mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-23 13:02:52 +02:00
19453c8511
Summary:
Specifically, when we perform runtime loop unrolling of a loop that
contains a convergent op, we can only unroll k times, where k divides
the loop trip multiple.
Without this change, we'll happily unroll e.g. the following loop
for (int i = 0; i < N; ++i) {
if (i == 0) convergent_op();
foo();
}
into
int i = 0;
if (N % 2 == 1) {
convergent_op();
foo();
++i;
}
for (; i < N - 1; i += 2) {
if (i == 0) convergent_op();
foo();
foo();
}.
This is unsafe, because we've just added a control-flow dependency to
the convergent op in the prelude.
In general, runtime unrolling loops that contain convergent ops is safe
only if we don't have emit a prelude, which occurs when the unroll count
divides the trip multiple.
Reviewers: resistor
Subscribers: llvm-commits, mzolotukhin
Differential Revision: http://reviews.llvm.org/D17526
llvm-svn: 263509
|
||
|---|---|---|
| .. | ||
| AArch64 | ||
| AMDGPU | ||
| PowerPC | ||
| X86 | ||
| 2004-05-13-DontUnrollTooMuch.ll | ||
| 2005-03-06-BadLoopInfoUpdate.ll | ||
| 2006-08-24-MultiBlockLoop.ll | ||
| 2007-04-16-PhiUpdate.ll | ||
| 2007-05-05-UnrollMiscomp.ll | ||
| 2007-05-09-UnknownTripCount.ll | ||
| 2007-11-05-Crash.ll | ||
| 2011-08-08-PhiUpdate.ll | ||
| 2011-08-09-IVSimplify.ll | ||
| 2011-08-09-PhiUpdate.ll | ||
| 2011-10-01-NoopTrunc.ll | ||
| 2012-04-09-unroll-indirectbr.ll | ||
| basic.ll | ||
| convergent.ll | ||
| ephemeral.ll | ||
| full-unroll-bad-cost.ll | ||
| full-unroll-crashers.ll | ||
| full-unroll-heuristics-2.ll | ||
| full-unroll-heuristics-cmp.ll | ||
| full-unroll-heuristics-phi-prop.ll | ||
| full-unroll-heuristics.ll | ||
| high-cost-trip-count-computation.ll | ||
| ignore-annotation-intrinsic-cost.ll | ||
| loop-remarks.ll | ||
| nsw-tripcount.ll | ||
| partial-unroll-optsize.ll | ||
| pr10813.ll | ||
| pr11361.ll | ||
| pr14167.ll | ||
| pr18861.ll | ||
| rebuild_lcssa.ll | ||
| runtime-loop1.ll | ||
| runtime-loop2.ll | ||
| runtime-loop3.ll | ||
| runtime-loop4.ll | ||
| runtime-loop5.ll | ||
| runtime-loop.ll | ||
| scevunroll.ll | ||
| shifted-tripcount.ll | ||
| tripcount-overflow.ll | ||
| unloop.ll | ||
| unroll-cleanup.ll | ||
| unroll-pragmas-disabled.ll | ||
| unroll-pragmas.ll | ||
| update-loop-info-in-subloops.ll | ||