a single missing character. Somehow, this had gone untested. I've added
tests for returns-twice logic specifically with the always-inliner that
would have caught this, and fixed the bug.
Thanks to Matt for the careful review and spotting this!!! =D
llvm-svn: 153832
on a per-callsite walk of the called function's instructions, in
breadth-first order over the potentially reachable set of basic blocks.
This is a major shift in how inline cost analysis works to improve the
accuracy and rationality of inlining decisions. A brief outline of the
algorithm this moves to:
- Build a simplification mapping based on the callsite arguments to the
function arguments.
- Push the entry block onto a worklist of potentially-live basic blocks.
- Pop the first block off of the *front* of the worklist (for
breadth-first ordering) and walk its instructions using a custom
InstVisitor.
- For each instruction's operands, re-map them based on the
simplification mappings available for the given callsite.
- Compute any simplification possible of the instruction after
re-mapping, and store that back int othe simplification mapping.
- Compute any bonuses, costs, or other impacts of the instruction on the
cost metric.
- When the terminator is reached, replace any conditional value in the
terminator with any simplifications from the mapping we have, and add
any successors which are not proven to be dead from these
simplifications to the worklist.
- Pop the next block off of the front of the worklist, and repeat.
- As soon as the cost of inlining exceeds the threshold for the
callsite, stop analyzing the function in order to bound cost.
The primary goal of this algorithm is to perfectly handle dead code
paths. We do not want any code in trivially dead code paths to impact
inlining decisions. The previous metric was *extremely* flawed here, and
would always subtract the average cost of two successors of
a conditional branch when it was proven to become an unconditional
branch at the callsite. There was no handling of wildly different costs
between the two successors, which would cause inlining when the path
actually taken was too large, and no inlining when the path actually
taken was trivially simple. There was also no handling of the code
*path*, only the immediate successors. These problems vanish completely
now. See the added regression tests for the shiny new features -- we
skip recursive function calls, SROA-killing instructions, and high cost
complex CFG structures when dead at the callsite being analyzed.
Switching to this algorithm required refactoring the inline cost
interface to accept the actual threshold rather than simply returning
a single cost. The resulting interface is pretty bad, and I'm planning
to do lots of interface cleanup after this patch.
Several other refactorings fell out of this, but I've tried to minimize
them for this patch. =/ There is still more cleanup that can be done
here. Please point out anything that you see in review.
I've worked really hard to try to mirror at least the spirit of all of
the previous heuristics in the new model. It's not clear that they are
all correct any more, but I wanted to minimize the change in this single
patch, it's already a bit ridiculous. One heuristic that is *not* yet
mirrored is to allow inlining of functions with a dynamic alloca *if*
the caller has a dynamic alloca. I will add this back, but I think the
most reasonable way requires changes to the inliner itself rather than
just the cost metric, and so I've deferred this for a subsequent patch.
The test case is XFAIL-ed until then.
As mentioned in the review mail, this seems to make Clang run about 1%
to 2% faster in -O0, but makes its binary size grow by just under 4%.
I've looked into the 4% growth, and it can be fixed, but requires
changes to other parts of the inliner.
llvm-svn: 153812
blocks in the function cloner. This removes the last case of trivially
dead code that I've been seeing in the wild getting inlined, analyzed,
re-inlined, optimized, only to be deleted. Nukes a FIXME from the
cleanup tests.
llvm-svn: 153572
aggressively. There are lots of dire warnings about this being expensive
that seem to predate switching to the TrackingVH-based value remapper
that is automatically updated on RAUW. This makes it easy to not just
prune single-entry PHIs, but to fully simplify PHIs, and to recursively
simplify the newly inlined code to propagate PHINode simplifications.
This introduces a bit of a thorny problem though. We may end up
simplifying a branch condition to a constant when we fold PHINodes, and
we would like to nuke any dead blocks resulting from this so that time
isn't wasted continually analyzing them, but this isn't easy. Deleting
basic blocks *after* they are fully cloned and mapped into the new
function currently requires manually updating the value map. The last
piece of the simplification-during-inlining puzzle will require either
switching to WeakVH mappings or some other piece of refactoring. I've
left a FIXME in the testcase about this.
llvm-svn: 153410
to instead rely on much more generic and powerful instruction
simplification in the function cloner (and thus inliner).
This teaches the pruning function cloner to use instsimplify rather than
just the constant folder to fold values during cloning. This can
simplify a large number of things that constant folding alone cannot
begin to touch. For example, it will realize that 'or' and 'and'
instructions with certain constant operands actually become constants
regardless of what their other operand is. It also can thread back
through the caller to perform simplifications that are only possible by
looking up a few levels. In particular, GEPs and pointer testing tend to
fold much more heavily with this change.
This should (in some cases) have a positive impact on compile times with
optimizations on because the inliner itself will simply avoid cloning
a great deal of code. It already attempted to prune proven-dead code,
but now it will be use the stronger simplifications to prove more code
dead.
llvm-svn: 153403
regressed seriously here, we are no longer removing allocas during
inline cleanup. This appears to be because of lifetime markers "using"
them. =/ I'll look into this shortly.
llvm-svn: 153394
It was added in 2007 as the first cut at supporting no-inline
attributes, but we didn't have function attributes of any form at the
time. However, it was added without any mention in the LangRef or other
documentation.
Later on, in 2008, Devang added function notes for 'inline=never' and
then turned them into proper function attributes. From that point
onward, as far as I can tell, the world moved on, and no one has touched
'llvm.noinline' in any meaningful way since.
It's time has now come. We have had better mechanisms for doing this for
a long time, all the frontends I'm aware of use them, and this is just
holding back progress. Given that it was never a documented feature of
the IR, I've provided no auto-upgrade support. If people know of real,
in-the-wild bitcode that relies on this, yell at me and I'll add it, but
I *seriously* doubt anyone cares.
llvm-svn: 152904
correlated pairs of pointer arguments at the callsite. This is designed
to recognize the common C++ idiom of begin/end pointer pairs when the
end pointer is a constant offset from the begin pointer. With the
C-based idiom of a pointer and size, the inline cost saw the constant
size calculation, and this provides the same level of information for
begin/end pairs.
In order to propagate this information we have to search for candidate
operations on a pair of pointer function arguments (or derived from
them) which would be simplified if the pointers had a known constant
offset. Then the callsite analysis looks for such pointer pairs in the
argument list, and applies the appropriate bonus.
This helps LLVM detect that half of bounds-checked STL algorithms
(such as hash_combine_range, and some hybrid sort implementations)
disappear when inlined with a constant size input. However, it's not
a complete fix due the inaccuracy of our cost metric for constants in
general. I'm looking into that next.
Benchmarks showed no significant code size change, and very minor
performance changes. However, specific code such as hashing is showing
significantly cleaner inlining decisions.
llvm-svn: 152752
candidate set for subsequent inlining, try to simplify the arguments to
the inner call site now that inlining has been performed.
The goal here is to propagate and fold constants through deeply nested
call chains. Without doing this, we loose the inliner bonus that should
be applied because the arguments don't match the exact pattern the cost
estimator uses.
Reviewed on IRC by Benjamin Kramer.
llvm-svn: 152556
introduced. Specifically, there are cost reductions for all
constant-operand icmp instructions against an alloca, regardless of
whether the alloca will in fact be elligible for SROA. That means we
don't want to abort the icmp reduction computation when we abort the
SROA reduction computation. That in turn frees us from the need to keep
a separate worklist and defer the ICmp calculations.
Use this new-found freedom and some judicious function boundaries to
factor the innards of computing the cost factor of any given instruction
out of the loop over the instructions and into static helper functions.
This greatly simplifies the code, and hopefully makes it more clear what
is happening here.
Reviewed by Eric Christopher. There is some concern that we'd like to
ensure this doesn't get out of hand, and I plan to benchmark the effects
of this change over the next few days along with some further fixes to
the inline cost.
llvm-svn: 152368
savings from a pointer argument becoming an alloca. Sometimes callees will even
compare a pointer to null and then branch to an otherwise unreachable block!
Detect these cases and compute the number of saved instructions, instead of
bailing out and reporting no savings.
llvm-svn: 148941
can't handle. Also don't produce non-zero results for things which won't be
transformed by SROA at all just because we saw the loads/stores before we saw
the use of the address.
llvm-svn: 148536
for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the
target indep prefetch change.
As usual, updating the testsuite is a PITA.
llvm-svn: 133337
pad, separating the exception and selector calls from the new lpad. Teaching
it not to do that, or to properly adjust the CFG afterwards, is out of
scope because it would require the other edges to the landing pad to be split
as well (effectively). Instead, just recover from the most likely cases
during inlining. The best long-term solution is to change the exception
representation and commit to either requiring or not requiring the more
complex edge-splitting logic; this is just a shorter-term hack.
llvm-svn: 132799
transformed by the inliner into a branch to the enclosing landing pad
(when inlined through an invoke). If not so optimized, it is lowered
DWARF EH preparation into a call to _Unwind_Resume (or _Unwind_SjLj_Resume
as appropriate). Its chief advantage is that it takes both the
exception value and the selector value as arguments, meaning that there
is zero effort in recovering these; however, the frontend is required
to pass these down, which is not actually particularly difficult.
Also document the behavior of landing pads a bit better, and make it
clearer that it's okay that personality functions don't always land at
landing pads. This is just a fact of life. Don't write optimizations that
rely on pushing things over an unwind edge.
llvm-svn: 132253
- the selector for the landing pad must provide all available information
about the handlers, filters, and cleanups within that landing pad
- calls to _Unwind_Resume must be converted to branches to the enclosing
lpad so as to avoid re-entering the unwinder when the lpad claimed it
was going to handle the exception in some way
This is quite specific to libUnwind-based unwinding. In an effort to not
interfere too badly with other unwinders, and with existing hacks in frontends,
this only triggers on _Unwind_Resume (not _Unwind_Resume_or_Rethrow) and does
nothing with selectors if it cannot find a selector call for either lpad.
llvm-svn: 132200
argument. The generated alloca has to have at least the alignment of the
byval, if not, the client may be making assumptions that the new alloca won't
satisfy.
llvm-svn: 122234
does normal initialization and normal chaining. Change the default
AliasAnalysis implementation to NoAlias.
Update StandardCompileOpts.h and friends to explicitly request
BasicAliasAnalysis.
Update tests to explicitly request -basicaa.
llvm-svn: 116720
halting analysis, it is illegal to delete a call to a read-only function.
The correct solution is almost certainly to add a "must halt" attribute and
only allow deletions in its presence.
XFAIL the relevant testcase for now.
llvm-svn: 102831
that can have a big effect :). The first is to enable the
iterative SCC passmanager juice that kicks in when the
scc passmgr detects that a function pass has devirtualized
a call. In this case, it will rerun all the passes it
manages on the SCC, up to the iteration count limit (4). This
is useful because a function pass may devirualize a call, and
we want the inliner to inline it, or pruneeh to infer stuff
about it, etc.
The second patch is to add *all* call sites to the
DevirtualizedCalls list the inliner uses. This list is
about to get renamed, but the jist of this is that the
inliner now reconsiders *all* inlined call sites as candidates
for further inlining. The intuition is this that in cases
like this:
f() { g(1); } g(int x) { h(x); }
We analyze this bottom up, and may decide that it isn't
profitable to inline H into G. Next step, we decide that it is
profitable to inline G into F, and do so, which means that F
now calls H. Even though the call from G -> H may not have been
profitable to inline, the call from F -> H may be (in this case
because a constant allows folding etc).
In my spot checks, this doesn't have a big impact on code. For
example, the LLC output for 252.eon grew from 0.02% (from
317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612
to 1520964 bytes). 252.eon never iterated in the SCC Passmgr,
176.gcc iterated at most 1 time.
llvm-svn: 102823
that appear due to inlining a callee as candidates for
futher inlining, but a recent patch made it do this if
those call sites were indirect and became direct.
Unfortunately, in bizarre cases (see testcase) doing this
can cause us to infinitely inline mutually recursive
functions into callers not in the cycle. Fix this by
keeping track of the inline history from which callsite
inline candidates got inlined from.
This shouldn't affect any "real world" code, but is required
for a follow on patch that is coming up next.
llvm-svn: 102822
were still inlining self-recursive functions into other functions.
Inlining a recursive function into itself has the potential to
reduce recursion depth by a factor of 2, inlining a recursive
function into something else reduces recursion depth by exactly
1. Since inlining a recursive function into something else is a
weird form of loop peeling, turn this off.
The deleted testcase was added by Dale in r62107, since then
we're leaning towards not inlining recursive stuff ever. In any
case, if we like inlining recursive stuff, it should be done
within the recursive function itself to get the algorithm
recursion depth win.
llvm-svn: 102798
that appear in the SCC as a result of inlining as candidates
for inlining. Change this so that it *does* consider call
sites that change from being indirect to being direct as a
result of inlining. This allows it to completely
"devirtualize" the testcase.
llvm-svn: 102146
Fix RefreshCallGraph to use CGN->replaceCallEdge instead of hand
rolling its own loop. replaceCallEdge properly maintains the
reference counts of the nodes, fixing a crash exposed by the
iterative callgraph stuff.
llvm-svn: 102120
we have RefreshCallGraph detect when a function pass devirtualizes
a call, and have CGSCCPassMgr iterate (up to a count) when this
happens. This allows (in the example) GVN to devirtualize the
call in foo, then the inliner to inline it away.
This is not currently enabled because I haven't done any analysis
on the (potentially substantial) code size or performance impact of
doing this, and guess what, it exposes callgraph updating bugs in
various passes. This is progress though, and you can play with it
by passing -max-cg-scc-iterations=5 to opt.
llvm-svn: 101973
recursive callsites, inlining can reduce the number of calls by
exponential factors, as it does in
MultiSource/Benchmarks/Olden/treeadd. More involved heuristics
will be needed.
llvm-svn: 101969
for the noinline attribute, and make the inliner refuse to
inline a call site when the call site is marked noinline even
if the callee isn't. This fixes PR6682.
llvm-svn: 99341
running IPSCCP early, and we run functionattrs interlaced with the inliner,
we often (particularly for small or noop functions) completely propagate
all of the information about a call to its call site in IPSSCP (making a call
dead) and functionattrs is smart enough to realize that the function is
readonly (because it is interlaced with inliner).
To improve compile time and make the inliner threshold more accurate, realize
that we don't have to inline dead readonly function calls. Instead, just
delete the call. This happens all the time for C++ codes, here are some
counters from opt/llvm-ld counting the number of times calls were deleted vs
inlined on various apps:
Tramp3d opt:
5033 inline - Number of call sites deleted, not inlined
24596 inline - Number of functions inlined
llvm-ld:
667 inline - Number of functions deleted because all callers found
699 inline - Number of functions inlined
483.xalancbmk opt:
8096 inline - Number of call sites deleted, not inlined
62528 inline - Number of functions inlined
llvm-ld:
217 inline - Number of allocas merged together
2158 inline - Number of functions inlined
471.omnetpp:
331 inline - Number of call sites deleted, not inlined
8981 inline - Number of functions inlined
llvm-ld:
171 inline - Number of functions deleted because all callers found
629 inline - Number of functions inlined
Deleting a call is much faster than inlining it, and is insensitive to the
size of the callee. :)
llvm-svn: 86975
with multiple return values it inserts a PHI to merge them all together.
However, if the return values are all the same, it ends up with a pointless
PHI and this pointless PHI happens to really block SRoA from happening in
at least a silly C++ example written by Doug, but probably others. This
fixes rdar://7339069.
llvm-svn: 85206
and that will make Caller too big to inline, see if it
might be better to inline Caller into its callers instead.
This situation is described in PR 2973, although I haven't
tried the specific case in SPASS.
llvm-svn: 83602
input filename so that opt doesn't print the input filename in the
output so that grep lines in the tests don't unintentionally match
strings in the input filename.
llvm-svn: 81537
indirect function pointer, inline it, then go to delete the body.
The problem is that the callgraph had other references to the function,
though the inliner had no way to know it, so we got a dangling pointer
and an invalid iterator out of the deal.
The fix to this is pretty simple: stop the inliner from deleting the
function by knowing that there are references to it. Do this by making
CallGraphNodes contain a refcount. This requires moving deletion of
available_externally functions to the module-level cleanup sweep where
it belongs.
llvm-svn: 80533
calls into a function and if the calls bring in arrays, try to merge
them together to reduce stack size. For example, in the testcase
we'd previously end up with 4 allocas, now we end up with 2 allocas.
As described in the comments, this is not really the ideal solution
to this problem, but it is surprisingly effective. For example, on
176.gcc, we end up eliminating 67 arrays at "gccas" time and another
24 at "llvm-ld" time.
One piece of concern that I didn't look into: at -O0 -g with
forced inlining this will almost certainly result in worse debug
info. I think this is acceptable though given that this is a case
of "debugging optimized code", and we don't want debug info to
prevent the optimizer from doing things anyway.
llvm-svn: 80215
llvm.dbg.region.end instrinsic. This nested llvm.dbg.func.start/llvm.dbg.region.end pair now enables DW_TAG_inlined_subroutine support in code generator.
llvm-svn: 69118
to support C99 inline, GNU extern inline, etc. Related bugzilla's
include PR3517, PR3100, & PR2933. Nothing uses this yet, but it
appears to work.
llvm-svn: 68940
functions that don't already have a (dynamic) alloca.
Dynamic allocas cause inefficient codegen and we shouldn't
propagate this (behavior follows gcc). Two existing tests
assumed such inlining would be done; they are hacked by
adding an alloca in the caller, preserving the point of
the tests.
llvm-svn: 61946
Chandler Carruth