[igt-dev] [PATCH] [PATCH i-g-t]tests/i915/gem_exec_create: Added __for_each_physical_engine to utilize all available engines.

[Tvrtko Ursulin]

On 06/03/2020 03:20, Ch, Sai Gowtham wrote:
> After fork we need to transfer the engine map from parent fd default
> context. (gem_context_copy_engines)
> 
> Also, since we have __for_each_physical_engine in the top level
> igt_fixture, I think _all_ subtests run with default ctx engine map. So
> all which submit to one from either all_engines or ppgtt_engines need to
> make sure engine maps are aligned.
> 
> it was suggested by tvrtko in one of the patch: https://patchwork.freedesktop.org/patch/351622/?series=72889&rev=1

I was sadly wrong there, it is not required to reopen and copy engines 
after fork. So __for_each_physical_engine should do it in this case.

Regards,

Tvrtko

P.S. You could also dynamically size (allocate) the engines array while 
at it to be more in line with the other recent cleanups.

Hm.. actually, do we even need the array? Andi, 
__for_each_physical_engine will always create engine map with 
consecutive e->flags integers, right? There can be no holes?

> 
> -----Original Message-----
> From: Dec, Katarzyna <katarzyna.dec at intel.com>
> Sent: Thursday, March 5, 2020 6:13 PM
> To: Ch, Sai Gowtham <sai.gowtham.ch at intel.com>; igt-dev at lists.freedesktop.org
> Subject: Re: [igt-dev] [PATCH] [PATCH i-g-t]tests/i915/gem_exec_create: Added __for_each_physical_engine to utilize all available engines.
> 
> On Tue, Feb 18, 2020 at 05:45:49PM +0530, sai gowtham wrote:
>> From: saigowth <sai.gowtham.ch at intel.com>
>>
>> Replaced the legacy for_each_engine* defines with the ones implemented
>> in the gem_engine_topology library.
>>
>> Cc: Dec Katarzyna <katarzyna.dec at intel.com>
>> Cc: Ursulin Tvrtko <tvrtko.ursulin at intel.com>
>> Signed-off-by: sai gowtham <sai.gowtham.ch at intel.com>
>> ---
>>   tests/i915/gem_exec_create.c | 29 ++++++++++++++++-------------
>>   1 file changed, 16 insertions(+), 13 deletions(-)
>>
>> diff --git a/tests/i915/gem_exec_create.c
>> b/tests/i915/gem_exec_create.c index 102a161d..c70572af 100644
>> --- a/tests/i915/gem_exec_create.c
>> +++ b/tests/i915/gem_exec_create.c
>> @@ -58,14 +58,15 @@ static double elapsed(const struct timespec
>> *start, const struct timespec *end)
>>   
>>   static void all(int fd, unsigned flags, int timeout, int ncpus)  {
>> +	const struct intel_execution_engine2 *e;
>>   	const uint32_t bbe = MI_BATCH_BUFFER_END;
>>   	struct drm_i915_gem_execbuffer2 execbuf;
>>   	struct drm_i915_gem_exec_object2 obj;
>>   	unsigned engines[16], nengine;
>>   
>>   	nengine = 0;
>> -	for_each_physical_engine(e, fd)
>> -		engines[nengine++] = eb_ring(e);
>> +	__for_each_physical_engine(fd, e)
>> +		engines[nengine++] = e->flags;
>>   	igt_require(nengine);
>>   
>>   	memset(&obj, 0, sizeof(obj));
>> @@ -86,38 +87,40 @@ static void all(int fd, unsigned flags, int
>> timeout, int ncpus)
>>   
>>   	intel_detect_and_clear_missed_interrupts(fd);
>>   	igt_fork(child, ncpus) {
>> +		int i915;
>>   		struct timespec start, now;
>>   		unsigned long count;
>>   		double time;
>> -
>> +		i915 = gem_reopen_driver(fd);
>> +		gem_context_copy_engines(fd, 0, i915, 0);
> Why do we need to reopen device and copy it's context? What's the purpose?
> Kasia
>>   		count = 0;
>>   		clock_gettime(CLOCK_MONOTONIC, &start);
>>   		do {
>>   			for (int loop = 0; loop < 1024; loop++) {
>>   				for (int n = 0; n < nengine; n++) {
>> -					obj.handle =  gem_create(fd, 4096);
>> -					gem_write(fd, obj.handle, 0, &bbe, sizeof(bbe));
>> +					obj.handle =  gem_create(i915, 4096);
>> +					gem_write(i915, obj.handle, 0, &bbe, sizeof(bbe));
>>   					execbuf.flags &= ~ENGINE_FLAGS;
>>   					execbuf.flags |= engines[n];
>> -					gem_execbuf(fd, &execbuf);
>> +					gem_execbuf(i915, &execbuf);
>>   					if (flags & LEAK)
>> -						gem_madvise(fd, obj.handle, I915_MADV_DONTNEED);
>> +						gem_madvise(i915, obj.handle, I915_MADV_DONTNEED);
>>   					else
>> -						gem_close(fd, obj.handle);
>> +						gem_close(i915, obj.handle);
>>   				}
>>   			}
>>   			count += nengine * 1024;
>>   			clock_gettime(CLOCK_MONOTONIC, &now);
>>   		} while (elapsed(&start, &now) < timeout); /* Hang detection ~120s */
>> -		obj.handle =  gem_create(fd, 4096);
>> -		gem_write(fd, obj.handle, 0, &bbe, sizeof(bbe));
>> +		obj.handle =  gem_create(i915, 4096);
>> +		gem_write(i915, obj.handle, 0, &bbe, sizeof(bbe));
>>   		for (int n = 0; n < nengine; n++) {
>>   			execbuf.flags &= ~ENGINE_FLAGS;
>>   			execbuf.flags |= engines[n];
>> -			gem_execbuf(fd, &execbuf);
>> +			gem_execbuf(i915, &execbuf);
>>   		}
>> -		gem_sync(fd, obj.handle);
>> -		gem_close(fd, obj.handle);
>> +		gem_sync(i915, obj.handle);
>> +		gem_close(i915, obj.handle);
>>   		clock_gettime(CLOCK_MONOTONIC, &now);
>>   
>>   		time = elapsed(&start, &now) / count;
>> --
>> 2.25.0
>>
>> _______________________________________________
>> igt-dev mailing list
>> igt-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/igt-dev
> _______________________________________________
> igt-dev mailing list
> igt-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/igt-dev
>