This patch fixes the issues on the initialization of buffers with
MapRead usage and mappedAtCreation == true.
1. The buffers with MapRead usage and mappedAtCreation == true can be
read from the CPU side just after the creation of the buffer,
however at that time the internal pending command buffers may not
be executed, thus causing the buffer is not cleared as is expected.
2. On D3D12 the buffer with MapRead and mappedAtCreation == true is
created on the READBACK heap, so all the data written in the CPU
side cannot be uploaded to the GPU memory. When the buffer is mapped
again all the original data written through the CPU pointer will be
overwritten by the data in the GPU memory (which means it is also
cleared to 0).
This patch fixes this issue by:
1. clearing the buffers with mappedAtCreation == true on the CPU side.
2. on D3D12 making the buffer with MapRead and mappedAtCreation == true
use the staging buffer instead of mapping itself.
Note that this change is only related to the code path with Toggle
"nonzero_clear_resources_on_creation_for_testing" enabled, currently
we don't plan to do the similar change when we enable Dawn wire.
BUG=dawn:414
TEST=dawn_end2end_tests
Change-Id: I2b3d0840333e8d99759800ab9fc141d0a7cf2f8d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27220
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
ClusterFuzz injects errors into the device which may not
initialize the resource manager which would later fail upon
shutdown. This adds a check to see if the resource manager
exists before releasing the pool.
BUG=chromium:1117595
Change-Id: Ie2b90f947f3ddd3906ecabc65e959915c1f7e386
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27040
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
More code is now shared across backends.
Bug: dawn:483
Change-Id: I7ca1b8cbc2f12e408c94fbe5bca9fd29e47e0004
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27021
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
WebGPU currently allows as many as 108 view descriptors per bind
group. This is too many to have one descriptor allocator per size,
so we need to bucket them by size.
Bug: dawn:443, dawn:488
Change-Id: I4fc8cf7cd0dc8292bb6a8488fd2ceb7575e1e5f7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24787
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Commit-Queue: Austin Eng <enga@chromium.org>
kMaxBindingsPerGroup is not a real WebGPU limit. Code in Dawn that
uses it now uses dynamically sized vectors, optimistic stack_vecs or
a different limit like kMaxDynamicBuffersPerPipelineLayout.
The CL introduces kMaxOptimalBindingsPerGroup=32 which is used to size
the static portion of stack_vecs, but the value is tentative.
Bug: dawn:443
Change-Id: I08e06bed6531bed8b4365f36cf2fc0579ac5f180
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/23502
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
This patch implements buffer lazy initialization before
CopyBufferToBuffer() behind the toggle LazyClearBufferOnFirstUse.
- If the source buffer is not initialized, it will be cleared to 0
before CopyBufferToBuffer().
- If the destination buffer is not initialized and the copy doesn't
overwrite the whole buffer, it will be cleared to 0 before
CopyBufferToBuffer(), otherwise the buffer shouldn't be cleared.
BUG=dawn:414
TEST=dawn_end2end_tests
Change-Id: I3d0512c6376a1ed8928e86f8e56fefebc16910fa
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24360
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This required changing DeviceBase::CreateBufferImpl to return
ResultOrError<Ref<BufferBase>>
Bug: chromium:1103154
Change-Id: I1a5811d293333b6ef29c988a08f2f1f84ac65702
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24500
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch implements buffer lazy initialization before writeBuffer():
if the buffer is not initialized and writeBuffer() doesn't cover the
whole buffer, the buffer will be cleared to 0, otherwise the buffer
shouldn't be cleared.
This patch also introduces a toggle LazyClearBufferOnFirstUse for the
development of buffer lazy initialization: before buffer lazy
initialization being completely supported, all the related code will
only be enabled behind this toggle to prevent the buffers with valid
content being unexpectedly cleared.
BUG=dawn:414
TEST=dawn_end2end_tests
Change-Id: I99a2aa98ca4b9b21d69c6b32080afb525e2c4ad3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24041
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Device shouldn't be used until after DeviceBase::Initialize so we need
to move the NextSerial call to after Device Initialization is completed.
Bug: chromium:1094477
Change-Id: I2ab317c1d75f853c022faac3fc0bbd66fbd7c14e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/23261
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
To avoid overly ticking, we only want to tick when:
1. the last submitted serial has moved beyond the completed serial
2. or the completed serial has not reached the future command serial added
by the trackers (MapRequestTracker, FenceSignalTracker, ErrorScopeTracker).
Bug: dawn:400
Change-Id: Ie7c65acc332846ac1a27f9a18f230149d96d2189
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19062
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
The windows.h macros were undefined only at the end of this platform
header previously but with the addition of d3d12sdklayers.h the
definition of ID3D12DebugQueue::GetMessage picked up the macro and
became GetMessageA or GetMessageW, but Dawn code referred to it as
GetMessage causing a compilation error.
Fix this by preemptively loading windows.h and undefing some of the
macros so that the D3D12 headers don't see them.
Bug:
Change-Id: I1985cc20a9bdec1d25619ac5088e918b2acf8ecb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/22400
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This will help remove noise in debug layer output. This patch also
promotes warnings and higher to Dawn errors.
BUG: dawn:363, dawn:418, dawn:419, dawn:421
Change-Id: I3112c94aff71fc7e76dff48c82bafe9e051ed3b4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21702
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Adds logic to lock residency for bound descriptor heaps, then unlock and
insert into the LRU cache when no longer bound. Adds a basic functional
test.
Bug: dawn:193
Change-Id: Idfaaee6b873374c07a0b94b1982ad65353218799
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21400
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
All the buffer backend files had basically the same implemenations
of MapRequestTracker and the tracker was owned by device backends.
This refactor puts MapRequestTracker into its own file
and has the tracker be owned by DeviceBase and BufferBase.
Bug: dawn:400
Change-Id: Id28422b575e9c04d4435d5f119e0ffe08c2d1ce8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21760
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Factor out common shader compilation logic to ShaderModuleD3D12
used by both RenderPipeline and ComputePipeline, and implement
a new compilation path using DXC when UseDXC toggle is enabled
Bug: dawn:402
Change-Id: I67d3ae0aecee11634af917735456ddbe10b3d86a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21840
Commit-Queue: Hugo Amiard <hugo.amiard@laposte.net>
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Allows bindgroups that use the same samplers to share
a descriptor heap allocation. This is particularly important
for sampler heaps which incur expensive pipeline flushes
due to the smaller size requiring more frequent switches.
The device dolls out entries to a sampler heap allocation cache.
When the BindGroup is created, it does a lookup and refs the
allocation. This ensures the cache does not grow unbounded
or needlessly store unused entires.
This change is a follow-up of de-coupling heaps.
BUG=dawn:155
Change-Id: I3ab6f1bdb13a40905cb990cd7a2139e73da30303
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/20783
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch adds the following:
- UseDXC toggle
- Loads DXC (and DXIL to sign the DXBC) in d3d12/PlatformFunctions
- Adds GetModuleDirectory to SystemUtils
GetModuleDirectory was added to prevent loading issues regarding dynamic libraries when the executable is not in the same path as the dawn module.
This patch doesn't add DXC to RenderPipelineD3D12 nor ComputePipelineD3D12.
Bug: dawn:402
Change-Id: I2b8e4a2b7df31b9c766c748f92e11050c0aec3a0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21420
Commit-Queue: Felix Maier <xilefmai@gmail.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Move mCompletedSerial and mLastSubmittedSerial to Device frontend and
add getters and setters for the device backend to access such.
This is to aid the Device in taking more ownership of Serials and Ticking.
Bug: dawn:400
Change-Id: Ifa53ac294a871e484716842a3d212373b57847c4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/20480
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Allows bindgroups to be populated by heap type. Previously, failing to
populate one type of GPU descriptor heap required both GPU heaps
to be switched out. This resulted in extra copies and heap allocations
should only one heap type overflow.
This change also simplifies GPU descriptor heap management:
- Allocator no longer needs to operate on both heaps.
- Sub-allocation tracking can be moved into handles.
A follow-up change will remove duplicated sampler heap allocations.
BUG=dawn:155
Change-Id: I1960cf3a8bc3d86d3e8b2775da3d0c92125bcf82
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19887
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
To avoid accidental memory leaks on account of using raw pointers,
use Ref<TextureBase> as method return type except at Dawn interface
boundaries.
Change-Id: I6459062ee28984de2cb1d5a2059bc70cf82b2faf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19580
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
Uses the staging descriptor allocator to enable
sub-allocation for RTV/DSVs.
This change also simplifies CPU descriptor heap
management for render-passes:
- Allocating slot-by-slot removes extra attachment
pass.
- No need for the slower direct allocation allocator.
- Move RP creation of view handles into RP builder.
BUG=dawn:155
Change-Id: I508492a2e56a897bf8c85f9a45cd13f62fa0a2ef
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/20042
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This makes all backends register the default queue at device
initialization time, so that the same queue is returned by
each call to GetDefaultQueue.
All usages of CreateQueue are replaced by GetDefaultQueue
except a couple ones that could use the queue initialized by
DawnTest::SetUp.
A deprecation warning mechanism is added so that users of Dawn
can now that they should upgrade their usage of the API. It also
comes with a backdoor so we can test that they are emitted.
New DeprecatedAPITests are added that will contain tests for
deprecated APIs, and will also check that deprecation warnings
are produced.
The special casing of GetDefaultQueue in the wire will be done
in a follow-up CL to ease the review. It happens to work through
the regular wire mechanisms at the moment but returns a different
object on each GetDefaultQueue call.
Bug: dawn:22
Change-Id: I78dc1fa474769674278d30040e8d05c658b88360
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19724
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Accessing the d3d device using COM needlessly
refcounts. This is a particular issue in areas
that frequently access the d3d device like
Populate().
BUG=dawn:155
Change-Id: I24e83093623afd02fa592d8ec0c404b4571c374b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19703
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Move the CPU descriptor allocators to the device and bucket them to
ensure only kMaxBindingsPerGroup exist rather than create them per BGL.
Also, renames NonShaderVisible => Staging.
BUG=dawn:155
Change-Id: If6dae368e7e2a2b349343bdf898041a049159038
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19001
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
What was previously the Device's loss status is now a state that also
contains the "being created" state. Its transitions are entirely
handled in the frontend which enforces somewhat uniform lifecycles
between backends.
The backend devices' ShutDownImpl() function is now guaranteed to be
called only during the destructor, which leads to further simplification.
Previously Destroy() could also be called when the device was first
lost. This require complications because, for example, a WGPUBuffer
could still exist, and would want to call some resource allocator
service after the call to Destroy(). Now destruction of the device's
backing API objects is deferred to the destructor. (that's ok as long
as the application can't submit any more work).
WaitForCompletion is now guaranteed to be called before ShutDownImpl() iff
the call to DeviceBase::Initialize was succesful and the backing device
not lost. The idea is that after DeviceBase::Initialize, the GPU can
have some work enqueued and we need to wait for it to complete before
deleting backing API objects. In the future we might also have backend
be reentrant, using WebGPU itself to implement parts of the backend.
Reentrant calls would only be allowed after DeviceBase::Initialize.
Also the DynamicUploader that was special-cased in all backends is
now handled entirely by the frontend.
Bug: dawn:373
Change-Id: I985417d67727ea3bc11849c999c5ef0e02403223
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18801
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
RefCounted (and derived) destructors should be protected on the class
to ensure the objects can ONLY be destructed by calling Release. This
avoids errors cause by destroying objects out from under code which
has an active reference count.
Unfortunately, many of the 'base' classes must continue having public
destructors because they are used as "blueprint" objects created on
the stack.
Added final on most-derived classes.
Ideas for future improvement:
- Change "base" objects to have protected destructors but create new
blueprint objects that privately derive from base objects. This
limits the blueprint object's usefulness to only be a blueprint.
- Modify createX methods to return Ref<Object> instead of Object*
Change-Id: I6f3b3b178118d135c4342cb912e982a3873d71af
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18780
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
In PIX's D3D12-only mode, there is no way to determine frame boundaries
for WebGPU since Dawn does not manage DXGI swap chains. Without
assistance, PIX will wait forever for a present that never happens.
If we know we're dealing with a swapbuffer texture, inform PIX we've
"presented" the texture so it can determine frame boundaries and use the
texture's contents for the UI.
Bug: dawn:364
Change-Id: I7eb628c460e1e7c446ad91b29b03dd7b54545afb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18060
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
We want to release pending commands in Device::WaitForIdleForDestruction()
so that when we call TickImpl(), we can reset the command allocators.
Bug: dawn:269
Change-Id: Ibd8fdd685a4e9ed7cce8176ba097bc67687bcd97
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18101
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Adds a toggle to force the use of small shader-visible heaps and
whitebox tests to verify bindgroup encoding correctness.
BUG=dawn:155
Change-Id: I4118b850d9f2cb445ae805aa68ebf4fab671261b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16960
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Use D3D12's QueryDeviceVideoMemoryInfo to get the OS-determined process
budget. Also introduces an export for reserving some amount of process
memory - which keeps Dawn from using the entire process's budget.
Bug: dawn:193
Change-Id: I6c17bd703d7cb24759bcee89c03add46944fec8c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16383
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Track what heaps are required to be resident upon command list
submission.
Bug: dawn:193
Change-Id: Icefe5aed047140f401d51018937c5364b3ed7c4f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16382
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Instead of counting descriptors to be allocated for the entire command
buffer in a pre-pass, the bindgroup state tracker is used to allocate
only dirty bindgroups upon recording draw/dispatch. If the heap has no
more room and must be changed, bindgroups will be re-created according
to the BGL.
A future change will address the CPU descriptors and removal of the
pre-pass.
BUG=dawn:256,dawn:307
Change-Id: I6603de17cfda713bd4512c46e1c93618ca01bb7b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/13400
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
The state-tracking of the webgpu.h swapchain is a bit complicated
because contrary to implementation-based swapchains, they have more
guarantees and a "replacing mechanism". For example instead of hoping
the implementation-based swapchain resize automatically, the
surface-based swapchain needs to be replaced by a new swapchain and
invalidated.
This mechanism of invalidation also needs to be triggered when the last
reference to the surface is lost because we don't want to risk the
application destroying the window from under us.
Adds tests for all the cases of invalidation I could think of apart from
device loss.
Bug: dawn:269
Change-Id: Id515dbb640e13c6e30bb1f1e93b8e54f1e2bba4b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/15400
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This commit changes wgpu::Device::CreateSwapChain to take an additional
wgpu::Surface argument. Passing nullptr is enough to stay on the
previous swapchain implementation, until the new one is ready.
In order to support both the "old" implementation-based swapchains and
the "new" surface-based swapchains. SwapChainBase is now split into
three abstract classes:
- SwapChainBase that has a virtual method for each of the
wgpu::SwapChain methods.
- OldSwapChainBase that corresponds to the implementation-based
swapchains.
- NewSwapChainBase that will contain the surface-based swapchain
implementation and will eventually just be renamed to SwapChainBase.
The interaction of the surface-based swapchains with the Surface objects
aren't implemented yet, neither are the swapchain methods. Only creation
works.
Validation tests for surface-based swapchain creation are added in the
end2end test target because they need to create OS windows.
Bug: dawn:269
Change-Id: I7e07d6c666479867b9a16d7b1b8c181d5dbd69a0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/15281
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Handle DeviceLostCallback once DeviceLost error occurs.
Disallow any other commands or actions on device to happen after device
has been lost.
Bug: dawn:68
Change-Id: Icbbbadf278cae5e6213050d00439118789c863dc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12801
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
