Re: [PATCH i-g-t v2 4/7] tests/gem_scheduler: Add gem_scheduler test

From: Dave Gordon <david.s.gordon@intel.com>
To: Chris Wilson <chris@chris-wilson.co.uk>,
	Derek Morton <derek.j.morton@intel.com>,
	intel-gfx@lists.freedesktop.org, daniel.vetter@ffwll.ch
Subject: Re: [PATCH i-g-t v2 4/7] tests/gem_scheduler: Add gem_scheduler test
Date: Mon, 7 Mar 2016 12:40:38 +0000	[thread overview]
Message-ID: <56DD76C6.3090704@intel.com> (raw)
In-Reply-To: <20160302201501.GP30782@nuc-i3427.alporthouse.com>

On 02/03/16 20:15, Chris Wilson wrote:
>> +static struct ring {
>> +	const char *name;
>> +	int id;
>> +	bool exists;
>> +} rings[] = {
>> +	{ "render", I915_EXEC_RENDER, false },
>> +	{ "bsd1",    I915_EXEC_BSD | 1<<13, false },
>> +	{ "bsd2",    I915_EXEC_BSD | 2<<13, false },
>
> This is wrong. The timeline is coupled to the exec_id, which is the same
> for "both" BSD engines. To fix that, properly split up the two rings
> with separate ids and deprecate HAS_BSD2.
>
>> +	{ "blt",    I915_EXEC_BLT, false },
>> +	{ "vebox",  I915_EXEC_VEBOX, false },
>> +};
>> +
>> +#define NBR_RINGS (sizeof(rings)/sizeof(struct ring))
>
> Also see intel_execution_engines to save on duplicating code.
> -Chris

exec_id is indeed the same for both BSD engines, but it is used ONLY in
i915_gem_busy_ioctl(), whose result can tell you that "a" BSD engine is 
busy but not which one -- probably callers don't care anyway.

.Dave.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx