Every render pass which invokes DrawIndexedIndirect, either directly or
through a RenderBundle execution, is now preceded immediately by at
least one validation pass.
All indirect buffer offests used with DII are validated, and their
validated values are copied into a separate scratch buffer (or zeroed
out there, in the case of validation failure). All encoded DII commands
are rewritten to use the validated parameters instead of the original
ones.
Bug: dawn:809
Change-Id: I5eead937f19536f84f89e2c8e6fed7f18f0aee9f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/63461
Commit-Queue: Ken Rockot <rockot@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Note that this is for direct dispatch calls only. Indirect dispatch
calls are still not validated.
Bug: dawn:1006
Change-Id: I061c15208a01dfb803923823ba4afd38667cad22
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/59122
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Ken Rockot <rockot@google.com>
Below are the list of all the individual changes, which are a good order
in which to review this CL.
Core changes:
- Change the tracking in the frontend to produce a synchronization
scope per dispatch instead of per compute pass. Some bindgroups might
not be part of any synchronization scopes so we also track all the
referenced resources on the side so they can be checked during
Queue::Submit validation.
- Fix clearing in the GL and Metal backends to use the per-dispatch
synchronization scopes.
- Fix the Vulkan backend to use the per dispatch synchronization scopes
to produce the correct pipeline barriers. This allows the removal of
previous logic that was subtly incorrect for Indirect buffer. This
allows the merging of the Compute and Render DescriptorSetTracker into
a single small helper class.
- D3D12 changes are similar to Vulkan, but the simplification is just a
the suppression of a branch with a lot of code in
BindGroupStateTracker.
Test changes:
- Fixup all the ResourceUsageTracking tests to follow the WebGPU spec
for synchronization scopes (fixing a lot of TODOs).
- Add additional tests checking that Indirect buffers are not allowed
to be used as a writeable storage in the same synchronization scope.
- Add tests for Queue::Submit validation correctly taking into account
resources that are bound but unused in compute passes.
- Add an end2end test for using a buffer as Indirect and Storage at the
same time in a DispatchIndirect, which would previously produce
incorrect barriers in the Vulkan and D3D12 backends.
Other small changes (that I was to lazy to put in a different CL):
- Add the utils::MakePipelineLayout helper function.
- Fix Indirect not being in the list of readonly buffer usages (caught
by a test added in this CL).
Bug: dawn:632
Change-Id: I77263c3535a4ba995faccbf26255da9a2f6ed3b5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/49887
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
This workaround was only necessary in the Metal backend but it was easy
to put it in the frontend so it was put there. However, this simplification
gets in the way of a validation change in the follow-up CL. So we move
the nooping to the Metal backend.
Bug: dawn:632
Change-Id: I0a9957a1bc9bdd7867eb8c60fcab710832c13af2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/49886
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This change is a preparation for making the compute pass track the
synchronization scope usages per dispatch instead of for the whole pass.
This CL just separates the Compute and RenderPassResourceUsage types.
This requires making the difference between
SyncScope/ComputePass/RenderPass ResourceUsageTracker instead of having
a single combined tracker.
This change also duplicates SetBindGroup by removing the common handling
in ProgrammablePassEncoder and putting it in ComputePassEncoder and
RenderEncoderBase. This is necessary because the UsageTracker types are
now split, but it will also help have different handling of SetBindGroup
for compute and render in follow-up CLs.
There are no functional changes.
Bug: dawn:632
Change-Id: I482c04483d8b734fb10e44e717071eedcff2f15f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/49884
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This change is a preparation for making the compute pass track the
synchronization scope usages per dispatch instead of for the whole pass.
- Split the tracking of render and compute passes usages.
- Rename PassTextureUsage to TextureSubresourceUsage since is it not
per-pass
- Add SyncScopeResourceUsage as a subclass of PassResourceUsage to
start modifying some of the code to work with synchronization scopes
(even if syncscope == pass at the moment).
There are no functional changes.
Bug: dawn:632
Change-Id: Ieeb6d70a44dc1c726f26989eebcd87e63e732785
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/49883
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This means that calling wgpu::Object::DoStuff will translate to a call
to dawn_native::ObjectBase::APIDoStuff. This will clarify the
difference between reentrant calls and internal calls in dawn_native.
Avoiding issues in the future.
This CL only changes the code generator to prefix with "API", performs
renames needed to make the code compile, and adds TODOs for things that
should be fixed in follow-up CLs.
Bug: dawn:723
Change-Id: Ie24471fa093adc4179d33d13323429847d076ecb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/45921
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Workaround for Metal becuase system crashes on 0 dispatches.
Bug: dawn:640
Change-Id: I5bd33b22242ddc31816a16acb019ce2f552808bb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/39521
Commit-Queue: Hao Li <hao.x.li@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This is the last piece of validation that was done in a separate
validation pass so the validation pass is removed.
Bug: dawn:635
Change-Id: I91ce5d5a512ac188f3dd56c90db9e69aee518844
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38845
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This also adds missing coverage for push/pop debug group in render
bundles. The RenderBundleEncoder didn't validate itself on Finish, so
add a regression test for that too.
The overarching goal with this CL is to do validation at encoding time
which will help produce SyncScopeResourceUsage in the frontend for
dispatch() calls so that they can be reused by the backends.
Bug: dawn:635
Change-Id: Ie5a2d987fda3854b3145ba4b7a34994ea605e820
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38842
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Also cache the boolean for IsValidationEnabled in the
ProgrammablePassEncoder.
Bug: dawn:635
Change-Id: I5f095d003d28b84eacbc488a7f8f3c346c646187
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38521
Commit-Queue: Austin Eng <enga@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
In a future CL the PassResourceUsage structure will become a
SyncScopeResourceUsage and will be used to validate at each
synchronization scope. For separation of concerns, the validation that
resource have the correct usage shouldn't be done at the sync scope
level but at each entrypoint that uses the resource.
The validation tests had no coverage of usage validation for pass usage
so validation tests are added for Indirct/Index/Vertex usages. (Uniform
and Storage are validated at bindgroup creation and already had
validation tests)
Bug: dawn:635
Change-Id: I5058ad30eb041809f0f60d9403f3cc2d5d7e7c96
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38380
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
The same query cannot be written twice in same render pass, so each
render pass also need to have its own query availability map.
Update timestamp query to only check the same query overwrite in same
render pass.
Bug: dawn:434
Change-Id: Icb070adf79a3d76c25367675f7432666eb0dd84f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/31180
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
Some APIs exposed by Dawn are not expected to be fully secured until
after the first Origin Trial of WebGPU. To prevent their usage we add a
new toggle that will be set by default by Chromium. This toggle throws a
validation error when an unsafe API is used.
Bug: chromium:1138528
Change-Id: I831db70bdac5128ebc32d36d55a0eaefc42c1807
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/31443
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
- There are some methods need to know which query indexes are used in
encoders: endOcclusionQuery, endPipelineStatisticsQuery, resolveQuerySet.
- On Vulkan, we also need to use the used query indexes to reset each
queries between uses. And because the reset command must be called
outside render pass, we need to check whether a query index is writen
twice on command encoder and render/compute encoders.
- Add validation on writeTimestamp for duplicate writes at same index.
Bug: dawn:434
Change-Id: I6d00dd91e565d960246b6d01ad434d2d5c095deb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27561
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
This patch enables one texture to be used as both write-only storage
texture and sampled texture in one compute pass.
Note that while we don't need to check the usage scope of a texture in
the whole compute pass scope, we still need to verify one texture cannot
be bound to multiple bindings that are used in one dispatch at the same
time. This check will be added in the following patches.
This patch also adds tests to ensure a texture can be used as the
following binding types in one compute pass:
- read-only storage + sampled
- write-only storage + sampled
- read-only storage + write-only storage
BUG=dawn:267
TEST=dawn_unittests
Change-Id: Ibff2b005a5269a0bfa254e0417de4920758add39
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/20120
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This patch factors resource usage tracking so it is separate from
command validation, allowing the bulk of command validation to be
completely skipped.
In DrawCallPerfRun/Vulkan_DynamicPipeline_DynamicBindGroup, disabling
validation cuts roughly 74 nanoseconds (20%) of CPU time, per draw.
In DrawCallPerfRun/Vulkan_DynamicBindGroup, disabling validation
cuts roughly 35 nanoseconds (17%) of CPU time, per draw.
In DrawCallPerfRun/Vulkan_MultipleBindGroups, disabling validation
cuts roughly 45 nanoseconds (14%) of CPU time, per draw.
Bug: dawn:271
Change-Id: I517b85840ba18c6a554b83f34a1d0aef1a8c56a6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/13520
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This was unnecessary verbosity. Fix this by having the ProcTable
generator using type aliases so all types appear like they have
"Base".
BUG=
Change-Id: I8c472fb924f6ce739e4e41038452381b4f727a2b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/13442
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch factors the CommandAllocator, CommandIterator, and error
handling out of CommandEncoderBase so it can later be used by the
RenderBundleEncoder.
Bug: dawn:154
Change-Id: Ia4f8c3ce7f432f0887b619bd8090aa9bec7330fc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9181
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This is needed so that RenderBundleEncoder and RenderPassEncoder can
share code. This patch also moves EndPass out of ProgrammablePassEncoder
and into both RenderPassEncoder and ComputePassEncoder. Render bundles
do not have EndPass.
Bug: dawn:154
Change-Id: Ib7126b2ba718b0b93e3d6f15c429ac910c0d5d31
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9180
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This is a reland of 7eb6be186b
Fixes casting issue on 32 bit machines
Original change's description:
> dawn_native: Indirect draw/dispatch
>
> Adds indirect draw and dispatch for all backends (without validation).
>
> Tests for opengl negative offset are skipped since there is no easy
> way to add the index buffer offset. Current idea is to use a compute
> shader to modify the indirect draw buffer.
>
> Change-Id: I1d3eec7c699b211423f4b911769cca17bfbcd045
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7481
> Commit-Queue: Idan Raiter <idanr@google.com>
> Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Change-Id: Id28c5658ee18ec5c030f721fb44d9f11ebe21ff9
Bug:dawn:54,chromium:972358
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7961
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Idan Raiter <idanr@google.com>
This reverts commit 7eb6be186b.
Reason for revert: This change is breaking Clusterfuzz bots. crbug.com/972358
Bug: chromium:972358
Original change's description:
> dawn_native: Indirect draw/dispatch
>
> Adds indirect draw and dispatch for all backends (without validation).
>
> Tests for opengl negative offset are skipped since there is no easy
> way to add the index buffer offset. Current idea is to use a compute
> shader to modify the indirect draw buffer.
>
> Change-Id: I1d3eec7c699b211423f4b911769cca17bfbcd045
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7481
> Commit-Queue: Idan Raiter <idanr@google.com>
> Reviewed-by: Kai Ninomiya <kainino@chromium.org>
TBR=cwallez@chromium.org,kainino@chromium.org,idanr@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I9b7b63de0900e20c4da8561ceb9b2dece77c3885
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7960
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Adds indirect draw and dispatch for all backends (without validation).
Tests for opengl negative offset are skipped since there is no easy
way to add the index buffer offset. Current idea is to use a compute
shader to modify the indirect draw buffer.
Change-Id: I1d3eec7c699b211423f4b911769cca17bfbcd045
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7481
Commit-Queue: Idan Raiter <idanr@google.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch is a follow-up of the descriptorization of render pass
descriptor. In this patch we changes the return value of
BeginComputePass from nullptr to an error compute pass encoder when
there is any error in BeginComputePass() to keep it consistent with what
we do in BeginRenderPass().
This patch also provides functions to create error render/compute pass
encoders. With this patch we can create a pass encoder in error by
specifying ErrorTag in the constructor, which is more staightforward
and human readable than the current implementation.
BUG=dawn:6
Change-Id: I1899ae65804f8cecd3079dc313e7e18acb88e37c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5140
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This removes the CommandBufferBuilder and copies all the logic into
CommandEncoderBase instead. No changes were done to the logic except for
the implementation of CommandEncoderBase::HandleError and Finish.
BUG=dawn:8
Change-Id: I7b6f44c3cf501477422f067bd277cef470073860
Reviewed-on: https://dawn-review.googlesource.com/c/4820
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Dawn used to return "nullptr" when an error happened while creating an
object. To match WebGPU we want to return valid pointers but to "error"
objects.
This commit implements WebGPU error handling for all "descriptorized"
objects and changes the nullptr error checks into "ValidateObject"
checks. This method is used both to check that the object isn't an
error, but also that all objects in a function call are from the same
device.
New validation is added to objects with methods (apart from Device) so
they check they aren't error objects.
A large number of ASSERTs were added to check that frontend objects
aren't accessed when they are errors, so that missing validation would
hit the asserts instead of crashing randomly.
The bind group validation tests were modified to test the behavior with
both nullptrs and error objects.
Future commits will change the CommandBufferBuilder to not be a builder
anymore but an "Encoder" instead with special-cased error handling. Then
once all objects are descriptorized, the notion of builders will be
removed from the code.
BUG=dawn:8
Change-Id: I8647712d5de3deb0e99e3bc58f34496f67710edd
Reviewed-on: https://dawn-review.googlesource.com/c/4360
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
To match the WebGPU idl, SetRenderPipeline and SetComputePipeline need to change
the name to SetPipeline.
This patch just rename these two APIs
BUG=dawn:52
Change-Id: I2545c94158ebd051027a018650bf760a25845d73
Reviewed-on: https://dawn-review.googlesource.com/c/3480
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
The fuzzer is able to trigger nullptr reads by failing to create objects
and then using the resulting nullptr in other operations. The proper fix
is to implement WebGPU error handling where creation failure returns a
valid but "error" object.
However implementing this error handling is a lot of work, so in the
meantime we use nullptr checks in relevant places to fix the fuzzer
issue. These checks will be removed once the error handling is changed.
BUG=dawn:8
Change-Id: I6777a7fa40383b3d2235e071c3f0109de7605a22
Reviewed-on: https://dawn-review.googlesource.com/c/2565
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This splits off part of CommandBufferBuilder in separate
RenderPassEncoder and ComputePassEncoder objects. To match the WebGPU
IDL and factor some code, both these encoders inherit from
ProgrammablePassEncoder.
These encoders are pure frontend objects and record into the
CommandBufferBuilder command allocator objects, so no changes to the
backends were needed.
Error handling is still ew, because the "builder" mechanism we had
doesn't allow for "split builders". Nicer error handling will have to
wait on Dawn matching WebGPU.
All the tests and samples were updated to the new structure.
BUG=dawn:5
Change-Id: I5f5d4ad866e2c07fedd1ba7a122258c6610941f1
Reviewed-on: https://dawn-review.googlesource.com/1543
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Ken Rockot
Corentin Wallez
Hao Li
Enrico Galli
Jiawei Shao
Austin Eng
Idan Raiter
Yan, Shaobo