Messages in this thread | | | Subject | Re: [PATCH v5 14/28] x86/fpu/xstate: Prevent unauthorised use of dynamic user state | From | "Liu, Jing2" <> | Date | Mon, 28 Jun 2021 18:11:38 +0800 |
| |
On 6/17/2021 3:28 AM, Dave Hansen wrote: > On 6/16/21 12:23 PM, Bae, Chang Seok wrote: >> On Jun 16, 2021, at 12:01, Hansen, Dave <dave.hansen@intel.com> wrote: >>> On 6/16/21 11:47 AM, Bae, Chang Seok wrote: >>>> Reading XINUSE via XGETBV is cheap but not free. I don't know spending a >>>> hundred cycles for this WARN is big deal but this is one of the most >>>> performance-critical paths. >>> Is XGETBV(1) really a hundred cycles? That seems absurdly high for a >>> non-serializing register read. >> This was checked to convince the benefit intended by PATCH25 -- >> https://lore.kernel.org/lkml/20210523193259.26200-26-chang.seok.bae@intel.com/ > That's odd. How is it possible that the performance of XGETBV(1) > informed the design of that patch without there being any mention of > XGETBV in the comments or changelog? Hi Chang,
I noticed the XGETBV(1) cycles you ran, however I calculated only ~16 cycles in the corresponding machine.
BRs, Jing
| |