[Intel-gfx] [igt-dev] [PATCH i-g-t 1/2] i915/gem_exec_latency: Normalize results into ns

[Antonio Argenziano]

On 29/01/19 01:55, Chris Wilson wrote:
> Present the latency results in nanoseconds not RCS cycles.
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
>   tests/i915/gem_exec_latency.c | 38 +++++++++++++++++++++++++++++++----
>   1 file changed, 34 insertions(+), 4 deletions(-)
> 
> diff --git a/tests/i915/gem_exec_latency.c b/tests/i915/gem_exec_latency.c
> index de16322a6..ea44adc14 100644
> --- a/tests/i915/gem_exec_latency.c
> +++ b/tests/i915/gem_exec_latency.c
> @@ -59,6 +59,7 @@
>   #define PREEMPT 0x2
>   
>   static unsigned int ring_size;
> +static double rcs_clock;
>   
>   static void
>   poll_ring(int fd, unsigned ring, const char *name)
> @@ -207,7 +208,7 @@ static void latency_on_ring(int fd,
>   		igt_cork_unplug(&c);
>   
>   	gem_set_domain(fd, obj[1].handle, I915_GEM_DOMAIN_GTT, 0);
> -	gpu_latency = (results[repeats-1] - results[0]) / (double)(repeats-1);
> +	gpu_latency = (results[repeats-1] - results[1]) / (double)(repeats-2);

How come you don't like the value at 0? Maybe adding a comment would 
make it clearer.

>   
>   	gem_set_domain(fd, obj[2].handle,
>   		       I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
> @@ -238,10 +239,11 @@ static void latency_on_ring(int fd,
>   	igt_assert(offset == obj[2].offset);
>   
>   	gem_set_domain(fd, obj[1].handle, I915_GEM_DOMAIN_GTT, 0);
> -	igt_info("%s: dispatch latency: %.2f, execution latency: %.2f (target %.2f)\n",
> +	igt_info("%s: dispatch latency: %.1fns, execution latency: %.1fns (target %.1fns)\n",
>   		 name,
> -		 (end - start) / (double)repeats,
> -		 gpu_latency, (results[repeats - 1] - results[0]) / (double)(repeats - 1));
> +		 (end - start) / (double)repeats * rcs_clock,
> +		 gpu_latency * rcs_clock,
> +		 (results[repeats - 1] - results[0]) / (double)(repeats - 1) * rcs_clock);
>   
>   	munmap(map, 64*1024);
>   	munmap(results, 4096);
> @@ -620,6 +622,30 @@ rthog_latency_on_ring(int fd, unsigned int engine, const char *name, unsigned in
>   	munmap(results, MMAP_SZ);
>   }
>   
> +static double clockrate(void)
> +{
> +	volatile uint32_t *reg;
> +	uint32_t r_start, r_end;
> +	struct timespec tv;
> +	uint64_t t_start, t_end;
> +	uint64_t elapsed;
> +
> +	reg = (volatile uint32_t *)((volatile char *)igt_global_mmio + RCS_TIMESTAMP);
> +
> +	t_start = igt_nsec_elapsed(&tv);
> +	r_start = *reg;
> +	elapsed = igt_nsec_elapsed(&tv) - t_start;
> +
> +	usleep(1000);
> +
> +	t_end = igt_nsec_elapsed(&tv);
> +	r_end = *reg;
> +	elapsed += igt_nsec_elapsed(&tv) - t_end;
> +
> +	elapsed = (t_end - t_start) + elapsed / 2;
> +	return (r_end - r_start) * 1e9 / elapsed;
> +}
> +
>   igt_main
>   {
>   	const struct intel_execution_engine *e;
> @@ -640,6 +666,10 @@ igt_main
>   			ring_size = 1024;
>   
>   		intel_register_access_init(intel_get_pci_device(), false, device);
> +		rcs_clock = clockrate();
> +		igt_info("RCS timestamp clock: %.3fKHz, %.1fns\n",
> +			 rcs_clock / 1e3, 1e9 / rcs_clock);
> +		rcs_clock = 1e9 / rcs_clock;
>   	}
>   
>   	igt_subtest("all-rtidle-submit")
>