[Intel-gfx] [PATCH i-g-t 3/3] intel_gpu_top: Show non-normalized client usage in numeric mode

Tvrtko Ursulin tvrtko.ursulin at linux.intel.com
Mon Apr 17 10:58:29 UTC 2023


Hi Kamil,

On 14/04/2023 16:28, Kamil Konieczny wrote:
> On 2023-03-28 at 13:54:29 +0100, Tvrtko Ursulin wrote:
>> From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>>
>> When numeric display is selected in the interactive mode it is more
>> descriptive to show client's non-normalized (by number of engines per
>> class) usage. Rendering of the visual representation (bar) is kept the
>> same.
>>
>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> 
> Reviewed-by: Kamil Konieczny <kamil.konieczny at linux.intel.com>

Thanks for reviewing all three and merging them too!

Regards,

Tvrtko

>> ---
>>   tools/intel_gpu_top.c | 20 +++++++++++---------
>>   1 file changed, 11 insertions(+), 9 deletions(-)
>>
>> diff --git a/tools/intel_gpu_top.c b/tools/intel_gpu_top.c
>> index 36da016c3df0..b6827b3de9bd 100644
>> --- a/tools/intel_gpu_top.c
>> +++ b/tools/intel_gpu_top.c
>> @@ -1275,14 +1275,14 @@ static void n_spaces(const unsigned int n)
>>   }
>>   
>>   static void
>> -print_percentage_bar(double percent, int max_len, bool numeric)
>> +print_percentage_bar(double percent, double max, int max_len, bool numeric)
>>   {
>>   	int bar_len, i, len = max_len - 2;
>>   	const int w = 8;
>>   
>>   	assert(max_len > 0);
>>   
>> -	bar_len = ceil(w * percent * len / 100.0);
>> +	bar_len = ceil(w * percent * len / max);
>>   	if (bar_len > w * len)
>>   		bar_len = w * len;
>>   
>> @@ -2010,7 +2010,8 @@ print_engine(struct engines *engines, unsigned int i, double t,
>>   			      engine->display_name, engine_items[0].buf);
>>   
>>   		val = pmu_calc(&engine->busy.val, 1e9, t, 100);
>> -		print_percentage_bar(val, con_w > len ? con_w - len : 0, false);
>> +		print_percentage_bar(val, 100.0, con_w > len ? con_w - len : 0,
>> +				     false);
>>   
>>   		printf("%s\n", buf);
>>   
>> @@ -2292,23 +2293,24 @@ print_client(struct client *c, struct engines *engines, double t, int lines,
>>   		       clients->max_name_len, c->print_name);
>>   
>>   		for (i = 0; c->samples > 1 && i < clients->num_classes; i++) {
>> -			double pct;
>> +			double pct, max;
>>   
>>   			if (!clients->class[i].num_engines)
>>   				continue; /* Assert in the ideal world. */
>>   
>> -			pct = (double)c->val[i] / period_us / 1e3 * 100 /
>> -			      clients->class[i].num_engines;
>> +			pct = (double)c->val[i] / period_us / 1e3 * 100;
>>   
>>   			/*
>>   			 * Guard against possible time-drift between sampling
>>   			 * client data and time we obtained our time-delta from
>>   			 * PMU.
>>   			 */
>> -			if (pct > 100.0)
>> -				pct = 100.0;
>> +			max = 100.0 * clients->class[i].num_engines;
>> +			if (pct > max)
>> +				pct = max;
>>   
>> -			print_percentage_bar(pct, *class_w, numeric_clients);
>> +			print_percentage_bar(pct, max, *class_w,
>> +					     numeric_clients);
>>   		}
>>   
>>   		putchar('\n');
>> -- 
>> 2.37.2
>>


More information about the Intel-gfx mailing list