Messages in this thread | | | Date | Tue, 9 Dec 2014 15:16:25 -0800 | Subject | Re: [PATCH] PCI: Clear bridge MEM_64 flag if one child does not support it | From | Yinghai Lu <> |
| |
On Tue, Dec 9, 2014 at 2:15 PM, Marek Kordík <kordikmarek@gmail.com> wrote: > On 12/09/2014 08:42 PM, Yinghai Lu wrote: >> >> On Tue, Dec 9, 2014 at 10:22 AM, Bjorn Helgaas <bhelgaas@google.com> >> wrote: >>> >>> On Tue, Dec 9, 2014 at 11:07 AM, Marek Kordík <kordikmarek@gmail.com> >>> wrote: >>>> >>>> I have built and booted your branch and it works well. Do you want me to >>>> attach some logs? >>>> (I am new here and I have read http://www.tux.org/lkml/ and I don't want >> >> Can you run some graphics benchmark program to check the performance >> between >> 1. >> >> git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git >> branch: for-pci-allocate-fit-3.18 >> >> 2. >> v 3.18 + clear mmio64 flags when children device does not support it >> >> > I have run Unigine Heaven benchmark on both versions (I tried version 2. > with and also without kernel parameters "debug ignore_loglevel pci=realloc") > and the performance of each version was the same (70-71 points). I tried to > run this benchmark also with kernel 3.15.10 and the result was 67 points. > Tomorrow I can try to run some more benchmarks, today I didn't have much > time.
so putting mem pref under bridge mem does not cause performance loss?
Thanks
Yinghai
| |