Currently the pass updates branch weights in the IR if the function has
any PGO info (entry frequency is set). However we could still have
regions of the CFG that does not have branch weights collected (e.g. a
cold region). In this case we'd use static estimates. Since static
estimates for branches are determined independently, they are
inconsistent. Updating them can "randomly" inflate block frequencies.
I've run into this in a completely cold loop of h264ref from
SPEC. -Rpass-with-hotness showed the loop to be completely cold during
inlining (before JT) but completely hot during vectorization (after JT).
The new testcase demonstrate the problem. We check array elements
against 1, 2 and 3 in a loop. The check against 3 is the loop-exiting
check. The block names should be self-explanatory.
In this example, jump threading incorrectly updates the weight of the
loop-exiting branch to 0, drastically inflating the frequency of the
loop (in the range of billions).
There is no run-time profile info for edges inside the loop, so branch
probabilities are estimated. These are the resulting branch and block
frequencies for the loop body:
check_1 (16)
(8) / |
eq_1 | (8)
\ |
check_2 (16)
(8) / |
eq_2 | (8)
\ |
check_3 (16)
(1) / |
(loop exit) | (15)
|
(back edge)
First we thread eq_1 -> check_2 to check_3. Frequencies are updated to
remove the frequency of eq_1 from check_2 and then from the false edge
leaving check_2. Changed frequencies are highlighted with * *:
check_1 (16)
(8) / |
eq_1~ | (8)
/ |
/ check_2 (*8*)
/ (8) / |
\ eq_2 | (*0*)
\ \ |
` --- check_3 (16)
(1) / |
(loop exit) | (15)
|
(back edge)
Next we thread eq_1 -> check_3 and eq_2 -> check_3 to check_1 as new
back edges. Frequencies are updated to remove the frequency of eq_1 and
eq_3 from check_3 and then the false edge leaving check_3 (changed
frequencies are highlighted with * *):
check_1 (16)
(8) / |
eq_1~ | (8)
/ |
/ check_2 (*8*)
/ (8) / |
/-- eq_2~ | (*0*)
(back edge) |
check_3 (*0*)
(*0*) / |
(loop exit) | (*0*)
|
(back edge)
As a result, the loop exit edge ends up with 0 frequency which in turn makes
the loop header to have maximum frequency.
There are a few potential problems here:
1. The profile data seems odd. There is a single profile sample of the
loop being entered. On the other hand, there are no weights inside the
loop.
2. Based on static estimation we shouldn't set edges to "extreme"
values, i.e. extremely likely or unlikely.
3. We shouldn't create profile metadata that is calculated from static
estimation. I am not sure what policy is but it seems to make sense to
treat profile metadata as something that is known to originate from
profiling. Estimated probabilities should only be reflected in BPI/BFI.
Any one of these would probably fix the immediate problem. I went for 3
because I think it's a good policy to have and added a FIXME about 2.
Differential Revision: https://reviews.llvm.org/D24118
llvm-svn: 280713
r280425 | dehao | 2016-09-01 16:15:50 -0700 (Thu, 01 Sep 2016) | 9 lines
Refactor LICM pass in preparation for LoopSink pass.
Summary: LoopSink pass uses some common function in LICM. This patch refactor the LICM code to make it usable by LoopSink pass (https://reviews.llvm.org/D22778).
r280429 | dehao | 2016-09-01 16:31:25 -0700 (Thu, 01 Sep 2016) | 9 lines
Refactor LICM to expose canSinkOrHoistInst to LoopSink pass.
Summary: LoopSink pass shares the same canSinkOrHoistInst functionality with LICM pass. This patch exposes this function in preparation of https://reviews.llvm.org/D22778
llvm-svn: 280453
Summary: LoopSink pass shares the same canSinkOrHoistInst functionality with LICM pass. This patch exposes this function in preparation of https://reviews.llvm.org/D22778
Reviewers: chandlerc, davidxl, danielcdh
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24171
llvm-svn: 280429
Summary: This is in preparation for LoopSink pass which calls replaceDominatedUsesWith to update after sinking.
Reviewers: chandlerc, davidxl, danielcdh
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24170
llvm-svn: 280427
Summary: LoopSink pass uses some common function in LICM. This patch refactor the LICM code to make it usable by LoopSink pass (https://reviews.llvm.org/D22778).
Reviewers: chandlerc, davidxl, danielcdh
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24168
llvm-svn: 280425
Previous change broke the C API for creating an EarlyCSE pass w/
MemorySSA by adding a bool parameter to control whether MemorySSA was
used or not. This broke the OCaml bindings. Instead, change the old C
API entry point back and add a new one to request an EarlyCSE pass with
MemorySSA.
llvm-svn: 280379
Summary:
Current implementation of LI verifier isn't ideal and fails to detect
some cases when LI is incorrect. For instance, it checks that all
recorded loops are in a correct form, but it has no way to check if
there are no more other (unrecorded in LI) loops in the function. This
patch adds a way to detect such bugs.
Reviewers: chandlerc, sanjoy, hfinkel
Subscribers: llvm-commits, silvas, mzolotukhin
Differential Revision: https://reviews.llvm.org/D23437
llvm-svn: 280280
Summary:
Use MemorySSA, if requested, to do less conservative memory dependency
checking.
This change doesn't enable the MemorySSA enhanced EarlyCSE in the
default pipelines, so should be NFC.
Reviewers: dberlin, sanjoy, reames, majnemer
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19821
llvm-svn: 280279
This is a first step towards supporting deopt value lowering and reporting entirely with the register allocator. I hope to build on this in the near future to support live-on-return semantics, but I have a use case which allows me to test and investigate code quality with just the live-in semantics so I've chosen to start there. For those curious, my use cases is our implementation of the "__llvm_deoptimize" function we bind to @llvm.deoptimize. I'm choosing not to hard code that fact in the patch and instead make it configurable via function attributes.
The basic approach here is modelled on what is done for the "Live In" values on stackmaps and patchpoints. (A secondary goal here is to remove one of the last barriers to merging the pseudo instructions.) We start by adding the operands directly to the STATEPOINT SDNode. Once we've lowered to MI, we extend the remat logic used by the register allocator to fold virtual register uses into StackMap::Indirect entries as needed. This does rely on the fact that the register allocator rematerializes. If it didn't along some code path, we could end up with more vregs than physical registers and fail to allocate.
Today, we *only* fold in the register allocator. This can create some weird effects when combined with arguments passed on the stack because we don't fold them appropriately. I have an idea how to fix that, but it needs this patch in place to work on that effectively. (There's some weird interaction with the scheduler as well, more investigation needed.)
My near term plan is to land this patch off-by-default, experiment in my local tree to identify any correctness issues and then start fixing codegen problems one by one as I find them. Once I have the live-in lowering fully working (both correctness and code quality), I'm hoping to move on to the live-on-return semantics. Note: I don't have any *known* miscompiles with this patch enabled, but I'm pretty sure I'll find at least a couple. Thus, the "experimental" tag and the fact it's off by default.
Differential Revision: https://reviews.llvm.org/D24000
llvm-svn: 280250
Reverse iterators to doubly-linked lists can be simpler (and cheaper)
than std::reverse_iterator. Make it so.
In particular, change ilist<T>::reverse_iterator so that it is *never*
invalidated unless the node it references is deleted. This matches the
guarantees of ilist<T>::iterator.
(Note: MachineBasicBlock::iterator is *not* an ilist iterator, but a
MachineInstrBundleIterator<MachineInstr>. This commit does not change
MachineBasicBlock::reverse_iterator, but it does update
MachineBasicBlock::reverse_instr_iterator. See note at end of commit
message for details on bundle iterators.)
Given the list (with the Sentinel showing twice for simplicity):
[Sentinel] <-> A <-> B <-> [Sentinel]
the following is now true:
1. begin() represents A.
2. begin() holds the pointer for A.
3. end() represents [Sentinel].
4. end() holds the poitner for [Sentinel].
5. rbegin() represents B.
6. rbegin() holds the pointer for B.
7. rend() represents [Sentinel].
8. rend() holds the pointer for [Sentinel].
The changes are #6 and #8. Here are some properties from the old
scheme (which used std::reverse_iterator):
- rbegin() held the pointer for [Sentinel] and rend() held the pointer
for A;
- operator*() cost two dereferences instead of one;
- converting from a valid iterator to its valid reverse_iterator
involved a confusing increment; and
- "RI++->erase()" left RI invalid. The unintuitive replacement was
"RI->erase(), RE = end()".
With vector-like data structures these properties are hard to avoid
(since past-the-beginning is not a valid pointer), and don't impose a
real cost (since there's still only one dereference, and all iterators
are invalidated on erase). But with lists, this was a poor design.
Specifically, the following code (which obviously works with normal
iterators) now works with ilist::reverse_iterator as well:
for (auto RI = L.rbegin(), RE = L.rend(); RI != RE;)
fooThatMightRemoveArgFromList(*RI++);
Converting between iterator and reverse_iterator for the same node uses
the getReverse() function.
reverse_iterator iterator::getReverse();
iterator reverse_iterator::getReverse();
Why doesn't iterator <=> reverse_iterator conversion use constructors?
In order to catch and update old code, reverse_iterator does not even
have an explicit conversion from iterator. It wouldn't be safe because
there would be no reasonable way to catch all the bugs from the changed
semantic (see the changes at call sites that are part of this patch).
Old code used this API:
std::reverse_iterator::reverse_iterator(iterator);
iterator std::reverse_iterator::base();
Here's how to update from old code to new (that incorporates the
semantic change), assuming I is an ilist<>::iterator and RI is an
ilist<>::reverse_iterator:
[Old] ==> [New]
reverse_iterator(I) (--I).getReverse()
reverse_iterator(I) ++I.getReverse()
--reverse_iterator(I) I.getReverse()
reverse_iterator(++I) I.getReverse()
RI.base() (--RI).getReverse()
RI.base() ++RI.getReverse()
--RI.base() RI.getReverse()
(++RI).base() RI.getReverse()
delete &*RI, RE = end() delete &*RI++
RI->erase(), RE = end() RI++->erase()
=======================================
Note: bundle iterators are out of scope
=======================================
MachineBasicBlock::iterator, also known as
MachineInstrBundleIterator<MachineInstr>, is a wrapper to represent
MachineInstr bundles. The idea is that each operator++ takes you to the
beginning of the next bundle. Implementing a sane reverse iterator for
this is harder than ilist. Here are the options:
- Use std::reverse_iterator<MBB::i>. Store a handle to the beginning of
the next bundle. A call to operator*() runs a loop (usually
operator--() will be called 1 time, for unbundled instructions).
Increment/decrement just works. This is the status quo.
- Store a handle to the final node in the bundle. A call to operator*()
still runs a loop, but it iterates one time fewer (usually
operator--() will be called 0 times, for unbundled instructions).
Increment/decrement just works.
- Make the ilist_sentinel<MachineInstr> *always* store that it's the
sentinel (instead of just in asserts mode). Then the bundle iterator
can sniff the sentinel bit in operator++().
I initially tried implementing the end() option as part of this commit,
but updating iterator/reverse_iterator conversion call sites was
error-prone. I have a WIP series of patches that implements the final
option.
llvm-svn: 280032
Summary:
While walking the use chain for identifying rematerializable values in RS4GC,
add the case where the current value and base value are the same PHI nodes.
This will aid rematerialization of geps and casts instead of relocating.
Reviewers: sanjoy, reames, igor
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23920
llvm-svn: 279975
Without invalidating the entries in the MD cache we would try to access instructions
that were removed in previous iterations of hoisting.
Differential Revision: https://reviews.llvm.org/D23927
llvm-svn: 279907
Summary: Dead store elimination gets very expensive when large numbers of instructions need to be analyzed. This patch limits the number of instructions analyzed per store to the value of the memdep-block-scan-limit parameter (which defaults to 100). This resulted in no observed difference in performance of the generated code, and no change in the statistics for the dead store elimination pass, but improved compilation time on some files by more than an order of magnitude.
Reviewers: dexonsmith, bruno, george.burgess.iv, dberlin, reames, davidxl
Subscribers: davide, chandlerc, dberlin, davidxl, eraman, tejohnson, mbodart, llvm-commits
Differential Revision: https://reviews.llvm.org/D15537
llvm-svn: 279833
We can't mark ORE (a function pass) preserved as required by the loop
passes because that is how we ensure that the required passes like
LazyBFI are all available any time ORE is used. See the new comments in
the patch.
Instead we use it directly just like the inliner does in D22694.
As expected there is some additional overhead after removing the caching
provided by analysis passes. The worst case, I measured was
LNT/CINT2006_ref/401.bzip2 which regresses by 12%. As before, this only
affects -Rpass-with-hotness and not default compilation.
llvm-svn: 279829
Summary:
This fixes pr29105. The reason is that lifetime marks creates new
aliasing pointers the original ones, but before this patch aliases
were not checked in performMemCpyToMemSetOptzn.
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23846
llvm-svn: 279769
It is invalid to hoist stores or loads if they are not executed on all paths
from the hoisting point to the exit of the function. In the testcase, there are
paths in the loop that do not execute the stores or the loads, and so hoisting
them within the loop is unsafe.
The problem is that the current implementation of hoistingFromAllPaths is
incomplete: it walks all blocks dominated by the hoisting point, and does not
return false when the loop contains a path on which the hoisted ld/st is
not executed.
Differential Revision: https://reviews.llvm.org/D23843
llvm-svn: 279732
I'm not sure if the `!isa<CallInst>(Inst) &&
!isa<TerminatorInst>(Inst))` bit is correct either, but this fixes the
case we know is broken.
llvm-svn: 279647
Summary:
This is part of a serious of patches to evolve ADCE.cpp to support
removing of unnecessary control flow.
This patch adds the ability to compute control dependences using
the iterated dominance frontier. We extend the liveness propagation
to alternate between data and control dependences until convergences.
Modify the pass manager intergation to compute the post-dominator tree
needed for iterator dominance frontier.
We still force all terminators live for now until we add code to
handlinge removing control flow in a later patch.
No changes to effective behavior with this patch
Previous patches:
D23225 [ADCE] Modify data structures to support removing control flow
D23065 [ADCE] Refactor anticipating new functionality (NFC)
D23102 [ADCE] Refactoring for new functionality (NFC)
Reviewers: nadav, majnemer, mehdi_amini
Subscribers: twoh, freik, llvm-commits
Differential Revision: https://reviews.llvm.org/D23559
llvm-svn: 279594
Summary:
In clang commit r268509 we started to invoke loop-unroll pass from the
driver even under -Os. However, we happen to not initialize optsize
thresholds properly, which si fixed with this change.
r268509 led to some big compile time regressions, because we started to
unroll some loops that we didn't unroll before. With this change I hope
to recover most of the regressions. We still are slightly slower than
before, because we do some checks here and there in loop-unrolling
before we bail out, but at least the slowdown is not that huge now.
Reviewers: hfinkel, chandlerc
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D23388
llvm-svn: 279585
Summary: GVNHoist: Use the pass version of MemorySSA and preserve it.
Reviewers: sebpop, george.burgess.iv
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23782
llvm-svn: 279504
This change needs to be reverted in order to revert -r278267 which cause performance regression on MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt from LNT and some other bechmarks.
See comments on https://reviews.llvm.org/D18777 for details.
llvm-svn: 279432
Currently nodes_iterator may dereference to a NodeType* or a NodeType&. Make them all dereference to NodeType*, which is NodeRef later.
Differential Revision: https://reviews.llvm.org/D23704
Differential Revision: https://reviews.llvm.org/D23705
llvm-svn: 279326
It causes a regression on our internal benchmark. Introduce cvp-dont-process flag and set it off by default while investigating the regression.
llvm-svn: 279082
Refactored so that a LSRUse owns its fixups, as oppsed to letting the
LSRInstance own them. This makes it easier to rate formulas for
LSRUses, since the fixups are available directly. The Offsets vector
has been removed since it was no longer necessary.
New target hook isFoldableMemAccessOffset(), which is used during formula
rating.
For SystemZ, this is useful to express that loads and stores with
float or vector types with a big/negative offset should be avoided in
loops. Without this, LSR will generate a lot of negative offsets that
would require extra instructions for loading the address.
Updated tests:
test/CodeGen/SystemZ/loop-01.ll
Reviewed by: Quentin Colombet and Ulrich Weigand.
https://reviews.llvm.org/D19152
llvm-svn: 278927
This is a mechanical change of comments in switches like fallthrough,
fall-through, or fall-thru to use the LLVM_FALLTHROUGH macro instead.
llvm-svn: 278902
IndVarSimplify::sinkUnusedInvariants calls
BasicBlock::getFirstInsertionPt on the ExitBlock and moves instructions
before it. This can return end(), so it's not safe to dereference. Add
an iterator-based overload to Instruction::moveBefore to avoid the UB.
llvm-svn: 278886
BasicBlock::Create isn't designed to take iterators (which might be
end()), but pointers (which might be nullptr). Fix the UB that was
converting end() to a BasicBlock* by calling BasicBlock::getNextNode()
in the first place.
llvm-svn: 278883
Summary:
This is part of a serious of patches to evolve ADCE.cpp to support
removing of unnecessary control flow.
This patch changes the data structures to hold liveness information to
support the additional information we will eventually need. In
particular we now have a notion of basic blocks being live because
they contain a live operations. This will eventually feed into control
dependence analysis of which branches are live. We cater to getting
from instructions to associated block information and from blocks to
information about their terminators.
This patch also changes the structure of the main loop of the
algorithm so that it alternates propagating liveness between
instructions and usign control dependence information to mark branches
live.
We force all terminators live for now until we add code to handlinge
removing control flow in a later patch.
No changes to effective behavior with this patch
Previous patches:
D23065 [ADCE] Refactor anticipating new functionality (NFC)
D23102 [ADCE] Refactoring for new functionality (NFC)
Reviewers: nadav, majnemer, mehdi_amini
Subscribers: freik, twoh, llvm-commits
Differential Revision: https://reviews.llvm.org/D23225
llvm-svn: 278807
If a loop is not rotated (for example when optimizing for size), the latch is not the backedge. If we promote an expression to post-inc form, we not only increase register pressure and add a COPY for that IV expression but for all IVs!
Motivating testcase:
void f(float *a, float *b, float *c, int n) {
while (n-- > 0)
*c++ = *a++ + *b++;
}
It's imperative that the pointer increments be located in the latch block and not the header block; if not, we cannot use post-increment loads and stores and we have to keep both the post-inc and pre-inc values around until the end of the latch which bloats register usage.
llvm-svn: 278658