_READ_BIT | + VK_ACCESS_SHADER_WRITE_BIT | + VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | + VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | + VK_ACCESS_TRANSFER_READ_BIT | + VK_ACCESS_TRANSFER_WRITE_BIT | + VK_ACCESS_HOST_READ_BIT | + VK_ACCESS_HOST_WRITE_BIT; + memoryBarrier.dstAccessMask = VK_ACCESS_INDIRECT_COMMAND_READ_BIT | + VK_ACCESS_INDEX_READ_BIT | + VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | + VK_ACCESS_UNIFORM_READ_BIT | + VK_ACCESS_INPUT_ATTACHMENT_READ_BIT | + VK_ACCESS_SHADER_READ_BIT | + VK_ACCESS_SHADER_WRITE_BIT | + VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | + VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | + VK_ACCESS_TRANSFER_READ_BIT | + VK_ACCESS_TRANSFER_WRITE_BIT | + VK_ACCESS_HOST_READ_BIT | + VK_ACCESS_HOST_WRITE_BIT; + + vkCmdPipelineBarrier( + (VkCommandBuffer)p_cmd_buffer.id, + VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, + VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, + 0, 1u, &memoryBarrier, 0u, nullptr, 0u, nullptr); + } + + // We write to a circular buffer. If you're getting barrier sync errors here, + // increase the value of BREADCRUMB_BUFFER_ENTRIES. + vkCmdFillBuffer((VkCommandBuffer)p_cmd_buffer.id, ((BufferInfo *)breadcrumb_buffer.id)->vk_buffer, breadcrumb_offset, sizeof(uint32_t), breadcrumb_id++); + vkCmdFillBuffer((VkCommandBuffer)p_cmd_buffer.id, ((BufferInfo *)breadcrumb_buffer.id)->vk_buffer, breadcrumb_offset + sizeof(uint32_t), sizeof(uint32_t), p_data); + breadcrumb_offset += sizeof(uint32_t) * 2u; + if (breadcrumb_offset >= BREADCRUMB_BUFFER_ENTRIES * sizeof(uint32_t) * 2u) { + breadcrumb_offset = 0u; + } +#endif } void RenderingDeviceDriverVulkan::on_device_lost() const { @@ -5061,68 +5317,142 @@ void RenderingDeviceDriverVulkan::on_device_lost() const { void RenderingDeviceDriverVulkan::print_lost_device_info() { #if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) - void *breadcrumb_ptr; - vmaFlushAllocation(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle, 0, sizeof(uint32_t)); - vmaInvalidateAllocation(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle, 0, sizeof(uint32_t)); - - vmaMapMemory(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle, &breadcrumb_ptr); - uint32_t last_breadcrumb = *(uint32_t *)breadcrumb_ptr; - vmaUnmapMemory(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle); - uint32_t phase = last_breadcrumb & uint32_t(~((1 << 16) - 1)); - uint32_t user_data = last_breadcrumb & ((1 << 16) - 1); - String error_msg = "Last known breadcrumb: "; - - switch (phase) { - case BreadcrumbMarker::ALPHA_PASS: - error_msg += "ALPHA_PASS"; - break; - case BreadcrumbMarker::BLIT_PASS: - error_msg += "BLIT_PASS"; - break; - case BreadcrumbMarker::DEBUG_PASS: - error_msg += "DEBUG_PASS"; - break; - case BreadcrumbMarker::LIGHTMAPPER_PASS: - error_msg += "LIGHTMAPPER_PASS"; - break; - case BreadcrumbMarker::OPAQUE_PASS: - error_msg += "OPAQUE_PASS"; - break; - case BreadcrumbMarker::POST_PROCESSING_PASS: - error_msg += "POST_PROCESSING_PASS"; - break; - case BreadcrumbMarker::REFLECTION_PROBES: - error_msg += "REFLECTION_PROBES"; - break; - case BreadcrumbMarker::SHADOW_PASS_CUBE: - error_msg += "SHADOW_PASS_CUBE"; - break; - case BreadcrumbMarker::SHADOW_PASS_DIRECTIONAL: - error_msg += "SHADOW_PASS_DIRECTIONAL"; - break; - case BreadcrumbMarker::SKY_PASS: - error_msg += "SKY_PASS"; - break; - case BreadcrumbMarker::TRANSPARENT_PASS: - error_msg += "TRANSPARENT_PASS"; - break; - case BreadcrumbMarker::UI_PASS: - error_msg += "UI_PASS"; - break; - default: - error_msg += "UNKNOWN_BREADCRUMB(" + itos((uint32_t)phase) + ')'; - break; + { + String error_msg = "Printing last known breadcrumbs in reverse order (last executed first)."; + if (!Engine::get_singleton()->is_accurate_breadcrumbs_enabled()) { + error_msg += "\nSome of them might be inaccurate. Try running with --accurate-breadcrumbs for precise information."; + } + _err_print_error(FUNCTION_STR, __FILE__, __LINE__, error_msg); } - if (user_data != 0) { - error_msg += " | User data: " + itos(user_data); - } + uint8_t *breadcrumb_ptr = nullptr; + VkResult map_result = VK_SUCCESS; + + vmaFlushAllocation(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle, 0, BREADCRUMB_BUFFER_ENTRIES * sizeof(uint32_t) * 2u); + vmaInvalidateAllocation(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle, 0, BREADCRUMB_BUFFER_ENTRIES * sizeof(uint32_t) * 2u); + { + void *ptr = nullptr; + map_result = vmaMapMemory(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle, &ptr); + breadcrumb_ptr = reinterpret_cast<uint8_t *>(ptr); + } + + if (breadcrumb_ptr && map_result == VK_SUCCESS) { + uint32_t last_breadcrumb_offset = 0; + { + _err_print_error_asap("Searching last breadcrumb. We've sent up to ID: " + itos(breadcrumb_id - 1u)); + + // Scan the whole buffer to find the offset with the highest ID. + // That means that was the last one to be written. + // + // We use "breadcrumb_id - id" to account for wraparound. + // e.g. breadcrumb_id = 2 and id = 4294967294; then 2 - 4294967294 = 4. + // The one with the smallest difference is the closest to breadcrumb_id, which means it's + // the last written command. + uint32_t biggest_id = 0u; + uint32_t smallest_id_diff = std::numeric_limits<uint32_t>::max(); + const uint32_t *breadcrumb_ptr32 = reinterpret_cast<const uint32_t *>(breadcrumb_ptr); + for (size_t i = 0u; i < BREADCRUMB_BUFFER_ENTRIES; ++i) { + const uint32_t id = breadcrumb_ptr32[i * 2u]; + const uint32_t id_diff = breadcrumb_id - id; + if (id_diff < smallest_id_diff) { + biggest_id = i; + smallest_id_diff = id_diff; + } + } - _err_print_error(FUNCTION_STR, __FILE__, __LINE__, error_msg); + _err_print_error_asap("Last breadcrumb ID found: " + itos(breadcrumb_ptr32[biggest_id * 2u])); + + last_breadcrumb_offset = biggest_id * sizeof(uint32_t) * 2u; + } + + const size_t entries_to_print = 8u; // Note: The value is arbitrary. + for (size_t i = 0u; i < entries_to_print; ++i) { + const uint32_t last_breadcrumb = *reinterpret_cast<uint32_t *>(breadcrumb_ptr + last_breadcrumb_offset + sizeof(uint32_t)); + const uint32_t phase = last_breadcrumb & uint32_t(~((1 << 16) - 1)); + const uint32_t user_data = last_breadcrumb & ((1 << 16) - 1); + String error_msg = "Last known breadcrumb: "; + + switch (phase) { + case BreadcrumbMarker::ALPHA_PASS: + error_msg += "ALPHA_PASS"; + break; + case BreadcrumbMarker::BLIT_PASS: + error_msg += "BLIT_PASS"; + break; + case BreadcrumbMarker::DEBUG_PASS: + error_msg += "DEBUG_PASS"; + break; + case BreadcrumbMarker::LIGHTMAPPER_PASS: + error_msg += "LIGHTMAPPER_PASS"; + break; + case BreadcrumbMarker::OPAQUE_PASS: + error_msg += "OPAQUE_PASS"; + break; + case BreadcrumbMarker::POST_PROCESSING_PASS: + error_msg += "POST_PROCESSING_PASS"; + break; + case BreadcrumbMarker::REFLECTION_PROBES: + error_msg += "REFLECTION_PROBES"; + break; + case BreadcrumbMarker::SHADOW_PASS_CUBE: + error_msg += "SHADOW_PASS_CUBE"; + break; + case BreadcrumbMarker::SHADOW_PASS_DIRECTIONAL: + error_msg += "SHADOW_PASS_DIRECTIONAL"; + break; + case BreadcrumbMarker::SKY_PASS: + error_msg += "SKY_PASS"; + break; + case BreadcrumbMarker::TRANSPARENT_PASS: + error_msg += "TRANSPARENT_PASS"; + break; + case BreadcrumbMarker::UI_PASS: + error_msg += "UI_PASS"; + break; + default: + error_msg += "UNKNOWN_BREADCRUMB(" + itos((uint32_t)phase) + ')'; + break; + } + + if (user_data != 0) { + error_msg += " | User data: " + itos(user_data); + } + + _err_print_error_asap(error_msg); + + if (last_breadcrumb_offset == 0u) { + // Decrement last_breadcrumb_idx, wrapping underflow. + last_breadcrumb_offset = BREADCRUMB_BUFFER_ENTRIES * sizeof(uint32_t) * 2u; + } + last_breadcrumb_offset -= sizeof(uint32_t) * 2u; + } + + vmaUnmapMemory(allocator, ((BufferInfo *)breadcrumb_buffer.id)->allocation.handle); + breadcrumb_ptr = nullptr; + } else { + _err_print_error(FUNCTION_STR, __FILE__, __LINE__, "Couldn't map breadcrumb buffer. VkResult = " + itos(map_result)); + } #endif on_device_lost(); } +inline String RenderingDeviceDriverVulkan::get_vulkan_result(VkResult err) { +#if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) + if (err == VK_ERROR_OUT_OF_HOST_MEMORY) { + return "VK_ERROR_OUT_OF_HOST_MEMORY"; + } else if (err == VK_ERROR_OUT_OF_DEVICE_MEMORY) { + return "VK_ERROR_OUT_OF_DEVICE_MEMORY"; + } else if (err == VK_ERROR_DEVICE_LOST) { + return "VK_ERROR_DEVICE_LOST"; + } else if (err == VK_ERROR_SURFACE_LOST_KHR) { + return "VK_ERROR_SURFACE_LOST_KHR"; + } else if (err == VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT) { + return "VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT"; + } +#endif + return itos(err); +} + /********************/ /**** SUBMISSION ****/ /********************/ @@ -5389,7 +5719,9 @@ RenderingDeviceDriverVulkan::RenderingDeviceDriverVulkan(RenderingContextDriverV } RenderingDeviceDriverVulkan::~RenderingDeviceDriverVulkan() { +#if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) buffer_free(breadcrumb_buffer); +#endif while (small_allocs_pools.size()) { HashMap<uint32_t, VmaPool>::Iterator E = small_allocs_pools.begin(); diff --git a/drivers/vulkan/rendering_device_driver_vulkan.h b/drivers/vulkan/rendering_device_driver_vulkan.h index 58f7a97ec0..06cd2a31be 100644 --- a/drivers/vulkan/rendering_device_driver_vulkan.h +++ b/drivers/vulkan/rendering_device_driver_vulkan.h @@ -143,6 +143,11 @@ class RenderingDeviceDriverVulkan : public RenderingDeviceDriver { #if defined(VK_TRACK_DEVICE_MEMORY) bool device_memory_report_support = false; #endif +#if defined(SWAPPY_FRAME_PACING_ENABLED) + // Swappy frame pacer for Android. + bool swappy_frame_pacer_enable = false; + uint8_t swappy_mode = 2; // See default value for display/window/frame_pacing/android/swappy_mode. +#endif DeviceFunctions device_functions; void _register_requested_device_extension(const CharString &p_extension_name, bool p_required); @@ -172,7 +177,12 @@ private: VmaPool _find_or_create_small_allocs_pool(uint32_t p_mem_type_index); private: +#if defined(DEBUG_ENABLED) || defined(DEV_ENABLED) + // It's a circular buffer. BufferID breadcrumb_buffer; + uint32_t breadcrumb_offset = 0u; + uint32_t breadcrumb_id = 0u; +#endif public: /*****************/ @@ -349,7 +359,11 @@ private: LocalVector<CommandQueue *> command_queues_acquired; LocalVector<uint32_t> command_queues_acquired_semaphores; RenderPassID render_pass; + int pre_transform_rotation_degrees = 0; uint32_t image_index = 0; +#ifdef ANDROID_ENABLED + uint64_t refresh_duration = 0; +#endif }; void _swap_chain_release(SwapChain *p_swap_chain); @@ -359,13 +373,24 @@ public: virtual Error swap_chain_resize(CommandQueueID p_cmd_queue, SwapChainID p_swap_chain, uint32_t p_desired_framebuffer_count) override final; virtual FramebufferID swap_chain_acquire_framebuffer(CommandQueueID p_cmd_queue, SwapChainID p_swap_chain, bool &r_resize_required) override final; virtual RenderPassID swap_chain_get_render_pass(SwapChainID p_swap_chain) override final; + virtual int swap_chain_get_pre_rotation_degrees(SwapChainID p_swap_chain) override final; virtual DataFormat swap_chain_get_format(SwapChainID p_swap_chain) override final; + virtual void swap_chain_set_max_fps(SwapChainID p_swap_chain, int p_max_fps) override final; virtual void swap_chain_free(SwapChainID p_swap_chain) override final; /*********************/ /**** FRAMEBUFFER ****/ /*********************/ + struct Framebuffer { + VkFramebuffer vk_framebuffer = VK_NULL_HANDLE; + + // Only filled in by a framebuffer created by a swap chain. Unused otherwise. + VkImage swap_chain_image = VK_NULL_HANDLE; + VkImageSubresourceRange swap_chain_image_subresource_range = {}; + bool swap_chain_acquired = false; + }; + virtual FramebufferID framebuffer_create(RenderPassID p_render_pass, VectorView<TextureID> p_attachments, uint32_t p_width, uint32_t p_height) override final; virtual void framebuffer_free(FramebufferID p_framebuffer) override final; @@ -463,7 +488,7 @@ private: struct UniformSetInfo { VkDescriptorSet vk_descriptor_set = VK_NULL_HANDLE; VkDescriptorPool vk_descriptor_pool = VK_NULL_HANDLE; - DescriptorSetPools::Iterator pool_sets_it = {}; + DescriptorSetPools::Iterator pool_sets_it; }; public: @@ -630,6 +655,7 @@ public: virtual void command_insert_breadcrumb(CommandBufferID p_cmd_buffer, uint32_t p_data) override final; void print_lost_device_info(); void on_device_lost() const; + static String get_vulkan_result(VkResult err); /********************/ /**** SUBMISSION ****/ diff --git a/drivers/wasapi/audio_driver_wasapi.cpp b/drivers/wasapi/audio_driver_wasapi.cpp index 4e7c9695a5..b800e5f1f4 100644 --- a/drivers/wasapi/audio_driver_wasapi.cpp +++ b/drivers/wasapi/audio_driver_wasapi.cpp @@ -85,7 +85,7 @@ public: _In_ const WAVEFORMATEX *pFormat, /* [annotation][in] */ _In_opt_ LPCGUID AudioSessionGuid) = 0; -}; +} __CRT_UUID_DECL(IAudioClient3, 0x7ED4EE07, 0x8E67, 0x4CD4, 0x8C, 0x1A, 0x2B, 0x7A, 0x59, 0x87, 0xAD, 0x42) #endif // __IAudioClient3_INTERFACE_DEFINED__ diff --git a/drivers/windows/dir_access_windows.cpp b/drivers/windows/dir_access_windows.cpp index f7632842ed..3ddbde72c4 100644 --- a/drivers/windows/dir_access_windows.cpp +++ b/drivers/windows/dir_access_windows.cpp @@ -230,7 +230,7 @@ String DirAccessWindows::get_current_dir(bool p_include_drive) const { return cdir; } else { if (_get_root_string().is_empty()) { - int pos = cdir.find(":"); + int pos = cdir.find_char(':'); if (pos != -1) { return cdir.substr(pos + 1); } @@ -282,7 +282,7 @@ Error DirAccessWindows::rename(String p_path, String p_new_path) { uint64_t id = OS::get_singleton()->get_ticks_usec(); while (true) { tmpfile_utf16 = (path + itos(id++) + ".tmp").utf16(); - HANDLE handle = CreateFileW((LPCWSTR)tmpfile_utf16.get_data(), GENERIC_WRITE, 0, NULL, CREATE_NEW, FILE_ATTRIBUTE_NORMAL, 0); + HANDLE handle = CreateFileW((LPCWSTR)tmpfile_utf16.get_data(), GENERIC_WRITE, 0, nullptr, CREATE_NEW, FILE_ATTRIBUTE_NORMAL, nullptr); if (handle != INVALID_HANDLE_VALUE) { CloseHandle(handle); break; @@ -344,7 +344,7 @@ String DirAccessWindows::get_filesystem_type() const { return "Network Share"; } - int unit_end = path.find(":"); + int unit_end = path.find_char(':'); ERR_FAIL_COND_V(unit_end == -1, String()); String unit = path.substr(0, unit_end + 1) + "\\"; diff --git a/drivers/windows/file_access_windows.cpp b/drivers/windows/file_access_windows.cpp index a8a2ea6b5e..4a0e5e5f49 100644 --- a/drivers/windows/file_access_windows.cpp +++ b/drivers/windows/file_access_windows.cpp @@ -64,7 +64,7 @@ bool FileAccessWindows::is_path_invalid(const String &p_path) { // Check for invalid operating system file. String fname = p_path.get_file().to_lower(); - int dot = fname.find("."); + int dot = fname.find_char('.'); if (dot != -1) { fname = fname.substr(0, dot); } @@ -127,7 +127,7 @@ Error FileAccessWindows::open_internal(const String &p_path, int p_mode_flags) { } #ifdef TOOLS_ENABLED - // Windows is case insensitive, but all other platforms are sensitive to it + // Windows is case insensitive in the default configuration, but other platforms can be sensitive to it // To ease cross-platform development, we issue a warning if users try to access // a file using the wrong case (which *works* on Windows, but won't on other // platforms), we only check for relative paths, or paths in res:// or user://, @@ -193,7 +193,7 @@ Error FileAccessWindows::open_internal(const String &p_path, int p_mode_flags) { uint64_t id = OS::get_singleton()->get_ticks_usec(); while (true) { tmpfile = path + itos(id++) + ".tmp"; - HANDLE handle = CreateFileW((LPCWSTR)tmpfile.utf16().get_data(), GENERIC_WRITE, 0, NULL, CREATE_NEW, FILE_ATTRIBUTE_NORMAL, 0); + HANDLE handle = CreateFileW((LPCWSTR)tmpfile.utf16().get_data(), GENERIC_WRITE, 0, nullptr, CREATE_NEW, FILE_ATTRIBUTE_NORMAL, nullptr); if (handle != INVALID_HANDLE_VALUE) { CloseHandle(handle); break; diff --git a/drivers/windows/file_access_windows_pipe.cpp b/drivers/windows/file_access_windows_pipe.cpp index 9bf0f4d852..7348b29a92 100644 --- a/drivers/windows/file_access_windows_pipe.cpp +++ b/drivers/windows/file_access_windows_pipe.cpp @@ -40,7 +40,7 @@ Error FileAccessWindowsPipe::open_existing(HANDLE p_rfd, HANDLE p_wfd, bool p_bl _close(); path_src = String(); - ERR_FAIL_COND_V_MSG(fd[0] != 0 || fd[1] != 0, ERR_ALREADY_IN_USE, "Pipe is already in use."); + ERR_FAIL_COND_V_MSG(fd[0] != nullptr || fd[1] != nullptr, ERR_ALREADY_IN_USE, "Pipe is already in use."); fd[0] = p_rfd; fd[1] = p_wfd; @@ -58,18 +58,18 @@ Error FileAccessWindowsPipe::open_internal(const String &p_path, int p_mode_flag _close(); path_src = p_path; - ERR_FAIL_COND_V_MSG(fd[0] != 0 || fd[1] != 0, ERR_ALREADY_IN_USE, "Pipe is already in use."); + ERR_FAIL_COND_V_MSG(fd[0] != nullptr || fd[1] != nullptr, ERR_ALREADY_IN_USE, "Pipe is already in use."); path = String("\\\\.\\pipe\\LOCAL\\") + p_path.replace("pipe://", "").replace("/", "_"); - HANDLE h = CreateFileW((LPCWSTR)path.utf16().get_data(), GENERIC_READ | GENERIC_WRITE, 0, nullptr, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); + HANDLE h = CreateFileW((LPCWSTR)path.utf16().get_data(), GENERIC_READ | GENERIC_WRITE, 0, nullptr, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, nullptr); if (h == INVALID_HANDLE_VALUE) { h = CreateNamedPipeW((LPCWSTR)path.utf16().get_data(), PIPE_ACCESS_DUPLEX, PIPE_TYPE_BYTE | PIPE_READMODE_BYTE | PIPE_NOWAIT, 1, 4096, 4096, 0, nullptr); if (h == INVALID_HANDLE_VALUE) { last_error = ERR_FILE_CANT_OPEN; return last_error; } - ConnectNamedPipe(h, NULL); + ConnectNamedPipe(h, nullptr); } fd[0] = h; fd[1] = h; @@ -79,19 +79,19 @@ Error FileAccessWindowsPipe::open_internal(const String &p_path, int p_mode_flag } void FileAccessWindowsPipe::_close() { - if (fd[0] == 0) { + if (fd[0] == nullptr) { return; } if (fd[1] != fd[0]) { CloseHandle(fd[1]); } CloseHandle(fd[0]); - fd[0] = 0; - fd[1] = 0; + fd[0] = nullptr; + fd[1] = nullptr; } bool FileAccessWindowsPipe::is_open() const { - return (fd[0] != 0 || fd[1] != 0); + return (fd[0] != nullptr || fd[1] != nullptr); } String FileAccessWindowsPipe::get_path() const { @@ -103,7 +103,7 @@ String FileAccessWindowsPipe::get_path_absolute() const { } uint64_t FileAccessWindowsPipe::get_buffer(uint8_t *p_dst, uint64_t p_length) const { - ERR_FAIL_COND_V_MSG(fd[0] == 0, -1, "Pipe must be opened before use."); + ERR_FAIL_COND_V_MSG(fd[0] == nullptr, -1, "Pipe must be opened before use."); ERR_FAIL_COND_V(!p_dst && p_length > 0, -1); DWORD read = 0; @@ -120,7 +120,7 @@ Error FileAccessWindowsPipe::get_error() const { } void FileAccessWindowsPipe::store_buffer(const uint8_t *p_src, uint64_t p_length) { - ERR_FAIL_COND_MSG(fd[1] == 0, "Pipe must be opened before use."); + ERR_FAIL_COND_MSG(fd[1] == nullptr, "Pipe must be opened before use."); ERR_FAIL_COND(!p_src && p_length > 0); DWORD read = -1; diff --git a/drivers/windows/file_access_windows_pipe.h b/drivers/windows/file_access_windows_pipe.h index 1eb3c6ef2f..5edf0500a5 100644 --- a/drivers/windows/file_access_windows_pipe.h +++ b/drivers/windows/file_access_windows_pipe.h @@ -39,7 +39,7 @@ #define WIN32_LEAN_AND_MEAN #include <windows.h> class FileAccessWindowsPipe : public FileAccess { - HANDLE fd[2] = { 0, 0 }; + HANDLE fd[2] = { nullptr, nullptr }; mutable Error last_error = OK; diff --git a/drivers/windows/ip_windows.cpp b/drivers/windows/ip_windows.cpp new file mode 100644 index 0000000000..4ff368d3f4 --- /dev/null +++ b/drivers/windows/ip_windows.cpp @@ -0,0 +1,164 @@ +/**************************************************************************/ +/* ip_windows.cpp */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ + +#if defined(WINDOWS_ENABLED) + +#include "ip_windows.h" + +#define WIN32_LEAN_AND_MEAN +#include <windows.h> +#include <winsock2.h> +#include <ws2tcpip.h> + +#include <iphlpapi.h> + +#include <stdio.h> + +#include <string.h> + +static IPAddress _sockaddr2ip(struct sockaddr *p_addr) { + IPAddress ip; + + if (p_addr->sa_family == AF_INET) { + struct sockaddr_in *addr = (struct sockaddr_in *)p_addr; + ip.set_ipv4((uint8_t *)&(addr->sin_addr)); + } else if (p_addr->sa_family == AF_INET6) { + struct sockaddr_in6 *addr6 = (struct sockaddr_in6 *)p_addr; + ip.set_ipv6(addr6->sin6_addr.s6_addr); + } + + return ip; +} + +void IPWindows::_resolve_hostname(List<IPAddress> &r_addresses, const String &p_hostname, Type p_type) const { + struct addrinfo hints; + struct addrinfo *result = nullptr; + + memset(&hints, 0, sizeof(struct addrinfo)); + if (p_type == TYPE_IPV4) { + hints.ai_family = AF_INET; + } else if (p_type == TYPE_IPV6) { + hints.ai_family = AF_INET6; + hints.ai_flags = 0; + } else { + hints.ai_family = AF_UNSPEC; + hints.ai_flags = AI_ADDRCONFIG; + } + hints.ai_flags &= ~AI_NUMERICHOST; + + int s = getaddrinfo(p_hostname.utf8().get_data(), nullptr, &hints, &result); + if (s != 0) { + print_verbose("getaddrinfo failed! Cannot resolve hostname."); + return; + } + + if (result == nullptr || result->ai_addr == nullptr) { + print_verbose("Invalid response from getaddrinfo."); + if (result) { + freeaddrinfo(result); + } + return; + } + + struct addrinfo *next = result; + + do { + if (next->ai_addr == nullptr) { + next = next->ai_next; + continue; + } + IPAddress ip = _sockaddr2ip(next->ai_addr); + if (ip.is_valid() && !r_addresses.find(ip)) { + r_addresses.push_back(ip); + } + next = next->ai_next; + } while (next); + + freeaddrinfo(result); +} + +void IPWindows::get_local_interfaces(HashMap<String, Interface_Info> *r_interfaces) const { + ULONG buf_size = 1024; + IP_ADAPTER_ADDRESSES *addrs; + + while (true) { + addrs = (IP_ADAPTER_ADDRESSES *)memalloc(buf_size); + int err = GetAdaptersAddresses(AF_UNSPEC, GAA_FLAG_SKIP_ANYCAST | GAA_FLAG_SKIP_MULTICAST | GAA_FLAG_SKIP_DNS_SERVER | GAA_FLAG_SKIP_FRIENDLY_NAME, + nullptr, addrs, &buf_size); + if (err == NO_ERROR) { + break; + } + memfree(addrs); + if (err == ERROR_BUFFER_OVERFLOW) { + continue; // Will go back and alloc the right size. + } + + ERR_FAIL_MSG("Call to GetAdaptersAddresses failed with error " + itos(err) + "."); + } + + IP_ADAPTER_ADDRESSES *adapter = addrs; + + while (adapter != nullptr) { + Interface_Info info; + info.name = adapter->AdapterName; + info.name_friendly = adapter->FriendlyName; + info.index = String::num_uint64(adapter->IfIndex); + + IP_ADAPTER_UNICAST_ADDRESS *address = adapter->FirstUnicastAddress; + while (address != nullptr) { + int family = address->Address.lpSockaddr->sa_family; + if (family != AF_INET && family != AF_INET6) { + continue; + } + info.ip_addresses.push_front(_sockaddr2ip(address->Address.lpSockaddr)); + address = address->Next; + } + adapter = adapter->Next; + // Only add interface if it has at least one IP. + if (info.ip_addresses.size() > 0) { + r_interfaces->insert(info.name, info); + } + } + + memfree(addrs); +} + +void IPWindows::make_default() { + _create = _create_unix; +} + +IP *IPWindows::_create_unix() { + return memnew(IPWindows); +} + +IPWindows::IPWindows() { +} + +#endif // WINDOWS_ENABLED diff --git a/drivers/windows/ip_windows.h b/drivers/windows/ip_windows.h new file mode 100644 index 0000000000..5788d33ee2 --- /dev/null +++ b/drivers/windows/ip_windows.h @@ -0,0 +1,54 @@ +/**************************************************************************/ +/* ip_windows.h */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ + +#ifndef IP_WINDOWS_H +#define IP_WINDOWS_H + +#if defined(WINDOWS_ENABLED) + +#include "core/io/ip.h" + +class IPWindows : public IP { + GDCLASS(IPWindows, IP); + + virtual void _resolve_hostname(List<IPAddress> &r_addresses, const String &p_hostname, Type p_type = TYPE_ANY) const override; + + static IP *_create_unix(); + +public: + virtual void get_local_interfaces(HashMap<String, Interface_Info> *r_interfaces) const override; + + static void make_default(); + IPWindows(); +}; + +#endif // WINDOWS_ENABLED + +#endif // IP_WINDOWS_H diff --git a/drivers/windows/net_socket_winsock.cpp b/drivers/windows/net_socket_winsock.cpp new file mode 100644 index 0000000000..3fe7fc619e --- /dev/null +++ b/drivers/windows/net_socket_winsock.cpp @@ -0,0 +1,613 @@ +/**************************************************************************/ +/* net_socket_winsock.cpp */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ + +#ifdef WINDOWS_ENABLED + +#include "net_socket_winsock.h" + +#include <winsock2.h> +#include <ws2tcpip.h> + +#include <mswsock.h> +// Workaround missing flag in MinGW +#if defined(__MINGW32__) && !defined(SIO_UDP_NETRESET) +#define SIO_UDP_NETRESET _WSAIOW(IOC_VENDOR, 15) +#endif + +size_t NetSocketWinSock::_set_addr_storage(struct sockaddr_storage *p_addr, const IPAddress &p_ip, uint16_t p_port, IP::Type p_ip_type) { + memset(p_addr, 0, sizeof(struct sockaddr_storage)); + if (p_ip_type == IP::TYPE_IPV6 || p_ip_type == IP::TYPE_ANY) { // IPv6 socket. + + // IPv6 only socket with IPv4 address. + ERR_FAIL_COND_V(!p_ip.is_wildcard() && p_ip_type == IP::TYPE_IPV6 && p_ip.is_ipv4(), 0); + + struct sockaddr_in6 *addr6 = (struct sockaddr_in6 *)p_addr; + addr6->sin6_family = AF_INET6; + addr6->sin6_port = htons(p_port); + if (p_ip.is_valid()) { + memcpy(&addr6->sin6_addr.s6_addr, p_ip.get_ipv6(), 16); + } else { + addr6->sin6_addr = in6addr_any; + } + return sizeof(sockaddr_in6); + } else { // IPv4 socket. + + // IPv4 socket with IPv6 address. + ERR_FAIL_COND_V(!p_ip.is_wildcard() && !p_ip.is_ipv4(), 0); + + struct sockaddr_in *addr4 = (struct sockaddr_in *)p_addr; + addr4->sin_family = AF_INET; + addr4->sin_port = htons(p_port); // Short, network byte order. + + if (p_ip.is_valid()) { + memcpy(&addr4->sin_addr.s_addr, p_ip.get_ipv4(), 4); + } else { + addr4->sin_addr.s_addr = INADDR_ANY; + } + + return sizeof(sockaddr_in); + } +} + +void NetSocketWinSock::_set_ip_port(struct sockaddr_storage *p_addr, IPAddress *r_ip, uint16_t *r_port) { + if (p_addr->ss_family == AF_INET) { + struct sockaddr_in *addr4 = (struct sockaddr_in *)p_addr; + if (r_ip) { + r_ip->set_ipv4((uint8_t *)&(addr4->sin_addr.s_addr)); + } + if (r_port) { + *r_port = ntohs(addr4->sin_port); + } + } else if (p_addr->ss_family == AF_INET6) { + struct sockaddr_in6 *addr6 = (struct sockaddr_in6 *)p_addr; + if (r_ip) { + r_ip->set_ipv6(addr6->sin6_addr.s6_addr); + } + if (r_port) { + *r_port = ntohs(addr6->sin6_port); + } + } +} + +NetSocket *NetSocketWinSock::_create_func() { + return memnew(NetSocketWinSock); +} + +void NetSocketWinSock::make_default() { + ERR_FAIL_COND(_create != nullptr); + + WSADATA data; + WSAStartup(MAKEWORD(2, 2), &data); + _create = _create_func; +} + +void NetSocketWinSock::cleanup() { + ERR_FAIL_COND(_create == nullptr); + + WSACleanup(); + _create = nullptr; +} + +NetSocketWinSock::NetSocketWinSock() { +} + +NetSocketWinSock::~NetSocketWinSock() { + close(); +} + +NetSocketWinSock::NetError NetSocketWinSock::_get_socket_error() const { + int err = WSAGetLastError(); + if (err == WSAEISCONN) { + return ERR_NET_IS_CONNECTED; + } + if (err == WSAEINPROGRESS || err == WSAEALREADY) { + return ERR_NET_IN_PROGRESS; + } + if (err == WSAEWOULDBLOCK) { + return ERR_NET_WOULD_BLOCK; + } + if (err == WSAEADDRINUSE || err == WSAEADDRNOTAVAIL) { + return ERR_NET_ADDRESS_INVALID_OR_UNAVAILABLE; + } + if (err == WSAEACCES) { + return ERR_NET_UNAUTHORIZED; + } + if (err == WSAEMSGSIZE || err == WSAENOBUFS) { + return ERR_NET_BUFFER_TOO_SMALL; + } + print_verbose("Socket error: " + itos(err) + "."); + return ERR_NET_OTHER; +} + +bool NetSocketWinSock::_can_use_ip(const IPAddress &p_ip, const bool p_for_bind) const { + if (p_for_bind && !(p_ip.is_valid() || p_ip.is_wildcard())) { + return false; + } else if (!p_for_bind && !p_ip.is_valid()) { + return false; + } + // Check if socket support this IP type. + IP::Type type = p_ip.is_ipv4() ? IP::TYPE_IPV4 : IP::TYPE_IPV6; + return !(_ip_type != IP::TYPE_ANY && !p_ip.is_wildcard() && _ip_type != type); +} + +_FORCE_INLINE_ Error NetSocketWinSock::_change_multicast_group(IPAddress p_ip, String p_if_name, bool p_add) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + ERR_FAIL_COND_V(!_can_use_ip(p_ip, false), ERR_INVALID_PARAMETER); + + // Need to force level and af_family to IP(v4) when using dual stacking and provided multicast group is IPv4. + IP::Type type = _ip_type == IP::TYPE_ANY && p_ip.is_ipv4() ? IP::TYPE_IPV4 : _ip_type; + // This needs to be the proper level for the multicast group, no matter if the socket is dual stacking. + int level = type == IP::TYPE_IPV4 ? IPPROTO_IP : IPPROTO_IPV6; + int ret = -1; + + IPAddress if_ip; + uint32_t if_v6id = 0; + HashMap<String, IP::Interface_Info> if_info; + IP::get_singleton()->get_local_interfaces(&if_info); + for (KeyValue<String, IP::Interface_Info> &E : if_info) { + IP::Interface_Info &c = E.value; + if (c.name != p_if_name) { + continue; + } + + if_v6id = (uint32_t)c.index.to_int(); + if (type == IP::TYPE_IPV6) { + break; // IPv6 uses index. + } + + for (const IPAddress &F : c.ip_addresses) { + if (!F.is_ipv4()) { + continue; // Wrong IP type. + } + if_ip = F; + break; + } + break; + } + + if (level == IPPROTO_IP) { + ERR_FAIL_COND_V(!if_ip.is_valid(), ERR_INVALID_PARAMETER); + struct ip_mreq greq; + int sock_opt = p_add ? IP_ADD_MEMBERSHIP : IP_DROP_MEMBERSHIP; + memcpy(&greq.imr_multiaddr, p_ip.get_ipv4(), 4); + memcpy(&greq.imr_interface, if_ip.get_ipv4(), 4); + ret = setsockopt(_sock, level, sock_opt, (const char *)&greq, sizeof(greq)); + } else { + struct ipv6_mreq greq; + int sock_opt = p_add ? IPV6_ADD_MEMBERSHIP : IPV6_DROP_MEMBERSHIP; + memcpy(&greq.ipv6mr_multiaddr, p_ip.get_ipv6(), 16); + greq.ipv6mr_interface = if_v6id; + ret = setsockopt(_sock, level, sock_opt, (const char *)&greq, sizeof(greq)); + } + ERR_FAIL_COND_V(ret != 0, FAILED); + + return OK; +} + +void NetSocketWinSock::_set_socket(SOCKET p_sock, IP::Type p_ip_type, bool p_is_stream) { + _sock = p_sock; + _ip_type = p_ip_type; + _is_stream = p_is_stream; +} + +Error NetSocketWinSock::open(Type p_sock_type, IP::Type &ip_type) { + ERR_FAIL_COND_V(is_open(), ERR_ALREADY_IN_USE); + ERR_FAIL_COND_V(ip_type > IP::TYPE_ANY || ip_type < IP::TYPE_NONE, ERR_INVALID_PARAMETER); + + int family = ip_type == IP::TYPE_IPV4 ? AF_INET : AF_INET6; + int protocol = p_sock_type == TYPE_TCP ? IPPROTO_TCP : IPPROTO_UDP; + int type = p_sock_type == TYPE_TCP ? SOCK_STREAM : SOCK_DGRAM; + _sock = socket(family, type, protocol); + + if (_sock == INVALID_SOCKET && ip_type == IP::TYPE_ANY) { + // Careful here, changing the referenced parameter so the caller knows that we are using an IPv4 socket + // in place of a dual stack one, and further calls to _set_sock_addr will work as expected. + ip_type = IP::TYPE_IPV4; + family = AF_INET; + _sock = socket(family, type, protocol); + } + + ERR_FAIL_COND_V(_sock == INVALID_SOCKET, FAILED); + _ip_type = ip_type; + + if (family == AF_INET6) { + // Select IPv4 over IPv6 mapping. + set_ipv6_only_enabled(ip_type != IP::TYPE_ANY); + } + + if (protocol == IPPROTO_UDP) { + // Make sure to disable broadcasting for UDP sockets. + // Depending on the OS, this option might or might not be enabled by default. Let's normalize it. + set_broadcasting_enabled(false); + } + + _is_stream = p_sock_type == TYPE_TCP; + + if (!_is_stream) { + // Disable windows feature/bug reporting WSAECONNRESET/WSAENETRESET when + // recv/recvfrom and an ICMP reply was received from a previous send/sendto. + unsigned long disable = 0; + if (ioctlsocket(_sock, SIO_UDP_CONNRESET, &disable) == SOCKET_ERROR) { + print_verbose("Unable to turn off UDP WSAECONNRESET behavior on Windows."); + } + if (ioctlsocket(_sock, SIO_UDP_NETRESET, &disable) == SOCKET_ERROR) { + // This feature seems not to be supported on wine. + print_verbose("Unable to turn off UDP WSAENETRESET behavior on Windows."); + } + } + return OK; +} + +void NetSocketWinSock::close() { + if (_sock != INVALID_SOCKET) { + closesocket(_sock); + } + + _sock = INVALID_SOCKET; + _ip_type = IP::TYPE_NONE; + _is_stream = false; +} + +Error NetSocketWinSock::bind(IPAddress p_addr, uint16_t p_port) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + ERR_FAIL_COND_V(!_can_use_ip(p_addr, true), ERR_INVALID_PARAMETER); + + sockaddr_storage addr; + size_t addr_size = _set_addr_storage(&addr, p_addr, p_port, _ip_type); + + if (::bind(_sock, (struct sockaddr *)&addr, addr_size) != 0) { + NetError err = _get_socket_error(); + print_verbose("Failed to bind socket. Error: " + itos(err) + "."); + close(); + return ERR_UNAVAILABLE; + } + + return OK; +} + +Error NetSocketWinSock::listen(int p_max_pending) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + + if (::listen(_sock, p_max_pending) != 0) { + _get_socket_error(); + print_verbose("Failed to listen from socket."); + close(); + return FAILED; + } + + return OK; +} + +Error NetSocketWinSock::connect_to_host(IPAddress p_host, uint16_t p_port) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + ERR_FAIL_COND_V(!_can_use_ip(p_host, false), ERR_INVALID_PARAMETER); + + struct sockaddr_storage addr; + size_t addr_size = _set_addr_storage(&addr, p_host, p_port, _ip_type); + + if (::WSAConnect(_sock, (struct sockaddr *)&addr, addr_size, nullptr, nullptr, nullptr, nullptr) != 0) { + NetError err = _get_socket_error(); + + switch (err) { + // We are already connected. + case ERR_NET_IS_CONNECTED: + return OK; + // Still waiting to connect, try again in a while. + case ERR_NET_WOULD_BLOCK: + case ERR_NET_IN_PROGRESS: + return ERR_BUSY; + default: + print_verbose("Connection to remote host failed."); + close(); + return FAILED; + } + } + + return OK; +} + +Error NetSocketWinSock::poll(PollType p_type, int p_timeout) const { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + + bool ready = false; + fd_set rd, wr, ex; + fd_set *rdp = nullptr; + fd_set *wrp = nullptr; + FD_ZERO(&rd); + FD_ZERO(&wr); + FD_ZERO(&ex); + FD_SET(_sock, &ex); + struct timeval timeout = { p_timeout / 1000, (p_timeout % 1000) * 1000 }; + // For blocking operation, pass nullptr timeout pointer to select. + struct timeval *tp = nullptr; + if (p_timeout >= 0) { + // If timeout is non-negative, we want to specify the timeout instead. + tp = &timeout; + } + + switch (p_type) { + case POLL_TYPE_IN: + FD_SET(_sock, &rd); + rdp = &rd; + break; + case POLL_TYPE_OUT: + FD_SET(_sock, &wr); + wrp = ≀ + break; + case POLL_TYPE_IN_OUT: + FD_SET(_sock, &rd); + FD_SET(_sock, &wr); + rdp = &rd; + wrp = ≀ + } + // WSAPoll is broken: https://daniel.haxx.se/blog/2012/10/10/wsapoll-is-broken/. + int ret = select(1, rdp, wrp, &ex, tp); + + if (ret == SOCKET_ERROR) { + return FAILED; + } + + if (ret == 0) { + return ERR_BUSY; + } + + if (FD_ISSET(_sock, &ex)) { + _get_socket_error(); + print_verbose("Exception when polling socket."); + return FAILED; + } + + if (rdp && FD_ISSET(_sock, rdp)) { + ready = true; + } + if (wrp && FD_ISSET(_sock, wrp)) { + ready = true; + } + + return ready ? OK : ERR_BUSY; +} + +Error NetSocketWinSock::recv(uint8_t *p_buffer, int p_len, int &r_read) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + + r_read = ::recv(_sock, (char *)p_buffer, p_len, 0); + + if (r_read < 0) { + NetError err = _get_socket_error(); + if (err == ERR_NET_WOULD_BLOCK) { + return ERR_BUSY; + } + + if (err == ERR_NET_BUFFER_TOO_SMALL) { + return ERR_OUT_OF_MEMORY; + } + + return FAILED; + } + + return OK; +} + +Error NetSocketWinSock::recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddress &r_ip, uint16_t &r_port, bool p_peek) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + + struct sockaddr_storage from; + socklen_t len = sizeof(struct sockaddr_storage); + memset(&from, 0, len); + + r_read = ::recvfrom(_sock, (char *)p_buffer, p_len, p_peek ? MSG_PEEK : 0, (struct sockaddr *)&from, &len); + + if (r_read < 0) { + NetError err = _get_socket_error(); + if (err == ERR_NET_WOULD_BLOCK) { + return ERR_BUSY; + } + + if (err == ERR_NET_BUFFER_TOO_SMALL) { + return ERR_OUT_OF_MEMORY; + } + + return FAILED; + } + + if (from.ss_family == AF_INET) { + struct sockaddr_in *sin_from = (struct sockaddr_in *)&from; + r_ip.set_ipv4((uint8_t *)&sin_from->sin_addr); + r_port = ntohs(sin_from->sin_port); + } else if (from.ss_family == AF_INET6) { + struct sockaddr_in6 *s6_from = (struct sockaddr_in6 *)&from; + r_ip.set_ipv6((uint8_t *)&s6_from->sin6_addr); + r_port = ntohs(s6_from->sin6_port); + } else { + // Unsupported socket family, should never happen. + ERR_FAIL_V(FAILED); + } + + return OK; +} + +Error NetSocketWinSock::send(const uint8_t *p_buffer, int p_len, int &r_sent) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + + int flags = 0; + r_sent = ::send(_sock, (const char *)p_buffer, p_len, flags); + + if (r_sent < 0) { + NetError err = _get_socket_error(); + if (err == ERR_NET_WOULD_BLOCK) { + return ERR_BUSY; + } + if (err == ERR_NET_BUFFER_TOO_SMALL) { + return ERR_OUT_OF_MEMORY; + } + + return FAILED; + } + + return OK; +} + +Error NetSocketWinSock::sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IPAddress p_ip, uint16_t p_port) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + + struct sockaddr_storage addr; + size_t addr_size = _set_addr_storage(&addr, p_ip, p_port, _ip_type); + r_sent = ::sendto(_sock, (const char *)p_buffer, p_len, 0, (struct sockaddr *)&addr, addr_size); + + if (r_sent < 0) { + NetError err = _get_socket_error(); + if (err == ERR_NET_WOULD_BLOCK) { + return ERR_BUSY; + } + if (err == ERR_NET_BUFFER_TOO_SMALL) { + return ERR_OUT_OF_MEMORY; + } + + return FAILED; + } + + return OK; +} + +Error NetSocketWinSock::set_broadcasting_enabled(bool p_enabled) { + ERR_FAIL_COND_V(!is_open(), ERR_UNCONFIGURED); + // IPv6 has no broadcast support. + if (_ip_type == IP::TYPE_IPV6) { + return ERR_UNAVAILABLE; + } + + int par = p_enabled ? 1 : 0; + if (setsockopt(_sock, SOL_SOCKET, SO_BROADCAST, (const char *)&par, sizeof(int)) != 0) { + WARN_PRINT("Unable to change broadcast setting."); + return FAILED; + } + return OK; +} + +void NetSocketWinSock::set_blocking_enabled(bool p_enabled) { + ERR_FAIL_COND(!is_open()); + + int ret = 0; + unsigned long par = p_enabled ? 0 : 1; + ret = ioctlsocket(_sock, FIONBIO, &par); + if (ret != 0) { + WARN_PRINT("Unable to change non-block mode."); + } +} + +void NetSocketWinSock::set_ipv6_only_enabled(bool p_enabled) { + ERR_FAIL_COND(!is_open()); + // This option is only available in IPv6 sockets. + ERR_FAIL_COND(_ip_type == IP::TYPE_IPV4); + + int par = p_enabled ? 1 : 0; + if (setsockopt(_sock, IPPROTO_IPV6, IPV6_V6ONLY, (const char *)&par, sizeof(int)) != 0) { + WARN_PRINT("Unable to change IPv4 address mapping over IPv6 option."); + } +} + +void NetSocketWinSock::set_tcp_no_delay_enabled(bool p_enabled) { + ERR_FAIL_COND(!is_open()); + ERR_FAIL_COND(!_is_stream); // Not TCP. + + int par = p_enabled ? 1 : 0; + if (setsockopt(_sock, IPPROTO_TCP, TCP_NODELAY, (const char *)&par, sizeof(int)) < 0) { + WARN_PRINT("Unable to set TCP no delay option."); + } +} + +void NetSocketWinSock::set_reuse_address_enabled(bool p_enabled) { + ERR_FAIL_COND(!is_open()); + + // On Windows, enabling SO_REUSEADDR actually would also enable reuse port, very bad on TCP. Denying... + // Windows does not have this option, SO_REUSEADDR in this magical world means SO_REUSEPORT +} + +bool NetSocketWinSock::is_open() const { + return _sock != INVALID_SOCKET; +} + +int NetSocketWinSock::get_available_bytes() const { + ERR_FAIL_COND_V(!is_open(), -1); + + unsigned long len; + int ret = ioctlsocket(_sock, FIONREAD, &len); + if (ret == -1) { + _get_socket_error(); + print_verbose("Error when checking available bytes on socket."); + return -1; + } + return len; +} + +Error NetSocketWinSock::get_socket_address(IPAddress *r_ip, uint16_t *r_port) const { + ERR_FAIL_COND_V(!is_open(), FAILED); + + struct sockaddr_storage saddr; + socklen_t len = sizeof(saddr); + if (getsockname(_sock, (struct sockaddr *)&saddr, &len) != 0) { + _get_socket_error(); + print_verbose("Error when reading local socket address."); + return FAILED; + } + _set_ip_port(&saddr, r_ip, r_port); + return OK; +} + +Ref<NetSocket> NetSocketWinSock::accept(IPAddress &r_ip, uint16_t &r_port) { + Ref<NetSocket> out; + ERR_FAIL_COND_V(!is_open(), out); + + struct sockaddr_storage their_addr; + socklen_t size = sizeof(their_addr); + SOCKET fd = ::accept(_sock, (struct sockaddr *)&their_addr, &size); + if (fd == INVALID_SOCKET) { + _get_socket_error(); + print_verbose("Error when accepting socket connection."); + return out; + } + + _set_ip_port(&their_addr, &r_ip, &r_port); + + NetSocketWinSock *ns = memnew(NetSocketWinSock); + ns->_set_socket(fd, _ip_type, _is_stream); + ns->set_blocking_enabled(false); + return Ref<NetSocket>(ns); +} + +Error NetSocketWinSock::join_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) { + return _change_multicast_group(p_multi_address, p_if_name, true); +} + +Error NetSocketWinSock::leave_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) { + return _change_multicast_group(p_multi_address, p_if_name, false); +} + +#endif // WINDOWS_ENABLED diff --git a/drivers/windows/net_socket_winsock.h b/drivers/windows/net_socket_winsock.h new file mode 100644 index 0000000000..5c3445b8cb --- /dev/null +++ b/drivers/windows/net_socket_winsock.h @@ -0,0 +1,102 @@ +/**************************************************************************/ +/* net_socket_winsock.h */ +/**************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/**************************************************************************/ +/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */ +/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/**************************************************************************/ + +#ifndef NET_SOCKET_WINSOCK_H +#define NET_SOCKET_WINSOCK_H + +#ifdef WINDOWS_ENABLED + +#include "core/io/net_socket.h" + +#include <winsock2.h> +#include <ws2tcpip.h> + +class NetSocketWinSock : public NetSocket { +private: + SOCKET _sock = INVALID_SOCKET; + IP::Type _ip_type = IP::TYPE_NONE; + bool _is_stream = false; + + enum NetError { + ERR_NET_WOULD_BLOCK, + ERR_NET_IS_CONNECTED, + ERR_NET_IN_PROGRESS, + ERR_NET_ADDRESS_INVALID_OR_UNAVAILABLE, + ERR_NET_UNAUTHORIZED, + ERR_NET_BUFFER_TOO_SMALL, + ERR_NET_OTHER, + }; + + NetError _get_socket_error() const; + void _set_socket(SOCKET p_sock, IP::Type p_ip_type, bool p_is_stream); + _FORCE_INLINE_ Error _change_multicast_group(IPAddress p_ip, String p_if_name, bool p_add); + +protected: + static NetSocket *_create_func(); + + bool _can_use_ip(const IPAddress &p_ip, const bool p_for_bind) const; + +public: + static void make_default(); + static void cleanup(); + static void _set_ip_port(struct sockaddr_storage *p_addr, IPAddress *r_ip, uint16_t *r_port); + static size_t _set_addr_storage(struct sockaddr_storage *p_addr, const IPAddress &p_ip, uint16_t p_port, IP::Type p_ip_type); + + virtual Error open(Type p_sock_type, IP::Type &ip_type) override; + virtual void close() override; + virtual Error bind(IPAddress p_addr, uint16_t p_port) override; + virtual Error listen(int p_max_pending) override; + virtual Error connect_to_host(IPAddress p_host, uint16_t p_port) override; + virtual Error poll(PollType p_type, int timeout) const override; + virtual Error recv(uint8_t *p_buffer, int p_len, int &r_read) override; + virtual Error recvfrom(uint8_t *p_buffer, int p_len, int &r_read, IPAddress &r_ip, uint16_t &r_port, bool p_peek = false) override; + virtual Error send(const uint8_t *p_buffer, int p_len, int &r_sent) override; + virtual Error sendto(const uint8_t *p_buffer, int p_len, int &r_sent, IPAddress p_ip, uint16_t p_port) override; + virtual Ref<NetSocket> accept(IPAddress &r_ip, uint16_t &r_port) override; + + virtual bool is_open() const override; + virtual int get_available_bytes() const override; + virtual Error get_socket_address(IPAddress *r_ip, uint16_t *r_port) const override; + + virtual Error set_broadcasting_enabled(bool p_enabled) override; + virtual void set_blocking_enabled(bool p_enabled) override; + virtual void set_ipv6_only_enabled(bool p_enabled) override; + virtual void set_tcp_no_delay_enabled(bool p_enabled) override; + virtual void set_reuse_address_enabled(bool p_enabled) override; + virtual Error join_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) override; + virtual Error leave_multicast_group(const IPAddress &p_multi_address, const String &p_if_name) override; + + NetSocketWinSock(); + ~NetSocketWinSock() override; +}; + +#endif // WINDOWS_ENABLED + +#endif // NET_SOCKET_WINSOCK_H |