The build system may store those files into a package.
Change-Id: I2ab54da5880d1979a19430520d36324629b8a7d8
Signed-off-by: Randy Li <randy.li@rock-chips.com>
The kernel uses termination character to divisive compatible.
Change-Id: I5db14e22aefa1ea54b8d32a4af7b112f22358dfb
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
Signed-off-by: leo.ding <leo.ding@rock-chips.com>
The external buffer would request the index of a buffer to
manage it in an array.
Change-Id: Ifbb169c75dfb6342b428b1fd640db9ec45ff5f73
Signed-off-by: Randy Li <randy.li@rock-chips.com>
I also fix a number of compiler warnings in this commmit.
A funtion is forgot to be exported and some functions should not
be exported.
Change-Id: I945456721a7507d448cfa0559418c356c6cb6ace
Signed-off-by: Randy Li <randy.li@rock-chips.com>
scan the stream just for one time in jpegd parser and
remove redundant codes.
Change-Id: I72920b15fc4ed15cfa1de55d9ed249d50cceebb2
Signed-off-by: timkingh.huang <timkingh.huang@rock-chips.com>
Use mpp_device to replace VPUClient interface. Then libmpp.so can
work without libvpu.so. And libvpu.so is only for legacy libvpuapi.
Remove iommu detection function in mpp. All buffer register will be
send to kernel with the form of low 10 bit file handle plus 22 bit
offset mode.
Remove vpu.h/vpu_api.h from libmpp.so
NOTE: mpp_frame.h is updated for external user.
Change-Id: I7ff2effd3c73c042ed65896847fc7458da29fc55
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
New chipset RK3228/RK3229/RK3399/RV1108/RK3328 is fully supported.
Old chipset RK30xx/RK31xx/RK3288/RK3368 is partly supported:
H.264/mjpeg/vp8/H.265 are supported if the old chipset has.
Change-Id: I33b3bd3286ab0b373d73bce50c83ee51e754ff51
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
In the future, the legacy vpu api won't be supported and the wrapper library is
not offered anymore.
Change-Id: I4019d689693e193d66fc222cc3e5662d7d6cc58d
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
The extra info magic is missing, fix it.
Change-Id: I63510a3036404ade547e8781e36590a88d99b4ee
Signed-off-by: timkingh.huang <timkingh.huang@rock-chips.com>
When multiplying width by height is larger than 4194303 (2^22 - 1),
22 bits of a register are not enough to pass luma and chroma offset
to kernel. So we add extra info including correct offset to ioctl
command.
Change-Id: Ica5faf4d6c8b149a491552d7a9b626534866aa3f
Signed-off-by: timkingh.huang <timkingh.huang@rock-chips.com>
Tips: when flush, only output the frame which can be output,
not to clear all frames. Such as, when infochange, then flush
is called, while current frame is for reference and should be
reserved, it would be error if it be cleared.
Change-Id: Ie33b0f51bbe54c1b6cad6d1d569b3409ca6414a2
Signed-off-by: leo.ding <leo.ding@rock-chips.com>
It is the correct to store multi char variable in a variable.
Finally we can't get rid of warning.
NOTE: Need to check the macro under different compiler.
Change-Id: Id43fabd9a6ad437938ba590c95dc0552ef4e5699
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
The old Video IP like VDPU1 only supports align with 16 bytes,
it would causing a numbers of memory problem without that.
Change-Id: I7570808824ab2e32602c8cae7f97c9b6b2947db9
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
Change-Id: Iedc60e981dc5bee3d49a779c34d693998ad3499b
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
Change-Id: I52554a3ca45747c13a7821a04bb1c01e45d9d44c
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
Change-Id: I92e9c672fb89c717a70e874679b52700a9595b89
Signed-off-by: Randy Li <randy.li@rock-chips.com>
Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
When retrieving frames current API only allows to define wether
the API should block without any timeout or not block at all.
This allows to specify a block timeout via the
MPP_SET_OUTPUT_BLOCK_TIMEOUT control operation.
Change-Id: Id35ad3d48f72881184009f830c406c3a655e1a90
Signed-off-by: LongChair <LongChair@hotmail.com>
Signed-off-by: Randy Li <randy.li@rock-chips.com>
I forget to include header files which would make the compiler
complain.
Change-Id: I745644f4577d05ef22adbd570dc94b6161eb65c8
Signed-off-by: Randy Li <randy.li@rock-chips.com>
The parser won't print the HDR information out neither.
Change-Id: Ie2c93221b3d6ae2aab7f1eb5ee0169088e45f1df
Signed-off-by: Randy Li <randy.li@rock-chips.com>
I didn't verify the PP fully, thanks to Timkingh fixed those
mistakes.
Also some register for ROI is added this time.
Change-Id: I713b59bdcb844b516804d5b17ebdfb01924fcc70
Signed-off-by: timkingh.huang <timkingh.huang@rock-chips.com>
Signed-off-by: Randy Li <randy.li@rock-chips.com>
When stream buf overflow happens, frame qp plus 3,
to reduce stream size.
Change-Id: If6270aae46cc09a3b8eec207ca441040e0330fdf
Signed-off-by: Lin Kesheng <lks@rock-chips.com>
since have little data to estimate qp, qp of frame in first gop is
nearly hor-line. now we add a re-encode mechanism to avoid this bug.
re-encode frame if it meets all the conditions below:
1. gop is the first gop
2. type is p frame
3. target_bis is larger than 1.5 * real_bits
and the qp_init will decrese 3 when re-encode.
TODO: maybe use sse to calculate a proper value instead of 3 or add
a much more suitable condition to start this re-encode process.
Change-Id: I2bea5dd31a3f0348a23b40650b2e11bce32ca2db
Signed-off-by: Jung Zhao <jung.zhao@rock-chips.com>
Signed-off-by: Lin Kesheng <lks@rock-chips.com>
Some video contain seq header at the header of file
only, so we use extra data to parse seq if need.
Change-Id: I82c6154bb61a8338f0875facee168510ba5e6dc2
Signed-off-by: fangxingwen <fxw@rock-chips.com>
