[igt-dev] [PATCH v4 05/14] xe_exec_balancer: Enable parallel submission and compute mode

[Souza, Jose]

On Thu, 2023-09-28 at 11:05 +0000, Francois Dugast wrote:
> From: Matthew Brost <matthew.brost at intel.com>
> 
> This is now supported. Test it.

Reviewed-by: José Roberto de Souza <jose.souza at intel.com>

> 
> Signed-off-by: Matthew Brost <matthew.brost at intel.com>
> Signed-off-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> ---
>  tests/intel/xe_exec_balancer.c | 21 +++++++++++++++------
>  1 file changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/tests/intel/xe_exec_balancer.c b/tests/intel/xe_exec_balancer.c
> index 0314b4cd2..a4a438db7 100644
> --- a/tests/intel/xe_exec_balancer.c
> +++ b/tests/intel/xe_exec_balancer.c
> @@ -383,6 +383,12 @@ test_exec(int fd, int gt, int class, int n_exec_queues, int n_execs,
>   * @virtual-userptr-rebind:		virtual userptr rebind
>   * @virtual-userptr-invalidate:		virtual userptr invalidate
>   * @virtual-userptr-invalidate-race:	virtual userptr invalidate racy
> + * @parallel-basic:			parallel basic
> + * @parallel-userptr:			parallel userptr
> + * @parallel-rebind:			parallel rebind
> + * @parallel-userptr-rebind:		parallel userptr rebind
> + * @parallel-userptr-invalidate:	parallel userptr invalidate
> + * @parallel-userptr-invalidate-race:	parallel userptr invalidate racy
>   */
>  
>  static void
> @@ -460,8 +466,8 @@ test_cm(int fd, int gt, int class, int n_exec_queues, int n_execs,
>  		};
>  		struct drm_xe_exec_queue_create create = {
>  			.vm_id = vm,
> -			.width = 1,
> -			.num_placements = num_placements,
> +			.width = flags & PARALLEL ? num_placements : 1,
> +			.num_placements = flags & PARALLEL ? 1 : num_placements,
>  			.instances = to_user_pointer(eci),
>  			.extensions = to_user_pointer(&ext),
>  		};
> @@ -470,6 +476,7 @@ test_cm(int fd, int gt, int class, int n_exec_queues, int n_execs,
>  					&create), 0);
>  		exec_queues[i] = create.exec_queue_id;
>  	}
> +	exec.num_batch_buffer = flags & PARALLEL ? num_placements : 1;
>  
>  	sync[0].addr = to_user_pointer(&data[0].vm_sync);
>  	if (bo)
> @@ -487,8 +494,12 @@ test_cm(int fd, int gt, int class, int n_exec_queues, int n_execs,
>  		uint64_t batch_addr = addr + batch_offset;
>  		uint64_t sdi_offset = (char *)&data[i].data - (char *)data;
>  		uint64_t sdi_addr = addr + sdi_offset;
> +		uint64_t batches[MAX_INSTANCE];
>  		int e = i % n_exec_queues;
>  
> +		for (j = 0; j < num_placements && flags & PARALLEL; ++j)
> +			batches[j] = batch_addr;
> +
>  		b = 0;
>  		data[i].batch[b++] = MI_STORE_DWORD_IMM_GEN4;
>  		data[i].batch[b++] = sdi_addr;
> @@ -500,7 +511,8 @@ test_cm(int fd, int gt, int class, int n_exec_queues, int n_execs,
>  		sync[0].addr = addr + (char *)&data[i].exec_sync - (char *)data;
>  
>  		exec.exec_queue_id = exec_queues[e];
> -		exec.address = batch_addr;
> +		exec.address = flags & PARALLEL ?
> +			to_user_pointer(batches) : batch_addr;
>  		xe_exec(fd, &exec);
>  
>  		if (flags & REBIND && i + 1 != n_execs) {
> @@ -661,9 +673,6 @@ igt_main
>  					test_exec(fd, gt, class, 1, 0,
>  						  s->flags);
>  
> -		if (s->flags & PARALLEL)
> -			continue;
> -
>  		igt_subtest_f("once-cm-%s", s->name)
>  			xe_for_each_gt(fd, gt)
>  				xe_for_each_hw_engine_class(class)