180 Commits

Author SHA1 Message Date
Austin Eng
cef68bc8b7 dawn_wire: Fix a bug with multiple injected devices
Device child objects were storing an *unstable* pointer to device
specific tracking information. Fix this by moving the tracking
information to a stable heap allocation.

Bug: dawn:565
Change-Id: I00ad72563ac66e29314603e77698718953fcbf15
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38280
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2021-01-21 23:15:43 +00:00
Austin Eng
8bcde8e394 dawn_wire: Add Reserve/InjectDevice
Now that the wire does enough tracking to prevent a malicious client
from freeing a device before its child objects, and the device is no
longer a "special" object with regard to reference/release, it is
safe to support multiple devices on the wire. The simplest way to
use this in WebGPU (to fix createReadyRenderPipeline validation)
is to add a reserve/inject device API similar to the one we use for
swapchain textures.

Bug: dawn:565
Change-Id: Ie956aff528c5610c9ecc5c189dab2d22185cb572
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/37800
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2021-01-19 19:27:52 +00:00
Carl Woffenden
1c84a6b160 Missing header for VS1029 (for commit 8c58491d)
This recently broke in VS2019 (weak_ptr and shared_ptr aren't included).

Change-Id: I88edffa3e965c02d0db83e2efd0977cf86c96800
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38023
Commit-Queue: Carl Woffenden <cwoffenden@gmail.com>
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
2021-01-15 18:44:19 +00:00
Austin Eng
8c58491d25 dawn_wire: Skip device inject error if the client drops the device
If the client drops the last reference to the device, it would
dereference an invalid pointer upon calling InjectError. So, skip the
call. We can't keep the device alive if the Buffer is still alive
because we intend to make all objects internally null if you delete
their device.

It is ok to skip error injection because if the client deletes the
device, it should not expect to receive any more error callbacks.

Bug: dawn:384
Change-Id: I4c694310e4395b06cd49603fc5d4cd846799decb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/37580
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2021-01-14 00:51:58 +00:00
Austin Eng
8ba0a01d1e Ensure all wire child objects are destroyed before their device
Destroying a device will implicit destroy all its child objects.
Attempting to use a child object after results in a fatal error.

Bug: dawn:384
Change-Id: I43c27c92cacde759be83cca79ac890f41bac3927
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/37002
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
2021-01-13 20:58:18 +00:00
Austin Eng
5ad5557667 Remove special-casing of device reference/release in the wire
The wire's device is externally owned so reference/release were no-ops.
To unify the code paths, remove the special casing and instead
take an extra ref on the device the wire server is created with. This
is functionally equivalent and will allow both the current wire code,
and the incoming change to allow multiple device/adapter creation to
both work.

This CL also makes it possible for the client to destroy the device
before child objects.
A follow-up CL will mitigate this on the server side.

Bug: dawn:384
Change-Id: Ic5427074469012dccf8689ec95a848e6ba2c1fc2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/37001
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2021-01-13 19:23:48 +00:00
Austin Eng
f0d7cc4f5a dawn_wire: Make all objects owned by the client
This removes the logic where the Client owns the Device and the
Device owns all other objects. Ownership should be tracked in
dawn_native either with refcounting or validation to disallow
operations after an object's parent has been destroyed.

This simplifies the wire client code in that the client only
tracks allocated handles and does not manage parent/child lifetimes.
This is an important simplification so we can support multiple WebGPU
instances, adapters, and devices on a single wire.

Bug: dawn:384
Change-Id: I8ecc7c368130b8917202150c467b5f0e7d4b753e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/37000
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2021-01-13 18:31:47 +00:00
Carl Woffenden
22eff1eb8c Fixes to build with MSVC 2019
When building with MSVC (using 'is_clang=false') various errors are
encountered. This fixes:

- multiple tests/end2end/ColorStateTests.cpp(309): error C2039:
  'transform': is not a member of 'std'

- dawn_wire/server/Server.h(171): error C2248:
  'dawn_wire::server::MapUserdata::MapUserdata': cannot access private
  member declared in class 'dawn_wire::server::MapUserdata'

- variants of the above in Server.h for all the other CallbackUserdata
  subclasses

Bug: dawn:602

Change-Id: I01b1db3fbf67ae170b21a7d890af6162b82a1725
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/37420
Commit-Queue: Carl Woffenden <cwoffenden@gmail.com>
Reviewed-by: Austin Eng <enga@chromium.org>
2021-01-12 19:47:13 +00:00
Carl Woffenden
e650b38aa6 Minor fix to build with MSVC: Gets content address instead of iterator
Fixes error C2679: binary '-': no operator found which takes a right-hand operand of type 'std::_Array_iterator<_Ty,23>' (or there is no acceptable conversion)

Change-Id: I566b05032da5f68674156eae2901a3f3e7a8acbb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/36082
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2021-01-11 20:02:21 +00:00
Austin Eng
200941c797 dawn_wire: Return early in callbacks after the server is destroyed
After the server is destroyed, the server's can't do anything like
forward callbacks to the client. Track this with a weak_ptr and
return early if it has expired.

It also updates device destruction in dawn_native so the lost
callback is always called, even on graceful destruction. This
is consistent with the rest of WebGPU where all callbacks are
guaranteed to be called in finite time.

Bug: chromium:1147416, chromium:1161943
Change-Id: Ib80dea36517401a2b8eafb01ded255ebbe757aef
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/35840
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
2021-01-05 08:37:08 +00:00
Austin Eng
e3fd026108 Use a common helper for std::nothrow
It's come up multiple times that ASAN doesn't support
std::nothrow which leads to OOM bugs filed by the fuzzers.
Use a common helper to avoid this and return nullptr for large
allocations when ASAN is enabled.

Bug: none
Change-Id: I492b4ff4e498cf82d4ca08ba849671d3d16b9cfb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/36280
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2021-01-05 07:40:48 +00:00
Stephan Hartmann
0b0e6e0579 use cstring instead of string.h
dawn codebase uses c-prefixed headers.
Follow up for https://dawn-review.googlesource.com/c/dawn/+/35001

Bug: None
Change-Id: I65b18f3ec8b2222f9a683c2cf81b710435dcf55c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/35017
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2020-12-08 14:14:53 +00:00
Stephan Hartmann
d63d562d1b IWYU: include string.h for memcpy
Bug: None
Change-Id: I978f49febf7e9e9e961957f62efd3ff9a166b81b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/35001
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-12-05 20:37:01 +00:00
Ivan Murashov
75a1f5234e Remove storage class specifier for the explicit template specialization
According to the http://www.eel.is/c++draft/temp.expl.spec:
An explicit specialization shall not use a storage-class-specifier
other than thread_local.
Clang doesn't claims about it, but GCC does.
An error example for GCC 8.4.0:
gen/third_party/dawn/src/dawn_wire/client/ApiObjects_autogen.h:25:5:
error: explicit template specialization cannot have a storage class

Bug: dawn:384
Change-Id: Iaf86722a943d19c9796a7f112885666ac88f20ca
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/33480
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2020-11-20 09:38:56 +00:00
Austin Eng
f2bc3b3edd Print warning if errors occur and no error callback has been set
Bug: none
Change-Id: Ibbd6f90be5192ad3975bb8b24f9b3053f7f9d74f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/32744
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2020-11-16 23:32:36 +00:00
Austin Eng
bc6a700bf5 dawn_wire: Protect against large allocations in ChunkedCommandHandler
Use std::nothrow on the allocation to catch failed allocations. Enforce
a max allocation limit a bit lower than ASAN's max 2GB allocation.

Bug: chromium:1145204
Change-Id: I91f2ddd5b58da6c39d4ab8bc447f7d9b7af8615f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/32340
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-11-13 01:49:12 +00:00
Austin Eng
7ceffe8511 dawn_wire: Reject new callbacks if the client is disconnected
If the wire client is disconnected, it will not receive any
messages from the server. Reject all callbacks that are created.

Bug: dawn:556
Change-Id: I2eb2c449b1ca6c8ea3e74040ef095abfc46a9061
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/31161
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
2020-11-12 18:03:42 +00:00
Austin Eng
b70a5b02e9 Reject all callbacks with DeviceLost on wire client disconnect
When the wire is disconnected, the client will not receive any
messages from the server. We need to manually reject all callbacks.

Bug: dawn:556
Change-Id: Ia03456b3209dbe0e1e54543d344180d11d4c6f1e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/31162
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
2020-11-11 21:01:18 +00:00
Austin Eng
3120d5ea0d Track and destroy all child objects on wire client destruction
This is needed so that:
 1. We can support multiple devices in the wire. The device will need
    to know how to destroy its child objects.
 2. The wire needs to be aware of all objects and their in-flight
    callbacks so that it can reject them if the wire is disconnnected.
    A future change will handle this.
 3. Fix leaks of objects on page teardown. When the page is torn down,
    the wire client is destroyed, and we skip calling release() for all
    objects since the object holding the proc table was also destroyed.

Bug: dawn:384, dawn:556

Change-Id: Ie23afe4e515b02e924fcfc2db92b749fd2257c9c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/31160
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-11-11 19:46:18 +00:00
Corentin Wallez
18c7b0e0b1 Add the -Wdeprecated-copy-dtor warning.
This should fix the warning triggering when using Dawn in Skia.

Bug: None
Change-Id: I045ebc87f9e8dbff035920fc6eaa409c2b70d0f2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30701
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-11-03 14:24:05 +00:00
Jiawei Shao
42103bc2e9 Fix crash when device is removed before CreateReady*Pipeline callback
This patch fixes a crash issue when the device is destroyed before
the callback of CreateReady{Render, Compute}Pipeline is called. Now
when the callback is called in DeviceBase::ShutDown(), the cached
pipeline object will also be destroyed before the callback returns.

BUG=dawn:529
TEST=dawn_end2end_tests

Change-Id: I91ec2608b53591d265c0648f5c02daf7fadac85e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30744
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
2020-10-24 03:11:43 +00:00
Jiawei Shao
03e1400fce Add the entry point of CreateReadyRenderPipeline
BUG=dawn:529
TEST=dawn_end2end_tests

Change-Id: I42ac0edc77e5b6119eb374da72698fca14596f7b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30540
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-10-21 04:37:41 +00:00
dan sinclair
e9c99e93e7 Add algorithm include.
This CL adds the algorithm include to, hopefully, satisfy the MSVC
build.

Change-Id: Ic0a8a9dfeffe7a02d94d748dbb2813ac576b5371
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30660
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
2020-10-20 20:30:11 +00:00
Jose Dapena Paz
1917dc27c2 IWYU: unique_ptr usage in ChunkedCommandSerializer requires including <memory>
Fix build breakage in libstdc++:
../../third_party/dawn/src/dawn_wire/ChunkedCommandSerializer.h:84:34: error: ‘unique_ptr’ is not a member of ‘std’
   84 |             auto cmdSpace = std::unique_ptr<char[]>(new (std::nothrow) char[requiredSize]);
      |                                  ^~~~~~~~~~
../../third_party/dawn/src/dawn_wire/ChunkedCommandSerializer.h:24:1: note: ‘std::unique_ptr’ is defined in header ‘<memory>’; did you forget to ‘#include <memory>’?

Bug: chromium:957519
Change-Id: I35b129225882a3f9b758c4f49d1a3cace22aacfe
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30600
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
2020-10-20 19:48:31 +00:00
Austin Eng
d676f44272 Reland "Replace the wire serializer with a no-op impl on disconnect"
This is a reland of e757c012befa5b2e7288da9ab5a27bbd4da0dd06

It changes a global noop serializer from a global static to a local
function static that's constructed on first use.

Original change's description:
> Replace the wire serializer with a no-op impl on disconnect
>
> Now that the command serialization knows to no-op if
> GetCmdSpace returns nullptr, when the wire is disconnected,
> we can replace it with a no-op serializer that always returns
> nullptr.
>
> Bug: chromium:951558
> Change-Id: I7363fd10f529119e515eda0e743e1a7839049b9b
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30000
> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
> Reviewed-by: Stephen White <senorblanco@chromium.org>
> Commit-Queue: Austin Eng <enga@chromium.org>

Bug: chromium:951558
Change-Id: I827cdbd212fa585b542fd4ea1eb9654eec6002c7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30420
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-10-19 15:59:21 +00:00
Jiawei Shao
ae5f950444 Add the entry point of CreateReadyComputePipeline
This patch adds the entry point of CreateReadyComputePipeline in both
dawn_native and dawn_wire.

TODOs:
1. Add more tests in dawn_unittests and dawn_end2end_tests.
2. Put the main logic of creating a pipeline into a separate thread.

BUG=dawn:529
TEST=dawn_end2end_tests

Change-Id: I7edd269a5422a8b85320a7f9173df925decba633
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30060
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-10-19 01:56:08 +00:00
Austin Eng
875091308b Cleanup old/deprecated wire and native APIs
Bug: none
Change-Id: Idf9a00d8dc675e1fbc83554c487edcb2dda498c8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30001
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-10-16 15:21:16 +00:00
Austin Eng
b1938273e4 Revert "Replace the wire serializer with a no-op impl on disconnect"
This reverts commit e757c012befa5b2e7288da9ab5a27bbd4da0dd06.

Reason for revert: Uses a static initializer

Original change's description:
> Replace the wire serializer with a no-op impl on disconnect
>
> Now that the command serialization knows to no-op if
> GetCmdSpace returns nullptr, when the wire is disconnected,
> we can replace it with a no-op serializer that always returns
> nullptr.
>
> Bug: chromium:951558
> Change-Id: I7363fd10f529119e515eda0e743e1a7839049b9b
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30000
> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
> Reviewed-by: Stephen White <senorblanco@chromium.org>
> Commit-Queue: Austin Eng <enga@chromium.org>

TBR=cwallez@chromium.org,senorblanco@chromium.org,enga@chromium.org

Change-Id: I6549cfb27c6c5812e067ea23c6a706e84c78e1a6
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:951558
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30380
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-10-16 03:03:58 +00:00
Austin Eng
e757c012be Replace the wire serializer with a no-op impl on disconnect
Now that the command serialization knows to no-op if
GetCmdSpace returns nullptr, when the wire is disconnected,
we can replace it with a no-op serializer that always returns
nullptr.

Bug: chromium:951558
Change-Id: I7363fd10f529119e515eda0e743e1a7839049b9b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30000
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-10-15 15:46:13 +00:00
Austin Eng
cac0442277 dawn_wire: Support chunked commands
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>
2020-10-13 22:35:34 +00:00
Corentin Wallez
4171134daf Buffer: Validate the offset is aligned to 8
This is to match the upstream WebGPU spec.

Bug: dawn:445

Change-Id: I1a511ed9a2a04c7b95368ce724d69c128158f097
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29360
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-10-07 17:19:53 +00:00
Corentin Wallez
3066cd3387 dawn_wire: Guarantee OOM on MAX_SIZE_T sized buffers.
This can be used to simplify some logic in Blink.

Bug: dawn:445

Change-Id: I9859c51bc95f564847035533426675188eb8ef99
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29362
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-10-07 15:03:33 +00:00
Austin Eng
16e01affcb Reland "Add a per-thread proc table using thread local storage"
This is a reland of b04a92f01be0e7dc2faf91e0f6f974f990e8b2f2
with the deletion of a duplicate exported function in dawn_wire that
was causing a compilation failure on Windows.

Original change's description:
> Add a per-thread proc table using thread local storage
>
> In situations where both dawn_wire and dawn_native are used on separate
> threads (Chrome with --single-process or --in-process-gpu), it's
> desirable to have a per-thread proc table so that the WebGPU C++ API can
> still be used. This eliminates classes of bugs with manual
> reference/release errors.
>
> This also changes many of the GetProcs functions to return const
> references to the static proc tables known at compile time, instead of a
> copy.
>
> Bug: none
> Change-Id: I8775bb715b312dd9476a1903fbd797d4b1302614
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29240
> Reviewed-by: Stephen White <senorblanco@chromium.org>
> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
> Commit-Queue: Austin Eng <enga@chromium.org>

Bug: none
Change-Id: Id90e5372132cd93a2f8631c8185d0e71b01bc1af
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29443
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
2020-10-06 16:13:42 +00:00
Kai Ninomiya
326e14f34b Revert "Add a per-thread proc table using thread local storage"
This reverts commit b04a92f01be0e7dc2faf91e0f6f974f990e8b2f2.

Reason for revert: Build failed on win-archive-dbg; see:
https://chromium-review.googlesource.com/c/chromium/src/+/2450792

Original change's description:
> Add a per-thread proc table using thread local storage
>
> In situations where both dawn_wire and dawn_native are used on separate
> threads (Chrome with --single-process or --in-process-gpu), it's
> desirable to have a per-thread proc table so that the WebGPU C++ API can
> still be used. This eliminates classes of bugs with manual
> reference/release errors.
>
> This also changes many of the GetProcs functions to return const
> references to the static proc tables known at compile time, instead of a
> copy.
>
> Bug: none
> Change-Id: I8775bb715b312dd9476a1903fbd797d4b1302614
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29240
> Reviewed-by: Stephen White <senorblanco@chromium.org>
> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
> Commit-Queue: Austin Eng <enga@chromium.org>

TBR=cwallez@chromium.org,senorblanco@chromium.org,enga@chromium.org

Change-Id: I4587b457b7b2dd5d3c7457065bf4e77b95af59d7
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: none
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29442
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
2020-10-05 22:35:40 +00:00
Austin Eng
b04a92f01b Add a per-thread proc table using thread local storage
In situations where both dawn_wire and dawn_native are used on separate
threads (Chrome with --single-process or --in-process-gpu), it's
desirable to have a per-thread proc table so that the WebGPU C++ API can
still be used. This eliminates classes of bugs with manual
reference/release errors.

This also changes many of the GetProcs functions to return const
references to the static proc tables known at compile time, instead of a
copy.

Bug: none
Change-Id: I8775bb715b312dd9476a1903fbd797d4b1302614
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29240
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
2020-10-05 19:53:58 +00:00
Corentin Wallez
df90930683 wire::client: Make Buffer::MapAsync surface server-side errors.
In the WebGPU specification, validation errors for mapAsync take
precedence over the early-unmap or early-destroy promise resolution.

Change the client to wait for the mapAsync status from the server before
sending the cancelation through the callback. If the server sends back
an error, then it takes precedence over the client-side status.

Also adds tests for the updated semantic.

Bug: dawn:445

Change-Id: I7bf1d8bbb3cb62d73ab19ecdf0aad2963e854964
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29300
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
2020-10-05 11:03:21 +00:00
Jiawei Shao
ed2b465f86 Report more detailed error information for the failures of mapAsync
This patch adds two new buffer map async status "destroyed before
callback" and "unmapped before callback" to replace the status "unknown"
so that the developers can get more details when meeting such errors in
the call of buffer mapAsync.

Note that this patch still preserves "unknown" as it is still being used
in Chromium.

BUG=dawn:533
TEST=dawn_unittests

Change-Id: I12deefb49311ea6adea72c24e4e40797dd7eb4a1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28883
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
2020-09-27 02:00:52 +00:00
Corentin Wallez
19b910d796 Make dawn components support CMake's BUILD_SHARED_LIBS
Convention in CMake land is that libraries default to STATIC or
SHARED based on BUILD_SHARED_LIBS. Do that for Dawn components
that users will link against, but make sure to correctly set
macros used for the export headers.

Bug: None

Change-Id: I6a96cb95706b63c18942424baada11ff81691866
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28503
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-09-16 19:03:18 +00:00
Austin Eng
0b89b27263 Move client-side OnCompletion callbacks to the server.
We need callbacks to be processed server-side so that callback
ordering can be made consistent.

Bug: dawn:516
Change-Id: Ie5590ca33fce6bda431f93ae9ff8e832468109c1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27481
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
2020-08-28 21:23:50 +00:00
Corentin Wallez
f7123d7463 Remove MapRead/WriteAsync and CreateBufferMapped
Bug: dawn:445

Change-Id: I0b0755b6bb754d1fff99aa59b08362f89950e300
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26301
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-08-20 14:22:29 +00:00
Corentin Wallez
13f3340173 Remove wgpu::Buffer::SetSubData
Bug: dawn:22

Change-Id: Id2bb79e84064c0d81d4dec5e85340d015806f9bc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26701
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-08-13 16:01:08 +00:00
Corentin Wallez
f93fa6acd9 dawn_wire: Implement CreateBufferMapped on top of mappedAtCreation
Reland with a fix for narrowing of uint64_t to size_t of the buffer size
in 32 bit mode.

This inverts the shimming to have the old mapping at creation method
be implemented on top of the new method..

Also updates Wire tests to use mappedAtCreation instead of
CreateBufferMapped.

TBR=senorblanco@chromium.org
Bug: dawn:445
Change-Id: I89fe84b6e0b5d0d4a5c6a2db7b38cb7d6cd063f0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25981
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-07-29 17:01:11 +00:00
Corentin Wallez
b88b1a15c7 Revert "dawn_wire: Implement CreateBufferMapped on top of mappedAtCreation"
This reverts commit 0811ecc7756f8913fdaf9c9bf589b8e498badd17.

Reason for revert: Makes the Dawn roll fail.

Original change's description:
> dawn_wire: Implement CreateBufferMapped on top of mappedAtCreation
> 
> This inverts the shimming to have the old mapping at creation method
> be implemented on top of the new method..
> 
> Also updates Wire tests to use mappedAtCreation instead of
> CreateBufferMapped.
> 
> Bug: dawn:445
> 
> Change-Id: I77dcfe72040e5bf187c41fe99c8dd785d5156a07
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25701
> Commit-Queue: Corentin Wallez <cwallez@chromium.org>
> Reviewed-by: Austin Eng <enga@chromium.org>

TBR=cwallez@chromium.org,senorblanco@chromium.org,enga@chromium.org

Change-Id: Ied940d505fdf576860697dee8df2548c73581eba
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: dawn:445
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25980
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-07-29 09:10:13 +00:00
Corentin Wallez
0811ecc775 dawn_wire: Implement CreateBufferMapped on top of mappedAtCreation
This inverts the shimming to have the old mapping at creation method
be implemented on top of the new method..

Also updates Wire tests to use mappedAtCreation instead of
CreateBufferMapped.

Bug: dawn:445

Change-Id: I77dcfe72040e5bf187c41fe99c8dd785d5156a07
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25701
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-07-28 15:30:07 +00:00
Corentin Wallez
5b007a519a dawn_wire: Implement MapRead/WriteAsync on top of MapAsync
This inverts the shimming to have the old mapping entrypoints be
implemented on top of the new mapping entrypoint.

Bug: dawn:445

Change-Id: I98c81b2873d506790b1b19048c5289c62d1b83ae
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25700
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-07-28 09:00:11 +00:00
Kai Ninomiya
0d158ac681 Fix some bugs in buffer mapping in the wire client
Bug: dawn:445
Change-Id: Ibfc55e26c2bc1757811669d84e54e9f775bee8f6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25440
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
2020-07-23 18:27:56 +00:00
Corentin Wallez
f6e7044697 Add offset and size to Get[Const]MappedRange.
Bug: dawn:445
Change-Id: I73758d95e61d1fbe69f328907a6170a1b27d785b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24983
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
2020-07-17 18:50:37 +00:00
Corentin Wallez
450b6f4f84 Use MapAsync callback for server->client map callbacks.
This changes the format of the server->client callback for async mapping
to match MapAsync. Previously there were two callbacks, one for
MapReadAsync and one for MapWriteAsync.

Bug: dawn:445

Change-Id: I3330c07ac8bb6d1fa9019563e9c946875e852639
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24821
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
2020-07-17 07:38:36 +00:00
Corentin Wallez
c15c6eb9d8 dawn_wire/server: Use MapAsync to do Map[Read/Write]Async
This changes the format of the client->server command for async mapping
to match MapAsync and updates the doer on the server to implement it via
MapAsync. Previously the command had an isWrite boolean that routed to
either MapReadAsync or MapWriteAsync.

Also updates the Wire tests using mock_webgpu.h for the new server-side
traces.

Bug: dawn:445

Change-Id: I673cbe5e08b7c71539e8678dbb07b4043a9fadb2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24820
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
2020-07-16 17:07:05 +00:00
Tomek Ponitka
9d66c5353f Implementing Queue::WriteTexture in Metal
Added implementation of writeTexture in Metal. It's using a
staging buffer instead of writing directly from the CPU to
the texture, because Dawn uses the private storage mode for
most of the Metal textures.

Bug: dawn:483
Change-Id: I6b85ee8bbe343881337bdb203a122dc1f1523177
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24581
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
2020-07-15 18:06:07 +00:00