|
9f9284b7
|
2023-07-17T15:41:27
|
|
Move ShareGroup to its own files
Bug: angleproject:8224
Change-Id: Id6d272018bb5ee8c3e35488f641efa4d99fa836d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4690003
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
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>
|
|
a8ba5112
|
2023-02-10T10:04:41
|
|
Capture/Replay: Deal with swap called in different contexts
Move recording the context setup of the main context to the
same location where all other context setups are recorded and
drop the assertion that checks that swap is always called from
the same context.
Invoke the context setup at the time the secondary contexts are set
up, and make sure the main context is correctly mapped.
Bug: angleproject:7911
Change-Id: I327bce318b1a0e26ffdbf096343f99cedd78c116
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4236541
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
a77e8e3a
|
2023-02-06T16:24:06
|
|
Limit logging when invalid calls are not captured.
An app was crashing during capture due to a huge volume of such calls:
% adb logcat -d | grep 'Not capturing invalid call' | wc -l
20609
There are a couple of cases where validation silently ignores "benign"
invalid calls, such as glUniform*(-1, ...):
https://crsrc.org/c/third_party/angle/src/libANGLE/validationES.cpp;drc=0c4306fc554c80506eb0f9b833a5d2a5fdd452d5;l=2815
Limit to (separately for active and inactive capture so that we still
see these after triggering mid-execution capture).
Example log after this CL:
02-07 11:54:45.869 7657 7749 I ANGLE : INFO: FrameCapture (capture inactive): Not capturing invalid call to glUniform1f
02-07 11:54:45.874 7657 7749 I ANGLE : INFO: FrameCapture (capture inactive): Not capturing invalid call to glUniform1f
02-07 11:54:45.882 7657 7749 I ANGLE : INFO: FrameCapture (capture inactive): Not capturing invalid call to glUniform1f (will no longer repeat for this entry point)
... (then I triggered capture) ...
02-07 11:55:13.049 7657 7749 I ANGLE : INFO: FrameCapture (capture active): Not capturing invalid call to glUniform1f
02-07 11:55:13.049 7657 7749 I ANGLE : INFO: FrameCapture (capture active): Not capturing invalid call to glUniform1f
02-07 11:55:13.050 7657 7749 I ANGLE : INFO: FrameCapture (capture active): Not capturing invalid call to glUniform1f (will no longer repeat for this entry point)
Bug: b/267795212
Change-Id: I2f150cfa5b4c74fc1ebe5abeb1201cc4caad80e3
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4224875
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Roman Lavrov <romanl@google.com>
|
|
f7b5d5d1
|
2022-12-15T10:52:07
|
|
Capture/Replay: Remove inline variable declarations.
This makes parsing easier for the "simplified C" interpreter.
We introduce a resource ID buffer as a way to manage a list
of resource IDs to replace the inline resource lists.
Turns on the Among Us trace in the interpreter tests.
Bug: angleproject:7775
Change-Id: I1bb9c0e9b087965a18691bc99b2e9947610b9eaf
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4128719
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
fdada9ee
|
2022-12-13T14:52:53
|
|
Re-land: "Make SyncIDs a packed type."
This re-land fixes the sync map size tracking.
This prepares syncs to use a simple resource map like other
types, which will make life easier in the trace interpreter.
Bug: angleproject:7775
Change-Id: If2114c51d5b68503890eacbf549182823667fedc
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4178012
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
8971a592
|
2023-01-18T14:35:09
|
|
Revert "Make SyncIDs a packed type."
This reverts commit 9de913077a5fcc3d2f2e327b56bbe30efe2fde96.
Reason for revert: Fails win-trace, somewhat flakily.
Original change's description:
> Make SyncIDs a packed type.
>
> This prepares syncs to use a simple resource map like other
> types, which will make life easier in the trace interpreter.
>
> Bug: angleproject:7775
> Change-Id: Ic2867f6133256f5ce2320eb2b322c1059266b201
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4103720
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Commit-Queue: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Cody Northrop <cnorthrop@google.com>
Bug: angleproject:7775
Change-Id: I29534b14c973fa34a4cb7457d534cd6156f33cd2
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4178010
Auto-Submit: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
|
|
|
9de91307
|
2022-12-13T14:52:53
|
|
Make SyncIDs a packed type.
This prepares syncs to use a simple resource map like other
types, which will make life easier in the trace interpreter.
Bug: angleproject:7775
Change-Id: Ic2867f6133256f5ce2320eb2b322c1059266b201
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4103720
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
7fe33996
|
2022-12-02T00:04:28
|
|
Capture/Replay: Regen/Restore shaders and programs
Add missing functionality that can recreate shaders and programs
that are deleted during the run.
Test: Plants vs. Zombies Heroes MEC
Bug: angleproject:5968
Bug: angleproject:7848
Change-Id: I2d41ebe9df1e2ad1beef831acd72fd3f06f4eb1f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4060241
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
0790a807
|
2022-12-01T23:58:36
|
|
Capture/Replay: Add ResourcesToDelete
For resources that are deleted but not recreated by the app,
we need to skip the delete call we've been making for
ResourcesToRegen. To support this, track which resources need
to be deleted.
We've gotten by without this for so long because most apps
will immediately recreate a resource after deleting it. When
they simply delete starting resources, we can't try to delete
them again.
Test: MEC of multiple apps
Bug: angleproject:4599
Change-Id: I226ba7887e2b7b31d4ce9a75d6a8d0a24f3f32ac
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4075486
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
|
b432c84c
|
2022-12-01T10:40:03
|
|
Capture/Replay: Fix collecting the initialized contexts
The main context initialization function will be emitted as
"Shared" and not based on the context number, so only add the
secondary contexts to the list of contexts that need additional
setup.
Fixes: b301b82235ca686d036a8d29380da22bb2060060
Capture/Replay: emit context setup for pre-MEC contexts only
Bug: angleproject:7805
Change-Id: I4ac770a303ac93a448c9b46bf8c4cd58900ddd54
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4068124
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
b301b822
|
2022-11-07T09:37:14
|
|
Capture/Replay: emit context setup for pre-MEC contexts only
When a context is made current for the first time it is added to
the shared context set. If this happens after the mid execution
capture has started, then the call to the context setup function
is created, but the according function implementations was not
emitted leading to compilation failure of the so created trace.
Since a context that was never current doesn't actually have any
setup that needs to be done when starting a MEC replay. Hence,
there is no need to emit the calls fot SetupReplayContextXX.
So track which context where actually in the shared context set
when MEC started, and only emit the according calls to the setup
functions.
Bug: angleproject:7805
Change-Id: I83f8714733ead5c0d71560013c360b5671f0822a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4008199
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
|
|
|
11de5157
|
2022-11-14T14:48:33
|
|
Capture/Replay: Reset Shaders
We've been resetting Programs correctly, but not Shaders. This hasn't
been a blocker so far, but it has been identified as a performance
problem for traces that compile mid-trace. We're not getting cache
hits because the persistent blob cache is not updated until a shader
is deleted.
Part of the challenge here is ShadersIDs and ProgramIDs share a
ResourceIDType, so separating them requires a bit more tracking.
To fix, start tracking which ShaderProgramIDs are actually
shaders and emit the correct glDelete* command.
Test: diablo_immortal MEC
Bug: angleproject:7823
Change-Id: I934f6eee243ab9681dca5fdebdad33ba31457cf5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4029226
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
e99e40c9
|
2022-10-11T12:27:18
|
|
FrameCapture: Implement shadow memory for coherent buffers.
On certain devices like the Pixel 6 it is not possible to mprotect
Vulkan allocated memory required for coherent buffer tracking.
To overcome this limitation implement a shadow memory for coherent
buffers when running FrameCapture that is exposed to the app and
syncronized with the Vulkan memory and can be mprotected for coherent
buffer tracking.
Add a test to determine whether memory protection can be used directly
or will require shadow memory. Run this test only on build
configurations with assertions enabled.
Determine the requirement of shadow memory through a deny list of
device manufacturers and models, which is checked against ANGLE's
SystemInfo.
Add ANGLE_CAPTURE_FORCE_SHADOW environment setting and Android
equivalent to force enable shadow memory.
Test: angle_end2end_tests --gtest_filter="BufferStorageTestES3.*/ES3_Vulkan"
Bug: angleproject:7402
Change-Id: I74b7930259d3bc1846ef96fffa782f8bc553b043
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3945018
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
109604b8
|
2022-11-07T09:26:05
|
|
Capture/Replay: remove unused context param from scanSetupCalls
Bug: angleproject:7805
Change-Id: I4e00876ad6174c15aaa9503cf9eeaa6ea041c2a6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4004416
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
ce62d52f
|
2022-10-27T18:41:11
|
|
Fix Mid Execution Capture for External Texture
How we capture eglCreateImage and eglCreateImageKHR
have changed recently. Fix the code that recreate
eglCreateImageKHR calls for external texture type
in Mid Execution Capture.
Bug: angleproject:7758
Change-Id: I2bd35ec3349f1d4f1aef1bf7f76ac50d4abca53d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3989645
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
06ecc918
|
2022-10-10T20:59:38
|
|
Capture/Replay: Optionally emit C sources.
This adds an option that controls if we can write out simplified C
replays. Once the work is finished we can remove the option and
only allow emitting C instead of both C and CPP.
Required emitting multi-line strings differently, as well as
conditionalizing a few other language differences.
Bug: angleproject:7713
Change-Id: I3303134316ed3fc1b4286bcd32961e8f7ecfbb06
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3953339
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
6193274a
|
2022-10-10T21:00:12
|
|
Capture/Replay: Redesign in-memory call capture replay.
This will allow the replay to use the call captures returned
by the interpreter's parser.
Bug: angleproject:7752
Change-Id: If1b281d9ce7ccfbdc23bea615e1e2258c8a029f9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3963367
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
|
|
|
4bfb749f
|
2022-10-10T20:59:48
|
|
Capture/Replay: Move shared trace code into src/common.
This will let them be accessible to the test harnesses. The
trace tests interpreter will need direct access to the classes
that we move in this CL.
This CL also moves the GLenum utils into the common folder,
where they were already used by some other tests.
Bug: angleproject:7752
Change-Id: I97ad607938ef29bc316f6d40098478e002ea8128
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3963362
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@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>
|
|
|
bb420af7
|
2022-10-10T20:59:24
|
|
Capture/Replay: Rewrite map buffer calls.
This wraps the map calls into simple helper functions that
hide the return value assignment, which in turn will simplify
C parsing.
Allows us to remove storing the mapped buffer ID in the params.
Bug: angleproject:7731
Change-Id: I756574cc0fa3d63cce4e0103b8f2861d093de186
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3953335
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>
|
|
|
c22f091d
|
2022-09-29T18:58:33
|
|
Replace Hard-Coded egl image attribs
This change is adapted from https://crrev.com/c/3838866
Added two hash maps to track below information:
EGLImage* --> EGLint* attrib
TextureID --> EGLImage*
During the CaptureShareGroupMidExecutionSetup,
when we iterate through all the textures the app created,
if the texture is bound to External target,
we will refer to the hash map TextureID --> EGLImage*
to find the EGLImage* pointer, and then use the result
as the key to refer to the other hash map
EGLImage* --> EGLint* attrib to find the attributes used
to create that EGLImage object. We can then use the
attributes to populate eglCreateImage or eglCreateImageKHR
calls in the trace calls.
Bug: angleproject:7570
Change-Id: I729de4ddea59242ccbe6243e036451f290545185
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3928212
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
|
|
|
5bafe449
|
2022-10-05T20:09:23
|
|
FrameCapture: Create default context reset calls
Add support for emitting default reset calls for some entrypoints
when recording a capture from the beginning.
Test: Lineage 2 Revolution MEC
Bug: angleproject:7741
Change-Id: I8e4e1184f3e3b68527a65283c459a43d135e95b4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3938442
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
|
6dc30c00
|
2022-09-29T08:41:53
|
|
FrameCapture: Reset default uniform arrays
If an application has updated a single uniform in the middle of an
array, we need to emit the Reset call for the entire array. We don't
track Reset calls per individual location in an array.
Test: Life is Strange MEC
Bug: angleproject:7711
Change-Id: Idec991ad060eb5e12272713a58aa921c5912f1cc
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3928253
Reviewed-by: Roman Lavrov <romanl@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
|
6ec89510
|
2022-08-31T11:10:45
|
|
Frame Capture: Remove MEC active variable.
We only used this in one place, where we can instead use a check
for frame capture being active generally.
Bug: angleproject:7621
Change-Id: Ic004e3021750a6c43443eb8935b59514e8c06978
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3868931
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
64f41972
|
2022-08-25T11:16:23
|
|
Use canonical gl.xml and update enum to string function.
This replaces our copy of gl.xml with the upstream canonical copy.
Note that one patch is required before we can remove ANGLE's copy:
https://github.com/KhronosGroup/OpenGL-Registry/pull/538
Because the upstream version uses a new method of enum groups, we
also update our enum-to-string generator to use the new groups.
This new code includes many more enums and groups in the mapping.
Bug: angleproject:6461
Change-Id: I1c0ab44c36afce8db04c9661b377bbe5762c913e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3856649
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
fd89020c
|
2022-07-18T20:26:38
|
|
FrameCapture: Reset VertexArray state
This change includes three things:
- Tracking resources by context
- Resetting VertexArray state
- Splitting ResetReplay into per-context
All three are needed to properly Reset Vertex Array objects.
Vertex Arrays are considered container objects, which means
they are not shared with other contexts. We've had a hole
in Reset where we did not take multiple contexts into account.
To address this, we now also track which contextID was used
to modify a resource. For objects that need to know which
context they are bound to, we track them in a map and
return the correct index when getTrackedResource is called.
To generate the calls to reset Vertex Array state, we are able
to mirror what has been done for other ResourceTypeIDs.
In order to properly Reset resources in multiple contexts,
we have to bind the correct context before emitting the
Reset calls. This has been done for Setup already, so we
have a template to follow. With this CL, we now emit
multiple functions to Reset state, similar to the following:
void ResetReplayContextShared()
{
// Reset shared objects
...
}
void ResetReplayContext5()
{
// Bind the right context
eglMakeCurrent(EGL_NO_DISPLAY, EGL_NO_SURFACE, EGL_NO_SURFACE, context5);
// Reset container objects for context5
...
}
void ResetReplayContext6()
{
// Bind the right context
eglMakeCurrent(EGL_NO_DISPLAY, EGL_NO_SURFACE, EGL_NO_SURFACE, context6);
// Reset container objects for context6
...
}
void ResetReplay()
{
EGLContext context = eglGetCurrentContext();
ResetReplayContextShared();
ResetReplayContext5();
ResetReplayContext6();
eglMakeCurrent(EGL_NO_DISPLAY, EGL_NO_SURFACE, EGL_NO_SURFACE, context);
}
Test: Ni no Kuni MEC
Bug: angleproject:4599
Bug: angleproject:7507
Change-Id: Ic585534f07172474aab0ce1d0396b0b064379f88
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3761874
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
|
|
|
fd9301c1
|
2022-06-17T12:25:10
|
|
Capture/Replay: Capture egl surface related calls
Creation from a pbuffer and generalized destruction are
implemented, as well as binding and releasing a TexImage
and making the context current with explicit draw and
read surfaces given.
Bug: angleproject:4964
Bug: angleproject:6180
Bug: angleproject:6512
Change-Id: Id8be6486125d45341905f3aabdbae4366cd568b7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3711741
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
ce78f5f8
|
2022-05-10T09:10:30
|
|
Capture/Replay: Reset default uniforms
Extend resource tracking to include default uniforms.
They are tracked per location, per program.
Setup for defaulft uniforms is unchanged, but Reset now
includes a section like this:
glUseProgram(gShaderProgramMap[6]);
UpdateCurrentProgram(6);
glUniform1uiv(gUniformLocations[gCurrentProgram][0], 1,
reinterpret_cast<const GLuint *>(&gBinaryData[30518768]));
glUniform1uiv(gUniformLocations[gCurrentProgram][1], 1,
reinterpret_cast<const GLuint *>(&gBinaryData[30518784]));
glUseProgram(gShaderProgramMap[9]);
UpdateCurrentProgram(9);
glUniform4fv(gUniformLocations[gCurrentProgram][1], 15,
reinterpret_cast<const GLfloat *>(&gBinaryData[30518960]));
glUniform4fv(gUniformLocations[gCurrentProgram][16], 3,
reinterpret_cast<const GLfloat *>(&gBinaryData[30519200]));
glUniformMatrix4fv(gUniformLocations[gCurrentProgram][20], 2, GL_FALSE,
reinterpret_cast<const GLfloat *>(&gBinaryData[30519248]));
...
Test: Animal Crossing MEC, Star Wars KOTOR MEC
Bug: angleproject:7307
Bug: angleproject:7353
Change-Id: Ic344767f5dd4ffc38c15ed6bf81a7d1daf274a7b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3657474
Reviewed-by: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
aac68adc
|
2022-05-21T22:27:22
|
|
Capture/Replay: Track dirty state per EntryPoint for Reset
This CL introduces a new concept of tracking whether an entry
point has been called. If so, emit Reset calls designed specifically
for that entrypoint.
This will likely only be useful for one dimensional state binding
calls that can easily be tracked and set back.
Only support for EntryPoint::GLUseProgram is added at the moment, but
any entry point can be added.
Reset will now set the active program back to what it was in Setup.
For example:
void ResetReplay()
{
...
glUseProgram(gShaderProgramMap[2]);
UpdateCurrentProgram(2);
}
Test: Animal Crossing MEC, Star Wars KOTOR MEC
Bug: angleproject:7307
Bug: angleproject:7353
Change-Id: I19d238ca13bbfb619e1cf8ed86f15ed6c4255a61
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3656827
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
bbf67e2e
|
2022-05-16T12:04:34
|
|
Capture/Replay: support capturing eglCreateImage/eglDestroyImage
Because we support only a few functions the supporting code is not
autogenerated. We don't capture the actual value of the display
variable, because we assume that there is only one display, and
the actual pointer to it is provided by the EGLWindow. The rest
of the capturing works just like with the GLES calls.
Bug: angleproject:4964
Bug: angleproject:5822
Bug: angleproject:6180
Bug: angleproject:6286
Bug: angleproject:6578
Bug: angleproject:7111
Change-Id: I385aa9648f93bf74706e9860e2aee5775eeba220
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3636062
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
f8ddca00
|
2022-03-03T15:34:15
|
|
FrameCapture: Use getCompressedTextImage.
Replace retrieveCachedTextureLevel with getCompressedTexImage.
Remove compressed texture shadow copy code.
Treat GLCopyImageSubData calls as texture updates by handling them in
trackTextureUpdate.
Bug: angleproject:5592
Bug: angleproject:5604
Bug: angleproject:6104
Bug: angleproject:6177
Bug: b/181797364
Change-Id: Ic6b2a41ce536e3e4b66497048954efdafa556d1e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3516377
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
ed590dc8
|
2022-03-11T12:54:44
|
|
FrameCapture: Add override for Glsizei* types.
Fix capturing glGetSynciv with null lengths on Linux.
When calling the default WriteParamValueReplay handler with a
nullptr, ostream on Linux will write `(nil)`.
This fixes retracing all traces containing glGetSynciv calls with null
lengths.
Add a NullLength test to FenceSyncTests that used to fail in the
capture_replay_tests.
Test: capture_replay_tests.py --gtest_filter=FenceSyncTest.NullLength/*
Bug: angleproject:7071
Bug: angleproject:6808
Change-Id: I0a2bd4db744ce2a026cd932ba0f2f4747672fcc8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3526653
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
7b202392
|
2022-03-02T17:32:45
|
|
Capture/Replay: Fix capture on Linux
Need to add support for printing "const GLint *" types.
We've been falling into the default handler and getting lucky with
how other platforms handle ostream with a null value. We use this
extensively with glShaderSource().
Test: MEC of an existing trace on Linux
Bug: angleproject:7071
Change-Id: I7e4f1fbf2876dddac49ea6583918dbc534070d75
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3501201
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
c80c7ae3
|
2022-02-24T10:43:15
|
|
Capture/Replay: Merge Deletes in Reset
Before this CL, when regenning resources, we would iterate through
each resource ID and issue a delete followed by a gen. The delete
is required to get resources back to their original state, often
caused by applications recreating them during the run.
This ran into problems when our resource maps had stale data in them,
i.e. a texture that had been deleted by the app but remained in our
gTextureMap. We could inadvertently delete a resource we had just
genned.
For example, in ResetReplay(), say we have two textures to delete:
// gTextureMap[1] start with 5
const GLuint glDeleteTextures_texturesPacked_1[] =
{ gTextureMap[1] };
glDeleteTextures(1, glDeleteTextures_texturesPacked_1);
// We just deleted texture 5, now create a new one
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
// The driver returned 15, which was unused at this time
UpdateTextureID(1, 0);
// gTextureMap[1] now contains 15
...
// gTextureMap[2] happens to start with 15, which was in use in
// the trace, but was deleted and is no longer in use. The
// deleted value is not cleared from gTextureMap (which is
// another possible way to solve this).
const GLuint glDeleteTextures_texturesPacked_2[] =
{ gTextureMap[2] };
glDeleteTextures(1, glDeleteTextures_texturesPacked_2);
// Whoops! We just deleted our brand new texture 15, even though
// it no longer maps to the original texture.
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
UpdateTextureID(2, 0);
// Now gTextureMap[2] contains whatever came back from the driver
...
// The first use of gTextureMap[1] will fail on loop because
// texture 15 no longer exists!
To avoid this problem we delete all resources up front before
genning any new ones.
const GLuint deleteTextures[] = {gTextureMap[1], gTextureMap[2]};
glDeleteTextures(2, deleteTextures);
// Now we no longer have any deletes in the create sequence
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
UpdateTextureID(1, 0);
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
UpdateTextureID(2, 0);
This is applied to all the resources we regen right now.
This CL:
* Merges the deletion of new resources (that didn't exist when the
trace starts) with resources that are being regenned. That means
there is just one big delete.
* Removes Delete from the call sequences we track for each resource
since we no longer need to keep them around.
* Adds a formatting helper to avoid code duplication.
Test: Blade&Soul Revolution MEC (UE4)
Bug: angleproject:4599
Bug: angleproject:7053
Change-Id: Ida3b7e1ad5d94c5e9860447d5cf959278f37ad47
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3492849
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
|
287057b3
|
2022-01-25T17:01:43
|
|
FrameCapture: Don't track coherent buffers before capture starts.
Even though coherent buffer storages were only captured after capture
was activated, the buffer tracker was incorrectly initialized with the
first call to glMapBufferRange, even though capture was not active.
This fixes corrupted rendering in War Planet Online due to all coherent
buffer storages being tracked at initialization of the app, when
FrameCapture was enabled but not active.
Introduce mMidExecutionCaptureActive, which exposes if MEC is currently
executed.
Bug: angleproject:5857
Bug: angleproject:6774
Change-Id: I4849d6497e8e08aefc60e68bae027915e9b2912f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3416227
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
c984dad5
|
2021-12-03T16:02:49
|
|
FrameCapture: Don't remove protection from shared pages.
Don't remove protection from pages that are shared when a
persistent coherent buffer storage is unmapped.
The shared page can still be relevant when the the other buffer is kept
and written to.
Add a PageSharingBuffers test that produces this behaviour.
Test: angle_end2end_tests --gtest_filter="BufferStorageTestES3.PageSharingBuffers/*"
Bug: angleproject:5857
Change-Id: I6927f25229d2dfe9f68ba9a993e9d3e994bc7ce0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3306623
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
d4b8cd5e
|
2021-09-14T16:34:31
|
|
FrameCapture: Implement capture of coherent bufferstorages.
Track buffers with the GL_MAP_COHERENT_BIT_EXT access flag in a
CoherentBufferTracker on a per page basis using the
PageFaultHandler and protection functions from system_utils.
Check for dirty memory and capture it on draw calls.
Bug: angleproject:5857
Change-Id: Ib098f96952db7583d3e6517570c179e7754f54b2
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3168629
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
f60b152c
|
2021-12-06T10:19:53
|
|
Frame Capture: Merge small frames into single sources.
By buffering up the replay functions before writing them out into
sources, we can reduce the number of replay files by a large
margin.
Bug: angleproject:5133
Change-Id: I0a556fd1a3a5f2dcc8a26b761e885c9b7e464cf5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3316200
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
e8b186c9
|
2021-12-09T09:45:20
|
|
Frame Capture: Don't unmap regenned buffers in reset.
These buffers are deleted and recreated, so unmapping them gives
a "resource not mapped" error during replay.
Found when working on the "world_war_doh" trace.
Bug: angleproject:5133
Change-Id: Id1e6318c44a2a09fa643ef0a72f1d03d8396e5fb
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3327862
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
d02d6a40
|
2021-12-06T09:26:14
|
|
Frame Capture: Track written files in ReplayWriter.
This means we don't need to recreate the list of written files in
the JSON where we list the trace files. It will allow for more
flexibility when we change what trace files we output.
Bug: angleproject:5133
Change-Id: I651b6a281c50040d32361a5a020ba3eaa4241bf0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3316199
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
0f09d378
|
2021-12-07T13:32:39
|
|
Frame Capture: Track return value in max resource IDs.
This adds handling for return value processing when it comes to
maximum resource IDs. Specifically for glCreateShader and
glCreateProgram.
Fixes MEC with re-capture with the blockman_go and probably other
traces that call CreateShader as the final operations during setup.
Bug: angleproject:5133
Change-Id: I799dfec776ca8c80a0b89c7f0c7c23423ef3896a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3320887
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
8313ffe2
|
2021-11-29T15:46:25
|
|
FrameCapture: Add ReplayWriter helper class.
This centralizes all of the replay file I/O into a single point.
Each function is stored separately in this class. Right now there's
some small file format changes:
- the namespaces for the capture label are gone. these had no
functional impact now that each trace was a separate module.
- the header is cleaned up into private and public functions. this
cleans up a few things like defining non-existent functions.
- inline variables are tracked through the replay. they are now
global instead of anonymous, and string sets get reused instead
of duplicated between frames.
There should be no functional changes with the naming of the trace
files or the trace behaviour.
This change sets the stage for the ReplayWriter buffering multiple
frames into a single cpp file on disk, which can potentially reduce
the number of object files in the compilation by a great deal. It
could also be extended to work with buffered I/O so we don't block
and hitch when we're writing large trace files.
Bug: angleproject:5133
Change-Id: Ib96685e1202ac8101a4b828fa79642236af2120a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3311940
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
3b6f80b7
|
2021-11-23T10:07:43
|
|
Capture/Replay: eliminate redundant parameters in GenOnBind
The template parameter already defines the parameter value name and
the parameter value type, so add some traits to simplify the call
interface of the function.
Bug: angleproject:6425
Change-Id: I037bed8e5cc2e7367d8faa3af5d4fba2bfbb52c3
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3295617
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
1d555687
|
2021-11-09T11:16:24
|
|
Capture/Replay: Set FBO ID when generated on bind
GLES allows FBO ID's to be reserved on bind, so if a FBO is bound with
and ID that was not yet reserved by a glGenFramebuffers call, update the
resource tracking and the resource map to account for this resources
that was created on bind.
Bug: angleproject:6425
Change-Id: I343fc17bfbbfd9c8c47d6fe207a4f3817acb835d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3190970
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
bb5dbd49
|
2021-10-05T11:57:44
|
|
Capture/Replay: Handle TvoidPointer in specialization
Bug: angleproject:6521
Change-Id: I863ad5307498e6394f5fbe3e650ce29c821083b1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3204960
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
043120ee
|
2021-09-29T15:56:46
|
|
Capture/Replay: Allow disabling program trimming.
The ANGLE_CAPTURE_TRIM_ENABLED variable disables program trimming
when set to 0. This is useful for trace validation, which expects
the GL state vectors to be identical.
Bug: angleproject:5133
Change-Id: I3ab41de80af7b223c4cb8a1b7e14f049717305fc
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3193417
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
3719c172
|
2021-09-20T17:33:14
|
|
FrameCapture: Filter program setup calls
This CL adds filtering to shared setup calls based on whether
a resource is used during the frame capture. Calls that are
filtered out will not be written to SetupReplayContextShared.
This eliminates setup for unused resources.
This has become critical for apps like Fortnite that compile
tens of thousands of programs before running any frames. This
can take multiple minutes to perform before rendering a frame.
To achieve this, we do the following:
- Add an active status to every call (default to TRUE).
- Mark setup calls for tracked resources to FALSE during MEC.
- During the run, track for any use of the resource, and mark
its setup calls back to TRUE.
- Enable tracking for ShaderPrograms.
Test: Fortnite MEC
Bug: b/180418810
Bug: angleproject:5658
Change-Id: I5070297d339de3b80f870640e8955077a2bc20aa
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3150387
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
a09dca68
|
2021-09-20T23:40:28
|
|
Fix app trace heap buffer overflow
During runMidExecutionCapture(), we need to update
mMaxAccessedResourceIDs[idType] for idType==ShaderProgram, so that we
have enought slot in gShaderProgramMap2 to store the
temp shader we create in GenerateLinkedProgram().
Changes in this CL:
1.Renaming Jamie's mMaxAccessResourceIDS array update function
updatePreCallResourceCounts() to updateResourceCounts().
2. Adding updateResourceCounts() to Cody's setup call post processing
function scanSetupCalls(). In GenerateLinkedProgram(),
we add the temp shader create call with its' tempShaderID
into the shareGroupSetupCalls.
After that we will run the function scanSetupCalls(),
where we will be able to loop through the tempShaderID
and update the mMaxAccessResourceIDS[ShaderProgram] with the correct
count.
Bug: angleproject:6398
Change-Id: Iee78b3b7c1908a90ee8456476902f5e52e4615dc
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3172005
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
|
|
|
0babcb08
|
2021-09-14T10:49:06
|
|
FrameCapture: Scan Setup instructions
The ability to add calls from a multitude of locations rather than a
centralized spot has become a useful feature. But some locations
were updating the max read buffer size via updateReadBufferSize,
while others weren't.
Rather than force everything to call a central spot for tracking, add
post process scanning of Setup calls to handle updates. This is
already done for frame calls in maybeCapturePreCallUpdates.
Test: Fortnite MEC
Bug: b/180418810
Bug: angleproject:5658
Change-Id: I51d3a39ea40073d2dca50339ef4b38712b81f4ca
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3160500
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
|
c12d2f37
|
2021-09-09T12:11:14
|
|
Capture/Replay: Remove unused function in header.
Bug: angleproject:5133
Change-Id: I80f22d95101263f2ff37835c9e0f634b96d41622
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3150258
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
c5a38976
|
2021-09-01T07:35:40
|
|
Capture/Replay: Add expression trigger validation calls.
Setting the environment variable "ANGLE_CAPTURE_VALIDATION_EXPR"
will make ANGLE's capture logic evaluate this expression every
captured call to see if it should insert a validation checkpoint.
The retracing script also accepts --validation-expr as an argument.
For instance, the expression:
((frame == 2) && (call < 1189) && (call > 1100) && ((call % 5) == 0))
Will insert validation checkpoints on frame 2, between calls 1100 and
1189 and will validate every 5th call. The 'call' here is the count of
captured calls, which are mostly GL calls with a few ANGLE replay
calls in the mix.
We add a small single-header library that can evaluate arthithmetic
expressions in order to parse these expressions, as well as an option
to the retracing script.
Bug: angleproject:5133
Change-Id: Ic369e85d8e905a3a7a32fa098f7d8ebe7baf4ab9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3136094
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
349d555d
|
2021-09-02T10:07:26
|
|
Capture/Replay: handle glGetBooleani_v
This also fixes a test.
Bug: angleproject:6180
Change-Id: Icad2d089738f4458c90b87620337e9798adb9141
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3139898
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
|
|
|
ec8418da
|
2021-07-02T09:55:39
|
|
Capture/Replay: Use plain arrays for resource maps.
Instead of using the templated wrapper class, use C++ arrays that
are dynamically allocated with "new". This saves a fair bit of binary
size spent in calling the operator[] function. To get the right size
this CL adds tracking to the capture for each resource type, then
updates trace initialization to store the maximum allocated resource
ID for each type.
Tested with Manhattan 3.0: on Windows with Release, binary size for
the trace shared object went from 7.053 MB to 6.125 MB. About a 13%
decrease in binary size. That will allow for longer captures and will
also reduce trace overhead.
Bug: angleproject:5133
Change-Id: I2cd7515262a9ec76112d5f790001e3a6e9acd3a9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3003383
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
9c05f55a
|
2021-07-15T10:58:10
|
|
Capture/Replay: Add replay state validation.
We can use this to validate traces that have slight differences
when retracing. The valdation works by embedding the "expected"
JSON into a string at the end of each captured frame. The replay
also embeds a callback which fires right before the swap in the
replay harness. The harness then gets the "actual" JSON and runs
a comparison. On a mismatch it calls "diff" externally.
Currently the diff call is hard-coded to work on Linux only.
Note that when running validation it's important to replay on
SwiftShader since that's what we use to retrace.
Bug: angleproject:5133
Change-Id: Icbf0031d07be8bd916607c537dec235f9a512c43
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3066008
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
58bb11ca
|
2021-04-16T12:28:50
|
|
Capture/Replay: Multi-Context Support
Add support for capturing and replaying multiple contexts.
1.) Create and initialize the Contexts in the share group during
SetupReplay().
2.) Track the Context the command stream is for, and if the Context ID
changes, inject an eglMakeCurrent() call to switch to the new Context.
3.) Intercept eglCreateContext() and eglMakeCurrent() to route to either
EGLWindow or WGLWindow, depending on the current platform.
Specifically, this enables capturing and replaying Asphalt 9.
Bug: angleproject:5878
Change-Id: I5bc9b7ece5388ce405ba3f9e9dc3967e78662000
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2830145
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
5f092f8b
|
2021-08-17T17:15:59
|
|
FrameCapture: Support glProgramBinary
This CL allows applications to use glProgramBinary. Normally this is
a problem because we need program source in order to recreate shaders
in use during mid-execution capture. Use of program binaries
means an app can start and render frames without having submitted
source for that run.
To support this, we will embed program source into ANGLE's binary
format. This will allow us to extract it when the app submits the
binary. We will only embed this when capture is enabled to prevent
increased binary size in the common case. Since this changes ANGLE's
binary format, apps will recreate binaries when capture is enabled.
Additionally, we can't allow captures to have glProgramBinary calls
in the middle of captured frame ranges, so intercept those calls and
replace with a full linking sequence.
Changes include:
- Add new frontend feauture enableProgramBinaryForCapture that allows
OES_get_program_binary during capture.
- Update ANGLE's binary format to include program source when
capture is enabled.
- Update maybeOverrideEntryPoint to handle multiple new calls
instead of a single call.
- Override calls to glProgramBinary that occur mid-capture
to instead emit a full GenerateLinkedProgram sequence.
- Add checks for created/attached shaders during FrameCapture since
they won't be available for programs populated by the app using
glProgramBinary.
Test: Fortnite MEC
Bug: b/180418810
Bug: angleproject:5658
Change-Id: Ib2a0e9e434d3ee0f384d128c48b2a7d4834f5b0f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3105390
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
839d5318
|
2021-08-16T08:29:28
|
|
Capture/Replay: Changes to MEC first frame capture.
Recent changes made it so it was impossible to trigger MEC *after*
the first frame and *before* the second frame. Instead it was only
possible to capture *before* the first frame or *after* the second
frame. Makes it possible for the retracing script to do identical
captures.
Also includes a number of refactorings to make it simpler to follow
the code, including renaming methods, and removing some extra
output files.
Bug: angleproject:5133
Change-Id: Ice6a189eb9f4d53e8ee1ba39beb537af2ef5fd9f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3097807
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
ad286e71
|
2021-08-17T10:43:47
|
|
Capture/Replay: Clean up ResourceTracker access.
This reduces the boilerplate needed to tracked genned/deleted
resources.
Refactoring change only.
Bug: angleproject:5133
Change-Id: I81f8877b2c308fe9d4136999f3ca63371a070720
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3100591
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
62cac8b5
|
2021-08-06T18:24:52
|
|
Capture/Replay: Add EGL support to generate_entry_points.py
Add EGL support to generate_entry_points.py. This is a pre-requisite to
frame capture being able to generate EGL calls, which is required for
multi-context support.
Bug: angleproject:5878
Bug: angleproject:4035
Change-Id: I3b8e9c91f9e2820b5207fc02d858e8109921f581
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3078993
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
3fd572c4
|
2021-08-04T12:45:09
|
|
FrameCapture: Fix setCaptureActive() and MEC timing
Frame capture is currently off by one when calling setCaptureActive()
and collecting the mid-execution capture state. The MEC state is
collected at frame N, but setCaptureActive() isn't called until frame
N+1, meaning all of the maybeCapturePreCallUpdates() and
maybeCapturePostCallUpdates() calls are not tracking resources correctly
for frame N.
This CL fixes when setCaptureActive() is called relative to when
setupSharedAndAuxReplay() is called, so the trace is active for the
first frame and all of the resources are correctly tracked.
Additionally, setCaptureActive() during the FrameCaptureShared() ctor if
frame capture starts with the first frame.
Bug: angleproject:6225
Change-Id: Id07e78b2da9c9d33779e20ab6a42f63cd103a6a6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3071940
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
35bdaf8d
|
2021-08-03T08:12:26
|
|
Capture/Replay: Minor cleanups.
Removes a duplicate assignment to the enabled variable. In the cpp
it's set to true, and in the header it was false. Also updates a
few variable names to be consistent with the group.
Bug: angleproject:5133
Change-Id: I3fb00ecc474191bea7c3a650ce23805b6f02d667
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3066007
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
262dc97c
|
2021-07-27T11:13:59
|
|
FrameCapture: Capture the EGL color space.
Add a SurfaceParams struct containing extents and color space so
a single map can be used for all draw surface parameters.
Persist the color space in the trace's header.
Bug: angleproject:5857
Change-Id: I327da8c8870f2d76d057954af57d977a10ebb59f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3056371
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
b5630d5c
|
2021-07-19T20:39:02
|
|
Capture/Replay: Add textures to resource tracking
This CL:
- Shifts resource reset tracking to be stored in an array of structs
indexed by type. This helps curb the complexity of adding more
resource types.
- Moves buffer and program tracking to the new layout.
- Adds textures to the resource tracker, using new layout.
Note that only aspects common to all ResourceIDTypes have been moved
to the new layout. Unique pieces of data, like buffer mapping state
and GLSync handling, remain in resource tracker alone.
Since texture setup is quite involved, this CL takes the approach of
having each call applied to two call chains at the same time;
SetupReplay calls and ResetReplay calls.
ResetReplay ends up with a sequence similar to buffers:
...
const GLuint deleteTextures[] = {
gTextureMap[1], gTextureMap[2], ... gTextureMap[n]};
glDeleteTextures(<count>, deleteTextures);
...
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
UpdateTextureID(1, 0);
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
UpdateTextureID(2, 0);
glGenTextures(1, reinterpret_cast<GLuint *>(gReadBuffer));
UpdateTextureID(3, 0);
...
glBindTexture(GL_TEXTURE_2D, gTextureMap[1]);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, 9729);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_R, 33071);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, 33071);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, 33071);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_COMPARE_FUNC, 519);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 0);
glTexImage2D(GL_TEXTURE_2D, 0, 6406, 512, 512, 0, GL_ALPHA, GL_UNSIGNED_BYTE,
reinterpret_cast<const GLubyte *>(&gBinaryData[183263280]));
glBindTexture(GL_TEXTURE_2D, gTextureMap[2]);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, 9987);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_COMPARE_FUNC, 519);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 6);
glTexStorage2D(GL_TEXTURE_2D, 7, GL_COMPRESSED_SRGB8_ETC2, 64, 64);
...
Test: PUBG Mobile MEC
Bug: b/180418573
Bug: angleproject:6087
Change-Id: I9f8e151c12aec5b2b7af376b8e0ff644ff9c61c4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3016114
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
1e75181f
|
2021-07-15T13:39:08
|
|
Capture/Replay: redesign capturing of arrays passed by pointer
Instead of doing a lookup of the number of passes array when the
call is written, store the number of array values in the
ParamCapture when the call is captured.
Bug: angleproject:6164
Change-Id: I87b0e2f776a6884b999cc50844e0777cda26b380
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3031543
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
|
|
|
223cd0ac
|
2021-06-15T18:46:07
|
|
Capture/Replay: Refactor shared context handling
This is the initial CL to enable capture/replay of multi-context
applications.
This CL refactors FrameCapture and FrameCaptureShared to move much of
the functionality into FrameCaptureShared, since most everything is
shared by Contexts in the share group. For example, the setup of the
majority of the GL objects is done in the new SetupReplayContextShared()
function in the new $LABEL_capture_context_shared_frame001.cpp file. The
setup is performed by (for example):
void SetupReplay()
{
$LABEL::InitReplay();
$LABEL::SetupReplayContextShared();
SetupReplayContext2();
}
This performs the shared setup first, followed by the context-specific
setup, which may reference shared objects careated by
LABEL::SetupReplayContextShared().
No re-capturing is required with this change, since the external APIs
(i.e., SetupReplay()) are still the same.
Bug: angleproject:5878
Test: Manual MEC and replay of Magic Tiles 3, Candy Crush Soda Saga, Temple Run 2
Change-Id: Iab7bfe651437e9be1dee83514cd97acc20c61d1d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2965780
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
5fb5f63b
|
2021-06-29T09:56:43
|
|
Capture: Write captured context for empty frames
When a frame is empty, still write the captured context for that
frame because the index file will contain the frame file, and the
replay will execute the according swapbuffer so that the replay will
query the context serialization when testing trace correctness.
With that enable RobustResourceInitTest.SurfaceInitializedAfterSwap/*
for capture/replay.
Bug: angleproject:5939
Change-Id: Ib280d55f739cc2bb8d2da0919e98c37a2cd576ba
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2987991
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
aec47ff6
|
2021-06-22T15:47:43
|
|
Capture/Replay: Update CopyImageSubData params
The parameters to glCopyImageSubData are flexible and
can take a TextureID or a RenderbufferID as a GLuint.
Our replay needs to remap those values, so we'll convert the GLuint
based on the target. This leads to a change like the following.
Before:
glCopyImageSubData(138, 0x0DE1, 0, 0, 0, 0,
642, 0x0DE1, 1, 0, 0, 0,
256, 256, 1);
After:
glCopyImageSubData(gTextureMap[138], 0x0DE1, 0, 0, 0, 0,
gTextureMap[642], 0x0DE1, 1, 0, 0, 0,
256, 256, 1);
Test: PUBG Mobile MEC
Bug: angleproject:6087
Bug: angleproject:6104
Change-Id: I5cd422e41ffbb4f08c8909e520bdce63e3008c5a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2981464
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
c8677c73
|
2021-06-12T17:23:42
|
|
Capture: Add a specialization for const GLuint *
Bug: angleproject:6058
Change-Id: I31cad1dd23d313aae625874e140c406747e0f4c4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2957953
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
78437770
|
2021-06-12T15:48:26
|
|
Capture: read four values for GL_TEXTURE_CROP_RECT_OES
This texture parameter describes a cropping region x,y, width, height.
Bug: angleproject:6057
Change-Id: I2c71f6aec1e536f42e06c19cb025889094b57a35
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2956510
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
22716d31
|
2021-06-11T13:51:41
|
|
Capture: Add specialization for ParamType::TGLubyte
Also enable "CurrentColorTest.Set/*" tests.
Bug: angleproject:6049
Change-Id: I65e62b6c0805687ad6744a12217e0c08bc97ca1a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2954262
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
90f9ddd9
|
2021-06-10T14:35:13
|
|
Capture: Add support for capturing (Get)ClipPlanef equation
In addition, correctly handle "const GLfloat *" when replaying,
and enable the according capture/replay tests
Bug: angleproject:6047
Change-Id: I5dd7510d1b621d4c941490c9d0261559a96dd681
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2951133
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
cb4f8a79
|
2021-06-08T10:39:47
|
|
Capture: Add a specialization for TGLeglImageOES
When writing out the parameter TGLeglImageOES an additional cast is
needed.
Bug: angleproject:6036
Change-Id: I7470a58a6ae17b0185d0c2e3fe1184b698f3efaf
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2944952
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
|
|
|
421c665c
|
2021-05-13T21:17:16
|
|
Capture: Make writeCppReplayIndexFiles a method of FrameCapture
This is useful to not pass a large amount of parameters around.
Bug: angleproject:5965
Change-Id: I3890a0ca852729bbf74725fb9e5f4062e20e9158
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2894485
Commit-Queue: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
8ea11eb7
|
2021-05-13T15:54:41
|
|
Capture/Replay: Reset programs on loop
Apps that create new programs during the run need to have them
deleted on loop. This CL handles that. It does not handle programs
that have been deleted and need to be recreated, or that have been
modified otherwise. An assert has been included to catch that for
future needs.
With this CL, entries like this will show up in ResetReplay:
...
glDeleteProgram(gShaderProgramMap[120]);
glDeleteProgram(gShaderProgramMap[121]);
...
Test: Pokemon Go MEC
Bug: b/188091629
Bug: angleproject:5968
Change-Id: I78c425e8fe95792fc626484641d067613dfae971
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2895326
Commit-Queue: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
7444466c
|
2021-04-19T18:44:13
|
|
Reland "Capture/Replay: Reset GL Fence Sync objects"
This is a reland of 9b4fd5456e3cc8c5412af96212d2840b422208b5
This fixes the original CL by moving where setCaptureActive() is called
to allow the first replay source file to be written.
Original change's description:
> Capture/Replay: Reset GL Fence Sync objects
>
> Asphalt 9 uses GL Fence Sync objects during rendering, which results in
> Fence Sync objects being created during setup and deleted while
> replaying frames. When the replay is restarted, the Fence Sync objects
> that were created during setup and deleted during the replay need to be
> recreated during the reset phase.
>
> Bug: angleproject:5883
> Bug: angleproject:4599
> Change-Id: I118f2b7208c4d512ab646b10f52b3a0936895089
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2838237
> Commit-Queue: Tim Van Patten <timvp@google.com>
> Reviewed-by: Cody Northrop <cnorthrop@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Bug: angleproject:5883
Bug: angleproject:4599
Bug: angleproject:5900
Change-Id: I5e1d901d8875007691699b7b973f3fb7db027337
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2850758
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
3839fa5b
|
2021-04-26T13:08:15
|
|
Revert "Capture/Replay: Reset GL Fence Sync objects"
This reverts commit 9b4fd5456e3cc8c5412af96212d2840b422208b5.
Reason for revert: Breaks capture:
INFO:root:Passed: 0, Comparison Failed: 0, Crashed: 0, CompileFailed 0, Skipped: 745, Timeout: 0
Bug: angleproject:5900
Original change's description:
> Capture/Replay: Reset GL Fence Sync objects
>
> Asphalt 9 uses GL Fence Sync objects during rendering, which results in
> Fence Sync objects being created during setup and deleted while
> replaying frames. When the replay is restarted, the Fence Sync objects
> that were created during setup and deleted during the replay need to be
> recreated during the reset phase.
>
> Bug: angleproject:5883
> Bug: angleproject:4599
> Change-Id: I118f2b7208c4d512ab646b10f52b3a0936895089
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2838237
> Commit-Queue: Tim Van Patten <timvp@google.com>
> Reviewed-by: Cody Northrop <cnorthrop@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Bug: angleproject:5883
Bug: angleproject:4599
Change-Id: I56c682e1d90dcdde7b18c941aad976825de3af42
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2851057
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
9b4fd545
|
2021-04-19T18:44:13
|
|
Capture/Replay: Reset GL Fence Sync objects
Asphalt 9 uses GL Fence Sync objects during rendering, which results in
Fence Sync objects being created during setup and deleted while
replaying frames. When the replay is restarted, the Fence Sync objects
that were created during setup and deleted during the replay need to be
recreated during the reset phase.
Bug: angleproject:5883
Bug: angleproject:4599
Change-Id: I118f2b7208c4d512ab646b10f52b3a0936895089
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2838237
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
fbfecbe2
|
2021-04-07T11:26:13
|
|
Capture/Replay: Add PPO/glProgramUniform support
Command and Conquer: Rivals requires additional frame capture API
support:
- Program Pipeline Objects
- glProgramUniform*
Bug: angleproject:5830
Change-Id: I159086f92d2dfead0a513cd17fadeda7df92f408
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2809891
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
fddbc9c7
|
2021-03-04T20:45:01
|
|
capture: Implement capturing GLES1 vertex pointers.
Implement capturing glVertexPointer, glTexCoordPointer, glNormalPointer,
glColorPointer and glPointSizePointerOES using a new
CaptureVertexPointerGLES1 helper function.
This is done by using fixed indices for gClientArrays that are retrieved
from the GLES1Renderer.
Bug: angleproject:5751
Change-Id: I6c774ff21942ea3422c4c77a615f88299901c0ab
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2773288
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
a70a6a9d
|
2021-03-09T10:38:13
|
|
Capture/Replay: Fix compressed cube textures
We were only tracking one image per level of cube map textures.
Instead, we need to track one per face (6).
Test: MEC for Extreme Car Driving Simulator
Bug: b/180419767
Bug: angleproject:5735
Change-Id: I59e6a5e83a60666a29f44d0a1e1993a1b461e8e7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2744293
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
|
6fc10389
|
2021-02-10T11:20:16
|
|
Move Frame Capture to capture/ folder.
This will make it easier to trigger the trace tests when these files
are modified.
Bug: angleproject:5530
Change-Id: I5f0c450595b380cd91b20c1477dc1845bee35dd9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2686120
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|