* fix triangle batch
since it would be too expensive to check the uniforms, simplify enable batching for built-in program.
* enable user update the materialID
* remove deprecated files
* remove some deprecated codes
* remove more deprecated codes
* remove ui deprecated codes
* remove more deprecated codes
* remove deprecated codes in ccmenuitem
* remove more deprecated codes in ui
* remove more deprecated codes in ui
* remove more deprecated codes in ui
* remove more deprecated codes
* remove more deprecated codes
* remove more deprecated codes
* remove vr related codes and ignore some modules
* remove allocator
* remove some config
* 【Feature】add back-end project file
* [Feature] add back-end file
* add pipeline descriptor and shader cache
* [Feature] support sprite for backend
* [Feature] remove unneeded code
* [Feature] according to es2.0 spec, you must use clamp-to-edge as texture wrap mode, and no mipmapping for non-power-of-two texture
* [Feature] set texture wrap mode to clamp-to-edge, and no mipmapping for non-power-of-two texture
* [Feature] remove macro define to .cpp file
* [Feature] add log info
* [Feature] add PipelineDescriptor for TriangleCommand
* [Feature] add PipelineDescriptor object as member of TriangleCommand
* [Feature] add getPipelineDescriptor method
* add renderbackend
* complete pipeline descriptor
* [Feature] add viewport in RenderCommand
* set viewport when rendrering
* [Feature] occur error when using RendererBackend, to be fixed.
* a workaround to fix black screen on macOS 10.14 (#19090)
* add rendererbackend init function
* fix typo
* [Feature] modify testFile
* [BugFix] modify shader path
* [Feature] set default viewport
* fix projection
* [Feature] modify log info
* [BugFix] change viewport data type to int
* [BugFix] add BindGroup to PipelienDescriptor
* [BugFix] change a_position to vec3 in sprite.vert
* [BugFix] set vertexLayout according to V3F_C4B_T2F structure
* [Feature] revert a_position to vec4
* [Feature] renderer should not use gl codes directly
* [Feature] it's better not use default value parameter
* fix depth test setting
* rendererbackend -> renderer
* clear color and depth at begin
* add metal backend
* metal support normalized attribute
* simplify codes
* update external
* add render pass desctriptor in pipeline descriptor
* fix warnings
* fix crash and memeory leak
* refactor Texture2D
* put pipeline descriptor into render command
* simplify codes
* [Feature] update Sprite
* fix crash when closing app
* [Feature] update SpriteBatchNode and TextureAtlas
* support render texture(not finish)
* [Feature] remove unused code
* make tests work on mac
* fix download-deps path error
* make tests work on iOS
* [Feature] support ttf under normal label effect
* refactor triangle command processing
* let renderer handle more common commands
* refactor backend
* make render texture work
* [Feature] refactor backend for GL
* [Feature]Renaming to make it easy to understand
* [Feature] change warp mode to CLAMP_TO_EDGE
* fix ghost
* simplify visit render queue logic
* support progress timer without rial mode
* support partcile system
* Feature/update label (#149)
* [BugFix] fix compile error
* [Feature] support outline effect in ios
* [Feature] add shader file
* [BugFix] fix begin and end RenderPass
* [Feature] update CustomCommand
* [Feature] revert project.pbxproj
* [Feature] simplify codes
* [BugFix] pack AI88 to RGBA8888 only when outline enable
* [Feature] support shadow effect in Label
* [Feature] support BMFont
* [Feature] support glow effect
* [Feature] simplify shader files
* LabelAtlas work
* handle blend function correctly
* support tile map
* don't share buffer in metal
* alloc buffer size as needed
* support more tilemap
* Merge branch 'minggo/metal-support' into feature/updateLabel
* minggo/metal-support:
support tile map
handle blend function correctly
LabelAtlas work
Feature/update label (#149)
support partcile system
# Conflicts:
# cocos/2d/CCLabel.cpp
# cocos/2d/CCSprite.cpp
# cocos/2d/CCSpriteBatchNode.cpp
# cocos/renderer/CCQuadCommand.cpp
# cocos/renderer/CCQuadCommand.h
* render texture work without saving file
* use global viewport
* grid3d works
* remove grabber
* tiled3d works
* [BugFix] fix label bug
* [Feature] add updateSubData for buffer
* [Feature] remove setVertexCount
* support depth test
* add callback command
* [Feature] add UITest
* [Feature] update UITest
* [Feature] remove unneeded codes
* fix custom command issue
* fix layer color blend issue
* [BugFix] fix iOS compile error
* [Feature] remove unneeded codes
* [Feature] fix updateVertexBuffer
* layerradial works
* add draw test back
* fix batch issue
* fix compiling error
* [BugFix] support ETC1
* [BugFix] get the correct pipelineDescriptor
* [BugFix] skip draw when backendTexture nullptr
* clipping node support
* [Feature] add shader files
* fix stencil issue in metal
* [Feature] update UILayoutTest
* [BugFix] skip drawing when vertexCount is zero
* refactor renderer
* add set global z order for stencil manager commands
* fix warnings caused by type
* remove viewport in render command
* [Feature] fix warnings caused by type
* [BugFix] clear vertexCount and indexCount for CustomComand when needed
* [Feature] update clear for CustomCommand
* ios use metal
* fix viewport issue
* fix LayerColorGradient crash
* [cmake] transport to android and windows (#160)
* save point 1
* compile on windows
* run on android
* revert useless change
* android set CC_ENABLE_CACHE_TEXTURE_DATA to 1
* add initGlew
* fix android crash
* add TODO new-renderer
* review update
* revert onGLFWWindowPosCallback
* fix android compiling error
* Impl progress radial (#162)
* progresstimer add radial impl
* default drawType to element
* dec invoke times of createVertexBuffer (#163)
* support depth/stencil format for gl backend
* simplify progress timer codes
* support motionstreak, effect is wrong
* fix motionstreak issue
* [Feature] update Scissor Test (#161)
* [Feature] update Scissor Test
* [Feature] update ScissorTest
* [Feature] rename function
* [Feature] get constant reference if needed
* [Feature] show render status (#164)
* improve performance
* fix depth state
* fill error that triangle vertex/index number bigger than buffer
* fix compiline error in release mode
* fix buffer conflict between CPU and GPU on iOS/macOS
* Renderer refactor (#165)
* use one vertes/index buffer with opengl
* fix error on windows
* custom command support index format config
* CCLayer: compact vertex data structure
* update comment
* fix doc
* support fast tilemap
* pass index format instead
* fix some wrong effect
* fix render texture error
* fix texture per-element size
* fix texture format error
* BlendFunc type refactor, GLenum -> backend::BlendFactor (#167)
* BlendFunc use backend::BlendFactor as inner field
* update comments
* use int to replace GLenum
* update xcode project fiel
* rename to GLBlendConst
* add ccConstants.h
* update xcode project file
* update copyright
* remove primitive command
* remove CCPrimitive.cpp/.h
* remove deprecated files
* remove unneeded files
* remove multiple view support
* remove multiple view support
* remove the usage of frame buffer in camera
* director don't use frame buffer
* remove FrameBuffer
* remove BatchCommand
* add some api reference
* add physics2d back
* fix crash when close app on mac
* improve render texture
* fix rendertexture issue
* fix rendertexture issue
* simplify codes
* CMake support for mac & ios (#169)
* update cmake
* fix compile error
* update 3rd libs version
* remove CCThread.h/.cpp
* remove ccthread
* use audio engine to implement simple audio engine
* remove unneeded codes
* remove deprecated codes
* remove winrt macro
* remove CC_USE_WIC
* set partcile blend function in more elegant way
* remove unneeded codes
* remove unneeded codes
* cmake works on windows
* update project setting
* improve performance
* GLFloat -> float
* sync v3 cmake improvements into metal-support (#172)
* pick: modern cmake, compile definitions improvement (#19139)
* modern cmake, use target_compile_definitions partly
* simplify macro define, remove USE_*
* modern cmake, macro define
* add physics 2d macro define into ccConfig.h
* remove USE_CHIPMUNK macro in build.gradle
* remove CocosSelectModule.cmake
* shrink useless define
* simplify compile options config, re-add if necessary
* update external for tmp CI test
* un-quote target_compile_options value
* add "-g" parameter only when debug mode
* keep single build type when generator Xcode & VS projecy
* update external for tmp CI tes
* add static_cast<char>(-1), fix -Wc++11-narrowing
* simplify win32 compile define
* not modify code, only improve compile options
# Conflicts:
# .gitignore
# cmake/Modules/CocosConfigDepend.cmake
# cocos/CMakeLists.txt
# external/config.json
# tests/cpp-tests/CMakeLists.txt
* modern cmake, improve cmake_compiler_flags (#19145)
* cmake_compiler_flags
* Fix typo
* Fix typo2
* Remove chanages from Android.mk
* correct lua template cmake build (#19149)
* don't add -Wno-deprecated into jsb target
* correct lua template cmake build
* fix win32 lua template compile error
* prevent cmake in-source-build friendly (#19151)
* pick: Copy resources to "Resources/" on win32 like in linux configuration
* add "/Z7" for cpp-tests on windows
* [cmake] fix iOS xcode property setting failed (#19208)
* fix iOS xcode property setting failed
* use search_depend_libs_recursive at dlls collect
* fix typo
* [cmake] add find_host_library into iOS toolchain file (#19230)
* pick: [lua android] use luajit & template cmake update (#19239)
* increase cmake stability , remove tests/CMakeLists.txt (#19261)
* cmake win32 Precompiled header (#19273)
* Precompiled header
* Fix
* Precompiled header for cocos
* Precompiled header jscocos2d
* Fix for COCOS2D_DEBUG is always 1 on Android (#19291)
Related #19289
* little build fix, tests cpp-tests works on mac
* sync v3 build related codes into metal-support (#173)
* strict initialization for std::array
* remove proj.win32 project configs
* modern cmake, cmake_cleanup_remove_unused_variables (#19146)
* Switch travis CI to xenial (#19207)
* Switch travis CI to xenial
* Remove language: android
* Set language: cpp
* Fix java problem
* Update sdkmanager
* Fix sdkmanger
* next sdkmanager fix
* Remove xenial from android
* revert to sdk-tools-{system}-3859397
* Remove linux cmake install
* Update before-install.sh
* Update .travis.yml
* Simplify install-deps-linux.sh, tested on Ubuntu 16.04 (#19212)
* Simplify install-deps-linux.sh
* Cleanup
* pick: install ninja
* update cocos2d-console submodule
* for metal-support alpha release, we only test cpp
* add HelloCpp into project(Cocos2d-x) for tmp test
* update extenal metal-support-4
* update uniform setting
* [Feature] update BindGroup
* [Feature] empty-test
* [Feature] cpp-test
* [Feature] fix GL compiler error
* [Feature] fix GL crash
* [Feature] empty-test
* [Feature] cpp-tests
* [feature] improve frameRate
* [feature] fix opengl compile error
* [feature] fix opengl compile error
* [BugFix] fix compute maxLocation error
* [Feature] update setting unifrom
* [Feature] fix namespace
* [Feature] remove unneeded code
* [Bugfix] fix project file
* [Feature] update review
* [texture2d] impl texture format support (#175)
* texture update
* update
* update texture
* commit
* compile on windows
* ddd
* rename
* rename methods
* no crash
* save gl
* save
* save
* rename
* move out pixel format convert functions
* metal crash
* update
* update android
* support gles compressed texture format
* support more compress format
* add more conversion methods
* ss
* save
* update conversion methods
* add PVRTC format support
* reformat
* add marco linux
* fix GL marcro
* pvrtc supported only by ios 8.0+
* remove unused cmake
* revert change
* refactor Texture2D::initWithData
* fix conversion log
* refactor Texture2D::initWithData
* remove some OpenGL constants for PVRTC
* add todo
* fix typo
* AutoTest works on mac/iOS by disable part cases, sync v3 bug fix (#174)
* review cpp-tests, and fix part issues on start auto test
* sync png format fix: Node:Particle3D abnormal texture effects #19204
* fix cpp-tests SpritePolygon crash, wrong png format (#19170)
* fix wrong png convert format from sRGB to Gray
* erase plist index if all frames was erased
* test_A8.png have I8 format, fix it
* [CCSpriteCache] allow re-add plist & add testcase (#19175)
* allow re-add plist & add testcase
* remove comments/rename method/update testcase
* fix isSpriteFramesWithFileLoaded & add testcase
* remove used variable
* remove unused variable
* fix double free issues when js/lua-tests exit on iOS (#19236)
* disable part cases, AutoTest works without crash on mac
* update cocos2dx files json, to test cocos new next
* fix spritecache plist parsing issue (#19269)
* [linux] Fix FileUtils::getContents with folder (#19157)
* fix FileUtils::getContents on linux/mac
* use stat.st_mode
* simplify
* [CCFileUtils] win32 getFileSize (#19176)
* win32 getFileSize
* fix stat
* [cpp test-Android]20：FileUtils/2 change title (#19197)
* sync #19200
* sync #19231
* [android lua] improve performance of lua loader (#19234)
* [lua] improve performance of lua loader
* remove cache fix
* Revert "fix spritecache plist parsing issue (#19269)"
This reverts commit f3a85ece4307a7b90816c34489d1ed2c8fd11baf.
* remove win32 project files ref in template.json
* add metal framework lnk ref into cpp template
* test on iOS, and disable part cases
* alBufferData instead of alBufferDataStatic for small audio file on Apple (#19227)
* changes AudioCache to use alBufferData instead of alBufferDataStatic
(also makes test 19 faster to trigger openal bugs faster)
The original problem: CrashIfClientProvidedBogusAudioBufferList
https://github.com/cocos2d/cocos2d-x/issues/18948
is not happening anymore, but there's still a not very frequent issue
that makes OpenAL crash with a call stack like this.
AudioCache::readDataTask > alBufferData > CleanUpDeadBufferList
It happes more frequently when the device is "cold", which means after
half an hour of not using the device (locked).
I could not find the actual source code for iOS OpenAL, so I used the
macOS versions:
https://opensource.apple.com/source/OpenAL/OpenAL-48.7/Source/OpenAL/oalImp.cpp.auto.html
They seem to use CAGuard.h to make sure the dead buffer list
has no threading issues. I'm worried because the CAGuard code I found
has macos and win32 define but no iOS, so I'm not sure. I guess the
iOS version is different and has the guard.
I could not find a place in the code that's unprotected by the locks
except the InitializeBufferMap() which should not be called more than
once from cocos, and there's a workaround in AudioEngine-impl for it.
I reduced the occurence of the CleanUpDeadBufferList crash by moving
the guard in ~AudioCache to cover the alDeleteBuffers call.
* remove hack method "setTimeout" on audio
* AutoTest works on iOS
* support set ios deployment target for root project
* enable all texture2d cases, since Jiang have fixed
* add CCTextureUtils to xcode project file (#176)
* add leak cases for SpriteFrameCache (#177)
* re-add SpriteFrameCache cases
* update template file json
* Update SpriteFrameCacheTest.cpp
* fix compiling error
* Copyright use python script updated. might be fix by manual later
* Revert "Copyright use python script updated. might be fix by manual later"
This reverts commit 49e99418e4da1fd02afda448ddeb16210f5e8c71.
* re modify copyright, consider utf-8-sig encoding
* another situation, add Copyright before "Permission is hereby granted"
* Revert "another situation, add Copyright before "Permission is hereby granted""
This reverts commit ee82591d32353c7ce2e146fe51ef447433b47571.
* another situation, add Copyright before Permission is hereby granted
* change "Copyright (c) 2016-2016 Chukong Technologies Inc." to " Copyright (c) 2016 Chukong Technologies Inc."
* script modify copyright, consider is cocos copyright or not
* change "Copyright (c) 2017 Chukong Technologies Inc." to "Copyright (c) 2016 Chukong Technologies Inc."
* manual fix, not fix audio related
* change "2016-2016" to "2016"
* fix: TriangleCommands with custom uniforms can be batched
TriangleCommands with custom uniforms can be batched together.
This improves the performance when using custom uniforms without
adding any penalties when not using them
Github issue #16224
* better tests
- more options for batching (faster)
- less flushing in between commands (faster)
- less renderer code (easier to maintain)
- less VBO/VAO (less memory)
- uses temporary buffer for faster rendering (more memory)
- QuadCommand is just a subclass of TriangleCommand (simpler)
- but statically it allocates indices internally (like any TriangleCommand) (more
memory, but not that much)
- adds more tests for SpritePolygon
- adds test for batching Tri + Quad command
- and fixes some other bugs
Squashed commit of the following:
commit cce893a9114da5dda91a895e4e661894f365e2c7
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Mon Feb 8 06:56:57 2016 -0800
QuadCommand: indices are static
thanks @Darinex
commit e5fd1c14e93d2f07b8266bfec03cf7206d348f59
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Sun Feb 7 19:18:42 2016 -0800
little improvements in xcode proj management
commit 6cd29764e01ad300f4e89e62aa6b399f73e94f01
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Sun Feb 7 19:05:52 2016 -0800
Renderer improvments: TriangleCommand and QuadCommand merged
- more options for batching (faster)
- less flushing in between commands (faster)
- less renderer code (easier to maintain)
- less VBO/VAO (less memory)
- uses temporary buffer for faster rendering (more memory)
and fixes some other bugs
Squashed commit of the following:
commit af7a209435c5f2430c31ab3ebcc7f45051eacb1a
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Sun Feb 7 18:55:27 2016 -0800
adds test for batching quads and tris
commit 25816e3eaeffd4b0cfe5fa15e97ff84e9013b8be
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Sun Feb 7 18:40:15 2016 -0800
cleanup renderer
commit 7f9785056681756c865d603516c7e14c46040e82
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Sun Feb 7 15:34:48 2016 -0800
clean up docs
commit 92a8b679dd404e4ca2839d85e2bd925195b12996
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Sun Feb 7 15:18:52 2016 -0800
QuadCommand converted to Triangles command
commit 7e21ec12d3781f6964b4d639cfa4bdc79eb8ae80
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Fri Feb 5 16:17:29 2016 -0800
adds missing files... new test as well
commit d1bd4391785515bdc2e1bebbeb3f321f983b51f2
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Fri Feb 5 14:35:41 2016 -0800
more tests for poly sprites
commit 5775a25b1046671bd0f5af4c5ab52b308b34bb39
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Fri Feb 5 00:30:13 2016 -0800
malloc instead of std::vector
commit 14a5d45cc978c55a182c03f181ca2bc603776561
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Thu Feb 4 21:40:46 2016 -0800
little optimization
commit 5667cb2b62416f8a439a54c3973d5841fbe1b68b
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Thu Feb 4 19:13:13 2016 -0800
quad works as expected
commit 462aef05baa6586591882faa366839bead4be068
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Thu Feb 4 15:22:59 2016 -0800
works
commit c789e6c42811278563c7682b5620296ae8d84aeb
Merge: d6c3729 de62924
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Thu Feb 4 14:08:49 2016 -0800
Merge branch 'v3' into uniform_matrix_auto
commit d6c37297954af752ebf42ec5f71e5e2fe28c9cc2
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Thu Feb 4 13:54:07 2016 -0800
works ok...
commit 45c288631989ea735cc31ec5aa683d9d41aacab3
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Thu Feb 4 12:58:34 2016 -0800
quad works!
commit 7137cff51df09adee7129738cea2b69da090756e
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Wed Feb 3 16:19:34 2016 -0800
more optimizations... WIP... doesn't work yet
commit a8324c1a9acb6b82102e05eee6e5e5f341306c7b
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Tue Feb 2 21:59:42 2016 -0800
compiles in release mode
more tests... better threshold
commit 884ffc756e7520fcbb9252260181538ec81b8368
Author: Ricardo Quesada <ricardoquesada@gmail.com>
Date: Tue Feb 2 17:21:32 2016 -0800
initial uniform_auto_matrix commit
Changes the CC_matrix uniform in runtime depending on whether the command can be batched or not.
Lot of things missing, though... like performance tests.
halx99