Differential D18401

amd64: align memmove buffers to 16 bytes before using rep movs
ClosedPublic
Actions

Authored by mjg on Dec 1 2018, 1:10 PM.

Details

Reviewers

kib

Commits

rS341364: amd64: align target memmove buffer to 16 bytes before using rep movs

Summary

Sample result from EPYC:

(256-511)... memmove_kernel_align16: 1.232132844 seconds.
(512-1023)... memmove_kernel_align16: 4.041251906 seconds.
(1024-2047)... memmove_kernel_align16: 11.302151396 seconds.
(2048-4095)... memmove_kernel_align16: 17.781004266 seconds.
(256-511)... memmove_kernel: 1.224122136 seconds.
(512-1023)... memmove_kernel: 4.494771289 seconds.
(1024-2047)... memmove_kernel: 13.450829032 seconds.
(2048-4095)... memmove_kernel: 23.493978908 seconds.

Diff Detail

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 21295

Event Timeline

mjg created this revision.Dec 1 2018, 1:10 PM

Herald added a subscriber: imp. · View Herald TranscriptDec 1 2018, 1:10 PM

Harbormaster completed remote builds in B21295: Diff 51468.Dec 1 2018, 1:10 PM

This aligns destination but does not align source. I would expect that the other side of movsb provides comparable improvement ?

sys/amd64/amd64/support.S
305	you can do testb
327	Is it a tab between instruction and register ?

This revision is now accepted and ready to land.Dec 1 2018, 1:35 PM

Closed by commit rS341364: amd64: align target memmove buffer to 16 bytes before using rep movs (authored by mjg). · Explain WhyDec 1 2018, 2:20 PM

This revision was automatically updated to reflect the committed changes.

I did basic tests with changing the alignment of src and slowdowns were very small compared to similarly misaligned dst, at least on EPYC. I may take a closer look later.

sys/amd64/amd64/support.S
327	yes

Revision Contents
Changeset List

Path

Size

sys/

amd64/

support.S

30 lines

Diff 51468

View Options

amd64: align memmove buffers to 16 bytes before using rep movsClosedPublicActions