« Concrete Blocks | Home | Internal Memory Bottlenecks and Their Removal »

November 13, 2009

"Look Ma, No Busywaits!"

When the CPU needs to do something which depends on a result which the GPU is currently working on, it has to wait for the GPU to catch up.  One of the biggest problems with the current architecture of xf86-video-glamo, both DRM and non-DRM versions, is that they do this waiting by spinning in a tight loop, each time checking the current status of the GPU, until it's caught up.  This isn't great for a few reasons.  It makes no use of the parallelism between the CPU and the GPU, so precious CPU time is being wasted while something more useful could be being done.  If there's nothing else to do, then the CPU could be sleeping - reducing power consumption.

Most GPUs, including Glamo, have a mechanism for being a little smarter.  The kernel can ask the chip to trigger an interrupt when a certain point in the command queue has been reached.  When a process needs to wait, the kernel can send it to sleep and watch out for the interrupt.  When it happens, the process can be quickly woken back up in a low-latency fashion, meaning that the process gets back to work with very little latency.

This week, I've been implementing this kind of thing for the Glamo DRM driver.  It goes a bit like this:

  • Process submits some rendering commands via one of the command submission ioctls.
  • Kernel driver places rendering commands on Glamo's command queue.
  • Process needs to wait for the GPU to catch up, so calls the wait ioctl.
  • Kernel driver puts an extra sequence of commands, called a fence, onto the command queue.  A unique number is associated with the fence.  The number is recorded by the kernel.
  • When the GPU processes the fence, it raises the interrupt and places a unique number into a certain register.
  • The interrupt handler checks this number, and wakes up the corresponding process.
I wrote a test program which tells Glamo to fill the whole screen with colour as fast as it can, waiting for the GPU to catch up each time.  The task was to make the program run with close to zero CPU usage while still getting the full framerate that I could get using busywaits.  The task was achieved successfully, and here's a screenshot to prove it.  The framerate - just below 50fps when doing fills of the entire VGA screen - was exactly the same with busywaits.  It even went up a little (to 50-51fps) when I improved the interrupt handling.

Things aren't always so great.  When the command sequence to be executed is very short, the overheads of fencing and scheduling become significant, and the overall rate drops.  However, it shouldn't be too difficult to design some kind of heuristic to use busywaits as a low-latency strategy in such cases.

There are still a few problems to iron out.  The fence mechanism seems to be able to fall out of sync with things, leading to processes waiting for too long (or even forever).  But when it works, some things do seem to feel a little faster in general use.

Geeks may be interested in the actual code.

5 Comments

Nice work! Thanks working on glamo. :)
Does this improvement could be bring to qtmoko (qtextended?) whereas it use fb and not a xserver?

Same question for the over work here:
https://www.bitwiz.org.uk/s/2009/11/internal-memory-bottlenecks-and-their-removal.html

THanks for your answer. :)

I was one of those asking for comments on your website... just to thank you for your work (hope this helps in feeling some interest in what you're doing).

d

Cool! Thanks for the insight.

Leave a comment