Device: Bump PIO and PCI latencies to more reasonable values

Information
Submitter:	Andreas Hansson
Repository:	gem5
Branch:	default
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Changeset 9198:213a0c0c3e52
---------------------------
Device: Bump PIO and PCI latencies to more reasonable values

This patch addresses a previously highlighted issue with the default
latencies used for PIO and PCI devices. The values are merely educated
guesses and might not represent the particular system you want to
model. However, the values in this patch are definitely far more
realistic than the previous ones.

In i8254xGBe, the writeConfig method is updated to use configDelay
instead of pioDelay.

A follow-up patch will update the regression stats.

Testing Done

util/regress all passing (disregarding t1000 and eio) with associated stats updates

Issue Summary

Description	From	Last Updated	Status
Can you elaborate on this change? It seems kind of random.	Steve Reinhardt	Sept. 9, 2012, 9:56 p.m.	Resolved

Ship It!

src/dev/i8254xGBe.cc (Diff revision 1)

Can you elaborate on this change?  It seems kind of random.

Show all issues

Andreas Hansson Sept. 7, 2012, 11:11 p.m. (Sept. 7, 2012, 11:11 p.m.)

Since the method is writeConfig I assume it was a typo that it used to say pioDelay. If that's wrong then I'm happy to change back. All other read/writeConfig end with a return configDelay.

Steve Reinhardt Sept. 8, 2012, 12:40 a.m. (Sept. 8, 2012, 12:40 a.m.)

OK, that makes sense... can you add a sentence to that effect to the commit message?

This leads me to the larger questions of (1) do we really need separate latencies for config-space accesses vs. non-config PIO accesses and (2) if so, couldn't we at least comment those variables in pcidev.hh, but those are outside the scope of this patch...

Andreas Hansson Sept. 9, 2012, 6:48 p.m. (Sept. 9, 2012, 6:48 p.m.)

I'll add a comment.

I have no personal preference regarding (1). I don't know if real devices have enough asymmetry between config-space and non-config space to warrant having two latencies as we do currently. I would think the best way forward is to prune the configDelay unless someone really thinks it is worth holding on to it.

Ali Saidi Sept. 9, 2012, 11:16 p.m. (Sept. 9, 2012, 11:16 p.m.)

I'd imagine they do, but config accesses are rare enough that it probably doesn't matter too much.

Description:

~		Changeset 9187:7f0d5ac4aefd
	~	Changeset 9198:213a0c0c3e52

		Device: Bump PIO and PCI latencies to more reasonable values

		This patch addresses a previously highlighted issue with the default
		latencies used for PIO and PCI devices. The values are merely educated
		guesses and might not represent the particular system you want to
		model. However, the values in this patch are definitely far more
		realistic than the previous ones.

	+	In i8254xGBe, the writeConfig method is updated to use configDelay
	+	instead of pioDelay.
	+
		A follow-up patch will update the regression stats.

Diff:

Revision 2 (+4 -6)

Show changes

	src/dev/Device.py
	src/dev/Pci.py
	src/dev/i8254xGBe.cc

Ship It!

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Screenshots

Files

Issue Summary

Description:

Changeset 9187:7f0d5ac4aefd

Changeset 9198:213a0c0c3e52

Diff:

Status: Closed (submitted)