[PATCH 8/8] amdgpu: use sync file for shared semaphores (v2.1)

[Dave Airlie]

On 11 April 2017 at 22:42, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> On Tue, Apr 11, 2017 at 01:22:20PM +1000, Dave Airlie wrote:
>> +static int amdgpu_sem_lookup_and_sync(struct amdgpu_cs_parser *p,
>> +                                   uint32_t handle)
>> +{
>> +     int r;
>> +     struct dma_fence *old_fence;
>> +
>> +     r = drm_syncobj_swap_fences(p->filp, handle, NULL, &old_fence);
>> +     if (r)
>> +             return r;
>
> I'm a bit puzzled over this interface as this means that all
> in-semaphores are not reusable. That seems a bit odd as it means
> userspace can't use the results from one engine on two+ parallel
> engines. Searching kronos for VkSemaphore
> https://www.khronos.org/registry/vulkan/specs/1.0/man/html/VkSemaphore.html
> isn't enlightening...

You need to look in the vulkan spec itself.

But yes this is the semantics, vulkan semaphores are explicitly 1:1 objects.

One wait must have one signal. It's a semaphore not a fence, so you can't
have multiple engines getting the results from one semaphore.

Dave.