kc3-lang/angle/src/libANGLE/renderer/vulkan/SyncVk.cpp

src/libANGLE/renderer/vulkan/SyncVk.cpp

Log

Author Commit Date CI Message
Shahbaz Youssefi 52fe3116 2023-07-17T16:20:54 Vulkan: Deduplicate share group's context set tracking Bug: angleproject:8224 Change-Id: I7a59a37229682fb91ff777f31e02e05d7ab2b80f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4690345 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Charlie Lao 16657963 2023-06-28T09:42:52 Vulkan: Improve SyncVk::serverWait If sync object has already finished, there is no need to insert pipeline barrier. With this CL, asphalt_9 average frame time reduces from 5.464 ms to 5.268 ms (-3.58%), making it 85% of native GLES (4.49 ms) on pixel 7 pro. Bug: b/277644512 Change-Id: I112ccef0619da3c05cba44066327edc0bb81c366 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4651599 Reviewed-by: Roman Lavrov <romanl@google.com> Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Igor Nazarov 54532241 2023-01-19T15:45:29 Vulkan: Remove mUse related code from SyncHelperNativeFence Since adding `rx::vk::ExternalFence` class, waiting on `mUse` is almost the same as waiting directly on `mExternalFence`. Because `mUse` is no longer required to release the fence and to reduce code complexity, `mUse` related code paths were removed. Since `SyncHelperNativeFence` no longer using `mUse` no need to inherit from `SyncHelper`. Added `SyncHelperInterface` abstract base instead. Bug: angleproject:8115 Change-Id: I2aabbc905d833f8bf6db36b3be356d6db40349d0 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4392880 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
Igor Nazarov 903d9fdf 2023-01-19T15:37:45 Vulkan: Implement ExternalFence for use in SyncHelperNativeFence `ExternalFence` allows concurrent usage in `CommandQueue` and `SyncHelperNativeFence` classes eliminating need of additional `vkQueueSubmit()` call. Waiting in `CommandQueue` on `QueueSerial` or `ResourceUse` will ensure corresponding state of the native FD (because `CommandQueue` will wait on the same FD instead of some other fence). After this change there will be only single `vkQueueSubmit()` call from the `SyncHelperNativeFence::initializeWithFd()` method. This CL and the follow-up is sufficient to fix the bugs below. Bug: angleproject:8115 Bug: angleproject:8117 Test: angle_end2end_tests --gtest_filter=EGLSyncTest.AndroidNativeFence_ExternalFenceWaitVVLBug* Change-Id: Ic562ecc71a95203454a1dc438589a13bcf3bff7f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4392879 Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Igor Nazarov <i.nazarov@samsung.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Igor Nazarov e71207de 2023-01-18T13:39:58 Vulkan: Fixes in rx::vk::SyncHelperNativeFence class Issues: - Resource::mUse was not updated in initializeWithFd() method, making code paths, that checking for "mUse" a dead-code. - mFenceWithFd may be potentially destroyed while still in use. (Because mUse is not valid). - missing checkCompletedCommands() in getStatus(), that may cause infinite loop in the application. Bug: angleproject:8115 Change-Id: I857d785f564b809eb83417fe5f212f7d0f12c5b5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4392878 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
Igor Nazarov e24f4519 2023-01-19T02:30:39 Vulkan: Add externalFence into submitCommands() Currently one-off fence in the `queueSubmitOneOff()` is used only in `SyncHelperNativeFence::initializeWithFd()` to submit external fence. Other `queueSubmitOneOff()` calls may use `QueueSerial` instead of a fence. Providing `fence` into `queueSubmitOneOff()` prevents tracking that submission with `QueueSerial`. Therefore using `mUse` to collecting `mFenceWithFd` as garbage will not work as intended. This CL removes `fence` from `queueSubmitOneOff()` and adds optional `externalFence` into `submitCommands()` instead. Providing `externalFence` will cause additional `vkQueueSubmit()` call: - first submission will submit everything as usual except using the `externalFence`. - second, will only submit internal `CommandQueue` fence for `QueueSerial` tracking. As the result of this CL, call to `initializeWithFd()` will always produce two (2) `vkQueueSubmit()` calls. Previously it may be one (1) or two (2) submissions. Future CL will reduce submission count to one (1). If add additional submission into `queueSubmitOneOff()` instead of `submitCommands()`, then maximum number of submissions will be three (3). Bug: angleproject:8117 Change-Id: I6f1ec12682aaab71bfc871e665fec2659df96b26 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4392877 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
Shahbaz Youssefi 5c04f18a 2023-04-24T16:18:24 Vulkan: Remove DisplayVk param from ToEGL It's no longer used Bug: angleproject:3041 Change-Id: I5063152d1598aa5d40d94bbf5c643a1288589037 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4470387 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Igor Nazarov <i.nazarov@samsung.com> Reviewed-by: Charlie Lao <cclao@google.com>
Charlie Lao d58fbf04 2023-04-05T12:32:09 Vulkan: Wait for surface ANI semaphore only if image is used Right now there is a bug that surface's ANI semaphore is added to context when WindowSurfaceVk::getAttachmentRenderTarget get called, which gets called from FramebufferVk::syncState, which is before we end the previous render pass, due to the endRenderpass usually is deferred until next renderPass starts. This caused ANI semaphore gets added to the previous render pass's submission, which does not use surface, and thus a bubble in GPU execution pipeline where the user FBO rendering gets unnecessarily blocked until ANI semaphore is signaled. This lowers GPU utilization and thus GPU frequency gets dropped and frame time increased. This CL stores ANI semaphore to ImageHelper object and when barrier is generated, the ANI semaphore is moved to CommandBuffer. When CommandBuffer gets flushed and submitted, it gets added to the waitSemaphores vector and submitted to vulkan. Since all use of swap chain image must go through barrier code first (you need at least change layout), this ensures ANI semaphore gets waited in exact and robust way. Only the submission that references the swap chain image will be waited. With this CL, professional_baseball_spirits reduces frame time from 3.8 ms to 2.7 ms, achieving parity with native GLES on pixel 7 pro. Bug: b/275624771 Change-Id: Ifa6cacf9e3bc147bfde54eb7def2fca48c50aca0 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4400011 Reviewed-by: Yuxin Hu <yuxinhu@google.com> Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Charlie Lao c402ea1c 2023-02-15T12:01:38 Vulkan: Rename hasUnfinishedUse to hasResourceUseFinished Most usage of hasUnfinishedUse is for !hasUnfinishedUse, and there was feedback that negative API is not preferred. This CL changes it to positive API name. Similarly renamed hasUnsubmittedUse to hasResourceUseSubmitted. Bug: b/267348918 Change-Id: Idb10b0f998ec50116ffb6aada19a98a516e87824 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4257105 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Igor Nazarov 8b9bd182 2023-02-08T16:17:10 Vulkan: Refactoring to use ProtectionType enum instead of bool This is a second stage of refactoring that was started here: Vulkan: Minor CommandQueue implementation refactoring. 0210b46d35b51ea04bddafb48ba406a87c39e58e Enumeration renamed: CommandContent -> ProtectionType. Currently interfaces of ContextVk/RendererVk/CommandQueue use "hasProtectedContent" boolean. Internally CommandQueue uses "vk::ProtectionType" enumeration to separate states related to Unprotected/Protected commands. This CL replaces boolean with enumeration for consistency. Bug: angleproject:7995 Change-Id: Ibb98cce661358d464be7c6a8367a1297d7093b1c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4232114 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Igor Nazarov <i.nazarov@samsung.com>
Charlie Lao 128eb01d 2023-01-25T15:55:04 Vulkan: Only call checkCompletedCommands if needed. SyncHelper::getStatus() is always calling checkCompletedCommands. This CL makes it only call RendererVk::checkCompletedCommands if sync is not finished. If it is already finished, we could just immediate return "signaled". Bug: b/266220198 Change-Id: Ie7be9ea9c07d40984945ae3f20a5276c10b3322d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4195834 Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com> Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Charlie Lao 1219f55a 2022-12-07T16:19:37 Vulkan: Remove Resource::isCurrentlyInUse Due to header file include order, this function can not directly made inline. This CL removes the function and replace it with renderer->getUnfinishedUse() to reduce one extra function call of one line function. Bug: b/262048658 Change-Id: Ied33b63d0ec88336a5ce42cf7726f16b2b883b86 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4089623 Reviewed-by: Yuxin Hu <yuxinhu@google.com> Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Charlie Lao 2e5ca217 2022-11-18T10:44:49 Vulkan: Let each current context has its own QueueSerial. This CL makes every current context has its own queueSerial. At context creation time or when context becomes current, it allocates a QueueIndex from renderer. When it becomes non-current, it releases QueueIndex for others to reuse. This way we significantly reduces the max number of QueueIndexs for reasonable usage. Each CommandBuffer has its own unique QueueSerial and we use that to determine if a resource is being used by the given CommandBuffer. The QueueSerial for RenderPassCommands is deferred until renderPass starts, and when we generate queueSerial for renderPassCommands, we also reserve a range of serials for outsideRenderPassCommands so that we can do incremental submission of outsideRenderPassCommands without need to close renderPassCommands. In rare situation, if that reserved serials runs out, we also close renderPassCommands to ensure the ordering of serials matches ordering of command buffers. With per current context queue serial, this CL is able to set resource queue serial as it is being used. This CL completely removes usage of ResourceUseList class since it was introduced due to deferred setSerial. This CL also get rid of refCount from ResourceUse since there we no longer add it to a ResourceUseList. With that, we also able to remove SharedResourceUse class since access to ResourceUse itself is now thread safe since we are able to make a copy of it when we add it to GarbageList. Because RenderPassCommands now has its own unique QueueSerial as it encodes command, we can use it to detect if a resource is being used by it or not, thus this CL also removes usage of CommandBufferID. Bug: b/255414841 Change-Id: I36dcbeaa7bc996f04e6c04bf9ad44cd0d630f61a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4038096 Reviewed-by: Yuxin Hu <yuxinhu@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Charlie Lao <cclao@google.com>
Charlie Lao 7dd8478e 2022-11-17T10:11:02 Vulkan: Make ResourceUse::serial an FastVector of Serials In preparation for per context queue serial, this CL makes ResourceUse::serial a FastVector of Serial. Right now we still limited to one serial index so that it still work the same way as before. This CL adds necessary data type and change the function names to reflect that tracking GPU progress needs a ResourceUse object instead of a single Serial number. Bug: b/255414841 Change-Id: Ic60cdf5ec8da45d1821f65a55947f5c553f65737 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4034548 Reviewed-by: Yuxin Hu <yuxinhu@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Charlie Lao <cclao@google.com>
Charlie Lao 8378032e 2022-11-15T11:47:55 Vulkan: Remove get API for mLastCompletedQueueSerial In preparation for per context queue serial, this CL makes mLastSubmittedQueueSerial and mLastCompletedQueueSerial private to CommandQueue. Before this CL, we have a get function to return the last submitted serial and last completed serial and passing these serials around. This works because the serial is a single uint64_t number. With per context queue serial, this will be an array of serials and there is potential risk associated with access it from different threads. This CL makes these serials private to CommandQueue and when you want to know if GPU is completed with resource, you ask RendererVk/CommandQueue directly. This way we can ensure they have thread safe access in the CommandQueue (no lock is necessary, but all access will be restricted to one class). Bug: b/255414841 Change-Id: Ica565decce4a80588e0b447e179a2b634b55d7c3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4021676 Reviewed-by: Yuxin Hu <yuxinhu@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Charlie Lao <cclao@google.com>
Geoff Lang 313956f2 2022-09-07T10:15:10 Vulkan: Don't store EGLSyncVk::mAttribs as a const ref. EGLSyncVk::mAttribs is a stack allocated variable in the parent scope. Its usage is currently safe because it is only used in EGLSyncVk::initialize but it is dangerous to have this member present. Bug: angleproject:7637 Change-Id: Idaf2f1a0306e55a5fe00e55c9e72cd71005e3dce Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3876889 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Yuxin Hu <yuxinhu@google.com> Commit-Queue: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi 0bc70e96 2022-06-10T22:27:04 Vulkan: Rename getShareGroupVk to getShareGroup + other miscellaneous clean up Bug: angleproject:7375 Change-Id: I25690860478a2fd181a67ce2b6cb4d7aac7dbaa7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3700197 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Yuxin Hu <yuxinhu@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Shahbaz Youssefi 3dfc8004 2022-06-08T14:24:48 Vulkan: Optimize sync followed by swap Previously, inserting a sync object immediately caused a submission. That was done in https://chromium-review.googlesource.com/c/angle/angle/+/3200274 to be able to wait until the sync object is signaled without having to wait for whatever is recorded after it until a flush naturally happens. Some applications issue a glFenceSync right before eglSwapBuffers. The submission incurred by glFenceSync disallowed the optimizations that eglSwapBuffers would have done, leading to performance degradations. This could have been avoided if glFenceSync was issued right after eglSwapBuffers, but that's not the case with a number of applications. In this change, when a fence is inserted: - For EGL sync objects, a submission is issued regardless - For GL sync objects, a submission is issued if there is no render pass open - For GL sync objects, the submission is deferred if there is an open render pass. This is done by marking the render pass closed, and flagging the context as having a deferred flash. If the context that issued the fence sync issues another draw call, the render pass is naturally closed and the submission is performed. If the context that issued the fence sync causes a submission, it would have a chance to modify the render pass before doing so. For example, it could apply swapchain optimizations before swapping, or add a resolve attachment for blit. If the context that issued the fence sync doesn't cause a submission before another context tries to access it (get status, wait, etc), the other context will flush its render pass and cause a submission on its behalf. This is possible because the deferral of submission is done only for GL sync objects, and those are only accessible by other contexts in the same share group. Bug: angleproject:7379 Change-Id: I3dd1c1bfd575206d730dd9ee2e33ba2254318521 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3695520 Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Amirali Abdolrashidi 37c53684 2022-04-05T15:03:14 Split Context ResourceUseList to RP Commandbuffers * Split the resource use lists retained in context between the render pass command buffer and the outside render pass command buffer, mostly in vk_helper.cpp and ContextVk.cpp. Bug: angleproject:7103 Change-Id: Ib696bdec2c545cd7df9ae4ab86c54a294041d908 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3573581 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Mohan Maiya 93186337 2022-02-14T11:15:04 Vulkan: Check Serial is valid in SyncHelperNativeFence The change 6d94471bc70 incorrectly checked for validity of mUse in the clientWait and getStatus methods. mUse is valid once the constructor is complete, instead check for validity of the serial. Bug: angleproject:2517 Test: EGLSyncTest.AndroidNativeFence_* Change-Id: Id71eddc4ecf27cea487be56cb362afd68db1fd65 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3465455 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: mohan maiya <m.maiya@samsung.com>
Mohan Maiya a0fbcb37 2022-01-13T14:26:58 Vulkan: Remove flush from SyncHelperNativeFence::serverWait The spec doesn't require a flush in native fence's server wait method Bug: angleproject:2517 Test: EGLSyncTest.AndroidNativeFence_* Change-Id: If6870fe1e25ddf3a1b724e776f5408523a7bdf50 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3388231 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: mohan maiya <m.maiya@samsung.com>
Shahbaz Youssefi acd8fc76 2021-12-16T01:05:02 Vulkan: Distinguish RP and outside-RP command buffer types What goes inside and outside a render pass command buffer is largely mutually exclusive. Moreover, the size and frequency of allocations is different between the two. This change distinguishes the C++ types used for inside and outside render pass command buffers: - The type now documents which command buffer a function is able to receive. - `isRenderPass` flag passing, checking and asserting is largely removed. - A follow up change experiments with using different (Vulkan vs ANGLE) secondary command buffers for inside and outside RP command buffers. - A future change could specialize the pool behaviors per command buffer type. Bug: angleproject:6811 Change-Id: Ia4bc669d26ac7e94e8a0dfb9b361666c82f42cc3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3344373 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 3128c055 2021-11-12T14:59:46 Vulkan: Add wait semaphores to queueSubmitOneOff For use in follow up change. Bug: angleproject:3966 Change-Id: I5bfac51ef9d47a6df5d52268d3ce4863b848b1d1 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3279226 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Lingfeng Yang <lfy@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi dbc0c646 2021-11-06T01:09:26 Vulkan: Output the reason for RP closure in command buffer To make it easier when viewing the command buffer in a graphics debugger, this change inserts a marker just before closing the render pass that specifies why the render pass was closed. Bug: angleproject:2472 Change-Id: I862e500cd58332d6e199c853315c560fe6a73dc2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3265609 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi d134581b 2021-10-07T16:09:25 Vulkan: Remove SyncHelper::mEvent Now that inserting a sync object incurs a flush, we can use the serial to perform synchronization. Bug: angleproject:6464 Bug: angleproject:6481 Change-Id: I8c82a12855b6497861ce34854d165af82956b384 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3212573 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Tim Van Patten 27e7c4dd 2021-10-01T13:43:57 Vulkan: Submit commands when Sync objects are initialized This CL calls ContextVk::flushImpl() after the sync object (VkEvent) has been created and initialized, so it has a valid Serial when it's waited on later. This change allows us to remove any tracking of pending sync objects, since all sync objects will be flushed and submitted by the end of SyncHelper::initialize(). Bug: angleproject:6464 Test: MultithreadingTest.CreateFenceThreadAClientWaitSyncThreadBDelayedFlush Change-Id: I47e7ced452727b434ed974368311fb3439a107c9 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3200274 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Tim Van Patten <timvp@google.com>
Jeff Vigil d2b659f9 2021-03-25T15:40:58 EGL: implement EGL_EXT_protected_content Context Add Validation check to Contexts and Surfaces Add Vulkan protected memory features and properties Add protected member to Surfaces and Contexts Implement hasProtectedContent in Vulkan Add QueueFamily helper, extent DeviceQueueMap Protected Swapchains always on for Android Add EGLProtectedContentTest Test: angle_end2end_test --gtest_filter=EGLProtectedContentTest Bug: angleproject:3965 Change-Id: I9352b1e594f71bb4e89cee7137a468940d186b1b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2800413 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten bd276beb 2021-02-12T10:58:43 Vulkan: Don't wait on unflushed sync objects without a Context The app "Car Parking Multiplayer" issues a ClientWaitSync() command without having already flushed the sync object and without an active context. We should return TIMEOUT immediately rather than attempting to wait on the sync object, since we can't flush it and it'll never be signalled. Bug: angleproject:5613 Bug: angleproject:5656 Test: MultithreadingTest.NoFlushNoContextReturnsTimeout Change-Id: Ieaf675ca9144f9c851c73b9ca399daaf4ed1cd0f Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2693375 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com>
Tim Van Patten 535d4783 2021-01-20T14:14:02 Vulkan: Flush if sync object is pending during SyncVk::getStatus() When a glGetSynciv() is performed for GL_SYNC_STATUS, we should flush any pending commands if a sync object is pending a flush, since the caller is interested in the status of a fence. This will guarantee that the work is submitted to the hardware and eventually completes. This is accomplished by moving mSyncObjectPendingFlush from ContextVk to ShareGroupVk, so that any sync objects used by any contexts within the share group are submitted to hardware and the required work completes. Bug: angleproject:5306 Bug: angleproject:5425 Test: FenceSyncTest.BasicOperations Change-Id: I2e2681ad01fda429ba37f061c9bac5eb91f800fd Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2641095 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Tim Van Patten <timvp@google.com>
Tim Van Patten bb766000 2021-01-20T16:58:16 Vulkan: Move Flush during SyncHelper::clientWait() earlier The checks and call to contextVk->flushImpl() needs to be moved before the timeout check to ensure that the vkEvent is flushed to HW eventually when the caller is waiting on it, even if they don't specify a timeout. Bug: chromium:1060139 Test: CopyOutputScalingPixelTest.ScaledCopyOfDrawnFrame/48 Change-Id: I725ea492aaf21e291fe3e5b9ae11ea009e59e019 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2640423 Commit-Queue: Alexis Hétu <sugoi@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Alexis Hétu <sugoi@chromium.org>
Tim Van Patten 909ea88b 2020-11-20T13:07:53 Reland "Vulkan: Ignore glFlush to reduce vkQueueSubmits in Asphalt 9" This is a reland of 5cf7472dd161bbda329dfc5e4e65bb6ce0c06fbd The ShareGroupVk::mResourceUseLists was not being cleared each call to RendererVk::submitFrame(), so it was growing indefinitely. Each vk::ResourceUseList within it was cleared, so it was holding an essentially "infinite" list of empty lists, but that caused the loop in RendererVk::submitFrame() to take more and more time until the tests timed out. The fix is to do 'resourceUseLists.clear()' once the loop to release all resources has completed, like releaseResourceUsesAndUpdateSerials() does for each individual list. Additionally, ASSERTs are added to guarantee that the lists are empty when the ContextVk and ShareGroupVk are destroyed. Original change's description: > Vulkan: Ignore glFlush to reduce vkQueueSubmits in Asphalt 9 > > Multithreaded apps can use the following pattern: > > glDrawElements() > glFenceSync() > glFlush() > glWaitSync() > > This currently results in a vkQueueSubmit for every glFlush() to ensure > that the work has landed in the command queue in the correct order. > However, ANGLE can instead avoid the vkQueueSubmit during the glFlush() > in this situation by instead flushing the ContextVk's commands and > ending the render pass to ensure the commands are submitted in the > correct order to the renderer. This improves performance for Asphalt 9 > by reducing frame times from 150-200msec to 35-55msec. > > Specifically, ANGLE will call flushCommandsAndEndRenderPass() when > there is a sync object pending a flush or if the ContextVk is currently > shared. > > Additionally, on all devices except Qualcomm, ANGLE can ignore all other > glFlush() calls entirely and return immediately. For Qualcomm devices, > ANGLE is still required to perform a full flush (resulting in a > vkQueueSubmit), since ignoring the glFlush() reduces the Manhattan 3.0 > offscreen score by ~3%. > > Bug: angleproject:5306 > Bug: angleproject:5425 > Change-Id: I9d747caf5bf306166be0fec630a78caf41208c27 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2552718 > Commit-Queue: Tim Van Patten <timvp@google.com> > Reviewed-by: Charlie Lao <cclao@google.com> > Reviewed-by: Jamie Madill <jmadill@chromium.org> Bug: angleproject:5306 Bug: angleproject:5425 Bug: angleproject:5470 Change-Id: I14ee424d032f22e5285d67accbec078ad1955dd0 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2595811 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Corentin Wallez a19bd601 2020-12-16T13:04:38 Revert "Vulkan: Ignore glFlush to reduce vkQueueSubmits in Asphalt 9" This reverts commit 5cf7472dd161bbda329dfc5e4e65bb6ce0c06fbd. Reason for revert: causes timeouts, see anglebug.com/5470 Original change's description: > Vulkan: Ignore glFlush to reduce vkQueueSubmits in Asphalt 9 > > Multithreaded apps can use the following pattern: > > glDrawElements() > glFenceSync() > glFlush() > glWaitSync() > > This currently results in a vkQueueSubmit for every glFlush() to ensure > that the work has landed in the command queue in the correct order. > However, ANGLE can instead avoid the vkQueueSubmit during the glFlush() > in this situation by instead flushing the ContextVk's commands and > ending the render pass to ensure the commands are submitted in the > correct order to the renderer. This improves performance for Asphalt 9 > by reducing frame times from 150-200msec to 35-55msec. > > Specifically, ANGLE will call flushCommandsAndEndRenderPass() when > there is a sync object pending a flush or if the ContextVk is currently > shared. > > Additionally, on all devices except Qualcomm, ANGLE can ignore all other > glFlush() calls entirely and return immediately. For Qualcomm devices, > ANGLE is still required to perform a full flush (resulting in a > vkQueueSubmit), since ignoring the glFlush() reduces the Manhattan 3.0 > offscreen score by ~3%. > > Bug: angleproject:5306 > Bug: angleproject:5425 > Change-Id: I9d747caf5bf306166be0fec630a78caf41208c27 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2552718 > Commit-Queue: Tim Van Patten <timvp@google.com> > Reviewed-by: Charlie Lao <cclao@google.com> > Reviewed-by: Jamie Madill <jmadill@chromium.org> TBR=timvp@google.com,jmadill@chromium.org,cclao@google.com Change-Id: I9886bf901a835d408b6a4b8be7ea408fa2121be0 No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: angleproject:5306 Bug: angleproject:5425 Bug: angleproject:5470 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2595032 Reviewed-by: Corentin Wallez <cwallez@chromium.org> Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Tim Van Patten 5cf7472d 2020-11-20T13:07:53 Vulkan: Ignore glFlush to reduce vkQueueSubmits in Asphalt 9 Multithreaded apps can use the following pattern: glDrawElements() glFenceSync() glFlush() glWaitSync() This currently results in a vkQueueSubmit for every glFlush() to ensure that the work has landed in the command queue in the correct order. However, ANGLE can instead avoid the vkQueueSubmit during the glFlush() in this situation by instead flushing the ContextVk's commands and ending the render pass to ensure the commands are submitted in the correct order to the renderer. This improves performance for Asphalt 9 by reducing frame times from 150-200msec to 35-55msec. Specifically, ANGLE will call flushCommandsAndEndRenderPass() when there is a sync object pending a flush or if the ContextVk is currently shared. Additionally, on all devices except Qualcomm, ANGLE can ignore all other glFlush() calls entirely and return immediately. For Qualcomm devices, ANGLE is still required to perform a full flush (resulting in a vkQueueSubmit), since ignoring the glFlush() reduces the Manhattan 3.0 offscreen score by ~3%. Bug: angleproject:5306 Bug: angleproject:5425 Change-Id: I9d747caf5bf306166be0fec630a78caf41208c27 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2552718 Commit-Queue: Tim Van Patten <timvp@google.com> Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tim Van Patten e8c5525c 2020-12-09T09:41:21 Vulkan: flushCommandsAndEndRenderPass during glFenceSync() The description for glFenceSync states: glFenceSync — create a new sync object and insert it into the GL command stream ANGLE needs to break the render pass and flush any prior commands to ensure that the newly created Fence object lands with the correct ordering. This will be done by calling ContextVk::flushCommandsAndEndRenderPass() before creating the Vulkan Event object. Bug: angleproject:5306 Change-Id: I3a460bb559acf8eb4b3041065ec743938d44d823 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2582199 Reviewed-by: Charlie Lao <cclao@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Tim Van Patten <timvp@google.com>
Courtney Goeltzenleuchter dcc0bebf 2020-12-10T13:43:03 Vulkan: clean up comments, rename per request Bug: b/170312581 Change-Id: Iefecb5776cac07477266e6a2e77af7d3c9f38ad6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2585745 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Courtney Goeltzenleuchter 6d94471b 2020-11-19T16:53:09 Vulkan: Eliminate fence wait from SyncVk Waiting on a fence is problematic if using threaded worker to submit work since code needs to know if the work has been submitted or not before doing the wait. Eliminate the wait by using a serial if available or if no serial check the file descriptor directly. If no serial it's because we are waiting on an imported file descriptor. Could call fence.wait in that case but can also wait on the file descriptor. Test: angle_end2end_tests --gtest_filter=EGLSyncTest.*/ES2_Vulkan Bug: b/170312581 Change-Id: I392a5e73ac8f851d0d5bc53f06063568c0c90d2c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2579042 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Courtney Goeltzenleuchter aafcb504 2020-12-08T09:32:10 Vulkan: Add ensureSubmission to queueSubmitOneOff Some callers of queueSubmitOneOff require that the command being queued to have been sent to the GPU. The new ensureSubmission parameter indicates that behavior and when running with threaded worker will wait for the worker queue to empty before returning. Bug: b/170312581 Change-Id: Ib620fb37f4b9b4431451ccbd10807c0dff1842af Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2579041 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Cody Northrop <cnorthrop@google.com> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill 84fce18c 2020-11-13T10:14:58 Vulkan: Give CommandQueue an abstract interface. This gives CommandQueue and CommandProcessor the exact same interface. This also moves the worker thread to be owned by CommandProcessor. Bug: b/172704839 Change-Id: Ife439bcf52d923e01a6a2166e0caaffce14fd086 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2537235 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Shahbaz Youssefi 9e7f08fc 2020-11-06T16:55:51 Vulkan: Fix incorrect reordering of barriers Take the following situation, a simple copy from image to buffer: ANGLE_TRY(contextVk->onBufferTransferWrite(buffer)); ANGLE_TRY(contextVk->onImageTransferRead(aspectFlags, image)); CommandBuffer &commandBuffer = contextVk->getOutsideRenderPassCommandBuffer(); commandBuffer.copyImageToBuffer(imageHandle, layout, bufferHandle, 1, regions); Both `onBufferTransferWrite` and `onImageTransferRead` may flush either the outsideRP or insideRP command buffers. If buffer is not previously used, but image is used: - onBufferTransferWrite: buffer usage is recorded in outsideRP1 - onImageTransferREad: outsiderRP1 is flushed, outsideRP2 is started - copyImageToBuffer: recorded on outsideRP2, but buffer usage not recorded there - A following command that uses the buffer and requires barrier doesn't close outsideRP2 as it believes it was not used there Bug: angleproject:5319 Change-Id: Ib8994083fbc21969a538cda3784adee57b089415 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2523388 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 49c13282 2020-11-08T10:51:54 Vulkan: Merge async command queue features. This merges the 'commandProcessor' and 'asyncCommandProcessing' features into one new 'asyncCommandQueue' features. Conceptually the command processor is asynchronous use of the CommandQueue. This renaming updates the code to reflect the usage. Bug: b/172704839 Change-Id: I1d7ca1187f802eed48823a01b57fb520527f2540 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2525144 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 6a1fb047 2020-11-07T19:04:43 Vulkan: Align CommandQueue and TaskProcessor. The intent is to merge these two classes and remove redundant code. Bug: b/172704839 Change-Id: I6eb66d7b4ddea3633b08df99464975cd59773b4e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2524550 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 65ee5168 2020-11-08T09:30:39 Vulkan: Align submit and serial management code. This progresses the goal of merging TaskProcessor and CommandQueue. Moving the serial management out of RendererVk allows these classes to have finer control over when thread synchronization locks happen. Note: device lost handling seems untested currently. Bug: b/172704839 Change-Id: I0cc61e1ffe41aad0b898d4146c8dbd08a2cebd3c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2525140 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill e6a302a0 2020-11-07T10:49:15 Vulkan: Move vk::CommandQueue to RendererVk. This consolidates most of the command processor vs queue logic in one place. It also has a number of incidental changes related to fences: - syncs now do not store a shared fence. instead they call command apis to wait for a particular serial with a timeout. this is not yet fully implemented in CommandProcessor. - surface swap history stores a serial instead of a fence. because the RendererVk class stores the command batches, we no longer have to do messy things with ContextVk. - it is no longer possible to ask for a wait on a serial that isn't in the command queue. Also adds mutex synchronization around the RendererVk methods. Bug: angleproject:5217 Bug: b/172704839 Change-Id: I5faf0e24bb6ede79a927ab149b80bfa8baca4620 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2524548 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Courtney Goeltzenleuchter 6bc362c4 2020-10-18T19:09:47 Allow single threaded CommandProcessor In replacing the legacy CommandQueue code with the threading capable CommandProcessor it would be good to be able to run the CommandProcessor in a single-thread environment. This CL changes the meaning of the feature flags for the commandProcessor and asynchronousCommandProcessing so that enabling commandProcess only changes the code paths to use the command processor but work it still done as part of the submitting thread (e.g. ContextVk). Enabling asynchronousCommandProcessing will cause a separate worker thread to be spawned which will asynchronously process the commands. This allows us to switch to the CommandProcessor without threading and then enable threading once performance issues are resolved. Bug: b/161912801 Bug: b/170329600 Change-Id: I534862b109a7e7708108190b7c3e894071d4c2ed Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2483580 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
Courtney Goeltzenleuchter 097f307e 2020-10-22T10:26:19 Vulkan: Fix segv referencing contextVk Tests dEQP-EGL.functional.fence_sync.* can cause SyncHelper::clientWait to be called with a null contextVk which we were trying to deference to get to the renderer. We have a pointer to the renderer already so just use that. Bug: b/154030730 Change-Id: If64ec2c960f39b893d2e47a11638d16f830aecfd Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2493181 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Courtney Goeltzenleuchter ed876984 2020-10-03T11:00:36 Vulkan: functionally complete worker thread Working on enhancing worker thread to completely own primary command buffers. This will include not only processing SCBs from main thread into a primary, but also submitting those command buffers to the queue. The CommandProcessor is a vk::Context so it can handle errors in the worker thread. When the main thread submits tasks to the worker thread it also syncs any outstanding errors from the worker. Include asynchronousCommandProcessing feature that will control whether the worker thread task does it's work in parallel or not. If false, we wait for the thread to complete it's work before letting the main thread continue. If true, the thread can execute in parallel with the main thread. Bug: b/154030730 Bug: b/161912801 Change-Id: I00f8f013d6cbb2af12a172c4f7927855db2f0ebf Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2328992 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Courtney Goeltzenleuchter 54dbd5e7 2020-09-23T16:10:12 Vulkan: Add mutex around queueSubmit related data There are several queueSubmit related members of RendererVk that can be accessed from multiple threads. Adding mutex around thoses accesses resolves race condition flagged by TSAN. Add stress test for QueueSerial handling Add test to catch race issue in garbage collection found by TSAN. Test: angle_end2end_tests MultithreadingTest.MultiContextDrawWithSwapBuffers angle_end2end_tests MultithreadingTest.MultiContextCreateAndDeleteResources Bug: b/168744561 Change-Id: I238cce9052476710778a3b35f8531891d90ddd6e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2415402 Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 68a5baeb 2020-09-23T22:13:03 Revert "Vulkan: Implement a SharedResourceUse pool" This reverts commit de335c16855f11d1f0a6f0b37bee30c8a09a6c1d. Reason for revert: Might actually regress CPU overhead perf. Unsure but it's possible the reported perf improvement was due to variance. Original change's description: > Vulkan: Implement a SharedResourceUse pool > > When adding a Resource to the ResourceUseList of ContextVk > we constructed a new SharedResourceUse object for tracking > and update of the Resource's Serial. We would then delete > it after releasing the resource. This incurs repeated > memory operation costs. > > Instead we now allocate a pool of SharedResourceUse objects > and acquire and release from this pool as needed. > > VTune profile of the Manhattan 30 offscreen benchmark > shows the CPU occupancy of bufferRead decrease from an > average of 0.9% -> 0.6% and imageRead decreases from > an average of 0.4% -> 0.3%. The bottleneck for both > these methods is the retain() method that leverages > the new SharedResourceUse pool. > > Bug: angleproject:4950 > Change-Id: Ib4f67c6f101d4b2de118014546e6cc14ad108703 > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2396597 > Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> > Reviewed-by: Jamie Madill <jmadill@chromium.org> > Commit-Queue: Mohan Maiya <m.maiya@samsung.com> TBR=syoussefi@chromium.org,jmadill@chromium.org,m.maiya@samsung.com # Not skipping CQ checks because original CL landed > 1 day ago. Bug: angleproject:4950 Change-Id: I40081551c3db67d6e55182fea40119946ed16ac3 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2426479 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Mohan Maiya de335c16 2020-09-14T12:04:20 Vulkan: Implement a SharedResourceUse pool When adding a Resource to the ResourceUseList of ContextVk we constructed a new SharedResourceUse object for tracking and update of the Resource's Serial. We would then delete it after releasing the resource. This incurs repeated memory operation costs. Instead we now allocate a pool of SharedResourceUse objects and acquire and release from this pool as needed. VTune profile of the Manhattan 30 offscreen benchmark shows the CPU occupancy of bufferRead decrease from an average of 0.9% -> 0.6% and imageRead decreases from an average of 0.4% -> 0.3%. The bottleneck for both these methods is the retain() method that leverages the new SharedResourceUse pool. Bug: angleproject:4950 Change-Id: Ib4f67c6f101d4b2de118014546e6cc14ad108703 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2396597 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Mohan Maiya <m.maiya@samsung.com>
Charlie Lao c03c4490 2020-08-28T11:03:20 Vulkan: Defer glFlush issued in middle of renderpass to endRenderpass Manhattan is calling glFlush in middle of a renderpass. This CL defers the flush that issued in the middle of renderpass to the end of renderpass. Bug: b/166475273 Change-Id: I6baa3898d5efc456e2205c44e13c64f3d79d1464 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2381942 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill eb85c021 2020-08-05T15:07:15 Vulkan: Refactor Context CommandBuffer query. getOutsideRenderPassCommandBuffer returns the command buffer directly since it's now stateless and cannot throw an error. All the RenderPass begin/end and flushing are done by the dependency functions (buffer/ image read/write). Bug: angleproject:4911 Change-Id: I5e7806be9d0e1b5e358524bd485298d660fac942 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2339544 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Jamie Madill 23daea2f 2020-08-06T10:40:00 Vulkan: Rename CommandBuffer get method. This method will no longer end the RenderPass in a following CL. Renaming it in a split CL makes the review diff simple. Bug: angleproject:4911 Change-Id: Id48257884dccb7c86f7de2cc9ca95e651fb68df7 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2340788 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Xiaoxuan Liu de4611c9 2020-06-04T13:44:17 Vulkan: Update eglDupNativeFenceFDANDROID() implementation. When exporting vkfence FD with vkGetFenceFdKHR() in SYNC_FD semantic, vkfence could be reset. Dup and store the SYNC_FD created vkfence's FD in SyncHelperNativeFence by exporting FD during initializeWithFd() meanwhile make the eglDupNativeFenceFDANDROID() only dup the mNativeFenceFd in SyncHelperNativeFence. Test: angle_end2end_tests --gtest_filter=EGLSyncTest.AndroidNativeFence_* Bug: angleproject:2517 Change-Id: I354185d26d0fda72baeb61702c879ed5665db6ec Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2230062 Reviewed-by: Jeff Vigil <j.vigil@samsung.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Charlie Lao 5891b091 2020-06-01T11:24:37 Vulkan: acquire swapchain image should use COLOR_ATTACHMENT_OUTPUT_BIT The mAcquireImageSemaphore is for achieving the maximum overlap between display engine and rendering engine. We should only need to block when we write to the swapchain image. Work that doesn't touch the visible buffer can proceed. Bug: b/157916459 Change-Id: Ic9a9a1f2a7648ef7f50f99578a0f0d674ae5e66c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2223826 Commit-Queue: Charlie Lao <cclao@google.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Tobin Ehlis 21c5af31 2020-04-20T10:29:13 Vulkan:Migrate events and queries to secondary Cmd Buffer Replace flushAndGetPrimaryCommandBuffer() function with endRenderPassAndGetCommandBuffer() for events and queries. The end result should be the same, but this allows a number of places that were putting commands directly into the primary to put the commands into ANGLE's custom SecondaryCommandBuffer (SCB) instead. This also fixes a couple of minor bugs related to command buffer ordering. flushAndBeginRenderPass() now flushes any outside RenderPass (RP) commands first. Also, when insideRP commands are flushed to the primary, set "mHasPrimaryCommands = true;" Bug: b/153666475 Change-Id: I68413f25d27175afed0a20bc49f22f4c8d01e4fb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2156932 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Charlie Lao <cclao@google.com> Commit-Queue: Tobin Ehlis <tobine@google.com>
Jeff Vigil e20f36f4 2020-04-21T15:45:44 EGL: Implement EGL_ANDROID_native_fence_sync on vulkan Check the following - 1. Vulkan ICD supports VK_KHR_external_fence_fd 2. ExternalFenceProperties and ExternalSemaphoreProperties support Android FD. eglCreateSync - if FD was provided import to VkFence, else create VkFence with new FD and then flush and submit VkFence on next vkQueueSubmit. eglGetSyncAttrib - for status call vkGetFenceStatus. eglDupNativeFenceFdANDROID - return FD from vkGetFenceFD call. eglClientWaitSync - call vkWaitForFences. eglWaitSync - dup FD, create VkSemaphore and import FD, then flush() and add VkSemaphore to next vkQueueSubmit as a waiting semaphore. Extended end2end test suite with nativefence test cases. Bug: angleproject:2517 Test: angle_end2end_tests --gtest_filter=EGLSyncTest.AndroidNativeFence_* Change-Id: I8f6a6f4c3d71d83007f662b78377aa015a740035 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2026177 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jeff Vigil 78dcba58 2020-04-21T15:20:13 Refactor SyncHelper with vk::Resource Set SyncVk's SyncHelper to inherit from vk::Resource. Bug: angleproject:2517 Change-Id: I499414d97e611d029abc12622f19436cbdb3e1f5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2160126 Commit-Queue: Mohan Maiya <m.maiya@samsung.com> Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill aa09ca69 2020-02-24T13:33:30 Vulkan: Update ContextVk command buffer access methods. Use the following names: - flushAndBeginRenderPass - endRenderPassAndGetCommandBuffer - flushAndGetPrimaryCommandBuffer This clarifies that a flush is part of the method. Bug: angleproject:4029 Change-Id: I6c870761339ea7aa39c83142200781ba39ad6a4b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2068129 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill c58458e6 2020-02-19T14:51:41 Vulkan: Remove CommandGraph code. Also updates relevant comments to no longer refer to a graph. Bug: angleproject:4029 Change-Id: Ic29716e9ae4926870f902947d49d8fee7af98662 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2057804 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com>
Jamie Madill 3c066404 2020-02-05T12:43:12 Vulkan: Command graph linearization (Step 4). Implements GLES 3.0 functionality for the new command recording style. Tested with angle_end2end_tests. dEQP/CTS testing will come later. Bug: angleproject:4029 Change-Id: Ibdc4f1355a7b880946699ec227fd9807f4a27af9 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2040197 Reviewed-by: Cody Northrop <cnorthrop@google.com> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 9681190d 2020-02-05T12:51:25 Make Debug marker functions return angle::Result. Allows error handling in these functions. Necessary for the Vulkan back-end. Bug: angleproject:4209 Change-Id: I2092e58e719c6ee562807e1c7e8ad26988342855 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2040196 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 441b72f0 2020-01-16T14:48:39 Vulkan: Make ContextVk own ResourceUseList. This moves the resource use tracking functionality out of CommandGraph. Making the list a separate class helps the implementation avoid tricky circular include problems. Bug: angleproject:4029 Change-Id: I3288fc685b21e949f12b0796109a2b7bb117c249 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2002931 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 4ff1651e 2020-01-16T12:42:27 Vulkan: Add ResourceUseList helper. This abstracts away the ResourceUse tracking from CommandGraph. Pulling it out will allow us to more easily swap resource use tracking into ContextVk. The command graph is eventually going away so we need to move functionality out of CommandGraph. Bug: angleproject:4029 Change-Id: Icc3bcc824510b03e91f8ff010a82751a81516bfa Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2002930 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill f10bf6bf 2019-09-26T10:27:18 Vulkan: Implement multi-threaded GL. The main component of this change is to make vk::BufferHelper, vk::ImageHelper and vk::SyncHelper use a common path. We introduce a new "vk::SharedGarbage" helper class that stores small lists of garbage from individual objects like an ImageHelper or BufferHelper. The SharedGarbage is stored in the RendererVk with the ResourceUse of the helper object. The ResourceUse tells RendererVk when it is safe to destroy the GarbageObjects. New "onGraphAccess" commands are added in a few places to enable the common garbage collection path. A couple Context-only resources like default attributes now are referenced where they were not before. Also reorganizes some functions so we can add a few helpful ASSERTs to our graph dependencies. Added "updateCurrentAccessNodes" for this. Also adds a "RendererScoped" helper to replace many uses of "ContextScoped". The multithreading EGL tests mostly pass but have some remaining flakiness so cannot yet be enabled. Bug: angleproject:2464 Change-Id: Ia3e3ae8848d731abf3f21ebe04c33e381e130be0 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1808444 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Jamie Madill 33ffed01 2019-09-26T10:19:35 Vulkan: Clean up garbage APIs. Instead of dumping resources to a context, we use the release APIs consistently. Refactoring/cleanup change only. Should have very litte impact on runtime behaviour. Bug: angleproject:2464 Change-Id: I2dc7f8316c466f7ccfad50a7b792ba0ee7bc2e49 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1804883 Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill b540da89 2019-09-19T14:19:12 Vulkan: Use ResourceUse to track object lifetime. With the new resource tracking scheme the CommandGraph, tracking a "Context serial" aka "current" serial is no longer necessary for CommandGraphResources. Serial tracking has been moved to the shared ResourceUse struct that gets updated on a command submission. Thus we don't need to store the serial as a current separate piece of info in BufferHelper/ImageHelper. Will lead to further redesign for the multi-threading support for Vulkan. Bug: angleproject:2464 Change-Id: I1ae4bcc27fcfb93422b4ab4c9682a458e482f295 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1785990 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi 57ad1e12 2019-08-23T14:53:26 Vulkan: Introduce ContextScoped Scoped is changed to DeviceScoped and a ContextScoped class is added. The former destroys objects at the end of scope, and the latter releases them. Bug: angleproject:3757 Change-Id: Ia0bbded53e94af9c8c72c77d07306d6724c6c409 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1769060 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill 83c24aa8 2019-08-21T12:28:18 Vulkan: Reference Context fences in FenceSyncVk. A particular test waits on a fence that was created in another Context. Then it immediately destroys all other resources. This triggered a bug where the VkEvent corresponding to the FenceSync was deleted while it was still in use by a command buffer. Fix this by referencing the other submit fence of the second ContextVk in the FenceSync class. Bug: angleproject:2464 Change-Id: I4ef096bf760aa8943db08d491006c68e36db993b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1757505 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Geoff Lang 69e46a18 2019-07-03T14:43:32 GL: Implement EGL_ANDROID_native_fence_sync This extension allows Chrome to use ANGLE on newer Android devices. BUG=angleproject:3643 Change-Id: I5456d61749399ca2bbc11cc5e98b9120f8702406 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1687121 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jonah Ryan-Davis <jonahr@google.com> Commit-Queue: Geoff Lang <geofflang@chromium.org>
Michael Spang ab6a59b2 2019-05-21T21:26:26 Vulkan: Submit semaphores from glWaitSemaphoreEXT & glSignalSemaphoreEXT Implement submission of client semaphores passed to glWaitSemaphoreEXT & glSignalSemaphoreEXT. This also relaxes the expectation that we will not flush() if there are no commands. Signaling semaphores in particular requires queue submission irrespective of whether there are any command buffers to submit. If there are neither commands nor semaphores, we can still skip queue submission. WebGL runs in Chrome with ANGLE & Vulkan interop as of this patch, albeit with incorrect synchronization due to texture barriers not being implemented yet. Quite a few flags are needed to try this: GN args: angle_vulkan_conformant_configs_only=true chrome \ --enable-features=UseSkiaRenderer,UiGpuRasterization \ --enable-gpu-rasterization \ --enable-oop-rasterization \ --enable-vulkan \ --use-gl=angle \ --use-angle=vulkan Bug: angleproject:3289 Change-Id: I3d49c230a2fbf0cd2a2b943b05ded0e4604cc313 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1623815 Commit-Queue: Michael Spang <spang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Geoff Lang e0956a16 2019-05-06T10:34:27 Vulkan: Rename getSubmitFence to getNextSubmitFence. Now that getLastSubmittedFence exists, the new name is more specific. BUG=angleproject:2464 Change-Id: I0bce1387b95b7b7b739e311d37e35e7d162d8f2d Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1595435 Commit-Queue: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Geoff Lang ee244c77 2019-05-06T10:30:18 Vulkan: Move command graph and garbage to ContextVk. To support multithreading, contexts should manage their own command graphs and garbage. This allows safe access to vulkan resources such as command pools without thread synchronization. BUG=angleproject:2464 Change-Id: I930149bc9f0793028761ee05ab50b8c0a4dec98a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1516515 Commit-Queue: Geoff Lang <geofflang@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Geoff Lang e755a537 2019-04-10T09:58:21 Vulkan: Add a new garbage type gated by fences. This allows Vulkan EGL objects such as EGL Syncs and EGL Images to give their garbage to the renderer before destroying. BUG=angleproject:2464 Change-Id: I59b8e1080e4292bd0856e59a928750c7e77a372e Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1562522 Commit-Queue: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Geoff Lang 892d1805 2019-03-27T14:21:34 Vulkan: Have the WindowSurfaceVk own the submit semaphores. Chaining of submit semaphores is only needed for window surfaces because they are required for the first usage of the swap chain image and final present of the image. Move ownership of the submit semaphores from RendererVk to WindowSurfaceVk and update all calls to finish and flush to be piped through a ContextVk which tracks the currently bound window surface. BUG=angleproject:2464 Change-Id: I4b3083124d7910a5dee297afc219e3a3f28057f2 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1542257 Commit-Queue: Geoff Lang <geofflang@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Geoff Lang 9049d321 2019-03-27T14:08:44 Vulkan: Pass the current context to egl Sync operations. The EGL_KHR_fence_sync spec says that if a flush is needed before waiting on the sync, it's done on the current context for the current thread. This helps simplify the multithreading design, we don't need to worry about flusing on a context that may no longer exist or is executing on a different thread. It does allow infinite waits because the context with the fence is never flushed but the spec allows this. BUG=angleproject:2464 Change-Id: I8bf2f93c408fee2dae95caa5bb9c76ba67687931 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1542256 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi 20ae6814 2019-02-27T17:11:58 Vulkan: Decouple EGLSync from renderer serial To support future work where RendererVk functionality is moved to ContextVk. Given multiple contexts, EGLSync can no longer rely on a single serial as it can be used in multiple contexts. Instead, the fence corresponding to the submission in which the EGLSync object signals is kept so it can be waited on. Introduces a `vk::Shared` class that includes a ref-counted reference to a Vulkan object (vk::Fence in this case). This is specially made to `destroy()` object when reference count reaches zero. Bug: angleproject:2464 Change-Id: I68c8229eea8df77974e28fcc2a9563dae5d204f9 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1493131 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill (use @chromium please) <jmadill@google.com> Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Jamie Madill c09ae15c 2019-02-01T14:16:32 Enable -Wextra-semi and -Wextra-semi-stmt. This will prevent users from accidentally making semicolon errors in the future. Bug: chromium:926235 Change-Id: I79a6fa376fb1ad8f0fcf1b65b1f572a035d1f4e9 Reviewed-on: https://chromium-review.googlesource.com/c/1446493 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Nico Weber <thakis@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Shahbaz Youssefi 82fddcb1 2019-01-18T14:27:43 Vulkan: Implement GLsync and EGLSync fence syncs That is required in GLES 3 for GLsync and EGL_KHR_fence_sync and EGL_KHR_wait_sync (or EGL 1.5) for EGLSync. The two constructs (GLsync and EGLSync) have similar semantics and share the implementation on the Vulkan backend. The implementation of a fence sync object is achieved through the combined use of a vkEvent and the implicit vkFence inserted at the end of every submission. Imagine the following command buffer: glDraw : Draw glCreateSync: Set Event <-- insertion of fence sync glDraw : Draw : Signal Fence <-- implicit fence at the end of submission glFlush : Submit Assume the serial S is associated to this submission. The following hold: - If event is set, the fence sync is signaled - If S is already finished, the fence sync is signaled - If client is waiting on the sync and S is not yet flushed, there will be a deadlock (unless multi-threaded and another thread performs the flush). The event is used to implement server waits (glWaitSync), as vkEvent is the only entity the GPU can signal and wait on within the command buffer. The wait is inserted in the command graph without incurring a flush, i.e. the wait can be within the same command buffer as event set. The event however does not support CPU waits (glClientWaitSync). vkFence is the only entity the CPU can wait on. For client wait therefore, the following algorithm is used: - If the event is already set, there's no wait -> already signaled - If timeout is zero, there's no wait -> timeout expired - If S is not flushed, flush it to ensure forward progress. - Wait until S is finished -> condition satisfied / timeout expired. Bug: angleproject:2466 Change-Id: I678995a6139dd9533fa8ad361a3d292b202c52a4 Reviewed-on: https://chromium-review.googlesource.com/c/1422552 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill 7c985f5c 2018-11-29T18:16:17 Make angle::Result an enum. This moves away from a class type to a value type. This should improve performance when using angle::Result as a return value. Previously the generated code would return a pointer instead of a value. Improves performance in the most targeted microbenchmark by 10%. In more realistic scanarios it will have a smaller improvement. Also simplifies the class implementation and usage. Includes some unrelated code generation changes. Bug: angleproject:2491 Change-Id: Ifcf86870bf1c00a2f73c39ea6e4f05ca705050aa Reviewed-on: https://chromium-review.googlesource.com/c/1356139 Commit-Queue: Jamie Madill <jmadill@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org>
Jamie Madill b980c563 2018-11-27T11:34:27 Reformat all cpp and h files. This applies git cl format --full to all ANGLE sources. Bug: angleproject:2986 Change-Id: Ib504e618c1589332a37e97696cdc3515d739308f Reviewed-on: https://chromium-review.googlesource.com/c/1351367 Reviewed-by: Jamie Madill <jmadill@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Jamie Madill b8eec4a4 2018-10-18T17:34:38 Use angle::Result in front-end (Part 7) Refactors the gl::FenceNV and gl::Sync classes. Bug: angleproject:2491 Change-Id: I0fe73d1ccf5407f460e173a3061735b330a88511 Reviewed-on: https://chromium-review.googlesource.com/c/1289712 Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill a0691b77 2018-07-25T10:41:22 Pass Context to Fence Impl methods. This is needed for the error refactoring and also for the Vulkan implementation. Bug: angleproject:2738 Change-Id: I4e1bed7f67ef17feb5554b5838a2ed5feb22bba0 Reviewed-on: https://chromium-review.googlesource.com/1150091 Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
Jamie Madill 70b5bb00 2017-08-28T13:32:37 Rename gl::FenceSync to gl::Sync. The spec refers to Sync objects, FenceSyncs being a subtype. The motivation for this fix is to clear up the FenceSync_ entry point for auto-generation. BUG=angleproject:1309 Change-Id: I94c440476d701628575e7a3eea68b6dd110f41c3 Reviewed-on: https://chromium-review.googlesource.com/636516 Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Jamie Madill <jmadill@chromium.org>
kmx.io     mail     discord kmxgit v0.4.0