Re: [PATCH v2 1/7] drm/xe: Adjust long-running workload timeslices to reasonable values

["Thomas Hellström" <thomas.hellstrom@linux.intel.com>]

On Fri, 2025-12-12 at 10:28 -0800, Matthew Brost wrote:
> A 10ms timeslice for long-running workloads is far too long and
> causes
> significant jitter in benchmarks when the system is shared. Adjust
> the
> value to 5ms for preempt-fencing VMs, as the resume step there is
> quite
> costly as memory is moved around, and set it to zero for pagefault
> VMs,
> since switching back to pagefault mode after dma-fence mode is
> relatively fast.
> 
> Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel
> GPUs")
> Cc: stable@vger.kernel.org
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>

Please add a comment in the commit message explaining why the type was
changed.

Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>


> ---
>  drivers/gpu/drm/xe/xe_vm.c       | 5 ++++-
>  drivers/gpu/drm/xe/xe_vm_types.h | 2 +-
>  2 files changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index c2012d20faa6..4648f8a458cf 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -1508,7 +1508,10 @@ struct xe_vm *xe_vm_create(struct xe_device
> *xe, u32 flags, struct xe_file *xef)
>  	INIT_WORK(&vm->destroy_work, vm_destroy_work_func);
>  
>  	INIT_LIST_HEAD(&vm->preempt.exec_queues);
> -	vm->preempt.min_run_period_ms = 10;	/* FIXME: Wire up to
> uAPI */
> +	if (flags & XE_VM_FLAG_FAULT_MODE)
> +		vm->preempt.min_run_period_ms = 0;
> +	else
> +		vm->preempt.min_run_period_ms = 5;
>  
>  	for_each_tile(tile, xe, id)
>  		xe_range_fence_tree_init(&vm->rftree[id]);
> diff --git a/drivers/gpu/drm/xe/xe_vm_types.h
> b/drivers/gpu/drm/xe/xe_vm_types.h
> index 3bf912bfbdcc..18bad1dd08e6 100644
> --- a/drivers/gpu/drm/xe/xe_vm_types.h
> +++ b/drivers/gpu/drm/xe/xe_vm_types.h
> @@ -263,7 +263,7 @@ struct xe_vm {
>  		 * @min_run_period_ms: The minimum run period before
> preempting
>  		 * an engine again
>  		 */
> -		s64 min_run_period_ms;
> +		unsigned int min_run_period_ms;
>  		/** @exec_queues: list of exec queues attached to
> this VM */
>  		struct list_head exec_queues;
>  		/** @num_exec_queues: number exec queues attached to
> this VM */