cpu: Add SMT support to MinorCPU

Information
Submitter:	Curtis Dunham
Repository:	gem5
Branch:	default
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

cpu: Add SMT support to MinorCPU

This patch adds SMT support to the MinorCPU.  Currently
RoundRobin or Random thread scheduling are supported.

Testing Done

Issue Summary

Description	From	Last Updated	Status
Style: all other class defns. have the { on the next line.	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Over indendation. Pull the first line up to the = ?	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Add [tid:] to this and other DPRINTFs like you've done in other cases.	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Drop open { on multi-line condition to next line to match the rest of the code.	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
minorTrace output really should reflect the threading. I notice you've added a thread index to (some?) blocks. Are you planning ...	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Doesn't this disappear into the thread info?	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Is this missing clause correct? The only_issue_microops flag is a bodgy way of handling draining, do you replace it with ...	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Id the full id comparison done to include the threadId? As the execSeqNums will be the same (as you didn't ...	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
"Unstalling fu: %d" ...	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Original spelling was correct.	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Next line before {	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
<nl> {	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Spelling	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
<nl> {	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
He he	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Style	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped
Add more than one set of thread structures to the default .pic file?	Andrew Bardsley	March 30, 2016, 4:50 a.m.	Dropped

src/cpu/minor/decode.hh (Diff revision 1)

Style: all other class defns. have the { on the next line.

Show all issues

src/cpu/minor/decode.hh (Diff revision 1)

Dropping Reservable is a good move.

src/cpu/minor/decode.cc (Diff revision 1)

Over indendation. Pull the first line up to the = ?

Show all issues

src/cpu/minor/decode.cc (Diff revision 1)

Add [tid:] to this and other DPRINTFs like you've done in other cases.

Show all issues

src/cpu/minor/decode.cc (Diff revision 1)

Drop open { on multi-line condition to next line to match the rest of the code.

Show all issues

src/cpu/minor/decode.cc (Diff revision 1)

minorTrace output really should reflect the threading. I notice you've added a thread index to (some?) blocks. Are you planning to dump all thread info in minorTrace() in another patch?

Show all issues

src/cpu/minor/execute.hh (Diff revision 1)

Doesn't this disappear into the thread info?

Show all issues

src/cpu/minor/execute.cc (Diff revision 1)

Is this missing clause correct? The only_issue_microops flag is a bodgy way of handling draining, do you replace it with another mechanism?

Show all issues

src/cpu/minor/execute.cc (Diff revision 1)

Id the full id comparison done to include the threadId? As the execSeqNums will be the same (as you didn't pass the condition on the previous clause), just comparing thread ids will work here, won't it?

Show all issues

src/cpu/minor/execute.cc (Diff revision 1)