Age | Commit message (Collapse) | Author |
|
Change-Id: I10e1b2924933073c116e7745a7b43d6cc5e5c9c9
|
|
|
|
Test: build
Change-Id: Idabb6dfecc81175ded771b910154f079ba40d341
|
|
|
|
|
|
|
|
This reverts commit 96f09a26cd8d68e434196db5d79d848a69ffa7a5.
Reason for revert: Bug: 227565325
Change-Id: I9846495ade9edc94546996558101fff509ee0901
|
|
We could be losing device lost details by using a direct
abort here versus VK_CHECK.
Bug: NONE
Test: Compile
Change-Id: I957c091b5008aba19f16616fe0c31370dc19a053
|
|
This adds a feature flag to AEMU/FEMU to allow hosts to
enable / disable async render control commands (e.g.
rcComposeAsync) as we saw some performance issue on FEMU
when using synchronous render control commands to compose
frames.
Bug: fxbug.dev/93349
Original-Change-Id: I5bc2046e417b2c82a6fc10b97baf7b4ddf4547f1
Change-Id: I5cab5b4770f18dcfae1532ad3efe24eebe1bd2e4
|
|
|
|
Bug: NONE
Test: Run script
Change-Id: I4183f124102f73e8a8f85933215c3dfdb8df0281
|
|
|
|
|
|
bug: 222700096
Change-Id: Ibcee535ee5084d28babb8502bbdbc9059a096f53
|
|
This uses the the new context_init variable.
BUG=202552093
TEST=run Cuttlefish
Change-Id: I72f756e52b87c6b138853bba2a3edd98ecb95170
|
|
Not used.
BUG=202552093
TEST=run Cuttlefish
Change-Id: I0eab3ae79fdc8dbce3642682495b6e153e585a56
|
|
When the host composer composites multiple frames simultaneously and
uses the same composition target buffer, there could be race condition
and leads to undefined behavior possibly including crash.
https://r.android.com/2045748 should have fixed that issue on the guest.
Adding logging on the host to make sure the issue doesn't happen at all.
Test: run the emulator
Change-Id: I26d71d1250880fc804352e1fc1a46fe1895afff6
|
|
|
|
Test: run with Vulkan feature disabled
Bug: b/226572716
Change-Id: I0538989ae9c5a354b328bc928175185b63cafc38
|
|
|
|
This fixes issues in some CI build environments where the tests might not be able to run right after being built
Change-Id: I629f02179e39c48c558b36d9553f6c0719abe4dc
|
|
These strings are printed in gfxstream_backend_init for diagnostic
purposes. Prior we were capturing GL-related info even when running
the Vulkan path.
With this change, I've gated the GL code behind a branch and populated
the correct Vulkan information in a Vulkan-specific branch.
Before we'd print:
```
gfxstream_backend_init:539 GL strings; [Google Inc. (NULL)] [ANGLE (NULL, NULL, )] [OpenGL ES 3.2.0 (ANGLE 2.1.16170 git hash: 40a8f2e6d3af)].
```
Now we print:
```
gfxstream_backend_init:537 GL strings; [NVIDIA(Vendor 0x10de)] [Quadro RTX 4000] [Vulkan 1.2.162 NVIDIA (Vendor 0x10de) 461.40 (Driver Version 0x734a0000)].
```
I haven't reworded the "GL Strings" portion of the text in case that
breaks anyones' workflows. If any code reviewers find that preferable to
rename, let me know. Likewise, I haven't renamed
android_getOpenglesHardwareStrings which we are using to fetch this.
Bug: Internal b/224997786
Test: See before/after GL strings print.
Change-Id: I849a922d4039049b928e48b6119154089b42ccfa
|
|
|
|
|
|
... which stores all the data on the stack, so that in a crash dump,
VkSwapchainCreateInfoKHR will be guaranteed to be captured.
Bug: b/222118078
Test: run the emulator
Change-Id: Ib63f1abda5c57894bc3edc238f0d3f870b1eb47b
|
|
|
|
Test: run the emulator
Bug: b/222117858
Change-Id: I2300742e95b9c010913d8aab9d9c555d0df46317
|
|
This is a script to process gfxstream crash dumps and print the last graphics API calls received from the guest.
Test: n/a
Change-Id: Ia449c2186f8465c2ac2928154b1eeb8582fbadb3
|
|
This will be used by downstream gfxsatream users to skip some tests from running on CI
Test: ran build script
Change-Id: Ic1da644ff0240f1b34d99baaa3879ff91b2f12f0
|
|
Change-Id: I75b245b53a0eb94d33aee643d35eb3083cca973a
|
|
Change-Id: Iacaf501b954e078de72b2eb4d0e866ab1ad4a13e
|
|
The virtio-gpu context ID is an unique identifier allocated by the
guest kernel, via idr_alloc(..). These IDs can be reused, but
virtio-gpu gurantees that re-use will only happens after a
particular context has been destroyed.
Many operations related to resource sharing between guest
processes, and between gfxstream and virtgpu are done in reference
to this context id.
gfxstream has a clear analogue to this already --> m_puid, which
is set via connecting to the "pipe:GLProcessPipe" service in the
guest.
The immediate goal is is make m_puid == ctx_id, so the guest
kernel can reference each host render thread via the ctx_id.
BUG=202552093
TEST=run Cuttlefish
Change-Id: Iadc0c288be6a869261658db87c5c569389f73f5b
|
|
- Useful for next patch.
BUG=202552093
TEST=run Cuttlefish
Change-Id: Ia6aca323ffa39e8739ac49325cca71981ab54caa
|
|
- Name change to better reflect meaning.
BUG=202552093
TEST=run Cuttlefish
Change-Id: Id4481a57f6b5e75cbec527fd890e76dfb4cb76cf
|
|
* Name change to better reflect meaning.
BUG=202552093
TEST=run Cuttlefish
Change-Id: Ib9de6769e4e74f060ae073a87ab45d8acb3cb1b0
|
|
Android crosvm and crosvm main have stopped using it.
Prefer stream_renderer_create_blob instead.
BUG=202552093
TEST=compile
Change-Id: I6f926bb29aaa6b1f6456712f3a3f81e06c638131
|
|
|
|
|
|
This patch allows the RenderDoc to capture for guest Vulkan instances if
the environment variable ANDROID_EMU_RENDERDOC is set.
Test: run Asphalt 9 on the emulator and enable a capture
Change-Id: I62a7dcc6c235b75cd09e26c550e356016cbf9b69
|
|
Set the ANDROID_EMU_RENDERDOC environment varaible will load the renderdoc
DLL in FrameBuffer::initialize().
Test: run the emulator with and without ANDROID_EMU_RENDERDOC set, and use renderdoc to capture some Asphalt 9 frames
Change-Id: I1e9b4abf1477f1ecdd8f3deea74c311478293578
|
|
Change-Id: If40a4e5e7fb3b5cc6e905eb74e29cc7d0e2dbe60
|
|
https: //cmake.org/cmake/help/git-stage/module/GoogleTest.html
Test: compile / run
Change-Id: I09986f31addf33b8668470cb7694fcb909379680
|
|
|
|
|
|
Change-Id: Ifa27653a061a62fab4283aa9edb7c105c3d87ded
|
|
Test: build gfxstream_backend and see pdb file generated
Change-Id: I05e48ec6ebb493cc24ed55fdea4568af890f5df8
|
|
* changes:
Vulkan: Redesign how QueueSignalReleaseImage synchronization works
SyncThread: directly send the task as part of the command
Refactor ThreadPool to eliminate broadcastIndexed and enqueueIndexed
|
|
Test: build on Linux with CC=clang CXX=clang++ cmake . ../ -G Ninja -D VIRGL_RENDERER_UNSTABLE_APIS=ON
Change-Id: Iaf0c32c6fd570e3efbe4bf9d9bd8998900922a49
|
|
|
|
|