Android Camera原理之camx hal架构

come from : https://www.jianshu.com/p/cfb1da9d4217

之前看过F7A与D2T的camera hal层源码,这两个手机的camera hal层源码结构是相当清晰的,都是放在hardware/目录中,就是通常说说的QCamera2架构,整体结构也比较简单,脉络比较清晰,但是代码糅合在一起,看上去很繁杂,分层思想体现的不明显。

目前主流的机型都使用camx架构,这个架构和之前架构的主要区别就是 芯片接口层的代码从hardware/qcom 迁移到 vendor/qcom/proprietary/下面我们主要关注的camera hal层的源码也是放在vendor/qcom/proprietary/camx/下面。

下面我们简单梳理一下各层次之间的结构图。

camx_arch.jpg

camx 编译:

首先在camx架构上,camx的核心目录是 vendor/qcom/proprietary/camx/src/ 目录下面:

drwxrwxr-x  9 jeffmony jeffmony 4096 12月 25 20:13 .
drwxrwxr-x  5 jeffmony jeffmony 4096 12月 27 11:00 ..
drwxrwxr-x  8 jeffmony jeffmony 4096 12月 25 20:25 core
drwxrwxr-x  7 jeffmony jeffmony 4096 12月 25 20:13 csl
drwxrwxr-x 14 jeffmony jeffmony 4096 12月 25 20:13 hwl
drwxrwxr-x  3 jeffmony jeffmony 4096 12月 25 20:13 lib
drwxrwxr-x  3 jeffmony jeffmony 4096 12月 25 20:13 osutils
drwxrwxr-x 11 jeffmony jeffmony 4096 12月 25 20:13 swl
drwxrwxr-x  4 jeffmony jeffmony 4096 12月 25 20:13 utils

核心的Android.mk在 ./lib/build/android/Android.mk 中。

其中包括的 静态库如下:这些静态库都是camx或者其他的目录下编译的,编译工程的时候,我们要先编译这些静态库,然后编译camx的动态库。动态库的名称是/vendor/lib/hw/camera.qcom.so

# Libraries to link
LOCAL_STATIC_LIBRARIES :=   \
    libcamxcore             \
    libcamxchi              \
    libcamxcsl              \
    libcamxofflinestats     \
    libnc                   \
    libcamxncs              \
    libstriping
 
LOCAL_WHOLE_STATIC_LIBRARIES := \
    libcamxdspstreamer          \
    libcamxhwlbps               \
    libcamxgenerated            \
    libcamxhal                  \
    libcamxhalutils             \
    libcamxhwlfd                \
    libcamxhwlife               \
    libcamxhwlipe               \
    libcamxhwliqmodule          \
    libcamxswlfdmanager         \
    libcamxswljpeg              \
    libcamxhwljpeg              \
    libcamxhwllrme              \
    libcamxswlransac            \
    libcamxhwltitan17x          \
    libcamxiqsetting            \
    libcamxosutils              \
    libcamxstats                \
    libcamxsensor               \
    libcamxutils

camera.provider中如何实现到camera hal层的跳跃,camera service调用到camera provider中的接口方法,现在调用到 camera provider中的 hardware/interfaces/camera/device/3.2/default/CameraDeviceSession.cpp 中的processCaptureRequest(...)方法,最终会调用到:

status_t ret = mDevice->ops->process_capture_request(mDevice, &halRequest);

这个mDevice->ops 就是 hardware/libhardware/include/hardware/camera3.h 中的 camera3_device_ops 结构体:

https://www.jianshu.com/p/099cc3b0ab25

typedef struct camera3_device_ops {
    int (*initialize)(const struct camera3_device *,
            const camera3_callback_ops_t *callback_ops);
    int (*configure_streams)(const struct camera3_device *,
            camera3_stream_configuration_t *stream_list);
    int (*register_stream_buffers)(const struct camera3_device *,
            const camera3_stream_buffer_set_t *buffer_set);
    const camera_metadata_t* (*construct_default_request_settings)(
            const struct camera3_device *,
            int type);
    int (*process_capture_request)(const struct camera3_device *,
            camera3_capture_request_t *request);
    void (*get_metadata_vendor_tag_ops)(const struct camera3_device*,
            vendor_tag_query_ops_t* ops);
    void (*dump)(const struct camera3_device *, int fd);
    int (*flush)(const struct camera3_device *);
 
    /* reserved for future use */
    void *reserved[8];
} camera3_device_ops_t;

这样找到在camera hal层的函数指针的映射关系。

映射到:vendor/qcom/proprietary/camx/src/core/hal/camxhal3entry.cpp 中的:

// Global dispatch
static Dispatch g_dispatchHAL3(&g_jumpTableHAL3);

看一下g_jumpTableHAL3 变量:在 vendor/qcom/proprietary/camx/src/core/hal/camxhal3.cpp 中定义的:


// Jump table for HAL3

JumpTableHAL3 g_jumpTableHAL3 =
{
    open,
    get_number_of_cameras,
    get_camera_info,
    set_callbacks,
    get_vendor_tag_ops,
    open_legacy,
    set_torch_mode,
    init,
    parallelQuery,
    setCallBack,
    get_tag_count,
    get_all_tags,
    get_section_name,
    get_tag_name,
    get_tag_type,
    close,
    initialize,
    configure_streams,
    construct_default_request_settings,
    process_capture_request,
    dump,
    flush,
    camera_device_status_change,
    torch_mode_status_change,
    process_capture_result,
    notify
};

这儿直接构成了指针函数的映射关系。

vendor/qcom/proprietary/camx/src/core/chi/camxchitypes.h中定义了CHIAppCallbacks结构体,如下:

struct CHIAppCallbacks
{
    /// @brief Called by the driver to get number of cameras
    INT(*CHIGetNumCameras)(
        UINT32* pNumFwCameras,
        UINT32* pNumLogicalCameras);
 
    /// @brief Called by the driver to get the camera info for the camera id
    CamxResult (*CHIGetCameraInfo)(
        UINT32      cameraId,
        CameraInfo* pCameraInfo);
 
    /// @brief Defines the prototype for the device status change callback method from to the framework. Please refer to
    ///        the camera_device_status_change documentation in hardware/camera_common.h.
    VOID (*CHIInitializeOverrideSession)(
        UINT32               cameraId,
        const Camera3Device* pCamera3Device,
        const HALCallbacks*  pHALCallbacks,
        Camera3StreamConfig* pStreamConfig,
        BOOL*                isOverrideEnabled,
        VOID**               ppPrivate);
 
    /// @brief Defines the prototype for the torch mode status change callback method from to the framework. Please refer to
    ///        the torch_mode_status_change documentation in hardware/camera_common.h.
    VOID (*CHIFinalizeOverrideSession)(
        const Camera3Device* pCamera3Device,
        UINT64*              pSession,
        VOID**               ppPrivate);
 
    /// @brief Called by the driver to inform about session closing
    VOID (*CHITeardownOverrideSession)(
        const Camera3Device* pCamera3Device,
        UINT64*              pSession,
        VOID*                pPrivate);
 
    /// @brief Called by the driver to pass on capture request call to CHI
    INT (*CHIOverrideProcessRequest)(
        const Camera3Device*    pCamera3Device,
        Camera3CaptureRequest*  pCaptureRequest,
        VOID*                   pPrivate);
 
    /// @brief Called by the driver to allow for additional override processing during open()
    INT(*CHIExtendOpen)(
        UINT32  cameraId,
        VOID*   pPrivateData);
 
    /// @brief Called by the driver to allow for additional override processing during close()
    INT(*CHIExtendClose)(
        UINT32  cameraId,
        VOID*   pPrivateData);
 
    /// @brief Called by the driver to allow override to remap special camera IDs into logical camera IDs
    UINT32(*CHIRemapCameraId)(
        UINT32              frameworkCameraId,
        CameraIdRemapMode   mode);
 
    /// @brief Interface to allow various override-specific settings to be toggled.
    UINT32(*CHIModifySettings)(
        VOID*   pPrivateData);
 
    /// @brief Get any vendor tag specific request settings the override wants to get added to the default settings
    VOID (*CHIGetDefaultRequestSettings)(
        UINT32           frameworkCameraId,
        INT              requestTemplate,
        const Metadata** pAdditionalMetadata);
 
    /// @brief Called by the driver to allow for flush()
    INT(*CHIOverrideFlush)(
        const Camera3Device*    pCamera3Device);
    INT(*CHIParallelQuery) (INT num, char* list[]);
    INT(*CHISetCallback) (void*);
 
};
 
typedef VOID(*CHIHALOverrideEntry)(
    CHIAppCallbacks* pCHIAppCallbacks);

这个结构体是函数指针,如何映射的,看看下面的关系:

vendor/qcom/proprietary/camx/src/core/hal/camxhal3module.h中定义的CHIAppCallbacks m_ChiAppCallbacks;

CHIAppCallbacks m_ChiAppCallbacks; ///< CHI HAL override entry

在 vendor/qcom/proprietary/camx/src/core/hal/camxhal3module.cpp中的 HAL3Module构造函数中,存在下面的执行语句:

CHIHALOverrideEntry funcCHIHALOverrideEntry =
    reinterpret_cast<CHIHALOverrideEntry>(
        CamX::OsUtils::LibGetAddr(m_hChiOverrideModuleHandle, "chi_hal_override_entry"));
 
if (NULL != funcCHIHALOverrideEntry)
{
    funcCHIHALOverrideEntry(&m_ChiAppCallbacks);
 
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIGetNumCameras);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIGetCameraInfo);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIFinalizeOverrideSession);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIInitializeOverrideSession);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIOverrideProcessRequest);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIOverrideFlush);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHITeardownOverrideSession);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIExtendOpen);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIExtendClose);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIRemapCameraId);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIModifySettings);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHIParallelQuery);
    CAMX_ASSERT(NULL != m_ChiAppCallbacks.CHISetCallback);
 
    if ((NULL != m_ChiAppCallbacks.CHIGetNumCameras)             &&
        (NULL != m_ChiAppCallbacks.CHIGetCameraInfo)             &&
        (NULL != m_ChiAppCallbacks.CHIFinalizeOverrideSession)   &&
        (NULL != m_ChiAppCallbacks.CHIInitializeOverrideSession) &&
        (NULL != m_ChiAppCallbacks.CHIOverrideProcessRequest)    &&
        (NULL != m_ChiAppCallbacks.CHIOverrideFlush)             &&
        (NULL != m_ChiAppCallbacks.CHITeardownOverrideSession)   &&
        (NULL != m_ChiAppCallbacks.CHIExtendOpen)                &&
        (NULL != m_ChiAppCallbacks.CHIExtendClose)               &&
        (NULL != m_ChiAppCallbacks.CHIRemapCameraId)             &&
        (NULL != m_ChiAppCallbacks.CHIModifySettings)            &&
        (NULL != m_ChiAppCallbacks.CHIParallelQuery)             &&
        (NULL != m_ChiAppCallbacks.CHISetCallback))
    {
        CAMX_LOG_WARN(CamxLogGroupHAL, "CHI Module library function pointers exchanged");
    }
}

m_ChiAppCallbacks 通过 funcCHIHALOverrideEntry 映射到 chi_hal_override_entry
这个 chi_hal_override_entry 就是指 vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxextensioninterface.cpp 中的 chi_hal_override_entry 函数,如下:


void chi_hal_override_entry(
    chi_hal_callback_ops_t* callbacks)
{
    ExtensionModule* pExtensionModule = ExtensionModule::GetInstance();
 
    CHX_ASSERT(NULL != callbacks);
 
    if (NULL != pExtensionModule)
    {
        callbacks->chi_get_num_cameras              = chi_get_num_cameras;
        callbacks->chi_get_camera_info              = chi_get_camera_info;
        callbacks->chi_initialize_override_session  = chi_initialize_override_session;
        callbacks->chi_finalize_override_session    = chi_finalize_override_session;
        callbacks->chi_override_process_request     = chi_override_process_request;
        callbacks->chi_teardown_override_session    = chi_teardown_override_session;
        callbacks->chi_extend_open                  = chi_extend_open;
        callbacks->chi_extend_close                 = chi_extend_close;
        callbacks->chi_remap_camera_id              = chi_remap_camera_id;
        callbacks->chi_modify_settings              = chi_modify_settings;
        callbacks->chi_get_default_request_settings = chi_get_default_request_settings;
        callbacks->chi_override_flush               = chi_override_flush;
        callbacks->chi_parallelquery                = chi_parallelquery;
        callbacks->chi_setcallback                  = chi_setcallback;
    }
}

这样就建立了 CHIAppCallbacks 中函数指针的一一映射关系。

 

camx_code.jpg

 

vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxextensionmodule.cpp 中的 ExtensionModule::OverrideProcessRequest 函数中执行了 m_pUsecaseFactory->CreateUsecaseObject,如下:

m_pSelectedUsecase[logicalCameraId] =
            m_pUsecaseFactory->CreateUsecaseObject(&m_logicalCameraInfo[logicalCameraId],
                static_cast<UsecaseId>(m_SelectedUsecaseId[logicalCameraId]),
                m_pStreamConfig[logicalCameraId]);

直接调用到: vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxusecaseutils.cpp 中的 UsecaseFactory::CreateUsecaseObject 函数:

Usecase* UsecaseFactory::CreateUsecaseObject(
    LogicalCameraInfo*              pLogicalCameraInfo,     ///< camera info
    UsecaseId                       usecaseId,              ///< Usecase Id
    camera3_stream_configuration_t* pStreamConfig)          ///< Stream config
{
    Usecase* pUsecase  = NULL;
    UINT     camera0Id = pLogicalCameraInfo->ppDeviceInfo[0]->cameraId;
    CHX_LOG_ERROR("UsecaseFactory::CreateUsecaseObject id = %d", usecaseId);
    switch (usecaseId)
    {
        case UsecaseId::PreviewZSL:
            pUsecase = AdvancedCameraUsecase::Create(pLogicalCameraInfo, pStreamConfig, usecaseId);
            break;
        case UsecaseId::MultiCamera:
            pUsecase = UsecaseMultiCamera::Create(pLogicalCameraInfo, pStreamConfig);
            break;
        case UsecaseId::MultiCameraVR:
            pUsecase = UsecaseMultiVRCamera::Create(pLogicalCameraInfo, pStreamConfig);
            break;
        case UsecaseId::MFNR:
            pUsecase = UsecaseMFNR::Create(camera0Id, pStreamConfig);
            break;
        case UsecaseId::QuadCFA:
            pUsecase = UsecaseQuadCFA::Create(pLogicalCameraInfo, pStreamConfig);
            break;
        case UsecaseId::Torch:
            pUsecase = UsecaseTorch::Create(camera0Id, pStreamConfig);
            break;
        default:
            pUsecase = AdvancedCameraUsecase::Create(pLogicalCameraInfo, pStreamConfig, usecaseId);
            break;
    }
 
    return pUsecase;
}
enum class UsecaseId
{
    NoMatch         = 0,
    Default         = 1,
    Preview         = 2,
    PreviewZSL      = 3,
    MFNR            = 4,
    MFSR            = 5,
    MultiCamera     = 6,
    QuadCFA         = 7,
    RawJPEG         = 8,
    MultiCameraVR   = 9,
    Torch           = 10,
    YUVInBlobOut    = 11,
    MaxUsecases     = 12,
};

前置摄像头的UsecaseId是 PreviewZSL,是单摄,后置摄像头的UsecaseId是 MultiCamera,是多摄。

camx-usecase

 

camx_usecase.jpg

 

vendor/qcom/proprietary/camx/src/core/chi/camxchi.cpp中的 ChiEntry函数如下:


/// ChiEntry

CAMX_VISIBILITY_PUBLIC VOID ChiEntry(
    ChiContextOps* pChiContextOps)
{
    if (NULL != pChiContextOps)
    {
        pChiContextOps->size                       = sizeof(ChiContextOps);
 
        pChiContextOps->majorVersion               = CHI_API_MAJOR_VERSION;
        pChiContextOps->minorVersion               = CHI_API_MINOR_VERSION;
        pChiContextOps->pOpenContext               = CamX::ChiOpenContext;
        pChiContextOps->pCloseContext              = CamX::ChiCloseContext;
        pChiContextOps->pGetNumCameras             = CamX::ChiGetNumCameras;
        pChiContextOps->pGetCameraInfo             = CamX::ChiGetCameraInfo;
        pChiContextOps->pEnumerateSensorModes      = CamX::ChiEnumerateSensorModes;
        pChiContextOps->pCreatePipelineDescriptor  = CamX::ChiCreatePipelineDescriptor;
        pChiContextOps->pDestroyPipelineDescriptor = CamX::ChiDestroyPipelineDescriptor;
        pChiContextOps->pCreateSession             = CamX::ChiCreateSession;
        pChiContextOps->pDestroySession            = CamX::ChiDestroySession;
        pChiContextOps->pFlushSession              = CamX::ChiFlushSession;
        pChiContextOps->pActivatePipeline          = CamX::ChiActivatePipeline;
        pChiContextOps->pDeactivatePipeline        = CamX::ChiDeactivatePipeline;
        pChiContextOps->pSubmitPipelineRequest     = CamX::ChiSubmitPipelineRequest;
        pChiContextOps->pTagOps                    = CamX::ChiGetTagOps;
    }
 
    // This is the workaround for presil HAL3test on Windows
    // On Device, set_camera_metadata_vendor_ops will be call the set the
    // static vendor tag operation in camera_metadata.c
    //
    // On Windows side, theoretically hal3test should mimic what Android framework
    // does and call the set_camera_metadata_vendor_ops function in libcamxext library
    // However, in Windows, if both hal3test.exe and hal.dll link to libcamxext library,
    // there are two different instance of static varibles sit in different memory location.
    // Even if set_camera_metadata_vendor_ops is called in hal3test, when hal try to
    // access to vendor tag ops, it is still not set.
    //
    // This is also a workaround to call vendor tag ops in Chi at GetNumCameras which happens to get called before
    // GetVendorTagOps
    CamX::g_vendorTagOps.get_all_tags     = CamX::ChiGetAllTags;
    CamX::g_vendorTagOps.get_section_name = CamX::ChiGetSectionName;
    CamX::g_vendorTagOps.get_tag_count    = CamX::ChiGetTagCount;
    CamX::g_vendorTagOps.get_tag_name     = CamX::ChiGetTagName;
    CamX::g_vendorTagOps.get_tag_type     = CamX::ChiGetTagType;
 
    set_camera_metadata_vendor_ops(&(CamX::g_vendorTagOps));
}

这个函数映射关系很重要,也在camx chi中比较常见,直接映射在此文件的CamxChi类中。都是从 vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxextensionmodule.cpp中调用过来的。

下面是预览时capture request 处理流程图:

camx_pipeline流程.jpg


check这段流程的时候我们最关注应该是5个重要的处理类型:

 

  • 1.UseCase , vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxusecase.h 上面有介绍类图。UseCase在camx中很有很多衍生类,这是camx针对不同的stream来建立不同的usecase对象,用来管理选择feature,并且创建 pipeline以及session
  • 2.ChiFeature, vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxfeature.h, usecase选择相应的feature,关联一组pipeline,收到request请求,根据request选择对应的feature
  • 3.Node , vendro/qcom/propriatary/camx/src/core/camxnode.h ,下面有类图。Node是camx中非常重要的一个父类,是camx中处理camera 请求的一个中间节点,用于处理pipeline下发的请求,下面有类图介绍,比较重要**的Node子类已经标出来了。
  • 4.pipeline , 一连串node的集合,通过pipeline下发给各个node处理。
  • 5.session , 若干个有关联的pipeline的集合,用来管理pipeline,使用pipeline处理请求。
    Node 节点在camx chi架构中至关重要,数据的处理都是通过封装好的Node节点来进行的。

    camxnode结构图

camx_node.jpg

 

node节点的创建地方在vendor/qcom/proprietary/camx/src/hwl/titian17x/camxtitian17xfactory.cpp

Node* Titan17xFactory::HwCreateNode(
    const NodeCreateInputData* pCreateInputData,
    NodeCreateOutputData*      pCreateOutputData
    ) const
{
    Node* pNode = NULL;
 
    switch (pCreateInputData->pNodeInfo->nodeId)
    {
        case AutoFocus:
            pNode = AutoFocusNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case BPS:
            pNode = BPSNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case IFE:
            pNode = IFENode::Create(pCreateInputData, pCreateOutputData);
            break;
        case IPE:
            pNode = IPENode::Create(pCreateInputData, pCreateOutputData);
            break;
        case Sensor:
            pNode = SensorNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case StatsProcessing:
            pNode = StatsProcessingNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case JPEG:
            pNode = JPEGEncNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case JPEGAggregator:
            pNode = JPEGAggrNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case StatsParse:
            pNode = StatsParseNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case ChiExternalNode:
            pNode = ChiNodeWrapper::Create(pCreateInputData, pCreateOutputData);
            break;
        case FDHw:
            pNode = FDHwNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case FDManager:
            pNode = FDManagerNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case OfflineStats:
            pNode = OfflineStatsNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case Torch:
            pNode = TorchNode::Create(pCreateInputData, pCreateOutputData);
            break;
        case LRME:
            pNode = LRMENode::Create(pCreateInputData, pCreateOutputData);
            break;
        case RANSAC:
            pNode = RANSACNode::Create(pCreateInputData, pCreateOutputData);
            break;
        default:
            CAMX_ASSERT_ALWAYS_MESSAGE("Unexpected node type");
            break;
    }
 
    return pNode;
}

Feature类图.jpg

camx_feature.jpg

在 vendor/qcom/proprietary/chi-cdk/vendor/chioverride/default/chxadvancedcamerausecase.cpp中的SelectFeatures(...)函数中有这些feature的创建代码,接下来讲解拍照的时候会详细分析一下。

现在说下拍照的场景,拍照的场景分为前置和后置,前置是单摄,后置是多摄,前面也有介绍,单摄和多摄使用的usecase是不同的,我们先讲解一下单摄。

前置拍照创建的pipeline有:

MiuiZSLSnapshotJpeg at index 0 for session 0, session's pipeline 0, camera id:1
MiuiZSLPreviewRaw at index 1 for session 1, session's pipeline 0, camera id:1
BinningZSLYuv2Jpeg at index 2 for session 2, session's pipeline 0, camera id:1
BinningMerge3YuvCustomTo1Yuv at index 3 for session 3, session's pipeline 0, camera id:1
ZSLSnapshotYUV at index 4 for session 4, session's pipeline 0, camera id:1
AdvancedAsdMeta at index 5 for session 5, session's pipeline 0, camera id:1
SWMFClearShotYuv at index 6 for session 6, session's pipeline 0, camera id:1
BinningZSLSnapshotYUV at index 7 for session 7, session's pipeline 0, camera id:1

后置拍照创建的pipeline有:


BackCameraJpegEncode at index 0 for session 0, session's pipeline 0, camera id:0
MfnrPrefilter at index 1 for session 0, session's pipeline 1, camera id:0
MfnrBlend at index 2 for session 0, session's pipeline 2, camera id:0
MfnrPostFilter at index 3 for session 0, session's pipeline 3, camera id:0
MfnrScale at index 4 for session 0, session's pipeline 4, camera id:0
Merge3YuvCustomTo1Yuv at index 5 for session 1, session's pipeline 0, camera id:0
ZSLSnapshotYUV at index 6 for session 2, session's pipeline 0, camera id:0
ZSLSnapshotYUVAux at index 7 for session 3, session's pipeline 0, camera id:3
SWMFSRYuv at index 8 for session 4, session's pipeline 0, camera id:0
AdvancedAsdMeta at index 9 for session 5, session's pipeline 0, camera id:0

pipeline在camx中的配置文件是:vendor/qcom/proprietary/chi-cdk/vendor/topology/default/titan17x_usecases.xml

 

  • 3
    点赞
  • 22
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值