G-API: oneVPL merge DX11 acceleration by sivanov-work · Pull Request #21049 · opencv/opencv

sivanov-work · 2021-11-12T12:59:33Z

Add DX11 allocation resources: surfaces
Implement actual DX11 acceleration with MFX allocator
Reuse IDeviceSelector interface for selecting device in DX11 accelerator policy

Add shared_lock & elastics_barrier to work out R-W pattern & provide lock-free access to MediaFrame data.
Add unit tests for lock-free primitives

Improve decode pipeline: add yield on queue decode operations for used up busy surfaces

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

I agree to contribute to the project under Apache 2 License.
To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
The PR is proposed to proper branch
There is reference to original bug report and related work
There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
The feature is well documented and sample code can be built with the project CMake

Build Configuration

force_builders=XCustom,Custom Win,Custom Mac
build_gapi_standalone:Linux x64=ade-0.1.1f
build_gapi_standalone:Win64=ade-0.1.1f
build_gapi_standalone:Mac=ade-0.1.1f
build_gapi_standalone:Linux x64 Debug=ade-0.1.1f

Xbuild_image:Custom=centos:7
Xbuildworker:Custom=linux-1
build_gapi_standalone:Custom=ade-0.1.1f

build_image:Custom=ubuntu-openvino-2021.3.0:20.04
Xbuild_image:Custom Win=openvino-2021.2.0
build_image:Custom Mac=openvino-2021.2.0

test_modules:Custom=gapi,python2,python3,java
test_modules:Custom Win=gapi,python2,python3,java
test_modules:Custom Mac=gapi,python2,python3,java

buildworker:Custom=linux-1
test_opencl:Custom=OFF
test_bigdata:Custom=1
test_filter:Custom=*

build_image:Custom Win=gapi-onevpl-2021.6.0
buildworker:Custom Win=windows-3
build_contrib:Custom Win=OFF

mpashchenkov

First wave (stupid comments).

mpashchenkov · 2021-11-17T08:45:01Z

modules/gapi/src/backends/render/grenderocv.cpp

@@ -157,19 +154,26 @@ GAPI_OCV_KERNEL_ST(RenderFrameOCVImpl, cv::gapi::wip::draw::GRenderFrame, Render
         *
         */


Is it necessary to align comment with other code in {}?

mpashchenkov · 2021-11-17T09:10:42Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

+    switch (FourCC) {
+        case MFX_FOURCC_I420:
+        case MFX_FOURCC_NV12:
+            nbytes = width * height +  2 * half_width * half_height;


Suggested change

nbytes = width * height + 2 * half_width * half_height;

nbytes = width * height + 2 * half_width * half_height;

mpashchenkov · 2021-11-17T09:11:13Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

+}
+
+surface_ptr_t create_surface_RGB4_(mfxFrameInfo frameInfo,
+                                  std::shared_ptr<void> out_buf_ptr,


One more space for parameters.

mpashchenkov · 2021-11-17T09:12:14Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

+}
+
+surface_ptr_t create_surface_other_(mfxFrameInfo frameInfo,
+                                   std::shared_ptr<void> out_buf_ptr,


mpashchenkov · 2021-11-24T11:05:42Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.cpp


 void VPLDX11AccelerationPolicy::init(session_t session) {
-    mfxStatus sts = MFXVideoCORE_GetHandle(session, MFX_HANDLE_D3D11_DEVICE, reinterpret_cast<mfxHDL*>(&hw_handle));
+    mfxStatus sts = MFXVideoCORE_SetHandle(session, MFX_HANDLE_D3D11_DEVICE, (mfxHDL) hw_handle);


(mfxHDL)
Why is it not some_cast<>()?

some kind of copy-paste from public VPL examples. I'll fix it later

mpashchenkov · 2021-11-24T11:50:09Z

modules/gapi/test/streaming/gapi_streaming_utils_test.cpp

+    auto read_work = [&lock, &shared_value](size_t count) {
+
+        auto old_shared_value = shared_value;
+        for (size_t i = 0; i < count; i ++) {


Suggested change

for (size_t i = 0; i < count; i ++) {

for (size_t i = 0; i < count; i++) {

mpashchenkov · 2021-11-24T11:51:19Z

modules/gapi/test/streaming/gapi_streaming_utils_test.cpp

+
+    // TEST invariant: last_visitor_id MUST be one from any LATTER worker visitor_id
+    bool one_of_available_ids_matched = false;
+    for (size_t id = 0; id < max_worker_count; id ++) {


Suggested change

for (size_t id = 0; id < max_worker_count; id ++) {

for (size_t id = 0; id < max_worker_count; id++) {

mpashchenkov · 2021-11-24T11:53:11Z

modules/gapi/test/streaming/gapi_streaming_vpl_core_test.cpp

+        if (adaptee.owns()) {
+            w_lock_counter ++;
+        } else {
+            r_lock_counter ++;


Remove extra space before increment ++ 648 and 650

mpashchenkov · 2021-11-24T11:53:26Z

modules/gapi/test/streaming/gapi_streaming_vpl_core_test.cpp

+        if (adaptee.owns()) {
+            w_unlock_counter ++;
+        } else {
+            r_unlock_counter ++;


Remove extra space before increment ++ 657 and 659

mpashchenkov · 2021-11-24T11:53:42Z

modules/gapi/test/streaming/gapi_streaming_vpl_core_test.cpp

+
+    mfxFrameData data;
+    const int exec_count = 123;
+    for (int i = 0; i < exec_count; i ++) {


Suggested change

for (int i = 0; i < exec_count; i ++) {

for (int i = 0; i < exec_count; i++) {

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.hpp

mpashchenkov · 2021-11-24T14:39:30Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.cpp

-    throw std::runtime_error("VPLDX11AccelerationPolicy::create_surface_pool() is not implemented");
+    // allocate textures by explicit request
+    mfxFrameAllocResponse mfxResponse;
+    //TODO


i think it must be removed

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.cpp

mpashchenkov · 2021-11-25T07:26:18Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.cpp

+    desc.MiscFlags      = 0;
+    std::vector<ID3D11Texture2D*> staging_textures;
+    staging_textures.reserve(request->NumFrameSuggested);
+    for (int i = 0; i < request->NumFrameSuggested; i ++ ) {


Suggested change

for (int i = 0; i < request->NumFrameSuggested; i ++ ) {

for (int i = 0; i < request->NumFrameSuggested; i++ ) {

mpashchenkov · 2021-11-25T07:59:00Z

modules/gapi/src/streaming/onevpl/accelerators/surface/dx11_frame_adapter.cpp

+                    pitch,
+                    pitch / 2,
+                    pitch / 2, 0u
+                };


mpashchenkov · 2021-11-25T08:10:21Z

modules/gapi/test/streaming/gapi_streaming_utils_test.cpp

+    auto work = [&lock, &shared_value](size_t count) {
+        for (size_t i = 0; i < count; i ++) {
+            lock.lock();
+            shared_value ++;


Suggested change

shared_value ++;

shared_value++;

mpashchenkov · 2021-11-25T08:10:46Z

modules/gapi/test/streaming/gapi_streaming_utils_test.cpp

+    };
+
+    std::thread worker_thread([&barrier, sync, work] () {
+


extra space

mpashchenkov · 2021-11-25T08:13:21Z

modules/gapi/test/streaming/gapi_streaming_vpl_core_test.cpp

+
+    std::function<mfxStatus(mfxMemId, mfxFrameData *)> lock =
+    [&lock_counter] (mfxMemId, mfxFrameData *) {
+        lock_counter ++;


Suggested change

lock_counter ++;

lock_counter++;

mpashchenkov · 2021-11-25T08:13:34Z

modules/gapi/test/streaming/gapi_streaming_vpl_core_test.cpp

+
+    mfxFrameData data;
+    const int exec_count = 123;
+    for (int i = 0; i < exec_count; i ++) {


Suggested change

for (int i = 0; i < exec_count; i ++) {

for (int i = 0; i < exec_count; i++) {

modules/gapi/src/streaming/onevpl/cfg_params_parser.cpp

mpashchenkov · 2021-11-29T13:35:02Z

modules/gapi/src/streaming/onevpl/accelerators/utils/shared_lock.cpp

+    bool pred_excl = exclusive_lock.load();
+    do {
+        if (!pred_excl) {
+            // if no exclusive lock than start shared lock transaction


than --> then

OrestChura

№ 2

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.hpp

OrestChura · 2021-11-29T12:25:39Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.hpp

+    VPLDX11AccelerationPolicy(device_selector_ptr_t selector);
    ~VPLDX11AccelerationPolicy();

+    using pool_t = CachedPool;


Can this using be a part of the parent class, or it is specific for cpu & dx11 only?

got it,
when new policies would be intorduced i suggest to make light refactoring here

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.cpp

anna-khakimova · 2021-12-01T18:00:37Z

Looks good for me. Great work!

alalek · 2021-12-01T22:54:39Z

Please configure builders with enabled oneVPL support (see previous oneVPL-related PRs)

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_dx11.cpp

modules/gapi/src/streaming/onevpl/accelerators/dx11_alloc_resource.cpp

modules/gapi/src/streaming/onevpl/accelerators/dx11_alloc_resource.hpp

OrestChura · 2021-12-02T17:31:23Z

modules/gapi/src/streaming/onevpl/accelerators/dx11_alloc_resource.cpp

+    return MFX_ERR_NONE;
+}
+
+DX11AllocationRecord::DX11AllocationRecord() = default;


can this be declared in .hpp?

modules/gapi/src/streaming/onevpl/accelerators/dx11_alloc_resource.hpp

modules/gapi/include/opencv2/gapi/streaming/onevpl/cfg_params.hpp

modules/gapi/src/streaming/onevpl/accelerators/utils/shared_lock.cpp

OrestChura · 2021-12-02T22:54:07Z

modules/gapi/src/streaming/onevpl/accelerators/utils/shared_lock.cpp

+            std::this_thread::yield();
+        }
+        prev_shared = shared_counter.load();
+    } while (prev_shared != 0 || !in_progress);


Did I understand correctly that when in_progress == true (and in the end of cycle body shared_counter == 0, execution should go out of cycle?
Then why it is named in_progress? I would expect that if some process is in progress, it means it should remain cycling.
More likely this variable should have been called like complete, because it means that the current action is complete and execution should move forward.
The same is for the method shared_lock above

it's locking with rollback in failed case:

When i'm write thread and want to make exclusive lock, then when i get current shared_counter and make sure no other competitors is here (shared_counter == 0) (---1----) then i will make exclusive_lock (----2----) and flag operation as in progress. Starting from this moment no one reader thread cannot take shared access because bool pred_excl = exclusive_lock.load(); will fail for them.

But let's imaging situation when after check (---1---) and right before (---2----) one of the competitors reader thread would be succeeded in obtaining shared lock... So when i finish operation (---2---) it is more readers thread obtained lock already... So i need to rolled back my (---2---) operation in this case and wait until all readers complete their works before continue...
So in_progress is using as flag that i must rollback extra exclusive lock when the second condition fails. And must proceed when both condition true: is_progress == true && `shared_lock ==0)

That's what I meant - in the end, in_progress == true, but you must proceed further. The comment is regarding naming. "In progress" itself means you are in the middle of some process and you cannot move further. It doesn't match with the meaning of this variable

If to suggest, this variable means that write access is gained, so it can be named like gained, succeeded; or, better, meaningful locked, because exclusive_lock was set true, but in the case you've described it should be rolled back. But it is locked that time.

this variable cannot be named as gained, succeeded of locked because resource is not locked until we finished final while. It actually means: we are in the middle of progress something and we must not make some assumption about succeed or not until we check the second condition shared_lock ==0

So in_progress means: I increased exclusive lock counter but i'm not sure that i'm allowed to continue with exclusive access until i check the second condition.
This is sync with using two independent atomic states: and we cannot proceed it both of them wouldn't reach the right state

OrestChura · 2021-12-02T23:06:10Z

modules/gapi/test/streaming/gapi_streaming_vpl_data_provider.cpp

    mfx_param_0.Data.U32 = provider_ptr->get_mfx_codec_id();
-    EXPECT_EQ(MFXSetConfigFilterProperty(cfg_inst_0,(mfxU8 *)"mfxImplDescription.mfxDecoderDescription.decoder.CodecID",
+    EXPECT_EQ(MFXSetConfigFilterProperty(cfg_inst_0,(mfxU8 *)CfgParam::decoder_id_name(),
                                                    mfx_param_0), MFX_ERR_NONE);


Please, remove the C-style cast

OrestChura · 2021-12-02T23:13:41Z

modules/gapi/test/streaming/gapi_streaming_tests.cpp

        auto args = cv::compile_args(use_only{pkg});
        if (cap) {
-            args += cv::compile_args(streaming::queue_capacity{cap.value()});
+            args += cv::compile_args(cv::gapi::streaming::queue_capacity{cap.value()});


Then, maybe using namespace cv::gapi; should be removed 3 lines upper

modules/gapi/src/streaming/onevpl/cfg_param_device_selector.cpp

modules/gapi/src/streaming/onevpl/accelerators/dx11_alloc_resource.cpp

modules/gapi/CMakeLists.txt

…masking

OrestChura · 2021-12-03T14:01:01Z

modules/gapi/src/streaming/onevpl/engine/decode/decode_engine_legacy.cpp

-        decRequest.NumFrameSuggested = preallocated_frames_count;
+        if (static_cast<size_t>(std::numeric_limits<mfxU16>::max()) < preallocated_frames_count) {
+            GAPI_LOG_WARNING(nullptr, "Cannot proceed with CfgParam \"" << CfgParam::frames_pool_size_name() << "\": " <<
+                                  preallocated_frames_count << ". It must not be equal than " <<


must not be greater I believe

OrestChura · 2021-12-03T14:51:45Z

modules/gapi/CMakeLists.txt

    ocv_target_link_libraries(opencv_test_gapi PRIVATE ${VPL_IMPORTED_TARGETS})
+    if(MSVC)
+        target_compile_options(opencv_test_gapi PUBLIC "/wd4201")
+    endif()


You've rewritten includes via pragma in tests, too - is it still necessary?

OrestChura

Thank you for all the efforts!

mpashchenkov

I didn't find places with possible nulls and going out of memory. I'm give up.

mpashchenkov · 2021-12-06T12:29:48Z

modules/gapi/src/streaming/onevpl/accelerators/surface/dx11_frame_adapter.cpp

+                                         static_cast<DX11AllocationItem::subresource_id_t>(
+                                            reinterpret_cast<uint64_t>(
+                                                reinterpret_cast<DX11AllocationItem::subresource_id_t *>(
+                                                    handle.second)))}};//,


Suggested change

handle.second)))}};//,

handle.second)))}};

mpashchenkov · 2021-12-06T12:47:13Z

modules/gapi/src/streaming/onevpl/onevpl_export.hpp

+#if defined(_MSC_VER)
+#pragma warning(push)
+#pragma warning(disable : 4201)
+#pragma warning(disable : 4302)
+#pragma warning(disable : 4311)
+#pragma warning(disable : 4312)
+#endif // defined(_MSC_VER)


Why is not:

if(MSVC) if(HAVE_GAPI_ONEVPL) <-- I'm not sure about the flag ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4201 /wd4302 /wd4311 /wd4312) endif()

in CMakeLists.txt?

One place of disabling can reduce problems in the future.

Because we should not disable warnings anywhere.
Problems with warnings should be suppressed locally.

sivanov-work · 2021-12-07T05:44:42Z

if no other comments are coming then I assume to apply them in the next PRs:

warning for cmake disabled for onevpl in unit tests only, regular build is unaffected: move them into header inclusion in tests
some typos in log & assert infos
to rename shared_lock : in_progress to is_pending etc.

dmatveev

Approved with comments

dmatveev · 2021-12-07T06:45:54Z

modules/gapi/CMakeLists.txt

    ocv_target_compile_definitions(opencv_test_gapi PRIVATE -DHAVE_ONEVPL)
    ocv_target_link_libraries(opencv_test_gapi PRIVATE ${VPL_IMPORTED_TARGETS})
+    if(MSVC)
+        target_compile_options(opencv_test_gapi PUBLIC "/wd4201")


Shouldn't this be PRIVATE?

dmatveev · 2021-12-07T06:48:22Z

modules/gapi/perf/streaming/gapi_streaming_source_perf_tests.cpp

    };

+    if (!type.empty()) {
+        cfg_params.push_back(CfgParam::create_decoder_id(type.c_str()));


BTW this looks odd from the user perspective - why do you ask users to provide you with c_str() while there's std::string ? Is that char* thing copied or should stay alive - for how long?

to retain public interface consistent (maybe for future): because STL classes are not ABI compatible in general

dmatveev · 2021-12-07T06:51:09Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

 namespace wip {
 namespace onevpl {
+namespace utils {
+mfxU32 GetSurfaceSize_(mfxU32 FourCC, mfxU32 width, mfxU32 height) {


alignment is still not counted here right?

note sometimes planes are aligned by up to 4096

it was get from VPL examples, may be we could add it when faces with the issue later

dmatveev · 2021-12-07T06:52:04Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

+surface_ptr_t create_surface_RGB4_(mfxFrameInfo frameInfo,
+                                   std::shared_ptr<void> out_buf_ptr,
+                                   size_t out_buf_ptr_offset,
+                                   size_t out_buf_size)


It may be a tiny detail but how this function is supposed to be used?

I see it creates something and returns, but that's the out_buf_ptr then?

it is shared across surfaces: each surface memory is sliced from out_buf_ptr
So all surfaces have reference to that buf and mapped on that buffer memory by offset

Referencing is required to keep out_buf_ptr alive when alive at least one surface

dmatveev · 2021-12-07T06:56:08Z

modules/gapi/src/streaming/onevpl/accelerators/accel_policy_cpu.cpp

+    // TODO more intelligent check
+    if (out_buf_size <=
+        out_buf_ptr_offset + (surfW * surfH) + ((surfW / 2) * (surfH / 2))) {
+        GAPI_LOG_WARNING(nullptr, "Not enough buffer, ptr: " << out_buf_ptr <<


-> Insufficient buffer size?

TODO in next PR

dmatveev · 2021-12-07T07:21:56Z

modules/gapi/src/streaming/onevpl/accelerators/surface/dx11_frame_adapter.cpp

+    if (mode == MediaFrame::Access::R) {
+        alloc_data->unlock_read(mid, data);
+    } else {
+        alloc_data->unlock_write(mid, data);


so can this be a plain abstract rwlock (e.g. like one from boost or pthreads)? Or something differs in the access at the VPL level?

adapter is quite different, but shared_mutex inside it should be reused from official sources further
P.S. is_own_locked can be put on upper level to do not customize official shared_mutex
P.P.S. official c++17 shared_mutex is mutex-based with wait-lock, but tbb has imple,entation with rw_spinlock to make busy wait

dmatveev · 2021-12-07T07:24:57Z

modules/gapi/src/streaming/onevpl/accelerators/utils/elastic_barrier.hpp

+namespace onevpl {
+
+template<typename Impl>
+class elastic_barrier {


Is the idea of this structure described anywhere?

(I'd expect to see it here)

May i add in the next PR such description?
In short overview:
pass only single one thread (the first one) to make some initialization: resource acquisition - another will waiting for the first thread job completion.
And make final deinitialization: resource release - only for last one( if no other threads hold resource or intendent to acquire it)

Some kind of double check lock patter in Gstreamer MediaResource Adapter - but support more scenarios( and more efficient?)

dmatveev · 2021-12-07T07:25:51Z

modules/gapi/src/streaming/onevpl/accelerators/utils/shared_lock.cpp

+}
+bool SharedLock::owns() const {
+    return exclusive_lock.load();
+}


why do we need our own shared lock here?

yes it is, but this can be easily emulated on upper level in LockAdapter

dmatveev · 2021-12-07T07:26:26Z

modules/gapi/src/streaming/onevpl/accelerators/utils/shared_lock.hpp

+namespace wip {
+namespace onevpl {
+
+class GAPI_EXPORTS SharedLock {


is it modelled after the C++14's shared_lock?

yes, homegrown solution...

dmatveev · 2021-12-07T07:27:00Z

modules/gapi/src/streaming/onevpl/cfg_params_parser.cpp

    return ret;
 }
+
+size_t strtoull_or_throw(const char* str) {


y use chars

to satisty c-string in interface functions (to make interface ABI consistent)

sivanov-work · 2021-12-08T05:38:57Z

@alalek Could you merge it please? Non addressed comment would be addressed at the next PR
Custom build for Win+gapi-onevpl is passed (warning fixed and disarmed for UT)

G-API: oneVPL merge DX11 acceleration * Merge DX11 initial * Fold conditions row in MACRO in utils * Inject DeviceSelector * Turn on DeviceSelector in DX11 * Change sharedLock logic & Move FMT checking in FrameAdapter c-tor * Move out NumSuggestFrame to configure params * Drain file source fix * Fix compilation * Force zero initializetion of SharedLock * Fix some compiler warnings * Fix integer comparison warnings * Fix integers in sample * Integrate Demux * Fix compilation * Add predefined names for some CfgParam * Trigger CI * Fix MultithreadCtx bug, Add Dx11 GetBlobParam(), Get rif of ATL CComPtr * Fix UT: remove unit test with deprecated video from opencv_extra * Add creators for most usable CfgParam * Eliminate some warnings * Fix warning in GAPI_Assert * Apply comments * Add VPL wrapped header with MSVC pragma to get rid of global warning masking

sivanov-work added category: g-api / gapi feature labels Nov 12, 2021

sivanov-work marked this pull request as draft November 12, 2021 13:01

sivanov-work requested review from AsyaPronina, dmatveev, mpashchenkov and smirnov-alexey November 17, 2021 06:18

sivanov-work marked this pull request as ready for review November 17, 2021 06:45

sivanov-work mentioned this pull request Nov 22, 2021

G-API: oneVPL Implement asynchronous MFP demux data provider #21022

Merged

6 tasks

sivanov-work removed the request for review from AsyaPronina November 22, 2021 09:46

sivanov-work added 13 commits November 23, 2021 09:33

Merge DX11 initial

77a1bf4

Fold conditions row in MACRO in utils

082ffdd

Inject DeviceSelector

d6295bb

Turn on DeviceSelector in DX11

ebd8dff

Change sharedLock logic & Move FMT checking in FrameAdapter c-tor

c56cda1

Move out NumSuggestFrame to configure params

406fbd5

Drain file source fix

4b36e60

Fix compilation

e5891c1

Force zero initializetion of SharedLock

63c7335

Fix some compiler warnings

681531a

Fix integer comparison warnings

036685b

Fix integers in sample

f3eb0a3

Integrate Demux

2a3aa90

sivanov-work force-pushed the vpl_dx11_merge branch from 766f801 to f3eb0a3 Compare November 23, 2021 12:02

sivanov-work added 3 commits November 23, 2021 15:11

Fix compilation

7cd7bc1

Add predefined names for some CfgParam

1a95269

Trigger CI

8413812

mpashchenkov reviewed Nov 24, 2021

View reviewed changes

mpashchenkov reviewed Nov 25, 2021

View reviewed changes

mpashchenkov requested a review from OrestChura November 26, 2021 11:53

mpashchenkov reviewed Nov 29, 2021

View reviewed changes

modules/gapi/src/streaming/onevpl/cfg_params_parser.cpp Show resolved Hide resolved

mpashchenkov reviewed Nov 29, 2021

View reviewed changes

OrestChura reviewed Nov 29, 2021

View reviewed changes

sivanov-work requested a review from anna-khakimova December 2, 2021 06:02

sivanov-work added 3 commits December 2, 2021 14:36

Fix MultithreadCtx bug, Add Dx11 GetBlobParam(), Get rif of ATL CComPtr

587bba8

Fix UT: remove unit test with deprecated video from opencv_extra

22ceef8

Add creators for most usable CfgParam

296240c

OrestChura reviewed Dec 2, 2021

View reviewed changes

OrestChura reviewed Dec 3, 2021

View reviewed changes

sivanov-work added 2 commits December 3, 2021 11:51

Eliminate some warnings

9b8a657

Fix warning in GAPI_Assert

3f990a3

OrestChura reviewed Dec 3, 2021

View reviewed changes

modules/gapi/CMakeLists.txt Outdated Show resolved Hide resolved

sivanov-work added 2 commits December 3, 2021 15:13

Apply comments

97e8a0f

Add VPL wrapped header with MSCV pragma to get rid of global warning …

8e090a7

…masking

OrestChura reviewed Dec 3, 2021

View reviewed changes

OrestChura approved these changes Dec 6, 2021

View reviewed changes

mpashchenkov approved these changes Dec 6, 2021

View reviewed changes

dmatveev self-assigned this Dec 7, 2021

dmatveev added this to the 4.5.5 milestone Dec 7, 2021

dmatveev approved these changes Dec 7, 2021

View reviewed changes

alalek merged commit 5c91f5b into opencv:4.x Dec 8, 2021

alalek mentioned this pull request Dec 9, 2021

G-API: broken Win32 builds #21225

Closed

alalek mentioned this pull request Dec 30, 2021

(5.x) Merge 4.x #21371

Merged

alalek mentioned this pull request Feb 22, 2022

(5.x) Merge 4.x #21651

Merged

		@@ -157,19 +154,26 @@ GAPI_OCV_KERNEL_ST(RenderFrameOCVImpl, cv::gapi::wip::draw::GRenderFrame, Render
		*
		*/

	nbytes = width * height + 2 * half_width * half_height;
	nbytes = width * height + 2 * half_width * half_height;

	for (size_t i = 0; i < count; i ++) {
	for (size_t i = 0; i < count; i++) {

	for (size_t id = 0; id < max_worker_count; id ++) {
	for (size_t id = 0; id < max_worker_count; id++) {

	for (int i = 0; i < exec_count; i ++) {
	for (int i = 0; i < exec_count; i++) {

	for (int i = 0; i < request->NumFrameSuggested; i ++ ) {
	for (int i = 0; i < request->NumFrameSuggested; i++ ) {

Uh oh!

Conversation

sivanov-work commented Nov 12, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Readiness Checklist

Build Configuration

Uh oh!

mpashchenkov left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

OrestChura left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

anna-khakimova commented Dec 1, 2021

Uh oh!

sivanov-work commented Nov 12, 2021 •

edited

Loading