This brings the old milepost submodule tree, along with its history, into the Orca repository, so we don't have to deal with git submodule issues anymore.
- Allow orca app to request either Canvas or GLES surfaces
- Add bounds check specifications to the json bindings spec format and to the bindings generator scripts.
- Generate GLES API json bindings spec from gl.xml
- Remove APIs that can't be bound with current wasm memory model (ie buffer mapping)
- Manually link remaining APIs, except glGetString
- Add fluid simulation sample
- Add abort messages for wasm loading and runtime fatal errors
- Adapt orca build tool to generate GLES json spec from gl.xml
- Adapt glesTriangle and fluid samples build scripts to new orca build tool
- update gles surface example
- add mp_dispatch_on_main_thread_sync() to osx threads
- add /experimental:c11atomics to glesTriangle build.bat
- mp_dispatch_on_main_thread_sync: win32 impl
- move mp_dispatch_on_main_thread_sync() on app layer
- configure dlmalloc to not assume contiguity. This allows us to reserve blocks on our end without trashing malloc-owned zones. This way we can add a block to store GL static strings when needed.
- implement deselect interface
- Log an error when an OpenGL function is called while no OpenGL API is selected, or if that function is not part of the selected API
- avoid dispatching calls to image or canvas API if surface is not selected. However, we could later allow it and temporarily select the surface _on behalf of the user_.
- Adding support for format string and optional parameters in assert macros
All the tooling has been built into a new command-line tool, `orca`. To use it, just run `orca -h`.
To build the samples, you must first install the orca tooling by running `orca dev install`. This simulates what it will be like to actually use the orca tooling as an end user, without the source checked out. Mostly. (I don't yet have it downloading pre-built copies of the Orca runtime libraries and `orca.exe`.)
Co-authored-by: Ben Visness <ben@bvisness.me>
Co-authored-by: Ben Visness <bvisness@gmail.com>
Co-authored-by: Martin Fouilleul <martinfouilleul@gmail.com>
Reviewed-on: #31
Co-authored-by: bvisness <ben@handmade.network>
Co-committed-by: bvisness <ben@handmade.network>
* Removes core_api and gles_api .txt bindings and replaces them with .json versions
* Rename bindgen2.py to bindgen.py, effectively deleting the old script
* Update build.bat to reflect new changes
This allows the renderer to use multiple source textures per batch, and thus create less batches (and less draw calls). It also doesn't create a separate batch for non-image fills/strokes. This greatly improves perf especially when interleaving images and solid color commands. When using lots of images though, it is still recommended to use an atlas, which ensures everything fits in a single batch.
Martin Fouilleul
Ilia Demianenko
bvisness
Ben Visness
Reuben Dunnington