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

sivanov-work · 2021-11-08T11:48:21Z

This PR is descendant of sync version:
#20926

the IDataProvider base interface changed: fetch_data was improved from raw memory to mfx bitstream, CodecId is not requried in this case

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=Custom,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

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

Xtest_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=ON
Xtest_bigdata:Custom=1
Xtest_filter:Custom=*
build_image:Custom=gapi-onevpl-2021.6.0

alalek · 2021-11-08T17:06:56Z

Please reuse CI instructions from #20926 to enable oneVPL builder configuration

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

smirnov-alexey · 2021-11-08T18:13:08Z

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

     * @return fetched bytes count.
     */
-    virtual size_t fetch_data(size_t out_data_bytes_size, void* out_data) = 0;
-
+#ifdef HAVE_ONEVPL


I don't think that using ifdefs in a public header works. Please, remove the guard and handle it in the sources

Have you added #else with assert? I think other reviewers need to check on that

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

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

smirnov-alexey · 2021-11-08T18:33:15Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+
+        // close reader
+        provider_state.store(State::Exhausted);
+        buffer_storage_non_empty_cond.notify_all();


Will notify_one work here?

Here and below

i had idea to upgrade fetch_bitstream for multithreading version, if required. but not for now.

Is there any drawback for notify all instead of one?

I don't think more than one thread should wake in your case, since others will be still blocked if I get it correctly. There is no major difference though. Feel free to resolve this comment

i do not want to debug code to find when it get stucked when something changed, semantically close reader means broadcast

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.hpp

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

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

sivanov-work · 2021-11-09T05:33:46Z

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

+        GAPI_LOG_WARNING(nullptr, "[" << this << "] " <<
+                                  "\"mfxImplDescription.mfxDecoderDescription.decoder.CodecID\" "
+                                  "is absent, total param count" << codec_params.size());
+        return;


must throw unsupported exception instead of return

sivanov-work · 2021-11-09T05:36:35Z

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

+        try {
+            provider = std::make_shared<MFPAsyncDemuxDataProvider>(file_path);
+            GAPI_LOG_INFO(nullptr, "MFP data provider created");
+        } catch (const DataProviderUnsupportedException& ex) {


it is supposed to throw Unsupported exception at least. Any other kind of exception must be considered as unpredictable error and must be not catched here but deliver to application level (missing file, access rights etc)

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

sivanov-work · 2021-11-09T05:45:03Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+    if (!is_stream_selected) {
+        GAPI_LOG_WARNING(nullptr, "[" << this << "] couldn't find video stream with supported params");
+        // TODO print codec supported list
+        throw DataProviderUnsupportedException("couldn't find supported video stream");


Need to print it ultimately

Or video codec?

stream is suppose :)
if only audio stream is present in container then it is unsupported stream, not codec

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.hpp

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

sivanov-work · 2021-11-09T07:10:05Z

modules/gapi/test/streaming/gapi_streaming_vpl_data_provider.cpp

+
+TEST_P(OneVPL_Source_MFPDispatcherTest, open_and_decode_file)
+{
+    if (!initTestDataPathSilent()) {


must be removed as agreement of opecv test implementation

sivanov-work · 2021-11-09T07:10:18Z

modules/gapi/test/streaming/gapi_streaming_vpl_data_provider.cpp

+    }
+    using namespace cv::gapi::wip::onevpl;
+
+    source_t path = findDataFile(std::get<0>(GetParam()), false);


must be always true

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

smirnov-alexey

Overall looks good to me. Please, seek additional reviewers

AsyaPronina

Part of files are only reviewed, will proceed tomorrow

AsyaPronina · 2021-11-15T14:33:09Z

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

+     * declare untagged struct mfxBitstream.
+     *
+     * IDataProvider makes sense only for HAVE_VPL is ON and to keep IDataProvider
+     * interface API/ABI compiant between core library and user application layer


compiant -> compliant

AsyaPronina · 2021-11-15T14:50:17Z

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

     */
-    virtual size_t fetch_data(size_t out_data_bytes_size, void* out_data) = 0;
+    virtual bool fetch_bitstream_data(std::shared_ptr<mfx_bitstream> &in_out_bitsream) = 0;


it is intended to be not copied, isn't it?

it's filled by Microsoft Foundation Primitives implementation there is no clear point: copy or not - it depends.
Anyway, this huge work happens in different thread: MFP async call, but current thread which is caller of fetch_bitstream_data just pulls for cooked up bitstream

AsyaPronina · 2021-11-15T14:51:26Z

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

@@ -7,10 +7,11 @@
 #ifndef GAPI_STREAMING_ONEVPL_ONEVPL_DATA_PROVIDER_INTERFACE_HPP
 #define GAPI_STREAMING_ONEVPL_ONEVPL_DATA_PROVIDER_INTERFACE_HPP
 #include <exception>
+#include <limits>


is it used somewhere?

it was, but no anymore. will be deleted

AsyaPronina · 2021-11-15T14:55:36Z

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

+#else // HAVE_ONEVPL
+struct IDataProvider::mfx_bitstream {
+    mfx_bitstream() {
+        GAPI_Assert(false && "Reject to create `mfxBitstream` till library compiled without VPL/MFX support");


Might be to replace till with more common word - because?

AsyaPronina · 2021-11-15T14:56:59Z

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

+namespace onevpl {
+
+IDataProvider::Ptr DataProviderDispatcher::create(const std::string& file_path,
+                                                  const std::vector<CfgParam> cfg_params) {


thanks, i missed it

AsyaPronina · 2021-11-15T16:54:56Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+
+        ComPtrGuard<IMFStreamDescriptor> stream_descriptor =
+                                    createCOMPtrGuard(stream_descriptor_tmp);
+        is_stream_selected = false; // deselect until find supported stream found


find is extra here

AsyaPronina · 2021-11-15T16:59:58Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+                          " - " << is_codec_supported)
+        } else {
+            GAPI_LOG_WARNING(nullptr, "[" << this << "] " <<
+                                      "Cannot media GUID subtype for stream by index: " <<


Cannot get ...

AsyaPronina · 2021-11-15T17:01:53Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+    if (!is_stream_selected) {
+        GAPI_LOG_WARNING(nullptr, "[" << this << "] couldn't find video stream with supported params");
+        // TODO print codec supported list
+        throw DataProviderUnsupportedException("couldn't find supported video stream");


Or video codec?

AsyaPronina · 2021-11-15T17:04:37Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+    hr = MFCreateSourceReaderFromMediaSource(source.get(), attributes.get(),
+                                             &source_reader_tmp);
+    if (FAILED(hr)) {
+        throw DataProviderSystemErrorException(HRESULT_CODE(hr), "Cannot create MFCreateSourceReaderFromMediaSource");


"Cannot create MFCreateSourceReaderFromMediaSource" -> "Cannot create IMFSourceReader"

AsyaPronina · 2021-11-15T17:08:25Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+        // Ask for the first sample.
+        request_next(hr, 0, 0);
+    } else {
+        throw DataProviderSystemErrorException(HRESULT_CODE(hr), "Cannot ReadSample");


"Cannot ReadSample"" or something which is already handled on the 457 line?

if (FAILED(hr)) { throw DataProviderSystemErrorException(HRESULT_CODE(hr), "Cannot create MFCreateSourceReaderFromMediaSource");

This is confusing. How can we reach this else statement at all?

agree, thanks

sivanov-work · 2021-11-16T13:37:40Z

UTs failed on CI, waiting #20995 to be merged

dmatveev · 2021-11-18T19:36:57Z

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

@@ -29,6 +29,23 @@ struct GAPI_EXPORTS DataProviderSystemErrorException : public DataProviderExcept
    std::string reason;
 };

+struct GAPI_EXPORTS DataProviderUnsupportedException : public DataProviderException {


should this be final?

i don't know at the moment, maybe

dmatveev · 2021-11-18T19:37:22Z

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

+    virtual const char* what() const noexcept override;
+
+private:
+    std::string reason;


Should reason be part of DataProviderException?

two commenters ask the same: then ok will move it to base

dmatveev · 2021-11-18T19:38:47Z

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

+        GAPI_Assert(false && "Reject to create `mfxBitstream` because library compiled without VPL/MFX support");
+    }
+};
+#endif // HAVE_ONEVPL


should the whole file body be under one big #ifdef?

I believe not, because mfxBitsream is VPL specific type and should not be exposed in public headers
so for NO_VPL configuration we have to have stub

Otherwise, it would have namespace-open-close too much

Another way which I considered before - it move all IDataProvider interface declaration & definition into lib private scope. but want to preserve customization from public scope in data provider implementaiton

In NO_VPL this whole streaming/onevpl directory is excluded isn't it?

actually no - it activates some stub-like public classed which throws assert on using
here is just wrapper on VPL bitstream which figured as public interface item

dmatveev · 2021-11-18T19:39:40Z

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

+
+    // Look-up CodecId from input params
+    // If set then raw data provider is preferred
+    GAPI_LOG_DEBUG(nullptr, "try find explicit cfg param\"mfxImplDescription.mfxDecoderDescription.decoder.CodecID\"");


should this sting be defined only once somewhere?

it make sense

will add in similar way as queue_capacity in future PRs
https://github.com/opencv/opencv/pull/21049/files#diff-3b97bff710e57efd4529da47be268daad0dddeeda82225e6354c18b93808b786R63

dmatveev · 2021-11-18T19:42:02Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+        if (FAILED(hr)) {
+            GAPI_LOG_WARNING(nullptr, "Cannot create MF_RESOLUTION_MEDIASOURCE from URI: " <<
+                                      url << ". Abort");
+            pSourceResolver->Release();


any RAII available here?

yes, thanks it was preserved from prototype version

dmatveev · 2021-11-18T19:48:31Z

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

+    if (!out_bitstream) {
+        out_bitstream = std::make_shared<mfx_bitstream>();
+        out_bitstream->MaxLength = bitstream_data_size;
+        out_bitstream->Data = (mfxU8 *)calloc(out_bitstream->MaxLength, sizeof(mfxU8));


if mxf_bitstream structure is yours, I'd avoid unnecessary allocations outside of it

If it is not, I'd find a way to associate an automatic buffer with a wrapper over this structure and set Data to that buffer pointer.

second option:
it allocates at once for whole source, effort is more that profit

allocating it once for the whole source changes the overall allocation model, what I ask for above is just making what's now automatic

I agree, that it would be quite worth for regular usage, but otherwise it doesn't use any pre-allocation in case of async demux, because it uses mapping from MFP buffers, so byte-receiving constructor would be useless for this case.
From my perspective, implementing smart-class mxf_bitstream for both cases looks redundant at the moment and should be done by demand when it would be used more than in 1-2 places

dmatveev · 2021-11-18T19:49:08Z

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

+            throw std::runtime_error("Cannot allocate bitstream.Data bytes: " +
+                                     std::to_string(out_bitstream->MaxLength * sizeof(mfxU8)));


..so using new could save you from extra handling like this

dmatveev · 2021-11-18T19:49:46Z

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

+    GAPI_LOG_DEBUG(nullptr, "bitstream after fetch, DataOffset: " << out_bitstream->DataOffset <<
+                            ", DataLength: " << out_bitstream->DataLength);


do we need that extensive debug logging here?

i used it to find no fully drained data problem and unconsumed bytes
In release is no Debug anyway isn't? moreover if(level==Debug) is nothing with syscall fread time

dmatveev · 2021-11-18T19:50:28Z

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

+    source_handle(nullptr, &fclose),
+    codec(std::numeric_limits<mfx_codec_id_type>::max()),
+    bitstream_data_size(bitstream_data_size_value) {
+    GAPI_Assert(false && "Unsupported: G-API compiled without `WITH_GAPI_ONEVPL=ON`");


in fact it is OpenCV compiled, maybe we should fix this message text elsewhere

it make sense. Will rename it when WITH_GAPI_ONEVPL tnasfrormed into WITH_MFX due to custom builders turning on

dmatveev · 2021-11-18T19:51:49Z

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

+const char* mfx_codec_id_to_cstr(mfxU32 mfx_id) {
+    switch(mfx_id) {
+        case MFX_CODEC_AVC:
+            return "MFX_CODEC_AVC";
+        case MFX_CODEC_HEVC:
+            return "MFX_CODEC_HEVC";
+        case MFX_CODEC_MPEG2:
+            return "MFX_CODEC_MPEG2";
+        case MFX_CODEC_VC1:
+            return "MFX_CODEC_VC1";
+        case MFX_CODEC_VP9:
+            return "MFX_CODEC_VP9";
+        case MFX_CODEC_AV1:
+            return "MFX_CODEC_AV1";
+        case MFX_CODEC_JPEG:
+            return "MFX_CODEC_JPEG";
+        default:
+            return "<unsupported>";
+    }
+}


A simple define cuts this code in half

let's waiting for DX11 PR...

AsyaPronina · 2021-11-18T08:40:06Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+        worker_key_to_buffer_mapping_storage.clear();
+    }
+
+    GAPI_LOG_INFO(nullptr, "Clean didn't used up storage, count: " <<


"Clean didn't used up storage, count: " -> "Clean up didn't used storage, count: " ?

use up -> phrasal vers -> drain to empty

Ok, let's make it quite simple:
Clean data storage, elapsed count:

AsyaPronina · 2021-11-18T08:54:48Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+            // Get the video frame buffer from the sample.
+            IMFMediaBuffer *buffer_ptr = nullptr;
+            hr = sample_ptr->ConvertToContiguousBuffer(&buffer_ptr);
+            if (FAILED(hr)) {


This check is extra

AsyaPronina · 2021-11-18T08:55:31Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+            IMFMediaBuffer *buffer_ptr = nullptr;
+            hr = sample_ptr->ConvertToContiguousBuffer(&buffer_ptr);
+            if (FAILED(hr)) {
+                GAPI_Assert(SUCCEEDED(hr) && "MFPAsyncDemuxDataProvider::OnReadSample");


Or you just may write GAPI_Assert(false && ...)
Message shouldn't contain a lot of details, am I right?

AsyaPronina · 2021-11-18T09:25:46Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+
+            hr = buffer_ptr->Lock(&staging_stream->Data, &max_buffer_size, &curr_size);
+            if (FAILED(hr)) {
+                GAPI_Assert(SUCCEEDED(hr) && "MFPAsyncDemuxDataProvider::OnReadSample");


You used SUCCEEDED(hr) just for double-check?
Thing that the message is the same as in error above is intended?

it seems like copy-paste. thanks

AsyaPronina · 2021-11-18T17:09:01Z

modules/gapi/src/streaming/onevpl/demux/async_mfp_demux_data_provider.cpp

+                                      "iteration: " << iterations);
+        } else {
+            GAPI_LOG_DEBUG(nullptr, "[" << this << "] is waiting for flush finishing, "
+                                "iteration: " << iterations);


AsyaPronina · 2021-11-18T19:47:32Z

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

@@ -17,13 +21,18 @@ namespace onevpl {
 struct FileDataProvider : public IDataProvider {

    using file_ptr = std::unique_ptr<FILE, decltype(&fclose)>;
-    FileDataProvider(const std::string& file_path);
+    FileDataProvider(const std::string& file_path,
+                     const std::vector<CfgParam> codec_params = {},


const std::vector<CfgParam>&

AsyaPronina · 2021-11-18T20:15:43Z

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

@@ -61,10 +62,15 @@ mfxResourceType cstr_to_mfx_resource_type(const char* cstr);

 mfxU32 cstr_to_mfx_codec_id(const char* cstr);

+const char* mfx_codec_id_to_cstr(mfxU32 mfx_id);
+
+const std::set<mfxU32> &get_supported_mfx_codec_id();


I would write get_supported_mfx_codec_ids, but it is up to you

AsyaPronina · 2021-11-18T22:24:17Z

modules/gapi/test/streaming/gapi_streaming_vpl_data_provider.cpp

+    }
+
+    EXPECT_TRUE(dd_result);
+    EXPECT_TRUE(std::dynamic_pointer_cast<MFPAsyncDemuxDataProvider>(provider_ptr));


might be != nullptr?

AsyaPronina · 2021-11-18T22:31:20Z

modules/gapi/test/streaming/gapi_streaming_vpl_data_provider.cpp

+        size_t total_produced_count = 0;
+        for (size_t i = 0; i < produce_buffer_count; i ++) {
+            std::shared_ptr<IDataProvider::mfx_bitstream> dummy_stream =
+                                std::make_shared<IDataProvider::mfx_bitstream>();


AsyaPronina · 2021-11-18T22:45:53Z

modules/gapi/test/streaming/gapi_streaming_vpl_data_provider.cpp

+}
+
+class OneVPL_Source_MFPAsyncDispatcherTest : public ::testing::TestWithParam<array_element_t> {};
+TEST_P(OneVPL_Source_MFPAsyncDispatcherTest, open_and_decode_file)


I don't understand differences with the first OneVPL_Source_MFPDispatcherTest :(
Could you please help?

retained from prev version for sync demux

sivanov-work · 2021-11-22T08:30:03Z

@alalek could you merge it please?
one more PR (#21049) and then I'll try WITH_MFX=ON with further integration PR

G-API: oneVPL Implement asynchronous MFP demux data provider * Add dummy dmux * Initial commit for draft versionn * Demux for low res file works * Add media source resolver to work over incorrect MIME * Add MFP Demux logger * stash changes * Extend IDataProvider with CodecId, Add troubleshooting info * Add IDapaProvider dispatcher * Add ComPtrGuard wrappers * Add new unit test scope for MFP demux & Add minor changes * Enhance UTs * Remove ATL header * Remove ATL another one * Fix build * Add static for some methods * Initial commit * Add async demuxing * Apply tdd idea * Intro IDataProvider changes: +fetch_bitstream, -fetch_data * Fix UTs * Remove IDataProvider::CodecId & Fix EOF hang * Remove sync demux * Remove mfp async dependencies * Remove VPL dependencies from IDataProvider declaration * Apply comments * Fix compilation * Suppress unused warning * Apply some comments * Apply some comments * Apply comments

sivanov-work added 23 commits November 2, 2021 15:58

Add dummy dmux

436ae78

Initial commit for draft versionn

0bf7e84

Demux for low res file works

644752e

Add media source resolver to work over incorrect MIME

dcc9956

Add MFP Demux logger

004d218

stash changes

2ff847b

Extend IDataProvider with CodecId, Add troubleshooting info

6b89257

Add IDapaProvider dispatcher

cb1c51f

Add ComPtrGuard wrappers

190123e

Add new unit test scope for MFP demux & Add minor changes

a002136

Enhance UTs

3c14a53

Remove ATL header

fb53ce9

Remove ATL another one

cbf11f8

Fix build

b03c979

Add static for some methods

21b9703

Initial commit

98afb86

Add async demuxing

ca14bcc

Apply tdd idea

a6f4775

Intro IDataProvider changes: +fetch_bitstream, -fetch_data

1897fd2

Fix UTs

1a7d215

Remove IDataProvider::CodecId & Fix EOF hang

8bcec43

Remove sync demux

8688e7c

Remove mfp async dependencies

a4bf60b

sivanov-work added category: g-api / gapi feature labels Nov 8, 2021

sivanov-work requested review from AsyaPronina, smirnov-alexey, TolyaTalamanov and dmatveev November 8, 2021 11:49

sivanov-work mentioned this pull request Nov 8, 2021

G-API: oneVPL - Implement simple MFP demux provider with auto provider selector #20926

Closed

6 tasks

sivanov-work changed the base branch from master to 4.x November 8, 2021 13:02

smirnov-alexey reviewed Nov 8, 2021

View reviewed changes

modules/gapi/include/opencv2/gapi/streaming/onevpl/data_provider_interface.hpp Outdated Show resolved Hide resolved

smirnov-alexey requested changes Nov 8, 2021

View reviewed changes

sivanov-work commented Nov 9, 2021

View reviewed changes

sivanov-work added 2 commits November 9, 2021 16:38

Remove VPL dependencies from IDataProvider declaration

6aee91a

Apply comments

6dab91c

sivanov-work requested a review from smirnov-alexey November 10, 2021 07:35

sivanov-work added 2 commits November 10, 2021 10:41

Fix compilation

3f2a3f3

Suppress unused warning

c6f519d

smirnov-alexey approved these changes Nov 11, 2021

View reviewed changes

sivanov-work removed the request for review from TolyaTalamanov November 15, 2021 09:57

AsyaPronina reviewed Nov 15, 2021

View reviewed changes

sivanov-work added 2 commits November 16, 2021 16:38

Apply some comments

1964208

Apply some comments

048d630

dmatveev self-assigned this Nov 18, 2021

dmatveev added this to the 4.5.5 milestone Nov 18, 2021

dmatveev reviewed Nov 18, 2021

View reviewed changes

AsyaPronina reviewed Nov 18, 2021

View reviewed changes

Apply comments

2f2b539

sivanov-work requested review from AsyaPronina and dmatveev November 19, 2021 12:53

dmatveev approved these changes Nov 22, 2021

View reviewed changes

alalek merged commit 02f0887 into opencv:4.x Nov 22, 2021

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

sivanov-work deleted the async_mfp_demux branch March 5, 2022 05:48

		throw std::runtime_error("Cannot allocate bitstream.Data bytes: " +
		std::to_string(out_bitstream->MaxLength * sizeof(mfxU8)));

		GAPI_LOG_DEBUG(nullptr, "bitstream after fetch, DataOffset: " << out_bitstream->DataOffset <<
		", DataLength: " << out_bitstream->DataLength);

Uh oh!

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

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

Uh oh!

Conversation

sivanov-work commented Nov 8, 2021 • edited by alalek Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Readiness Checklist

Build Configuration

Uh oh!

alalek commented Nov 8, 2021

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!

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!

Uh oh!

Uh oh!

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!

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!

smirnov-alexey left a comment

Choose a reason for hiding this comment

Uh oh!

AsyaPronina 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!

sivanov-work commented Nov 8, 2021 •

edited by alalek

Loading

AsyaPronina Nov 15, 2021 •

edited

Loading