Original question:
Hi,
I have set HDMI to Vout1 / DP1 according to the following link, but the clock(pclk) is always about 600 MHz, How to modify it(148.5MHz)?
Looking forward to your reply, thanks a lot!
Hi, I use VP2 connected to vout1,as following
For HDMI output, vision apps connects VP2 output to vout0 (DP0) and pixel clock is 148.5MHz. When I switch to Vout1 / DP1 (use VP2), the clock is wrong ,is not 148.5MHz.
How can I change it?
Hi,
I'm sure it's changed. My amendment is as follows:
1、pinmux（vision_apps/utils/misc/src/app_pinmux.c）
2、appDssDefaultInit (vision_apps/utils/dss/src/app_dss_defaults.c)
3、According to the above content, the clock will be abnormal (It is different from the following (vision_apps/utils/dss/include/app_dctrl.h) configuration, but vout0/dp0 is ok).
so, what's wrong with it?
I wonder if the following (vision_apps/utils/dss/src/app_dss_j721e.c) should be revised as well. If so, can you tell me how to do it.
Look forward to your reply!
Sorry to reply so long，
I use k3config tool in Linux system，as follows:
It should be the wrong usage, K3config tool can read registers？Is there any other way to dump?
Look forward to your reply!
Hi,
I can read the value in the register with devmem2 tool. Now I can transfer the picture with bt1120, but the result is wrong，as follow
Original picture（YUV422）：
Picture after transmission：
The value of the DSS0_VP_CONFIG register is：0x03200000
Look forward to your reply!
Hi,
There is another problem, as follow
·Is there an instance of writeback in vision_apps, or how to use it?
look forward to your reply!
Hi,
I tried two ways, no change, as follow
The value of the DSS0_VP_POL_FREQ register is：0x000040000 or 0x00000000
But, I made the following attempts
· Enable or Disable Color Space Conversion RGB to YUV
·Color Space Conversion full or limit range setting
The results are as follows:
·Original picture:
·Enable CSC and full range setting
·Enable CSC and limit range setting
·Disable CSC
How to locate? Where are the problems？
look forward to your reply！
Hi,
For other reasons, I need to modify videoport. When I switched to VP4, Vout1 was always problematic. The following is my debugging record:
1、app_dss_defaults.c
2、app_dss_j721e.c
3、The value of the associated register
·DSS0_VP_CONFIG(04AE0000): 03200000
·DSS0_COMMON_DISPC_CONNECTIONS(04A000E4): 00000080
·DSS0_VP_POL_FREQ(04AE004C): 00000000
·DSS0_VP_CONTROL(04AE0004): 00000141
Because bt1120 will send background information when the system starts, I measured the data lines of two kinds of Vout, found a phenomenon ,as follow:
·data0, data4 and data15 in VOUT0 have data signals, but data2 and data6 in VOUT1 have data signals.
Is it the above phenomenon that causes Vout1 to be wrong all the time? If so, how to solve it? If not, how to locate the cause of the problem?
Looking forward to your reply!
Hi,
Thank you for your reply！
I use the pinmux configured by the ti-pinmux tool, I think there should be no mistakes.
I changed not only the videoport, but also the clock. The clk settings for VP2 in vision_apps are as follows:
My changes as following, Can the following configuration be used as the CLK of VP4？
And what do the notes in the document mean? as following:
Looking forward to your reply!
Hi,
The result of previous debugging is still problematic, though there is no difference in the naked eye (a picture is bright and a picture is dark). We compared the hexadecimal system and found that the two images (All 0 yuv images) are completely different, as follows:
Y component(The original picture is all 0, and the tda4 transmitter is 42):
UV component(The original picture is full 0, and the tda4 transmitter becomes 594F):
The contrast between the two images is as follows(a picture is bright and a picture is dark):
What has been checked is as follows:
1、pinmux:all right
2、gamma correction:disable
3、color space conversion full range setting: full range setted
look forward to your reply！
Hi，thank you for your reply!
Yes, It's the same project.
OK, I will close that thread.
After the above phenomenon (the hexadecimal format of the two yuv images is different), I did the following checks：
1、Each data line is measured with an oscilloscope，all 0 YUV image transmission should not appear the following waveform：
2、Checked pinmux setting (padconfig of vout0), as following：
·The value of each register is 0x0001000a
So I think pinmux is right
3、gamma correction、color space conversion full range setting
·gamma correction:disable
·color space conversion full range setting: full range setted
4、
I modified it in file：ti-processor-sdk-rtos-j721e-evm-07_02_00_06/vision_apps/utils/dss/src/app_dss_defaults.c.
I tried 12bit, 16bit, 24bit, but the received YUV image did not change (including the converted hexadecimal file)，So I don't think this is the place to change bt.1120 mode (16 bit mode or 20 bit mode). Is my understanding right? Where can I change bt.1120 mode？
Finally, what other settings will change the data flow and how to check？
Hi, thank you for your reply!
Yes, I am using full range out. The format of my input data is YUV422 (yuyv).
I am familiar with the data stream of DSS on one side and find that there are many data conversions in video pipeline and video port.
For example, in the following transformation, the offset value of RGB is not 0, although it has been set to 0 in the driver file.
Is there a way to skip these transitions?
Looking forward to your reply！
Hi, thank you for your reply!
No, I am not doing any brightness control. I dump the value of the register as follows. It is not the same as what is configured in the driver.
·Configuration in driver file
The value of dump:
| register | value |
| 0x04AA0008 | 0x00810042 |
| 0x04AA000C | 0x07DA0019 |
| 0x04AA0010 | 0x007007B6 |
| 0x04AA005C | 0x07A20070 |
| 0x04AA0060 | 0x000007EE |
| 0x04AA0068 | 0x00000000 |
| 0x04AA006C | 0x40004000 |
·Configuration in driver file
The value of dump:
| register | value |
| 0x04A60040 | 0x015F0100 |
| 0x04A60044 | 0x01000000 |
| 0x04A60048 | 0x07AA074D |
| 0x04A6004C | 0x00000100 |
| 0x04A60050 | 0x000001BB |
| 0x04A60054 | 0xC0000000 |
| 0x04A60058 | 0x0000C000 |
| 0x04A60023 | 0x00280AFD |
From the values of the above registers, it is found that, the offset value of RGB is not 0, although it has been set to 0 in the driver file.
How to locate the problem？
Looking forward to your reply！
Hi, thank you for your reply!
OK, I will check the value of the register (0x04A6023C).
Can't dpi0 be configured in bt.1120-16bit mode？If I can, how can I confirm that I am configuring 16 bit mode? I always thought it was determined by the following(ti-processor-sdk-rtos-j721e-evm-07_02_00_06/vision_apps/utils/dss/src/app_dss_defaults.c) parameters, but later I found that it works on the following configuration.Does this configuration determine the mode of BT（bt1120.16bit or bt1120.20bit）?
Driver file configuration：
DSS0_VP_CONTROL Register:
Looking forward to your reply！
Hi, thank you for your reply!
So far, I have checked the following links：
1、
·VC1 range mapping disabled
·Luma key operation is disabled
·Gamma inversion is disabled
·The matrix configuration of CSC is the same as that in the driver file
·Full range
The value of DSS0_VID_ATTRIBUTES register (0x04A60020) is : 0x00280AFD
The value of DSS0_VID_ATTRIBUTES2 register (0x04A60024) is: 0x3C000000
2、
·Full range
·The matrix configuration of CSC is the same as that in the driver file
·Gamma disabled
·APP_DCTRL_VIFW_24BIT
The value of DSS0_VP_CONFIG register (0x04AA0000) is: 0x03200000
The value of DSS0_COMMON_DISPC_CONNECTIONS register (0x04A000E4) is: 0x00000002
The value of DSS0_VP_CONTROL register (0x04AA0004) is: 0x00000361
3、Hardware circuit connection：
use 16bit, connect D2 to D9 and D12 to D19 data lines
4、YUV test procedure is as follows
#include "shm.h"
#include <TI/j7.h>
#include <TI/tivx.h>
#include <VX/vx.h>
#include <string.h>
#include <sys/types.h>
#include <sys/ipc.h>
#include <sys/shm.h>
#include <stdio.h>
#include <signal.h>
#include <dirent.h>
#include <errno.h>
#include <time.h>
#define DISPLAY_NUM_RUN_COUNT 100
//#define FILE_DIR "/opt/vision_apps/test_data/dttest"
typedef struct {
const char *name;
uint32_t pipeId;
uint32_t dataFormat;
uint32_t inWidth;
uint32_t inHeight;
uint32_t bpp;
uint32_t pitchY;
uint32_t pitchUV;
uint32_t outWidth;
uint32_t outHeight;
uint32_t posX;
uint32_t posY;
uint32_t loopCount;
} Arg;
/**
* @brief Add time stamp to bt1120 image
* @param addr: Bt1120 binary image sequence
* @param timestamp: 64 bit timestamp, small end representation
*/
static int fill_time_stamp(void *addr, unsigned long long *timestamp) {
char *addrp = (char *)addr;
int i = 0;
for (i = 15; i >= 0; i--)
addrp[15 - i] =
0x50 |
((uint8_t)(((*timestamp) & ((uint64_t)0xf << (i * 4))) >> (i * 4)));
return 0;
}
#if 1
static vx_status readYUV422Input(char *file_name, unsigned char *data_ptr) {
vx_status status = VX_SUCCESS;
if (status == VX_SUCCESS) {
FILE *fp = fopen(file_name, "rb");
if (fp == NULL) {
printf("Unable to open file %s \n", file_name);
return (VX_FAILURE);
}
vx_uint32 img_width = 1920;
vx_uint32 img_height = 1080;
// Copy YuYv
vx_uint32 num_bytes = fread(data_ptr, 1, img_width * img_height * 2, fp);
if (num_bytes != (img_width * img_height * 2))
printf("CbCr bytes read = %d, expected = %d \n", num_bytes,
img_width * img_height * 2);
fclose(fp);
}
return (status);
}
#endif
vx_status app_tidl_od_main(vx_int32 argc, vx_char *argv[]) {
vx_status status = VX_SUCCESS;
unsigned char gTiovxCtDisplayArray1[1920 * 1080 * 2] = {0};
Arg arg_;
arg_.pipeId = 2;
arg_.dataFormat = VX_DF_IMAGE_YUYV;
arg_.inWidth = 1920;
arg_.inHeight = 1080;
arg_.bpp = 2;
arg_.pitchY = 1920;
arg_.pitchUV = 0;
arg_.outWidth = 1920;
arg_.outHeight = 1080;
arg_.posX = 0;
arg_.posY = 0;
vx_context context = vxCreateContext();
status = vxGetStatus((vx_reference)context);
if (status != VX_SUCCESS) {
printf("vxGetStatus is failed! \n");
}
vx_image disp_image = 0;
vx_imagepatch_addressing_t image_addr;
tivx_display_params_t params;
vx_user_data_object param_obj;
vx_graph graph = 0;
vx_node node = 0;
//uint32_t loop_count = arg_.loopCount;
if (vx_true_e == tivxIsTargetEnabled(TIVX_TARGET_DISPLAY1))
{
tivxHwaLoadKernels(context);
disp_image = vxCreateImage(context, arg_.inWidth, arg_.inHeight, arg_.dataFormat);
image_addr.dim_x = arg_.inWidth;
image_addr.dim_y = arg_.inHeight;
image_addr.stride_x = arg_.bpp;
image_addr.stride_y = arg_.pitchY * 2;
image_addr.scale_x = VX_SCALE_UNITY;
image_addr.scale_y = VX_SCALE_UNITY;
image_addr.step_x = 1;
image_addr.step_y = 1;
vx_rectangle_t rect;
rect.start_x = 0;
rect.start_y = 0;
rect.end_x = arg_.inWidth;
rect.end_y = arg_.inHeight;
DIR * dir;
char *FILE_DIR = "/opt/vision_apps/test_data/dttest";
struct dirent * ptr;
dir = opendir(FILE_DIR);
if(NULL == dir)
{
printf("opendir error...!\n");
printf("%s\n",strerror(errno));
return -1;
}
char file_name[512] = {0};
loop1:
ptr = readdir(dir);
sprintf(file_name,"%s/%s",FILE_DIR,ptr->d_name);
if(!(strcmp(ptr->d_name,".")) || !(strcmp(ptr->d_name,"..")))
{
goto loop1;
}
printf("filename is-->%s\n",file_name);
if(ptr != NULL)
{
status = readYUV422Input(file_name,gTiovxCtDisplayArray1);
if (status != VX_SUCCESS)
{
printf("readYUV422Input is failed! \n");
}
}
long long current_time = time(0);
fill_time_stamp((void *)gTiovxCtDisplayArray1,(unsigned long long*)¤t_time);
status = vxCopyImagePatch(disp_image, &rect, 0, &image_addr,
(void *)gTiovxCtDisplayArray1, VX_WRITE_ONLY,
VX_MEMORY_TYPE_HOST);
if (status != VX_SUCCESS) {
printf("vxCopyImagePatch 0 is failed! \n");
}
memset(¶ms, 0, sizeof(tivx_display_params_t));
params.opMode = TIVX_KERNEL_DISPLAY_BUFFER_COPY_MODE;
params.pipeId = arg_.pipeId;
params.outWidth = arg_.outWidth;
params.outHeight = arg_.outHeight;
params.posX = arg_.posX;
params.posY = arg_.posY;
param_obj = vxCreateUserDataObject(context, "tivx_display_params_t",
sizeof(tivx_display_params_t), ¶ms);
graph = vxCreateGraph(context);
node = tivxDisplayNode(graph, param_obj, disp_image);
status = vxSetNodeTarget(node, VX_TARGET_STRING, TIVX_TARGET_DISPLAY1);
if (status != VX_SUCCESS) {
printf("vxSetNodeTarget is failed! \n");
}
status = vxVerifyGraph(graph);
if (status != VX_SUCCESS) {
printf("vxVerifyGraph is failed! \n");
}
while(1)
{
while ((ptr=readdir(dir))!= NULL)
{
if(!(strcmp(ptr->d_name,".")) || !(strcmp(ptr->d_name,"..")))
{
continue;
}
sprintf(file_name,"%s/%s",FILE_DIR,ptr->d_name);
printf("filename is-->%s\n",file_name);
status = readYUV422Input(file_name,gTiovxCtDisplayArray1);
if(status != VX_SUCCESS)
{
printf("readYUV422Input is failed! \n");
}
current_time = time(0);
fill_time_stamp((void *)gTiovxCtDisplayArray1,(unsigned long long*)¤t_time);
status = vxProcessGraph(graph);
if (status != VX_SUCCESS) {
printf("vxProcessGraph is failed!\n");
} else {
printf("vxProcessGraph is successed!\n");
}
status = vxCopyImagePatch(disp_image, &rect, 0, &image_addr,
(void *)gTiovxCtDisplayArray1, VX_WRITE_ONLY,
VX_MEMORY_TYPE_HOST);
if (status == VX_SUCCESS) {
printf("vxCopyImagePatch 1 is success! \n");
} else {
printf("in while(1),vxCopyImagePatch error\n");
//goto exit;
}
}
dir = opendir(FILE_DIR);
if(NULL == dir)
{
printf("opendir2 error...!\n");
printf("%s\n",strerror(errno));
printf("break while(1)...!!!\n");
break;
}
}
vxReleaseNode(&node);
vxReleaseGraph(&graph);
vxReleaseImage(&disp_image);
vxReleaseUserDataObject(¶m_obj);
tivxHwaUnLoadKernels(context);
}
return status;
}
#if 0
TEST_WITH_ARG(tivxHwaDisplay, testZeroBufferCopyMode, Arg, PARAMETERS)
{
vx_context context = context_->vx_context_;
vx_image disp_image[2];
vx_imagepatch_addressing_t image_addr;
tivx_display_params_t params;
vx_user_data_object param_obj;
vx_graph graph = 0;
vx_node node = 0;
vx_image disp_image_temp;
uint32_t num_refs;
uint32_t loop_count = arg_.loopCount;
vx_graph_parameter_queue_params_t graph_parameters_queue_params_list[1];
if (vx_true_e == tivxIsTargetEnabled(TIVX_TARGET_DISPLAY1))
{
tivxHwaLoadKernels(context);
CT_RegisterForGarbageCollection(context, ct_teardown_hwa_kernels, CT_GC_OBJECT);
ASSERT_VX_OBJECT(disp_image[0] = vxCreateImage(context, arg_.inWidth, arg_.inHeight, arg_.dataFormat), VX_TYPE_IMAGE);
ASSERT_VX_OBJECT(disp_image[1] = vxCreateImage(context, arg_.inWidth, arg_.inHeight, arg_.dataFormat), VX_TYPE_IMAGE);
image_addr.dim_x = arg_.inWidth;
image_addr.dim_y = arg_.inHeight;
image_addr.stride_x = arg_.bpp;
image_addr.stride_y = arg_.pitchY;
image_addr.scale_x = VX_SCALE_UNITY;
image_addr.scale_y = VX_SCALE_UNITY;
image_addr.step_x = 1;
image_addr.step_y = 1;
vx_rectangle_t rect;
rect.start_x = 0;
rect.start_y = 0;
rect.end_x = arg_.inWidth;
rect.end_y = arg_.inHeight;
vxCopyImagePatch(disp_image[0],
&rect,
0,
&image_addr,
(void *)gTiovxCtDisplayArray1,
VX_WRITE_ONLY,
VX_MEMORY_TYPE_HOST
);
vxCopyImagePatch(disp_image[1],
&rect,
0,
&image_addr,
(void *)gTiovxCtDisplayArray2,
VX_WRITE_ONLY,
VX_MEMORY_TYPE_HOST
);
memset(¶ms, 0, sizeof(tivx_display_params_t));
params.opMode=TIVX_KERNEL_DISPLAY_ZERO_BUFFER_COPY_MODE;
params.pipeId=arg_.pipeId;
params.outWidth=arg_.outWidth;
params.outHeight=arg_.outHeight;
params.posX=arg_.posX;
params.posY=arg_.posY;
ASSERT_VX_OBJECT(param_obj = vxCreateUserDataObject(context, "tivx_display_params_t", sizeof(tivx_display_params_t), ¶ms), (enum vx_type_e)VX_TYPE_USER_DATA_OBJECT);
ASSERT_VX_OBJECT(graph = vxCreateGraph(context), VX_TYPE_GRAPH);
ASSERT_VX_OBJECT(node = tivxDisplayNode(graph, param_obj, disp_image[0]), VX_TYPE_NODE);
VX_CALL(vxSetNodeTarget(node, VX_TARGET_STRING, TIVX_TARGET_DISPLAY1));
/* Image is parameter number 1 for Display Node and becomes graph parameter 0 */
add_graph_parameter_by_node_index(graph, node, 1);
/* Set graph schedule config such that graph parameter @ index 0 is enqueuable */
graph_parameters_queue_params_list[0].graph_parameter_index = 0;
graph_parameters_queue_params_list[0].refs_list_size = 2; /* Two images */
graph_parameters_queue_params_list[0].refs_list = (vx_reference*)&disp_image[0];
/* Schedule mode auto is used, here we don't need to call vxScheduleGraph
* Graph gets scheduled automatically as refs are enqueued to it
*/
vxSetGraphScheduleConfig(graph,
VX_GRAPH_SCHEDULE_MODE_QUEUE_AUTO,
1,
graph_parameters_queue_params_list
);
/* Explicitly set graph pipeline depth */
ASSERT_EQ_VX_STATUS(VX_SUCCESS, set_graph_pipeline_depth(graph, 2));
VX_CALL(vxVerifyGraph(graph));
/* Enqueue input references */
vxGraphParameterEnqueueReadyRef(graph, 0, (vx_reference*)&disp_image[0], 1);
vxGraphParameterEnqueueReadyRef(graph, 0, (vx_reference*)&disp_image[1], 1);
/* Wait for the loop count */
while(loop_count-- > 0)
{
/* Dequeue one image buffer */
vxGraphParameterDequeueDoneRef(graph, 0, (vx_reference*)&disp_image_temp, 1, &num_refs);
/* Enqueue the same image buffer */
vxGraphParameterEnqueueReadyRef(graph, 0, (vx_reference*)&disp_image_temp, 1);
}
/* ensure all graph processing is complete */
vxWaitGraph(graph);
VX_CALL(vxReleaseNode(&node));
VX_CALL(vxReleaseGraph(&graph));
VX_CALL(vxReleaseImage(&disp_image[0]));
VX_CALL(vxReleaseImage(&disp_image[1]));
VX_CALL(vxReleaseUserDataObject(¶m_obj));
ASSERT(node == 0);
ASSERT(graph == 0);
ASSERT(disp_image[0] == 0);
ASSERT(disp_image[1] == 0);
ASSERT(param_obj == 0);
tivxHwaUnLoadKernels(context);
}
}
TESTCASE_TESTS(tivxHwaDisplay,
testBufferCopyMode,
testZeroBufferCopyMode
)
#endif
Hi, thank you for your reply!
This is my overlay configuration：
The oscilloscope detects that there is a signal in Data0、Data4、Data15.
Received pictures (after YUV422 to nv12), as follows：
Y component：
UV component：
What is the reason？
Looking forward to your reply！
Hi, thank you for your reply!
First of all, the input of DSS is a yuv422i-yuyv image with full 0 (Y component and UV component are both 0).
Then there are three ways to detect the output.
1、The first one is to use the oscilloscope to check the waveform of 30us (there are about 1080 pixels when the clock is 74.25mhz), and find a lot of non-zero data, as follows:
Why do 0 and 1 have different trigger waveforms?
2、With YUV image viewer, it is found that the original image is different from the received image.
After all 0 pictures are transmitted, the colors are different，as follows:
As shown below, the color in the red box is different：
3、Open the file in hexadecimal and find that the Y and UV components have changed after transmission
