Commit Graph

3323 Commits

Author SHA1 Message Date
Hajime Hoshi
c61be3c532 internal/ui: bug fix: getMetal might return a nil pointer as a non-nil interface 2022-03-24 13:36:53 +09:00
Hajime Hoshi
eeb5687b73 internal/graphicsdriver/metal, internal/graphicsdriver/opengl: change the return type to pointers
On second thought, returning pointers is more natural. Handling
nil is a caller's responsibility.
2022-03-24 13:20:36 +09:00
Hajime Hoshi
7bb7e45522 internal/graphicsdriver/metal, internal/graphicsdriver/opengl: change the return type to interface
Converting a nil with a pointer type to an interface type is no longer
a nil-nil interface, and then comparing with nil causes an unexpected
results.
2022-03-24 03:24:58 +09:00
Hajime Hoshi
3074dca670 internal/ui: add the EBITEN_GRAPHICS_LIBRARY environment variable
The `ebitengl` build tag is gone instead.

Closes #2007
2022-03-24 02:23:37 +09:00
Hajime Hoshi
151dfffad6 internal/restorable: simplify canDetectContextLostExplicitly
This constant was set with some wrong assumptions:

 1. On Android, recovering was needed.
 2. On iOS, OpenGL ES was used when
    a. The architecture was 386 or amd64 == an emulator is used
    b. The build tag ebitengl was not specified
    c. gomobile-build was used
 3. On browsers, recovering was needed.

1., 2b, and 2c are correct.

2a. is not correct: Now emulators are available on all the
architectures with both Metal and OpenGL.

3. is not correct: Ebiten no longer recovers the contest lost.

Now, Ebiten can detect a context lost explicitly when

 1. On Android
 2. On iOS and on gomobile-build
    (When gomobile-build is used, OpenGL should always be used)

Based on this fact, this change changes the constant to a variable,
and fixes the logic to set the variable.
2022-03-23 03:01:59 +09:00
Hajime Hoshi
85daef7436 internal/ui: bug fix: OpenGL must be used when gomobile-build is used
Now the logic of graphicsDriver() is updated, this can return a Metal
graphics driver on iOS even when gomobile-build is used.

This change fixes this issue.
2022-03-23 02:58:56 +09:00
Hajime Hoshi
09b0a0ffaf internal/ui: refactoring: simplify the selection of graphics libraries 2022-03-23 02:21:53 +09:00
Hajime Hoshi
7e6e022065 cmd/ebitenmobile: support Metal for iOS simulators 2022-03-23 01:34:58 +09:00
Hajime Hoshi
a74e7b1578 Revert "internal/graphicsdriver/metal: refactoring"
This reverts commit 1c57393393.

Reason: nothing is rendered on iOS.

Closes #2022
2022-03-23 01:22:39 +09:00
Hajime Hoshi
8e6907c64a internal/ui: reduce the calls of graphicsDriver() 2022-03-22 02:17:41 +09:00
Hajime Hoshi
15548b4c74 internal/ui: add userInterfaceImpl.graphicsDriver 2022-03-22 00:13:22 +09:00
Hajime Hoshi
be1836339b internal/ui: define the common struct UserInterface for all the environments
The existing UserInterface structs became userInterfaceImpl structs.
2022-03-21 23:10:27 +09:00
Hajime Hoshi
99437944bc internal/ui: remove graphicsDrivre() calls from image.go 2022-03-21 22:49:47 +09:00
Hajime Hoshi
cd57bccbfc internal/graphicsdriver: let some functions return an error
This is a preparation for the DirectX driver.

Updates #1007
2022-03-21 22:23:12 +09:00
Hajime Hoshi
0dcf43b350 internal/graphicsdriver: reorder the member 2022-03-21 22:05:21 +09:00
Hajime Hoshi
5e973ab419 Revert "internal/graphicscommand: clear the queue regardless of an error"
This reverts commit e21636fbb9.

Reason: Simply this is no longer needed. We gave up testing when an error
occurs in a graphics command queue.
2022-03-21 21:57:59 +09:00
Hajime Hoshi
31104c4e79 internal/processtest: remove TestShaderNoMain and add shadernomain.go
A shader compilation error breaks the state of the graphics command
queue, and this cannot be reused. Thus, a process test is appropriated.
2022-03-21 21:52:38 +09:00
Hajime Hoshi
81b9f91f86 internal/graphicscommand: compile shaders lazily
With DirectX, the graphics driver cannot be determined until the
main loop starts, as a transparent window cannot be treated with
DirectX so far. On the other hand, compiling shaders requires a
graphics driver as it requires information about Y directions of
NDCs and framebuffers.

This change delays compiling shaders until the graphics commands
are actually executed in the main loop.

Updates #1007
Updates #2019
2022-03-21 21:09:02 +09:00
Hajime Hoshi
e21636fbb9 internal/graphicscommand: clear the queue regardless of an error
There was a potential issue that commands in the queue are never
reset when an error happens and the queue continues to send errors.
This is problematic especially for testings.

This change fixes the issue by Go's defer.
2022-03-21 20:10:03 +09:00
Hajime Hoshi
ad0e0e3e66 internal/graphicscommand: remove an unused member 2022-03-21 19:50:05 +09:00
Hajime Hoshi
9b1adf799d internal/shader: move syntax tests to internal/shader 2022-03-21 19:31:06 +09:00
Hajime Hoshi
b5bb576a58 internal/ui: refactoring: unexport ConvertUniforms 2022-03-21 17:05:50 +09:00
Hajime Hoshi
f4ad12987d internal/ui: bug fix: compile error with the cbackend tag 2022-03-21 16:59:53 +09:00
Hajime Hoshi
12ce5ae83a internal/ui: remove the call of graphicsDriver() from the context 2022-03-21 16:48:33 +09:00
Hajime Hoshi
fc96eb30a1 internal/atlas: simplify the logic by adding paddings to the mask 2022-03-21 16:13:21 +09:00
Hajime Hoshi
870a18e8f5 internal/buffered: remove unnecessary conditions from resolvePendingPixels 2022-03-21 15:27:21 +09:00
Hajime Hoshi
4b31983b6d internal/buffered: English 2022-03-21 15:20:52 +09:00
Hajime Hoshi
c316aaae72 internal/buffered: simplify the API 2022-03-21 15:19:06 +09:00
Hajime Hoshi
5c3f8915d1 internal/atlas: add an argument check 2022-03-21 15:11:38 +09:00
Hajime Hoshi
29f7a45ccc internal/buffer: remove the graphics-driver argument from ReplacePartialRegionPixels
This is necessary to remove the graphics driver usage from
(*ebiten.Image).At. And this is necessary to determine the graphics
driver after the window becomes transparent or not.

Unfortunately, it is not obvious to make a transparent window with
DirectX. Then, the determination of a graphics driver should be delayed.

Updates #1007
2022-03-21 05:59:43 +09:00
Hajime Hoshi
4f070915b2 internal/atlas: bug fix: do not use the padding when a mask is specified
With paddings, a mask size was not appropriate.

As a simple solution, let's not use paddings when a mask is specified.
2022-03-21 05:50:31 +09:00
Hajime Hoshi
dfc9eeaaec internal/restorable: bug fix: the parameter mask was not passed correctly 2022-03-21 05:40:13 +09:00
Hajime Hoshi
fbcbd2a001 internal/buffered: refactoring 2022-03-21 04:36:32 +09:00
Hajime Hoshi
b3f4d6c522 internal/atlas: add a new parameter specifying a mask for ReplacePixels 2022-03-21 03:56:04 +09:00
Hajime Hoshi
5c79b86412 internal/restorable: add a new parameter specifying a mask for ReplacePixels 2022-03-21 03:08:58 +09:00
Hajime Hoshi
e55dbbf3cd internal/restorable: add more tests 2022-03-21 03:04:52 +09:00
Hajime Hoshi
72a6ab8da4 internal/restorable: add more tests 2022-03-21 02:44:48 +09:00
Hajime Hoshi
924f7ea932 internal/restorable: refactoring 2022-03-21 01:26:48 +09:00
Hajime Hoshi
5fe6791b5d internal/graphicscommand: add a new paramter 'mask' to ReplacePixels 2022-03-21 01:11:01 +09:00
Hajime Hoshi
c1a0d83f8d internal/restorable: simplify the implementation of rect
Allow overlapped regions at pixelsRecords.

This removes the members `last*` for caching. These were introduced
at 7e7751bd43, and apparently these
are no longer needed.
2022-03-21 00:06:30 +09:00
Hajime Hoshi
ea81d4abf4 internal/buffered: refactoring 2022-03-20 19:03:41 +09:00
Hajime Hoshi
14c327c89b internal/atlas: replace Pixels with At to reduce unnecessary slice allocations 2022-03-20 18:39:17 +09:00
Hajime Hoshi
11ff0ab48c internal/atlas: refactoring: remove arguments from Pixels 2022-03-20 18:28:57 +09:00
Hajime Hoshi
367a9ec5bf internal/atlas: refactoring 2022-03-20 18:07:29 +09:00
Hajime Hoshi
54b4e87506 internal/restorable: bug fix: ReplacePixels on a sub-image might panic on Android
If regions by ReplacePixel are overlapped, this can panics. This can
happen only on Android, where a context lost can happen.

Thus, a sub-image cannot call a direct ReplacePixels. internal/buffer
has to care this.
2022-03-20 18:01:37 +09:00
Hajime Hoshi
696bbc088f internal/ui: rename variables 2022-03-20 16:51:23 +09:00
Hajime Hoshi
c3e855ab02 internal/buffered: refactoring 2022-03-20 16:44:11 +09:00
Hajime Hoshi
673556d03f internal/ui: move the error handlings to the ui package 2022-03-20 16:26:26 +09:00
Hajime Hoshi
bd07f6246f internal/buffered: remove mysterious comments 2022-03-20 05:17:43 +09:00
Hajime Hoshi
cc574ad67e internal/atlas: rename functions: Area -> Region 2022-03-20 04:20:23 +09:00
Hajime Hoshi
b59dd45239 internal/buffered: separate ReplacePixels with the large-area and small-area versions
For the large-area version, this doesn't require a graphics driver.
This is necessary to ensure that ReplacePixels never needs a graphics
driver.
2022-03-20 04:13:31 +09:00
Hajime Hoshi
4cbce71b2b internal/graphicscommand: move the choice of graphics drivers to internal/ui 2022-03-20 02:51:31 +09:00
Hajime Hoshi
1d9982ee6d internal/ui: move the dependency on graphicscommand from ebiten to ui 2022-03-20 02:00:44 +09:00
Hajime Hoshi
e78f34aa26 internal/ui: add Image
This is a preparation for a refactoring. Image will be a proxy to
pass a graphics driver to the lower layer.
2022-03-20 01:39:05 +09:00
Hajime Hoshi
3e44a20b22 internal/graphicscommand: swap arguments 2022-03-20 00:09:17 +09:00
Hajime Hoshi
8e3576b404 internal/mipmap: remove unused functions 2022-03-19 23:45:15 +09:00
Hajime Hoshi
7c2300b352 internal/graphicscommand: reduce the usages of the graphicsDriver function 2022-03-19 23:03:34 +09:00
Hajime Hoshi
58366ce64d internal/glfw: make the file operation atomic
If multiple Ebiten processes start at the same time, the file operation
can cause a race condition. Add a file lock to avoid this.
2022-03-18 20:32:25 +09:00
Hajime Hoshi
2e6bb8c4e2 internal/graphicsdriver/metal: add comments 2022-03-14 22:02:44 +09:00
Hajime Hoshi
a617576879 internal/shaderir: replace Mul with ComponentWiseMul and MatrixMul
This is a preparation for DirectX / HLSL.

Updates #1007
2022-03-13 19:17:46 +09:00
Hajime Hoshi
044d41dd2d internal/shaderir: change Op from string to int
This is a preparation to distinguish Hadamard product and matrix
product for HLSL.

Updates #1007
2022-03-13 16:20:04 +09:00
Hajime Hoshi
7d0f95e9be internal/graphicsdriver: refactoring: replace Uniform with []float32
Closes #2016
2022-03-13 03:42:13 +09:00
Hajime Hoshi
35a5c88901 internal/graphics: rename constants 2022-03-12 04:01:48 +09:00
Hajime Hoshi
98e2d86ea3 internal/shaderir/msl: refactoring 2022-03-12 00:18:55 +09:00
Hajime Hoshi
5715539a0f internal/shaderir/glsl, internal/shaderir/msl: refactoring 2022-03-11 15:18:38 +09:00
Hajime Hoshi
e63709a2a5 internal/shaderir/glsl: refactoring 2022-03-11 02:03:15 +09:00
Hajime Hoshi
6f00221051 internal/shaderir/msl: rename the package name
Updates #2010
2022-03-10 16:25:11 +09:00
Hajime Hoshi
5f97bdc1d3 internal/shader: fix typo
Updates #2011
2022-03-10 02:54:32 +09:00
Hajime Hoshi
5614e9d7eb internal/shader: fix misspellings 2022-03-10 02:52:44 +09:00
Hajime Hoshi
880dd0682e internal/shader: bug fix: fix some failing tests 2022-03-10 02:24:29 +09:00
Hajime Hoshi
d3b72d4ef9 internal/shader: add test files
Updates #2011
2022-03-10 01:02:51 +09:00
Hajime Hoshi
de05ede27e internal/shader: bug fix: evaluate the builtin float/int functions in the pre-compile phase correctly
Closes #2011
2022-03-10 00:47:16 +09:00
Hajime Hoshi
47bfd0db9e internal/shaderir/metal: rename to msl
Closes #2010
2022-03-10 00:14:06 +09:00
Hajime Hoshi
5e5e3c64ac internal/shaderir/metal: remove a redundant case 2022-03-09 02:54:31 +09:00
Hajime Hoshi
15064dbdc1 internal/shaderir/glsl, internal/shaderir/metal: simplify function names 2022-03-09 01:43:36 +09:00
Hajime Hoshi
e4fba8b047 bug fix: compile error with Go 1.16 and older 2022-03-08 00:08:53 +09:00
Hajime Hoshi
86ba7719eb internal/ui: bug fix: use CompareAndSwap 2022-03-07 23:51:10 +09:00
Hajime Hoshi
c37ffecafd internal/graphicsdriver/metal: reorder uniform variables
This might improve memory allocations.

This makes the code more consistent with HLSL's packing rule [1]

[1] https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/dx-graphics-hlsl-packing-rules
2022-03-05 22:12:08 +09:00
Hajime Hoshi
9c6b7aaca3 internal/ui: catch the error at At correctly 2022-03-05 01:44:59 +09:00
Hajime Hoshi
df2133186d internal/shaderir: bug fix: test failures 2022-03-04 19:01:58 +09:00
Hajime Hoshi
3c1e64dd67 internal/shaderir/metal: bug fix: define the 'mod' function correctly
Closes #2006
2022-03-04 18:05:44 +09:00
Hajime Hoshi
d4d4b9c070 internal/graphicsdriver/metal, internal/graphicsdriver/opengl: more efficient modulo 2022-03-04 03:47:55 +09:00
Hajime Hoshi
ed22052e5f internal/graphicscommand: add a comment 2022-02-28 00:15:52 +09:00
Hajime Hoshi
dd8900ea48 internal/graphicsdriver: refactoring: choose the graphics driver at this package 2022-02-27 23:51:19 +09:00
Hajime Hoshi
99f8e335ee ebiten: refactoring 2022-02-27 22:21:57 +09:00
Hajime Hoshi
77f765d483 internal/testing: bug fix: considering the case when Y is inverted
Closes #2003
2022-02-27 22:14:32 +09:00
Hajime Hoshi
4ca1220671 internal/graphicscommand: bug fix: compile error, test failures 2022-02-27 20:13:21 +09:00
Hajime Hoshi
b22309a0e5 internal/graphicsdriver: replace Pixels with ReadPixels
Now preparing a byte slice is the caller's responsibility.
2022-02-27 20:03:13 +09:00
Hajime Hoshi
1cb7633ff6 internal/ui: bug fix: atomic.Value.Store cannot be called when a differnet type value is already stored 2022-02-27 17:41:19 +09:00
Hajime Hoshi
566957dc1c internal/graphicsdriver/metal: invert Y at the vertex shader
Instead of using a negative height in the viewport, invert the Y direction
at the vertex shader. This is a little more readable as a negative height
is hacky.

This is a preparation for DirectX 12. DirectX 12's coodination system
is very similar to Metal, but doesn't treat a negative height in its
viewport unfortunately.

Updates #1007
2022-02-27 01:53:03 +09:00
divVerent
ea35296be7
ebiten: provide SetScreenFilterEnabled, a way to disable the default screen filter (#1997)
Using FilterNearest rather than filterScreen or FilterLinear speeds up a
Dell E6500 from 23fps->36fps and 27fps->48fps.

Updates #1772
2022-02-24 03:46:27 +09:00
divVerent
f2209a0b51
internal/atlas: optimization: send premultiplied alpha from vertex to fragment shader. (#1996)
Note that this applies only to the builtin shaders - interface for Kage stays
unchanged for compatibility.

Minor compatibility delta: when interpolating alpha values, previous code has
created nonsense values, such as, when interpolating from
fully-transparent-black (0,0,0,0) to opaque-white (1,1,1,1), something like
half-transparent-grey (0.25,0.25,0.25,0.5) where half-transparent-white
(0.5,0.5,0.5,0.5) is used by the new code.

I assume this is a strict improvement, however this may warrant some testing.

Possible later improvement could be moving the premultiplication from fragment
shader to CPU. Did not do this as it makes the code rather inconsistent of Kage
vs built-in shader usage.

Updates #1772
2022-02-24 02:27:50 +09:00
divVerent
476f4e3f9a
internal/gamepad: fix SetWindowLongPtrW crash on 32-bit Windows (#1999)
Closes #1998
2022-02-23 23:18:20 +09:00
Hajime Hoshi
8b7273b74a internal/graphicsdriver/metal: call PresentDrawable only when necessary 2022-02-22 01:37:13 +09:00
Hajime Hoshi
ee911fd892 internal/graphicsdriver/metal: fix an error message 2022-02-20 22:01:36 +09:00
Hajime Hoshi
bf33658e64 internal/gamepad: respect original names more 2022-02-15 04:07:11 +09:00
Hajime Hoshi
f17d222a4b internal/atlas: resolve deferred every frame instead of every tick
Resolving defferred functions doesn't have to be done every tick. Rather,
this should be done every frame as this is related to rendering.
2022-02-14 03:45:58 +09:00
Hajime Hoshi
4949c49cd0 internal/ui: refactoring 2022-02-14 03:39:02 +09:00
Hajime Hoshi
871121abe3 internal/ui: refactoring: remove (*contextImpl).layout 2022-02-14 02:59:02 +09:00
Hajime Hoshi
cffa123405 internal/ui: refactoring: remove sizeChanged for js
It is safe to call Layout every frame.
2022-02-14 02:50:32 +09:00