|
f5f419ec
|
2024-09-04T20:22:20
|
|
Vulkan: Add verify-restore in CompressAndStorePipelineCacheVk()
Change fixes following problem:
When blob cache is full, but still allows to store the current
pipeline cache data, storing next chunk may trigger eviction of
already stored items. Depending on the blob cache implementation,
eviction process may choose to evict chunks from the current pipeline
cache data. As the result: blob cache will not contain all chunks.
Solution:
The `StorePipelineCacheVkChunks()` function now check what chunks
exist in the blob cache (if `lastNumStoredChunks` parameter is positive)
and restores any missing chunks if their number is less than
`lastNumStoredChunks`.
First call to `StorePipelineCacheVkChunks()` passes zero (0) to the
`lastNumStoredChunks`, which will skip the check and store all chunks
unconditionally.
If "verifyPipelineCacheInBlobCache" feature is enabled, the
`StorePipelineCacheVkChunks()` will be called again to restore possibly
missing chunks.
If "hasBlobCacheThatEvictsOldItemsFirst" feature is disabled,
the function will be called repeatedly in a loop (opposed to only once),
until all chunks are present or until number of missing chunk stops
decreasing.
Bug: angleproject:4722
Change-Id: Ibbbdb2f878d6aafd7ab5586fc654e2b9be900a9a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5837042
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
|
|
|
8469debb
|
2024-09-04T20:04:10
|
|
Vulkan: Add dual slots in CompressAndStorePipelineCacheVk()
Change fixes following problem:
Currently, each call to `CompressAndStorePipelineCacheVk()` stores
chunks in order, starting from 0. This overrides previously stored
chunks. In case of app termination (kill) in the middle of this
process, the entire cache data will be corrupted, since it will
partially contain chunks from the new and old caches.
Solution:
In order to fix this problem, this change introduces `slotIndex` into
the `chunkCacheHash` calculation. Slot index is managed by
`vk::Renderer::getNextPipelineCacheBlobCacheSlotIndex()` method, which
will alternate between 0 and 1 when "useDualPipelineBlobCacheSlots"
feature is enabled, and always 0 otherwise.
Additionally, chunk storing order is reversed: last chunk is stored
first and the first (0 chunk) - last. This is done because 0 chunk is
the first that is loaded in `GetAndDecompressPipelineCacheVk()` and used
as indication that there is data in the cache. Writing it last, ensures
that other chunks will be also available.
When "useDualPipelineBlobCacheSlots" is enabled, each call to
`CompressAndStorePipelineCacheVk()` will use slot index opposed to the
slot that is stored in the cache, avoiding damaging existing data. After
writing all chunks for a brief moment there may be 2 instances of the
data. However, data for the previous slot will be immediately erased
(by writing 1/0-sized blobs) starting from the 0 chunk.
To control if erasing of old pipeline cache data will be erased by using
0-sized or 1-sized blobs blobs, added
`useEmptyBlobsToEraseOldPipelineCacheFromBlobCache` feature.
The `GetAndDecompressPipelineCacheVk()` function will iterate over each
available slot index checking only 0 chunk until data is found.
In case of the OpenCL API, features will always have following values:
- "useDualPipelineBlobCacheSlots" -> false
- "useEmptyBlobsToEraseOldPipelineCacheFromBlobCache" -> true
Note: this solution requires 2X pipeline cache size space in the blob
cache to work as expected, otherwise it will exacerbate other problem:
When blob cache is full, but still allows to store the current
pipeline cache data, storing next chunk may trigger eviction of
already stored items. Depending on the blob cache implementation,
eviction process may choose to evict chunks from the current pipeline
cache data. As the result: blob cache will not contain all chunks.
The above problem will be addressed in the follow up CL.
Bug: angleproject:4722
Change-Id: I2920bc3d89263280cdfe0466446fca26415e2b25
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5756576
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
|
|
d4f93407
|
2024-03-25T18:14:26
|
|
CL: Conditional enable of Vulkan Validation layers
Rather than force YesIfAvailable always,
we restrict the Vk validation layers if
user specifies it in the GN args:
angle_enable_vulkan_validation_layers
Bug: angleproject:356320155
Change-Id: I9035d0e7d220ce59db39f1a86f31b403c44088dd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5753765
Commit-Queue: Austin Annestrand <a.annestrand@samsung.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
3813e805
|
2024-06-21T14:29:55
|
|
Add env var/property control of API dump layer
Layer is now included with all debug/assert builds or when
specified through GN arg.
VK_VULKAN_API_DUMP_LAYER on desktop, or for android set
debug.angle.enable_vulkan_api_dump_layer to 1.
Also included some renaming since layers other than
validation are supported.
Bug: angleproject:8623
Change-Id: I2bb976dede53712f781a531d0c4f3ce4d33f052c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5649937
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Mark Łobodziński <mark@lunarg.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
c3a1cae4
|
2024-04-15T14:58:55
|
|
Use angle::SimpleMutex everywhere in libGLESv2
Only cases left that use std::mutex are:
- Share group and the context ErrorSet mutexes as they need try_lock()
- Anywhere mutexes are used in conjunction with std::condition_variables
(as they explicitly require std::mutex)
Bug: angleproject:8667
Change-Id: Ib6d68938b0886f9e7c43e023162557990ecfb300
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5453294
Reviewed-by: Roman Lavrov <romanl@google.com>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
bb70f5d1
|
2024-03-13T06:02:40
|
|
CL/VK: Setup platform as provider of vk::Context
Setup the CLPlatformVk to be the implementer of `vk::Context`.
Initialize the Renderer instance during platform setup. This removes the
dependency on the DisplayVk instance and as a consequence any associated
egl state.
Bug: angleproject:8515
Change-Id: If1848b002d6309911305ca3cb93c316ed7f48965
Signed-off-by: Gowtham Tammana <g.tammana@samsung.com>
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5406612
Commit-Queue: Austin Annestrand <a.annestrand@samsung.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
a8e9aa26
|
2024-03-11T23:50:18
|
|
CL/VK: Setup GlobalOps for the platform
Setup CLPlatformVk as a provider of vk::GlobalOps, and use the
vk::GlobalOps associated thread pool for async tasks.
Bug: angleproject:8515
Change-Id: I3e1ae069afabeadfdfa02f4a2f99419882f91c6a
Signed-off-by: Gowtham Tammana <g.tammana@samsung.com>
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5403225
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
47ca7552
|
2024-03-07T14:19:06
|
|
CL/VK: Select Vulkan Angle as platform
Ask for Vulkan Angle platform when setting up the platform. Else it is
picking from the default one, which might not be Vulkan.
Bug: angleproject:8515
Change-Id: I1f43a5de3eb2c72900631695e2b3a5ffef3d966d
Signed-off-by: Gowtham Tammana <g.tammana@samsung.com>
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5403224
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
914fe61b
|
2024-03-15T13:20:49
|
|
Vulkan: Rename RendererVk.* to vk_renderer.*
Done in a separate CL from the move to namespace vk to avoid possible
rebase-time confusion with the file name change.
Bug: angleproject:8564
Change-Id: Ibab79029834b88514d4466a7a4c076b1352bc450
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5370107
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Commit-Queue: Amirali Abdolrashidi <abdolrashidi@google.com>
|
|
|
60aaf4a0
|
2024-03-14T12:58:56
|
|
Vulkan: Move renderer to namespace vk
This class is agnostic of EGL. This change moves it to namespace vk for
use with the OpenCL implementation
Bug: angleproject:8564
Change-Id: I57f7807d6af8b3d5d7f8efbaf8b5d537a930f881
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5371324
Reviewed-by: Austin Annestrand <a.annestrand@samsung.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
af56ca61
|
2024-01-11T13:14:32
|
|
OpenCL/Vulkan: Initial support for context
Introduce OpenCL context object creation for
Vulkan backend in ANGLE.
Bug: angleproject:8499
Change-Id: I56794649c87d5cf1490e496525e6534c8f90ce32
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5233366
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Austin Annestrand <a.annestrand@samsung.com>
|
|
|
c673c837
|
2024-01-08T19:05:38
|
|
OpenCL/Vulkan: Initial support for platform/device
Introduce OpenCL platform and device object creation
for Vulkan backend in ANGLE.
Bug: angleproject:8498
Change-Id: I5af9a23ed343648115bd9f0e81cd7a53080e2982
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5233365
Commit-Queue: Austin Annestrand <a.annestrand@samsung.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
66d29149
|
2023-10-16T12:43:43
|
|
OpenCL: Add scaffolding to renderer/vulkan code
Start off ANCLE (OpenCL on ANGLE) with backend scaffolding code with
placeholder API error returns until they are later implemented.
Bug: angleproject:8377
Change-Id: I223d3482cce097ccb298e121fc03ec416e8958fd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4950556
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
ed7c31c7
|
2023-11-29T18:31:43
|
|
OpenCL: Replace error-passing APIs to use angle::Result
Part 2 of Issue 8435:
This part replaces existing error handling for all CL APIs that
return CL objects to now use "angle::Result".
Bug: angleproject:8435
Change-Id: I75e41c7cbb06220cabec7cc9f9cb5107a3b6bd6d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5075773
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
a9512a9f
|
2023-11-29T14:56:29
|
|
OpenCL: Replace error-return APIs to use angle::Result
The motivation of converting existing CL APIs using
"cl_int errorCode" to now use "angle::Result" is for better
interop with internal ANGLE plumbing/abstractions.
To facilitate this, we now make use of a TLS errorCode
object that we set internally and read from entrypoint/stubs
layer so that we can propagate back to user.
Bug: angleproject:8435
Change-Id: I7c2a786019c33c56649be4b6e8b61bc318497bb1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5075772
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
d8c18ac6
|
2021-10-08T13:00:19
|
|
Re-land: "Isolate commit_id.h from other code."
Re-land limits the angle_version deps to Windows to fix an iOS
visibility rule violation in GN.
Instead of recompiling multiple source files, we can seal off the
files that include commit_id to prevent recompilations when the
commit changes and source files don't change.
Bug: angleproject:2551
Change-Id: I033f00ec7afe4bfd01e29e0eea8848eea27747a0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3233899
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
7defdb60
|
2021-10-19T14:07:07
|
|
Revert "Isolate commit_id.h from other code."
This reverts commit 6d7ebf656d06f1152967aa5f4457d3787a24732c.
Reason for revert: Suspected reason for failing ANGLE roll into Chromium
Original change's description:
> Isolate commit_id.h from other code.
>
> Instead of recompiling multiple source files, we can seal off the
> files that include commit_id to prevent recompilations when the
> commit changes and source files don't change.
>
> Bug: angleproject:2551
> Change-Id: I98800c5a8e04ec7d4c3e57ea0eb3b636d3b224bb
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3212895
> Commit-Queue: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Geoff Lang <geofflang@chromium.org>
> Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Bug: angleproject:2551
Bug: angleproject:6588
Change-Id: I19f41141f57d0683efb9a4ec043e078a9f7dc925
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3226056
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
|
|
|
6d7ebf65
|
2021-10-08T13:00:19
|
|
Isolate commit_id.h from other code.
Instead of recompiling multiple source files, we can seal off the
files that include commit_id to prevent recompilations when the
commit changes and source files don't change.
Bug: angleproject:2551
Change-Id: I98800c5a8e04ec7d4c3e57ea0eb3b636d3b224bb
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3212895
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
|
|
69562546
|
2021-06-11T19:12:11
|
|
CL: Refactor info structs and fix conformance bug
- Remove variable name prefix from Info structs to be more consistent
with other ANGLE structs.
- Fix CL object validation check with magics, since the Mesa solution
doesn't work without RTTI.
- Add support for some extensions required by OpenCL 1.1 and for some
optional extensions.
- Fix more conformance bugs.
Bug: angleproject:6015
Change-Id: I41b1c45d95059a9994f5dc78bf9b74476cc6f2d4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2956349
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
82654e5d
|
2021-06-06T15:46:53
|
|
CL: Add program build commands
Add program build commands to front end and pass-through back end.
Bug: angleproject:6015
Change-Id: I1175646d35ff050796b718d7f0e269520292b319
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2940839
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
baca10b7
|
2021-05-29T16:26:57
|
|
CL: Remove object cache and fix multi-threading
All CL front end objects used to be cached to be able to determine
if an object has been created by the front end to check its validity.
The validity is now checked with the existence of an intrinsic value
(the dispatch table pointer), which is consistent with the patterns
found in Mesa and clvk (though clvk uses a magic value).
This allows the removal of all cached objects.
The cached objects were stored with std::unique_ptr. These are now
gone and all remaining pointers are now custom intrinsic reference
count pointers.
Also remove global lock which causes deadlocks, e.g. when CL API
is called from a separate thread to unlock a blocking call with
a user event. Most of the front end is constant and already
thread-safe. The ref count is also thread-safe now (atomic).
A few remaining locks will follow.
Without the global lock it was now possible to make the API
reentrant, and to remove the workaround with the Khronos ICD
loader to skip ANGLE's OpenCL library.
Bug: angleproject:6001
Change-Id: I7d3b52db9011a02cb7ea9ebdeb6e22c4c702ef5b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2927395
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
9459456b
|
2021-05-27T13:43:41
|
|
CL: kernel creation for front end & passthrough
Add kernel object to back end and implement creation and info query.
Bug: angleproject:6001
Change-Id: I6e3fdd2b35d7e73ed56144dc938bf21436d3559f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2922150
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
51ab3dc3
|
2021-05-25T16:54:11
|
|
CL: Improve validation readablity
Improve validation readablity by introducing a bit field class
and by changing how the CL error code is returned.
Bug: angleproject:6001
Change-Id: I51deb745454e5281de725481edef85eb30be28c7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2919690
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: John Plate <jplate@google.com>
|
|
|
10a8182c
|
2021-05-19T21:00:05
|
|
CL: Fix querying default device if non exists
Bug: angleproject:5992
Change-Id: Ie43f905fbb9cf41b0f6f88b2db27ebe9de9c37e8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2906993
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
c2fd3388
|
2021-05-15T22:28:27
|
|
CL: Add front end object references to back end objects
Add front end object references to back end objects, which requires
a significant amount of refactoring, because the back end objects
have to be constructed during the construction of the front end
objects, so that the references can be passed to the back end objects,
which then can be passed to the front end member initialization.
That would have been easier with inheritance than with PImpl.
Bug: angleproject:5904
Change-Id: Ib58e6a698e76987bdd63cd8088f923424d6c622b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2897249
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
48b835cf
|
2021-05-09T22:47:00
|
|
CL: Implement context for front end and passthrough
Bug: angleproject:5904
Change-Id: I23b764bba87be3a51a1b5b44b13968fc572efde9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2883773
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: John Plate <jplate@google.com>
|
|
|
93aff937
|
2021-05-08T16:46:22
|
|
Refactor CL platform object
Move GetPlatformInfo implementation from stubs to cl::Platform,
because stubs are meant to be small.
Also move CLPlatformImpl::Info instance to cl::Platform as it is
more efficient to be accessed from there, and more consistent with
cl::Device.
Bug: angleproject:5904
Change-Id: I4c0ba6390467d6ef357e38a5ef08f54b0a88d423
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2880663
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
e369a282
|
2021-05-08T14:08:28
|
|
Implement CL device object for front end and passthrough
Bug: angleproject:5904
Change-Id: I17f6d10ed59c62fe9126d29b7de2c127c5d0039a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2880662
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: John Plate <jplate@google.com>
|
|
|
dfe208f1
|
2021-04-28T18:26:00
|
|
Add loader for CL pass-through back end
Bug: angleproject:5904
Change-Id: If4960f3150f6bbc85a30f0f6ac5c9e668e6ff756
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2864022
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
16a919b1
|
2021-04-27T11:36:11
|
|
Add stubs for CL platform layer back ends
Bug: angleproject:5904
Change-Id: If23c7f76013a17c3c67c13194566438035beb3d1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2853582
Commit-Queue: John Plate <jplate@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|