* Add a feature to select mipmap range to image asset.
* fix format
* fix format
* modify comments, remove uncessary methods
* finx comments.
* fix logic problem
* fix mipmap range problems
* Add validation information incase someone wants to generate mipmap for compressed texture
* remove some validations to make gfx bakend clear
* fix a stupid typo
* Add validation information for setMipRange
* use assertID to validate the mipmap range
* draft
* fix some stupid bugs introduced in the lase commit
* fixed bug when init webgl's texture.
* draft
* fix some bugs
* fix some warnings
* fix undefined type cast
* For Vulkan device, query the storage feature through optimal tiling instead of linear tiling.
* fix some bugs appear when resize texture, clean the changes. all tests passed.
* remove debug codes, improve performance, correct the framebuffer size.
* make some input parameters readonly
* Make the gfx function's parameter's access priority the same as native engine
* remove uncessary members and interface for ES3, move getGLSampler into IWebGL2GPUSampler, delete WebGL2CmfFuncCreateSampler...
* create a seperated branch to solve the Issue 8203
* Add format feature for query
* fix bugs and add formatFeature info for metal backend.
* 1. fix some bugs; 2. correct some condition judgements; 3. remove some uncessary comments;
* fix a bug of textureExclusive, fix some typos.
* clean the code, correct some formatFeatures.
* fix bugs in metal backend, fix bugs found on windows.
* Use a proper log funciton in DeviceValidator. Fix code format.
* fix perfect-forwarding in CondVar interface
For there is no unified way to reliably manage the process across
backends. Right from the beginning the input information from system
event is just not reliable enough to be used as ground truth.
This also means the asynchronicity between device & render threads will break
whenever a resize event happens.
* sampler wrap mode validation
* fg: fewer hashing per frame
* better interface names for query pool
* auto update fbo after rt resize
so that the framegraph reuse logic can work
* vk: better subpass dependency deduction
* GL: better subpass support
* uint -> uint32_t
* officially make viewport/scissor dynamic by default
* gfx: states store hashes
* gfx: defend against usages after destroy