This is espcially important because the new SROA pass goes to great
lengths to provide helpful names for debugging, and as a consequence
they can become very slow to render.
Good for between 5% and 15% of the SROA runtime on some slow test cases
such as the one in PR15412.
llvm-svn: 177495
We don't want to write out >1000 files at the same time. That could make things
prohibitively expensive. Instead, register the "writeout" function so that it's
emitted serially.
<rdar://problem/12439551>
llvm-svn: 177437
- it is trivially known to be used inside the loop in a way that can not be optimized away
- there is no use outside of the loop which can take advantage of the computation hoisting
llvm-svn: 177432
For each compile unit, we want to register a function that will flush that
compile unit. Otherwise, __gcov_flush() would only flush the counters within the
current compile unit, and not any outside of it.
PR15191 & <rdar://problem/13167507>
llvm-svn: 177340
Rules include:
1)1 x*y +/- x*z => x*(y +/- z)
(the order of operands dosen't matter)
2) y/x +/- z/x => (y +/- z)/x
The transformation is disabled if the new add/sub expr "y +/- z" is a
denormal/naz/inifinity.
rdar://12911472
llvm-svn: 177088
The fundamental problem is that SROA didn't allow for overly wide loads
where the bits past the end of the alloca were masked away and the load
was sufficiently aligned to ensure there is no risk of page fault, or
other trapping behavior. With such widened loads, SROA would delete the
load entirely rather than clamping it to the size of the alloca in order
to allow mem2reg to fire. This was exposed by a test case that neatly
arranged for GVN to run first, widening certain loads, followed by an
inline step, and then SROA which miscompiles the code. However, I see no
reason why this hasn't been plaguing us in other contexts. It seems
deeply broken.
Diagnosing all of the above took all of 10 minutes of debugging. The
really annoying aspect is that fixing this completely breaks the pass.
;] There was an implicit reliance on the fact that no loads or stores
extended past the alloca once we decided to rewrite them in the final
stage of SROA. This was used to encode information about whether the
loads and stores had been split across multiple partitions of the
original alloca. That required threading explicit tracking of whether
a *use* of a partition is split across multiple partitions.
Once that was done, another problem arose: we allowed splitting of
integer loads and stores iff they were loads and stores to the entire
alloca. This is a really arbitrary limitation, and splitting at least
some integer loads and stores is crucial to maximize promotion
opportunities. My first attempt was to start removing the restriction
entirely, but currently that does Very Bad Things by causing *many*
common alloca patterns to be fully decomposed into i8 operations and
lots of or-ing together to produce larger integers on demand. The code
bloat is terrifying. That is still the right end-goal, but substantial
work must be done to either merge partitions or ensure that small i8
values are eagerly merged in some other pass. Sadly, figuring all this
out took essentially all the time and effort here.
So the end result is that we allow splitting only when the load or store
at least covers the alloca. That ensures widened loads and stores don't
hurt SROA, and that we don't rampantly decompose operations more than we
have previously.
All of this was already fairly well tested, and so I've just updated the
tests to cover the wide load behavior. I can add a test that crafts the
pass ordering magic which caused the original PR, but that seems really
brittle and to provide little benefit. The fundamental problem is that
widened loads should Just Work.
llvm-svn: 177055
constructs default arguments. It can now take default arguments from
cl::opt'ions. Add a new -default-gcov-version=... option, and actually test it!
Sink the reverse-order of the version into GCOVProfiling, hiding it from our
users.
llvm-svn: 177002
emitProfileNotes(), similar to emitProfileArcs(). Also update its comment.
Also add a comment on Version[4] (there will be another comment in clang later),
and compress lines that exceeded 80 columns.
llvm-svn: 176994
Nadav reported a performance regression due to the work I did to
merge the library call simplifier into instcombine [1]. The issue
is that a new LibCallSimplifier object is being created whenever
InstCombiner::runOnFunction is called. Every time a LibCallSimplifier
object is used to optimize a call it creates a hash table to map from
a function name to an object that optimizes functions of that name.
For short-lived LibCallSimplifier instances this is quite inefficient.
Especially for cases where no calls are actually simplified.
This patch fixes the issue by dropping the hash table and implementing
an explicit lookup function to correlate the function name to the object
that optimizes functions of that name. This avoids the cost of always
building and destroying the hash table in cases where the LibCallSimplifier
object is short-lived and avoids the cost of building the table when no
simplifications are actually preformed.
On a benchmark containing 100,000 calls where none of them are simplified
I noticed a 30% speedup. On a benchmark containing 100,000 calls where
all of them are simplified I noticed an 8% speedup.
[1] http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20130304/167639.html
llvm-svn: 176840
We want vectorization to happen at -g. Ignore calls to the dbg.value intrinsic
and don't transfer them to the vectorized code.
radar://13378964
llvm-svn: 176768
it. Fortunately, versions of gcov that predate the extra checksum also ignore
any extra data, so this isn't a problem. There will be a matching commit in
compiler-rt.
llvm-svn: 176745
into the actual gcov file.
Instead of using the bottom 4 bytes as the function identifier, use a counter.
This makes the identifier numbers stable across multiple runs.
llvm-svn: 176616
Fixes rdar:13349374.
Volatile loads and stores need to be preserved even if the language
standard says they are undefined. "volatile" in this context means "get
out of the way compiler, let my platform handle it".
Additionally, this is the only way I know of with llvm to write to the
first page (when hardware allows) without dropping to assembly.
llvm-svn: 176599
When considering folding a bitcast of an alloca into the alloca itself,
make sure we don't shrink the amount of memory being allocated, or
things rapidly go sideways.
rdar://13324424
llvm-svn: 176547
* Only apply divide bypass optimization when not optimizing for size.
* Fixed bug caused by constant for 0 value of type Int32,
used dividend type to generate the constant instead.
* For atom x86-64 apply the divide bypass to use 16-bit divides instead of
64-bit divides when operand values are small enough.
* Added lit tests for 64-bit divide bypass.
Patch by Tyler Nowicki!
llvm-svn: 176442
The LoopVectorizer often runs multiple times on the same function due to inlining.
When this happens the loop vectorizer often vectorizes the same loops multiple times, increasing code size and adding unneeded branches.
With this patch, the vectorizer during vectorization puts metadata on scalar loops and marks them as 'already vectorized' so that it knows to ignore them when it sees them a second time.
PR14448.
llvm-svn: 176399
The instcombine recognized pattern looks like:
a = b * c
d = a +/- Cst
or
a = b * c
d = Cst +/- a
When creating the new operands for fadd or fsub instruction following the related fmul, the first operand was created with the second original operand (M0 was created with C1) and the second with the first (M1 with Opnd0).
The fix consists in creating the new operands with the appropriate original operand, i.e., M0 with Opnd0 and M1 with C1.
llvm-svn: 176300
Shadow checks are disabled and memory loads always produce fully initialized
values in functions that don't have a sanitize_memory attribute. Value and
argument shadow is propagated as usual.
This change also updates blacklist behaviour to match the above.
llvm-svn: 176247
This properly asks TargetLibraryInfo if a call is available and if it is, it
can be translated into the corresponding LLVM builtin. We don't vectorize sqrt()
yet because I'm not sure about the semantics for negative numbers. The other
intrinsic should be exact equivalents to the libm functions.
Differential Revision: http://llvm-reviews.chandlerc.com/D465
llvm-svn: 176188
enhancement done the trivial way; by extending inputs and truncating outputs
which is addequate for targets with little or no support for integer arithmetic
on integer types less than 32 bits.
llvm-svn: 176139
These are two related changes (one in llvm, one in clang).
LLVM:
- rename address_safety => sanitize_address (the enum value is the same, so we preserve binary compatibility with old bitcode)
- rename thread_safety => sanitize_thread
- rename no_uninitialized_checks -> sanitize_memory
CLANG:
- add __attribute__((no_sanitize_address)) as a synonym for __attribute__((no_address_safety_analysis))
- add __attribute__((no_sanitize_thread))
- add __attribute__((no_sanitize_memory))
for S in address thread memory
If -fsanitize=S is present and __attribute__((no_sanitize_S)) is not
set llvm attribute sanitize_S
llvm-svn: 176075
This is a common pattern with dyn_cast and similar constructs, when the
PHI no longer depends on the select it can often be turned into a simpler
construct or even get hoisted out of the loop.
PR15340.
llvm-svn: 175995
The 'nobuiltin' attribute is applied to call sites to indicate that LLVM should
not treat the callee function as a built-in function. I.e., it shouldn't try to
replace that function with different code.
llvm-svn: 175835
Storing the load/store instructions with the values
and inspect them using Alias Analysis to make sure
they don't alias, since the GEP pointer operand doesn't
take the offset into account.
Trying hard to not add any extra cost to loads and stores
that don't overlap on global values, AA is *only* calculated
if all of the previous attempts failed.
Using biggest vector register size as the stride for the
vectorization access, as we're being conservative and
the cost model (which calculates the real vectorization
factor) is only run after the legalization phase.
We might re-think this relationship in the future, but
for now, I'd rather be safe than sorry.
llvm-svn: 175818
This patch makes asan instrument memory accesses with unusual sizes (e.g. 5 bytes or 10 bytes), e.g. long double or
packed structures.
Instrumentation is done with two 1-byte checks
(first and last bytes) and if the error is found
__asan_report_load_n(addr, real_size) or
__asan_report_store_n(addr, real_size)
is called.
Also, call these two new functions in memset/memcpy
instrumentation.
asan-rt part will follow.
llvm-svn: 175507
This fixes PR15289. This bug was introduced (recently) in r175215; collecting
all std::vector references for candidate pairs to delete at once is invalid
because subsequent lookups in the owning DenseMap could invalidate the
references.
bugpoint was able to reduce a useful test case. Unfortunately, because whether
or not this asserts depends on memory layout, this test case will sometimes
appear to produce valid output. Nevertheless, running under valgrind will
reveal the error.
llvm-svn: 175397
(or (bool?A:B),(bool?C:D)) --> (bool?(or A,C):(or B,D))
By the time the OR is visited, both the SELECTs have been visited and not
optimized and the OR itself hasn't been transformed so we do this transform in
the hopes that the new ORs will be optimized.
The transform is explicitly disabled for vector-selects until "codegen matures
to handle them better".
Patch by Muhammad Tauqir!
llvm-svn: 175380
Several functions and variable names used the term 'tree' to refer
to what is actually a DAG. Correcting this mistake will, hopefully,
prevent confusion in the future.
No functionality change intended.
llvm-svn: 175278
It enables to work with a smaller constant, which is target friendly for those which can compare to immediates.
It also avoids inserting a shift in favor of a trunc, which can be free on some targets.
This used to work until LLVM-3.1, but regressed with the 3.2 release.
llvm-svn: 175270
For some basic blocks, it is possible to generate many candidate pairs for
relatively few pairable instructions. When many (tens of thousands) of these pairs
are generated for a single instruction group, the time taken to generate and
rank the different vectorization plans can become quite large. As a result, we now
cap the number of candidate pairs within each instruction group. This is done by
closing out the group once the threshold is reached (set now at 3000 pairs).
Although this will limit the overall compile-time impact, this may not be the best
way to achieve this result. It might be better, for example, to prune excessive
candidate pairs after the fact the prevent the generation of short, but highly-connected
groups. We can experiment with this in the future.
This change reduces the overall compile-time slowdown of the csa.ll test case in
PR15222 to ~5x. If 5x is still considered too large, a lower limit can be
used as the default.
This represents a functionality change, but only for very large inputs
(thus, there is no regression test).
llvm-svn: 175251
All instances of std::multimap have now been replaced by
DenseMap<K, std::vector<V> >, and this yields a speedup of 5% on the
csa.ll test case from PR15222.
No functionality change intended.
llvm-svn: 175216
This is another commit on the road to removing std::multimap from
BBVectorize. This gives an ~1% speedup on the csa.ll test case
in PR15222.
No functionality change intended.
llvm-svn: 175215
It's possible (e.g. after an LTO build) that an internal global may be used for
debugging purposes. If that's the case appending a '.b' to it makes it hard to
find that variable. Steal the name from the old GV before deleting it so that
they can find that variable again.
llvm-svn: 175104
When building the pairable-instruction dependency map, don't search
past the last pairable instruction. For large blocks that have been
divided into multiple instruction groups, searching past the last
instruction in each group is very wasteful. This gives a 32% speedup
on the csa.ll test case from PR15222 (when using 50 instructions
in each group).
No functionality change intended.
llvm-svn: 174915
This map is queried only for instructions in pairs of pairable
instructions; so make sure that only pairs of pairable
instructions are added to the map. This gives a 3.5% speedup
on the csa.ll test case from PR15222.
No functionality change intended.
llvm-svn: 174914
This eliminates one more linear search over a range of
std::multimap entries. This gives a 22% speedup on the
csa.ll test case from PR15222.
No functionality change intended.
llvm-svn: 174893
This flag makes asan use a small (<2G) offset for 64-bit asan shadow mapping.
On x86_64 this saves us a register, thus achieving ~2/3 of the
zero-base-offset's benefits in both performance and code size.
Thanks Jakub Jelinek for the idea.
llvm-svn: 174886
This removes the last of the linear searches over ranges of std::multimap
iterators, giving a 7% speedup on the doduc.bc input from PR15222.
No functionality change intended.
llvm-svn: 174859
Profiling suggests that getInstructionTypes is performance-sensitive,
this cleans up some double-casting in that function in favor of
using dyn_cast.
No functionality change intended.
llvm-svn: 174857
By itself, this does not have much of an effect, but only because in the default
configuration the full cycle checks are used only for small problem sizes.
This is part of a general cleanup of uses of iteration over std::multimap
ranges only for the purpose of checking membership.
No functionality change intended.
llvm-svn: 174856
This reverts r171041. This was a nice idea that didn't work out well.
Clang warnings need to be associated with warning groups so that they can
be selectively disabled, promoted to errors, etc. This simplistic patch didn't
allow for that. Enhancing it to provide some way for the backend to specify
a front-end warning type seems like overkill for the few uses of this, at
least for now.
llvm-svn: 174748
This is a follow-up to the cost-model change in r174713 which splits
the cost of a memory operation between the address computation and the
actual memory access. In r174713, this cost is always added to the
memory operation cost, and so BBVectorize will do the same.
Currently, this new cost function is used only by ARM, and I don't
have any ARM test cases for BBVectorize. Assistance in generating some
good ARM test cases for BBVectorize would be greatly appreciated!
llvm-svn: 174743
isn't using the default calling convention. However, if the transformation is
from a call to inline IR, then the calling convention doesn't matter.
rdar://13157990
llvm-svn: 174724
Adds a function to target transform info to query for the cost of address
computation. The cost model analysis pass now also queries this interface.
The code in LoopVectorize adds the cost of address computation as part of the
memory instruction cost calculation. Only there, we know whether the instruction
will be scalarized or not.
Increase the penality for inserting in to D registers on swift. This becomes
necessary because we now always assume that address computation has a cost and
three is a closer value to the architecture.
radar://13097204
llvm-svn: 174713
We don't want too many classes in a pass and the classes obscure the details. I
was going a little overboard with object modeling here. Replace classes by
generic code that handles both loads and stores.
No functionality change intended.
llvm-svn: 174646
1. Moved a comment from ObjCARCOpts.cpp -> ObjCARCContract.cpp.
2. Removed a comment from ObjCARCOpts.cpp that was already moved to
ObjCARCAliasAnalysis.h/.cpp.
llvm-svn: 174581
In the loop vectorizer cost model, we used to ignore stores/loads of a pointer
type when computing the widest type within a loop. This meant that if we had
only stores/loads of pointers in a loop we would return a widest type of 8bits
(instead of 32 or 64 bit) and therefore a vector factor that was too big.
Now, if we see a consecutive store/load of pointers we use the size of a pointer
(from data layout).
This problem occured in SingleSource/Benchmarks/Shootout-C++/hash.cpp (reduced
test case is the first test in vector_ptr_load_store.ll).
radar://13139343
llvm-svn: 174377
says, but that's a defect (to be filed). "Cls::purevfn()" is still an odr use.
Also fixes a bug in the previous patch that caused us to not mark the function
referenced just because we didn't want to mark it odr used.
llvm-svn: 174240
the SCEV vector size in LoopStrengthReduce. It is observed that
the BaseRegs vector size is 4 in most cases,
and elements are frequently copied when it is initialized as
SmallVector<const SCEV *, 2> BaseRegs.
Our benchmark results show that the compilation time performance
improved by ~0.5%.
Patch by Wan Xiaofei.
llvm-svn: 174219
This is a re-worked version of r174048.
Given source IR:
call void @llvm.dbg.declare(metadata !{i32* %argc.addr}, metadata !14), !dbg !15
we used to generate
call void @llvm.dbg.declare(metadata !27, metadata !28), !dbg !29!27 = metadata !{null}
With this patch, we will correctly generate
call void @llvm.dbg.declare(metadata !{i32* %argc.addr}, metadata !27), !dbg !28
Looking up %argc.addr in ValueMap will return null, since %argc.addr is already
correctly set up, we can use identity mapping.
rdar://problem/13089880
llvm-svn: 174093
Given source IR:
call void @llvm.dbg.declare(metadata !{i32* %argc.addr}, metadata !14), !dbg !15
we used to generate
call void @llvm.dbg.declare(metadata !27, metadata !28), !dbg !29!27 = metadata !{null}
With this patch, we will correctly generate
call void @llvm.dbg.declare(metadata !{i32* %argc.addr}, metadata !27), !dbg !28
Looking up %argc.addr in ValueMap will return null, since %argc.addr is already
correctly set up, we can use identity mapping.
llvm-svn: 173946
This was missed since whenever I was including ObjCARCAliasAnalysis.h, I
was including ObjCARC.h before it which included these includes
(resulting in no compilation breakage).
llvm-svn: 173764
This also required adding 2x headers Dependency Analysis.h/Provenance Analysis.h
and a .cpp file DependencyAnalysis.cpp to unentangle the dependencies inbetween
ObjCARCContract and ObjCARCOpts.
llvm-svn: 173760
Because BBVectorize may significantly shorten a loop body, unroll
again after vectorization. This is especially important when using
runtime or partial unrolling.
llvm-svn: 173730
It is way too slow. Change the default option value to 0.
Always do exact shadow propagation for unsigned ICmp with constants, it is
cheap (under 1% cpu time) and required for correctness.
llvm-svn: 173682
When flipping the pair of subvectors that form a vector, if the
vector length is 2, we can use the SK_Reverse shuffle kind to get
more-accurate cost information. Also we can use the SK_ExtractSubvector
shuffle kind to get accurate subvector extraction costs.
The current cost model implementations don't yet seem complex enough
for this to make a difference (thus, there are no test cases with this
commit), but it should help in future.
Depending on how the various targets optimize and combine shuffles in
practice, we might be able to get more-accurate costs by combining the
costs of multiple shuffle kinds. For example, the cost of flipping the
subvector pairs could be modeled as two extractions and two subvector
insertions. These changes, however, should probably be motivated
by specific test cases.
llvm-svn: 173621
This name change does the following:
1. Causes the function name to use proper ARC terminology.
2. Makes it clear what the function truly does.
llvm-svn: 173609
In the future, AttributeWithIndex won't be used anymore. Besides, it exposes the
internals of the AttributeSet to outside users, which isn't goodness.
llvm-svn: 173602
In the future, AttributeWithIndex won't be used anymore. Besides, it exposes the
internals of the AttributeSet to outside users, which isn't goodness.
llvm-svn: 173601
In the future, AttributeWithIndex won't be used anymore. Besides, it exposes the
internals of the AttributeSet to outside users, which isn't goodness.
llvm-svn: 173600
The 'getSlot' function and its ilk allow introspection into the AttributeSet
class. However, that class should be opaque. Allow access through accessor
methods instead.
llvm-svn: 173522
Only for integers, pointers, and vectors of those. No floats.
Instrumentation seems very heavy, and may need to be replaced
with some approximation in the future.
llvm-svn: 173452
loops over instructions in the basic block or the use-def list of the
value, neither of which are really efficient when repeatedly querying
about values in the same basic block.
What's more, we already know that the CondBB is small, and so we can do
a much more efficient test by counting the uses in CondBB, and seeing if
those account for all of the uses.
Finally, we shouldn't blanket fail on any such instruction, instead we
should conservatively assume that those instructions are part of the
cost.
Note that this actually fixes a bug in the pass because
isUsedInBasicBlock has a really terrible bug in it. I'll fix that in my
next commit, but the fix for it would make this code suddenly take the
compile time hit I thought it already was taking, so I wanted to go
ahead and migrate this code to a faster & better pattern.
The bug in isUsedInBasicBlock was also causing other tests to test the
wrong thing entirely: for example we weren't actually disabling
speculation for floating point operations as intended (and tested), but
the test passed because we failed to speculate them due to the
isUsedInBasicBlock failure.
llvm-svn: 173417
Original commit message:
Plug TTI into the speculation logic, giving it a real cost interface
that can be specialized by targets.
The goal here is not to be more aggressive, but to just be more accurate
with very obvious cases. There are instructions which are known to be
truly free and which were not being modeled as such in this code -- see
the regression test which is distilled from an inner loop of zlib.
Everywhere the TTI cost model is insufficiently conservative I've added
explicit checks with FIXME comments to go add proper modelling of these
cost factors.
If this causes regressions, the likely solution is to make TTI even more
conservative in its cost estimates, but test cases will help here.
llvm-svn: 173357
that can be specialized by targets.
The goal here is not to be more aggressive, but to just be more accurate
with very obvious cases. There are instructions which are known to be
truly free and which were not being modeled as such in this code -- see
the regression test which is distilled from an inner loop of zlib.
Everywhere the TTI cost model is insufficiently conservative I've added
explicit checks with FIXME comments to go add proper modelling of these
cost factors.
If this causes regressions, the likely solution is to make TTI even more
conservative in its cost estimates, but test cases will help here.
llvm-svn: 173342
a cost fuction that seems both a bit ad-hoc and also poorly suited to
evaluating constant expressions.
Notably, it is missing any support for trivial expressions such as
'inttoptr'. I could fix this routine, but it isn't clear to me all of
the constraints its other users are operating under.
The core protection that seems relevant here is avoiding the formation
of a select instruction wich a further chain of select operations in
a constant expression operand. Just explicitly encode that constraint.
Also, update the comments and organization here to make it clear where
this needs to go -- this should be driven off of real cost measurements
which take into account the number of constants expressions and the
depth of the constant expression tree.
llvm-svn: 173340
terms of cost rather than hoisting a single instruction.
This does *not* change the cost model! We still set the cost threshold
at 1 here, it's just that we track it by accumulating cost rather than
by storing an instruction.
The primary advantage is that we no longer leave no-op intrinsics in the
basic block. For example, this will now move both debug info intrinsics
and a single instruction, instead of only moving the instruction and
leaving a basic block with nothing bug debug info intrinsics in it, and
those intrinsics now no longer ordered correctly with the hoisted value.
Instead, we now splice the entire conditional basic block's instruction
sequence.
This also places the code for checking the safety of hoisting next to
the code computing the cost.
Currently, the only observable side-effect of this change is that debug
info intrinsics are no longer abandoned. I'm not sure how to craft
a test case for this, and my real goal was the refactoring, but I'll
talk to Dave or Eric about how to add a test case for this.
llvm-svn: 173339
Chandler Carruth