|
607d398e
|
2022-03-14T16:32:21
|
|
Vulkan: Optimize resolve of multisample swapchains
* Resolves the multisampled image if the last render pass
draws into the default framebuffer.
* Added test to check the number of resolves in the optimization
subpass (credit: Xinyi He)
* Added test to check the number of resolves outside the subpass.
* Added disabled test to see if the subpass resolve works.
Bug: angleproject:6762
Change-Id: I86a8db3387851ab97d5f7a3d8a0ff26961254c14
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3523062
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Amirali Abdolrashidi <abdolrashidi@google.com>
|
|
fe28a429
|
2022-03-30T15:34:49
|
|
Vulkan: Create buffer for vertex array if robust enabled
If robust access is enabled (i.e., chrome), we want to ensure vulkan
driver never access beyond that OpenGL buffer boundary. But with
suballocation from BufferPool, we are using the same VkBuffer for all
suballocations from the same BufferBlock. this combined with the fact
that there is no size information in the vkCmdBindVertexBuffers, it
means vulkan driver can not properly ensure vertex access not go beyond
the subrange. It can only guarantee not access beyond the entire
VkBuffer size. This CL creates a dedicated vkBuffer object and bind it
to the suballocation of the vkDeviceMemory so that vulkan driver will
see the exact range of the subrange instead of entire buffer. Since we
may allocated more memory than actual requested size and the extra
paddings are not zero filled , user size is used to create this
vkBuffer. This is only enabled when robust access is enabled.
This CL also ported webgl conformance test
out-of-bounds-index-buffers.html and out-of-bounds-array-buffers.html
to end2end test.
Bug: chromium:1310038
Change-Id: I3499ae600028149b1039082e5011232b3e4e5e80
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3553940
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
078da78f
|
2022-03-17T13:06:58
|
|
Vulkan: Support GL_EXT_EGL_image_storage extension
This extension provides a mechanism for creating texture objects that
are both EGLImage targets and immutable and removes the possibility of
implicit orphaning. EGL images created from external sources now
support types other than 2D. Tests covering the new feature were added
to ImageTest.cpp.
Bug: angleproject:6346
Change-Id: Id3e328f352deb1af47062be232384229a8b1c341
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3530489
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: mohan maiya <m.maiya@samsung.com>
|
|
2d54b68e
|
2022-03-22T17:32:06
|
|
Vulkan: Use storeOp=None more opportunistically
Previously, storeOp=None was used when the attachment was in "read-only
mode" and storeOp=Store. With this change, storeOp=None is used more
opportunistically when it's deemed that the attachment was not written
to, regardless of if it was put in "read-only mode" (a construct added
to support read-only depth/stencil feedback loops).
Bug: angleproject:5048
Change-Id: I10832d4e2b97793ea1347a47175cbf8ce9af57d6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3556368
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
ce964d66
|
2022-03-26T18:56:50
|
|
Vulkan: Add tessellation to GetPipelineStage helper
The tessellation stages were missing from a helper, which was silently
returning an invalid value.
Add a test and an assert, which fires before the fix.
Test: GFXBench Car Chase
Test: GLSLTest_ES31.TessellationTextureBufferAccess
Bug: angleproject:7135
Bug: b/218314686
Change-Id: I2bc8d374300fc1470e52affabab7491698c99cee
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3554575
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Cody Northrop <cnorthrop@google.com>
|
|
a02fa8a0
|
2022-03-24T22:14:44
|
|
Vulkan: mark image contents defined on clear-invalidate-clear
The backend was ignoring "repeated clears" with an Invalidate between
them, which marked the image contents as undefined. When a clear happens
after invalidate, verify that the clear parameters were the same, and
simply mark the image contents defined if they aren't already.
For example, in this scenario:
- glBindFramebuffer(fbo)
- glInvalidateFramebuffer(color + depth)
- glClear(color + depth)
- depth only render
- glInvalidateFramebuffer(depth)
- glBindFramebuffer(0)
The color clear got skipped entirely because it was cleared with that
color in a previous frame and no other color draws happened since. This
caused sampling from the FBO's texture to return garbage data.
Bug: angleproject:7127
Change-Id: I4ffe65c67375931ab63f07f27fa59ed0a4b90cd9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3551297
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
041c4c6d
|
2022-03-25T16:30:03
|
|
Vulkan: Track color attachment usage like D/S in render pass
That is in preparation for optimizing mid-render-pass clears, which
requires an answer to the following query: "has this color image been
read from / written to so far in the render pass?"
With this change, a future CL will also be able to optimize color
attachment invalidates, which currently break the render pass
unconditionally, the same way depth/stencil is optimized.
Bug: angleproject:5048
Change-Id: I3d3ee40d8444e6861c06340d5d52b17f5ee895b4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3542989
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Lingfeng Yang <lfy@google.com>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
3f331fae
|
2022-03-22T15:23:38
|
|
Vulkan: Dirty bits for depth/stencil access and feedback loop
In preparation for doing the same for color, the depth/stencil render
pass access and feedback loop modes are now updated with ContextVk dirty
bits.
This change also fixes clear after read-only depth/stencil feedback
loop. The render pass wasn't broken in that case.
Bug: angleproject:5048
Change-Id: I40f9b49593f9e6f35f42408e41c9d6267edb375e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3542988
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
20e7bbb7
|
2022-03-18T17:03:18
|
|
Vulkan: Fix invalidate of attachments with emulated format
Some image formats may be emulated such that the emulated format has
more channels than the original. ANGLE clears the image once so that
these extra channels contain valid values, and carefully ensures they
are never modified.
For swapchain images with such formats, as they are automatically
invalidated at the end of the frame, a workaroud was added to make sure
they are re-cleared in the beginning of the next frame. This however
doesn't fix the issue of glInvalidateFramebuffer resulting in the
contents of attachments with such formats to be discarded (even if the
following render pass clears it, the contents are invalid in between).
This change instead makes sure invalidate of images with emulated
formats that have extra channels are handled appropriately:
- On IMR hardware, the invalidate is dropped altogether as it provides
little to no benefit.
- On TBR hardware, a clear is automatically staged on the invalidated
image.
The latter replaces the workaround that was added to make the following
render pass use loadOp=CLEAR, by adding a clear that's respected
regardless of what the future usage is.
This change also paves the way for a future change where the invalidate
of color attachments is tracked in render passes similarly to how
depth/stencil currently is. With this change, the image is no longer in
an inconsistent state where its contents are considered invalid, even
though some channels are meant to remain valid.
Bug: angleproject:6860
Change-Id: Iec5b4854dfbe3a0bf93cd5aa82c19fe116065744
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3536389
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
730c1271
|
2022-01-07T13:41:08
|
|
Vulkan: Submit queue more often for texture data
Outside command buffers should be flushed more often in order
to prevent the texture data accumulation just before the first
render pass when they are referenced.
* Added a tracker next to copyBufferToImage() for texture size
(in ContextVk). When its value passes kMaxBufferToImageCopySize,
the outside command buffer operations should be submitted and
the tracker would be reset. Currently, the threshold value is
set to 1 << 28 = 256M.
* Added a variation of submitFrame() to be used in outside
command buffer submission. The main difference is that it
copies mResourceUseList into GetShareGroupVk() rather than
move it.
* Refactored the two functions into submitFrameImpl().
* Added a helper function to submit the outside command
buffer.
* Added explicit copy functions for ResourceUseList and
SharedResourceUse. The counter in the copied object is
incremented by 1.
* Added a test to make sure submitting the outside command
buffer does not break the render pass.
Bug: angleproject:6354
Change-Id: Ia1d4f857fcbd06934609c94622ccbf675b3b1c72
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3379231
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Amirali Abdolrashidi <abdolrashidi@google.com>
|
|
|
b74d6768
|
2022-03-17T16:48:21
|
|
Vulkan: Refactor D/S usage tracking in render pass
This is in preparation for doing the same for color images. That is in
turn in preparation for optimizing mid-render-pass clears, which require
an answer to the following query: "has this color image been read from /
written to so far in the render pass?"
Bug: angleproject:5048
Change-Id: I8abe97919897b8fbf644482071bf7326ca3325d6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3533143
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
e8ee847d
|
2022-02-23T12:25:39
|
|
Vulkan: Add UpdateDescriptorSetsBuilder.
This helper class encapsulates the vkUpdateDescriptorSets caching.
As part of the refactor, we switch passing a ContextVk to passing
a vk::Context with some mutable variables. This helps encapsulate
ContextVk. Since we use the perf counters in many places, this CL
moves the perf counters to vk::Context, so we can access them
everywhere.
Refactoring change only.
Bug: angleproject:6776
Change-Id: Id529962b2f425bece6f9b3bd0cd1698c692e58cb
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3484980
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
fe4fa1cb
|
2022-03-08T11:10:16
|
|
Vulkan: Tune pruneEmptyBuffers() to be a bit more aggressive
Right now we check every second and if we find a buffer is empty and
remains empty for 4 checks, we free the buffer. This means we may keep
peak memory usage for 4 seconds. This CL reduces the check to 1/4
seconds, so that a buffer will gets freed if not used for 1 second. Also
added a threshold to keep maximum count of empty buffers to 16. This CL
also optimizes pruneEmptyBuffers() function to avoid erase calls on each
empty buffer block. We set the pointer to null as we free the empty
buffer and a new loop is added at the end to remove all null elements at
once.
Bug: b/223428306
Change-Id: Iff93d6eb404ca22399b26c7adb7efe5c4b87270d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3511311
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
ea70300b
|
2022-03-01T14:55:00
|
|
Fix base level changes not updating FBO completeness check.
Bug: chromium:1299264
Change-Id: I0881a4916c3eeb9ee023d28d207795899417d530
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3498282
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Auto-Submit: Jamie Madill <jmadill@chromium.org>
|
|
|
36ce2646
|
2022-03-04T11:37:08
|
|
Vulkan: Align the size before calling into VMA's allocator
The size we saved in BufferSuballocation object may used for
vkFlushMappedMemory() call. This will ensure that we save the aligned
size to avoid VVL error for unaligned size in that flush call.
bug: b/219974369
Change-Id: Iccf3199f43cd346dab80083d5494fbaeee39948f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3504000
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
1cef917c
|
2022-03-02T23:50:11
|
|
Reland "Remove ImageViewHelper from ResourceUseList"
This is a reland of 126e967de3978f0dc4f6cf0fa3f9749366ea39ba
Original change's description:
> Remove ImageViewHelper from ResourceUseList
>
> Instead of keeping a dedicated SharedResourceUse mUse for
> ImageViewHelper, let ImageViewHelper take ImageHelper.mUse
> for lifetime tracking. ImageViewHelper no longer needs to
> add its' own mUse in the ResourceUseList through retain()
> calls, and this should cut the ResourceUseList size by
> almost half. For instance, in the trace aztec_ruins,
> the maximum ResourceUseList size before the change is 3643,
> and the maximum ResourceUseList size after the change is 1694.
> Since ImageViewHelper no longer needs to keep SharedResourceUse
> mUse as a class memberWe can remove the inheritance from
> Resource class, and make ImageViewHelper simply a NonCopyable class.
> Bug: angleproject:6717
> Change-Id: I460e83f5f3c1d6ef9722b9f3c9a5ba9552563cb9
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3449450
> Reviewed-by: Charlie Lao <cclao@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Commit-Queue: Yuxin Hu <yuxinhu@google.com>
Bug: angleproject:6717
Change-Id: I3aa9785d7dcdc8db82847f1586f8cd7d5c838d7c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3501194
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
|
|
|
8390196a
|
2022-02-26T02:36:22
|
|
Revert "Remove ImageViewHelper from ResourceUseList"
This reverts commit 126e967de3978f0dc4f6cf0fa3f9749366ea39ba.
Reason for revert: Hits an ASSERT about too much garbage being accumulated. anglebug.com/7063
Original change's description:
> Remove ImageViewHelper from ResourceUseList
>
> Instead of keeping a dedicated SharedResourceUse mUse for
> ImageViewHelper, let ImageViewHelper take ImageHelper.mUse
> for lifetime tracking. ImageViewHelper no longer needs to
> add its' own mUse in the ResourceUseList through retain()
> calls, and this should cut the ResourceUseList size by
> almost half. For instance, in the trace aztec_ruins,
> the maximum ResourceUseList size before the change is 3643,
> and the maximum ResourceUseList size after the change is 1694.
> Since ImageViewHelper no longer needs to keep SharedResourceUse
> mUse as a class memberWe can remove the inheritance from
> Resource class, and make ImageViewHelper simply a NonCopyable class.
>
> Bug: angleproject:6717
> Change-Id: I460e83f5f3c1d6ef9722b9f3c9a5ba9552563cb9
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3449450
> Reviewed-by: Charlie Lao <cclao@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Commit-Queue: Yuxin Hu <yuxinhu@google.com>
Bug: angleproject:6717
Change-Id: Iea3d40458e2cc5be6ab0257ba6df8b82a4eeecda
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3491345
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
126e967d
|
2022-02-09T01:46:25
|
|
Remove ImageViewHelper from ResourceUseList
Instead of keeping a dedicated SharedResourceUse mUse for
ImageViewHelper, let ImageViewHelper take ImageHelper.mUse
for lifetime tracking. ImageViewHelper no longer needs to
add its' own mUse in the ResourceUseList through retain()
calls, and this should cut the ResourceUseList size by
almost half. For instance, in the trace aztec_ruins,
the maximum ResourceUseList size before the change is 3643,
and the maximum ResourceUseList size after the change is 1694.
Since ImageViewHelper no longer needs to keep SharedResourceUse
mUse as a class memberWe can remove the inheritance from
Resource class, and make ImageViewHelper simply a NonCopyable class.
Bug: angleproject:6717
Change-Id: I460e83f5f3c1d6ef9722b9f3c9a5ba9552563cb9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3449450
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Yuxin Hu <yuxinhu@google.com>
|
|
|
f996d4d8
|
2022-02-16T09:27:04
|
|
Vulkan: Fix DynamicBuffer cache reuse.
There is a regression bug that the DynamicBuffer's BufferHelper's size
is no longer the size of entire buffer block's size, but the size of
suballocation. This caused buffer reuse logic to screw up since it
relies on the size to make the decision to keep or free the
mFreeBufferList. This CL fixes this bug by using the buffer block's
size.
This CL also removes DynamicBufferPolicy since it is not used any more
since most of DynamicBuffer usages that triggered that
DynamicBufferPolicy implementation have been switched to BufferPool. The
remaining DynamicBuffer usage is only for attributes and uniforms which
are generally small and per context.
Bug: angleproject:6980
Change-Id: I9e013cbd67dd74f5e7fc5bb7d9c9696bd1f69965
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3469714
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
1315bc81
|
2022-02-09T11:25:58
|
|
Vulkan: Refactor DynamicBuffer::allocate.
Reduces duplicate code by re-using allocateFromCurrentBuffer.
Bug: angleproject:6980
Change-Id: I3f1126a53136be8f68a3a79008dd9adae45c86b8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3449279
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
8d966f7d
|
2022-02-08T11:05:09
|
|
Vulkan: Simplify SubAllocation data types.
We no longer need to bootstrap on the wrapper classes because we
use a separate garbage list. This simplifies the code for the
allocation tracking considerably. Also we remove a few mutable
accessor to fortify our OOP design.
(Credit to Jamie since it is splited from his CL)
Bug: angleproject:6980
Change-Id: I2025ec4d141531273e824bb586957a38efc30be1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3469715
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
b1075951
|
2022-02-09T17:35:56
|
|
vk_helpers: Adjust CopyImageToBuffer extent to block size.
When using readPixels with compressed formats, the VkBufferImageCopy
bufferRowLength and bufferImageHeight need to be a multiple of the block
size.
This prevents the following Vk validation errors:
VUID-vkCmdCopyImageToBuffer-bufferRowLength-00203
VUID-vkCmdCopyImageToBuffer-bufferImageHeight-00204
Add tests with texture sizes that are not multiple of block size.
Test: angle_end2end_tests --gtest_filter="GetImageTest.CompressedTexImageNotBlockMultiple*"
Test: angle_end2end_tests --gtest_filter="GetImageTest.CompressedTexImageCubeMap*"
Test: angle_end2end_tests --gtest_filter="GetImageTestES32.CompressedTexImageCubeMapArray*"
Bug: angleproject:6177
Change-Id: I34c788efece285753857500801e88abeaa6a8a78
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3452111
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
bdedf146
|
2022-01-14T13:35:30
|
|
TextureVk: Implement getCompressedTexImage.
Extend ImageHelper::readPixels to support compressed formats.
Implement readPixelsForCompressedGetImage.
Add compression tests to GetImageTest.
Implement CaptureGetCompressedTexImageANGLE_pixels in
capture_gles_ext_params.
Test: angle_end2end_tests --gtest_filter="GetImageTest*.CompressedTexImage*"
Bug: angleproject:6177
Change-Id: I6e5b6d746b9c6cc9b885fa44f75a985660d8fb58
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3452110
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Lubosz Sarnecki <lubosz.sarnecki@collabora.com>
|
|
|
527ceb73
|
2022-02-07T18:25:02
|
|
Vulkan: Switch XFB counter buffer to suballocation
Bug: b/205337962
Change-Id: I2e26fa3ab150b858f07665459fa108440af988d5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3402333
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Ian Elliott <ianelliott@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
f7f26956
|
2022-02-10T13:33:32
|
|
Vulkan: Fix two bugs exposed by moving mEmptyBuffer to suballocation
There are two bugs exposed when I try to switch mEmptyBuffer to use
suballocation. 1) BufferHelper::initializeNonZeroMemory() for device
local memory uses GPU to do zero fill, but it is missing an
vkEndCommandBuffer call. 2) glTexSubImage2D call try to pick
glCopyuBufferToImage code path and we are not checking against pixel
format. Vulkan spec requires offset must be number of pixel size and we
are using "4" for pixel size which does not work for RGB format.
Bug: b/205337962
Change-Id: Ibca6a60ebd6afc5e2bca07b9cc6a43ebdd240f15
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3449348
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Ian Elliott <ianelliott@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
419bca3f
|
2022-01-19T18:22:56
|
|
Vulkan: Use Vulkan API directly for BufferPool's buffer allocation
There are two motivations in this CL. 1) There are two layers of
suballocator right now. BufferPool provides first suballocation. It
tries to allocate from one of the buffers in the pool. If that failed,
it try to create a new BufferBlock (i.e, a VkBuffer). Right now that
calls into VMA which creates another pool to allocate a buffer. We
really only need one layer of suballocation. And 2) Because we uses VMA
to do actual VkBuffer allocation, we have to use Allocator object. But
VMA can not handle external buffers, so we end up having a BufferMemory
class just to handle two different cases. This CL attempts to clean up
this by let ANGLE calling into vulkan driver directly for the actual
buffer allocation, just like we did for VkImages. By doing so, we able
to remove BufferHelper::mMemory data member as well as BufferMemory
class all together. External memory is now treated exactly the same at
BufferHelper.
Bug: b/205337962
Change-Id: I7c183ab0fd7d9aceb6cf416b0214c300798bc010
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3402740
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
734365a7
|
2022-01-27T06:05:56
|
|
Vulkan: improve command buffer reset performance
When trying to profile why a game's framerate was decreasing over time,
I found that memset() was taking the most self time out of any function
in the process.
The top two callers of memset() were:
RenderPassCommandBufferHelper::reset
OutsideRenderPassCommandBufferHelper::flushToPrimary
Profiling specifically called out mUsedBuffers.clear() and
mRenderPassUsedImages.clear() as being the worst offending lines.
The problem is that FastIntegerSet (and by proxy, FastIntegerMap, where
FastIntegerSet manages the keys) perform best when all keys are close to
zero. With large key values (in this case, large image and buffer
serials), significantly more memory gets allocated, causing
mUsedBuffers.clear() and mRenderPassUsedImages.clear() during command
buffer reset to take more and more time the longer the process runs.
This change is essentially a partial revert of f9a062c9754.
Bug: angleproject:6954
Change-Id: Id2542f6425f0845fe81d393ecf6b614b474c53c8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3420925
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: mohan maiya <m.maiya@samsung.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
6207fe88
|
2022-02-03T14:11:30
|
|
Vulkan: Minor cleanup to DynamicDescriptorPool.
In preparation for future changes.
Bug: angleproject:6776
Change-Id: I0a2ab004883ad482f4ed19075c5f5f4e9c451ae0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3437413
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
f65e7c3c
|
2022-02-02T14:23:31
|
|
Vulkan: Take base level into account in image views
Similarly to how max level changes are handled, this change also handles
base level changes. In either case, the old views are retained and new
views are created for the [base, max] level range.
Test provided by m.maiya@samsung.com
Bug: angleproject:6967
Test: MipmapTestES31.UpdateBaseLevel/*
Change-Id: Iee8b2f035c1cc0fa916b987a91939b8ae65d27c5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3441074
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: mohan maiya <m.maiya@samsung.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@google.com>
|
|
|
1608a956
|
2022-02-02T13:54:53
|
|
Vulkan: Revert client vertex data streaming to use DynamicBuffer
In early CL crrev.com/c/3352489, I switched client vertex data streaming
from using DynamicBuffer to sub-allocating from the buffer pool. That
caused CPU overhead regression due to extra cost of handling the
suballocation object creating and garbage collection etc. Even after all
other optimizations I did since then that significantly improved garbage
collection performance, there is still 6% CPU time regression as
measured with gardenscape. This CL moves StreamVertexData() back to use
DynamicBuffer. In order to do that, I have cleaned up DynamicBuffer
interface to be consistent with suballocation interface by storing the
current allocated offset/size in the suballocation object. With that,
the BufferHelper object that returned from DynamicBuffer will be able to
pass around and referenced exactly like it comes from suballocation code
path, and you can retrieve offset/size from that BufferHelper object
instead of having to pass offset around between various function calls.
Given that streaming vertex data from client memory is only possible for
default vertex array and there is only one default vertex array for each
context, this stream vertex data dynamic buffer is essentially a per
context object. So the other change I made here is that I have merged
mDynamicVertexData with default attribute (which uses per context
dynamic buffers) code to use the same sets of dynamic buffers, since you
will only use one or the other but not both.
Bug: b/205337962
Change-Id: I0ceca5b854069f00afdb9544ee86953b9b773821
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3434645
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
bc3be5a8
|
2022-01-27T12:12:58
|
|
Vulkan: Add a dedicated suballocation garbage list
Suballocations are the most common garbage objects in most usage cases.
The current garbage collection code will construct a garbage object from
suballocation and then construct a SharedGarbage object with a
std::vector that holds only one element. And then it adds this
SharedGarbage to the garbage list. This CL tries to avoid create
std::vector with just one element and avoid the cost of switch statement
for each garbage object by adding a new dedicated garbage list that only
holds the suballocation garbages, which is the most common garbages in
the system. With gardenscapes running offscreen with --minimum-gpu-work,
it reduces CPU overhead from 2.55ms to 2.20ms on Pixel6.
Bug: b/215768827
Change-Id: Ia2872442462917c0caadb263769a1cbf3dd7366f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3414356
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
945a09d1
|
2022-01-30T13:57:32
|
|
Vulkan: Fix off-by-one in DynamicBuffer.
allocateFromCurrentBuffer was using >= when it should have been
using >. This prevented filling a buffer to the exact maximum.
Bug: angleproject:6776
Change-Id: I9656e398080381f97d06912ff10fe62e7e7333ac
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3426882
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
665ddccd
|
2021-12-09T23:06:39
|
|
Vulkan: Emulate dithering
Dithering in OpenGL is vaguely defined, to the extent that no dithering
is also a valid dithering algorithm. Dithering is enabled by default,
but emulating it has a non-negligible cost. Similarly to some other
GLES drivers, ANGLE enables dithering only on low-bit formats where
visual banding is particularly common; namely RGBA4444, RGBA5551 and
RGB565.
Dithering is emulated in the fragment shader and is controlled by a spec
constant. Every 2 bits of the spec constant correspond to one
attachment, with the value indicating:
- 00: No dithering
- 01: Dither for RGBA4444
- 10: Dither for RGBA5551
- 11: Dither for RGB565
The translator appends code to the shader that, based on the format
specified by the specialization constant, adds dithering to each color
attachment output. A 2x2 Bayer matrix is used for dithering, indexed by
gl_FragCoord.xy % 2.
Bug: angleproject:6755
Change-Id: Ib45da5938e299b6626bff921119d63e7357dd353
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3374261
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
168e0ae5
|
2022-01-27T16:50:00
|
|
Vulkan: Refactor descriptor pool code.
This uses vk::Context to init DynamicDescriptorPool, and removes
some accesses to ProgramExecutableVk member variables so the methods
can more easily be moved into another class.
The changes should have no effective behaviour difference.
Note that the immutable sampler descriptor count is entirely set by
the format of the immutable samplers. However we can't determine this
count by looking at the descriptor set layout desc with the current
implementation.
Bug: angleproject:6776
Change-Id: I0a323d19fa3720b5afd8300401ed9c30b105c0ef
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3424657
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
701d51b1
|
2022-01-25T15:44:39
|
|
Vulkan: Add SmallBufferPool for small allocations
The small pool uses buddy algorithm that is much faster. The only
downside is that it rounds size to power of two. For small allocations
that rounding does not generate much waste and avoid fragmentation as
well. This CL adds a small pool for host visible non-coherent pool. My
testing with gardenscape shows that on top of other CLs, this reduces
CPU overhead from 1.77ms to 1.55ms as measured with --minimize-gpu-work
with offscreen.
Bug: b/215768827
Change-Id: I68434931f238c4e980b77d3df46d762260ef1db5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3415211
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Ian Elliott <ianelliott@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
d6dd0cb5
|
2021-09-21T15:26:13
|
|
EGL: EGL_KHR_lock_surface3 backend
Add vulkan implementation
Add test: EGLLockSurface3Test
Test: angle_end2end_test --gtest_filter=EGLLockSurface3Test
Bug: angleproject:6062
Change-Id: Id5bfe37895b550392d11e9e9cc1262c1f0288c42
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3174323
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
d2354968
|
2022-01-20T10:59:05
|
|
Vulkan: Rename BufferHelper::initFor* to allocateFor*
Simply name change per feedback from other CL's review.
Bug: b/205337962
Change-Id: Ieb53ed9a2922d09716a1219eb340fe273e5f1807
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3402882
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
1b5efe51
|
2022-01-19T14:04:35
|
|
Vulkan: Rename SubAllocation to Suballocation
Simply a name change to make it one word. No functional change is
expected.
Bug: b/205337962
Change-Id: Ic505536821f18141c0d036b13d9aa81554a8bafd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3403158
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
15439f8e
|
2022-01-13T14:58:41
|
|
Vulkan: Remove BufferMemoryAllocator
This class was added in crrev.com/c/3036256. The original intention was
to use VMA to implement buffer suballocation. Because VMA itself does
not support buffer suballocation, I was thinking to use VMA custom pool
to implement it and this class was intended to wrap all these
functionality into one class. But now thanks to Jamie's effort, VMA
exported generic suballocation algorithm via API and we have implemented
buffer suballocation using that virtual allocation API. So this
BufferMemoryAllocator class is really no longer useful. This CL mostly
reverted that CL and flatten out the buffer allocation call to directly
use VMA's Allocator object.
Bug: b/205337962
Change-Id: I0336056e440f39e2ff49fee8e0ff4b1f355cefe4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3244022
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
9d11a341
|
2021-12-16T13:47:04
|
|
Vulkan: Fix Vulkan SCB vs multiple subpasses
vkCmdNextSubpass must be called on the primary command buffer, so the
render pass command buffers need to be split on subpass boundaries.
This is only done when using Vulkan secondary command buffers.
Bug: angleproject:6811
Change-Id: I087fff305c757c78e87bfde4410e7de6bd1a6ba6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3344774
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
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>
|
|
|
caf6433a
|
2022-01-04T14:37:29
|
|
Vulkan: Rename releaseStagingBuffer to releaseStagedUpdate
There is no more per ImageHelper dynamic buffer for staging, and this
function is not just removing staging buffer, but removing all staged
updates. Rename it to reflect what it actually does.
Bug: b/208323792
Change-Id: I5a2182291848a9e137ed29b888f31215664b46fd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3367078
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
93159c02
|
2021-12-30T16:01:52
|
|
Vulkan: Remove DynamicShadowBuffer implementation
This is no longer used since mShadowBuffer was removed from BufferVk in
https://chromium-review.googlesource.com/c/angle/angle/+/3313333
Bug: b/208323792
Change-Id: I0bb1d66c2f0f58bac74139a274c957681edf2ad4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3362504
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
3263eb01
|
2021-12-28T10:10:50
|
|
Vulkan: Switch GPU translated vertex buffer to NonHostVisible
When we translate using GPU, there is no need to use host visible
memory. Use device local memory instead.
Bug: b/205337962
Change-Id: Ic76dcb28bde2f079f6ac406d846518bf5f369d74
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3340553
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
8589c456
|
2021-12-22T17:13:33
|
|
Vulkan: Remove mStagingBuffer from ImageHelper and ContextVk
This CL removes ImageHelper::mStagingBuffer and
ContextVk::mStagingBuffer and uses per shared group staging buffer pool
instead.
Bug: b/208323792
Change-Id: I46b5fd542697eb860def9c6d592dd147ad4dc973
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3354090
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
6da1ac81
|
2021-12-22T10:13:43
|
|
Vulkan: Switch ReadPixels from DynamicBuffer to shared pool
FramebufferVk::readPixelsImpl() and ImageHelper::copyImageDataToBuffer()
use per FramebufferVk DynamicBuffer. This CL removes this and uses
shared buffer pool to allocate a temporary staging buffer for readPixels
as needed and frees it immediately afterwards.
Bug: b/208323792
Change-Id: I65ddf9bf9f1f14578d9def63f5287cb1a4121dff
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3354038
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
ad27d5d6
|
2021-12-21T11:22:30
|
|
Reland "Vulkan: Consolidate all vertex conversion buffers to shared pool"
This is a reland of cca412cd8b349b7281727c50f2a59d115fd90a05
Further inspection shows it was red-herring. The original CL does not
have the un-intended diff that I saw in the commit email. This is
try to reland the original CL without any modification.
Original change's description:
> Vulkan: Consolidate all vertex conversion buffers to shared pool
>
> There are various conversion buffers that holds converted vertex or
> element or index data. They are DynamicBuffer for now. This CL switches
> them to use the shared group buffer pool. With this change, all
> allocation is represented by a BufferHelper object instead of an offset.
> I am able to remove the offset arguments from a lot of APIs.
>
> Bug: b/208323792
> Change-Id: Ib611beb0c16cddbdd9ddf7b8961c439da9fa5180
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3352489
> Reviewed-by: Tim Van Patten <timvp@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Commit-Queue: Charlie Lao <cclao@google.com>
Bug: b/208323792
Change-Id: I90852ad38c2b9ac423800bb6854757bcc17cd166
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3370602
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
a61a7f59
|
2021-12-28T18:16:40
|
|
Vulkan: Use correct aspectMask
Update ImageHelper::stageSubresourceUpdateFromImage() to use the correct
aspectMask flags based on the image's format, rather than assuming only
color formats.
Additionally, only call ensureMutable() for sRGB textures that indicate
srgbDecode == GL_SKIP_DECODE_EXT.
Bug: angleproject:6852
Test: DepthStencilFormatsTest.DepthTextureRender
Change-Id: Iaf1f167612144f836c5574b0f561cc1b58c3cb6a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3360090
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
4e85bdd9
|
2022-01-06T17:06:25
|
|
Revert "Vulkan: Consolidate all vertex conversion buffers to shared pool"
This reverts commit cca412cd8b349b7281727c50f2a59d115fd90a05.
Reason for revert: There is accidental code merge bug left in.
Original change's description:
> Vulkan: Consolidate all vertex conversion buffers to shared pool
>
> There are various conversion buffers that holds converted vertex or
> element or index data. They are DynamicBuffer for now. This CL switches
> them to use the shared group buffer pool. With this change, all
> allocation is represented by a BufferHelper object instead of an offset.
> I am able to remove the offset arguments from a lot of APIs.
>
> Bug: b/208323792
> Change-Id: Ib611beb0c16cddbdd9ddf7b8961c439da9fa5180
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3352489
> Reviewed-by: Tim Van Patten <timvp@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Commit-Queue: Charlie Lao <cclao@google.com>
Bug: b/208323792
Change-Id: I18bba207d1d8bb76dff32d9855a744dba93bc6d6
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3370601
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
cca412cd
|
2021-12-21T11:22:30
|
|
Vulkan: Consolidate all vertex conversion buffers to shared pool
There are various conversion buffers that holds converted vertex or
element or index data. They are DynamicBuffer for now. This CL switches
them to use the shared group buffer pool. With this change, all
allocation is represented by a BufferHelper object instead of an offset.
I am able to remove the offset arguments from a lot of APIs.
Bug: b/208323792
Change-Id: Ib611beb0c16cddbdd9ddf7b8961c439da9fa5180
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3352489
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
fd014271
|
2021-12-16T20:44:42
|
|
Vulkan: Match chromaFilter to min/mag for YUV AHBs
From the VVL:
VUID-VkSamplerCreateInfo-minFilter VkCreateSampler:
VK_FORMAT_FEATURE_SAMPLED_IMAGE_YCBCR_CONVERSION_SEPARATE_RECONSTRUCTION_FILTER_BIT
specifies that the format can have different chroma, min, and mag
filters. However,
VK_FORMAT_FEATURE_SAMPLED_IMAGE_YCBCR_CONVERSION_SEPARATE_RECONSTRUCTION_FILTER_BIT
is not supported for VkSamplerYcbcrConversionCreateInfo.format =
VK_FORMAT_UNDEFINED so chromaFilter needs to be equal to
minFilter/magFilter.
We don't know what the min/mag filters are when the AHB is imported, so
we need to update the YcbcrConversionDesc with the current min/mag
filter value before creating the VkSamplerYcbcrConversion when the
ImageView is created. Additionally, if the min/mag filters are updated
later, TextureVk::syncState() needs to recreate the ImageViews to ensure
the chromaFilter matches the new min/magFilters.
Test: atest android.media.cts.DecodeAccuracyTest#testGLViewDecodeAccuracy[0]
Test: ImageTestES3.SourceYUVAHBTargetExternalYUVSampleLinearFiltering
Bug: b/210526871
Change-Id: I95dbd9738f6e3fd0870e484518eee105e995f93a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3346394
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
0e49a3dd
|
2022-01-04T11:23:54
|
|
Vulkan: Add std::move support for BufferHelper
There are needs to support std::move for BufferHelpers in other CLs (See
crrev.com/c/3352489). Without this support, we can not store
BufferHelper into std::vector. This CL adds move support for
BufferHelper class.
Bug: b/208323792
Change-Id: I93f79490715750abc1bcedd41b683ad0c2460ebb
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3366855
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
43b0e92b
|
2021-12-13T15:38:32
|
|
Vulkan: Consolidate mHostVisibleBufferPool and mMapInvalidate
BufferVk::mHostVisibleBufferPool is allocated when BufferVk::mBuffer is
not hostvisible and we need to map it. In that case
mHostVisibleBufferPool is allocated and data copied from mBuffer to it
and the pointer to mHostVisibleBufferPool is returned to user.
BufferVk::mMapInvalidateRangeStagingBuffer is used when map is called on
a small range. In this case we allocate memory for the small range of
buffer and return that intead of waiting for entire buffer for GPU to
finish. Also when BufferSubData is called, we also needs to allocate a
staging buffer and issue a copyBuffer from staging buffer to main
buffer. This CL consolidate all these three usage cases into one
mStagingBuffer. It removes mHostVisibleBufferPool and
mMapInvalidateRangeStagingBuffer from BufferVk class. This makes overall
logic of managing data consistency much simpler as well since we only
have two buffers: The main buffer storage mBuffer or mStagingBuffer. And
mIsStagingBufferMapped tracks if mStagingBuffer is the one actually
mapped to user or not so that at unmap time we know if we should flush
the data to mBuffer or not.
Bug: b/208323792
Change-Id: I4f0c79a2d86da1a43844ed2ba83ddeb7dd4a5c0b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3335216
Reviewed-by: Lingfeng Yang <lfy@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
bc9d2d7d
|
2021-12-07T16:16:47
|
|
Vulkan: Consolidate BufferHelper's code paths into one
There are cases that BufferHelper may not sub-allocate from the pool.
For example, when buffer is created from external memory. Right now we
have two different code paths in the BufferHelper object and checking if
suballocation object is valid or not and pick different code path. This
CL consolidates both code path into one by making non sub-allocated case
also creates BufferSUbAllocation object and owns a BufferBlock object.
Bug: b/205337962
Change-Id: Iac82bdb0b69d424e4147b52d458ced6274e106a1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3322100
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
6c894e82
|
2021-11-04T14:49:41
|
|
Vulkan: Replace BufferVk::getBufferAndOffset() with getBuffer()
Now BufferHelper class already keeps offset information. There is no
reason for BufferVk to have that information any more.
Bug: b/205337962
Change-Id: I6e014fb480bfcd5018ef9231b0fb87a50021f179
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3266147
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
83a670ab
|
2021-10-29T09:12:26
|
|
Vulkan: Implement BufferPool using VMA's virtual allocator
VMA's allocation calls used to be sub-allocating a pool of memory. What
we really want is sub-allocate a VkBuffer object. VMA recently added
support to expose the underlying range allocation algorithm via APIs,
which user can use it to sub-allocate any object. This CL uses that new
virtual allocation API to sub-allocate from a pool of VkBuffers.
In this CL we only switched BufferVk::mBuffer to sub-allocate from the
BufferPool object.
Bug: b/205337962
Change-Id: Ia6ef00c22e58687e375b31bc12ac515fd89f3488
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3266146
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
6e5fd816
|
2021-12-15T18:31:13
|
|
Support more ImageLayout in ConvertImageLayoutToGLImageLayout
Bug: chromium:1264439
Change-Id: I35938a65981edf495bd9aafb8bfb17aa3ec9a66c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3341640
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Peng Huang <penghuang@chromium.org>
|
|
|
36fcf80b
|
2021-12-11T20:55:53
|
|
Vulkan: Consolidate SamplerYcbcrConversionCache
Now that SamplerDesc and ImageHelper have the fully qualified
description of a VkSamplerYcbcrConversion, we can lazy allocate
the conversion handle along with streamlining the interface to the
SamplerYcbcrConversionCache class.
Bug: angleproject:6732
Test: Texture2DTestES3.*Yuv*Vulkan
Change-Id: I7deaa0220031bc4c500d88021459017bf937218c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3333629
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: mohan maiya <m.maiya@samsung.com>
|
|
|
32f572b7
|
2021-12-03T17:58:51
|
|
Support creating EGLImage from VkImage
Bug: chromium:1264439
Change-Id: I520182143e748f25b44d0725f3f171b7b33a85d8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3311131
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Peng Huang <penghuang@chromium.org>
|
|
24241724
|
2021-11-22T13:26:53
|
|
Reland: Vulkan: Use vkResetQueryPoolEXT when available
This is a reland of 97aa5187ed164f1e9a22b64c966f2a747f651bfb
Original change's description:
> ANGLE currently uses vkCmdResetQueryPool to reset a query pool. As
> reported by Arm, this can cause GPU bubbles. It is better to use
> vkResetQueryPoolEXT() when available (either by Vulkan 1.2 or by the
> VK_EXT_host_query_reset extension).
>
> Bug: angleproject:6692
> Change-Id: I650d7d0f55bdf6587b75be43cc7bed2dc3eb935a
> Reviewed-on:
> https://chromium-review.googlesource.com/c/angle/angle/+/3294061
> Commit-Queue: Ian Elliott <ianelliott@google.com>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Reviewed-by: Charlie Lao <cclao@google.com>
Bug: angleproject:6692
Change-Id: Idff87977bd39ccd1d05684e48929f4cd3fa733c7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3313382
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Ian Elliott <ianelliott@google.com>
|
|
8815245a
|
2021-12-01T08:58:01
|
|
Vulkan: Don't retain in buffer/imageRead unnecessarily
mUsedBuffers/images already tracks whether a buffer/image is used in a
render pass, so we can add to ContextVk's resource use list only once.
Bug: angleproject:6717
Change-Id: Idd528c7f0f3fcb4d8d8879ab6cc7055260ecb45d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3310523
Commit-Queue: Lingfeng Yang <lfy@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
8f6f5a4b
|
2021-11-30T23:48:30
|
|
Vulkan: Fix image respecify's usage tracking
When respecifying an image due to mip level count changes, the previous
image is staged as an update to the new image. The resource usage info
was not being transferred to the image being staged as an update,
causing it to be prematurely deleted.
Test based on one authored by sugoi@google.com.
Bug: chromium:1270658
Bug: angleproject:4835
Change-Id: I215c65ba700d7be608d0910d3cb37fcfdf297a2a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3308921
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
44d0ef69
|
2021-11-25T14:37:38
|
|
Vulkan: Enhance SamplerDesc for YUV formats
In order to uniquely identify a VkSamplerYcbcrConversion, having just
the format is insufficient. We need to account for conversion model
and color component range. Refactor SamplerDesc by encapsulating YUV
related fields into its own YcbcrConversionDesc.
Bug: angleproject:6732
Test: Texture2DTestES3.*Yuv*Vulkan
Change-Id: I23114c50646232dccde988a52b0649778ad72a90
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3301899
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Mohan Maiya <m.maiya@samsung.com>
|
|
|
a5e391d7
|
2021-11-19T17:35:06
|
|
Add GL_ANGLE_vulkan_image extension
glAcquireTextures & glReleaseTextures are added for sharing
texture ownership with an external API.
Bug: chromium:1264439
Change-Id: If46d8d230b4f611768b5ff1187674509e42f01e0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3293921
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Peng Huang <penghuang@chromium.org>
|
|
7f1ab9be
|
2021-11-25T21:59:51
|
|
Vulkan: add explicit template specialization for barrierImpl
ImageHelper::barrierImpl with primary command buffer is used in
other translation units (SurfaceVk::acquireNextSwapchainImage()).
Linking fails if compiler has optimized out symbol for primary
command buffer.
Bug: chromium:819294
Change-Id: I91d79788ae51d24472e8bd1d20cdb1a375e86300
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3301705
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
e00ad443
|
2021-11-15T20:16:16
|
|
Add EGL_ANGLE_vulkan_image extension
This extension is for exporting VkImage from EGLImage.
The VkImage must be used with the same VkDevice used by ANGLE
Vulkan backend.
Bug: chromium:1264439
Change-Id: I222d900465cf2716d94fc64f06e240390ec518ac
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3285025
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Peng Huang <penghuang@chromium.org>
|
|
|
2fa1adf0
|
2021-11-24T18:29:13
|
|
Revert "Vulkan: Use vkResetQueryPoolEXT when available"
This reverts commit 97aa5187ed164f1e9a22b64c966f2a747f651bfb.
Reason for revert: This apparently caused crbug.com/1273344
Original change's description:
> Vulkan: Use vkResetQueryPoolEXT when available
>
> ANGLE currently uses vkCmdResetQueryPool to reset a query pool. As
> reported by Arm, this can cause GPU bubbles. It is better to use
> vkResetQueryPoolEXT() when available (either by Vulkan 1.2 or by the
> VK_EXT_host_query_reset extension).
>
> Bug: angleproject:6692
> Change-Id: I650d7d0f55bdf6587b75be43cc7bed2dc3eb935a
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3294061
> Commit-Queue: Ian Elliott <ianelliott@google.com>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Reviewed-by: Charlie Lao <cclao@google.com>
Bug: angleproject:6692
Change-Id: I70d5eeb0b98f016fd262935194f8f4abee2a72b9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3300309
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Reviewed-by: Ian Elliott <ianelliott@google.com>
Reviewed-by: Lingfeng Yang <lfy@google.com>
Commit-Queue: Ian Elliott <ianelliott@google.com>
|
|
|
924c8d8a
|
2021-11-18T22:58:04
|
|
Vulkan: Fix image tiling when DRM format modifiers are specified
Bug: angleproject:6679
Change-Id: I5198a180608535131b127c135423fb80c6478140
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3292027
Reviewed-by: Craig Stout <cstout@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
8b60855b
|
2021-09-15T15:16:10
|
|
EGL: implement EGL_KHR_mutable_render_buffer
Pass render buffer mode change to WindowSurfaceVk.
On mode change trigger OUT_OF_DATE. Then in CreateSwapchain,
if new mode, set the Presentation mode and the Image count.
OffscreenSurfaceVk ignores mode change.
Add MUTABLE_RENDER_BUFFER_BIT to GenerateDefaultConfig.
Test: dEQP-EGL.functional.mutable_render_buffer.*
Bug: angleproject:3966
Change-Id: I7b59708514bcda10f8d45ce5f9528aa840fcccfa
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3171822
Commit-Queue: Brandon Schade <b.schade@samsung.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
97aa5187
|
2021-11-22T13:26:53
|
|
Vulkan: Use vkResetQueryPoolEXT when available
ANGLE currently uses vkCmdResetQueryPool to reset a query pool. As
reported by Arm, this can cause GPU bubbles. It is better to use
vkResetQueryPoolEXT() when available (either by Vulkan 1.2 or by the
VK_EXT_host_query_reset extension).
Bug: angleproject:6692
Change-Id: I650d7d0f55bdf6587b75be43cc7bed2dc3eb935a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3294061
Commit-Queue: Ian Elliott <ianelliott@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
|
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>
|
|
|
6fe13477
|
2021-11-11T00:47:14
|
|
Vulkan: Add external's pNext to ANGLE_external_objects_flags
ANGLE was chaining VkImageFormatListCreateInfoKHR to
VkImageCreateInfo::pNext to support sRGB extensions. For external
images, it was unknown whether that was valid because there was no way
to know if external used an identical chain of pNexts. This was causing
a discrepancy between images created by Chrome and those created by
ANGLE as part of an import.
This change updates ANGLE_external_objects_flags to take in the pNext
chain external has used to create the image so ANGLE could create the
image identically.
Bug: chromium:1266094
Change-Id: I479b9e7ff39d437425dc91c79834880749766f99
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3274177
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Peng Huang <penghuang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
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>
|
|
|
d2d3a546
|
2021-11-11T12:22:04
|
|
Vulkan: Write perf warnings in command buffer
It's much easier to understand what command the perf warning refers to
when it's visible in the command buffer using a graphics API debugger.
This change creates ANGLE_VK_PERF_WARNING which gives the warning both
to the application (through ANGLE_PERF_WARNING) and inserts it in the
command buffer.
Bug: angleproject:2472
Change-Id: Ie84feed53eca5cda93e1f2bc653fcbf9bcd57b56
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3275839
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
558981c1
|
2021-11-10T23:13:03
|
|
Vulkan: Make write-after-invalidate checks more precise
Previously, the size of the command buffer was used as indication for
whether the render pass attachments might have been modified after
glInvalidateFramebuffer. In that case, the invalidate was undone. This
is made more precise by making sure only vkCmdClearAttachments and
vkCmdDraw* calls are counted for this purpose.
For example, inserting event markers after glInvalidateFramebuffer now
retains the invalidation.
Note that this can be even further optimized by tracking real writes to
attachments. For example, currently a draw call with depth test
disabled still undoes the invalidation of the depth buffer, but it
shouldn't.
Bug: angleproject:5079
Change-Id: I6257b4116a73213884b919bc7f3c86ff39b6aeed
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3274176
Reviewed-by: Ian Elliott <ianelliott@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
53371cc0
|
2021-11-01T20:25:17
|
|
Vulkan: Use optimalBufferCopyOffsetAlignment
optimalBufferCopyOffsetAlignment is the optimal buffer offset alignment
in bytes for vkCmdCopyBufferToImage2KHR, vkCmdCopyBufferToImage,
vkCmdCopyImageToBuffer2KHR, and vkCmdCopyImageToBuffer. The per texel
alignment requirements are enforced, but applications should use the
optimal alignment for optimal performance and power use.
To improve efficiency, this CL updates ContextVk::mStagingBuffer's
alignment to the max of:
- minMemoryMapAlignment
- nonCoherentAtomSize
- optimalBufferCopyOffsetAlignment
On ARM, this is not expected to have any affect, since all three values
are 0x40, but other platforms may see a benefit.
Bug: angleproject:4297
Change-Id: I9185da111e09c5d782eb1dedb10369727cb9bf51
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3256007
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Charlie Lao <cclao@google.com>
|
|
|
5d8a89e4
|
2021-11-02T19:57:04
|
|
Vulkan: Always override surface format GL_RGB8 to GL_RGBA8
If an app requests to create a surface with GL_RGB8, override it to be
GL_RGBA8 for Android.
This is to workaround an issue with the Android Vulkan loader which
limits which formats can be used with swapchains.
This CL also adds GL_RGB8 back to DisplayVkAndroid::generateConfigs(),
effectively reverting the following CL:
https://chromium-review.googlesource.com/c/angle/angle/+/3235466
This is being done with this CL (rather than reverting) since these
changes are required to handle surfaces created with GL_RGB8.
Bug: angleproject:6277
Bug: angleproject:6651
Change-Id: Iad78ea0d7bdf12e1e309ed6a7181f08fac38b9de
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3258143
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Ian Elliott <ianelliott@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
4fd4eb50
|
2021-11-05T11:14:47
|
|
Vulkan: Minor cleanups.
Makes some VkFlags into VkAccessFlags to aid with code search. Also
corrects a few typos.
Bug: angleproject:6566
Change-Id: Ia4d14d38bb6d4e8c4c71fed6038c12edae112be9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3264224
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
afaa1287
|
2021-10-26T19:29:44
|
|
Vulkan: Don't submit XFB queries when XFB is inactive
Consider the following GL commands:
glBeginQuery(GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN)
glBeginTransformFeedback()
glDrawArrays()
glPauseTransformFeedback()
glDrawArrays()
glResumeTransformFeedback()
glDrawArrays()
glEndTransformFeedback()
glEndQuery(GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN)
The renderpass is broken for each of the draw calls, leading to
pauseRenderPassQueriesIfActive() and resumeRenderPassQueriesIfActive()
being called when the RP is stopped/started. This leads ANGLE to
generate 3 TransformFeedbackPrimitivesWritten queries when looping
through the "active" queries, since no consideration is made for XFB
actually being active/unpaused.
ARM doesn't support the TransformFeedbackPrimitivesWritten query when
XFB is disabled. Instead, the query is never completed and the results
are never available. It could probably be argued that this is an ARM
bug, and they should just output '0' if XFB isn't active during the
query.
Regardless, ANGLE shouldn't be issuing the queries when XFB is no
active, since it's wasteful.
Bug: angleproject:6622
Test: dEQP.GLES3/functional_transform_feedback*
Test: TransformFeedbackTest.TransformFeedbackQueryPausedDrawThenResume
Change-Id: I380fb60405d7f538b802bb0df72bf609f0bc58e9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3246532
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
32b3c963
|
2021-10-28T11:00:03
|
|
Vulkan: Remove MSRTT assertion in the presence of OOM
When the MSRTT transient multisampled image is created, the emulated
clear staged update is removed (as it's also staged for the single
sampled image). This was done after allocating memory for the image,
which can fail, leaving the update staged and leading to an ASSERT
failure when the texture is used.
Technically, the texture should not be used, and that's a Chrome bug
where the OOM result from ANGLE is being ignored. This change makes
sure the staged update is removed before allocating memory so the
multisampled image is in a more consistent state.
Bug: chromium:1263046
Change-Id: Ida8923ccbc517f740d76b4c2880086a039fa8a1f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3251580
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
bae19e06
|
2021-10-26T13:35:57
|
|
Vulkan: Avoid unnecessary wait if mapBufferRange indicates read only
When we call BufferVk::mapRangeImpl(), both from internal code paths for
data reads or due to glMapBufferRange call, we are not passing the
access bit to the call. This CL passes the proper access bits to the
call and only wait for GPU writes to finish if access is for read only.
This CL also adds access bitfield to the BufferVk::mapImpl() API and
have various callers pass in the proper access bits as well.
Bug: b/203582620
Change-Id: Ica8493c902dbd7b15996266c81ce0fd4dbfc2520
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3245487
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
191c236a
|
2021-10-14T17:01:45
|
|
Vulkan: Remove "current queue serial".
Queries, semaphores, and pipelines instead use the normal
vk::Resource design to track their lifetimes. Removes the current
serial APIs from all classes. Current serials are still tracked
internally in the command queue classes.
Bug: b/169788986
Change-Id: Idcd2c2a93bc8225c6f3f7c247eb8fcfb76be1030
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3223644
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
421dbf20
|
2021-10-26T19:06:19
|
|
Revert "Vulkan: Use different strategy for buffer memory allocation"
This reverts commit 0dbe308d91a3526e28fecd9014d873c649f4ef71.
Reason for revert: crbug.com/1253325
Original change's description:
> Vulkan: Use different strategy for buffer memory allocation
>
> This CL uses different memory allocation strategy based on the requested
> size. If the requested size exceeds 1M, we use dedicated memory
> allocation to avoid memory waste associated with the sub-allocator.
> Otherwise we uses VMA's sub-allocator pool. This CL creates two sets of
> customized pool so that we uses different allocation strategy for each
> set of pool: the small pool uses buddy algorithm which favors speed over
> memory and large pool uses default algorithm that favors memory saving
> over speed. This CL also replaces vmaFindMemoryTypeIndexForBufferInfo
> with vmaFindMemoryTypeIndex to avoid create and destroy VkBuffer object
> just try to find memoryTypeIndex.
>
> Bug: b/195588159
> Change-Id: I2bddbfffd77ba2ce6b9389d83a31051c4b748c4d
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2939490
> Commit-Queue: Charlie Lao <cclao@google.com>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Reviewed-by: Tim Van Patten <timvp@google.com>
Bug: b/195588159
Change-Id: I2c8fe8cb2930d16f5212570d32be68e7c6a6e5f3
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3244258
Reviewed-by: Charlie Lao <cclao@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
|
|
|
9ac2409e
|
2021-10-26T00:55:06
|
|
Reland "Vulkan: Flush descriptor set updates during flush*Commands()"
This reverts commit 420e77a64e890ad4c585d72789ca59501be3ac75.
Reason for revert: The crash is in ProgramExecutableVk::getOrAllocateShaderResourcesDescriptorSet(), which is unrelated to this change. Creating a CL to reland this (the reland button is failing), to try and investigate further.
Original change's description:
> Revert "Vulkan: Flush descriptor set updates during flush*Commands()"
>
> This reverts commit 02b73c2fd738b237f1ab3ecd400feec222903a48.
>
> Reason for revert: Causes test to crash on Linux/Intel: anglebug.com/6591
>
> Original change's description:
> > Vulkan: Flush descriptor set updates during flush*Commands()
> >
> > The intent of this CL is to reduce the number of descriptor set updates
> > by delaying the work until all of the GLES commands that could trigger a
> > re-update have been performed and the command stream is being flushed.
> > To achieve this, flushDescriptorSetUpdates() is being moved from
> > setupDraw()/setupDispatch() to
> > flushRenderPassCommands()/flushOutsideRPCommands().
> >
> > This change also exposed an issue where the BufferView handles were not
> > being preserved until flushDescriptorSetUpdates() was called. To resolve
> > this, flushDescriptorSetUpdates() is also being called during
> > BufferViewHelper::release() before the BufferView memory is released.
> >
> > Bug: angleproject:5706
> > Change-Id: I61e19af9c0fac891aa2115d72391459b80d22f19
> > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2939385
> > Reviewed-by: Charlie Lao <cclao@google.com>
> > Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> > Reviewed-by: Jamie Madill <jmadill@chromium.org>
> > Commit-Queue: Tim Van Patten <timvp@google.com>
>
> Bug: angleproject:5706
> Bug: angleproject:6591
> Change-Id: I9d1ee3fcb3d1aebc86e60896e0065cab847a92b4
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3233901
> Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Bug: angleproject:5706
Bug: angleproject:6591
Change-Id: Ibc7e934b82e6cd90e766f43135583f7c45acfe39
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3244024
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
0dbe308d
|
2021-07-16T14:26:20
|
|
Vulkan: Use different strategy for buffer memory allocation
This CL uses different memory allocation strategy based on the requested
size. If the requested size exceeds 1M, we use dedicated memory
allocation to avoid memory waste associated with the sub-allocator.
Otherwise we uses VMA's sub-allocator pool. This CL creates two sets of
customized pool so that we uses different allocation strategy for each
set of pool: the small pool uses buddy algorithm which favors speed over
memory and large pool uses default algorithm that favors memory saving
over speed. This CL also replaces vmaFindMemoryTypeIndexForBufferInfo
with vmaFindMemoryTypeIndex to avoid create and destroy VkBuffer object
just try to find memoryTypeIndex.
Bug: b/195588159
Change-Id: I2bddbfffd77ba2ce6b9389d83a31051c4b748c4d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2939490
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
52934a5d
|
2021-10-04T22:19:30
|
|
Vulkan: Use the UNDEFINED layout for ExternalPreInitialized
According to the spec, the UNDEFINED layout does not invalidate the
memory of external images.
Bug: angleproject:6460
Change-Id: I682e8c3501cc52431268ff2c3a063467ba9d2a65
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3203796
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
420e77a6
|
2021-10-20T15:02:45
|
|
Revert "Vulkan: Flush descriptor set updates during flush*Commands()"
This reverts commit 02b73c2fd738b237f1ab3ecd400feec222903a48.
Reason for revert: Causes test to crash on Linux/Intel: anglebug.com/6591
Original change's description:
> Vulkan: Flush descriptor set updates during flush*Commands()
>
> The intent of this CL is to reduce the number of descriptor set updates
> by delaying the work until all of the GLES commands that could trigger a
> re-update have been performed and the command stream is being flushed.
> To achieve this, flushDescriptorSetUpdates() is being moved from
> setupDraw()/setupDispatch() to
> flushRenderPassCommands()/flushOutsideRPCommands().
>
> This change also exposed an issue where the BufferView handles were not
> being preserved until flushDescriptorSetUpdates() was called. To resolve
> this, flushDescriptorSetUpdates() is also being called during
> BufferViewHelper::release() before the BufferView memory is released.
>
> Bug: angleproject:5706
> Change-Id: I61e19af9c0fac891aa2115d72391459b80d22f19
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2939385
> Reviewed-by: Charlie Lao <cclao@google.com>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Commit-Queue: Tim Van Patten <timvp@google.com>
Bug: angleproject:5706
Bug: angleproject:6591
Change-Id: I9d1ee3fcb3d1aebc86e60896e0065cab847a92b4
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3233901
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
02b73c2f
|
2021-07-14T15:45:47
|
|
Vulkan: Flush descriptor set updates during flush*Commands()
The intent of this CL is to reduce the number of descriptor set updates
by delaying the work until all of the GLES commands that could trigger a
re-update have been performed and the command stream is being flushed.
To achieve this, flushDescriptorSetUpdates() is being moved from
setupDraw()/setupDispatch() to
flushRenderPassCommands()/flushOutsideRPCommands().
This change also exposed an issue where the BufferView handles were not
being preserved until flushDescriptorSetUpdates() was called. To resolve
this, flushDescriptorSetUpdates() is also being called during
BufferViewHelper::release() before the BufferView memory is released.
Bug: angleproject:5706
Change-Id: I61e19af9c0fac891aa2115d72391459b80d22f19
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2939385
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Tim Van Patten <timvp@google.com>
|
|
|
eb1d2474
|
2021-10-04T20:23:26
|
|
Vulkan: Add R8G8B8_UNORM to vk_format_map.json
The map of R8G8B8_UNORM -> VK_FORMAT_R8G8B8_UNORM is missing from
vk_format_map.json. This leads to attempting to get format bits for
angle::FormatID::NONE / VK_FORMAT_UNDEFINED.
This CL adds R8G8B8_UNORM to vk_format_map.json, so devices that support
it, like ARM, can use it. This allows those devices to import AHBs
created with R8G8B8_UNORM.
This CL also adds some additional ASSERT() and UNREACHABLE() calls to
catch missing formats sooner.
This allows the EGL tests to pass on ARM devices, but the tests are
still skipped due to Qualcomm not having the necessary support for
optimal tiling.
Bug: angleproject:6277
Test: dEQP-EGL.functional.image*
Change-Id: I41b7ead97a10ae80be4b6048d36d0d85ad71784f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3204152
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Charlie Lao <cclao@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
22ecd454
|
2021-07-19T17:28:05
|
|
Vulkan: Move BufferMemory class from vk_helpers.h to vk_utils.h
In preparation for the StagingBuffer may also use BufferMemory, this CL
moves the class to vk_utils.h so that it can be used by any class
without create kinks on header file inclusion.
Bug: b/195588159
Change-Id: I97ab136be691322fe6284fd1b4b7336fdf1f904f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3040106
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
8ef1b2a2
|
2021-07-15T15:27:15
|
|
Vulkan: Add BufferMemoryAllocator class
BufferMemoryAllocator class will be responsible for memory allocation
for vkBuffers. The original Allocator class remains the wrapper for
vmaAllocator.
Bug: b/195588159
Change-Id: I26d016e2f8d13ba459d0aa0bd4e1ebf4d0d48251
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3036256
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
6793d50e
|
2021-10-06T15:48:55
|
|
Vulkan: Remove ImageHelper::setFirstAllocatedLevel and Add assertion
ImageHelper::mFirstAllocatedLevel is set via ImageHelper::initExternal.
There isn't a good use to set the mFirstAllocatedLevel when mImage is
invalid, and semantically it is wrong to try to get mFirstAllocatedLevel
or format while its storage is invalid (i.e., not allocated). This CL
removes ImageHelper::setFirstAllocatedLevel() API all together and added
assertion that image is valid when ImageHelper::getFirstAllocatedLevel
and getActualFormat and various other get APIs whose value are set
during initExternal call.
Bug: angleproject:6532
Change-Id: I9c97f488b3e959443fec632540e7f0906a49dc13
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3209838
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
185e48a7
|
2021-10-06T18:53:08
|
|
Vulkan: Make TextureVk::mRedefinedLevels bitmask of gl::LevelIndex
We have gl::LevelIndex and vk::LevelIndex. gl::LevelIndex means level
index at OpenGL concept and is used in TextureVk code. vk::LevelIndex
means already adjusted with baseLevel and is used by ImageHelper object.
This mRedefinedLevel is gl::TexLevelMask type, but it actually is a bit
mask of vk::LevelIndex right now. This is inconsistent and confusing.
This CL changes it to bit mask of gl::LevelIndex to be consistent with
the namespace of "gl".
Bug: angleproject:6532
Change-Id: Ic88e373f70d3bd3e3fea676765c82fd0941e245c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3210726
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
6cb599f9
|
2021-10-06T13:07:27
|
|
Vulkan: Change dest to dst for consistency
Cleanup only, no functional change. dst aligns better with src.
Bug: angleproject:6502
Change-Id: I69821b1aae50a7ce647c7cc876468b6de309eec8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3208514
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>
|
|
|
74b2886f
|
2021-09-22T13:49:57
|
|
Vulkan: Try to use CPU to copy when preserving old buffer data
When glBufferSubData is called, we may acquire a new buffer if buffer is
still GPU busy. When this happened, we have to preserve buffer content
if old buffer has valid data in it. Instead of always use GPU to do
copy, this CL will check if GPU is not writing to the buffer, we will
just use CPU to do the copy form old buffer to new buffer from the
ranges outside subData, controlled by the feature flag
preferCPUWhenPreservingBufferData.
Bug: b/200067929
Change-Id: I42053104b2be8da5f399cca92e934254988f2fd8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3177322
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
4f0e7e7f
|
2021-10-04T14:50:10
|
|
Vulkan: Remove extraCopyBufferRegion workaround.
Bug: angleproject:2838
Change-Id: I53ba3cf307f930ae9550d1bfee4650df6e826930
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3202475
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
2f1d503b
|
2021-09-16T21:49:14
|
|
Vulkan: Fix Vulkan secondary command buffers
This change abstracts initialization, begin/end, reset and recycle of
command buffers such that both Vulkan and ANGLE secondary command
buffers are supported.
Bug: angleproject:6100
Change-Id: I8c79764ac98b599fda08fe45cf8c4f0a6573f0f9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2987873
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
a097ee2e
|
2021-09-27T15:18:02
|
|
Vulkan: Fix clearing external images with emulated format
External images may already have data, and clearing them due to their
format being emulated is incorrect.
This change makes sure that only the emulated channels are cleared.
The RGBXDataPreservedAHB test is based on one contributed by
Jason Macnak <natsu@google.com>.
Bug: b/192315789
Change-Id: Ibc8953fdac356f2a62a5b46512a51e1916b4a1b6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3193416
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
|
|
|
d9882db5
|
2021-04-16T21:47:25
|
|
Vulkan: Destroy unused resources in a timely manner
Deletion of garbage that's eligible for immediate cleanup is currently
delayed until the next submission.
Fix this by checking if garbage can be freed immediately and deleting it
right away if it's not used. Some missing retain() calls needed to
be added to semaphore operations in order to avoid freeing resources
in use by a memory barrier.
Based on a change by spang@chromium.org
Bug: angleproject:4628
Change-Id: I034c24d7631c973b4546952f7718a0cd4114763f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2832002
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
dc99db91
|
2021-09-16T21:45:35
|
|
Vulkan: Allow Vulkan secondary CBs to count their commands
This is used by render pass related code to determine if some command
has been issued since a previous event. With ANGLE
SecondaryCommandBuffer, the memory pointer where the commands are being
written to are used for this purpose. For Vulkan secondary command
buffers the code simply increments a counter.
Bug: angleproject:6100
Change-Id: I85320f8453bd3325793df85aabde3d42b1eeb22a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3167214
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|