|
40f4de8f
|
2023-12-15T10:17:32
|
|
Vulkan: Ensure we use cached memory for readPixels stagingBuffer
Previous CL crrev.com/c/5112759 does not solve the performance issue for
ChromeOS. The reason is that on more recent intel GPU, there is no
hostVisibleCachedCoherent heap. When we allocate staging buffer, we
specify CachedCoherent as the preferredFlags instead of requiredFlags.
This means we still end up getting UncachedCoherent since VMA tries to
respect coherent bits as first priority. This CL Changes CachedCoherent
to CachedPreferCoherent, and made Cached as required bit, thus ensures
the memory allocated is cached. Since coherent bit may not be honored,
thus we have to call invalidate/flush (which underline implementation
will check the bit and early out if no need).
Somehow on ARM GPU using cachedNonCoherent staging buffer causing many
test failures, even though we do call invalidate() after allocation, and
tests pass on all other GPUs. It almost indicates ARM driver have a bug
with invalidate() that it is not doing expected. But before I can be
sure and fixed, I added feature bit to keep ARM the old behavior, which
uses UnCached memory for readPixels which should suffer the performance
as well.
Bug: b/315836169
Bug: b/310701311
Change-Id: I1eec6105ce74275faa893b0206be8470f0cde72f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5122318
Commit-Queue: Charlie Lao <cclao@google.com>
Auto-Submit: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
3439e421
|
2023-12-15T11:37:53
|
|
Vulkan: Remove AHB workaround for camera app
This is now fixed in camera app main branch. The workaround is removed
here.
Bug: b/239181279
Change-Id: I782baf7a42a8c9a897f1947d8b2f6feba606ce29
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5126931
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
6d9f0aee
|
2023-11-30T00:00:00
|
|
Metal: Ensure helper invocations when derivatives are used
Set sample coverage mask to ensure that
fragment shader helper invocations are
performed when derivatives are used.
Fixed the following tests on Apple silicon:
dEQP-GLES2.functional.texture.mipmap
.2d.projected.nearest_linear_repeat
.2d.projected.linear_linear_repeat
.cube.projected.linear_nearest
.cube.projected.linear_linear
dEQP-GLES3.functional.texture.mipmap
.2d.projected.nearest_nearest_mirror
.2d.projected.linear_nearest_clamp
.2d.projected.linear_nearest_repeat
.2d.projected.linear_nearest_mirror
.2d.projected.nearest_linear_repeat
.2d.projected.linear_linear_clamp
.2d.projected.linear_linear_repeat
.3d.projected.nearest_nearest_mirror
.3d.projected.linear_nearest_clamp
.3d.projected.linear_nearest_repeat
.3d.projected.nearest_linear_clamp
.3d.projected.nearest_linear_mirror
.3d.projected.linear_linear_clamp
.3d.projected.linear_linear_repeat
Fixed: angleproject:8443
Fixed: angleproject:7023
Change-Id: I13fc52532b94de098dcf3e9f5f02d48a6c8913c7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5089131
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
|
|
fb6b960c
|
2023-11-27T14:24:28
|
|
Remove GL_CHROMIUM_texture_filtering_hint
This was using an unregistered Vulkan extension to set the precision of
SwiftShader's internal filtering for the sake of Chrome. That's baked
in at build instead.
Bug: angleproject:8349
Bug: chromium:726075
Change-Id: I12849d2d29d99626f22a92ee9d74366f78658476
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5063344
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
edcfab40
|
2023-11-23T00:00:00
|
|
Metal: Implement textureCubeGrad transformation for AGX
Wrapped all affected built-in function calls
with helpers that transform derivative values.
Fixed all *.texturegrad.* dEQP failures on Apple silicon.
Fixed: angleproject:7021
Fixed: angleproject:8433
Change-Id: I16b023840ad267ab72d31fde3cb0fa7048e5310c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5071254
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
|
|
|
3680a5dc
|
2023-11-17T13:51:07
|
|
Vulkan: Let program warmup continue passed link
The warmup task does not actually affect the link results, so there is
no reason to wait for it when the application queries the link status.
This change allows the warm up task to continue in parallel until the
program is used at draw time. This allows the warm up to be more
efficient when the link itself is not parallelized.
For applications that create programs in the middle of every frame, it's
still likely best to disable warm up (as the following immediate draw
will already effectively do the warm up).
Note that currently the warm up code in the Vulkan backend is not
completely thread-safe, and so the program still blocks on that task
before the first draw can happen (or the program is modified in any
way).
Bug: angleproject:8417
Change-Id: I0877fef39a0585c3279e32699ce817d4643d7cd6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5037538
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
d842f3db
|
2023-11-29T06:53:58
|
|
Vulkan: add a disableSeparateShaderObjects feature
Dota Underlords uses one PPO and keeps swapping the attached program.
Though we get cache hit, the hundreds pipeline creation calls still
amount to non-trivial CPU overhead via ANGLE. The use of SSO has been
disabled by the app in certain cases, and this feature is added to
support entirely disabling SSO in ANGLE on all Vulkan backend impls.
Bug: b/309028728
Test: Dota Underlords easily hits target FPS on ANV and RADV
Change-Id: I54532975237218563b8750f1878a83e386fe776d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5073414
Auto-Submit: Yiwei Zhang <zzyiwei@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
2155534b
|
2023-11-06T11:49:20
|
|
Don't set dirty bits for attribs that are out of range.
PrivateState::setAllDirtyBits sets all bits in mDirtyCurrentValues.
When the context has fewer max attibutes than MAX_VERTEX_ATTRIBS, this
can cause out-of-bounds access to
PrivateState::mVertexAttribCurrentValues if the dirty bits are iterated
over without range validation.
Bug: chromium:1496378
Change-Id: I65481c432263a6e353a9361bba741b97dc5e20b1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5008034
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
|
|
|
12cc04d0
|
2023-11-07T10:26:18
|
|
Vulkan: Workaround camera AHB's layerCount being random value
We are seeing AHB used by google's camera app has random value in the
layerCount. In my reproduced case, it is 0, which translates to
layerCount in the VkImage. Sometimes I am also seeing a huge number in
the layerCount as well. This causes problem in ARM driver that end up
with assertion and other problems that end up with crash later. We need
to root cause the underlying bug which is tracked by b/239181279. But
before we have an actual fix in AHB creation code, this CL forces
com.google.android.GoogleCamera AHB's layerCount to be 1 as a temporary
app specific bug workaround so that we can move forward to get camera
app up running with ANGLE.
Bug: b/309480316
Bug: b/239181279
Change-Id: I219f6f89863a2cb8dee6c5efb0812389272373ba
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5010082
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
5a71be90
|
2023-11-03T15:07:06
|
|
D3D: Make non-constant loop indexing limitation a feature flag
... instead of exposing it as a front-end limitation only to set a
shader flag in the front-end.
Bug: angleproject:1130
Change-Id: Idbd9a1b7b79d231bd9fc1c0061c768df4aaab92d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5004847
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
e7ce481f
|
2023-11-01T22:44:22
|
|
Consolidate parallel compilation in front-end
This cleans up the multiple compilation task implementations in the
backends and consolidates them in the front-end. The front-end is then
able to do the compilation in an unlocked tail call instead if
desired (in a future change).
This change is in preparation for having the program link tasks directly
wait on the shader compilation tasks. As a result, the "shader resolve"
should not be needed to access the shader compilation results; it should
be enough to wait for the compilation job. This change therefore moves
post-processing of results to the compilation job itself as they did not
need to actually be done after compilation is done (merely after
translation is done). As a side effect, shader substition and other
debug features should now work for the GL backend as they are now done
before back-end compilation.
Bug: angleproject:8297
Change-Id: Ib9274b1149fadca7545956a864d6635b6cba5c3a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4994655
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
3b555e8d
|
2023-10-26T00:00:00
|
|
GL: Add emulateClipOrigin workaround
Older Adreno drivers apply GL_CLIP_ORIGIN_EXT
to the framebuffer instead of the clip space,
thus causing various side-effects.
Used a uniform to emulate the correct behavior.
Drive-by:
* Fixed a typo in ClipControlTest.OriginFrontFacing
Fixed: angleproject:8392
Change-Id: I2f7145977ab0e11dc88e8dbfec2cd32c4c31f830
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4987326
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
|
|
65f6c2ea
|
2023-10-22T18:07:55
|
|
Metal: always implicitly resolve MSAA render buffers on tiled GPUs.
On tile based GPUs, implicitly resolving MSAA render buffers to
single-sampled is preferred. Because it would save bandwidth by avoiding
storing the MSAA textures to memory. Furthermore resolving as
StoreAction is almost a free operation on these GPUs.
Traditional desktop GPUs almost always store MSAA textures to memory
anyway, so this feature would have no benefit besides adding additional
resolve step as well as memory overhead of the hidden single-sampled
textures.
Bug: chromium:1486094
Change-Id: I5eb3b1314560024dd5c0834b0c0b43a6b4d3d51b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4962114
Commit-Queue: Quyen Le <lehoangquyen@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
6258acae
|
2023-10-19T00:00:00
|
|
GL: Add resyncDepthRangeOnClipControl workaround
Some drivers do not apply GL_CLIP_ORIGIN_EXT and
GL_CLIP_DEPTH_MODE_EXT state changes until other
transformation states are updated.
As a workaround, force-resync the depth range on
clip control updates.
Fixed: angleproject:8381
Change-Id: I26f95108e9419be30ea0b562e02ab4bae78f65bb
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4966525
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
cfc40d2b
|
2023-07-19T17:47:13
|
|
Vulkan: Adjust clear color precision for GL_RGB5_A1 FBO
When clear color has more precision than the framebuffer
format can hold, dithering is automatically applied on
some hardware.
This issue causes below dEQP tests to fail when the FBO
color attachment format is RGB5_A1:
KHR-GLES31.core.draw_buffers_indexed.color_masks
KHR-GLES32.core.draw_buffers_indexed.color_masks
Adjust the clear color precision for RGB5_A1 format to
workaround the issue.
We can remove this workaround once the vulkan driver
fixes the auto-dithering problem.
Bug: b/292282210
Change-Id: Ic3ffebd2d20c8782612619a60d1ec2cc6d613c22
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4937472
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Roman Lavrov <romanl@google.com>
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
|
|
|
57fcee54
|
2023-10-11T14:48:53
|
|
Vulkan: Add feature flag for externalFormatResolve
This CL adds supportsExternalFormatResolve feature flag and enable the
vulkan extension if existed.
This CL is split from Chris Forbes's CL on android gerrit.
Bug: b/223456677
Change-Id: I731f35dc629c12c7290c6ed24ff54967e2e4eab7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4932592
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
|
|
|
ffb32380
|
2023-10-02T16:01:37
|
|
Vulkan: Fix device creation on multi-queue devices
When multiple eligible queue families are exposed by the device, ANGLE
defers devices creation until a surface is made current, at which point
it selects a queue family that supports presentation to that surface.
This path was largely untested and was broken:
- Some post-device-creation initialization was not deferred until the
device was actually created
- Some cap calculation depended on the chosen queue family index
- Query of device capabilities was done too late, such that ANGLE could
not correctly determine the level of support for ES versions.
Bug: angleproject:8300
Change-Id: I2a35396d1fd08ed26d217dff07d10e9a9c1ac55f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4907895
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Roman Lavrov <romanl@google.com>
|
|
|
97a48912
|
2023-09-26T10:31:58
|
|
Turn is-link-thread-safe feature into a positive condition
Chrome's --disable-gpu-driver-bug-workarounds currently sets every
feature to false, which breaks the GL backend.
Bug: angleproject:8297
Change-Id: I284d0699e356d7c1a362eb992cdc0d052f9ea7c2
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4887598
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Roman Lavrov <romanl@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
141bada9
|
2023-09-22T10:25:30
|
|
Vulkan: add prefer cached memory type for dynamic buffer usage.
Bug: b/288119108
Change-Id: I0fb5d91780e83af06762b9f3e6122313e76624da
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4886846
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Hailin Zhang <hailinzhang@google.com>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
cd8495b0
|
2023-09-14T20:00:32
|
|
Re-enable Metal on AMD FirePro.
Instead of disabling Metal altogether, test whether it works to just
disable the rescobeGlobalVariables compiler workaround on FirePro
devices.
Bug: angleproject:8317
Change-Id: Ia93a1fef04a387c5756d14660a2da4eb39b8db08
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4864732
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
ffd6ec26
|
2023-09-07T15:39:19
|
|
Reland "Make egl surface uncurrent when being destroyed"
This relands commit 497440cdcb7d2ee59bca612dd07fc13cf09a6a57.
This is to workaround errors when app does below behaviors:
1) while there is a context still bound to the current
rendering thread and the surface, call eglDestroySurface()
2) create a new surface eglCreateWindowSurface()
3) call eglMakeCurrent() with the surface created in step 2)
4) does work on the new surface
The old surface won't be destroyed in step 1) because
it was still bound by the context of the current rendering
thread. When creating new surface on step 2), some hardware
will return error code EGL_BAD_ALLOC, because the old egl
surface is still associated with the native window.
To workaround, when destroying surface, if the surface
is still bound by the context of the current rendering
thread, release the context and surface by passing
EGL_NO_CONTEXT and EGL_NO_SURFACE to eglMakeCurrent().
The workaround is controlled by a frontend feature
uncurrentEglSurfaceUponSurfaceDestroy. This feature
is only enabled on vulkan backends.
Bug: b/292285899
Change-Id: Id4c47f1b20e0f90d1013a68893fd70e917c030e2
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4867066
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
|
|
|
b4852ef9
|
2023-02-08T14:18:06
|
|
Vulkan: Drop support for Vulkan 1.0
Bug: angleproject:7959
Change-Id: Ib673679ea1a503af22b37092dbff1ee1fd34fba6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4233092
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Ian Elliott <ianelliott@google.com>
|
|
|
68bfa1ed
|
2023-08-22T22:02:15
|
|
Support for link to be entirely parallelized
The link job is split as such:
- Front-end link
- Back-end link
- Independent back-end link subtasks (typically native driver compile
jobs)
- Post-link finalization
Each step depends on the previous. These steps are executed as such:
1. Program::link calls into ProgramImpl::link
- ProgramImpl::link runs whatever needs the Context, such as releasing
resources
- ProgramImpl::link returns a LinkTask
2. Program::link implements a closure that calls the front-end link and
passes the results to the backend's LinkTask.
3. The LinkTask potentially returns a set of LinkSubTasks to be
scheduled by the worker pool
4. Once the link is resolved, the post-link finalization is run
In the above, steps 1 and 4 are done under the share group lock. Steps
2 and 3 can be done in threads or without holding the share group lock
if the backend supports it. Step 2 is not yet made independent of the
Context on some backends, and a frontend feature is used to make that
step either run on the main thread or as a worker thread.
Bug: angleproject:8297
Change-Id: I12f1e6bbaf365543dfcac969e166e0b5aa622104
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4808191
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
5afc76e3
|
2023-09-12T11:05:08
|
|
Disable Metal on AMD FirePro devices.
These older drivers are crashing in the driver for unknown reasons.
Only 0.5% of Chrome users are on these devices, so we'll disable
Metal on them.
Bug: angleproject:8317
Change-Id: Ia3ca5c3fa8e96aab903e15d989cfbd459679b15d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4859020
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Kenneth Russell <kbr@chromium.org>
|
|
|
8a7ad933
|
2023-09-13T15:11:46
|
|
Revert "Make egl surface uncurrent when being destroyed"
This reverts commit 497440cdcb7d2ee59bca612dd07fc13cf09a6a57.
Reason for revert: this caused chromium webview tests failures: https://chromium-review.googlesource.com/c/chromium/src/+/4860891.
Original change's description:
> Make egl surface uncurrent when being destroyed
>
> This is to workaround errors when app does below behaviors:
>
> 1) while there is a context still bound to the current
> rendering thread and the surface, call eglDestroySurface()
> 2) create a new surface eglCreateWindowSurface()
> 3) call eglMakeCurrent() with the surface created in step 2)
> 4) does work on the new surface
>
> The old surface won't be destroyed in step 1) because
> it was still bound by the context of the current rendering
> thread. When creating new surface on step 2), some hardware
> will return error code EGL_BAD_ALLOC, because the old egl
> surface is still associated with the native window.
>
> To workaround, when destroying surface, if the surface
> is still bound by the context of the current rendering
> thread, release the context and surface by passing
> EGL_NO_CONTEXT and EGL_NO_SURFACE to eglMakeCurrent().
>
> The workaround is controlled by a frontend feature
> uncurrentEglSurfaceUponSurfaceDestroy. This feature
> is only enabled on vulkan and gl backends.
>
> Bug: b/292285899
> Change-Id: I872d2e116ba6860f58d1176f011a5ef7c5a5af4e
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4851255
> Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
> Reviewed-by: Charlie Lao <cclao@google.com>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Commit-Queue: Yuxin Hu <yuxinhu@google.com>
Bug: b/292285899
Change-Id: I760054d856294e6691e79e165fd73ce9e560621f
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4862958
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Auto-Submit: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
|
|
|
497440cd
|
2023-09-07T15:39:19
|
|
Make egl surface uncurrent when being destroyed
This is to workaround errors when app does below behaviors:
1) while there is a context still bound to the current
rendering thread and the surface, call eglDestroySurface()
2) create a new surface eglCreateWindowSurface()
3) call eglMakeCurrent() with the surface created in step 2)
4) does work on the new surface
The old surface won't be destroyed in step 1) because
it was still bound by the context of the current rendering
thread. When creating new surface on step 2), some hardware
will return error code EGL_BAD_ALLOC, because the old egl
surface is still associated with the native window.
To workaround, when destroying surface, if the surface
is still bound by the context of the current rendering
thread, release the context and surface by passing
EGL_NO_CONTEXT and EGL_NO_SURFACE to eglMakeCurrent().
The workaround is controlled by a frontend feature
uncurrentEglSurfaceUponSurfaceDestroy. This feature
is only enabled on vulkan and gl backends.
Bug: b/292285899
Change-Id: I872d2e116ba6860f58d1176f011a5ef7c5a5af4e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4851255
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
|
|
|
b185c3ea
|
2023-09-11T12:28:43
|
|
Vulkan: Add host-image-copy usage to images when optimal
The change currently doesn't actually copy on host, but prepares the
image for it.
Bug: angleproject:8341
Change-Id: I4458712dca46ef9872020e158a3f902e94f5eb93
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4856146
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
|
e234c182
|
2023-09-08T16:19:29
|
|
Add features for VK_EXT_host_image_copy
Bug: angleproject:8341
Change-Id: Id076e6c9e040f3c19a1eb221f6099ac11d4d091a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4852280
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
0688a940
|
2023-08-21T00:00:00
|
|
GL: Disable EXT_texture_mirror_clamp_to_edge on Mesa
Although the extension string is exposed,
the new enum is currently rejected when
using an OpenGL ES context.
Bug: angleproject:8319
Change-Id: I17c4105344fe7ca3038a79e0e09528db1d96376b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4807744
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
|
|
|
9f258f8a
|
2023-08-21T00:00:00
|
|
GL: Add disableRenderSnorm workaround
Disable the extension on Mesa versions that
clamp negative snorm values to zero on read.
Fixed: angleproject:8315
Change-Id: I5459db40cb08c546fba15f5e6d70578029a8218a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4804324
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
|
|
|
d8f088e0
|
2023-08-14T11:07:24
|
|
GL: Remove parallel compile/link without driver support
This feature was practically disabled everywhere due to various bugs,
and is complicating the code. In effect, the code was always spawning a
thread for the compilation and link jobs, immediately fail it (due to a
workaround), then do the job when compile/link is resolved (much closer
to draw time). This leads to bad user experience, but also is racy
because the shaders may get recompiled in the meantime and there is
little the GL backend could do to stop that (efficiently).
After this change, parallel compile/link
is either done by the driver (if supported), or it isn't done.
This is a partial revert of a100d8f471f79b9f88d387164992cc5bd9c6ee9f.
Bug: angleproject:3031
Bug: chromium:922936
Bug: chromium:1184692
Bug: chromium:1202928
Change-Id: I6348bee3249ccb3828bb98ac2a69dc7d305f821c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4774785
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
59f158c1
|
2023-08-10T00:00:00
|
|
GL: Add explicitFragmentLocations workaround
Some drivers produce incorrect results when
a fragment output has an implicit location
and gl_SampleMask[] is written to.
Fixed: angleproject:8308
Change-Id: I615952ef61b1cb611984ec7defb189d89ab3281c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4777702
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
|
|
|
16cfa28e
|
2023-08-08T22:08:24
|
|
Vulkan: Basic infra for parallel link
This change moves pipeline warm up to a parallelizable task, mostly as
an exercise to put in the infrastructure for parallel link in the Vulkan
backend. Follow up changes will move more of the link step to this
task.
The end goal is to be able to make the link task independent of
ContextVk, which would allow it to be run as an UnlockedTailCall, even
if not using a worker thread.
Bug: angleproject:8297
Change-Id: I17047162b2a41f0d681d9e3ee33f2e0239b4280d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4764231
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
ffe81dd3
|
2023-08-03T14:53:32
|
|
Vulkan: allow opt-in to MSRTT emulation via overrides
The existing enableMultisampledRenderToTexture feature combines
conditions for device feature support and the policy decision of whether
to allow MSRTT emulation.
This change splits it into two features, allowing application developers
to control the policy condition for the emulation path without impacting
the device capability checks.
Bug: angleproject:8291
Change-Id: Ic1525c878906b10df777c582e44b931028aae928
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4749525
Auto-Submit: Steven Noonan <steven@uplinklabs.net>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
d2236b58
|
2023-07-29T23:50:05
|
|
Support substituting translated shaders.
Add dumpTranslatedShaders and enableTranslatedShaderSubstitution
frontend ANGLE features, which allow ANGLE developers to prototype
optimizations and other transforms without fully implementing them in
the shader translator.
Tested on macOS with ANGLE's Metal backend, but should work with the
other source-level translator backends.
Add documentation for pre-existing substitution of shader sources, and
of translated shaders added in this CL.
Fixed: angleproject:8280
Change-Id: I24d5ef88a479b23e81cc8169fe813c263acfc71f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4731553
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Kenneth Russell <kbr@chromium.org>
|
|
|
9aadc7aa
|
2023-07-13T11:44:51
|
|
Metal: Require MSL 2.1.
We had previously shipped both MSL 2.0 and 2.1 shaders to support
MacOS 10.13 (MSL 2.0 only) while also supporting the stencil blit
fast path that MSL 2.1 provides with the has_shader_stencil_output
feature.
Each configuration of precompiled shaders is ~300kb so we will drop
support for MacOS 10.13 and only ship the MSL 2.1 shaders.
Rework of the disablement for NVIDIA and GPU family 1 to be more
readable. The features themselves are always true because "disable on
NVIDIA" is always true and the check for "is NVIDIA" is moved to the
code which fails the initialization.
Bug: angleproject:8258
Change-Id: Icc8c69540e43fd2b0b237fffbfe170bb3422903f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4681130
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
4cab5e94
|
2023-07-12T10:04:10
|
|
Vulkan: Enable timeline semaphores if supported by device
This is needed for a similar change in Chromium to function with
`--enable-features=VulkanFromANGLE`.
Bug: angleproject:8253
Change-Id: I422964ea010c650439dd4797ee8ba7b533f26a87
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4675807
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
4e6124da
|
2023-06-23T14:46:28
|
|
GL: Ensure all instanced attributes have a buffer with data
Apple OpenGL drivers sometimes crash when given an instanced draw with
a buffer that has never been given data.
It's not efficient to check if the attribute is both zero-sized and
instanced so just ensure that every time a zero-sized buffer is bound
to an attribute, it gets initialized with some data.
Bug: chromium:1456243
Change-Id: I66b7c7017843153db2df3bc50010cba765d03c5f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4642048
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
42c3a9fe
|
2023-06-23T03:53:09
|
|
Vulkan: support creating context with MoltenVK on macOS
The Vulkan loader somewhat recently introduced a requirement that
clients must opt-in to using portability implementations of Vulkan (such
as MoltenVK). Since there is no native Vulkan driver for macOS (and
therefore no alternative), unconditionally enable the portability
enumeration extension there.
Bug: angleproject:8229
Change-Id: I24f0f24e25abd277855ed9ac4de370cfb47d3266
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4639495
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Steven Noonan <steven@uplinklabs.net>
Reviewed-by: Charlie Lao <cclao@google.com>
Auto-Submit: Steven Noonan <steven@uplinklabs.net>
|
|
|
ac263582
|
2023-06-19T12:26:44
|
|
Vulkan: add workaround for VK_EXT_full_screen_exclusive on AMD
On outdated (but recent) AMD drivers, the Windows-only Vulkan extension
VK_EXT_full_screen_exclusive appeared to be implicitly enabled and set
to VK_FULL_SCREEN_EXCLUSIVE_APPLICATION_CONTROLLED_EXT mode. Even though
ANGLE did not enable or interact with this extension at all, the driver
was incorrectly returning VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
error codes on various swapchain operations when the full screen window
focus was lost (i.e. alt-tab out and back in). Naturally, ANGLE was not
expecting these error codes and did not know how to handle them.
Depending on where the errors occurred, ANGLE might crash or retry
creating the swapchain repeatedly.
Treating the unexpected VK_ERROR_FULL_SCREEN_EXCLUSIVE_MODE_LOST_EXT
error code as VK_ERROR_OUT_OF_DATE_KHR/VK_SUBOPTIMAL_KHR was not
sufficient, because the driver would repeat the error on every swapchain
operation, apparently expecting the error to be handled by
a vkAcquireFullScreenExclusiveModeEXT call (even though that would make
no sense, since the extension was not enabled).
The incorrect driver behavior was reported to AMD and was fixed in
recent driver releases. The earliest driver I've tested and know to be
working is AMD's Adrenaline driver version 23.5.2
(VkPhysicalDeviceProperties calls this driverVersion 2.0.262/0x800106).
The last known bad version was 0x8000e9.
The simplest workaround on these older AMD graphics drivers is to
explicitly enable the extension, but set it to
VK_FULL_SCREEN_EXCLUSIVE_DISALLOWED_EXT mode. On newer drivers we do not
need to do anything with the extension and can ignore it.
Bug: angleproject:8215
Change-Id: I7c58d47a0350f4b0bc1a77f200c1e2f72fcde8d8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4627279
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
88c6dfcc
|
2023-06-02T16:09:24
|
|
Mali: Make the placeholder FBO complete for timer queries
Change the bindFramebufferForTimerQueries workaround to ensure
the framebuffer is complete.
Bug: chromium:1356053,b/269068358
Change-Id: Ief3198fdc4800468670a3f4323a1ffd3ca083c6f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4582606
Auto-Submit: Kramer Ge <fangzhoug@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
|
|
21f16cb1
|
2023-06-09T17:30:38
|
|
Disable clang-format on ANGLE features autogen outputs
Updates the script to produce reasonably formatted code without
clang-format.
Autogen files moved to autogen/ sub-directories because clang-format
does not support per-file settings ;(
This allows to run this codegen very quickly
(~50ms on my machine)
Bug: angleproject:8193
Change-Id: Ie84282090d574ebb4debe3edcfd82f983f27a5ff
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4604578
Commit-Queue: Roman Lavrov <romanl@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|