Page MenuHomeFreeBSD
Differential D1968

Make the LSFR in sched_ule a function and use it always.
ClosedPublic
Actions

Authored by imp on Feb 26 2015, 12:11 AM.
Tags
None
Referenced Files
F106663005: D1968.diff
Fri, Jan 3, 2:10 PM
Unknown Object (File)
Wed, Jan 1, 1:38 AM
Unknown Object (File)
Fri, Dec 13, 8:38 AM
Unknown Object (File)
Nov 27 2024, 4:29 AM
Unknown Object (File)
Nov 20 2024, 10:59 PM
Unknown Object (File)
Oct 10 2024, 8:55 PM
Unknown Object (File)
Oct 7 2024, 4:45 AM
Unknown Object (File)
Oct 7 2024, 4:44 AM
View All Files
Subscribers
rpokala-panasas.com

Details

Reviewers
kmacy
kib
alfred
jmg
jhb
• ian
Commits
rS279349: Create sched_rand() and move the LCG code into that. Call this when
Summary

Create sched_rand() and move the LCG code into that. Call this when
we need randomness in ULE. This removes random() call from the
rebalance interval code.

Submitted by Harrison Grundy

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Skipped
Unit
Tests Skipped

Event Timeline

imp updated this revision to Diff 3983.Feb 26 2015, 12:11 AM
imp retitled this revision from to Make the LSFR in sched_ule a function and use it always..
imp updated this object.
imp edited the test plan for this revision. (Show Details)
imp added reviewers: alfred, jhb, jmg, kib, • ian, kmacy.Feb 26 2015, 12:18 AM
rpokala-panasas.com added a subscriber: rpokala-panasas.com.Feb 26 2015, 12:20 AM
This comment was removed by rpokala-panasas.com.
rpokala-panasas.com added inline comments.Feb 26 2015, 12:22 AM
sys/kern/sched_ule.c
368 ↗(On Diff #3983)

Could we use defined constants for the magic values?

imp updated this revision to Diff 3985.Feb 26 2015, 12:42 AM

This is a LCG not an LSFR, add comments about it.
Also, fix randomness in the rebalancing stuff to be
the full range.

imp added a comment.Feb 26 2015, 12:44 AM
In D1968#5, @rpokala-panasas.com wrote:

Could we use defined constants for the magic values?

Since they are used only once, I updated sched_random() to include where they
came from. Please let me know if you think this is a good balance. They aren't
really meant to be changed since a good LCG is hard to find.

imp updated this object.Feb 26 2015, 12:46 AM
kib added inline comments.Feb 26 2015, 9:30 AM
sys/kern/sched_ule.c
371 ↗(On Diff #3985)

I agree with the gang that prefers to avoid such side-effects in the expressions. It is ugly and does not add anything to the code concisenesses.

Since we move the lines around, I do not see a reason to keep the obfuscation in the code. It already caused the bug when Harrison refactored the code.

720 ↗(On Diff #3985)

There was stray space added at the beginning of line in the original patch. It seems it was imported there as well.

imp updated this revision to Diff 3993.Feb 26 2015, 3:48 PM

Nits from kib

kib edited edge metadata.Feb 26 2015, 4:48 PM
In D1968#11, @imp wrote:

Nits from kib

I am unable to see any changes WRT original patch.

imp updated this revision to Diff 3996.Feb 26 2015, 5:14 PM
imp edited edge metadata.

Update with Kib's changes.

imp updated this revision to Diff 3997.Feb 26 2015, 5:16 PM

No, this time it will really work!

imp added a comment.Feb 26 2015, 5:19 PM
In D1968#13, @kostikbel wrote:
In D1968#11, @imp wrote:

Nits from kib

I am unable to see any changes WRT original patch.

don't know what happened, but they appear to be there now.

kib accepted this revision.Feb 26 2015, 5:21 PM
kib edited edge metadata.
kib added inline comments.
sys/kern/sched_ule.c
367 ↗(On Diff #3997)

Function name should be on the new line.

370 ↗(On Diff #3997)

There should be a blank line after local variables declarations.

373 ↗(On Diff #3997)

There should be a space before '('.

This revision is now accepted and ready to land.Feb 26 2015, 5:21 PM
imp closed this revision.Feb 27 2015, 2:57 AM
imp updated this revision to Diff 4004.

Closed by commit rS279349 (authored by @imp).

Revision Contents
Changeset List

PathSize
head/
sys/
kern/
31 lines

Diff 4004

View Options

head/sys/kern/sched_ule.c

Loading...