This CL implements RequestDevice and also has changes for
Dawn to internally use wgpu::FeatureName enums, instead of
strings. Some of the string handling is kept for now to
support the deprecated creation path. GetFeatureInfo is added
to the instance to get a name and description of the feature,
for reporting in about://gpu.
Dawn device toggles are now passed in an extension struct off
of the device descriptor. This is only supported in dawn_native,
and not dawn_wire, for now, since dawn_wire doesn't have a way
to serialize lists of null-terminated const char*.
To enable the client to check whether the toggle descriptor is
supported, a `dawn-native` feature is added which is supported
all the time with dawn_native, but not supported with dawn_wire.
Feature `dawn-native` also enables a synchronous version of
CreateDevice for convenience.
Bug: dawn:160
Change-Id: Ifc195e7ea808c6c319021528ef4b36bd65583bff
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72020
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
This patch sets 'is_wire_transparent' on the structures whose members
are all wire transparent and are not pointers, so that we can use
memcpy when serializing and deserializing these structures:
- GPUBlendComponent
- GPUColor
- GPUExtent3D
- GPULimits
- GPUOrigin3D
- GPUStencilFaceState
- GPUVertexAttribute
- GPUBlendState
In the next patch we will support memcpy on the qualified structures
whose members contain pointers (e.g. GPUVertexBufferLayout).
BUG=chromium:1266727
Change-Id: If46289f2d10cc7b17e6f5330cd2c2d4dc481f8b9
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/73000
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch tries to use the data directly in deserialized buffer
when the data is "data-only" and doesn't affect the control flow instead
of allocating and copying into a temporary buffer. Due to the protection
of TOCTOU attacks, currently we only set "wire_is_data_only" on the
parameter "data" in Queue.WriteBuffer() and Queue.WriteTexture().
With this patch, the performance of dawn_perf_tests
BufferUploadPerf.Run/*_WriteBuffer_BufferSize_* with "-w" will
be greatly improved (~20%) when the upload buffer size is greater than
1MB.
BUG=chromium:1266727
Change-Id: I7a9d54c9b505975235ee37aa72ee97f082ad3aa3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72063
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch removes the support of annotation "const*const*" in
the template WireCmd.cpp as it is not used in current Dawn wire
implementation.
BUG=chromium:1266727
Change-Id: I4c4d68ccc050edb5f855094910440e90aef2bc4a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72040
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch uses memcpy to copy the arrays in basic types instead
of iterating every elements in a for-loop.
In the next step we will copy specific structures with memcpy.
With this patch, the performance of dawn_perf_tests
BufferUploadPerf.Run/*_WriteBuffer_BufferSize_* with "-w" will
be greatly improved (~30%).
BUG=chromium:1266727
Change-Id: I7c6fb0fafa63bd6b602eeef8cf2c0ae0cfc7b0be
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71180
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This adds overloads for WireCmd [De]serialization that don't take the
object id provider/resolvers and produce a fatal error as soon as an
object is encountered.
Bug: dawn:1186
Change-Id: I13e796a5d8f59c26279b9079d4496390506c739a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/68941
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
- Puts all the macros first, then all the instantiation of macros for
more clarity of what code actually gets generated.
- Move ErrorObjectProvider and friend in the anonymous namespace.
- Give a name to a boolean parameter.
Bug: dawn:1186
Change-Id: I2662ba16bfff23b53342d352d9b7fbd62aabb8c3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/68940
Commit-Queue: Loko Kung <lokokung@google.com>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Adds a way to store the limits on the Adapter and the
Device. For now, adapter limits are always the default
limits, and device limits are stored but not used.
This CL also adds usage of an ErrorObjectIdResolver and
Provider in the WGPUDeviceProperties serialization and
deserialization helpers. Serializing/deserializing this
struct should never have objects.
Bug: dawn:685
Change-Id: I1479b4407b0f9ec9f9b2bff62cad7caa693c99d7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/63983
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Chained structs *may* contain objects which means
deserialization may need an ObjectIdResolver.
However, in practice, we never need to send chained
structs from the server to the client that contain objects for
a valid command.
The one upcoming need for chained server->client structs is to
serialize limit structs.
Because limit structs never need objects, we provide a dummy
implementation of the ObjectIdResolver which always yields an error.
An analogous change is done for ObjectIdProvider.
These classes will be used in a follow-up CL.
Bug: dawn:685
Change-Id: I1c0f3f2d080377f2e1a77bc6e896f24d3d9ab931
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/63981
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This didn't make sense since these members can be specified to have no
data by setting the length to 0.
- Prevent uses of this patten by adding an assert in WireCmd.cpp's
generator.
- Fix SetBindGroup dynamicOffset to not be optional but default to
nullptr instead.
This issues would cause a read of uninitialized pointers becaus the
generator code looked like this:
SetBindGroupCmd cmd;
cmd.dynamicOffsetCount = record.dynamicOffsetCount; // 1
bool has_dynamicOffset = record.has_dynamicOffsetl // false
if (has_dynamicOffset) {
cmd.dynamicOffsets = ...;
}
// Oh no! dynamicOffsets contains garbage even if dynamicOffsetCount
// is set to 1. dawn_native will happily read it.
Bug: chromium:1220036
Change-Id: I5c468b639f671cef3be2fa64667a0bf114fc902b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/54643
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Bug: dawn:680
Change-Id: I6d57280ab11381649deef51ee7babf5ca73f359b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/42340
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
This makes the primitives in the serialized wire protocol
the same across platforms and architectures which is better
for both fuzzing and remoting Dawn.
Commands that used size_t are updated to use uint64_t, and
the server-side implementation checks if conversion to
size_t would narrow.
Bug: dawn:680
Change-Id: Icef9dc11a72699685ed7191c34d6a922b652c887
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/41582
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
- Encapsulate deserialize buffer and size into a DeserializeBuffer
class. This limits the possible operations so we can be sure
buffer/size are not manually mutated such that we consume more
bytes than available.
- Ensure that memberLength (on deserialization) doesn't narrow (or
widen). Previously, values were always implicitly cast to size_t.
- Slight optimization that removes "= nullptr" initialization for
pointers written by DeserializeBuffer::Read. These
are always written to on success, so we don't need to initialize
to nullptr.
Bug: dawn:680
Change-Id: I3779a343e85ff90810707148a952c6ba27cf9d22
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/41521
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
String deserialization requires one more byte than the length
computed from strlen. This CL guards against the case when
length+1 overflows.
Also include a slight optimization to memcpy string contents
instead of using std::copy. It's safe to cast away volatile
qualifiers here since the string data doesn't affect control flow.
Bug: dawn:680
Change-Id: Icaf319ea6c5aedcf0c33d17a0ea7c253f4f249e1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/41800
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
BufferConsumer wraps a buffer pointer and size and exposes a
limited number of operations to get data while decrementing
the remaining available size. This makes it so that code
reading or writing into a buffer cannot easily consume more
bytes than available.
This CL guards against serialization overflows using
BufferConsumer, and it implements GetPtrFromBuffer
(for deserialization) on top of BufferConsumer. A future patch
will make the rest of the deserialization code use BufferConsumer.
Bug: dawn:680
Change-Id: Ic2bd6e7039e83ce70307c2ff47aaca9891c16d91
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/41780
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
This allows deserialization to fail if the buffer is not large enough.
Before, we simply assumed the buffer was at least the size of
WGPUDeviceProperties.
Bug: none
Change-Id: I24e1f84c583f48d4e32c35276e5508e257e9f530
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/39861
Auto-Submit: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This CL makes the wire gracefully handle all invalid and unknown
sTypes. All unknown sType structs are serialized and deserialized
as the base WGPUChainedStruct with sType Invalid.
Bug: dawn:369, dawn:654
Change-Id: Ia2571df81fc96e2c672d3ea13c03237a2d5fa5c1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/39460
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This CL adds support for chunking large commands by first serializing
large commands first into a separate buffer, and then sending the
buffer data chunk by chunk.
This code path is used for large writeBuffer and writeTexture, as well
as the inline memory transfer service for buffer mapping. The transfer
for writeBuffer and writeTexture will be optimized further in Chrome,
and the inline memory transfer service is currently used only in tests.
Bug: chromium:1123861, chromium:951558
Change-Id: I02491a44e653e2383174958d9c3d4a4db6fd7bde
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28882
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
The new warnings are:
- -Wdeprecated-copy
- -Winvalid-offsetof
- -Wpessimizing-move
And the list of warnings was sorted alphabetically.
Bug: chromium:1072449
Change-Id: I9f3eecae645455c481ecc2e0be4df350e1453907
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/20381
Reviewed-by: Zhenyao Mo <zmo@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This is a more accurate name and conflicts less with the callback
request serials.
Bug: none
Change-Id: I0f9660c24468064dadffb3ab9b3392d403f93c41
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19260
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This CL also adds a couple of dummy extensions in dawn.json so that
the serialization/deserialization in the wire can be tested.
Bug: dawn:369
Change-Id: I5ec3853c286f45d9b04e8bf9d04ebd9176dc917b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18520
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
webgpu.h is the "official" header for WebGPU in native and is being
developed in https://github.com/webgpu-native/webgpu-headers
dawn.h and dawncpp.h are changed to become webgpu.h and webgpu_cpp.h
respectively and use the new naming convention. New dawn.h and dawncpp.h
headers are created that just proxy the types, constants and functions
to their WebGPU counterpart.
Almost no naming change is done in Dawn in this commit, which help check
that the proxying headers work correctly. A couple changes were
necessary, in particular for tests of the internal of dawncpp.h, and a
workaround for a standard library bug for std::underlying_type was
removed because it is no longer needed and got in the way.
Finally since some templates were renamed to match the name of the file
they create instead of using the generic "api" name.
BUG=dawn:22
Change-Id: I12ee22d0b02ccb5b8a52ceccabb3e63ce74da007
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12480
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch implements the serialization and deserialization of
DawnDeviceProperties in dawn_wire for the use of serializing this type
of object in Chromium.
BUG=chromium:996713
TEST=dawn_unittests
Change-Id: I1678627a017079540689d8529a1a7e1c975aae61
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12240
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This prevents bugs where the compiler assumes a piece of memory
will be the same if read from twice.
Bug: dawn:230
Change-Id: Ib3358e56b6cf8f1fbf449c5d564ef85c969d695b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11840
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
We were generating has_member for all pointer members which meant
it was possible to incorrectly pass nullptr for required arguments.
Bug: chromium:1009166
Change-Id: I5e32bab5ccae010468e55897ae9840ab8fda684f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11760
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Optional labels were added to command buffer descriptor and compute pass
descriptor in https://dawn-review.googlesource.com/c/dawn/+/11601.
Sadly, the wire was not updated. This CL fixes support for optional
const char*.
Bug: dawn:22
Change-Id: I23211456f5e3d5f3c0344d8b8cc604bab20015f6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11660
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This is to match the work in progress webgpu.h header.
Also contains a fix for the wire where it wouldn't GetExtraRequiredSize
for structures that are by-value members of other structures.
BUG=dawn:22
Change-Id: I3c706bf9cd7a550d40fd667877f032c860d0a032
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9382
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This patch adds MemoryTransfer client/server interfaces and
uses it to implement data transfers for buffer mapping.
This patch also provides a default "inline" implementation of
the MemoryTransfer which is used if the embedder does not
provide one on initialization.
Because implementations of MemoryTransfer perform their own
serialization, a skip_serialize option is added to WireCmd records.
Bug: dawn:156
Change-Id: I2fa035517628a3ad465b0bc18a6ffc477e2bd67f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8642
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This requires some changes in the wire because WireCmd doesn't know the
size of void. This also adds a handwritten implementation of the Wire
commands for SetSubData that internally converts to uint8_t so that
WireCmd can generate the de/serialization.
BUG=dawn:160
Change-Id: Icbf0fd7dd841639ee6f67333844e027b27a8afcc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7780
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This is the first command to return a struct. This patch also
updates the code generator to support structure return values.
Bug: dawn:7
Change-Id: Ie8acec895c0ec88429672138ffc900032fbbc447
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/4780
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This removes blocks of code that were obviously builder-specific but
also removes the ObjectStorage::valid member that was used to implement
the maybe monad on the wire server side. This is no longer needed since
dawn_native handles the maybe monad internally now.
BUG=dawn:125
Change-Id: I8c30daae9fc70853bc1996d85a860b4877c5976c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6161
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This will allow implementing the following part of the WebGPU IDL using
a nullable pointer instead of an extra hasDepthStencilState boolean:
partial interface GPURenderPipelineDescriptor {
GPUDepthStencilStateDescriptor? depthStencilState;
};
BUG=dawn:102
Change-Id: Iae709831ad857fcef073f18753ab39567a8797da
Reviewed-on: https://dawn-review.googlesource.com/c/4500
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Yunchao He <yunchao.he@intel.com>
Unify code generation for Client->Server and Server->Client commands.
Methods in dawn.json are converted into command records and additional
commands are specified in dawn_wire.json. This can then be used to
completely generate the command handlers and command struct definitions.
Bug: dawn:88
Change-Id: Ic796796ede0aafe02e14f1f96790324dad92f4c0
Reviewed-on: https://dawn-review.googlesource.com/c/3800
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Without this the serialize step failed to increment the pointer to
the buffer to account for the recording of "main" in the render
pipeline descriptor:
descriptor->vertexStage->entryPoint = "main";
BUG=dawn:4
Change-Id: Ic3829787af2213577ac67ee93625679afdae1cc9
Reviewed-on: https://dawn-review.googlesource.com/c/2880
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
In preparation for the descriptorization of BindGroup, support was added
to treat wire ID 0 as nullptr for a bunch of objects. Now that we have a
fuzzer for the wire+frontend, we need to validate when we have a 0 id.
Either the wire needs to reject the ID or the frontend needs to validate
against nullptrs. Since only few entrypoints will have a use for
nullptrs (bind groups, render pass resolve textures), we require an
opt-in in the JSON file for a structure member or an argument to be
optional.
This disables the tests related to ID 0 = nullptr, because we don't yet
have optional argument/members in dawn.json.
BUG=chromium:905273
BUG=chromium:906418
BUG=chromium:908678
Change-Id: If9a3c4857db43ca26a90abff2437e1cebb0ab79b
Reviewed-on: https://dawn-review.googlesource.com/c/2704
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
BUG=dawn:13
Change-Id: I7a224503d0a33ef148e63b8327a6a53df1b3868a
Reviewed-on: https://dawn-review.googlesource.com/1520
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Also moves the TerribleCommandBuffer to utils:: because it isn't part of
the implementation of the wire, renames dawn::wire to dawn_wire, moves
src/wire to src/dawn_wire and puts the interface of dawn_wire in
src/include/dawn_wire.
Austin Eng
Jiawei Shao
Corentin Wallez
Ben Clayton
Zhenyao Mo
François Beaufort
Yunchao He
Stephen White