derek/gem5
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

cpu: disallow speculative update of branch predictor tables (o3)

Browse Source 
The Minor and o3 cpu models share the branch prediction
code. Minor relies on the BPredUnit::squash() function
to update the branch predictor tables on a branch mispre-
diction. This is fine because Minor executes in-order, so
the update is on the correct path. However, this causes the
branch predictor to be updated on out-of-order branch
mispredictions when using the o3 model, which should not
be the case.

This patch guards against speculative update of the branch
prediction tables. On a branch misprediction, BPredUnit::squash()
calls BpredUnit::update(..., squashed = true). The underlying
branch predictor tests against the value of squashed. If it is
true, it restores any speculatively updated internal state
it might have (e.g., global/local branch history), then returns.
If false, it updates its prediction tables. Previously, exist-
ing predictors did not test against the "squashed" parameter.

To accomodate for this change, the Minor model must now call
BPredUnit::squash() then BPredUnit::update(..., squashed = false)
on branch mispredictions. Before, calling BpredUnit::squash()
performed the prediction tables update.

The effect is a slight MPKI improvement when using the o3
model. A further patch should perform the same modifications
for the indirect target predictor and BTB (less critical).

Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
This commit is contained in:
Arthur Perais
2016-12-21 15:07:16 -06:00
parent 34065f8d5f
commit 497cc2d373
9 changed files with 142 additions and 170 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
src/cpu/pred/bpred_unit.cc
View File

@@ -330,13 +330,9 @@ BPredUnit::update(const InstSeqNum &done_sn, ThreadID tid)
while (!predHist[tid].empty() &&
predHist[tid].back().seqNum <= done_sn) {
// Update the branch predictor with the correct results.
if (!predHist[tid].back().wasSquashed) {
update(tid, predHist[tid].back().pc,
predHist[tid].back().predTaken,
predHist[tid].back().bpHistory, false);
} else {
retireSquashed(tid, predHist[tid].back().bpHistory);
}
update(tid, predHist[tid].back().pc,
predHist[tid].back().predTaken,
predHist[tid].back().bpHistory, false);
predHist[tid].pop_back();
}
@@ -430,12 +426,23 @@ BPredUnit::squash(const InstSeqNum &squashed_sn,
tid, hist_it->seqNum);
}
// Have to get GHR here because the update deletes bpHistory
// Get the underlying Global History Register
unsigned ghr = getGHR(tid, hist_it->bpHistory);
// There are separate functions for in-order and out-of-order
// branch prediction, but not for update. Therefore, this
// call should take into account that the mispredicted branch may
// be on the wrong path (i.e., OoO execution), and that the counter
// counter table(s) should not be updated. Thus, this call should
// restore the state of the underlying predictor, for instance the
// local/global histories. The counter tables will be updated when
// the branch actually commits.
// Remember the correct direction for the update at commit.
pred_hist.front().predTaken = actually_taken;
update(tid, (*hist_it).pc, actually_taken,
pred_hist.front().bpHistory, true);
hist_it->wasSquashed = true;
if (actually_taken) {
if (hist_it->wasReturn && !hist_it->usedRAS) {