<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<div dir="auto">
<div><br>
<div class="gmail_extra"><br>
<div class="gmail_quote">Am 02.08.2019 07:38 schrieb Lionel Landwerlin <lionel.g.landwerlin@intel.com>:<br type="attribution">
<blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>
<div>On 02/08/2019 08:21, Koenig, Christian wrote:<br>
</div>
<blockquote>
<div dir="auto">
<div><br>
<div><br>
<div class="elided-text">Am 02.08.2019 07:17 schrieb Lionel Landwerlin <a href="mailto:lionel.g.landwerlin@intel.com">
<lionel.g.landwerlin@intel.com></a>:<br type="attribution">
<blockquote style="margin:0 0 0 0.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>
<div>On 02/08/2019 08:08, Koenig, Christian wrote:<br>
</div>
<blockquote>
<div dir="auto">Hi Lionel,
<div dir="auto"><br>
</div>
<div dir="auto">Well that looks more like your test case is buggy.</div>
<div dir="auto"><br>
</div>
<div dir="auto">According to the code the ctx1 queue always waits for sem1 and ctx2 queue always waits for sem2.</div>
</div>
</blockquote>
<p><br>
</p>
<p>That's supposed to be the same underlying syncobj because it's exported from one VkDevice as opaque FD from sem1 and imported into sem2.<br>
</p>
</div>
</blockquote>
</div>
</div>
</div>
<div dir="auto"><br>
</div>
<div dir="auto">Well than that's still buggy and won't synchronize at all.</div>
<div dir="auto"><br>
</div>
<div dir="auto">When ctx1 waits for a semaphore and then signals the same semaphore there is no guarantee that ctx2 will run in between jobs.</div>
<div dir="auto"><br>
</div>
<div dir="auto">It's perfectly valid in this case to first run all jobs from ctx1 and then all jobs from ctx2.</div>
</div>
</blockquote>
<p><br>
</p>
<p>That's not really how I see the semaphores working.</p>
<p>The spec describe VkSemaphore as an interface to an internal payload opaque to the application.</p>
<p><br>
</p>
<p>When ctx1 waits on the semaphore, it waits on the payload put there by the previous iteration.</p>
</div>
</blockquote>
</div>
</div>
</div>
<div dir="auto"><br>
</div>
<div dir="auto">And who says that it's not waiting for it's own previous payload?</div>
<div dir="auto"><br>
</div>
<div dir="auto">See if the payload is a counter this won't work either. Keep in mind that this has the semantic of a semaphore. Whoever grabs the semaphore first wins and can run, everybody else has to wait.</div>
<div dir="auto"><br>
</div>
<div dir="auto">
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>
<p>Then it proceeds to signal it by replacing the internal payload.</p>
</div>
</blockquote>
</div>
</div>
</div>
<div dir="auto"><br>
</div>
<div dir="auto">That's an implementation detail of our sync objects, but I don't think that this behavior is part of the Vulkan specification.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Regards,</div>
<div dir="auto">Christian.</div>
<div dir="auto"></div>
<div dir="auto">
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>
<p><br>
</p>
<p>ctx2 then waits on that and replaces the payload again with the new internal synchronization object.</p>
<p><br>
</p>
<p>The internal payload is a dma fence in our case and signaling just replaces a dma fence by another or puts one where there was none before.</p>
<p>So we should have created a dependecy link between all the submissions and then should be executed in the order of QueueSubmit() calls.<br>
</p>
<p><br>
</p>
<p>-Lionel<br>
</p>
<p><br>
</p>
<blockquote>
<div dir="auto">
<div dir="auto"><br>
</div>
<div dir="auto">It only prevents running both at the same time and as far as I can see that still works even with threaded submission.</div>
<div dir="auto"><br>
</div>
<div dir="auto">You need at least two semaphores for a tandem submission.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Regards,</div>
<div dir="auto">Christian.</div>
<div dir="auto">
<div>
<div class="elided-text">
<blockquote style="margin:0 0 0 0.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>
<p><br>
</p>
<blockquote>
<div dir="auto">
<div dir="auto"><br>
</div>
<div dir="auto">This way there can't be any Synchronisation between the two.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Regards,</div>
<div dir="auto">Christian.</div>
</div>
<div><br>
<div class="elided-text">Am 02.08.2019 06:55 schrieb Lionel Landwerlin <a href="mailto:lionel.g.landwerlin@intel.com">
<lionel.g.landwerlin@intel.com></a>:<br type="attribution">
</div>
</div>
<div>
<div>Hey Christian,</div>
<div><br>
</div>
<div>The problem boils down to the fact that we don't immediately create dma fences when calling vkQueueSubmit().</div>
<div>This is delayed to a thread.</div>
<div><br>
</div>
<div>From a single application thread, you can QueueSubmit() to 2 queues from 2 different devices.</div>
<div>Each QueueSubmit to one queue has a dependency on the previous QueueSubmit on the other queue through an exported/imported semaphore.</div>
<div><br>
</div>
<div>From the API point of view the state of the semaphore should be changed after each QueueSubmit().</div>
<div>The problem is that it's not because of the thread and because you might have those 2 submission threads tied to different VkDevice/VkInstance or even different applications (synchronizing themselves outside the vulkan API).</div>
<div><br>
</div>
<div>Hope that makes sense.</div>
<div>It's not really easy to explain by mail, the best explanation is probably reading the test :
<a href="https://gitlab.freedesktop.org/mesa/crucible/blob/master/src/tests/func/sync/semaphore-fd.c#L788">
https://gitlab.freedesktop.org/mesa/crucible/blob/master/src/tests/func/sync/semaphore-fd.c#L788</a></div>
<div><br>
</div>
<div>Like David mentioned you're not running into that issue right now, because you only dispatch to the thread under specific conditions.</div>
<div>But I could build a case to force that and likely run into the same issue.<br>
</div>
<div><br>
</div>
<div>-Lionel<br>
</div>
<div><br>
</div>
<div>On 02/08/2019 07:33, Koenig, Christian wrote:<br>
</div>
<blockquote>
<div>
<div dir="auto">Hi Lionel,
<div dir="auto"><br>
</div>
<div dir="auto">Well could you describe once more what the problem is?</div>
<div dir="auto"><br>
</div>
<div dir="auto">Cause I don't fully understand why a rather normal tandem submission with two semaphores should fail in any way.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Regards,</div>
<div dir="auto">Christian.</div>
</div>
<div><br>
<div>Am 02.08.2019 06:28 schrieb Lionel Landwerlin <a href="mailto:lionel.g.landwerlin@intel.com">
<lionel.g.landwerlin@intel.com></a>:<br type="attribution">
</div>
</div>
</div>
<font size="2"><span style="font-size:11pt">
<div>There aren't CTS tests covering the issue I was mentioning.<br>
But we could add them.<br>
<br>
I don't have all the details regarding your implementation but even with <br>
the "semaphore thread", I could see it running into the same issues.<br>
What if a mix of binary & timeline semaphores are handed to vkQueueSubmit()?<br>
<br>
For example with queueA & queueB from 2 different VkDevice :<br>
vkQueueSubmit(queueA, signal semA);<br>
vkQueueSubmit(queueA, wait on [semA, timelineSemB]); with <br>
timelineSemB triggering a wait before signal.<br>
vkQueueSubmit(queueB, signal semA);<br>
<br>
<br>
-Lionel<br>
<br>
On 02/08/2019 06:18, Zhou, David(ChunMing) wrote:<br>
> Hi Lionel,<br>
><br>
> By the Queue thread is a heavy thread, which is always resident in driver during application running, our guys don't like that. So we switch to Semaphore Thread, only when waitBeforeSignal of timeline happens, we spawn a thread to handle that wait. So we
don't have your this issue.<br>
> By the way, I already pass all your CTS cases for now. I suggest you to switch to Semaphore Thread instead of Queue Thread as well. It works very well.<br>
><br>
> -David<br>
><br>
> -----Original Message-----<br>
> From: Lionel Landwerlin <a href="mailto:lionel.g.landwerlin@intel.com"><lionel.g.landwerlin@intel.com></a><br>
> Sent: Friday, August 2, 2019 4:52 AM<br>
> To: dri-devel <a href="mailto:dri-devel@lists.freedesktop.org"><dri-devel@lists.freedesktop.org></a>; Koenig, Christian
<a href="mailto:Christian.Koenig@amd.com"><Christian.Koenig@amd.com></a>; Zhou, David(ChunMing)
<a href="mailto:David1.Zhou@amd.com"><David1.Zhou@amd.com></a>; Jason Ekstrand <a href="mailto:jason@jlekstrand.net">
<jason@jlekstrand.net></a><br>
> Subject: Threaded submission & semaphore sharing<br>
><br>
> Hi Christian, David,<br>
><br>
> Sorry to report this so late in the process, but I think we found an issue not directly related to syncobj timelines themselves but with a side effect of the threaded submissions.<br>
><br>
> Essentially we're failing a test in crucible :<br>
> func.sync.semaphore-fd.opaque-fd<br>
> This test create a single binary semaphore, shares it between 2 VkDevice/VkQueue.<br>
> Then in a loop it proceeds to submit workload alternating between the 2 VkQueue with one submit depending on the other.<br>
> It does so by waiting on the VkSemaphore signaled in the previous iteration and resignaling it.<br>
><br>
> The problem for us is that once things are dispatched to the submission thread, the ordering of the submission is lost.<br>
> Because we have 2 devices and they both have their own submission thread.<br>
><br>
> Jason suggested that we reestablish the ordering by having semaphores/syncobjs carry an additional uint64_t payload.<br>
> This 64bit integer would represent be an identifier that submission threads will WAIT_FOR_AVAILABLE on.<br>
><br>
> The scenario would look like this :<br>
> - vkQueueSubmit(queueA, signal on semA);<br>
> - in the caller thread, this would increment the syncobj additional u64 payload and return it to userspace.<br>
> - at some point the submission thread of queueA submits the workload and signal the syncobj of semA with value returned in the caller thread of vkQueueSubmit().<br>
> - vkQueueSubmit(queueB, wait on semA);<br>
> - in the caller thread, this would read the syncobj additional<br>
> u64 payload<br>
> - at some point the submission thread of queueB will try to submit the work, but first it will WAIT_FOR_AVAILABLE the u64 value returned in the step above<br>
><br>
> Because we want the binary semaphores to be shared across processes and would like this to remain a single FD, the simplest location to store this additional u64 payload would be the DRM syncobj.<br>
> It would need an additional ioctl to read & increment the value.<br>
><br>
> What do you think?<br>
><br>
> -Lionel<br>
<br>
<br>
</div>
</span></font></blockquote>
<p><br>
</p>
</div>
</blockquote>
<p><br>
</p>
</div>
</blockquote>
</div>
<br>
</div>
</div>
</div>
</blockquote>
<p><br>
</p>
</div>
</blockquote>
</div>
<br>
</div>
</div>
</div>
</body>
</html>