For the store of a wide value merged from a pair of values, especially int-fp pair,
sometimes it is more efficent to split it into separate narrow stores, which can
remove the bitwise instructions or sink them to colder places.
Now the feature is only enabled on x86 target, and only store of int-fp pair is
splitted. It is possible that the application scope gets extended with perf evidence
support in the future.
Differential Revision: https://reviews.llvm.org/D22840
llvm-svn: 280505
The motivating case occurs with SSE/AVX scalar intrinsics, so this is a first step towards
shrinking that to a single shufflevector.
Note that the transform is intentionally limited to shuffles that are equivalent to vector
selects to avoid creating arbitrary shuffle masks that may not lower well.
This should solve PR29126:
https://llvm.org/bugs/show_bug.cgi?id=29126
Differential Revision: https://reviews.llvm.org/D23886
llvm-svn: 280504
For uniform instructions, we're only required to generate a scalar value for
the first vector lane of each unroll iteration. Thus, if we have a reverse
interleaved group, computing the member index off the scalar GEP corresponding
to the last vector lane of its pointer operand technically makes the GEP
non-uniform. We should compute the member index off the first scalar GEP
instead.
I've added the updated member index computation to the existing reverse
interleaved group test.
llvm-svn: 280497
This patch fixes a crash caused by an incorrect folding of an ordered comparison
between a packed floating point vector and a splat vector of NaN.
An ordered comparison between a vector and a constant vector of NaN, should
always be folded into a constant vector where each element is i1 false.
Since revision 266175, SimplifyFCmpInst folds the ordered fcmp into a scalar
'false'. Later on, this would cause an assertion failure, since the value type
of the folded value doesn't match the expected value type of the uses of the
original instruction: "Assertion failed: New->getType() == getType() &&
"replaceAllUses of value with new value of different type!".
This patch fixes the issue and adds a test case to the already existing test
InstSimplify/floating-point-compares.ll.
Differential Revision: https://reviews.llvm.org/D24143
llvm-svn: 280488
We're sinking stores, which is a good thing, but in the process creating selects for the store address operand, which SROA/Mem2Reg can't look through, which caused serious regressions.
The real fix is in SROA, which I'll be looking into.
llvm-svn: 280470
While removing a scalar shackle from an icmp fold, I noticed that I couldn't find any tests to trigger
this code path.
The 'and' shrinking transform should be handled by InstCombiner::foldCastedBitwiseLogic()
or eliminated with InstSimplify. The icmp narrowing is part of InstCombiner::foldICmpWithCastAndCast().
Differential Revision: https://reviews.llvm.org/D24031
llvm-svn: 280370
This was a real restriction in the original version of SinkIfThenCodeToEnd. Now it's been rewritten, the restriction can be lifted.
As part of this, we handle a very common and useful case where one of the incoming branches is actually conditional. Consider:
if (a)
x(1);
else if (b)
x(2);
This produces the following CFG:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ | /
[ end ]
[end] has two unconditional predecessor arcs and one conditional. The conditional refers to the implicit empty 'else' arc. This same pattern can also be caused by an empty default block in a switch.
We can't sink the call to x() down to end because no call to x() happens on the third incoming arc (assume that x() has sideeffects for the sake of argument; if something is safe to speculate we could indeed sink nevertheless but this cannot happen in the general case and causes many extra selects).
We are now able to detect this case and split off the unconditional arcs to a common successor:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ / |
[sink.split] |
\ /
[ end ]
Now we can sink the call to x() into %sink.split. This can cause significant code simplification in many testcases.
llvm-svn: 280364
r279460 rewrote this function to be able to handle more than two incoming edges and took pains to ensure this didn't regress anything.
This time we change the logic for determining if an instruction should be sunk. Previously we used a single pass greedy algorithm - sink instructions until one requires more than one PHI node or we run out of instructions to sink.
This had the problem that sinking instructions that had non-identical but trivially the same operands needed extra logic so we sunk them aggressively. For example:
%a = load i32* %b %d = load i32* %b
%c = gep i32* %a, i32 0 %e = gep i32* %d, i32 1
Sinking %c and %e would naively require two PHI merges as %a != %d. But the loads are obviously equivalent (and maybe can't be hoisted because there is no common predecessor).
This is why we implemented the fairly complex function areValuesTriviallySame(), to look through trivial differences like this. However it's just not clever enough.
Instead, throw areValuesTriviallySame away, use pointer equality to check equivalence of operands and switch to a two-stage algorithm.
In the "scan" stage, we look at every sinkable instruction in isolation from end of block to front. If it's sinkable, we keep track of all operands that required PHI merging.
In the "sink" stage, we iteratively sink the last non-terminator in the source blocks. But when calculating how many PHIs are actually required to be inserted (to work out if we should stop or not) we remove any values that have already been sunk from the set of PHI-merges required, which allows us to be more aggressive.
This turns an algorithm with potentially recursive lookahead (looking through GEPs, casts, loads and any other instruction potentially not CSE'd) to two linear scans.
llvm-svn: 280351
As discussed in https://reviews.llvm.org/D22666, our current mechanism to
support -pg profiling, where we insert calls to mcount(), or some similar
function, is fundamentally broken. We insert these calls in the frontend, which
means they get duplicated when inlining, and so the accumulated execution
counts for the inlined-into functions are wrong.
Because we don't want the presence of these functions to affect optimizaton,
they should be inserted in the backend. Here's a pass which would do just that.
The knowledge of the name of the counting function lives in the frontend, so
we're passing it here as a function attribute. Clang will be updated to use
this mechanism.
Differential Revision: https://reviews.llvm.org/D22825
llvm-svn: 280347
-fprofile-dir=path allows the user to specify where .gcda files should be
emitted when the program is run. In particular, this is the first flag that
causes the .gcno and .o files to have different paths, LLVM is extended to
support this. -fprofile-dir= does not change the file name in the .gcno (and
thus where lcov looks for the source) but it does change the name in the .gcda
(and thus where the runtime library writes the .gcda file). It's different from
a GCOV_PREFIX because a user can observe that the GCOV_PREFIX_STRIP will strip
paths off of -fprofile-dir= but not off of a supplied GCOV_PREFIX.
To implement this we split -coverage-file into -coverage-data-file and
-coverage-notes-file to specify the two different names. The !llvm.gcov
metadata node grows from a 2-element form {string coverage-file, node dbg.cu}
to 3-elements, {string coverage-notes-file, string coverage-data-file, node
dbg.cu}. In the 3-element form, the file name is already "mangled" with
.gcno/.gcda suffixes, while the 2-element form left that to the middle end
pass.
llvm-svn: 280306
Summary:
Use MemorySSA, if requested, to do less conservative memory dependency
checking.
This change doesn't enable the MemorySSA enhanced EarlyCSE in the
default pipelines, so should be NFC.
Reviewers: dberlin, sanjoy, reames, majnemer
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19821
llvm-svn: 280279
This is a first step towards supporting deopt value lowering and reporting entirely with the register allocator. I hope to build on this in the near future to support live-on-return semantics, but I have a use case which allows me to test and investigate code quality with just the live-in semantics so I've chosen to start there. For those curious, my use cases is our implementation of the "__llvm_deoptimize" function we bind to @llvm.deoptimize. I'm choosing not to hard code that fact in the patch and instead make it configurable via function attributes.
The basic approach here is modelled on what is done for the "Live In" values on stackmaps and patchpoints. (A secondary goal here is to remove one of the last barriers to merging the pseudo instructions.) We start by adding the operands directly to the STATEPOINT SDNode. Once we've lowered to MI, we extend the remat logic used by the register allocator to fold virtual register uses into StackMap::Indirect entries as needed. This does rely on the fact that the register allocator rematerializes. If it didn't along some code path, we could end up with more vregs than physical registers and fail to allocate.
Today, we *only* fold in the register allocator. This can create some weird effects when combined with arguments passed on the stack because we don't fold them appropriately. I have an idea how to fix that, but it needs this patch in place to work on that effectively. (There's some weird interaction with the scheduler as well, more investigation needed.)
My near term plan is to land this patch off-by-default, experiment in my local tree to identify any correctness issues and then start fixing codegen problems one by one as I find them. Once I have the live-in lowering fully working (both correctness and code quality), I'm hoping to move on to the live-on-return semantics. Note: I don't have any *known* miscompiles with this patch enabled, but I'm pretty sure I'll find at least a couple. Thus, the "experimental" tag and the fact it's off by default.
Differential Revision: https://reviews.llvm.org/D24000
llvm-svn: 280250
We check that a sinking candidate is used by only one PHI node during our legality checks. However for instructions that are used by other sinking candidates our heuristic is less conservative. This can result in a candidate actually being illegal when we come to sink it because of how we sunk a predecessor. Do the used-by-only-one-PHI checks again during sinking to ensure we don't crash.
llvm-svn: 280228
We're sinking stores, which is a good thing, but in the process creating selects for the store address operand, which SROA/Mem2Reg can't look through, which caused serious regressions.
The real fix is in SROA, which I'll be looking into.
llvm-svn: 280219
This was a real restriction in the original version of SinkIfThenCodeToEnd. Now it's been rewritten, the restriction can be lifted.
As part of this, we handle a very common and useful case where one of the incoming branches is actually conditional. Consider:
if (a)
x(1);
else if (b)
x(2);
This produces the following CFG:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ | /
[ end ]
[end] has two unconditional predecessor arcs and one conditional. The conditional refers to the implicit empty 'else' arc. This same pattern can also be caused by an empty default block in a switch.
We can't sink the call to x() down to end because no call to x() happens on the third incoming arc (assume that x() has sideeffects for the sake of argument; if something is safe to speculate we could indeed sink nevertheless but this cannot happen in the general case and causes many extra selects).
We are now able to detect this case and split off the unconditional arcs to a common successor:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ / |
[sink.split] |
\ /
[ end ]
Now we can sink the call to x() into %sink.split. This can cause significant code simplification in many testcases.
llvm-svn: 280217
r279460 rewrote this function to be able to handle more than two incoming edges and took pains to ensure this didn't regress anything.
This time we change the logic for determining if an instruction should be sunk. Previously we used a single pass greedy algorithm - sink instructions until one requires more than one PHI node or we run out of instructions to sink.
This had the problem that sinking instructions that had non-identical but trivially the same operands needed extra logic so we sunk them aggressively. For example:
%a = load i32* %b %d = load i32* %b
%c = gep i32* %a, i32 0 %e = gep i32* %d, i32 1
Sinking %c and %e would naively require two PHI merges as %a != %d. But the loads are obviously equivalent (and maybe can't be hoisted because there is no common predecessor).
This is why we implemented the fairly complex function areValuesTriviallySame(), to look through trivial differences like this. However it's just not clever enough.
Instead, throw areValuesTriviallySame away, use pointer equality to check equivalence of operands and switch to a two-stage algorithm.
In the "scan" stage, we look at every sinkable instruction in isolation from end of block to front. If it's sinkable, we keep track of all operands that required PHI merging.
In the "sink" stage, we iteratively sink the last non-terminator in the source blocks. But when calculating how many PHIs are actually required to be inserted (to work out if we should stop or not) we remove any values that have already been sunk from the set of PHI-merges required, which allows us to be more aggressive.
This turns an algorithm with potentially recursive lookahead (looking through GEPs, casts, loads and any other instruction potentially not CSE'd) to two linear scans.
llvm-svn: 280216
This was deliberately disabled during my rewrite of SinkIfThenToEnd to keep behaviour
at least vaguely consistent with the previous version and keep it as close to NFC as
I could.
There's no real reason not to merge sideeffect calls though, so let's do it! Small fixup
along the way to ensure we don't create indirect calls.
Should fix PR28964.
llvm-svn: 280215
Summary:
1) CoroEarly now lowers llvm.coro.promise intrinsic that allows to obtain
a coroutine promise pointer from a coroutine frame and vice versa.
2) CoroFrame now interprets Promise argument of llvm.coro.begin to
place CoroutinPromise alloca at a deterministic offset from the coroutine frame.
Now, the coroutine promise example from docs\Coroutines.rst compiles and produces expected result (see test/Transform/Coroutines/ex4.ll).
Reviewers: majnemer
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23993
llvm-svn: 280184
Summary:
LSV was using two vector sets (heads and tails) to track pairs of adjiacent position to vectorize.
A recent optimization is trying to obtain the longest chain to vectorize and assumes the positions
in heads(H) and tails(T) match, which is not the case is there are multiple tails for the same head.
e.g.:
i1: store a[0]
i2: store a[1]
i3: store a[1]
Leads to:
H: i1
T: i2 i3
Instead of:
H: i1 i1
T: i2 i3
So the positions for instructions that follow i3 will have different indexes in H/T.
This patch resolves PR29148.
This issue also surfaced the fact that if the chain is too long, and TLI
returns a "not-fast" answer, the whole chain will be abandoned for
vectorization, even though a smaller one would be beneficial.
Added a testcase and FIXME for this.
Reviewers: tstellarAMD, arsenm, jlebar
Subscribers: mzolotukhin, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D24057
llvm-svn: 280179
We don't need to limit predication to blocks that have a single incoming
edge, we just need to use the right mask.
This fixes PR30172.
Differential Revision: https://reviews.llvm.org/D24009
llvm-svn: 280148
Summary:
Fix a couple issues limiting the application of indirect call promotion
in ThinLTO mode:
- Invoke indirect call promotion before globalopt, since it may
eliminate imported functions which appear unreferenced.
- Invoke indirect call promotion with InLTO=true so that the PGOFuncName
metadata is used to get the name for locals which would have been
renamed during promotion.
Reviewers: davidxl, mehdi_amini
Subscribers: Prazek, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D24004
llvm-svn: 280024
After r279649 when getting a vector value from VectorLoopValueMap, we create an
insertelement sequence on-demand if the value has been scalarized instead of
vectorized. We previously inserted this insertelement sequence before the
value's first vector user. However, this insert location is problematic if that
user is the phi node of a first-order recurrence. With this patch, we move the
insertelement sequence after the last scalar instruction we created when
scalarizing the value. Thus, the value's vector definition in the new loop will
immediately follow its scalar definitions. This should fix PR30183.
Reference: https://llvm.org/bugs/show_bug.cgi?id=30183
llvm-svn: 280001
Summary:
While walking the use chain for identifying rematerializable values in RS4GC,
add the case where the current value and base value are the same PHI nodes.
This will aid rematerialization of geps and casts instead of relocating.
Reviewers: sanjoy, reames, igor
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23920
llvm-svn: 279975
Assuming the default FP env, we should not treat fdiv and frem any differently in terms of
trapping behavior than any other FP op. Ie, FP ops do not trap with the default FP env.
This matches how we treat the fdiv/frem in IR with isSafeToSpeculativelyExecute() and in
the backend after:
https://reviews.llvm.org/rL279970
llvm-svn: 279973
Summary:
[Coroutines] Part 9: Add cleanup subfunction.
This patch completes coroutine heap allocation elision. Now, the heap elision example from docs\Coroutines.rst compiles and produces expected result (see test/Transform/Coroutines/ex3.ll)
Intrinsic Changes:
* coro.free gets a token parameter tying it to coro.id to allow reliably discovering all coro.frees associated with a particular coroutine.
* coro.id gets an extra parameter that points back to a coroutine function. This allows to check whether a coro.id describes the enclosing function or it belongs to a different function that was later inlined.
CoroSplit now creates three subfunctions:
# f$resume - resume logic
# f$destroy - cleanup logic, followed by a deallocation code
# f$cleanup - just the cleanup code
CoroElide pass during devirtualization replaces coro.destroy with either f$destroy or f$cleanup depending whether heap elision is performed or not.
Other fixes, improvements:
* Fixed buglet in Shape::buildFrame that was not creating coro.save properly if coroutine has more than one suspend point.
* Switched to using variable width suspend index field (no longer limited to 32 bit index field can be as little as i1 or as large as i<whatever-size_t-is>)
Reviewers: majnemer
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23844
llvm-svn: 279971
Fixed a bug in run-time checks for possible memory conflicts inside loop.
The bug is in Low <-> High boundaries calculation. The High boundary should be calculated as "last memory access pointer + element size".
Differential revision: https://reviews.llvm.org/D23176
llvm-svn: 279930
Summary:
This is obviously an interesting case because it may motivate code
restructuring or LTO.
Reporting this requires instantiation of ORE in the loop where the call
sites are first gathered. I've checked compile-time
overhead *with* -Rpass-with-hotness and the worst slow-down was 6% in
mcf and quickly tailing off. As before without -Rpass-with-hotness
there is no overhead.
Because this could be a pretty noisy diagnostics, it is currently
qualified as 'verbose'. As of this patch, 'verbose' diagnostics are
only emitted with -Rpass-with-hotness, i.e. when the output is expected
to be filtered.
Reviewers: eraman, chandlerc, davidxl, hfinkel
Subscribers: tejohnson, Prazek, davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D23415
llvm-svn: 279860
Summary:
This fixes pr29105. The reason is that lifetime marks creates new
aliasing pointers the original ones, but before this patch aliases
were not checked in performMemCpyToMemSetOptzn.
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23846
llvm-svn: 279769
It is invalid to hoist stores or loads if they are not executed on all paths
from the hoisting point to the exit of the function. In the testcase, there are
paths in the loop that do not execute the stores or the loads, and so hoisting
them within the loop is unsafe.
The problem is that the current implementation of hoistingFromAllPaths is
incomplete: it walks all blocks dominated by the hoisting point, and does not
return false when the loop contains a path on which the hoisted ld/st is
not executed.
Differential Revision: https://reviews.llvm.org/D23843
llvm-svn: 279732
This patch unifies the data structures we use for mapping instructions from the
original loop to their corresponding instructions in the new loop. Previously,
we maintained two distinct maps for this purpose: WidenMap and ScalarIVMap.
WidenMap maintained the vector values each instruction from the old loop was
represented with, and ScalarIVMap maintained the scalar values each scalarized
induction variable was represented with. With this patch, all values created
for the new loop are maintained in VectorLoopValueMap.
The change allows for several simplifications. Previously, when an instruction
was scalarized, we had to insert the scalar values into vectors in order to
maintain the mapping in WidenMap. Then, if a user of the scalarized value was
also scalar, we had to extract the scalar values from the temporary vector we
created. We now aovid these unnecessary scalar-to-vector-to-scalar conversions.
If a scalarized value is used by a scalar instruction, the scalar value is used
directly. However, if the scalarized value is needed by a vector instruction,
we generate the needed insertelement instructions on-demand.
A common idiom in several locations in the code (including the scalarization
code), is to first get the vector values an instruction from the original loop
maps to, and then extract a particular scalar value. This patch adds
getScalarValue for this purpose along side getVectorValue as an interface into
VectorLoopValueMap. These functions work together to return the requested
values if they're available or to produce them if they're not.
The mapping has also be made less permissive. Entries can be added to
VectorLoopValue map with the new initVector and initScalar functions.
getVectorValue has been modified to return a constant reference to the mapped
entries.
There's no real functional change with this patch; however, in some cases we
will generate slightly different code. For example, instead of an insertelement
sequence following the definition of an instruction, it will now precede the
first use of that instruction. This can be seen in the test case changes.
Differential Revision: https://reviews.llvm.org/D23169
llvm-svn: 279649
I'm not sure if the `!isa<CallInst>(Inst) &&
!isa<TerminatorInst>(Inst))` bit is correct either, but this fixes the
case we know is broken.
llvm-svn: 279647
div/rem instructions in basic blocks that require predication currently prevent
vectorization. This patch extends the existing mechanism for predicating stores
to handle other instructions and leverages it to predicate divs and rems.
Differential Revision: https://reviews.llvm.org/D22918
llvm-svn: 279620
Summary:
This patch adds coroutine frame building algorithm. Now, simple coroutines such as ex0.ll and ex1.ll (first examples from docs\Coroutines.rst can be compiled).
Documentation and overview is here: http://llvm.org/docs/Coroutines.html.
Upstreaming sequence (rough plan)
1.Add documentation. (https://reviews.llvm.org/D22603)
2.Add coroutine intrinsics. (https://reviews.llvm.org/D22659)
...
7. Split coroutine into subfunctions. (https://reviews.llvm.org/D23461)
8. Coroutine Frame Building algorithm <= we are here
9. Add f.cleanup subfunction.
10+. The rest of the logic
Reviewers: majnemer
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D23586
llvm-svn: 279609
The test case included with r279125 exposed an existing signed integer
overflow. Since getTreeCost can return INT_MAX, we can't sum this cost together
with other costs, such as getReductionCost.
This patch removes the possibility of assigning a cost of INT_MAX. Since we
were previously using INT_MAX as an indicator for "should not vectorize", we
now explicitly check this condition with "isTreeTinyAndNotFullyVectorizable"
before computing a cost.
This patch adds a run-line to the test case used for r279125 that ensures we
don't vectorize. Previously, this line would vectorize the test case by chance
due to undefined behavior in the cost calculation.
Differential Revision: https://reviews.llvm.org/D23723
llvm-svn: 279562
...because like the corresponding code, this is just too big to keep adding to.
And the next step is to add a vector version of each of these tests to show
missed folds.
Also, auto-generate CHECK lines and add comments for the tests that correspond to
the source code.
llvm-svn: 279530
[Recommitting now an unrelated assertion in SROA is sorted out]
The new version has several advantages:
1) IMSHO it's more readable and neater
2) It handles loads and stores properly
3) It can handle any number of incoming blocks rather than just two. I'll be taking advantage of this in a followup patch.
With this change we can now finally sink load-modify-store idioms such as:
if (a)
return *b += 3;
else
return *b += 4;
=>
%z = load i32, i32* %y
%.sink = select i1 %a, i32 5, i32 7
%b = add i32 %z, %.sink
store i32 %b, i32* %y
ret i32 %b
When this works for switches it'll be even more powerful.
Round 4. This time we should handle all instructions correctly, and not replace any operands that need to be constant with variables.
This was really hard to determine safely, so the helper function should be put into the Instruction API. I'll do that as a followup.
llvm-svn: 279460
Summary: We can allow sinking if the single user block has only one unique predecessor, regardless of the number of edges. Note that a switch statement with multiple cases can have the same destination.
Reviewers: mcrosier, majnemer, spatel, reames
Subscribers: reames, mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D23722
llvm-svn: 279448
The new version has several advantages:
1) IMSHO it's more readable and neater
2) It handles loads and stores properly
3) It can handle any number of incoming blocks rather than just two. I'll be taking advantage of this in a followup patch.
With this change we can now finally sink load-modify-store idioms such as:
if (a)
return *b += 3;
else
return *b += 4;
=>
%z = load i32, i32* %y
%.sink = select i1 %a, i32 5, i32 7
%b = add i32 %z, %.sink
store i32 %b, i32* %y
ret i32 %b
When this works for switches it'll be even more powerful.
Round 4. This time we should handle all instructions correctly, and not replace any operands that need to be constant with variables.
This was really hard to determine safely, so the helper function should be put into the Instruction API. I'll do that as a followup.
llvm-svn: 279443
This change cause performance regression on MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt from LNT and some other bechmarks.
See https://reviews.llvm.org/D18777 for details.
llvm-svn: 279433
This change needs to be reverted in order to revert -r278267 which cause performance regression on MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt from LNT and some other bechmarks.
See comments on https://reviews.llvm.org/D18777 for details.
llvm-svn: 279432
The test case included in r279125 exposed existing undefined behavior in the
SLP vectorizer that it did not introduce. This patch reapplies the original
patch, but modifies the test case to avoid hitting the undefined behavior. This
allows us to close PR28330 while keeping the UBSan bot happy. The undefined
behavior the original test uncovered will be addressed in a follow-on patch.
Reference: https://llvm.org/bugs/show_bug.cgi?id=28330
llvm-svn: 279370
This is a partial enablement (move the ConstantInt guard down) because there are many
different folds here and one of the later ones will require reworking 'isSignBitCheck'.
llvm-svn: 279339
The intended transform is:
// Simplify icmp eq (or (ptrtoint P), (ptrtoint Q)), 0
// -> and (icmp eq P, null), (icmp eq Q, null).
P and Q are both pointer types, but may have different types. We need
two calls to getNullValue() to make the icmps.
llvm-svn: 279271
CGSCC use a WeakVH to track call sites. RAUW a call within a function
can result in that WeakVH getting confused about whether or not the call
site is still around.
llvm-svn: 279268
Of course, we really need to refactor and fix all of the cmp predicates,
but this one is interesting because without it, we later perform an
information-losing transform of icmp (shl 1, Y), C, and we can't recover
the better fold.
llvm-svn: 279263
The new version has several advantages:
1) IMSHO it's more readable and neater
2) It handles loads and stores properly
3) It can handle any number of incoming blocks rather than just two. I'll be taking advantage of this in a followup patch.
With this change we can now finally sink load-modify-store idioms such as:
if (a)
return *b += 3;
else
return *b += 4;
=>
%z = load i32, i32* %y
%.sink = select i1 %a, i32 5, i32 7
%b = add i32 %z, %.sink
store i32 %b, i32* %y
ret i32 %b
When this works for switches it'll be even more powerful.
llvm-svn: 279229
We abort building vectorizable trees in some cases (e.g., if the maximum
recursion depth is reached, if the region size is too large, etc.). If this
happens for a reduction, we can be left with a root entry that needs to be
gathered. For these cases, we need make sure we actually set VectorizedValue to
the resulting vector.
This patch ensures we properly set VectorizedValue, and it also ensures the
insertelement sequence generated for the gathers is inserted at the correct
location.
Reference: https://llvm.org/bugs/show_bug.cgi?id=28330
Differential Revison: https://reviews.llvm.org/D23410
llvm-svn: 279125
It causes a regression on our internal benchmark. Introduce cvp-dont-process flag and set it off by default while investigating the regression.
llvm-svn: 279082
Also, add a scalar test to demonstrate one of the intermediate folds that
is necessary to accomplish the existing, multi-step test. And simplify
the vector tests to only check the final piece of that multi-step transform.
llvm-svn: 278995
minimal and boring form than the old pass manager's version.
This pass does the very minimal amount of work necessary to inline
functions declared as always-inline. It doesn't support a wide array of
things that the legacy pass manager did support, but is alse ... about
20 lines of code. So it has that going for it. Notably things this
doesn't support:
- Array alloca merging
- To support the above, bottom-up inlining with careful history
tracking and call graph updates
- DCE of the functions that become dead after this inlining.
- Inlining through call instructions with the always_inline attribute.
Instead, it focuses on inlining functions with that attribute.
The first I've omitted because I'm hoping to just turn it off for the
primary pass manager. If that doesn't pan out, I can add it here but it
will be reasonably expensive to do so.
The second should really be handled by running global-dce after the
inliner. I don't want to re-implement the non-trivial logic necessary to
do comdat-correct DCE of functions. This means the -O0 pipeline will
have to be at least 'always-inline,global-dce', but that seems
reasonable to me. If others are seriously worried about this I'd like to
hear about it and understand why. Again, this is all solveable by
factoring that logic into a utility and calling it here, but I'd like to
wait to do that until there is a clear reason why the existing
pass-based factoring won't work.
The final point is a serious one. I can fairly easily add support for
this, but it seems both costly and a confusing construct for the use
case of the always inliner running at -O0. This attribute can of course
still impact the normal inliner easily (although I find that
a questionable re-use of the same attribute). I've started a discussion
to sort out what semantics we want here and based on that can figure out
if it makes sense ta have this complexity at O0 or not.
One other advantage of this design is that it should be quite a bit
faster due to checking for whether the function is a viable candidate
for inlining exactly once per function instead of doing it for each call
site.
Anyways, hopefully a reasonable starting point for this pass.
Differential Revision: https://reviews.llvm.org/D23299
llvm-svn: 278896
It is pretty easy to get it down to O(nlogn + mlogm). This
implementation has the added benefit of automatically deduplicating
entries between the two sets.
llvm-svn: 278837
I have audited all the callers of concatenate and none require duplicate
entries to service concatenation.
These duplicates serve no purpose but to needlessly embiggen the IR.
N.B. Layering getMostGenericAliasScope on top of concatenate makes it
O(nlogn + mlogm) instead of O(n*m).
llvm-svn: 278836
Summary:
This patch adds simple coroutine splitting logic to CoroSplit pass.
Documentation and overview is here: http://llvm.org/docs/Coroutines.html.
Upstreaming sequence (rough plan)
1.Add documentation. (https://reviews.llvm.org/D22603)
2.Add coroutine intrinsics. (https://reviews.llvm.org/D22659)
...
7. Split coroutine into subfunctions <= we are here
8. Coroutine Frame Building algorithm
9. Handle coroutine with unwinds
10+. The rest of the logic
Reviewers: majnemer
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23461
llvm-svn: 278830
This reverts commit r278660.
It causes downstream assertion failure in InstCombine on shuffle
instructions. Comes up in __mm_swizzle_epi32.
llvm-svn: 278672
The new version has several advantages:
1) IMSHO it's more readable and neater
2) It handles loads and stores properly
3) It can handle any number of incoming blocks rather than just two. I'll be taking advantage of this in a followup patch.
With this change we can now finally sink load-modify-store idioms such as:
if (a)
return *b += 3;
else
return *b += 4;
=>
%z = load i32, i32* %y
%.sink = select i1 %a, i32 5, i32 7
%b = add i32 %z, %.sink
store i32 %b, i32* %y
ret i32 %b
When this works for switches it'll be even more powerful.
llvm-svn: 278660
If a loop is not rotated (for example when optimizing for size), the latch is not the backedge. If we promote an expression to post-inc form, we not only increase register pressure and add a COPY for that IV expression but for all IVs!
Motivating testcase:
void f(float *a, float *b, float *c, int n) {
while (n-- > 0)
*c++ = *a++ + *b++;
}
It's imperative that the pointer increments be located in the latch block and not the header block; if not, we cannot use post-increment loads and stores and we have to keep both the post-inc and pre-inc values around until the end of the latch which bloats register usage.
llvm-svn: 278658
IRCE has the ability to further version pre-loops and post-loops that it
created, but this isn't useful at all. This change teaches IRCE to
leave behind some metadata in the loops it creates (by cloning the main
loop) so that these new loops are not re-processed by IRCE.
Today this bug is hidden by another bug -- IRCE does not update LoopInfo
properly so the loop pass manager does not re-invoke IRCE on the loops
it split out. However, once the latter is fixed the bug addressed in
this change causes IRCE to infinite-loop in some cases (e.g. it splits
out a pre-loop, a pre-pre-loop from that, a pre-pre-pre-loop from that
and so on).
llvm-svn: 278617
The (negative) test case is supposed to check that IRCE does not muck
with range checks it cannot handle, not that it does the right thing in
the absence of profiling information.
llvm-svn: 278612
Loops containing `indirectbr` may not be in simplified form, even after
running LoopSimplify. Reject then gracefully, instead of tripping an
assert.
llvm-svn: 278611
Summary:
Refactor the existing support into a LoopDataPrefetch implementation
class and a LoopDataPrefetchLegacyPass class that invokes it.
Add a new LoopDataPrefetchPass for the new pass manager that utilizes
the LoopDataPrefetch implementation class.
Reviewers: mehdi_amini
Subscribers: sanjoy, mzolotukhin, nemanjai, llvm-commits
Differential Revision: https://reviews.llvm.org/D23483
llvm-svn: 278591
`IVVisitor::visitCast` used to have the invariant that if the
instruction it was passed was a sext or zext instruction, the result of
the instruction would be wider than the induction variable. This is no
longer true after rL275037, so this change teaches `IndVarSimplify` s
implementation of `IVVisitor::visitCast` to work with the relaxed
invariant.
A corresponding change to SimplifyIndVar to preserve the said invariant
after rL275037 would also work, but given how `IVVisitor::visitCast` is
spelled (no indication of said invariant), I figured the current fix is
cleaner.
Fixes PR28935.
llvm-svn: 278584
InnerLoopVectorizer shouldn't handle a loop with cycles inside the loop
body, even if that cycle isn't a natural loop.
Fixes PR28541.
Differential Revision: https://reviews.llvm.org/D22952
llvm-svn: 278573
They aren't static, and moving them to the entry block across something
else will only result in tears.
Root cause of http://crbug.com/636558.
llvm-svn: 278571
Summary: The refined propagation algorithm is more accurate and robust.
Reviewers: davidxl, dnovillo
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23224
llvm-svn: 278522
Rewrite Visited[Cond] = getValueFromConditionImpl(..., Visited) statement which can lead to a memory corruption since getValueFromConditionImpl changes Visited map and invalidates the iterators.
llvm-svn: 278514
Take range metadata into account for conditions like this:
%length = load i32, i32* %length_ptr, !range !{i32 0, i32 2147483647}
%cmp = icmp ult i32 %a, %length
This is a common pattern for range checks where the length of the array is dynamically loaded.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D23267
llvm-svn: 278496
Currently LVI can only gather value constraints from comparisons like:
* icmp <pred> Val, ...
* icmp ult (add Val, Offset), ...
In fact we can handle any predicate in latter comparisons.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D23357
llvm-svn: 278493
Summary:
1. Make coroutine representation more robust against optimization that may duplicate instruction by introducing coro.id intrinsics that returns a token that will get fed into coro.alloc and coro.begin. Due to coro.id returning a token, it won't get duplicated and can be used as reliable indicator of coroutine identify when a particular coroutine call gets inlined.
2. Move last three arguments of coro.begin into coro.id as they will be shared if coro.begin will get duplicated.
3. doc + test + code updated to support the new intrinsic.
Reviewers: mehdi_amini, majnemer
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D23412
llvm-svn: 278481
Summary:
This patch adds IsVariadicFunction bit to summary in order
to not import variadic functions. Inliner doesn't inline
variadic functions because it is hard to reason about it.
This one small fix improves Importer by about 16%
(going from 86% to 100% of imported functions that are
inlined anywhere)
on some spec benchmarks like 'int' and others.
Reviewers: eraman, mehdi_amini, tejohnson
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D23339
llvm-svn: 278432
When legal, extending trip count in the loop control logic generates better code compared to truncating IV. This is because
(1) extending trip count is a loop invariant operation (see genLoopLimit where we prove trip count is loop invariant).
(2) Scalar Evolution seems to have problems understanding trunc when computing loop trip count. So removing them allows better analysis performed in Scalar Evolution. (In particular this fixes PR 28363 which is the motivation for this change).
I am not going to perform any performance test. Any degradation caused by this should be an indication of a bug elsewhere.
To prove legality, we rely on SCEV to prove zext(trunc(IV)) == IV (or similarly for sext). If this holds, we can prove equivalence of trunc(IV)==ExitCnt (1) and IV == zext(ExitCnt). Simply take zext of boths sides of (1) and apply the proven equivalence.
This commit contains changes in a newly added testcase which was not included in the previous commit (which was reverted later on).
https://reviews.llvm.org/D23075
llvm-svn: 278421
Summary:
This is an extension of the fix in r271424. That fix dealt with builder
insert points being moved by SCEV expansion, but only for the lifetime
of the expand call. This change modifies the interface so that LSR can
safely call expand multiple times at the same insert point and do the
right thing if one of the expansions decides to move the original insert
point.
This is a fix for PR28719.
Reviewers: sanjoy
Subscribers: llvm-commits, mcrosier, mzolotukhin
Differential Revision: https://reviews.llvm.org/D23342
llvm-svn: 278413
Summary:
This fixes PR 28933 by making sure GVNHoist does not try to recreate memory
accesses when it has not actually moved them.
Reviewers: sebpop
Subscribers: llvm-commits, george.burgess.iv
Differential Revision: https://reviews.llvm.org/D23411
llvm-svn: 278401
Summary:
Keep track of all methods for which we have devirtualized at least
one call and then print them sorted alphabetically. That allows to
avoid duplicates and also makes the order deterministic.
Add optimization names into the remarks, so that it's easier to
understand how has each method been devirtualized.
Fix a bug when wrong methods could have been reported for
tryVirtualConstProp.
Reviewers: kcc, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23297
llvm-svn: 278389
When legal, extending trip count in the loop control logic generates better code compared to truncating IV. This is because
(1) extending trip count is a loop invariant operation (see genLoopLimit where we prove trip count is loop invariant).
(2) Scalar Evolution seems to have problems understanding trunc when computing loop trip count. So removing them allows better analysis performed in Scalar Evolution. (In particular this fixes PR 28363 which is the motivation for this change).
I am not going to perform any performance test. Any degradation caused by this should be an indication of a bug elsewhere.
To prove legality, we rely on SCEV to prove zext(trunc(IV)) == IV (or similarly for sext). If this holds, we can prove equivalence of trunc(IV)==ExitCnt (1) and IV == zext(ExitCnt). Simply take zext of boths sides of (1) and apply the proven equivalence.
https://reviews.llvm.org/D23075
llvm-svn: 278334
Change --no-pgo-warn-missing to -pgo-warn-missing-function
and negate the default. /NFC
Add more test to make sure the warning is off by default
llvm-svn: 278314
Summary:
A particular coroutine usage pattern, where a coroutine is created, manipulated and
destroyed by the same calling function, is common for coroutines implementing
RAII idiom and is suitable for allocation elision optimization which avoid
dynamic allocation by storing the coroutine frame as a static `alloca` in its
caller.
coro.free and coro.alloc intrinsics are used to indicate which code needs to be suppressed
when dynamic allocation elision happens:
```
entry:
%elide = call i8* @llvm.coro.alloc()
%need.dyn.alloc = icmp ne i8* %elide, null
br i1 %need.dyn.alloc, label %coro.begin, label %dyn.alloc
dyn.alloc:
%alloc = call i8* @CustomAlloc(i32 4)
br label %coro.begin
coro.begin:
%phi = phi i8* [ %elide, %entry ], [ %alloc, %dyn.alloc ]
%hdl = call i8* @llvm.coro.begin(i8* %phi, i32 0, i8* null,
i8* bitcast ([2 x void (%f.frame*)*]* @f.resumers to i8*))
```
and
```
%mem = call i8* @llvm.coro.free(i8* %hdl)
%need.dyn.free = icmp ne i8* %mem, null
br i1 %need.dyn.free, label %dyn.free, label %if.end
dyn.free:
call void @CustomFree(i8* %mem)
br label %if.end
if.end:
...
```
If heap allocation elision is performed, we replace coro.alloc with a static alloca on the caller frame and coro.free with null constant.
Also, we need to make sure that if there are any tail calls referencing the coroutine frame, we need to remote tail call attribute, since now coroutine frame lives on the stack.
Documentation and overview is here: http://llvm.org/docs/Coroutines.html.
Upstreaming sequence (rough plan)
1.Add documentation. (https://reviews.llvm.org/D22603)
2.Add coroutine intrinsics. (https://reviews.llvm.org/D22659)
3.Add empty coroutine passes. (https://reviews.llvm.org/D22847)
4.Add coroutine devirtualization + tests.
ab) Lower coro.resume and coro.destroy (https://reviews.llvm.org/D22998)
c) Do devirtualization (https://reviews.llvm.org/D23229)
5.Add CGSCC restart trigger + tests. (https://reviews.llvm.org/D23234)
6.Add coroutine heap elision + tests. <= we are here
7.Add the rest of the logic (split into more patches)
Reviewers: mehdi_amini, majnemer
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D23245
llvm-svn: 278242
Teach LVI how to gather information from conditions in the form of (cond1 && cond2). Our out-of-tree front-end emits range checks in this form.
Reviewed By: sanjoy
Differential Revision: http://reviews.llvm.org/D23200
llvm-svn: 278231
This is a resubmission of previously reverted r277592. It was hitting overly strong assertion in getConstantRange which was relaxed in r278217.
Use LVI to prove that adds do not wrap. The change is motivated by https://llvm.org/bugs/show_bug.cgi?id=28620 bug and it's the first step to fix that problem.
Reviewed By: sanjoy
Differential Revision: http://reviews.llvm.org/D23059
llvm-svn: 278220
Hal pointed out that the semantic of our intrinsic and the libc
call are slightly different. Add a comment while I'm here to
explain why we can't emit an intrinsic. Thanks Hal!
llvm-svn: 278200
Summary:
The inliner not being a function pass requires the work-around of
generating the OptimizationRemarkEmitter and in turn BFI on demand.
This will go away after the new PM is ready.
BFI is only computed inside ORE if the user has requested hotness
information for optimization diagnostitics (-pass-remark-with-hotness at
the 'opt' level). Thus there is no additional overhead without the
flag.
Reviewers: hfinkel, davidxl, eraman
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D22694
llvm-svn: 278185
Summary:
This hopefully fixes PR28825. The problem now was that a value from the
original loop was used in a subloop, which became a sibling after separation.
While a subloop doesn't need an lcssa phi node, a sibling does, and that's
where we broke LCSSA. The most natural way to fix this now is to simply call
formLCSSA on the original loop: it'll do what we've been doing before plus
it'll cover situations described above.
I think we don't need to run formLCSSARecursively here, and we have an assert
to verify this (I've tried testing it on LLVM testsuite + SPECs). I'd be happy
to be corrected here though.
I also changed a run line in the test from '-lcssa -loop-unroll' to
'-lcssa -loop-simplify -indvars', because it exercises LCSSA
preservation to the same extent, but also makes less unrelated
transformation on the CFG, which makes it easier to verify.
Reviewers: chandlerc, sanjoy, silvas
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23288
llvm-svn: 278173
Summary:
We teach alias analysis that invariant.start is readonly.
This helps with GVN and memcopy optimizations that currently treat.
invariant.start as a clobber.
We need to treat this as readonly, so that DSE does not incorrectly
remove stores prior to the invariant.start
Reviewers: sanjoy, reames, majnemer, dberlin
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23214
llvm-svn: 278138
no prof data for func warning is turned off by default
due to its high verbosity and minimal usefulness.
Differential Revision: http://reviews.llvm.org/D23295
llvm-svn: 278127
Summary:
In the use optimizer, we need to keep of whether the lower bound still
dominates us or else we may decide a lower bound is still valid when it
is not due to intervening pushes/pops. Fixes PR28880 (and probably a
bunch of other things).
Reviewers: george.burgess.iv
Subscribers: MatzeB, llvm-commits, sebpop
Differential Revision: https://reviews.llvm.org/D23237
llvm-svn: 277978
Summary:
The correctness fix here is that when we CSE a load with another load,
we need to combine the metadata on the two loads. This matches the
behavior of other passes, like instcombine and GVN.
There's also a minor optimization improvement here: for load PRE, the
aliasing metadata on the inserted load should be the same as the
metadata on the original load. Not sure why the old code was throwing
it away.
Issue found by inspection.
Differential Revision: http://reviews.llvm.org/D21460
llvm-svn: 277977
Xinliang David Li