imx VPU解码分析1-wrap

主要有二个库，一个是imx-vpuwrap，另一个是imx-vpu-hantro。从资料上是这么描述的

我是这么理解的，imx-vpu-hantro是更底层的代码，imx-vpuwrap是封装层实现。

本篇介绍vpuwrap的解码，参阅vpu_wrapper.h。

1.基本的数据类型。

返回值。

VpuDecRetCode value	Description
VPU_DEC_RET_SUCCESS	Success.
VPU_DEC_RET_FAILURE	Fail.
VPU_DEC_RET_INVALID_PARAM	Parameter is invalid.
VPU_DEC_RET_INVALID_HANDLE	Handle is invalid.
VPU_DEC_RET_INVALID_FRAME_BUFFER	Frame buffer is invalid.
VPU_DEC_RET_INSUFFICIENT_FRAME_BUFFERS	Frame buffers is insufficient.
VPU_DEC_RET_INVALID_STRIDE	Stride is invalid.
VPU_DEC_RET_WRONG_CALL_SEQUENCE	State of the object is not correct.
VPU_DEC_RET_FAILURE_TIMEOUT	The wait for HW/SW finish has timed out.

主要关注这个结构体VpuDecBufRetCode，在解码过程中返回的最重要的信息。

VpuDecBufRetCode value	Description
VPU_DEC_INPUT_NOT_USED	Input data has been consumed.
VPU_DEC_INPUT_USED	Input data hasn’t been consumed.
VPU_DEC_OUTPUT_EOS	Got eos.
VPU_DEC_OUTPUT_DIS	Got one frame to output.
VPU_DEC_OUTPUT_NODIS	Got no frame to output.
VPU_DEC_OUTPUT_REPEAT	One frame is output repeatly.
VPU_DEC_OUTPUT_DROPPED	Got one frame to drop.
VPU_DEC_OUTPUT_MOSAIC_DIS	Got one mosaic frame to output.
VPU_DEC_NO_ENOUGH_BUF	No enough buffer to hold for output.
VPU_DEC_NO_ENOUGH_INBUF	No enough input buffer.
VPU_DEC_INIT_OK	Initialization of decoding is ok.
VPU_DEC_SKIP	Skip to decode one frame.
VPU_DEC_ONE_FRM_CONSUMED	One frame has been decoded.
VPU_DEC_RESOLUTION_CHANGED	Resolution of the frame is changed.
VPU_DEC_FLUSH	Flush the decoder.

VpuDecConfig的配置参数

VpuDecConfig value	Description
VPU_DEC_CONF_SKIPMODE = 0	parameter value: VPU_DEC_SKIPNONE(default); VPU_DEC_SKIPPB; VPU_DEC_SKIPB; VPU_DEC_SKIPALL; VPU_DEC_ISEARCH.
VPU_DEC_CONF_INPUTTYPE	parameter value: VPU_DEC_IN_NOMAL: normal(default); VPU_DEC_IN_KICK: kick -- input data/size in VPU_DecDecodeBuf() will be ignored; VPU_DEC_IN_DRAIN: drain -- stream reach end, and input data/size in VPU_DecDecodeBuf() will be ignored.
VPU_DEC_CONF_BUFDELAY	For stream mode. The parameter represents buf size(unit: bytes), buffer size==0 indicate no any delay.
VPU_DEC_CONF_INIT_CNT_THRESHOLD	At seqinit stage. vpu will report error if data count reach the threshold.
VPU_DEC_CONF_ENABLE_TILED	Configure output frame to tiled after parsed sequence header and before register frame buffer.

VpuColorFormat的格式，一般VPU解码出来是YUV420格式数据。

typedef enum

{

VPU_COLOR_420=0,

VPU_COLOR_422H=1,

VPU_COLOR_422V=2,

VPU_COLOR_444=3,

VPU_COLOR_400=4,

VPU_COLOR_422YUYV=13,

VPU_COLOR_422UYVY=14,

VPU_COLOR_ARGB8888=15,

VPU_COLOR_BGRA8888=16,

VPU_COLOR_RGB565=17,

VPU_COLOR_RGB555=18,

VPU_COLOR_BGR565=19,

}VpuColorFormat;

VpuCodStd的视频格式

typedef enum {

VPU_V_MPEG4 = 0,

VPU_V_DIVX3,

VPU_V_DIVX4,

VPU_V_DIVX56,

VPU_V_XVID,

VPU_V_H263,

VPU_V_AVC,        //h264

VPU_V_AVC_MVC,

VPU_V_VC1,

VPU_V_VC1_AP,

VPU_V_MPEG2,

VPU_V_RV,

VPU_V_MJPG,

VPU_V_AVS,

VPU_V_VP8,

VPU_V_VP9,

VPU_V_HEVC,        //h265

VPU_V_SORENSON,

VPU_V_VP6,

VPU_V_WEBP,

} VpuCodStd;

VpuDecOpenParam的打开参数

Members	Type	Description
CodecFormat	VpuCodStd	Type of video.
nReorderEnable	int	Whether to enable record.
nChromaInterleave	int	Should be set to 1 when (nMapType!=0).
nMapType	int	Registered frame buffer type: 0--linear; 1--frame tile; 2--field tile.
nTiled2LinearEnable	int	Output frame(only valid when nMapType!=0) : 0--tile; 1--linear(not supported).
nPicWidth	int	Width of the decoded image.
nPicHeight	int	Height of the decoded image.
nEnableFileMode	int	Whether filemode is enabled.
nEnableVideoCompressor	int	Whether video compressor is enabed.
nPixelFormat	int	Output 10 bit or cut to 8 bit for Hantro G2. 0 for output 10 bit, 1 for cut to 8 bit.
nAdaptiveMode	int	Whether in adaptive mode.
pAppCxt	void *	Reserved for future application extension.
nSecureMode	int	Whether in secure mode.
nSecureBufferAllocSize	int	Allocated buffer size in secure mode.

VpuFrameBuffer帧buffer基本结构体

Members	Type	Description
nStrideY	unsigned int	Luma stride information.
nStrideC	unsigned int	Chroma stride information.
pbufY	unsigned char *	Physical assress of luma frame pointer or top field pointer.
pbufCb	unsigned char *	Physical assress of chroma frame pointer or top field pointer.
pbufCr	unsigned char *
pbufMvCol	unsigned char *
pbufY_tilebot	unsigned char *	For field tile: physical assress of luma bottom pointer.
pbufCb_tilebot	unsigned char *	For field tile: physical assress of chroma bottom pointer.
pbufVirtY	unsigned char *	Virtual address of luma frame pointer or top field pointer.
pbufVirtCb	unsigned char *	Virtual address of chroma frame pointer or top field pointer.
pbufVirtCr	unsigned char *
pbufVirtMvCol	unsigned char *
pbufVirtY_tilebot	unsigned char *	For field tile: virtual address of luma bottom pointer.
pbufVirtCb_tilebot	unsigned char *	For field tile: virtual address of chroma bottom pointer.
nReserved[5]	int	Reserved for future extension.
pPrivate	void *	Reserved for future special extension.

上层的包装结构体VpuDecOutFrameInfo

Members	Type	Description
pDisplayFrameBuf	VpuFrameBuffer *	Pointer to VpuFrameBuffer struct which contains the display information of the frame.
ePicType	VpuPicType	Type of frame.
eFieldType	VpuFieldType	Field type of vpu.
nMVCViewID	int	Extended info: support dynamic resolution, ...
pExtInfo	VpuFrameExtInfo *	Luma offset.
nReserved[2]	int	Reserved for future extension.
pPrivate	void *	Reserved for future special extension.

VpuDecInitInfo一些初始化信息

Members	Type	Description
nPicWidth	int	Aligned width of image.
nPicHeight	int	Aligned height of image.
nFrameRateRes	int	Numerator of framerate.
nFrameRateDiv	int	Denominator of framerate.
PicCropRect	VpuRect	VpuRect struct that contains crop information of image.
nMinFrameBufferCount	int	Minimum frame buffer count in vpu.
nMjpgSourceFormat	int	Source color format of jpeg
nInterlace	int	Whether video is interlaced.
nQ16ShiftWidthDivHeightRatio	unsigned int	Fixed point for width/height: 1: 0x10000; 0.5: 0x8000;...
nConsumedByte	int	Reserved to record sequence length: value -1 indicate unknow.
nAddressAlignment	int	Address alignment for Y/Cb/Cr (unit: bytes).
nFrameSize	int	Hantro video decoder append DMV and compression table in pixel buffer.
nBitDepth	int	Bit depth of video.
nReserved[3]	int	Reserved for future extension.
pSpecialInfo	void *	Reserved for future special extension.
hasColorDesc	int	Whether has color description.
hasHdr10Meta	int	Whether has hdr10 meta data.
Hdr10Meta	VpuHDR10Meta	Hdr10Meta struct which contains hdr10 meta data.
ColourDesc	VpuColourDesc	ColourDesc struct which contains color description.
ChromaLocInfo	VpuChromaLocInfo	VpuChromaLocInfo struct which contains chroma information.

VpuCodecData私有解码数据，数据和大小。这个我的理解是给特定解码器用的。一般不需要使用到。

Members	Type	Description
pData	unsigned char *	Codec data virtual address.
nSize	unsigned int	Codec data length.

上一层的包装结构体VpuBufferNode，包含地址。这个结构体就是实质传入给解码器的。

nSize表示传入解码器的数据大小，pVirAddr表示传入解码器的数据指针。

Members	Type	Description
pPhyAddr	unsigned char *	Buffer physical base address.
pVirAddr	unsigned char *	Buffer virtual base address.
nSize	unsigned int	Length of data.
sCodecData	VpuCodecData	VpuCodecData struct that contains codec data information.
nReserved[2]	int	Reserved for future extension.
pPrivate	void *	Reserved for future special extension

2.基本API

a.VPU的打开和关闭

VPU_DecGetVersionInfo

 VPU_DecGetWrapperVersionInfo

 VPU_DecGetInitialInfo

VPU_DecConfig

VPU_DecOpen  ：打开一个 vpu handle

VPU_DecGetCapability

VPU_DecDisCapability

VPU_DecGetErrInfo

VPU_DecGetNumAvailableFrameBuffers 

VPU_DecLoad        ：加载VPU

VPU_DecUnLoad

VPU_DecReset

VPU_DecClose

VPU_DecFlushAll

b.VPU解码

VPU_DecDecodeBuf： 解码一帧数据，结果存储在pOutBufRetCode中。

VPU_DecGetOutputFrame：解码帧信息

VPU_DecGetConsumedFrameInfo：帧buffer信息

​​​​​​​VPU_DecOutFrameDisplayed：清理帧显示

c.内存申请和释放

VPU_DecQueryMem

VPU_DecGetMem

VPU_DecFreeMem

VPU_DecRegisterFrameBuffer

3.VPU解码流程

 完整步骤：

1.加载VPU

VPU_DecLoad()

2.获取vpulib和vpu wrapper版本信息

VPU_DecGetVersionInfo()

VPU_DecGetWrapperVersionInfo()

3.申请buffer

VPU_DecQueryMem()

4.分配内存给vpu

VPU_DecGetMem()

5.打开一个VPU解码器

VPU_DecOpen()

6.设置解码器

VPU_DecGetCapability()

VPU_DecConfig()

7.开始解码

VPU_DecDecodeBuf()

8.解码后，根据输出状态执行

VPU_DecGetOutputFrame()

VPU_DecGetConsumedFrameInfo()

VPU_DecFlushAll()

• If (pOutBufRetCode & VPU_DEC_ONE_FRM_CONSUMED) is not 0, then call VPU_DecGetConsumedFrameInfo() to get the frame information in vpu.
• If (pOutBufRetCode & VPU_DEC_OUTPUT_DIS) is not 0, then call VPU_DecGetOutputFrame() to get the output frame information.
• If (pOutBufRetCode & VPU_DEC_FLUSH) is not 0, then call VPU_DecFlushAll() to flush the decoder.

9.显示帧后将帧缓冲区释放

VPU_DecOutFrameDisplayed()

10.关闭VPU

VPU_DecClose()

VPU_DecUnLoad()

11.释放VPU内存

VPU_DecFreeMem()

实际流程中按需要处理。参考wrap的示例程序即可。

在后面的文章中描述了

imx VPU解码分析3-wrap的示例-CSDN博客