Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Use native dawn dev package for development #338

Merged
merged 31 commits into from
Jul 21, 2020
Merged

Use native dawn dev package for development #338

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
31 commits
Select commit Hold shift + click to select a range
badd437
Add Dawn dependecy file and change CMakeLists.txt
Jun 7, 2020
5c03f7b
Add basic component
Jun 7, 2020
29a30a6
Fix buffer issue
Jun 8, 2020
c0a4475
Fix buffer issue and add op_webgpu module
Jun 12, 2020
aab50e7
Add softmax op
Jun 19, 2020
6525774
Modify some initialize func
Jun 22, 2020
c5eb33d
Modify some initialize func
Jun 22, 2020
0611de2
Delete fence cause queue.submit will create fence internelly
Jun 22, 2020
dbb915f
Modify op_webgpu
Jun 22, 2020
ddc74b8
Add webgpu_cpp.cpp file
Jun 23, 2020
d865652
Rename setDirty
Jun 24, 2020
78a7a82
Modify dnn.cpp
Jun 24, 2020
f5c9edf
Modify some func
Jun 28, 2020
0f48045
Fix initialize func
Jun 28, 2020
0ce21b5
Unmap buffer before submit cmd_buffer
Jun 28, 2020
7d073a9
Add tensor readAsync
Jun 29, 2020
58de7e4
Add unit test
Jun 29, 2020
1df08b3
Async read wgpu buffer
Jun 30, 2020
b3911e8
And async read buffer method
Jul 1, 2020
91eb988
Keep consistent coding style
Jul 1, 2020
e6fb2ef
Async read buffer
Jul 5, 2020
eb1c26b
Made some changes as requested
Jul 8, 2020
8e47b6d
Modify unMap
Jul 8, 2020
359dac6
Modify buffer usage
Jul 9, 2020
fa95ba0
Change Buffer usage
Jul 9, 2020
8377efd
Modify uniform buffer
Jul 12, 2020
d76d2b2
Modify uniform buffer
Jul 12, 2020
99b3eac
Fix coding style issue
Jul 13, 2020
9a64e09
Change Coding style
Jul 15, 2020
ec67c9a
Correct the wrong indents of paramters
Jul 16, 2020
a9b65f1
Release uniform buffer
Jul 18, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions modules/dnn/src/webgpu/include/buffer.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,11 +13,11 @@ class Buffer
public:
Buffer(const std::shared_ptr<wgpu::Device> device);
Buffer(const std::shared_ptr<wgpu::Device> device,
const void* data, size_t size,
wgpu::BufferUsage usage = wgpu::BufferUsage::Storage |
wgpu::BufferUsage::CopyDst | wgpu::BufferUsage::CopySrc );
Buffer( const void* data, size_t size,
wgpu::BufferUsage usage = wgpu::BufferUsage::Uniform | wgpu::BufferUsage::CopyDst);
const void* data, size_t size,
wgpu::BufferUsage usage = wgpu::BufferUsage::Storage |
wgpu::BufferUsage::CopyDst | wgpu::BufferUsage::CopySrc);
Buffer(const void* data, size_t size,
wgpu::BufferUsage usage = wgpu::BufferUsage::Uniform | wgpu::BufferUsage::CopyDst);
~Buffer()
{
buffer_.Release();
Expand Down
4 changes: 2 additions & 2 deletions modules/dnn/src/webgpu/include/tensor.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ class Buffer;
class Tensor{
wzw-intel marked this conversation as resolved.
Show resolved Hide resolved
public:
Tensor(Format fmt = wFormatFp32);
Tensor( const void* data, std::vector<int>& shape,
Tensor(const void* data, std::vector<int>& shape,
Format fmt = wFormatFp32);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Every Tensor should support both read and write, the "usage" parameter should be eliminated. Use "wgpu::BufferUsage::Storage | wgpu::BufferUsage::CopyDst | wgpu::BufferUsage::CopySrc" for internal buffer's usage.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

plase correct the wrong indents of paramters, not only for this function.

Copy link
Collaborator Author

@NALLEIN NALLEIN Jul 16, 2020 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

plase correct the wrong indents of paramters, not only for this function.

Thanks for your patience, I've modified it. By the way, I've add the layer test for softmax in this repo, please help to check whether it works on your device.

git clone https://github.com/NALLEIN/opencv.git
cd opencv/
git checkout -b layerTest origin/layerTest
mkdir build
cd build
cmake -D CMAKE_BUILD_TYPE=Release -DWITH_WEBGPU=ON -D CMAKE_INSTALL_PREFIX=/usr/local ..
make -j8
$(PATH_TO_OPENCV)/build/bin/opencv_test_dnn --gtest_filter=Layer_Test_Softmax.Accuracy

The test case is here. If there is nothing wrong with the test, I will start to complete other ops for evaluation 2.
@huningxin @wzw-intel PTAL.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

test passed on my machine with nVidia GPU.

const void* mapRead();
void unMap();
Expand All @@ -23,7 +23,7 @@ class Tensor{
// Change shape and format to as passed in.
// Copy data if data != NULL
// Allocate new internal buffer if new size > old size or alloc flag is true
Tensor reshape( const void* data, const std::vector<int>& shape,
Tensor reshape(const void* data, const std::vector<int>& shape,
bool alloc = false,
Format fmt = wFormatInvalid);
Tensor fillData(const void * data);
Expand Down
4 changes: 2 additions & 2 deletions modules/dnn/src/webgpu/src/buffer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@ Buffer::Buffer(std::shared_ptr<wgpu::Device> device)
usage_ = wgpu::BufferUsage::Storage;
}

Buffer::Buffer( std::shared_ptr<wgpu::Device> device,
Buffer::Buffer(std::shared_ptr<wgpu::Device> device,
const void* data, size_t size,
wgpu::BufferUsage usage)
{
Expand All @@ -25,7 +25,7 @@ Buffer::Buffer( std::shared_ptr<wgpu::Device> device,
if(data) buffer_.SetSubData(0, size_, data);
}

Buffer::Buffer( const void* data, size_t size,
Buffer::Buffer(const void* data, size_t size,
wgpu::BufferUsage usage)
{
createContext();
Expand Down
10 changes: 5 additions & 5 deletions modules/dnn/src/webgpu/src/op_softmax.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,11 +89,11 @@ bool OpSoftmax::forward(Tensor& in, Tensor& out)
uniformBuffer_ = new Buffer(&param, sizeof(SoftmaxParam));
}

bindTensor( in, 0, bgEntries);
bindTensor( *max_tensor_, 1, bgEntries);
bindTensor( *sum_tensor_, 2, bgEntries);
bindTensor( out, 3, bgEntries);
bindUniform( *uniformBuffer_, 4, bgEntries);
bindTensor(in, 0, bgEntries);
bindTensor(*max_tensor_, 1, bgEntries);
bindTensor(*sum_tensor_, 2, bgEntries);
bindTensor(out, 3, bgEntries);
bindUniform(*uniformBuffer_, 4, bgEntries);

createBindGroup();
createCommandBuffer();
Expand Down