cpu: Resolve targets of predicted 'taken' conditional direct branches at decode (o3)

Information
Submitter:	Arthur Perais
Repository:	gem5
Branch:
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Changeset 11707:1d085f66c4ca

The target of taken conditional direct branches does not
need to be resolved in IEW: the target can be computed at
decode, usually using the decoded instruction word and the PC.

The higher-than-necessary penalty is taken only on conditional
branches that are predicted taken but miss in the BTB. Thus,
this is mostly inconsequential on IPC if the BTB is big/associative
enough (fewer capacity/conflict misses). Nonetheless, what gem5
simulates is not representative of how conditional branch targets
can be handled.

Testing Done

util/regress --modes=se

build/NULL/tests/opt/quick/se/51.memcheck/null/none/memcheck: FAILED! (Python says "NameError: name 'TrafficGen' is not defined"). I'm guessing this is OK as I get the same error without my patch.

Summary:

-	cpu: Resolve targets of predicted 'taken' conditional direct branches at decode (O3)
+	cpu: Resolve targets of predicted 'taken' conditional direct branches at decode (o3)

Description:

		Changeset 11707:1d085f66c4ca

~		cpu: Resolve targets of predicted 'taken' conditional direct branches at decode (O3)
	~	cpu: Resolve targets of predicted 'taken' conditional direct branches at decode (o3)

		The target of taken conditional direct branches does not
		need to be resolved in IEW: the target can be computed at
		decode, usually using the decoded instruction word and the PC.

		The higher-than-necessary penalty is taken only on conditional
		branches that are predicted taken but miss in the BTB. Thus,
		this is mostly inconsequential on IPC if the BTB is big/associative
		enough (fewer capacity/conflict misses). Nonetheless, what gem5
		simulates is not representative of how conditional branch targets
		can be handled.

Ship It!

The error you're seeing is because you don't have Google's protobuf library installed. Though it has no affect on what you're seeing.

Can you run the full system linux boot regression with O3 to be sure that it still works correctly? This is a more rigorous test than any SE mode tests.

Arthur Perais Nov. 23, 2016, 6:44 a.m. (Nov. 23, 2016, 6:44 a.m.)

Here it is. As expected, nothing changes regarding performance or correctness (as proxied by sim_ints and sim_ops).
Of course, the number of branches resolved at Decode increases dramatically (omitted here).

***** build/ALPHA/tests/opt/long/fs/10.linux-boot/alpha/linux/tsunami-minor: passed.

--- long/fs/10.linux-boot/alpha/linux/tsunami-o3 ---
*** gem5: OK
*** stat_diff: FAILURE: Statistics mismatch
Statistics mismatch
Maximum error magnitude: +609.982116%

                                  Reference  New Value   Abs Diff   Pct Chg
Key statistics:

  host_inst_rate                     239114     252487      13373    +5.59%
  host_mem_usage                     338260     509164     170904   +50.52%
  sim_insts                        52969539   52969539          0    +0.00%
  sim_ops                          52969539   52969539          0    +0.00%
  sim_ticks                      1865011607500 1865011901500     294000    +0.00%
  system.cpu.commit.committedInsts   56159642   56159642          0    +0.00%
  system.cpu.commit.committedOps   56159642   56159642          0    +0.00%
  system.cpu.committedInsts        52969539   52969539          0    +0.00%
  system.cpu.committedOps          52969539   52969539          0    +0.00%
  system.cpu.ipc                   0.408632   0.408630  -0.000002    -0.00%
  system.cpu.ipc_total             0.408632   0.408630  -0.000002    -0.00%

***** build/ALPHA/tests/opt/long/fs/10.linux-boot/alpha/linux/tsunami-o3: CHANGED!

--- long/fs/10.linux-boot/alpha/linux/tsunami-o3-dual ---
*** gem5: OK
*** stat_diff: FAILURE: Statistics mismatch
Statistics mismatch
Maximum error magnitude: +654.222399%

                                  Reference  New Value   Abs Diff   Pct Chg
Key statistics:

  host_inst_rate                     237406     253837      16431    +6.92%
  host_mem_usage                     342872     515808     172936   +50.44%
  sim_insts                        56676315   56676490        175    +0.00%
  sim_ops                          56676315   56676490        175    +0.00%
  sim_ticks                      1907549438500 1907550384500     946000    +0.00%
  system.cpu0.commit.committedInsts   49598051   49598186        135    +0.00%
  system.cpu0.commit.committedOps   49598051   49598186        135    +0.00%
  system.cpu0.committedInsts       46714728   46714853        125    +0.00%
  system.cpu0.committedOps         46714728   46714853        125    +0.00%
  system.cpu0.ipc                  0.390977   0.390991   0.000014    +0.00%
  system.cpu0.ipc_total            0.390977   0.390991   0.000014    +0.00%
  system.cpu1.commit.committedInsts   10447204   10447254         50    +0.00%
  system.cpu1.commit.committedOps   10447204   10447254         50    +0.00%
  system.cpu1.committedInsts        9961587    9961637         50    +0.00%
  system.cpu1.committedOps          9961587    9961637         50    +0.00%
  system.cpu1.ipc                  0.567818   0.567762  -0.000056    -0.01%
  system.cpu1.ipc_total            0.567818   0.567762  -0.000056    -0.01%

***** build/ALPHA/tests/opt/long/fs/10.linux-boot/alpha/linux/tsunami-o3-dual: CHANGED!

***** build/ALPHA/tests/opt/long/fs/10.linux-boot/alpha/linux/tsunami-switcheroo-full: passed.

Jason Lowe-Power Nov. 23, 2016, 7:21 a.m. (Nov. 23, 2016, 7:21 a.m.)
```
Thanks!
```

Ship It!

Ship It!

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Changeset 11707:1d085f66c4ca

Screenshots

Files

Summary:

Description:

Changeset 11707:1d085f66c4ca

Status: Closed (submitted)