When building GPU vertices from paths, we call stroke.SplitCubic once
per OpCubic. Before this change, each call to stroke.SplitCubic would
allocate a slice, which we would only use to iterate over.
This allocation can be easily avoided by reusing the slice. We can
conveniently store it in gpu.quadSplitter.
In a real application that renders hundreds of paths with tens of
rounded rectangles per path, this saved roughly 4500 allocations (or 1
MB worth) per frame.
Signed-off-by: Dominik Honnef <dominik@honnef.co>
All GPU APIs except OpenGL ES 2 can generate mipmaps for textures.
This trades 33% more GPU memory use for improved rendering quality
and speed for downscaled images.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
splitTransform func was creating multiple copies of f32.Affine2D
due to not having access to the internal and passing around non-pointer.
This makes splitTransform from ~8ns to .Split ~2ns.
It seems that the non-pointer receiver was ~0.6ns slower in this case.
Signed-off-by: Egon Elbre <egonelbre@gmail.com>
There are no public API that uses f32.Rectangle anymore. Move Rectangle
to an internal package for internal use.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Creating an image.Rectangle from a f32.Rectangle is used by two packages in Gio
and about to be used for a third. Add a Round method to f32.Rectangle to avoid
duplicating the implementation.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Ever since commit 8ff654628, the headless implementation has used two
GPU backend (not renderer) instances, one for the renderer and one for
creating the offscreen texture to render into. This arrangment leaks
resources because the backends only clear temporary storage at
BeginFrame, which is not called when reading pixel data from renders.
This change adds an internal constructor, gpu.NewWithDevice, to allow
headless.Window to share its device with the renderer, fixing the leak.
It also makes the code simpler (took me a while to debug this issue); in
fact I'm surprised it even works.
This is not a great fix: it adds an exported yet internal constructor,
and the ownership transfer of the device is surprising enough to warrant
two comments.
Fixes gio#322
Signed-off-by: Elias Naur <mail@eliasnaur.com>
This change is a follow-up to bcf3ff77ff, fixing the two renderers to
properly render images with non-zero origins.
Signed-off-by: Pierre Curto <pierre.curto@gmail.com>
We're about to use clip.Ops for pointer areas; this change makes the
decoding accessible from the router package.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The Add method was a compatibility stop-gap.
API change. Use clip.Op.Push and Pop the return value to explicitly mark
the region affected by the clip operation.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The default renderer caches pre-processed paths for efficient re-use,
but failed to distinguish outline paths from stroke paths.
Fixes gio#294
Signed-off-by: Elias Naur <mail@eliasnaur.com>
In a discussion with Raph Levien, the author of our compute renderer
implementation, it became clear to me that it's not at all certain that
complex strokes will ever be efficiently supported by a GPU renderer.
At the same time, the machinery for converting a complex stroke to a
GPU-friendly outline has a significant maintenance cost. Further, it is
surprising to users that complex strokes are significantly slower and
allocate memory.
This change removes support for complex strokes, leaving only
round-capped, round-joined strokes supported by the compute renderer.
The default renderer still converts all strokes to outline, but it also
caches the result.
This is an API change. The complex stroke conversion code has been moved
to the external gioui.org/x/stroke package, with a similar API.
Updats gio#282 (Inkeliz brought up the allocation issue)
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The op.Save and Load methods exist to support the need for
transformation, clip, pointer area state to behave as stacks. For
example, layout needs to apply an offset to its children but not
subsequent operations.
Before this change, op.Save and Load were used to save and restore the
state:
ops := new(op.Ops)
// Save state.
state := op.Save(ops)
// Apply offset.
op.Offset(...).Add(ops)
// Draw with offset applied.
draw(ops)
// Restore state.
state.Load()
A drawback with the op.Save mechanism is that there is no direct
connection between the state change and the saving and loading of state.
This causes confusion as to when a Save/Load is needed and who is
responsible for performing them, which leads to subtle bugs and over-use
of Save/Loads.
This change gets rid of the general state stack and replaces it with
per-state stacks. There is now a stack for transformation, clip, pointer
areas, and they can only be restored by the code pushing state to them.
The example above now becomes:
ops := new(op.Ops)
// Push offset to the transformation stack.
stack := op.Offset(...).Push(ops)
// Draw with offset applied.
draw(ops)
// Restore state.
stack.Pop()
For convenience, transformation also be Add'ed if the stack operation is
not required.
Simple state such as the current material no longer has a way to be
restored; it is assumed the client of a PaintOp adds their desired
material operation before it.
API change: replace op.Save/Load with explicit Push/Pop scopes for
op.TransformOps, pointer.AreaOps, clip.Ops.
To ease porting, this change retains a version of op.Save/Load that
saves and restores the transformation and clip stacks. It also retains
an Add method for clip.Op.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The clip field tracks the intersection of the clip stack, and belongs in
the clip stack entries. This is a refactor that makes it easier to
implement scoped clip operations.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The rect field tracks whether the clip stack can be represented by a
pixel-aligned rectangle. It is easier to track as part of pathOp instead
of the global drawState.
Remove an obsolete comment while here.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
This change implements a Vulkan port for the two renderers, old and
compute. Run with GIORENDERER=forcecompute to test the compute renderer.
To shake out bugs faster, it is also made the default on systems that
support it. To disable Vulkan and force the use of OpenGL, use the
`novulkan` tag:
$ go run -tags novulkan gioui.org/example/kitchen
Don't forget to file an issue describing the issue that prompted the use
of the tag.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
driver.Device.NewFramebuffer doesn't provide additional information over
driver.Device.NewTexture, so Texture can hold its (optional) framebuffer
on behalf of the renderers. Metal don't even need a separate framebuffer
object.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Vulkan textures (VkImage) are always in a particular layout, where each
layout is optimized for a particular use (transfer, sampling, compute
storage). Vulkan allows layout transitions everywhere except inside
render passes. This change adds driver.Device.PrepareTexture for
instructing the driver to switch a texture to a layout for sampling
in preparation for using it in a render pass.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Vulkan doesn't support changing uniforms during a render pass. However,
push constants *can* be changed. The gio-shader repository was changed
to use push constants instead of uniforms, this change implements the
corresponding driver and renderer change.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Modern GPU API such as Metal and Vulkan use explicit render passes
and command buffers for recording rendering commands. They don't have
global state; each render pass starts with a clean set of bound
textures, pipeline etc.
Change our GPU abstraction to better match newer API and modify our two
renderers to explicitly describe their render passes.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
There's no meaningful reason to have them separate. The intention was to
enable rendering concurrent with other processing, but that's gaining
framerate at the expense of input latency and complicating ImageOp
semantics.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Changes to the gioui.org/shader module are generally breaking changes,
which means that a particular version of gioui.org must be build with a
particular version of gioui.org/shader. Until now, the gpu package
checked the module version against an expected version and would fail at
runtime if there's a mismatch.
This change replaces all that complexity with a simple procedural
change: bump the module major version of gioui.org/shader at each such
incompatible change. It doesn't matter that we'll eventually reach
gioui.org/v1234/shader; the module is internal and won't break clients.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The OpenGL (ES) implementations on Apple platforms are deprecated and
don't support GPU compute programs. This change adds support for the
replacement, the Metal GPU API.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Modern API such as Metal and Vulkan want clients to compile expensive
state changes into pipeline objects. Change our GPU driver abstraction
to match, thereby paving the way for future drivers.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The default renderer tracks opaque rectangle draw operations and render
them front-to-back with a z-buffer to omit overdraw. However,
- Overlapping opaque rectangles are rare in a GUI, and the most common
instance, a solid window background, is already optimized to a glClear.
- A z-buffer is memory heavy.
- We conservatively assume a 16-bit depth buffer, which limits the
number of drawing operations to 64k (#127).
- Depth buffer support makes GPU ports more complex, especially for
upcoming ports (Metal, Vulkan).
- The compute renderer doesn't use z-buffers.
This change removes the optimization; a follow-up removes GPU backend
support.
Fixes gio#127
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Both the OpenGL and the Direct3D API are stateful and gpu.GPU renders to
the render target current when Frame is called.
Modern GPU API such as Metal don't have a concept of a current render
target, and the target even changes each frame.
Add RenderTarget and add an explicit target argument to GPU.Frame as
well as the underlying driver.Device.BeginFrame.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
This change avoids the hard dependency on GPU support for sRGB encoded
textures in the compute renderer.
With this change and the previously added CPU fallback, Gio no longer
rely on any GPU functionality outside the OpenGL ES 2.0 level.
Fixes gio#49
Fixes gio#154
Fixes gio#97
Fixes gio#36
Fixes gio#172
Signed-off-by: Elias Naur <mail@eliasnaur.com>
This change adds a CPU fallback for devices that don't support the old
renderer nor have GPU support for compute programs.
Most of the hard work is implemented in the gioui.org/cpu module. It
uses the SwiftShader project with light modification to output
statically compiled CPU .o files for each compute program.
The CPU fallback only covers Linux and Android on arm, arm64, amd64
architectures. There is no fundamental reason support can't be extended
to other platforms:
- macOS and iOS are probably easy, but it's likely that virtually every
device has GPU support for compute shaders.
- Windows needs a Cgo-less port, or a build constraint to require a C
compiler (Gio core doesn't).
- FreeBSD and OpenBSD are probably also easy to do because they're so
similar to Linux.
- The 386 binaries didn't work properly in my tests, so fixes to
SwiftShader is probably needed. However, I expect virtually every
Intel device can run amd64 binaries.
Updates gio#49
Fixes gio#228
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Dominik Honnef