Messages in this thread | | | Date | Fri, 9 Jan 2015 16:47:52 -0800 | From | Greg Kroah-Hartman <> | Subject | Re: [PATCH char-misc-next 02/13] misc: mic: SCIF ring buffer infrastructure |
| |
On Fri, Jan 09, 2015 at 04:28:19PM -0800, Sudeep Dutt wrote: > On Fri, 2015-01-09 at 15:05 -0800, Greg Kroah-Hartman wrote: > > On Wed, Dec 10, 2014 at 11:47:42AM -0800, Sudeep Dutt wrote: > > > SCIF ring buffer is a single producer, single consumer byte stream ring > > > buffer optimized for avoiding reads across the PCIe bus. The ring buffer is > > > used to implement a receive queue for SCIF driver messaging between nodes and > > > for byte stream messaging between SCIF endpoints. Each SCIF node has a > > > receive queue for every other SCIF node, and each connected endpoint has a > > > receive queue for messages from its peer. This pair of receive queues is > > > referred to as a SCIF queue pair. > > > > And the reason you aren't using the built-in kernel ring buffer code is > > why? > > > > This simple byte stream ring buffer is in our performance data path for > small messages and is optimized to avoid reads across the PCIe bus while > adding the required barriers and hardware workarounds for the MIC > Coprocessor. I will add some more documentation here in the v2 which I > will post early next week. > > We did not find other ring buffers in the kernel which were tailored for > our use case across PCIe. I am guessing (please correct me) that you are > referring to the ring buffer in include/linux/ring_buffer.h. It does not > seem to be designed for being used between two independent OS's across > the PCIe bus. Please let me know your thoughts.
No, it's designed for an in-kernel ringbuffer, so you need to document exactly this.
thanks,
greg k-h
| |