|
a70ef7fc
|
2024-04-25T14:19:50
|
|
Add EGL_ANGLE_no_error for disabling EGL validation.
Chrome makes many small EGL calls that can have proportionally
expensive validation.
Bug: angleproject:8434
Change-Id: I4f4d0e6eff64839f76a0f7bf48e5c94b8df9d809
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5491459
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
038ad6c7
|
2024-04-23T13:23:04
|
|
EGL: Use unlocked tail calls for surface, sync and img calls
EGL surface and sync calls can be expensive to do while holding the
global lock on EGL. They are safe because the parameters are captured
by value and the underlying EGL driver is also thread safe.
EGL image creation also tends to be expensive and is called freqently
by Chrome.
Bug: angleproject:8434
Change-Id: I7e554fe2e3700d98469de267f7bbff1e96358c78
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5478229
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
|
|
|
884dc380
|
2024-04-24T16:46:01
|
|
EGL: Make eglGetCurrent(Context|Surface|Display) lockless.
These functions access only egl::Thread state.
Bug: angleproject:8434
Change-Id: I3dd6cd1f4fd145613f0be824e4f6e13815422997
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5485526
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
c2ad86bc
|
2024-01-09T15:39:26
|
|
eglGetError is now a lockless entrypoint
eglGetError returns last recorded error in a thread,
there is no need to lock the global mutex.
Bug: b/318921454
Change-Id: I9e5192becc67dc81b54abfb63d2b32283ebc69d8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5181881
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
b46cf698
|
2023-11-30T11:00:18
|
|
Vulkan: Fix Android deadlock with querying buffer age
Similar to eglSwapBuffers, eglQuerySurface with EGL_BUFFER_AGE_EXT can
cause a call to vkAcquireNextImageKHR.
Bug: angleproject:6851
Bug: b/313975825
Change-Id: If3f0521219cab9aba2aeb2b70958bf0f197bc96a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5077406
Reviewed-by: Igor Nazarov <i.nazarov@samsung.com>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
ce263437
|
2023-09-11T12:25:28
|
|
Vulkan: Perform CPU wait in clientWait outside the global lock
Leverage UnlockedTailCall and move the CPU side wait
during a clientWait outside of the global mutex lock.
Bug: angleproject:8340
Tests: FenceSyncTest.BasicOperations*
Tests: EGLSyncTest.EglClientWaitSync*
Change-Id: I8c05e62e74cc64d38bf8797d28faaf49135e71fc
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4851649
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: mohan maiya <m.maiya@samsung.com>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
|
9740b01b
|
2023-09-08T16:30:08
|
|
Enhance UnlockedTailCall run method
UnlockedTailCall::CallType is now std::function<void(void *)>
This is in preparation for upcoming changes where unlocked
tail calls need access to objects outside block and namespace
scope.
Bug: angleproject:8340
Tests: UnlockedTailCall*
Change-Id: Ida6822b701c5c11ce4b8f6e3aae53108755e2cad
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4852021
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: mohan maiya <m.maiya@samsung.com>
|
|
36c3e0f5
|
2023-01-17T17:42:59
|
|
Implement "Shared Context Mutex" functionality.
Existing implementation uses single `GlobalMutex` for
- EGL calls
- GL calls for Contexts with concurrent access.
This CL introduces abstract `egl::ContextMutex` with two
implementations:
- SingleContextMutex;
- SharedContextMutex<Mutex>;
Note:
`std::mutex` is used in this commit. It is very easy to change mutex
type either at compile-time or at run-time (single type per Display).
When Context:
- is not Shared;
- does not use `EGLImage`s;
- does not use EGL_DISPLAY_TEXTURE_SHARE_GROUP_ANGLE
- does not use EGL_DISPLAY_SEMAPHORE_SHARE_GROUP_ANGLE
then it will be using `SingleContextMutex` with minimal overhead.
Before such Context is used as `shareContext` or uses `EGLImage`
its mutex replaced by `SharedContextMutex<Mutex>`.
The `GlobalMutex` is only used for EGL calls, while `egl::ContextMutex`
implementations for GL calls. Because some EGL calls use Context,
explicit `egl::ContextMutex` lock is required. This is implemented by
generating "egl_context_mutex_autogen.h" header, and insertion of
`ANGLE_EGL_SCOPED_CONTEXT_LOCK()` macro before `ANGLE_EGL_VALIDATE()`
in each EGL entry point. Implementation in "egl_context_lock_impl.h"
returns lock for required APIs. Special cases of `egl::ContextMutex`
lock handled separately. `std::unique_lock<>` is not used for
performance reasons.
`egl::ContextMutex` explicitly locked when capturing EGL calls.
Fixes EGLImage problem:
https://chromium.googlesource.com/angle/angle/+/e18240d136d15e5cdfa4fa4a6355ca21c8d807b6
Mark contexts as shared when importing EGL images.
Details:
- EGLImage inherits Context's mutex when created.
Mutex is used when the EGLImage accessed or destroyed.
- When EGLImage is used in Context with other `egl::ContextMutex`,
two mutexes are merged into one.
- After the mutex merge, Context Groups will remain separate,
but will not be able to run in parallel.
Fixes race when checking `context->isShared()` in the
`SCOPED_SHARE_CONTEXT_LOCK()` macro. One Context may start executing GL
call while not "Shared", but become "Shared" inside the call. New
(second) "Shared" Context may immediately start using GL and potentially
corrupt some "Shared" state.
Possible performance benefit: allows parallel execution in some cases,
when single `GlobalMutex` would block.
Important note:
Process of replacing the `SingleContextMutex` by
`SharedContextMutex<Mutex>` is not 100% safe. This mean that
original Context may still be using `SingleContextMutex` after
activating `SharedContextMutex<Mutex>`. However, this was always
the case before introduction of this CL. Old `Context::mShared`
member update was not synchronized in any way at all. In other
words, this solution does not 100% fix the original problem.
For 100% safe solution `SingleContextMutex` should not be used
(always pass `SharedContextMutex<Mutex>` to the `gl::Context`
constructor). See `lockAndActivateSharedContextMutex()` for more
details.
CL adds new build option:
angle_enable_shared_context_mutex = true
Behavior with other build options:
- When:
`angle_enable_shared_context_mutex` is disabled or
`angle_enable_share_context_lock` is disabled or
`angle_force_context_check_every_call` is enabled,
Contexts will always have `SingleContextMutex`, however it will be
only used in special cases. `SCOPED_SHARE_CONTEXT_LOCK()` will use
`GlobalMutex` when applicable.
- Otherwise, `SCOPED_SHARE_CONTEXT_LOCK()` will use `egl::ContextMutex`.
Some GFXBench "1080p Driver Overhead 2 Offscreen" performance numbers.
Tested on S906B (Samsung Galaxy S22+) on old ANGLE base:
https://chromium.googlesource.com/angle/angle/+/807c94ea85e046c6f279d081d99f0fb1bcf1191a
Capture/Replay: Adjust tests do adhere to capture limits
Each test result is an average frame number from 6 runs.
SingleContextMutex 6579 ( +0.13%)
(old) GetContextLock() (mShared is false) 6570
Forced `mShared = true` or NOT using `SingleContextMutex`.
SharedContextMutex<std::mutex> FORCE 5061 (-22.97%)
(old) GetContextLock() FORCE 4766 (-27.46%)
Bug: angleproject:6957
Bug: chromium:1336126
Change-Id: Idcd919f9d4bf482b9ae489bd8b4415ec96048e32
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4374545
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
7fd01d03
|
2023-04-19T00:54:24
|
|
Vulkan: Throttle the CPU without holding the global lock
CPU throttling goes through CommandQueue and is thread-safe. Performing
it in an unlocked tail call allows other unrelated EGL calls to go
through.
Bug: angleproject:8135
Change-Id: Idb3841be5d8ea8c4b76217f6707be26b28ea39c2
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4444027
Reviewed-by: Igor Nazarov <i.nazarov@samsung.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
|
cd6a58f2
|
2023-04-18T12:45:10
|
|
Vulkan: Make eglPrepareSwapBuffersANGLE less special
This function now uses the UnlockedTailCall mechanism so it doesn't
require as much special-case code generation.
This change does not fix the bug that this function is doing too much
work without holding any locks. That will be done in a follow up.
Bug: angleproject:6851
Bug: angleproject:8133
Change-Id: I77f4d514ff4aeef85bc1cc59214f7caa23aca7df
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4443186
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
1328f2f3
|
2023-04-17T16:43:12
|
|
Vulkan: Destroy the surface without holding the EGL lock
This change defers surface destruction to the end of the entry point
that causes it so that it is done without holding the EGL lock. This
works around a specific deadlock in Android. On this platform:
- For EGL applications, parts of surface creation and destruction are
handled by the platform, and parts of it are done by the native EGL
driver. Namely, on surface destruction, native_window_api_disconnect
is called outside the EGL driver.
- For Vulkan applications, vkDestroySurfaceKHR takes full responsibility
for destroying the surface, including calling
native_window_api_disconnect.
Unfortunately, native_window_api_disconnect may use EGL sync objects and
can lead to calling into the EGL driver. For ANGLE, this is
particularly problematic because it is simultaneously a Vulkan
application and the EGL driver, causing `vkDestroySurfaceKHR` to call
back into ANGLE and attempt to reacquire the EGL lock.
Since there are no users of the surface when calling
vkDestroySurfaceKHR, it is safe for ANGLE to destroy it without holding
the EGL lock.
Note that only eglDestroySurface and eglMakeCurrent may lead to the
destruction of a window surface.
Bug: b/275176234
Bug: angleproject:8127
Change-Id: I02dc52e53e150943457e3f503e7ef30469f96b05
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4428754
Reviewed-by: Charlie Lao <cclao@google.com>
Auto-Submit: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
3b57e999
|
2023-04-17T15:38:47
|
|
Scope global lock in entry points
This change is a no-op. It's in preparation for adding calls at the end
of the entry point after the lock is unlocked.
Bug: angleproject:8127
Change-Id: I4cd79ff8e5f20f87f36040afbd1ed9f16406d519
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4436589
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
|
be3d7e6b
|
2023-03-30T14:35:18
|
|
Remove redundant g_SurfaceMutex.
This mutex become redundant after recent CL:
https://chromium.googlesource.com/angle/angle/+/23ad4fa2be35aed303cbb8c2d632a04714354ef0
Don't hold global surface lock during AcquireNextImage
Every place in the code where `ANGLE_SCOPED_GLOBAL_SURFACE_LOCK()`
is used also uses `ANGLE_SCOPED_GLOBAL_LOCK()`.
Bug: angleproject:6851
Change-Id: I464b6ca74743c9ee9fa23caad216f0e26c480655
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4385293
Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
a52c0a6c
|
2023-02-06T16:01:27
|
|
Capture/Replay: Add and handle new resource type for EGLSync
So far calls involving EGLSync were not tracking the actual
sync objects, and this may lead to race conditions in
multi-threaded and multi-context scenarios.
This CL adds the type EGLSyncID and some specialized code
handling of egl::Sync to distinguish EGLSync from the already
existing GLSync objects in order to track them separately.
Bug: angleproject:7911
Change-Id: I91b188a41069bc0620f51c55ee516d23b55bdd38
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4200095
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
8403e4c5
|
2022-10-10T20:59:29
|
|
EGL: Resource IDs for Surface, Context and EGL Image.
This will make these classes play nicely with resource maps. As these
objects are used in a lot of places, and simplified C can't handle
unordered_map, it's necessary to index the maps by simple packed IDs
in capture/replay code. This indirection will also have increased
safety as we validate EGL resource ID handle values before accessing
the memory directly.
Also hides some of the other EGL capture methods behind helper methods
to simplify the C code and hide assignments and other complex maps.
Bug: angleproject:7758
Change-Id: Ibc7bb56430d3068bd38877c9dfb011979d4ea234
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3957164
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
|
|
|
2265e37b
|
2022-10-12T09:27:16
|
|
Capture/Replay: Auto-generate EGL capture code.
Replaces the custom code in the EGL stubs. Skips a few "Get"
entry points because this CL doesn't implement pointer capture
like we do for all the GL entry points.
Includes a new state in the AttributeMap that indicates which
type of attribute values we used when initializing the map.
Bug: angleproject:4035
Change-Id: I272eac5e4068602ce710ef66c9a1dce5387943a5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3949911
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
7c4dc253
|
2022-10-12T08:38:46
|
|
Capture/Replay: Clean up EGL capture.
This switches the EGL capture types to ANGLE-casted pointers since
that's what we receive in the capture layer. Note that even if the
capture layer were used as a pure layer, not an EGL implementation,
we'd still have these types for state tracking.
This also prefixes each EGL class in the entry points with the egl
namespace for consistency and for simplifying the ParamType code.
Required changing to non-const gl::Context * in a few places. Also
changes the gSurfaceMap to be indexed by the raw pointer value,
which cleans up the code somewhat.
Bug: angleproject:4035
Change-Id: Id800c1ba25e5819ac7ea1df8aab806bc393fe192
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3949910
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
926b43e7
|
2022-01-06T13:31:54
|
|
Reland: Frontend: separate lock in swap prep
Swapchain-based backends like Vulkan might block a lot in
vkAcquireNextImageKHR, which is bad for overall fast progress if we also
hold the global EGL lock there.
This CL starts to split the global EGL lock. We release the EGL lock
when performing vkAcquireNextImageKHR, and only maintain a lock for
surfaces.
This is done via a new custom entry point, EGL_PrepareSwapBuffers, so
that we can control how the global lock is used throughout the entire
call.
Bug: angleproject:6851
Change-Id: I095cd8b3bdbb13c842cab0a46148e2122582cdfd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3373426
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Lingfeng Yang <lfy@google.com>
|
|
|
9d668d6c
|
2022-01-07T18:18:23
|
|
Revert "Frontend: separate lock in swap prep"
This reverts commit 40c5cb255c0a07bdab574aa076ee603e7d791ab3.
Reason for revert: Regression in TSAN caused by this CL:
https://ci.chromium.org/ui/p/angle/builders/ci/linux-tsan-test/352/overview
Original change's description:
> Frontend: separate lock in swap prep
>
> Swapchain-based backends like Vulkan might block a lot in
> vkAcquireNextImageKHR, which is bad for overall fast progress if we also
> hold the global EGL lock there.
>
> This CL starts to split the global EGL lock. We release the EGL lock
> when performing vkAcquireNextImageKHR, and only maintain a lock for
> surfaces.
>
> Bug: angleproject:6851
> Change-Id: I329d5c4c579718a4980c4261590f77099ce1400e
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3361249
> Reviewed-by: Charlie Lao <cclao@google.com>
> Reviewed-by: Geoff Lang <geofflang@chromium.org>
> Commit-Queue: Lingfeng Yang <lfy@google.com>
Bug: angleproject:6851
Change-Id: Ie03b784021f7b8b5c1ef95a911ef7da4029abd46
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3373165
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
40c5cb25
|
2022-01-06T13:31:54
|
|
Frontend: separate lock in swap prep
Swapchain-based backends like Vulkan might block a lot in
vkAcquireNextImageKHR, which is bad for overall fast progress if we also
hold the global EGL lock there.
This CL starts to split the global EGL lock. We release the EGL lock
when performing vkAcquireNextImageKHR, and only maintain a lock for
surfaces.
Bug: angleproject:6851
Change-Id: I329d5c4c579718a4980c4261590f77099ce1400e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3361249
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Lingfeng Yang <lfy@google.com>
|
|
b5424bb4
|
2021-01-14T15:09:41
|
|
Generate internal gl entry point functions as C functions.
Some internal GL functions are exported to our libGLESv1_CM library
and to properly export them, they must be C functions.
Bug: angleproject:5534
Change-Id: I37280312f73fd5e55166e4fa36659267d657a50b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2628139
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
|
|
|
234fccfb
|
2020-12-01T10:59:16
|
|
Entry Points: Add egl:: namespace to Display.
This will prevent symbol collision with X11. This fixes the
build integration with Skia.
Bug: angleproject:2621
Bug: angleproject:5416
Change-Id: I6949a375cf9fcdd790b4c40ffb82c7c25bc15315
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2567644
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
3f564fc7
|
2020-11-23T13:59:17
|
|
EGL: Generate entry points.
This adds two final auto-generated files for the EGL and EXT extension
entry points. It adds a new data file that stores a mapping between
object types and entry points for associating labeled objects with
certain methods. When we generate errors we record the associated
object in the debug message output.
This places the remainder of the hand-written code in "stub" files.
Going forward the work for implementing new extension entry points
for EGL will be to update the registry XML files and then implement
the corresponding stub methods. Event logging, parameter packing,
and validation are all handled by the auto-generated code.
Bug: angleproject:2621
Change-Id: I28153432802c37b929ff2ea1e1a3e3ce9de91605
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2562680
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
|