[Intel-gfx] [PATCH i-g-t] overlay/intel-gpu-overlay Fix power consumption not being displayed

Vlad, Marius C marius.c.vlad at intel.com
Wed Nov 18 04:50:10 PST 2015


Just trying to figure out what we have currently. I can redo with with -1, if that's 
OK with you.

Out of curiosity noticed that I915_PERF_ENERGY is not available on my machine, 
is that related to i915_oa? (Haswell).

-----Original Message-----
From: Chris Wilson [mailto:chris at chris-wilson.co.uk] 
Sent: Wednesday, November 18, 2015 2:45 PM
To: Vlad, Marius C <marius.c.vlad at intel.com>
Cc: marius.vlad0 at gmail.com; intel-gfx at lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH i-g-t] overlay/intel-gpu-overlay Fix power consumption not being displayed

On Wed, Nov 18, 2015 at 02:36:22PM +0200, Marius Vlad wrote:
> The power metter was not showing up due to a check over I915_PERF_ENERGY.
> ENOENT is returned when I915_PERF_ENERGY is not available, and we use 
> that for relaying on debugfs i915_energy_uJ.
> 
> Signed-off-by: Marius Vlad <marius.c.vlad at intel.com>
> ---
>  overlay/power.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/overlay/power.c b/overlay/power.c index 6c5c374..6873e7f 
> 100644
> --- a/overlay/power.c
> +++ b/overlay/power.c
> @@ -60,7 +60,7 @@ int power_init(struct power *power)
>  	memset(power, 0, sizeof(*power));
>  
>  	power->fd = perf_open();
> -	if (power->fd != -1)
> +	if (power->fd != -ENOENT)

Nope. The bug is

diff --git a/overlay/power.c b/overlay/power.c index 6c5c374..2f1521b 100644
--- a/overlay/power.c
+++ b/overlay/power.c
@@ -45,7 +45,7 @@ static int perf_open(void)
 
        attr.type = i915_type_id();
        if (attr.type == 0)
-               return -ENOENT;
+               return -1;
        attr.config = I915_PERF_ENERGY;
 
        attr.read_format = PERF_FORMAT_TOTAL_TIME_ENABLED;

and the more pressing concern is that we still don't have support for the more accurate tracking from the kernel.
-Chris

--
Chris Wilson, Intel Open Source Technology Centre


More information about the Intel-gfx mailing list