through PHI nodes across iterations.
This patch teaches the new advanced loop unrolling heuristics to propagate
constants into the loop from the preheader and around the backedge after
simulating each iteration. This lets us brute force solve simple recurrances
that aren't modeled effectively by SCEV. It also makes it more clear why we
need to process the loop in-order rather than bottom-up which might otherwise
make much more sense (for example, for DCE).
This came out of an attempt I'm making to develop a principled way to account
for dead code in the unroll estimation. When I implemented
a forward-propagating version of that it produced incorrect results due to
failing to propagate *cost* between loop iterations through the PHI nodes, and
it occured to me we really should at least propagate simplifications across
those edges, and it is quite easy thanks to the loop being in canonical and
LCSSA form.
Differential Revision: http://reviews.llvm.org/D11706
llvm-svn: 243900
Since r241097, `DIBuilder` has only created distinct `DICompileUnit`s.
The backend is liable to start relying on that (if it hasn't already),
so make uniquable `DICompileUnit`s illegal and automatically upgrade old
bitcode. This is a nice cleanup, since we can remove an unnecessary
`DenseSet` (and the associated uniquing info) from `LLVMContextImpl`.
Almost all the testcases were updated with this script:
git grep -e '= !DICompileUnit' -l -- test |
grep -v test/Bitcode |
xargs sed -i '' -e 's,= !DICompileUnit,= distinct !DICompileUnit,'
I imagine something similar should work for out-of-tree testcases.
llvm-svn: 243885
Remove the fake `DW_TAG_auto_variable` and `DW_TAG_arg_variable` tags,
using `DW_TAG_variable` in their place Stop exposing the `tag:` field at
all in the assembly format for `DILocalVariable`.
Most of the testcase updates were generated by the following sed script:
find test/ -name "*.ll" -o -name "*.mir" |
xargs grep -l 'DILocalVariable' |
xargs sed -i '' \
-e 's/tag: DW_TAG_arg_variable, //' \
-e 's/tag: DW_TAG_auto_variable, //'
There were only a handful of tests in `test/Assembly` that I needed to
update by hand.
(Note: a follow-up could change `DILocalVariable::DILocalVariable()` to
set the tag to `DW_TAG_formal_parameter` instead of `DW_TAG_variable`
(as appropriate), instead of having that logic magically in the backend
in `DbgVariable`. I've added a FIXME to that effect.)
llvm-svn: 243774
The patch changes the SLPVectorizer::vectorizeStores to choose the immediate
succeeding or preceding candidate for a store instruction when it has multiple
consecutive candidates. In this way it has better chance to find more slp
vectorization opportunities.
Differential Revision: http://reviews.llvm.org/D10445
llvm-svn: 243666
Summary:
As added initially, statepoints required their call targets to be a
constant pointer null if ``numPatchBytes`` was non-zero. This turns out
to be a problem ergonomically, since there is no way to mark patchable
statepoints as calling a (readable) symbolic value.
This change remove the restriction of requiring ``null`` call targets
for patchable statepoints, and changes PlaceSafepoints to maintain the
symbolic call target through its transformation.
Reviewers: reames, swaroop.sridhar
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11550
llvm-svn: 243502
Summary:
Make Scalar Evolution able to propagate NSW and NUW flags from instructions to SCEVs in some cases. This is based on reasoning about when poison from instructions with these flags would trigger undefined behavior. This gives a 13% speed-up on some Eigen3-based Google-internal microbenchmarks for NVPTX.
There does not seem to be clear agreement about when poison should be considered to propagate through instructions. In this analysis, poison propagates only in cases where that should be uncontroversial.
This change makes LSR able to create induction variables for expressions like &ptr[i + offset] for loops like this:
for (int i = 0; i < limit; ++i) {
sum += ptr[i + offset];
}
Here ptr is a 64 bit pointer and offset is a 32 bit integer. For NVPTX, LSR currently creates an induction variable for i + offset instead, which is not as fast. Improving this situation is what brings the 13% speed-up on some Eigen3-based Google-internal microbenchmarks for NVPTX.
There are more details in this discussion on llvmdev.
June: http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-June/thread.html#87234
July: http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-July/thread.html#87392
Patch by Bjarke Roune
Reviewers: eliben, atrick, sanjoy
Subscribers: majnemer, hfinkel, jingyue, meheff, llvm-commits
Differential Revision: http://reviews.llvm.org/D11212
llvm-svn: 243460
The 'common' section TLS is not implemented.
Current C/C++ TLS variables are not placed in common section.
DWARF debug info to get the address of TLS variables is not generated yet.
clang and driver changes in http://reviews.llvm.org/D10524
Added -femulated-tls flag to select the emulated TLS model,
which will be used for old targets like Android that do not
support ELF TLS models.
Added TargetLowering::LowerToTLSEmulatedModel as a target-independent
function to convert a SDNode of TLS variable address to a function call
to __emutls_get_address.
Added into lib/Target/*/*ISelLowering.cpp to call LowerToTLSEmulatedModel
for TLSModel::Emulated. Although all targets supporting ELF TLS models are
enhanced, emulated TLS model has been tested only for Android ELF targets.
Modified AsmPrinter.cpp to print the emutls_v.* and emutls_t.* variables for
emulated TLS variables.
Modified DwarfCompileUnit.cpp to skip some DIE for emulated TLS variabls.
TODO: Add proper DIE for emulated TLS variables.
Added new unit tests with emulated TLS.
Differential Revision: http://reviews.llvm.org/D10522
llvm-svn: 243438
Summary:
If a scale or a base register can be rewritten as "Zext({A,+,1})" then
LSR will now consider a formula of that form in its normal cost
computation.
Depends on D9180
Reviewers: qcolombet, atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9181
llvm-svn: 243348
Summary:
Was D9784: "Remove loop variant range check when induction variable is
strictly increasing"
This change re-implements D9784 with the two differences:
1. It does not use SCEVExpander and does not generate new
instructions. Instead, it does a quick local search for existing
`llvm::Value`s that it needs when modifying the `icmp`
instruction.
2. It is more general -- it deals with both increasing and decreasing
induction variables.
I've added all of the tests included with D9784, and two more.
As an example on what this change does (copied from D9784):
Given C code:
```
for (int i = M; i < N; i++) // i is known not to overflow
if (i < 0) break;
a[i] = 0;
}
```
This transformation produces:
```
for (int i = M; i < N; i++)
if (M < 0) break;
a[i] = 0;
}
```
Which can be unswitched into:
```
if (!(M < 0))
for (int i = M; i < N; i++)
a[i] = 0;
}
```
I went back and forth on whether the top level logic should live in
`SimplifyIndvar::eliminateIVComparison` or be put into its own
routine. Right now I've put it under `eliminateIVComparison` because
even though the `icmp` is not *eliminated*, it no longer is an IV
comparison. I'm open to putting it in its own helper routine if you
think that is better.
Reviewers: reames, nicholas, atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11278
llvm-svn: 243331
Now that we are generating sane codegen for vector sext/zext nodes on SSE targets, this patch uses instcombine to replace the SSE41/AVX2 pmovsx and pmovzx intrinsics with the equivalent native IR code.
Differential Revision: http://reviews.llvm.org/D11503
llvm-svn: 243303
The pointer size of the addrspacecasted pointer might not have matched,
so this would have hit an assert in accumulateConstantOffset.
I think this was here to allow constant folding of a load of an
addrspacecasted constant. Accumulating the offset through the
addrspacecast doesn't make much sense, so something else is necessary
to allow folding the load through this cast.
llvm-svn: 243300
Summary:
Fix the cost of interleaved accesses for ARM/AArch64.
We were calling getTypeAllocSize and using it to check
the number of bits, when we should have called
getTypeAllocSizeInBits instead.
This would pottentially cause the vectorizer to
generate loads/stores and shuffles which cannot
be matched with an interleaved access instruction.
No performance changes are expected for now since
matching/generating interleaved accesses is still
disabled by default.
Reviewers: rengolin
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D11524
llvm-svn: 243270
r243250 appeared to break clang/test/Analysis/dead-store.c on one of the build
slaves, but I couldn't reproduce this failure locally. Probably a false
positive as I saw this test was broken by r243246 or r243247 too but passed
later without people fixing anything.
llvm-svn: 243253
Summary:
This patch updates TargetTransformInfoImplCRTPBase::getGEPCost to consider
addressing modes. It now returns TCC_Free when the GEP can be completely folded
to an addresing mode.
I started this patch as I refactored SLSR. Function isGEPFoldable looks common
and is indeed used by some WIP of mine. So I extracted that logic to getGEPCost.
Furthermore, I noticed getGEPCost wasn't directly tested anywhere. The best
testing bed seems CostModel, but its getInstructionCost method invokes
getAddressComputationCost for GEPs which provides very coarse estimation. So
this patch also makes getInstructionCost call the updated getGEPCost for GEPs.
This change inevitably breaks some tests because the cost model changes, but
nothing looks seriously wrong -- if we believe the new cost model is the right
way to go, these tests should be updated.
This patch is not perfect yet -- the comments in some tests need to be updated.
I want to know whether this is a right approach before fixing those details.
Reviewers: chandlerc, hfinkel
Subscribers: aschwaighofer, llvm-commits, aemerson
Differential Revision: http://reviews.llvm.org/D9819
llvm-svn: 243250
Summary:
This patch improves trivial loop unswitch.
The current trivial loop unswitch only checks if loop header's terminator contains a trivial unswitch condition. But if the loop header only has one reachable successor (due to intentionally or unintentionally missed code simplification), we should consider the successor as part of the loop header. Therefore, instead of stopping at loop header's terminator, we should keep traversing its successors within loop until reach a *real* conditional branch or switch (whose condition can not be constant folded). This change will enable a single -loop-unswitch pass to unswitch multiple trivial conditions (unswitch one trivial condition could open opportunity to unswitch another one in the same loop), while the old implementation can unswitch only one per pass.
Reviewers: reames, broune
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11481
llvm-svn: 243203
Add a verifier check that `DILocalVariable`s of tag
`DW_TAG_arg_variable` always have a non-zero 'arg:' field, and those of
tag `DW_TAG_auto_variable` always have a zero 'arg:' field. These are
the only configurations that are properly understood by the backend.
(Also, fix the bad examples in LangRef and test/Assembler, and fix the
bug in Kaleidoscope Ch8.)
A large number of testcases seem to have bitrotted their way forward
from some ancient version of the debug info hierarchy that didn't have
`arg:` parameters. If you have out-of-tree testcases that start failing
in the verifier and you don't care enough to get the `arg:` right, you
may have some luck just calling:
sed -e 's/, arg: 0/, arg: 1/'
or some such, but I hand-updated the ones in tree.
llvm-svn: 243183
This patch extend LoopReroll pass to hand the loops which
is similar to the following:
while (len > 1) {
sum4 += buf[len];
sum4 += buf[len-1];
len -= 2;
}
llvm-svn: 243171
The names for instructions inserted were previous dependent on iteration order. By deriving the names from the original instructions, we can avoid instability in tests without resorting to ordered traversals. It also makes the IR mildly easier to read at large scale.
llvm-svn: 243140
Summary:
Resolving a branch allows us to ignore blocks that won't be executed, and thus make our estimate more accurate.
This patch is intended to be applied after D10205 (though it could be applied independently).
Reviewers: chandlerc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10206
llvm-svn: 243084
Summary:
Scalarizer has two data structures that hold information about changes
to the function, Gathered and Scattered. These are cleared in finish()
at the end of runOnFunction() if finish() detects any changes to the
function.
However, finish() was checking for changes by only checking if
Gathered was non-empty. The function visitStore() only modifies
Scattered without touching Gathered. As a result, Scattered could have
ended up having stale data if Scalarizer only scalarized store
instructions. Since the data in Scattered is used during the execution
of the pass, this introduced dangling pointer errors.
The fix is to check whether both Scattered and Gathered are empty
before deciding what to do in finish(). This also fixes a problem
where the Function can be modified although the pass returns false.
Reviewers: rnk
Subscribers: rnk, srhines, llvm-commits
Differential Revision: http://reviews.llvm.org/D10459
llvm-svn: 243040
The MSVC ABI requires that we generate an alias for the vtable which
means looking through a GlobalAlias which cannot be overridden improves
our ability to devirtualize.
Found while investigating PR20801.
Patch by Andrew Zhogin!
Differential Revision: http://reviews.llvm.org/D11306
llvm-svn: 242955
Currently, a load from an alloca that is used in as single block and is not preceded
by a store is replaced by undef. This is not always correct if the single block is
inside a loop.
Fix the logic so that:
1) If there are no stores in the block, replace the load with an undef, as before.
2) If there is a store (regardless of where it is in the block w.r.t the load), bail
out, and let the rest of mem2reg handle this alloca.
Patch by: gil.rapaport@intel.com
Differential Revision: http://reviews.llvm.org/D11355
llvm-svn: 242884
Summary: The current code in LoopUnswtich::processCurrentLoop() mixes trivial loop unswitch and non-trivial loop unswitch together. It goes over all basic blocks in the loop and checks if a condition is trivial or non-trivial unswitch condition. However, trivial unswitch condition can only occur in the loop header basic block (where it controls whether or not the loop does something at all). This refactoring separate trivial loop unswitch and non-trivial loop unswitch. Before going over all basic blocks in the loop, it checks if the loop header contains a trivial unswitch condition. If so, unswitch it. Otherwise, go over all blocks like before but don't check trivial condition any more since they are not possible to be in the other blocks. This code has no functionality change.
Reviewers: meheff, reames, broune
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11276
llvm-svn: 242873
types and loads, loads or stores widened past the size of an alloca,
etc.
This started off with a bug report about big-endian behavior with
bitfields and loads and stores to a { i32, i24 } struct. An initial
attempt to fix this was sent for review in D10357, but that didn't
really get to the root of the problem.
The core issue was that canConvertValue and convertValue in SROA were
handling different bitwidth integers by doing a zext of the integer. It
wouldn't do a trunc though, only a zext! This would in turn lead SROA to
form an i24 load from an i24 alloca, zext it to i32, and then use it.
This would at least produce the wrong value for big-endian systems.
One of my many false starts here was to correct the computation for
big-endian systems by shifting. But this doesn't actually work because
the original code has a 64-bit store to the entire 8 bytes, and a 32-bit
load of the last 4 bytes, and because the alloc size is 8 bytes, we
can't lose that last (least significant if bigendian) byte! The real
problem here is that we're forming an i24 load in SROA which is actually
not sufficiently wide to load all of the necessary bits here. The source
has an i32 load, and SROA needs to form that as well.
The straightforward way to do this is to disable the zext logic in
canConvertValue and convertValue, forcing us to actually load all
32-bits. This seems like a really good change, but it in turn breaks
several other parts of SROA.
First in the chain of knock-on failures, we had places where we were
doing integer-widening promotion even though some of the integer loads
or stores extended *past the end* of the alloca's memory! There was even
a comment about preventing this, but it only prevented the case where
the type had a different bit size from its store size. So I added checks
to handle the cases where we actually have a widened load or store and
to avoid trying to special integer widening promotion in those cases.
Second, we actually rely on the ability to promote in the face of loads
past the end of an alloca! This is important so that we can (for
example) speculate loads around PHI nodes to do more promotion. The bits
loaded are garbage, but as long as they aren't used and the alignment is
suitable high (which it wasn't in the test case!) this is "fine". And we
can't stop promoting here, lots of things stop working well if we do. So
we need to add specific logic to handle the extension (and truncation)
case, but *only* where that extension or truncation are over bytes that
*are outside the alloca's allocated storage* and thus totally bogus to
load or store.
And of course, once we add back this correct handling of extension or
truncation, we need to correctly handle bigendian systems to avoid
re-introducing the exact bug that started us off on this chain of misery
in the first place, but this time even more subtle as it only happens
along speculated loads atop a PHI node.
I've ported an existing test for PHI speculation to the big-endian test
file and checked that we get that part correct, and I've added several
more interesting big-endian test cases that should help check that we're
getting this correct.
Fun times.
llvm-svn: 242869
We insert a bitcast which obfuscates the getCalledFunction for the utility
function which looks up attributes from the called function. Loosing ABI
changing parameter attributes is a bad thing.
rdar://21516488
llvm-svn: 242807
A patch by Chakshu Grover!
This patch allows constfolding of trunc,rint,nearbyint,ceil and floor intrinsics using APFloat class.
Differential Revision: http://reviews.llvm.org/D11144
llvm-svn: 242763
Summary:
SpeculativeExecution enables a series straight line optimizations (such
as SLSR and NaryReassociate) on conditional code. For example,
if (...)
... b * s ...
if (...)
... (b + 1) * s ...
speculative execution can hoist b * s and (b + 1) * s from then-blocks,
so that we have
... b * s ...
if (...)
...
... (b + 1) * s ...
if (...)
...
Then, SLSR can rewrite (b + 1) * s to (b * s + s) because after
speculative execution b * s dominates (b + 1) * s.
The performance impact of this change is significant. It speeds up the
benchmarks running EigenFloatContractionKernelInternal16x16
(ba68f42fa6/unsupported/Eigen/CXX11/src/Tensor/TensorContractionCuda.h?at=default#cl-526)
by roughly 2%. Some internal benchmarks that have the above code pattern
are improved by up to 40%. No significant slowdowns are observed on
Eigen CUDA microbenchmarks.
Reviewers: jholewinski, broune, eliben
Subscribers: llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D11201
llvm-svn: 242437
Internalizing an individual comdat group member without also internalizing
the other members of the comdat can break comdat semantics. For example,
if a module contains a reference to an internalized comdat member, and the
linker chooses a comdat group from a different object file, this will break
the reference to the internalized member.
This change causes the internalizer to only internalize comdat members if all
other members of the comdat are not externally visible. Once a comdat group
has been fully internalized, there is no need to apply comdat rules to its
members; later optimization passes (e.g. globaldce) can legally drop individual
members of the comdat. So we drop the comdat attribute from all comdat members.
Differential Revision: http://reviews.llvm.org/D10679
llvm-svn: 242423
Self-referential constants containing references to a merged function
no longer cause the MergeFunctions pass to infinite loop. Also adds a
reproduction IR which would otherwise fail, which was isolated from a similar
issue in Chromium.
Author: jrkoenig
Reviewers: nlewycky, jfb
Subscribers: llvm-commits, nlewycky, jfb
Differential Revision: http://reviews.llvm.org/D11208
llvm-svn: 242337
During estimation of unrolling effect we should be able to propagate
constants through casts.
Differential Revision: http://reviews.llvm.org/D10207
llvm-svn: 242257
Sometimes an incidentally created instruction can duplicate a Value used
elsewhere. It then often doesn't end up in the leader table. If it's later
removed, we attempt to remove it from the leader table and segfault.
Instead we should just ignore the removal request, which won't cause any
problems. The reverse situation, where the original instruction is replaced by
the new one (which you might think could leave the leader table empty) cannot
occur, because the incidental instruction will never be found in the first
place.
llvm-svn: 242199
Volatile loads and stores are made visible in global state regardless of
what memory is involved. It is not correct to disregard the ordering
and synchronization scope because it is possible to synchronize with
memory operations performed by hardware.
This partially addresses PR23737.
llvm-svn: 242126
Previously we would refrain from attempting to increase the linkage of
available_externally globals because they were considered weak for the
linker. Now they are treated more like a declaration instead of a weak
definition.
This was causing SSE alignment faults in Chromuim, when some code
assumed it could increase the alignment of a dllimported global that it
didn't control. http://crbug.com/509256
llvm-svn: 242091
This test case was breaking the hexagon elf bot. The failing lines
were actually unnecessary as checking that the store still reads the
correct value demonstrates that everything is working fine now.
llvm-svn: 242073
When spotting that a loop can use ctpop, we were incorrectly replacing all uses of a value with a value derived from ctpop.
The bug here was exposed because we were replacing a use prior to the ctpop with the ctpop value and so we have a use before def, i.e., we changed
%tobool.5 = icmp ne i32 %num, 0
store i1 %tobool.5, i1* %ptr
br i1 %tobool.5, label %for.body.lr.ph, label %for.end
to
store i1 %1, i1* %ptr
%0 = call i32 @llvm.ctpop.i32(i32 %num)
%1 = icmp ne i32 %0, 0
br i1 %1, label %for.body.lr.ph, label %for.end
Even if we inserted the ctpop so that it dominates the store here, that would still be incorrect. The store doesn’t want the result of ctpop.
The fix is very simple, and involves replacing only the branch condition with the ctpop instead of all uses.
Reviewed by Hal Finkel.
llvm-svn: 242068
Enable runtime unrolling for loops with unroll count metadata ("#pragma unroll N")
and a runtime trip count. Also, do not unroll loops with unroll full metadata if the
loop has a runtime loop count. Previously, such loops would be unrolled with a
very large threshold (pragma-unroll-threshold) if runtime unrolled happened to be
enabled resulting in a very large (and likely unwise) unroll factor.
llvm-svn: 242047
Summary:
This at least saves compile time. I also encountered a case where
ephemeral values affect whether other variables are promoted, causing
performance issues. It may be a bug in LSR, but I didn't manage to
reduce it yet. Anyhow, I believe it's in general not worth considering
ephemeral values in LSR.
Reviewers: atrick, hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11115
llvm-svn: 242011
There is no suitable basic block to sink instructions in loops without
exits. The only way an instruction in a loop without exits can be used
is as an incoming value to a PHI. In such cases, the incoming block for
the corresponding value is unreachable.
This fixes PR24013.
Differential Revision: http://reviews.llvm.org/D10903
llvm-svn: 241987
After changes in rL231820 loop re-rotation is performed even in -Oz mode. Since loop rotation is disabled for -Oz, it seems loop re-rotation should be disabled too.
Differential Revision: http://reviews.llvm.org/D10961
llvm-svn: 241897
Not doing this can lead to misoptimizations down the line, e.g. because
of range metadata on the replacing load excluding values that are valid
for the load that is being replaced.
llvm-svn: 241886
Summary:
In RewriteLoopExitValues, before expanding out an SCEV expression using
SCEVExpander, try to see if an existing LLVM IR expression already
computes the value we're interested in. If so use that existing
expression.
Apart from reducing IndVars' reliance on the rest of the compilation
pipeline, this also prevents IndVars from concluding some expressions as
"high cost" when they're not. For instance,
`InductiveRangeCheckElimination` often emits code of the following form:
```
len = umin(len_A, len_B)
loop:
...
if (i++ < len)
goto loop
outside_loop:
use(i)
```
`SCEVExpander` refuses to rewrite the use of `i` in `outside_loop`,
since it thinks the value of `i` on loop exit, `len`, is a high cost
expansion since it contains an `umax` in it. With this change,
`IndVars` can see that it can re-use `len` instead of creating a new
expression to compute `umin(len_A, len_B)`.
I considered putting this cleverness in `SCEVExpander`, but I was
worried that it may then have a deterimental effect on other passes
that use it. So I decided it was better to just do this in the one
place where it seems like an obviously good idea, with the intent of
generalizing later if needed.
Reviewers: atrick, reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10782
llvm-svn: 241838
Summary:
Often filter-like loops will do memory accesses that are
separated by constant offsets. In these cases it is
common that we will exceed the threshold for the
allowable number of checks.
However, it should be possible to merge such checks,
sice a check of any interval againt two other intervals separated
by a constant offset (a,b), (a+c, b+c) will be equivalent with
a check againt (a, b+c), as long as (a,b) and (a+c, b+c) overlap.
Assuming the loop will be executed for a sufficient number of
iterations, this will be true. If not true, checking against
(a, b+c) is still safe (although not equivalent).
As long as there are no dependencies between two accesses,
we can merge their checks into a single one. We use this
technique to construct groups of accesses, and then check
the intervals associated with the groups instead of
checking the accesses directly.
Reviewers: anemet
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10386
llvm-svn: 241673
Summary:
Initially, these intrinsics seemed like part of a family of "frame"
related intrinsics, but now I think that's more confusing than helpful.
Initially, the LangRef specified that this would create a new kind of
allocation that would be allocated at a fixed offset from the frame
pointer (EBP/RBP). We ended up dropping that design, and leaving the
stack frame layout alone.
These intrinsics are really about sharing local stack allocations, not
frame pointers. I intend to go further and add an `llvm.localaddress()`
intrinsic that returns whatever register (EBP, ESI, ESP, RBX) is being
used to address locals, which should not be confused with the frame
pointer.
Naming suggestions at this point are welcome, I'm happy to re-run sed.
Reviewers: majnemer, nicholas
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11011
llvm-svn: 241633
This checks subtarget feature compatibility for inlining by verifying
that the callee is a strict subset of the caller's features. This includes
the cpu as part of the subtarget we can get via the incoming functions as
the backend takes CPUs as feature sets.
This allows us to inline things like:
int foo() { return baz(); }
int __attribute__((target("sse4.2"))) bar() {
return foo();
}
so that generic code can be inlined into specialized functions.
llvm-svn: 241221
TwoAddressInstructionPass stops after a successful commuting but 3 Addr
conversion might be good for some cases.
Consider:
int foo(int a, int b) {
return a + b;
}
Before this commit, we emit:
addl %esi, %edi
movl %edi, %eax
ret
After this commit, we try 3 Addr conversion:
leal (%rsi,%rdi), %eax
ret
Patch by Volkan Keles <vkeles@apple.com>!
Differential Revision: http://reviews.llvm.org/D10851
llvm-svn: 241206
This is mostly an NFC, which increases code readability (instead of
saving old terminator, generating new one in front of old, and deleting
old, we just call a function). However, it would additionaly copy
the debug location from old instruction to replacement, which
would help PR23837.
llvm-svn: 241197
We would create a phi node with a zero initialized operand instead of
undef in the case where no value was originally available. This was
problematic for x86_mmx which has no null value.
llvm-svn: 241143
Surprisingly, this is a correctness issue: the mmx type exists for
calling convention purposes, LLVM doesn't have a zero representation for
them.
This partially fixes PR23999.
llvm-svn: 241142
Summary:
nsw are flaky and can often be removed by optimizations. This patch enhances
nsw by leveraging @llvm.assume in the IR. Specifically, NaryReassociate now
understands that
assume(a + b >= 0) && assume(a >= 0) ==> a +nsw b
As a result, it can split more sext(a + b) into sext(a) + sext(b) for CSE.
Test Plan: nary-gep.ll
Reviewers: broune, meheff
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D10822
llvm-svn: 241139
Set debug location for terminator instruction in loop backedge block
(which is an unconditional jump to loop header). We can't copy debug
location from original backedges, as there can be several of them,
with different debug info locations. So, we follow the approach of
SplitBlockPredecessors, and copy the debug info from first non-PHI
instruction in the header (i.e. destination block).
This is yet another change for PR23837.
llvm-svn: 240999
If we are dealing with a pointer induction variable, isInductionPHI
gives back a step value of Stride / size of pointer. However, we might
be indexing with a legal type wider than the pointer width.
Handle this by inserting casts where appropriate instead of crashing.
This fixes PR23954.
llvm-svn: 240877
The PruneEH pass tries to annotate functions as 'noreturn' if it doesn't
see a ReturnInst. However, a naked function containing inline assembly
can contain control flow leaving the function.
This fixes PR23971.
llvm-svn: 240876
It is possible for a global to be substituted with another global of a
different type or a different kind (i.e. an alias) at IR link time. One
example of this scenario is when a Microsoft ABI vtable is substituted with
an alias referring to a larger vtable containing an RTTI reference.
This will cause the global to be RAUW'd with a possibly bitcasted reference
to the other global. This will of course also affect any references to the
global in bitset metadata.
The right way to handle such metadata is simply to ignore it. This is sound
because the linked module should contain another copy of the bitset entries as
applied to the new global.
llvm-svn: 240866
This change extends the detection of base pointers for vector constructs to handle arbitrary phi and select nodes. The existing non-vector code already handles those, so this is basically just extending the vector special case to be less special cased. It still isn't generalized vector handling since we can't handle arbitrary vector instructions (e.g. shufflevectors), but it's a lot closer.
The general structure of the change is as follows:
* Extend the base defining value relation over a subset of vector instructions and vector typed phi & select instructions.
* Move scalarization from before base pointer rewriting to after base pointer rewriting. The extension of the BDV relation is sufficient to find vector base phis for vector inputs.
* Preserve the existing special case logic for when the base of a vector element is locally obvious. This general idea could be extended to the scalar case as well.
Differential Revision: http://reviews.llvm.org/D10461#inline-84275
llvm-svn: 240850
If we have a caller that knows a particular argument can never be null, we can exploit this fact while simplifying values in the inline cost analysis. This has the effect of reducing the cost for inlining when a null check is present in the callee, but the value is known non null in the caller. In particular, any dependent control flow can be discounted from the cost estimate.
Note that we use the parameter attributes at the call site to memoize the analysis within the caller's code. The setting of this attribute is done in InstCombine, the inline cost analysis just consumes it. This is intentional and important because we want the inline cost analysis results to be easily cachable themselves. We're not currently doing so, but initial results on LTO indicate this will quickly become important.
Differential Revision: http://reviews.llvm.org/D9129
llvm-svn: 240828
Summary:
Fixes PR23809. Without passing the context to SimplifyICmpInst, we would
use the assume to prove that the condition feeding the assume is
trivially true (see isValidAssumeForContext in ValueTracking.cpp),
causing the removal of the assume which may be useful for later
optimizations.
Test Plan: pr23800.ll
Reviewers: hfinkel, majnemer
Reviewed By: hfinkel
Subscribers: henryhu, llvm-commits, wengxt, broune, meheff, eliben
Differential Revision: http://reviews.llvm.org/D10695
llvm-svn: 240683
This previously caused miscompilations as a result of phi nodes receiving
undef incoming values from blocks dominated by such successors.
Differential Revision: http://reviews.llvm.org/D10726
llvm-svn: 240670
We performed a simple, but incomplete, intersection when it came time to
CSE instructions. It didn't handle, for example, the 'exact' flag.
This fixes PR23922.
llvm-svn: 240595
Reassociate mutated existing instructions in order to form negations
which would create additional reassociate opportunities.
This fixes PR23926.
llvm-svn: 240593
Summary:
Because LSR happens at a late stage where mul of a power of 2 is
typically canonicalized to shl, this canonicalization emits code that
can be better CSE'ed.
Test Plan:
Transforms/LoopStrengthReduce/shl.ll shows how this change makes GVN more
powerful. Fixes some existing tests due to this change.
Reviewers: sanjoy, majnemer, atrick
Reviewed By: majnemer, atrick
Subscribers: majnemer, llvm-commits
Differential Revision: http://reviews.llvm.org/D10448
llvm-svn: 240573
With option OptForSize enabled, the Loop Vectorizer is not supposed to
create tail loop. The condition checking that was invalid and was not
matching to the comment above.
Patch by Marianne Mailhot-Sarrasin.
llvm-svn: 240556
As with the previous patch, the goal is to turn the class into a general
loop-versioning class. This patch removes any references to loop
distribution.
llvm-svn: 240352
This avoids creating an unnecessary undefined reference on targets such as
NVPTX that require such references to be declared in asm output.
llvm-svn: 240321
The personality routine currently lives in the LandingPadInst.
This isn't desirable because:
- All LandingPadInsts in the same function must have the same
personality routine. This means that each LandingPadInst beyond the
first has an operand which produces no additional information.
- There is ongoing work to introduce EH IR constructs other than
LandingPadInst. Moving the personality routine off of any one
particular Instruction and onto the parent function seems a lot better
than have N different places a personality function can sneak onto an
exceptional function.
Differential Revision: http://reviews.llvm.org/D10429
llvm-svn: 239940
The original change broke clang side tests. I will be submitting those momentarily. This change includes post commit feedback on the original change from from Pete Cooper.
Original Submission comments:
If a parameter to a function is known non-null, use the existing parameter attributes to record that fact at the call site. This has no optimization benefit by itself - that I know of - but is an enabling change for http://reviews.llvm.org/D9129.
Differential Revision: http://reviews.llvm.org/D9132
llvm-svn: 239849
Any combination of +-inf/+-inf is NaN so it's already ignored with
nnan and we can skip checking for ninf. Also rephrase logic in comments
a bit.
llvm-svn: 239821
If a parameter to a function is known non-null, use the existing parameter attributes to record that fact at the call site. This has no optimization benefit by itself - that I know of - but is an enabling change for http://reviews.llvm.org/D9129.
Differential Revision: http://reviews.llvm.org/D9132
llvm-svn: 239795
This patch adds the safe stack instrumentation pass to LLVM, which separates
the program stack into a safe stack, which stores return addresses, register
spills, and local variables that are statically verified to be accessed
in a safe way, and the unsafe stack, which stores everything else. Such
separation makes it much harder for an attacker to corrupt objects on the
safe stack, including function pointers stored in spilled registers and
return addresses. You can find more information about the safe stack, as
well as other parts of or control-flow hijack protection technique in our
OSDI paper on code-pointer integrity (http://dslab.epfl.ch/pubs/cpi.pdf)
and our project website (http://levee.epfl.ch).
The overhead of our implementation of the safe stack is very close to zero
(0.01% on the Phoronix benchmarks). This is lower than the overhead of
stack cookies, which are supported by LLVM and are commonly used today,
yet the security guarantees of the safe stack are strictly stronger than
stack cookies. In some cases, the safe stack improves performance due to
better cache locality.
Our current implementation of the safe stack is stable and robust, we
used it to recompile multiple projects on Linux including Chromium, and
we also recompiled the entire FreeBSD user-space system and more than 100
packages. We ran unit tests on the FreeBSD system and many of the packages
and observed no errors caused by the safe stack. The safe stack is also fully
binary compatible with non-instrumented code and can be applied to parts of
a program selectively.
This patch is our implementation of the safe stack on top of LLVM. The
patches make the following changes:
- Add the safestack function attribute, similar to the ssp, sspstrong and
sspreq attributes.
- Add the SafeStack instrumentation pass that applies the safe stack to all
functions that have the safestack attribute. This pass moves all unsafe local
variables to the unsafe stack with a separate stack pointer, whereas all
safe variables remain on the regular stack that is managed by LLVM as usual.
- Invoke the pass as the last stage before code generation (at the same time
the existing cookie-based stack protector pass is invoked).
- Add unit tests for the safe stack.
Original patch by Volodymyr Kuznetsov and others at the Dependable Systems
Lab at EPFL; updates and upstreaming by myself.
Differential Revision: http://reviews.llvm.org/D6094
llvm-svn: 239761
It is valid for globals to be unnamed, but aliases must have a name. To avoid
creating invalid IR, we need to assign names to any aliases we create that
point to unnamed objects that have been moved into combined globals.
llvm-svn: 239590
Summary:
A side effect of this change is that it IRBuilder now automatically
created debug info locations for new instructions, which is the
same as debug location of insertion point. This is fine for the
functions in questions (GetStoreValueForLoad and
GetMemInstValueForLoad), as they are used in two situations:
* GVN::processLoad, which tries to eliminate a load. In this case
new instructions would have the same debug location as the load they
eventually replace;
* MaterializeAdjustedValue, which adds new instructions to the end
of the basic blocks, which could later be used to replace the load
definition. In this case we don't yet know the way the load would
be eventually replaced (either by assembling the precomputed values
via PHI, or by using them directly), so just using the basic block
strategy seems to be reasonable. There is also a special case
in the code that *would* adjust the location of the last
instruction replacing the load definition to the location of the
load.
Test Plan: regression test suite
Reviewers: echristo, dberlin, dblaikie
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10405
llvm-svn: 239585
This improves debug locations in passes that do a lot of basic block
transformations. Important case is LoopUnroll pass, the test for correct
debug locations accompanies this change.
Test Plan: regression test suite
Reviewers: dblaikie, sanjoy
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10367
llvm-svn: 239551
This only updates one of the uses. The other is used in cases
that may never touch memory, so I'm not sure why this is even
calling it. Perhaps there should be a new, similar hook for such
cases or pass -1 for unknown address space.
llvm-svn: 239540
If the first argument to a function is a 'this' argument and the second
has the sret attribute, the ArgumentPromotion pass may promote the 'this'
argument to more than one argument, violating the IR constraint that 'sret'
may only be applied to the first or second argument.
Although this IR constraint is arguably unnecessary, it highlighted the fact
that ArgPromotion does not need to preserve this attribute. Dropping the
attribute reduces register pressure in the backend by avoiding the register
copy required by sret. Because sret implies noalias, we also replace the
former with the latter.
Differential Revision: http://reviews.llvm.org/D10353
llvm-svn: 239488
Determining proper debug locations for instructions created in
PHITransAddr is tricky. We use a simple approach here and simply copy
debug locations from instructions computing load address to
"corresponding" instructions re-creating the address computation
in predecessor basic blocks.
This may not always be correct, given all the rearrangement and
simplification going on, and debug locations may jump around a lot,
as the basic blocks we copy locations between may be very far from
each other.
Still, this would work good in most simple cases (e.g. when chain
of address computing instruction is short, or our mapping turns out
to be 1-to-1), and we desire to have *some* reasonable debug locations
associated with newly inserted instructions.
See http://reviews.llvm.org/D10351 review thread for more details.
Test Plan: regression test suite
Reviewers: spatel, dblaikie
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10351
llvm-svn: 239479
Test Plan: regression test suite
Reviewers: eugenis, dblaikie
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10343
llvm-svn: 239438
We don't want to replace function A by Function B in one module and Function B
by Function A in another module.
If these functions are marked with linkonce_odr we would end up with a function
stub calling B in one module and a function stub calling A in another module. If
the linker decides to pick these two we will have two stubs calling each other.
rdar://21265586
llvm-svn: 239367
Interleaved memory accesses are grouped and vectorized into vector load/store and shufflevector.
E.g. for (i = 0; i < N; i+=2) {
a = A[i]; // load of even element
b = A[i+1]; // load of odd element
... // operations on a, b, c, d
A[i] = c; // store of even element
A[i+1] = d; // store of odd element
}
The loads of even and odd elements are identified as an interleave load group, which will be transfered into vectorized IRs like:
%wide.vec = load <8 x i32>, <8 x i32>* %ptr
%vec.even = shufflevector <8 x i32> %wide.vec, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
%vec.odd = shufflevector <8 x i32> %wide.vec, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
The stores of even and odd elements are identified as an interleave store group, which will be transfered into vectorized IRs like:
%interleaved.vec = shufflevector <4 x i32> %vec.even, %vec.odd, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
store <8 x i32> %interleaved.vec, <8 x i32>* %ptr
This optimization is currently disabled by defaut. To try it by adding '-enable-interleaved-mem-accesses=true'.
llvm-svn: 239291
Summary:
canUnrollCompletely takes `unsigned` values for `UnrolledCost` and
`RolledDynamicCost` but is passed in `uint64_t`s that are silently
truncated. Because of this, when `UnrolledSize` is a large integer
that has a small remainder with UINT32_MAX, LLVM tries to completely
unroll loops with high trip counts.
Reviewers: mzolotukhin, chandlerc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10293
llvm-svn: 239218
CVP wants to analyze the condition operand of a select along an edge.
It succeeds in getting back a Constant but not a ConstantInt. Instead,
it gets a ConstantExpr. It then assumes that the Constant must be equal
to false because it isn't equal to true.
Instead, perform an additional comparison.
This fixes PR23752.
llvm-svn: 239217
If we have (select a, b, c), it is sometimes valid to simplify this to a
single select operand. However, doing so is only valid if the
computation doesn't inject poison into the computation.
It might be helpful to consider the following example:
(select (icmp ne %i, INT_MAX), (add nsw %i, 1), INT_MIN)
The select is equivalent to (add %i, 1) but not (add nsw %i, 1).
Self hosting on x86_64 revealed that this occurs very, very rarely so
bailing out is hopefully pretty reasonable.
llvm-svn: 239215
This reverts commit r239141. This commit was an attempt to reintroduce
a previous patch that broke many self-hosting bots with clang timeouts,
but it still has slowdown issues, at least on ARM, increasing the
compilation time (stage 2, clang's) by 5x.
llvm-svn: 239175
The new naming is (to me) much easier to understand. Here is a summary
of the new state of the world:
- '*Threshold' is the threshold for full unrolling. It is measured
against the estimated unrolled cost as computed by getUserCost in TTI
(or CodeMetrics, etc). We will exceed this threshold when unrolling
loops where unrolling exposes a significant degree of simplification
of the logic within the loop.
- '*PercentDynamicCostSavedThreshold' is the percentage of the loop's
estimated dynamic execution cost which needs to be saved by unrolling
to apply a discount to the estimated unrolled cost.
- '*DynamicCostSavingsDiscount' is the discount applied to the estimated
unrolling cost when the dynamic savings are expected to be high.
When actually analyzing the loop, we now produce both an estimated
unrolled cost, and an estimated rolled cost. The rolled cost is notably
a dynamic estimate based on our analysis of the expected execution of
each iteration.
While we're still working to build up the infrastructure for making
these estimates, to me it is much more clear *how* to make them better
when they have reasonably descriptive names. For example, we may want to
apply estimated (from heuristics or profiles) dynamic execution weights
to the *dynamic* cost estimates. If we start doing that, we would also
need to track the static unrolled cost and the dynamic unrolled cost, as
only the latter could reasonably be weighted by profile information.
This patch is sadly not without functionality change for the new unroll
analysis logic. Buried in the heuristic management were several things
that surprised me. For example, we never subtracted the optimized
instruction count off when comparing against the unroll heursistics!
I don't know if this just got lost somewhere along the way or what, but
with the new accounting of things, this is much easier to keep track of
and we use the post-simplification cost estimate to compare to the
thresholds, and use the dynamic cost reduction ratio to select whether
we can exceed the baseline threshold.
The old values of these flags also don't necessarily make sense. My
impression is that none of these thresholds or discounts have been tuned
yet, and so they're just arbitrary placehold numbers. As such, I've not
bothered to adjust for the fact that this is now a discount and not
a tow-tier threshold model. We need to tune all these values once the
logic is ready to be enabled.
Differential Revision: http://reviews.llvm.org/D9966
llvm-svn: 239164
isInductionPHI wants to calculate the stride based on the pointee size.
However, this is not possible when the pointee is zero sized.
This fixes PR23763.
llvm-svn: 239143
I don't have the IR which is causing the build bot breakage but I can
postulate as to why they are timing out:
1. SimplifyWithOpReplaced was stripping flags from the simplified value.
2. visitSelectInstWithICmp was overriding SimplifyWithOpReplaced because
it's simplification wasn't correct.
3. InstCombine would revisit the add instruction and note that it can
rederive the flags.
4. By modifying the value, we chose to revisit instructions which reuse
the value. One of the instructions is the original select, causing
LLVM to never reach fixpoint.
Instead, strip the flags only when we are sure we are going to perform
the simplification.
llvm-svn: 239141
We cleverly handle cases where computation done in one argument of a select
instruction is suitable for the other operand, thus obviating the need
of the select and the comparison. However, the other operand cannot
have flags.
This fixes PR23757.
llvm-svn: 239115
We don't need to go through LSR to trigger this bug. Instead,
hand-craft a tricky GEP and get the constant folder to hack on it when
parsing the IR.
llvm-svn: 239017
Summary:
Once a gc.statepoint has been rewritten to relocate live references, the
SSA values represent physical pointers instead of logical references.
Logical dereferencability does not imply physical dereferencability and
after RewriteStatepointsForGC has run any attributes that imply
dereferencability of the logical references need to be stripped.
This current approach is conservative, and can be made more precise
later if needed. For starters, we need to strip dereferencable
attributes only from pointers that live in the GC address space.
Reviewers: reames, pgavlin
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10105
llvm-svn: 238883
Unreachable values may use themselves in strange ways due to their
dominance property. Attempting to translate through them can lead to
infinite recursion, crashing LLVM. Instead, claim that we weren't able
to translate the value.
This fixes PR23096.
llvm-svn: 238702
The patch evaluates the expansion cost of exitValue in indVarSimplify pass, and only does the rewriting when the expansion cost is low or loop can be deleted with the rewriting. It provides an option "-replexitval=" to control the default aggressiveness of the exitvalue rewriting. It also fixes some missing cases in SCEVExpander::isHighCostExpansionHelper to enhance the evaluation of SCEV expansion cost.
Differential Revision: http://reviews.llvm.org/D9800
llvm-svn: 238507
Currently we only fold a BitCast into a Load when the BitCast is its
only user.
Do the same for any no-op cast.
Differential Revision: http://reviews.llvm.org/D9152
llvm-svn: 238452
Canonicalizing 'x [+-] (-Constant * y)' is not a win if we don't *know*
we will open up CSE opportunities.
If the multiply was 'nsw', then negating 'y' requires us to clear the
'nsw' flag. If this is actually worth pursuing, it is probably more
appropriate to do so in GVN or EarlyCSE.
This fixes PR23675.
llvm-svn: 238397
Summary:
This patch made two improvements to NaryReassociate and the NVPTX pipeline
1. Run EarlyCSE/GVN after NaryReassociate to get rid of redundant common
expressions.
2. When adding an instruction to SeenExprs, maps both the SCEV before and after
reassociation to that instruction.
Test Plan: updated @reassociate_gep_nsw in nary-gep.ll
Reviewers: meheff, broune
Reviewed By: broune
Subscribers: dberlin, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D9947
llvm-svn: 238396
model the dense vector instruction bonuses.
Previously, this code really didn't effectively compute the density of
inlined vector instructions and apply the intended inliner bonus. It
would try to compute it repeatedly while analyzing the function and
didn't handle the case where future vector instructions would tip the
scales back towards the bonus.
Instead, speculatively apply all possible bonuses to the threshold
initially. Once we *know* that a certain bonus can not be applied,
subtract it. This should delay early bailout enough to get much more
consistent results without actually causing us to analyze huge swaths of
code. I expect some (hopefully mild) compile time hit here, and some
swings in performance, but this was definitely the intended behavior of
these bonuses.
This also dramatically simplifies the computation of the bonuses to not
interact with each other in confusing ways. The previous code didn't do
a good job of this and the values for bonuses may be surprising but are
at least now clearly written in the code.
Finally, fix code to be in line with comments and use zero as the
bailout condition.
Patch by Easwaran Raman, with some comment tweaks by me to try and
further clarify what is going on with this code.
http://reviews.llvm.org/D8267
llvm-svn: 238276
Summary:
In case of functions that have a pointer argument and only pass it to
each other, the function attributes pass deduces that the pointer should
get the readnone attribute, but fails to remove a readonly attribute
that may already have been present.
Reviewers: nlewycky
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9995
llvm-svn: 238152
This patch extends EarlyCSE to take advantage of the information that a controlling branch gives us about the value of a Value within this and dominated basic blocks. If the current block has a single predecessor with a controlling branch, we can infer what the branch condition must have been to execute this block. The actual change to support this is downright simple because EarlyCSE's existing scoped hash table logic deals with most of the complexity around merging.
The patch actually implements two optimizations.
1) The first is analogous to JumpThreading in that it enables EarlyCSE's CSE handling to fold branches which are exactly redundant due to a previous branch to branches on constants. (It doesn't actually replace the branch or change the CFG.) This is pretty clearly a win since it enables substantial CFG simplification before we start trying to inline.
2) The second is analogous to CVP in that it exploits the knowledge gained to replace dominated *uses* of the original value. EarlyCSE does not otherwise reason about specific uses, so this is the more arguable one. It does enable further simplication and constant folding within the rest of the visit by EarlyCSE.
In both cases, the added code only handles the easy dominance based case of each optimization. The general case is deferred to the existing passes.
Differential Revision: http://reviews.llvm.org/D9763
llvm-svn: 238071
InstCombine transforms A *nsw B +nsw A *nsw C to A *nsw (B + C).
This is incorrect -- e.g. if A = -1, B = 1, C = INT_SMAX. Then
nothing in the LHS overflows, but the multiplication in RHS overflows.
We need to first make sure that we won't multiple by INT_SMAX + 1.
Test case `add_of_mul` contributed by Sanjoy Das.
This fixes PR23635.
Differential Revision: http://reviews.llvm.org/D9629
llvm-svn: 238066
This change does a few things:
- Move some InstCombine transforms to InstSimplify
- Run SimplifyCall from within InstCombine::visitCallInst
- Teach InstSimplify to fold [us]mul_with_overflow(X, undef) to 0.
llvm-svn: 237995
PR23608 pointed out that using the preheader to gain a context instruction isn't always legal because a loop might not have a preheader. When looking into that, I realized that using the preheader to determine legality for sinking is questionable at best. Given no test covers that case and the original commit didn't seem to intend it, I restructured the code to only ask context sensative queries for hoising of loads and stores. This is effectively a partial revert of 237593.
llvm-svn: 237985
Summary:
x = &a[i];
y = &a[i + j];
=>
y = x + j;
along with some refactoring work such as extracting method
findClosestMatchingDominator.
Depends on D9786 which provides the ScalarEvolution::getGEPExpr interface.
Test Plan: nary-gep.ll
Reviewers: meheff, broune
Reviewed By: broune
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D9802
llvm-svn: 237971
On X86 (and similar OOO cores) unrolling is very limited, and even if the
runtime unrolling is otherwise profitable, the expense of a division to compute
the trip count could greatly outweigh the benefits. On the A2, we unroll a lot,
and the benefits of unrolling are more significant (seeing a 5x or 6x speedup
is not uncommon), so we're more able to tolerate the expense, on average, of a
division to compute the trip count.
llvm-svn: 237947
In effect a partial revert of r237858, which was a dumb shortcut.
Looking at the dependencies of the destination should be the proper
fix: if the new memset would depend on anything other than itself,
the transformation isn't correct.
llvm-svn: 237874
Fixes PR23599, another miscompile introduced by r235232: when there is
another dependency on the destination of the created memset (i.e., the
part of the original destination that the memcpy doesn't depend on)
between the memcpy and the original memset, we would insert the created
memset after the memcpy, and thus after the other dependency.
Instead, insert the created memset right after the old one.
llvm-svn: 237858
Make sure if we're truncating a constant that would then be sign extended
that the sign extension of the truncated constant is the same as the
original constant.
> Canonicalize min/max expressions correctly.
>
> This patch introduces a canonical form for min/max idioms where one operand
> is extended or truncated. This often happens when the other operand is a
> constant. For example:
>
> %1 = icmp slt i32 %a, i32 0
> %2 = sext i32 %a to i64
> %3 = select i1 %1, i64 %2, i64 0
>
> Would now be canonicalized into:
>
> %1 = icmp slt i32 %a, i32 0
> %2 = select i1 %1, i32 %a, i32 0
> %3 = sext i32 %2 to i64
>
> This builds upon a patch posted by David Majenemer
> (https://www.marc.info/?l=llvm-commits&m=143008038714141&w=2). That pass
> passively stopped instcombine from ruining canonical patterns. This
> patch additionally actively makes instcombine canonicalize too.
>
> Canonicalization of expressions involving a change in type from int->fp
> or fp->int are not yet implemented.
llvm-svn: 237821
This change adds a new GC strategy for supporting the CoreCLR runtime.
This strategy is currently identical to Statepoint-example GC,
but is necessary for several upcoming changes specific to CoreCLR, such as:
1. Base-pointers not explicitly reported for interior pointers
2. Different format for stack-map encoding
3. Location of Safe-point polls: polls are only needed before loop-back edges and before tail-calls (not needed at function-entry)
4. Runtime specific handshake between calls to managed/unmanaged functions.
llvm-svn: 237753
We were special casing a handful of intrinsics as not needing a safepoint before them. After running into another valid case - memset - I took a closer look and realized that almost no intrinsics need to have a safepoint poll before them. Restructure the code to make that apparent so that we stop hitting these bugs. The only intrinsics which need a safepoint poll before them are ones which can run arbitrary code.
llvm-svn: 237744
Summary:
Introduce dereferenceable, dereferenceable_or_null metadata for loads
with the same semantic as corresponding attributes.
This patch depends on http://reviews.llvm.org/D9253
Patch by Artur Pilipenko!
Reviewers: hfinkel, sanjoy, reames
Reviewed By: sanjoy, reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9365
llvm-svn: 237720
Summary: When PlaceSafepoints pass replaces old return result with gc_result from statepoint, it asserts that gc_result can not have preceding phis in its parent block. This is only true on invoke statepoint, which terminates the block and puts its result at the beginning of the normal successor block. Call statepoint does not terminate the block and thus its result is in the same block with it. There should be no restriction on whether there are phis or not.
Reviewers: reames, igor-laevsky
Reviewed By: igor-laevsky
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9803
llvm-svn: 237597
Summary:
Allow hoisting of loads from values marked with dereferenceable_or_null
attribute. For values marked with the attribute perform
context-sensitive analysis to determine whether it's known-non-null or
not.
Patch by Artur Pilipenko!
Reviewers: hfinkel, sanjoy, reames
Reviewed By: reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9253
llvm-svn: 237593
At the present time, we don't have a way to represent general dependency
relationships, so everything is represented using memory dependency. In order
to preserve the data dependency of a READ_REGISTER on WRITE_REGISTER, we need
to model WRITE_REGISTER as writing (which we had been doing) and model
READ_REGISTER as reading (which we had not been doing). Fix this, and also the
way that the chain operands were generated at the SDAG level.
Patch by Nicholas Paul Johnson, thanks! Test case by me.
llvm-svn: 237584
"Store to invariant address..." is moved as the last line. This is not
the prime result of the analysis. Plus it simplifies some of the tests.
llvm-svn: 237573
SimplifyDemandedBits was "simplifying" a constant by removing just sign bits.
This caused a canonicalization race between different parts of instcombine.
Fix and regression test added - third time lucky?
llvm-svn: 237539
The AArch64 LNT bot is unhappy - I've found that the problem is in
SimpliftDemandedBits, but that's going to require another code review
so reverting in the meantime.
llvm-svn: 237528
... I'd copied the check-next lines from a previous test so they were
slightly wrong, and had managed to test the wrong source tree. D'oh!
llvm-svn: 237521
The test timeouts were due to instcombine fighting itself. Regression test added.
Original log message:
Canonicalize min/max expressions correctly.
This patch introduces a canonical form for min/max idioms where one operand
is extended or truncated. This often happens when the other operand is a
constant. For example:
%1 = icmp slt i32 %a, i32 0
%2 = sext i32 %a to i64
%3 = select i1 %1, i64 %2, i64 0
Would now be canonicalized into:
%1 = icmp slt i32 %a, i32 0
%2 = select i1 %1, i32 %a, i32 0
%3 = sext i32 %2 to i64
This builds upon a patch posted by David Majenemer
(https://www.marc.info/?l=llvm-commits&m=143008038714141&w=2). That pass
passively stopped instcombine from ruining canonical patterns. This
patch additionally actively makes instcombine canonicalize too.
Canonicalization of expressions involving a change in type from int->fp
or fp->int are not yet implemented.
llvm-svn: 237520
There's no point in copying around constants, so, when all else fails,
we can still transform memcpy of memset into two independent memsets.
To quote the example, we can turn:
memset(dst1, c, dst1_size);
memcpy(dst2, dst1, dst2_size);
into:
memset(dst1, c, dst1_size);
memset(dst2, c, dst2_size);
When dst2_size <= dst1_size.
Like r235232 for copy constructors, this can occur in move constructors.
Differential Revision: http://reviews.llvm.org/D9682
llvm-svn: 237506
Summary:
This is a pass for speculative execution of instructions for simple if-then (triangle) control flow. It's aimed at GPUs, but could perhaps be used in other contexts. Enabling this pass gives us a 1.0% geomean improvement on Google benchmark suites, with one benchmark improving 33%.
Credit goes to Jingyue Wu for writing an earlier version of this pass.
Patched by Bjarke Roune.
Test Plan:
This patch adds a set of tests in test/Transforms/SpeculativeExecution/spec.ll
The pass is controlled by a flag which defaults to having the pass not run.
Reviewers: eliben, dberlin, meheff, jingyue, hfinkel
Reviewed By: jingyue, hfinkel
Subscribers: majnemer, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D9360
llvm-svn: 237459
This reverts r237453 - it was causing timeouts on some bots. Reverting
while I investigate (it's probably InstCombine fighting itself...)
llvm-svn: 237458
Summary:
Consider (B | i) * S as (B + i) * S if B and i have no bits set in
common.
Test Plan: @or in slsr-mul.ll
Reviewers: broune, meheff
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9788
llvm-svn: 237456
This patch introduces a canonical form for min/max idioms where one operand
is extended or truncated. This often happens when the other operand is a
constant. For example:
%1 = icmp slt i32 %a, i32 0
%2 = sext i32 %a to i64
%3 = select i1 %1, i64 %2, i64 0
Would now be canonicalized into:
%1 = icmp slt i32 %a, i32 0
%2 = select i1 %1, i32 %a, i32 0
%3 = sext i32 %2 to i64
This builds upon a patch posted by David Majenemer
(https://www.marc.info/?l=llvm-commits&m=143008038714141&w=2). That pass
passively stopped instcombine from ruining canonical patterns. This
patch additionally actively makes instcombine canonicalize too.
Canonicalization of expressions involving a change in type from int->fp
or fp->int are not yet implemented.
llvm-svn: 237453
This has caused some local failures. Updating the test case to be more
like the majority of the similar test cases.
Committing on behalf of Hubert Tong (hstong@ca.ibm.com).
llvm-svn: 237449
Transfer the calling convention from the invoke being replaced by
PlaceStatepoints to the new invoke to gc.statepoint created. Add a test
case that would have caught this issue.
llvm-svn: 237414
rL236672 would generate all invoke statepoints with deopt args set to a
list containing the single element "0", instead of an empty list.
Also add a test case that would have caught this.
llvm-svn: 237413
Summary:
Extract method haveNoCommonBitsSet so that we don't have to duplicate this logic in
InstCombine and SeparateConstOffsetFromGEP.
This patch also makes SeparateConstOffsetFromGEP more precise by passing
DominatorTree to computeKnownBits.
Test Plan: value-tracking-domtree.ll that tests ValueTracking indeed leverages dominating conditions
Reviewers: broune, meheff, majnemer
Reviewed By: majnemer
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D9734
llvm-svn: 237407
This is to cleanup some redundency generated by LoopUnroll pass. Such redundency may not be cleaned up by existing passes after LoopUnroll.
Differential Revision: http://reviews.llvm.org/D9777
llvm-svn: 237395
Function 'ConstantFoldScalarCall' (in ConstantFolding.cpp) works under the
wrong assumption that a call to 'convert.from.fp16' returns a value of
type 'float'.
However, intrinsic 'convert.from.fp16' can be overloaded; for example, we
can call 'convert.from.fp16.f64' to convert from half to double; etc.
Before this patch, the following example would have triggered an assertion
failure in opt (with -constprop):
```
define double @foo() {
entry:
%0 = call double @llvm.convert.from.fp16.f64(i16 0)
ret double %0
}
```
This patch fixes the problem in ConstantFolding.cpp. When folding a call to
convert.from.fp16, we perform a different kind of conversion based on the call
return type.
Added test 'Transform/ConstProp/convert-from-fp16.ll'.
Differential Revision: http://reviews.llvm.org/D9771
llvm-svn: 237377
Summary:
This implements the initial version as was proposed earlier this year
(http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-January/080462.html).
Since then Loop Access Analysis was split out from the Loop Vectorizer
and was made into a separate analysis pass. Loop Distribution becomes
the second user of this analysis.
The pass is off by default and can be enabled
with -enable-loop-distribution. There is currently no notion of
profitability; if there is a loop with dependence cycles, the pass will
try to split them off from other memory operations into a separate loop.
I decided to remove the control-dependence calculation from this first
version. This and the issues with the PDT are actively discussed so it
probably makes sense to treat it separately. Right now I just mark all
terminator instruction required which keeps identical CFGs for each
distributed loop. This seems to be working pretty well for 456.hmmer
where even though there is an empty if-then block in the distributed
loop initially, it gets completely removed.
The pass keeps DominatorTree and LoopInfo updated. I've tested this
with -loop-distribute-verify with the testsuite where we distribute ~90
loops. SimplifyLoop is violated in some cases and I have a FIXME
covering this.
Reviewers: hfinkel, nadav, aschwaighofer
Reviewed By: aschwaighofer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8831
llvm-svn: 237358
Summary:
This patch teaches the PlaceSafepoints pass about two `CallSite`
function attributes:
* "statepoint-id": if the string value of this attribute can be parsed
as an integer, then it is propagated to the ID parameter of the
statepoint created.
* "statepoint-num-patch-bytes": if the string value of this attribute
can be parsed as an integer, then it is propagated to the `num patch
bytes` parameter of the statepoint created.
This change intentionally does not assert on a malformed value for these
attributes, given that they're not "official" attributes.
Reviewers: reames, pgavlin
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9735
llvm-svn: 237286
Avoid running forever by checking we are not reassociating an expression into
the same form.
Tested with @avoid_infinite_loops in nary-add.ll
llvm-svn: 237269
This patch uses the new function profile metadata "function_entry_count"
to annotate entry counts from sample profiles.
In a sampling profile, the total samples collected at the function entry
are an approximation for the number of times that function was invoked.
llvm-svn: 237265
Summary:
This change adds two new parameters to the statepoint intrinsic, `i64 id`
and `i32 num_patch_bytes`. `id` gets propagated to the ID field
in the generated StackMap section. If the `num_patch_bytes` is
non-zero then the statepoint is lowered to `num_patch_bytes` bytes of
nops instead of a call (the spill and reload code remains unchanged).
A non-zero `num_patch_bytes` is useful in situations where a language
runtime requires complete control over how a call is lowered.
This change brings statepoints one step closer to patchpoints. With
some additional work (that is not part of this patch) it should be
possible to get rid of `TargetOpcode::STATEPOINT` altogether.
PlaceSafepoints generates `statepoint` wrappers with `id` set to
`0xABCDEF00` (the old default value for the ID reported in the stackmap)
and `num_patch_bytes` set to `0`. This can be made more sophisticated
later.
Reviewers: reames, pgavlin, swaroop.sridhar, AndyAyers
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9546
llvm-svn: 237214
Summary:
If the branch that leads to the PHI node and the Select instruction
depend on correlated conditions, we might be able to directly use the
corresponding value from the Select instruction as the incoming value
for the PHI node, allowing later removal of the select instruction.
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9051
llvm-svn: 237201
When relocating a pointer, we need to determine a base pointer for the derived pointer being relocated. We have limited support for handling a pointer extracted from a vector; the current code only handled the case where the entire vector was known to contain base pointers. This patch extends the reasoning to handle chains of insertelements where the indices are constants. This case turns out to be fairly common in vectorized code. We can now handle vectors which contains mixtures of base and derived pointers provided the insertelements use constant indices.
Note that this doesn't solve the general problem. To handle variable indexed insertelements, we'd need to scalarize and introduce conditional branching based on the index. Alternatively, we could eagerly scalarize, but the code structure doesn't currently make either fix easy. The patch also doesn't handle shufflevector or other vector manipulation for much the same reasons. I plan to defer this work until I have a motivating test case.
Differential Revision: http://reviews.llvm.org/D9676
llvm-svn: 237200
As a step towards getting rid of internal pass manager hack entirely, remove the need for loop simplify to run in the inner pass manager. The new code does produce slightly different loop structures, so this isn't technically NFC.
Differential Revision: http://reviews.llvm.org/D9585
llvm-svn: 237172
Summary:
This patch reimplements heuristic that tries to estimate optimization beneftis
from complete loop unrolling.
In this patch I kept the minimal changes - e.g. I removed code handling
branches and folding compares. That's a promising area, but now there
are too many questions to discuss before we can enable it.
Test Plan: Tests are included in the patch.
Reviewers: hfinkel, chandlerc
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8816
llvm-svn: 237156
This fixes another miscompile introduced by r235232: when there was a
dependency on the memcpy destination other than the memset, we would
ignore it, because we only looked at the source dependency.
It was a mistake to use SrcDepInfo. Instead, just use DepInfo.
llvm-svn: 237066
Summary:
In RewriteStatepointsForGC pass, we create a gc_relocate intrinsic for
each relocated pointer, and the gc_relocate has the same type with the
pointer. During the creation of gc_relocate intrinsic, llvm requires to
mangle its type. However, llvm does not support mangling of all possible
types. RewriteStatepointsForGC will hit an assertion failure when it
tries to create a gc_relocate for pointer to vector of pointers because
mangling for vector of pointers is not supported.
This patch changes the way RewriteStatepointsForGC pass creates
gc_relocate. For each relocated pointer, we erase the type of pointers
and create an unified gc_relocate of type i8 addrspace(1)*. Then a
bitcast is inserted to convert the gc_relocate to the correct type. In
this way, gc_relocate does not need to deal with different types of
pointers and the unsupported type mangling is no longer a problem. This
change would also ease further merge when LLVM erases types of pointers
and introduces an unified pointer type.
Some minor changes are also introduced to gc_relocate related part in
InstCombineCalls, CodeGenPrepare, and Verifier accordingly.
Patch by Chen Li!
Reviewers: reames, AndyAyers, sanjoy
Reviewed By: sanjoy
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9592
llvm-svn: 237009
The QPX single-precision load/store intrinsics have implied
truncation/extension from/to the declared value type of <4 x double> to the
memory type of <4 x float>. When we can prove the alignment of the pointer
argument, and thus replace the intrinsic with a regular load or store, we need
to load or store the correct data type (<4 x float>) instead of (<4 x double>).
llvm-svn: 236973
This changes the shape of the statepoint intrinsic from:
@llvm.experimental.gc.statepoint(anyptr target, i32 # call args, i32 unused, ...call args, i32 # deopt args, ...deopt args, ...gc args)
to:
@llvm.experimental.gc.statepoint(anyptr target, i32 # call args, i32 flags, ...call args, i32 # transition args, ...transition args, i32 # deopt args, ...deopt args, ...gc args)
This extension offers the backend the opportunity to insert (somewhat) arbitrary code to manage the transition from GC-aware code to code that is not GC-aware and back.
In order to support the injection of transition code, this extension wraps the STATEPOINT ISD node generated by the usual lowering lowering with two additional nodes: GC_TRANSITION_START and GC_TRANSITION_END. The transition arguments that were passed passed to the intrinsic (if any) are lowered and provided as operands to these nodes and may be used by the backend during code generation.
Eventually, the lowering of the GC_TRANSITION_{START,END} nodes should be informed by the GC strategy in use for the function containing the intrinsic call; for now, these nodes are instead replaced with no-ops.
Differential Revision: http://reviews.llvm.org/D9501
llvm-svn: 236888
Summary:
I noticed this bug when deubging a WIP on LSR. I wonder whether and how we
should add a regression test for this.
Test Plan: no tests failed.
Reviewers: atrick
Subscribers: hfinkel, llvm-commits
Differential Revision: http://reviews.llvm.org/D9536
llvm-svn: 236887
Summary:
One step further getting aggregate loads and store being optimized
properly. This will only handle struct with one element at this point.
Test Plan: Added unit tests for the new supported cases.
Reviewers: chandlerc, joker-eph, joker.eph, majnemer
Reviewed By: majnemer
Subscribers: pete, llvm-commits
Differential Revision: http://reviews.llvm.org/D8339
Patch by Amaury Sechet.
From: Amaury Sechet <amaury@fb.com>
llvm-svn: 236695
If we have recognized that a conditional is constant at a particular location in the code (while trying to decide if we can simplify a conditional branch), we can eagerly replace that condition with a constant if it's definition is post dominated by the branch in question.
In practice, this ends up being a compile time savings at most. JumpThreading would have visited each using branch anyways. CVP would have visited the cmp itself again. Unless LVI gives up early, we shouldn't gain any addition power by doing this transformation early. What we do gain is simplicity and compile time.
Differential Revision: http://reviews.llvm.org/D9312
llvm-svn: 236684
options.
This commit fixes a bug in llc and opt where "-mcpu" and "-mattr" wouldn't
override function attributes "-target-cpu" and "-target-features" in the IR.
Differential Revision: http://reviews.llvm.org/D9537
llvm-svn: 236677
Renames the original CreateGCStatepoint to CreateGCStatepointCall, and
moves invoke creating functionality from PlaceSafepoints.cpp to
IRBuilder.cpp.
This changes the labels generated for PlaceSafepoints/invokes.ll so use
a regex there to make the basic block labels more resilient.
llvm-svn: 236672
Summary:
When computing branch weights in BPI, we used to disallow branches with
weight 0. This is a minor nuisance, because a branch with weight 0 is
different to "don't have information". In the context of
instrumentation, it may mean "never executed", in the context of
sampling, it means "never or seldom executed".
In allowing 0 weight branches, I ran into issues with the switch
expansion code in selection DAG. It is currently hardwired to not handle
branches with weight 0. To maintain the current behaviour, I changed it
to use 1 when it finds 0, but perhaps the algorithm needs changes to
tolerate branches with weight zero.
Reviewers: hansw
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9533
llvm-svn: 236617
The patch disabled unrolling in loop vectorization pass when VF==1 on x86 architecture,
by setting MaxInterleaveFactor to 1. Unrolling in loop vectorization pass may introduce
the cost of overflow check, memory boundary check and extra prologue/epilogue code when
regular unroller will unroll the loop another time. Disable it when VF==1 remove the
unnecessary cost on x86. The same can be done for other platforms after verifying
interleaving/memory bound checking to be not perf critical on those platforms.
Differential Revision: http://reviews.llvm.org/D9515
llvm-svn: 236613
COMDAT groups which have become rendered unused because of inline are
discardable if we can prove that we've made the group empty.
This fixes PR22285.
llvm-svn: 236539
It got this in some cases (if one of them was an identified object), but not in all cases.
This caused stores to undef to block load-forwarding in some cases, etc.
Added test to Transforms/GVN to verify optimization occurs as expected.
llvm-svn: 236511
When optimizing demanded bits of the operands of an Add we have to
remove the nsw/nuw flags as we have no guarantee anymore that we don't
wrap. This is legal here because the top bit is not demanded. In fact
this operaion was already performed but missed in the case of an Add
with a constant on the right side. To fix this this patch refactors the
code to unify the code paths in SimplifyDemandedUseBits() handling of
Add/Sub:
- The transformation of Add->Or is removed from the simplify demand
code because the equivalent transformation exists in
InstCombiner::visitAdd()
- KnownOnes/KnownZero are not adjusted for Add x, C anymore as
computeKnownBits() already performs these computations.
- The simplification of the operands is unified. In this new version
constant on the right side of a Sub are shrunk now as I could not find
a reason why not to do so.
- The special case for clearing nsw/nuw in ShrinkDemandedConstant() is
not necessary anymore as the caller does that already.
Differential Revision: http://reviews.llvm.org/D9415
llvm-svn: 236269
Summary:
Optimizing these well are especially interesting for IRCE since it
"clamps" values by generating this sort of pattern through SCEV
expressions.
Depends on D9352.
Reviewers: majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9353
llvm-svn: 236203
Summary:
After this change `MatchSelectPattern` recognizes the following form
of SMIN:
Y >s C ? ~Y : ~C == ~Y <s ~C ? ~Y : ~C = SMIN(~Y, ~C)
Reviewers: majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9352
llvm-svn: 236202
Finish off PR23080 by renaming the debug info IR constructs from `MD*`
to `DI*`. The last of the `DIDescriptor` classes were deleted in
r235356, and the last of the related typedefs removed in r235413, so
this has all baked for about a week.
Note: If you have out-of-tree code (like a frontend), I recommend that
you get everything compiling and tests passing with the *previous*
commit before updating to this one. It'll be easier to keep track of
what code is using the `DIDescriptor` hierarchy and what you've already
updated, and I think you're extremely unlikely to insert bugs. YMMV of
course.
Back to *this* commit: I did this using the rename-md-di-nodes.sh
upgrade script I've attached to PR23080 (both code and testcases) and
filtered through clang-format-diff.py. I edited the tests for
test/Assembler/invalid-generic-debug-node-*.ll by hand since the columns
were off-by-three. It should work on your out-of-tree testcases (and
code, if you've followed the advice in the previous paragraph).
Some of the tests are in badly named files now (e.g.,
test/Assembler/invalid-mdcompositetype-missing-tag.ll should be
'dicompositetype'); I'll come back and move the files in a follow-up
commit.
llvm-svn: 236120
Summary:
This patch adds constant folding of insertelement instruction to undef value when index operand is constant and is not less than vector size or is undef.
InstCombine does not support this case, but I'm happy to add it there also if this change is accepted.
Test Plan: Unittests and regression tests for ConstProp pass.
Reviewers: majnemer
Reviewed By: majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9287
llvm-svn: 235854
There can be various constant pointers in the IR which do not get relocated at a safepoint. One example is the address of a global variable. Another example is a pointer created via inttoptr. Note that the optimizer itself likes to create such inttoptrs when locally propagating constants through dynamically dead code.
To deal with this, we need to exclude uses of constants from contributing to the liveness of a safepoint which might reach that use. At some later date, it might be worth exploring what could be done to support the relocation of various special types of "constants", but that's future work.
Differential Revision: http://reviews.llvm.org/D9236
llvm-svn: 235821
llvm.frameescape() intrinsic is not a real call. The intrinsic can only exist in the entry block. Inserting a gc.statepoint() before llvm.frameescape() may split the entry block, and push the intrinsic out of the entry block.
Patch by: Swaroop.Sridhar@microsoft.com
Differential Revision: http://reviews.llvm.org/D8910
llvm-svn: 235820
This is a follow-on to D8833 (insertps optimization when the zero mask is not used).
In this patch, we check for the case where the zmask is used, but both input vectors
to the insertps intrinsic are the same operand or the zmask overrides the destination
lane. This lets us replace the 2nd shuffle input operand with the zero vector.
Differential Revision: http://reviews.llvm.org/D9257
llvm-svn: 235810
When using bit tests for hole checks, we call AddPredecessorToBlock to give the
phi node a value from the bit test block. This would break if we've
previously called removePredecessor on the default destination because the
switch is fully covered.
Test case by Mark Lacey.
llvm-svn: 235771
Same as r235145 for the call instruction - the justification, tradeoffs,
etc are all the same. The conversion script worked the same without any
false negatives (after replacing 'call' with 'invoke').
llvm-svn: 235755
Summary:
We pick this order because SeparateConstOffsetFromGEP may create more
opportunities for SLSR.
Test Plan:
reassociate-geps-and-slsr.ll
no performance regression on internal benchmarks
Reviewers: meheff
Subscribers: llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D9230
llvm-svn: 235632
Only clear out the NSW/NUW flags if we are optimizing 'add'/'sub' while
taking advantage that the sign bit is not set. We do this optimization
to further shrink the mask but shrinking the mask isn't NSW/NUW
preserving in this case.
llvm-svn: 235558
An nsw/nuw operation relies on the values feeding into it to not
overflow if 'poison' is not to be produced. This means that
optimizations which make modifications to the bottom of a chain (like
SimplifyDemandedBits) must strip out nsw/nuw if they cannot ensure that
they will be preserved.
This fixes PR23309.
llvm-svn: 235544
https://llvm.org/bugs/show_bug.cgi?id=23163.
Gep merging sometimes behaves like a reverse CSE/LICM optimization,
which has negative impact on performance. In this patch we restrict
gep merging to happen only when the indexes to be merged are both consts,
which ensures such merge is always beneficial.
The patch makes gep merging only happen in very restrictive cases.
It is possible that some analysis/optimization passes rely on the merged
geps to get better result, and we havn't notice them yet. We will be ready
to further improve it once we see the cases.
Differential Revision: http://reviews.llvm.org/D8911
llvm-svn: 235455
https://llvm.org/bugs/show_bug.cgi?id=23163.
Gep merging sometimes behaves like a reverse CSE/LICM optimizations,
which has negative impact on performance. In this patch we restrict
gep merging to happen only when the indexes to be merged are both consts,
which ensures such merge is always beneficial.
The patch makes gep merging only happen in very restrictive cases.
It is possible that some analysis/optimization passes rely on the merged
geps to get better result, and we havn't notice them yet. We will be ready
to further improve it once we see the cases.
Differential Revision: http://reviews.llvm.org/D9007
llvm-svn: 235451
MemIntrinsic::getDest() looks through pointer casts, and using it
directly when building the new GEP+memset results in stuff like:
%0 = getelementptr i64* %p, i32 16
%1 = bitcast i64* %0 to i8*
call ..memset(i8* %1, ...)
instead of the correct:
%0 = bitcast i64* %p to i8*
%1 = getelementptr i8* %0, i32 16
call ..memset(i8* %1, ...)
Instead, use getRawDest, which just gives you the i8* value.
While there, use the memcpy's dest, as it's live anyway.
In most cases, when the optimization triggers, the memset and memcpy
sizes are the same, so the built memset is 0-sized and eliminated.
The problem occurs when they're different.
Fixes a regression caused by r235232: PR23300.
llvm-svn: 235419
Summary:
After we rewrite a candidate, the instructions used by the old form may
become unused. This patch cleans up these unused instructions so that we
needn't run DCE after SLSR.
Test Plan: removed -dce in all the SLSR tests
Reviewers: broune, meheff
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9101
llvm-svn: 235410
Summary: so that we needn't run DCE after this pass.
Test Plan: removed -dce from the commandline in split-gep.ll and split-gep-and-gvn.ll
Reviewers: meheff
Subscribers: llvm-commits, HaoLiu, hfinkel, jholewinski
Differential Revision: http://reviews.llvm.org/D9096
llvm-svn: 235409
This commit fixes the code which adds lifetime markers in InlineFunction to skip
zero-sized allocas instead of asserting on them.
rdar://problem/20531155
llvm-svn: 235312
Harden r235258 to support any integer bitwidth. The quick glance at
the reference made me think only i32 and i64 were valid types, but
they're not special, so any overload is legal.
Thanks to David Majnemer for noticing!
llvm-svn: 235261
Followup to r235232, which caused PR23278.
We can't assume the memset and memcpy sizes have the same type, as
nothing in the language reference prevents that.
Instead, zext both to i64 if they disagree.
While there, robustify tests by using i8 %c rather than i8 0 for the
memset character.
llvm-svn: 235258
A common idiom in some code is to do the following:
memset(dst, 0, dst_size);
memcpy(dst, src, src_size);
Some of the memset is redundant; instead, we can do:
memcpy(dst, src, src_size);
memset(dst + src_size, 0,
dst_size <= src_size ? 0 : dst_size - src_size);
Original patch by: Joel Jones
Differential Revision: http://reviews.llvm.org/D498
llvm-svn: 235232
Summary:
An alternative is to use a worklist approach. However, that approach
would break the traversing order so that we couldn't lookup SeenExprs
efficiently. I don't see a clear winner here, so I picked the easier approach.
Along with two minor improvements:
1. preserves ScalarEvolution by forgetting instructions replaced
2. removes dead code locally avoiding the need of running DCE afterwards
Test Plan: add to slsr-add.ll a test that requires multiple iterations
Reviewers: broune, dberlin, atrick, meheff
Reviewed By: atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9058
llvm-svn: 235151
See r230786 and r230794 for similar changes to gep and load
respectively.
Call is a bit different because it often doesn't have a single explicit
type - usually the type is deduced from the arguments, and just the
return type is explicit. In those cases there's no need to change the
IR.
When that's not the case, the IR usually contains the pointer type of
the first operand - but since typed pointers are going away, that
representation is insufficient so I'm just stripping the "pointerness"
of the explicit type away.
This does make the IR a bit weird - it /sort of/ reads like the type of
the first operand: "call void () %x(" but %x is actually of type "void
()*" and will eventually be just of type "ptr". But this seems not too
bad and I don't think it would benefit from repeating the type
("void (), void () * %x(" and then eventually "void (), ptr %x(") as has
been done with gep and load.
This also has a side benefit: since the explicit type is no longer a
pointer, there's no ambiguity between an explicit type and a function
that returns a function pointer. Previously this case needed an explicit
type (eg: a function returning a void() function was written as
"call void () () * @x(" rather than "call void () * @x(" because of the
ambiguity between a function returning a pointer to a void() function
and a function returning void).
No ambiguity means even function pointer return types can just be
written alone, without writing the whole function's type.
This leaves /only/ the varargs case where the explicit type is required.
Given the special type syntax in call instructions, the regex-fu used
for migration was a bit more involved in its own unique way (as every
one of these is) so here it is. Use it in conjunction with the apply.sh
script and associated find/xargs commands I've provided in rr230786 to
migrate your out of tree tests. Do let me know if any of this doesn't
cover your cases & we can iterate on a more general script/regexes to
help others with out of tree tests.
About 9 test cases couldn't be automatically migrated - half of those
were functions returning function pointers, where I just had to manually
delete the function argument types now that we didn't need an explicit
function type there. The other half were typedefs of function types used
in calls - just had to manually drop the * from those.
import fileinput
import sys
import re
pat = re.compile(r'((?:=|:|^|\s)call\s(?:[^@]*?))(\s*$|\s*(?:(?:\[\[[a-zA-Z0-9_]+\]\]|[@%](?:(")?[\\\?@a-zA-Z0-9_.]*?(?(3)"|)|{{.*}}))(?:\(|$)|undef|inttoptr|bitcast|null|asm).*$)')
addrspace_end = re.compile(r"addrspace\(\d+\)\s*\*$")
func_end = re.compile("(?:void.*|\)\s*)\*$")
def conv(match, line):
if not match or re.search(addrspace_end, match.group(1)) or not re.search(func_end, match.group(1)):
return line
return line[:match.start()] + match.group(1)[:match.group(1).rfind('*')].rstrip() + match.group(2) + line[match.end():]
for line in sys.stdin:
sys.stdout.write(conv(re.search(pat, line), line))
llvm-svn: 235145
Summary:
This fixes a left-over efficiency issue in D8950.
As Andrew and Daniel suggested, we can store the candidates in a stack
and pop the top element when it does not dominate the current
instruction. This reduces the worst-case time complexity to O(n).
Test Plan: a new test in nary-add.ll that exercises this optimization.
Reviewers: broune, dberlin, meheff, atrick
Reviewed By: atrick
Subscribers: llvm-commits, sanjoy
Differential Revision: http://reviews.llvm.org/D9055
llvm-svn: 235129
This is very similar to D8486 / r232852 (vperm2). If we treat insertps intrinsics
as shufflevectors, we can optimize them better.
I've left all but the full zero case of the zero mask variants out of this patch.
I don't think those can be converted into a single shuffle in all cases, but I'd
be happy to be proven wrong as I was for vperm2f128.
Either way, we'd need to support whatever sequence we come up with for those cases
in the backend before converting them here.
Differential Revision: http://reviews.llvm.org/D8833
llvm-svn: 235124
Remove 'inlinedAt:' from MDLocalVariable. Besides saving some memory
(variables with it seem to be single largest `Metadata` contributer to
memory usage right now in -g -flto builds), this stops optimization and
backend passes from having to change local variables.
The 'inlinedAt:' field was used by the backend in two ways:
1. To tell the backend whether and into what a variable was inlined.
2. To create a unique id for each inlined variable.
Instead, rely on the 'inlinedAt:' field of the intrinsic's `!dbg`
attachment, and change the DWARF backend to use a typedef called
`InlinedVariable` which is `std::pair<MDLocalVariable*, MDLocation*>`.
This `DebugLoc` is already passed reliably through the backend (as
verified by r234021).
This commit removes the check from r234021, but I added a new check
(that will survive) in r235048, and changed the `DIBuilder` API in
r235041 to require a `!dbg` attachment whose 'scope:` is in the same
`MDSubprogram` as the variable's.
If this breaks your out-of-tree testcases, perhaps the script I used
(mdlocalvariable-drop-inlinedat.sh) will help; I'll attach it to PR22778
in a moment.
llvm-svn: 235050
Add missing `!dbg` attachments to `@llvm.dbg.*` intrinsics. I updated
these using a script (add-dbg-to-intrinsics.sh) that I'll attach to
PR22778 for posterity.
llvm-svn: 235040
Summary:
With this patch, SLSR may rewrite
S1: X = B + i * S
S2: Y = B + i' * S
to
S2: Y = X + (i' - i) * S
A secondary improvement: if (i' - i) is a power of 2, emit Y as X + (S << log(i' - i)). (S << log(i' -i)) is in a canonical form and thus more likely GVN'ed than (i' - i) * S.
Test Plan: slsr-add.ll
Reviewers: hfinkel, sanjoy, meheff, broune, eliben
Reviewed By: eliben
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8983
llvm-svn: 235019
Inlining such intrinsics is very difficult, since you need to
simultaneously transform many calls to llvm.framerecover and potentially
duplicate the functions containing them. Normally this intrinsic isn't
added until EH preparation, which is part of the backend pass pipeline
after inlining. However, if it were to get fed through the inliner,
this change will ensure that it doesn't break the code.
llvm-svn: 234937
Summary:
This transformation reassociates a n-ary add so that the add can partially reuse
existing instructions. For example, this pass can simplify
void foo(int a, int b) {
bar(a + b);
bar((a + 2) + b);
}
to
void foo(int a, int b) {
int t = a + b;
bar(t);
bar(t + 2);
}
saving one add instruction.
Fixes PR22357 (https://llvm.org/bugs/show_bug.cgi?id=22357).
Test Plan: nary-add.ll
Reviewers: broune, dberlin, hfinkel, meheff, sanjoy, atrick
Reviewed By: sanjoy, atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8950
llvm-svn: 234855
Summary:
Runtime unrolling of loops needs to emit an expression to compute the
loop's runtime trip-count. Avoid runtime unrolling if this computation
will be expensive.
Depends on D8993.
Reviewers: atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8994
llvm-svn: 234846
Summary:
Teach `isHighCostExpansion` to consider divisions by power-of-two
constants as cheap and add a test case. This change is needed for a new
user of `isHighCostExpansion` that will be added in a subsequent change.
Depends on D8995.
Reviewers: atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8993
llvm-svn: 234845
Sanjoy Das