Public bug reported:
[Feature Description]
Starting in ICL/ICX there is a performance improvement to the "rep movsb" copy
instruction for short strings.
>From the optimization guide:
3.7.6.1 Fast Short REP MOVSB
Beginning with processors based on Ice Lake Client microarchitecture,
REP MOVSB performance of short operations is enhanced. The enhancement
applies to string lengths between 1 and 128 bytes long. Support for
fast-short REP MOVSB is enumerated by the CPUID feature flag: CPUID
[EAX=7H, ECX=0H).EDX.FAST_SHORT_REP_MOVSB[bit 4] = 1. There is no change
in the REP STOS performance.
Linux could check for this CPUID bit and make the flag "fsrm" appear
in /proc/cpuinfo flags.
#define X86_FEATURES_FSRM (18*32+ 4) /* Fast short rep movsb */
At some future point copy functions could be modified to check for this
feature and have some alternate copy loop for short strings on CPUs that
do NOT have FSRM (benchmarking needed to see if this actually helps).
Target Kernel: 5.6
Target Release: 20.10
[HW/SW Information]
IceLake Server
[Business Justification]
performance
** Affects: intel
Importance: Undecided
Status: New
** Affects: ubuntu
Importance: Undecided
Status: New
** Tags: intel-kernel-20.10
** Also affects: ubuntu
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1883630
Title:
[ICX]ICL/ICX Fast Short Rep Movs
To manage notifications about this bug go to:
https://bugs.launchpad.net/intel/+bug/1883630/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
