Re: [PATCH 3/7] RDMA/qedr: eliminate duplicate barriers on weakly-ordered archs

2018-03-15 Thread Jason Gunthorpe
On Tue, Mar 13, 2018 at 11:20:24PM -0400, Sinan Kaya wrote: > Code includes wmb() followed by writel() in multiple places. writel() > already has a barrier on some architectures like arm64. > > This ends up CPU observing two barriers back to back before executing the > register write. > > Since

Re: [PATCH 3/7] RDMA/qedr: eliminate duplicate barriers on weakly-ordered archs

2018-03-14 Thread okaya
On 2018-03-14 00:12, Jason Gunthorpe wrote: On Tue, Mar 13, 2018 at 11:20:24PM -0400, Sinan Kaya wrote: Code includes wmb() followed by writel() in multiple places. writel() already has a barrier on some architectures like arm64. This ends up CPU observing two barriers back to back before

Re: [PATCH 3/7] RDMA/qedr: eliminate duplicate barriers on weakly-ordered archs

2018-03-13 Thread Jason Gunthorpe
On Tue, Mar 13, 2018 at 11:20:24PM -0400, Sinan Kaya wrote: > Code includes wmb() followed by writel() in multiple places. writel() > already has a barrier on some architectures like arm64. > > This ends up CPU observing two barriers back to back before executing the > register write. > > Since

[PATCH 3/7] RDMA/qedr: eliminate duplicate barriers on weakly-ordered archs

2018-03-13 Thread Sinan Kaya
Code includes wmb() followed by writel() in multiple places. writel() already has a barrier on some architectures like arm64. This ends up CPU observing two barriers back to back before executing the register write. Since code already has an explicit barrier call, changing writel() to