Re: [PATCH v2 2/5] media: visl: Add a stable_output parameter

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 22/11/2023 17:49, Detlev Casanova wrote:
> On Wednesday, November 22, 2023 11:03:53 A.M. EST Hans Verkuil wrote:
>> On 24/10/2023 21:09, Detlev Casanova wrote:
>>> This parameter is used to ensure that for a given input, the output
>>> frames are always identical so that it can be compared against
>>> a reference in automatic tests.
>>>
>>> Reviewed-by: Daniel Almeida <daniel.almeida@xxxxxxxxxxxxx>
>>> Signed-off-by: Detlev Casanova <detlev.casanova@xxxxxxxxxxxxx>
>>> ---
>>>
>>>  drivers/media/test-drivers/visl/visl-core.c |   5 +
>>>  drivers/media/test-drivers/visl/visl-dec.c  | 125 +++++++++++---------
>>>  drivers/media/test-drivers/visl/visl.h      |   1 +
>>>  3 files changed, 77 insertions(+), 54 deletions(-)
>>>
>>> diff --git a/drivers/media/test-drivers/visl/visl-core.c
>>> b/drivers/media/test-drivers/visl/visl-core.c index
>>> df6515530fbf..d28d50afec02 100644
>>> --- a/drivers/media/test-drivers/visl/visl-core.c
>>> +++ b/drivers/media/test-drivers/visl/visl-core.c
>>> @@ -88,6 +88,11 @@ module_param(bitstream_trace_nframes, uint, 0);
>>>
>>>  MODULE_PARM_DESC(bitstream_trace_nframes,
>>>  
>>>  		 " the number of frames to dump the bitstream through 
> debugfs");
>>>
>>> +bool stable_output;
>>> +module_param(stable_output, bool, 0644);
>>> +MODULE_PARM_DESC(stable_output,
>>> +		 " only write stable data for a given input on the 
> output frames");
>>> +
>>>
>>>  static const struct visl_ctrl_desc visl_fwht_ctrl_descs[] = {
>>>  
>>>  	{
>>>  	
>>>  		.cfg.id = V4L2_CID_STATELESS_FWHT_PARAMS,
>>>
>>> diff --git a/drivers/media/test-drivers/visl/visl-dec.c
>>> b/drivers/media/test-drivers/visl/visl-dec.c index
>>> 318d675e5668..61cfca49ead9 100644
>>> --- a/drivers/media/test-drivers/visl/visl-dec.c
>>> +++ b/drivers/media/test-drivers/visl/visl-dec.c
>>> @@ -197,19 +197,30 @@ static void visl_tpg_fill_sequence(struct visl_ctx
>>> *ctx,> 
>>>  {
>>>  
>>>  	u32 stream_ms;
>>>
>>> -	stream_ms = jiffies_to_msecs(get_jiffies_64() -
>>> ctx->capture_streamon_jiffies); -
>>> -	scnprintf(buf, bufsz,
>>> -		  "stream time: %02d:%02d:%02d:%03d sequence:%u 
> timestamp:%lld
>>> field:%s", -		  (stream_ms / (60 * 60 * 1000)) % 24,
>>> -		  (stream_ms / (60 * 1000)) % 60,
>>> -		  (stream_ms / 1000) % 60,
>>> -		  stream_ms % 1000,
>>> -		  run->dst->sequence,
>>> -		  run->dst->vb2_buf.timestamp,
>>> -		  (run->dst->field == V4L2_FIELD_ALTERNATE) ?
>>> -		  (run->dst->field == V4L2_FIELD_TOP ?
>>> -		  " top" : " bottom") : "none");
>>> +	if (!stable_output) {
>>> +		stream_ms = jiffies_to_msecs(get_jiffies_64() -
>>> ctx->capture_streamon_jiffies); +
>>> +		scnprintf(buf, bufsz,
>>> +			  "stream time: %02d:%02d:%02d:%03d 
> sequence:%u timestamp:%lld
>>> field:%s", +			  (stream_ms / (60 * 60 * 
> 1000)) % 24,
>>> +			  (stream_ms / (60 * 1000)) % 60,
>>> +			  (stream_ms / 1000) % 60,
>>> +			  stream_ms % 1000,
>>
>> How useful is this 'stream time' anyway? I don't think this adds anything
>> useful.
> 
> I suppose that the more debug information is shown, the better.
> 
>>> +			  run->dst->sequence,
>>> +			  run->dst->vb2_buf.timestamp,
>>> +			  (run->dst->field == V4L2_FIELD_ALTERNATE) ?
>>> +			  (run->dst->field == V4L2_FIELD_TOP ?
>>> +			  " top" : " bottom") : "none");
>>> +	} else {
>>> +		scnprintf(buf, bufsz,
>>> +			  "sequence:%u timestamp:%lld field:%s",
>>> +			  run->dst->sequence,
>>> +			  run->dst->vb2_buf.timestamp,
>>> +			  (run->dst->field == V4L2_FIELD_ALTERNATE) ?
>>> +			  (run->dst->field == V4L2_FIELD_TOP ?
>>> +			  " top" : " bottom") : "none");
>>> +
>>> +	}
>>>
>>>  }
>>>  
>>>  static void visl_tpg_fill(struct visl_ctx *ctx, struct visl_run *run)
>>>
>>> @@ -244,15 +255,17 @@ static void visl_tpg_fill(struct visl_ctx *ctx,
>>> struct visl_run *run)> 
>>>  	frame_dprintk(ctx->dev, run->dst->sequence, "");
>>>  	line++;
>>>
>>> -	visl_get_ref_frames(ctx, buf, TPG_STR_BUF_SZ, run);
>>
>> This function shows both the ts of the ref frames and the buffer
>> index. Is it just the buffer index that causes the problem? If so,
>> then wouldn't it be better to either never show the buffer index
>> or only if !stable_output.
> 
> Indeed, the buffer index is the issue, but I did not check if the ref frames ts 
> are stable. I'll do some tests with it and keep the ref frames in stable 
> output mode if they are stable.
> 
>>> +	if (!stable_output) {
>>> +		visl_get_ref_frames(ctx, buf, TPG_STR_BUF_SZ, run);
>>>
>>> -	while ((line_str = strsep(&tmp, "\n")) && strlen(line_str)) {
>>> -		tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 
> 16, line_str);
>>> -		frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", 
> line_str);
>>> -	}
>>> +		while ((line_str = strsep(&tmp, "\n")) && 
> strlen(line_str)) {
>>> +			tpg_gen_text(&ctx->tpg, basep, line++ * 
> line_height, 16, line_str);
>>> +			frame_dprintk(ctx->dev, run->dst->sequence, 
> "%s\n", line_str);
>>> +		}
>>>
>>> -	frame_dprintk(ctx->dev, run->dst->sequence, "");
>>> -	line++;
>>> +		frame_dprintk(ctx->dev, run->dst->sequence, "");
>>> +		line++;
>>> +	}
>>>
>>>  	scnprintf(buf,
>>>  	
>>>  		  TPG_STR_BUF_SZ,
>>>
>>> @@ -280,28 +293,30 @@ static void visl_tpg_fill(struct visl_ctx *ctx,
>>> struct visl_run *run)> 
>>>  		frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", 
> buf);
>>>  	
>>>  	}
>>>
>>> -	line++;
>>> -	frame_dprintk(ctx->dev, run->dst->sequence, "");
>>> -	scnprintf(buf, TPG_STR_BUF_SZ, "Output queue status:");
>>> -	tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 16, buf);
>>> -	frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", buf);
>>> +	if (!stable_output) {
>>> +		line++;
>>> +		frame_dprintk(ctx->dev, run->dst->sequence, "");
>>> +		scnprintf(buf, TPG_STR_BUF_SZ, "Output queue status:");
>>> +		tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 
> 16, buf);
>>> +		frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", 
> buf);
>>>
>>> -	len = 0;
>>> -	for (i = 0; i < out_q->num_buffers; i++) {
>>> -		char entry[] = "index: %u, state: %s, request_fd: %d, 
> ";
>>> -		u32 old_len = len;
>>> -		char *q_status = visl_get_vb2_state(out_q->bufs[i]-
>> state);
>>> +		len = 0;
>>> +		for (i = 0; i < out_q->num_buffers; i++) {
>>> +			char entry[] = "index: %u, state: %s, 
> request_fd: %d, ";
>>> +			u32 old_len = len;
>>> +			char *q_status = visl_get_vb2_state(out_q-
>> bufs[i]->state);
>>>
>>> -		len += scnprintf(&buf[len], TPG_STR_BUF_SZ - len,
>>> -				 entry, i, q_status,
>>> -				 to_vb2_v4l2_buffer(out_q-
>> bufs[i])->request_fd);
>>> +			len += scnprintf(&buf[len], TPG_STR_BUF_SZ - 
> len,
>>> +					 entry, i, q_status,
>>> +					 
> to_vb2_v4l2_buffer(out_q->bufs[i])->request_fd);
>>>
>>> -		len += visl_fill_bytesused(to_vb2_v4l2_buffer(out_q-
>> bufs[i]),
>>> -					   &buf[len],
>>> -					   TPG_STR_BUF_SZ - 
> len);
>>> +			len += 
> visl_fill_bytesused(to_vb2_v4l2_buffer(out_q->bufs[i]),
>>> +						   
> &buf[len],
>>> +						   
> TPG_STR_BUF_SZ - len);
>>>
>>> -		tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 
> 16,
>>> &buf[old_len]);
>>> -		frame_dprintk(ctx->dev, run->dst->sequence, "%s", 
> &buf[old_len]);
>>> +			tpg_gen_text(&ctx->tpg, basep, line++ * 
> line_height, 16,
>>> &buf[old_len]); +			frame_dprintk(ctx->dev, run-
>> dst->sequence, "%s",
>>> &buf[old_len]); +		}
>>>
>>>  	}
>>>  	
>>>  	line++;
>>>
>>> @@ -333,25 +348,27 @@ static void visl_tpg_fill(struct visl_ctx *ctx,
>>> struct visl_run *run)> 
>>>  		frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", 
> buf);
>>>  	
>>>  	}
>>>
>>> -	line++;
>>> -	frame_dprintk(ctx->dev, run->dst->sequence, "");
>>> -	scnprintf(buf, TPG_STR_BUF_SZ, "Capture queue status:");
>>> -	tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 16, buf);
>>> -	frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", buf);
>>> +	if (!stable_output) {
>>> +		line++;
>>> +		frame_dprintk(ctx->dev, run->dst->sequence, "");
>>> +		scnprintf(buf, TPG_STR_BUF_SZ, "Capture queue 
> status:");
>>> +		tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 
> 16, buf);
>>> +		frame_dprintk(ctx->dev, run->dst->sequence, "%s\n", 
> buf);
>>>
>>> -	len = 0;
>>> -	for (i = 0; i < cap_q->num_buffers; i++) {
>>> -		u32 old_len = len;
>>> -		char *q_status = visl_get_vb2_state(cap_q->bufs[i]-
>> state);
>>> +		len = 0;
>>> +		for (i = 0; i < cap_q->num_buffers; i++) {
>>> +			u32 old_len = len;
>>> +			char *q_status = visl_get_vb2_state(cap_q-
>> bufs[i]->state);
>>>
>>> -		len += scnprintf(&buf[len], TPG_STR_BUF_SZ - len,
>>> -				 "index: %u, status: %s, 
> timestamp: %llu, is_held: %d",
>>> -				 cap_q->bufs[i]->index, q_status,
>>> -				 cap_q->bufs[i]->timestamp,
>>> -				 to_vb2_v4l2_buffer(cap_q-
>> bufs[i])->is_held);
>>> +			len += scnprintf(&buf[len], TPG_STR_BUF_SZ - 
> len,
>>> +					 "index: %u, status: 
> %s, timestamp: %llu, is_held: %d",
>>> +					 cap_q->bufs[i]-
>> index, q_status,
>>> +					 cap_q->bufs[i]-
>> timestamp,
>>> +					 
> to_vb2_v4l2_buffer(cap_q->bufs[i])->is_held);
>>>
>>> -		tpg_gen_text(&ctx->tpg, basep, line++ * line_height, 
> 16,
>>> &buf[old_len]);
>>> -		frame_dprintk(ctx->dev, run->dst->sequence, "%s", 
> &buf[old_len]);
>>> +			tpg_gen_text(&ctx->tpg, basep, line++ * 
> line_height, 16,
>>> &buf[old_len]); +			frame_dprintk(ctx->dev, run-
>> dst->sequence, "%s",
>>> &buf[old_len]); +		}
>>>
>>>  	}
>>>  
>>>  }
>>>
>>> diff --git a/drivers/media/test-drivers/visl/visl.h
>>> b/drivers/media/test-drivers/visl/visl.h index 31639f2e593d..5a81b493f121
>>> 100644
>>> --- a/drivers/media/test-drivers/visl/visl.h
>>> +++ b/drivers/media/test-drivers/visl/visl.h
>>> @@ -85,6 +85,7 @@ extern unsigned int visl_dprintk_nframes;
>>>
>>>  extern bool keep_bitstream_buffers;
>>>  extern int bitstream_trace_frame_start;
>>>  extern unsigned int bitstream_trace_nframes;
>>>
>>> +extern bool stable_output;
>>>
>>>  #define frame_dprintk(dev, current, fmt, arg...) \
>>>  
>>>  	do { \
>>
>> Should stable_output perhaps be 1 by default?
> 
> In that case, why not use the visl_debug parameter and show the unstable data 
> only when it is set to one ?

I don't think that's a good idea. That parameter enables driver debugging output,
and is meant to track down driver issues. It shouldn't be mixed with changing
driver behavior.

Regards,

	Hans

> 
> --
> Detlev





[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]

  Powered by Linux