This patch teaches LICM's implementation of store promotion to exploit the fact that the memory location being accessed might be provable thread local. The fact it's thread local weakens the requirements for where we can insert stores since no other thread can observe the write. This allows us perform store promotion even in cases where the store is not guaranteed to execute in the loop.
Two key assumption worth drawing out is that this assumes a) no-capture is strong enough to imply no-escape, and b) standard allocation functions like malloc, calloc, and operator new return values which can be assumed not to have previously escaped.
In future work, it would be nice to generalize this so that it works without directly seeing the allocation site. I believe that the nocapture return attribute should be suitable for this purpose, but haven't investigated carefully. It's also likely that we could support unescaped allocas with similar reasoning, but since SROA and Mem2Reg should destroy those, they're less interesting than they first might seem.
Differential Revision: http://reviews.llvm.org/D16783
llvm-svn: 263072
When checking whether an smin is positive, we can move the comparison to one of the inputs if the other is known positive. If the known positive one is the min, then the other can't be negative. If the other is the min, then we compute the min.
Differential Revision: http://reviews.llvm.org/D17873
llvm-svn: 263059
I somehow missed this. The case in GCC (global_alloc) was similar to
the new testcase except it had an array of structs rather than a two
dimensional array.
Fixes RP26885.
llvm-svn: 263058
As part of r251146 InstCombine was extended to call computeKnownBits on
every value in the function to determine whether it happens to be
constant. This increases typical compiletime by 1-3% (5% in irgen+opt
time) in my measurements. On the other hand this case did not trigger
once in the whole llvm-testsuite.
This patch introduces the notion of ExpensiveCombines which are only
enabled for OptLevel > 2. I removed the check in InstructionSimplify as
that is called from various places where the OptLevel is not known but
given the rarity of the situation I think a check in InstCombine is
enough.
Differential Revision: http://reviews.llvm.org/D16835
llvm-svn: 263047
Original commit message:
calculate builtin_object_size if argument is a removable pointer
This patch fixes calculating correct value for builtin_object_size function
when pointer is used only in builtin_object_size function call and never
after that.
Patch by Strahinja Petrovic.
Differential Revision: http://reviews.llvm.org/D17337
Reland the original change with a small modification (first do a null check
and then do the cast) to satisfy ubsan.
llvm-svn: 263011
TSan instrumentation functions for atomic stores, loads, and cmpxchg work on
integer value types. This patch adds casts before calling TSan instrumentation
functions in cases where the value is a pointer.
Differential Revision: http://reviews.llvm.org/D17833
llvm-svn: 262876
This lets select sub-targets enable this pass. The patch implements the
idea from the recent llvm-dev thread:
http://thread.gmane.org/gmane.comp.compilers.llvm.devel/94925
The goal is to enable the LoopDataPrefetch pass for the Cyclone
sub-target only within Aarch64.
Positive and negative tests will be included in an upcoming patch that
enables selective prefetching of large-strided accesses on Cyclone.
llvm-svn: 262844
This reverts commit r262250.
It causes SPEC2006/gcc to generate wrong result (166.s) in AArch64 when
running with *ref* data set. The error happens with
"-Ofast -flto -fuse-ld=gold" or "-O3 -fno-strict-aliasing".
llvm-svn: 262839
This code has been successfully used to bootstrap libc++ in a no-asserts
mode for a very long time, so the code that follows cannot be completely
incorrect. I've added a test that shows the current behavior for this
kind of code with DFSan. If it is desirable for DFSan to do something
special when processing an invoke of a variadic function, it can be
added, but we shouldn't keep an assert that we've been ignoring due to
release builds anyways.
llvm-svn: 262829
Given that we're not actually reducing the instruction count in the included
regression tests, I think we would call this a canonicalization step.
The motivation comes from the example in PR26702:
https://llvm.org/bugs/show_bug.cgi?id=26702
If we hoist the bitwise logic ahead of the bitcast, the previously unoptimizable
example of:
define <4 x i32> @is_negative(<4 x i32> %x) {
%lobit = ashr <4 x i32> %x, <i32 31, i32 31, i32 31, i32 31>
%not = xor <4 x i32> %lobit, <i32 -1, i32 -1, i32 -1, i32 -1>
%bc = bitcast <4 x i32> %not to <2 x i64>
%notnot = xor <2 x i64> %bc, <i64 -1, i64 -1>
%bc2 = bitcast <2 x i64> %notnot to <4 x i32>
ret <4 x i32> %bc2
}
Simplifies to the expected:
define <4 x i32> @is_negative(<4 x i32> %x) {
%lobit = ashr <4 x i32> %x, <i32 31, i32 31, i32 31, i32 31>
ret <4 x i32> %lobit
}
Differential Revision: http://reviews.llvm.org/D17583
llvm-svn: 262645
This patch provides the following infrastructure for PGO enhancements in inliner:
Enable the use of block level profile information in inliner
Incremental update of block frequency information during inlining
Update the function entry counts of callees when they get inlined into callers.
Differential Revision: http://reviews.llvm.org/D16381
llvm-svn: 262636
Summary: With discriminator, LineLocation can uniquely identify a callsite without the need to specifying callee name. Remove Callee function name from the key, and put it in the value (FunctionSamples).
Reviewers: davidxl, dnovillo
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D17827
llvm-svn: 262634
The vectorization of first-order recurrences (r261346) caused PR26734. When
detecting these recurrences, we need to ensure that the previous value is
actually defined inside the loop. This patch includes the fix and test case.
llvm-svn: 262624
Summary: This is the last step toward supporting aggregate memory access in instcombine. This explodes stores of arrays into a serie of stores for each element, allowing them to be optimized.
Reviewers: joker.eph, reames, hfinkel, majnemer, mgrang
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D17828
llvm-svn: 262530
Summary: This is another step toward improving fca support. This unpack load of array in a series of load to array's elements.
Reviewers: chandlerc, joker.eph, majnemer, reames, hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15890
llvm-svn: 262521
parts of the AA interface out of the base class of every single AA
result object.
Because this logic reformulates the query in terms of some other aspect
of the API, it would easily cause O(n^2) query patterns in alias
analysis. These could in turn be magnified further based on the number
of call arguments, and then further based on the number of AA queries
made for a particular call. This ended up causing problems for Rust that
were actually noticable enough to get a bug (PR26564) and probably other
places as well.
When originally re-working the AA infrastructure, the desire was to
regularize the pattern of refinement without losing any generality.
While I think it was successful, that is clearly proving to be too
costly. And the cost is needless: we gain no actual improvement for this
generality of making a direct query to tbaa actually be able to
re-use some other alias analysis's refinement logic for one of the other
APIs, or some such. In short, this is entirely wasted work.
To the extent possible, delegation to other API surfaces should be done
at the aggregation layer so that we can avoid re-walking the
aggregation. In fact, this significantly simplifies the logic as we no
longer need to smuggle the aggregation layer into each alias analysis
(or the TargetLibraryInfo into each alias analysis just so we can form
argument memory locations!).
However, we also have some delegation logic inside of BasicAA and some
of it even makes sense. When the delegation logic is baking in specific
knowledge of aliasing properties of the LLVM IR, as opposed to simply
reformulating the query to utilize a different alias analysis interface
entry point, it makes a lot of sense to restrict that logic to
a different layer such as BasicAA. So one aspect of the delegation that
was in every AA base class is that when we don't have operand bundles,
we re-use function AA results as a fallback for callsite alias results.
This relies on the IR properties of calls and functions w.r.t. aliasing,
and so seems a better fit to BasicAA. I've lifted the logic up to that
point where it seems to be a natural fit. This still does a bit of
redundant work (we query function attributes twice, once via the
callsite and once via the function AA query) but it is *exactly* twice
here, no more.
The end result is that all of the delegation logic is hoisted out of the
base class and into either the aggregation layer when it is a pure
retargeting to a different API surface, or into BasicAA when it relies
on the IR's aliasing properties. This should fix the quadratic query
pattern reported in PR26564, although I don't have a stand-alone test
case to reproduce it.
It also seems general goodness. Now the numerous AAs that don't need
target library info don't carry it around and depend on it. I think
I can even rip out the general access to the aggregation layer and only
expose that in BasicAA as it is the only place where we re-query in that
manner.
However, this is a non-trivial change to the AA infrastructure so I want
to get some additional eyes on this before it lands. Sadly, it can't
wait long because we should really cherry pick this into 3.8 if we're
going to go this route.
Differential Revision: http://reviews.llvm.org/D17329
llvm-svn: 262490
As noted in the code comment, I don't think we can do the same transform that we do for
*scalar* integers comparisons to *vector* integers comparisons because it might pessimize
the general case.
Exhibit A for an incomplete integer comparison ISA remains x86 SSE/AVX: it only has EQ and GT
for integer vectors.
But we should now recognize all the variants of this construct and produce the optimal code
for the cases shown in:
https://llvm.org/bugs/show_bug.cgi?id=26701
llvm-svn: 262424
Summary: SampleProfile pass needs to be performed after InstructionCombiningPass, which helps eliminate un-inlinable function calls.
Reviewers: davidxl, dnovillo
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D17742
llvm-svn: 262419
Summary:
This adds the beginning of an update API to preserve MemorySSA. In particular,
this patch adds a way to remove memory SSA accesses when instructions are
deleted.
It also adds relevant unit testing infrastructure for MemorySSA's API.
(There is an actual user of this API, i will make that diff dependent on this one. In practice, a ton of opt passes remove memory instructions, so it's hopefully an obviously useful API :P)
Reviewers: hfinkel, reames, george.burgess.iv
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D17157
llvm-svn: 262362
This patch fixes calculating correct value for builtin_object_size function
when pointer is used only in builtin_object_size function call and never
after that.
Patch by Strahinja Petrovic.
Differential Revision: http://reviews.llvm.org/D17337
llvm-svn: 262337
The intended effect of this patch in conjunction with:
http://reviews.llvm.org/rL259392http://reviews.llvm.org/rL260145
is that customers using the AVX intrinsics in C will benefit from combines when
the load mask is constant:
__m128 mload_zeros(float *f) {
return _mm_maskload_ps(f, _mm_set1_epi32(0));
}
__m128 mload_fakeones(float *f) {
return _mm_maskload_ps(f, _mm_set1_epi32(1));
}
__m128 mload_ones(float *f) {
return _mm_maskload_ps(f, _mm_set1_epi32(0x80000000));
}
__m128 mload_oneset(float *f) {
return _mm_maskload_ps(f, _mm_set_epi32(0x80000000, 0, 0, 0));
}
...so none of the above will actually generate a masked load for optimized code.
This is the masked load counterpart to:
http://reviews.llvm.org/rL262064
llvm-svn: 262269
Summary:
I re-benchmarked this and results are similar to original results in
D13259:
On ARM64:
SingleSource/Benchmarks/Polybench/linear-algebra/solvers/dynprog -59.27%
SingleSource/Benchmarks/Polybench/stencils/adi -19.78%
On x86:
SingleSource/Benchmarks/Polybench/linear-algebra/solvers/dynprog -27.14%
And of course the original ~20% gain on SPECint_2006/456.hmmer with Loop
Distribution.
In terms of compile time, there is ~5% increase on both
SingleSource/Benchmarks/Misc/oourafft and
SingleSource/Benchmarks/Linkpack/linkpack-pc. These are both very tiny
loop-intensive programs where SCEV computations dominates compile time.
The reason that time spent in SCEV increases has to do with the design
of the old pass manager. If a transform pass does not preserve an
analysis we *invalidate* the analysis even if there was *no*
modification made by the transform pass.
This means that currently we don't take advantage of LLE and LV sharing
the same analysis (LAA) and unfortunately we recompute LAA *and* SCEV
for LLE.
(There should be a way to work around this limitation in the case of
SCEV and LAA since both compute things on demand and internally cache
their result. Thus we could pretend that transform passes preserve
these analyses and manually invalidate them upon actual modification.
On the other hand the new pass manager is supposed to solve so I am not
sure if this is worthwhile.)
Reviewers: hfinkel, dberlin
Subscribers: dberlin, reames, mssimpso, aemerson, joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D16300
llvm-svn: 262250
Philip Reames