| bacae1c546a604cd0dcf6cc16c9201083d94e9dd |
|
08-Jan-2017 |
zongwave <wei.zong@intel.com> |
cl_kernel: fix CL build double precision constant casting to float warning
* OpenCL compiler report warning:
"double precision constant requires cl_khr_fp64, casting to single precision"
the CL build will be failed at some hardware without double support.
adding suffix 'f' at the end of literal floating numbers to define float format
Signed-off-by: Zong Wei <wei.zong@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 06f757dd1455a9178b9d9765d8f9ecafb61345ec |
|
25-Feb-2016 |
Wind Yuan <feng.yuan@intel.com> |
code style: format kernel cl_bayer_pipe.cl
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 0205f49b222b89406ac8f6b05a50e07bae86573c |
|
18-Feb-2016 |
Yinhang Liu <yinhangx.liu@intel.com> |
cl-bayer-ee: fix flicker issue on the edge
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| d0ecb8288cc90b0c9e993613111c216491a60d24 |
|
24-Dec-2015 |
Wangfei <feix.w.wang@intel.com> |
cl-bayer-pipe: reduce noise on low light edge
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 1f72e1bfc3d9c0dd55965c708abae00226a378aa |
|
28-Dec-2015 |
wujunkai166 <wujunkai166@gmail.com> |
cl-bayer: fix bug that color bar exist on top of image.
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| ac5ec2cb3691eea1a4370da4983cc7fea5d1b569 |
|
17-Dec-2015 |
wujunkai166 <wujunkai166@gmail.com> |
cl-bayer-pipe: clean up and optimize performance
* Use 'if' statement to replace 'for' statement outside slm
load function.
* Put the calculation of 'coff0' out of dot_denoise() function
to reduce repeat slm read.
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| cd1edab603945df06213343e3023a8b34612c97a |
|
14-Dec-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer-pipe: fix hard code of input image
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 93f57e368765670bc4f5de0dc90f2e657d20df0e |
|
14-Dec-2015 |
Wangfei <feix.w.wang@intel.com> |
cl-bayer-ee: picture quality improve.
* adjust ee algorithm and args to get a higher picture quality.
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 66a677ec4817859937a8886ff3f48fc79daa8430 |
|
09-Dec-2015 |
wujunkai166 <wujunkai166@gmail.com> |
cl-bayer: improve work group size for performance.
* workgroup cell of 64x4 is better than 8x8 for performance
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| bccca4d20846a3fe8aabb9b743e3db0248ef6155 |
|
07-Dec-2015 |
wujunkai166 <wujunkai166@gmail.com> |
cl-bayer: optimize EE calculation for performance
* Performance improved
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 61a5b0c713a8fac73508f583cc6a5f601a11e323 |
|
25-Nov-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: clean up
* reduce macro define in bayer kernel
* remove unnecessary funcs
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 39190e675b34fa2c8e4f9cefcb3e2acb17aef1e1 |
|
30-Nov-2015 |
Wangfei <feix.w.wang@intel.com> |
cl-bayer: enable edge enhancement in bayer-pipe
* enable bayer-EE
* add bayer NR/EE tuning parameters in ciq module
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| a8cdc36f78608f91055b06ba5f684ba4557196e1 |
|
18-Nov-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: separate bayer into basic and demosaic 2 parts
* bayer-basic: works for blc/wb/gamma/3a-stats,
input SGRBG16-packed, output: SGRBG16-planar
* bayer-pipe: works for demosaic/denoise
input SGRBG16-planar, output: RGB16-planar
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| bd68e435ad2fac007b92f92aa7d73f070be9c75b |
|
19-Nov-2015 |
Wangfei <feix.w.wang@intel.com> |
cl-bayer: add spatial delta coeff into bayer NR
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 107b77884a1e779f75f5b59bc06d7b61b2af5393 |
|
18-Nov-2015 |
Wangfei <feix.w.wang@intel.com> |
cl-bayer: suport bilateral filter in bayer NR
* add range coeff table to reduce calculation
* load to slm to improve performance
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| e9a759f7f91acdb053be19cac1413d2dc04396d8 |
|
18-Nov-2015 |
Yinhang Liu <yinhangx.liu@intel.com> |
histogram: fixed 3a-stats issue with half data zero
Commited-by: Wind Yuan <feng.yuan@intel.com>
Verified-by Yinhang Liu <yinhangx.liu@intel.com>
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 4484b1343c14becbacb99e9246ca6fa202cc4820 |
|
30-Oct-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: change input order from R to RGBA
* change channle order R to RGBA even read more horizontal edge
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 5bd3ae6ff7887be6dacc9ccc96487967b36963b3 |
|
29-Oct-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: output planar RGB and slm vector load
* bayer pipe default change output to XCAM_PIX_FMT_RGB48_planar
* demosaic/denoise load slm buffer in advance to reduce frequent
buffer loading
* demosaic/denoise process 2-cell pixels on each workitem
* use mad operator instead of multiply/add as many as possible
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 044508605a55591376806ef4003fd1d5c02f9877 |
|
24-Sep-2015 |
wujunkai166 <wujunkai166@gmail.com> |
cl-bayer: change sum from binary to quarternary
* quarternay sum brings better perf than binary sum. this can
save barrier waiting time
* verified-by: Wind Yuan <feng.yuan@intel.com>
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 50b4ff703dd10a737a2571029c33458e88b53682 |
|
22-Sep-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: fix 3a-stats calculation
* reduce sum on 3a-stats may get wrong index if l_id_x + count larger
than STATS_3A_GRID_SIZE which is not the buf-pitch
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 3444cc212927f6609f3fd6a74f778f4504d64a23 |
|
21-Sep-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: set gamma table as optional
* gamma table size 256 is not good enough for every case. need to disable
it on special case like WDR mode
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 8792cf7d1ab983d59e3caac0e86f6e4f6a7eb330 |
|
11-Sep-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: tune nr delta coff
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 08ade226edc3aaf039269207c7b8e699b5229754 |
|
20-Aug-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl-bayer: add new denoise in denosaic as optional
* add new noise-reduction feature in bayer-pipe
Out = (In(x1,y1)*w_delta(x1,y1) + In(x2,y2)*w_delta(x2,y2) +
In(x3,y3)*w_delta(x3,y3) + In(x4,y4)*w_delta(x4,y4) +
In(avg) * w_delta(avg)) /
(w_delta(x1,y1) + w_delta(x2,y2) + w_delta(x3,y3) +
w_delta(x4,y4) + w_delta(avg))
* performance: change float4 array to 4 float array to
reduce bank conflicts
* performance: use float4 vector operator in 3a-stats(sum)
to reduce bank conflicts
* set noise reduction as optional, disable by default.
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| d5f2ea30a234af3339b91930177b780d229f6cef |
|
02-Sep-2015 |
Jia Meng <jia.meng@intel.com> |
cl-bayer: clamp value in 3a grid statistics
* avg_y stat may exceed 8-bit range after wb calculation, so
saturated conversion is used to clamp out-of-range values.
Signed-off-by: Wind Yuan <feng.yuan@intel.com>
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|
| 4b2a37a447ed08f158b1da4e467fcb47eac8706f |
|
21-Jul-2015 |
Wind Yuan <feng.yuan@intel.com> |
cl pipe: merge all bayer kernel together to improve performance
* merged blc, wb, gamma, 3a-stats and demosaic together
* use share local memory for 3a-stats and demosaic
* use reduce method to do average-calculation on 3a stats
* gamma-table reduce workload
/external/libxcam/cl_kernel/kernel_bayer_pipe.cl
|