DemandedBits and BDCE currently only support scalar integers. This
patch extends them to also handle vector integer operations. In this
case bits are not tracked for individual vector elements, instead a
bit is demanded if it is demanded for any of the elements. This matches
the behavior of computeKnownBits in ValueTracking and
SimplifyDemandedBits in InstCombine.
The getDemandedBits() method can now only be called on instructions that
have integer or vector of integer type. Previously it could be called on
any sized instruction (even if it was not particularly useful). The size
of the return value is now always the scalar size in bits (while
previously it was the type size in bits).
Differential Revision: https://reviews.llvm.org/D55297
llvm-svn: 348549
Partial Redundancy Elimination of GEPs prevents CodeGenPrepare from
sinking the addressing mode computation of memory instructions back
to its uses. The problem comes from the insertion of PHIs, which
confuse CGP and make it bail.
I've autogenerated the check lines of an existing test and added a
store instruction to demonstrate the motivation behind this change.
The store is now using the gep instead of a phi.
Differential Revision: https://reviews.llvm.org/D55009
llvm-svn: 348496
This reverts commit r348457.
The original commit causes clang to crash when doing an instrumented
build with a new pass manager. Reverting to unbreak our integrate.
llvm-svn: 348484
I was finally able to quantify what i thought was missing in the fix,
it was vector constants. If we have a scalar (and %x, -1),
it will be instsimplified before we reach this code,
but if it is a vector, we may still have a -1 element.
Thus, we want to avoid the fold if *at least one* element is -1.
Or in other words, ignoring the undef elements, no sign bits
should be set. Thus, m_NonNegative().
A follow-up for rL348181
https://bugs.llvm.org/show_bug.cgi?id=39861
llvm-svn: 348462
This patch teaches LoopSimplifyCFG to delete loop blocks that have
become unreachable after terminator folding has been done.
Differential Revision: https://reviews.llvm.org/D54023
Reviewed By: anna
llvm-svn: 348457
This change caused SEGVs in instcombine. (The r347934 change seems to me to be a
precipitating cause, not a root cause. Details are on the llvm-commits thread
for r347934.)
llvm-svn: 348426
Extracting from a splat constant is always handled by InstSimplify.
Move the test for this from InstCombine to InstSimplify to make
sure that stays true.
llvm-svn: 348423
Treat terminators which resume exception propagation as returning instructions
(at least, for the purposes of marking outlined functions `noreturn`). This is
to avoid inserting traps after calls to outlined functions which unwind.
rdar://46129950
llvm-svn: 348404
Summary: debug intrinsics might be marked norecurse to enable the caller function to be norecurse and optimized if needed. This avoids code gen optimisation differences when -g is used, as in globalOpt.cpp:processInternalGlobal checks.
Reviewers: chandlerc, jmolloy, aprantl
Reviewed By: aprantl
Subscribers: aprantl, llvm-commits
Differential Revision: https://reviews.llvm.org/D55187
llvm-svn: 348381
The tests here are based on the motivating cases from D54827.
More background:
1. We don't get these cases in general with SimplifyCFG because the root
of the pattern match is an icmp, not a branch. I'm not sure how often
we encounter this pattern vs. the seemingly more likely case with
branches, but I don't see evidence to leave the minimal pattern
unoptimized.
2. This has a chance of increasing compile-time because we're using a
ValueTracking call to handle the match. The motivating cases could be
handled with a simpler pair of calls to isImpliedTrueByMatchingCmp/
isImpliedFalseByMatchingCmp, but I saw that we have a more
comprehensive wrapper around those, so we might as well use it here
unless there's evidence that it's significantly slower.
3. Ideally, we'd handle the fold to constants in InstSimplify, but as
with the existing code here, we could extend this to handle cases
where the result is not a constant, but a new combined predicate.
That would mean splitting the logic across the 2 passes and possibly
duplicating the pattern-matching cost.
4. As mentioned in D54827, this seems like the kind of thing that should
be handled in Correlated Value Propagation, but that pass is currently
limited to dealing with instructions with constant operands, so extending
this bit of InstCombine is the smallest/easiest way to get these patterns
optimized.
llvm-svn: 348367
Ideally, we would fold all of these in InstSimplify in a
similar way to rL347896, but this is a bit awkward when
we're trying to simplify a compare directly because the
ValueTracking API expects the compare as an input, but
in InstSimplify, we just have the operands of the compare.
Given that we can do transforms besides just simplifications,
we might as well just extend the code in InstCombine (which
already does simplifications with constant operands).
llvm-svn: 348312
If a PHI node out of extracted region has multiple incoming values from it,
split this PHI on two parts. First PHI has incomings only from region and
extracts with it (they are placed to the separate basic block that added to the
list of outlined), and incoming values in original PHI are replaced by first
PHI. Similar solution is already used in CodeExtractor for PHIs in entry block
(severSplitPHINodes method). It covers PR39433 bug.
Patch by Sergei Kachkov!
Differential Revision: https://reviews.llvm.org/D55018
llvm-svn: 348205
When we have a shuffle that extends a source vector with undefs
and then do some binop on that, we must make sure that the extra
elements remain undef with that binop if we reverse the order of
the binop and shuffle.
'or' is probably the easiest example to show the bug because
'or C, undef --> -1' (not undef). But there are other
opcode/constant combinations where this is true as shown by
the 'shl' test.
llvm-svn: 348191
These are the baseline tests for D54827.
Patch based on code originally written by: @yinyuefengyi (luo xionghu)
Differential Revision: https://reviews.llvm.org/D54994
llvm-svn: 348151
Add '-k 1' to 'sort -b' calls in SimpleLoopUnswitch tests, as required
for sort implementation on NetBSD. The '-b' modifier is ineffective
if specified without any key. Per the manpage:
Note that the -b option has no effect unless key fields are specified.
Differential Revision: https://reviews.llvm.org/D55168
llvm-svn: 348097
We were adding the entire scalarization extraction cost for reductions, which returns the total cost of extracting every element of a vector type.
For reductions we don't need to do this - we just need to extract the 0'th element after the reduction pattern has completed.
Fixes PR37731
Rebased and reapplied after being reverted in rL347541 due to PR39774 - which was fixed by D54955/rL347759 and D55017/rL347997
Differential Revision: https://reviews.llvm.org/D54585
llvm-svn: 348076
Extend ssub.sat(X, C) -> sadd.sat(X, -C) canonicalization to also
support non-splat vector constants. This is done by generalizing
the implementation of the isNotMinSignedValue() helper to return
true for constants that are non-splat, but don't contain any
signed min elements.
Differential Revision: https://reviews.llvm.org/D55011
llvm-svn: 348072
Summary:
Follow up to D54270, which allowed importing of var args functions
unless they called va_start. As pointed out in the post-commit comments
on that patch, the inliner can handle functions that call va_start in
certain situations as well. Go ahead and enable importing of all var
args functions. Measurements on a large binary show that this increases
imports and binary size by an insignificant amount.
Reviewers: davidxl
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D54607
llvm-svn: 348068
These tests should probably go under a separate test file because they
should fold with just -constprop, but they're similar to the scalar
tests already in here.
llvm-svn: 348045
Summary:
When mem2reg inserts phi nodes in blocks with unreachable predecessors,
it adds undef operands for those incoming edges. When there are
multiple such predecessors, the order is currently based on the address
of the BasicBlocks. This change fixes that by using the BBNumbers in
the sort/search predicates, as is done elsewhere in mem2reg to ensure
determinism.
Also adds a testcase with a bunch of unreachable preds, which
(nodeterministically) fails without the fix.
Reviewers: majnemer
Reviewed By: majnemer
Subscribers: mgrang, llvm-commits
Differential Revision: https://reviews.llvm.org/D55077
llvm-svn: 348024
Summary:
An additional fix for PR39774. Need to update the references for the
RedcutionRoot instruction when it is replaced during the vectorization
phase to avoid compiler crash on reduction vectorization.
Reviewers: RKSimon, spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D55017
llvm-svn: 347997
Adding a new reduction pattern match for vectorizing code similar
to TSVC s3111:
for (int i = 0; i < N; i++)
if (a[i] > b)
sum += a[i];
This patch adds support for fadd, fsub and fmull, as well as multiple
branches and different (but compatible) instructions (ex. add+sub) in
different branches.
The difference from the previous patch(https://reviews.llvm.org/D49168)
is as follows:
- Added check of fast-math property of fp-instruction to the
previous patch
- Fix/add some pattern for if-reduction.ll
Differential Revision: https://reviews.llvm.org/D54464
Patch by Takahiro Miyoshi <takahiro.miyoshi@linaro.org>
and Masakazu Ueno <masakazu.ueno@linaro.org>
llvm-svn: 347989
Terminator folding transform lacks MemorySSA update for memory Phis,
while they exist within MemorySSA analysis. They need exactly the same
type of updates as regular Phis. Failing to update them properly ends up
with inconsistent MemorySSA and manifests in various assertion failures.
This patch adds Memory Phi updates to this transform.
Thanks to @jonpa for finding this!
Differential Revision: https://reviews.llvm.org/D55050
Reviewed By: asbirlea
llvm-svn: 347979
r320789 suppressed moving the insertion point of SCEV expressions with
dev/rem operations to the loop header in non-loop-invariant situations.
This, and similar, hoisting is also unsafe in the loop-invariant case,
since there may be a guard against a zero denominator. This is an
adjustment to the fix of r320789 to suppress the movement even in the
loop-invariant case.
This fixes PR30806.
Differential Revision: https://reviews.llvm.org/D54713
llvm-svn: 347934
Also revert fix r347876
One of the buildbots was reporting a failure in some relevant tests that I can't
repro or explain at present, so reverting until I can isolate.
llvm-svn: 347911
This is an almost direct move of the functionality from InstCombine to
InstSimplify. There's no reason not to do this in InstSimplify because
we never create a new value with this transform.
(There's a question of whether any dominance-based transform belongs in
either of these passes, but that's a separate issue.)
I've changed 1 of the conditions for the fold (1 of the blocks for the
branch must be the block we started with) into an assert because I'm not
sure how that could ever be false.
We need 1 extra check to make sure that the instruction itself is in a
basic block because passes other than InstCombine may be using InstSimplify
as an analysis on values that are not wired up yet.
The 3-way compare changes show that InstCombine has some kind of
phase-ordering hole. Otherwise, we would have already gotten the intended
final result that we now show here.
llvm-svn: 347896
This commit caused a large compile-time slowdown in some cases when NDEBUG is
off due to the dominator tree verification it added. Fix this by only doing
dominator tree and loop info verification when something has been hoisted.
Differential Revision: https://reviews.llvm.org/D52827
llvm-svn: 347889
Summary:
When splitting musttail calls, the split blocks' original terminators
get removed; inform the DTU when this happens.
Also add a testcase that fails an assertion in the DTU without this fix.
Reviewers: fhahn, junbuml
Reviewed By: fhahn
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D55027
llvm-svn: 347872
TFE and LWE support requires extra result registers that are written in the
event of a failure in order to detect that failure case.
The specific use-case that initiated these changes is sparse texture support.
This means that if image intrinsics are used with either option turned on, the
programmer must ensure that the return type can contain all of the expected
results. This can result in redundant registers since the vector size must be a
power-of-2.
This change takes roughly 6 parts:
1. Modify the instruction defs in tablegen to add new instruction variants that
can accomodate the extra return values.
2. Updates to lowerImage in SIISelLowering.cpp to accomodate setting TFE or LWE
(where the bulk of the work for these instruction types is now done)
3. Extra verification code to catch cases where intrinsics have been used but
insufficient return registers are used.
4. Modification to the adjustWritemask optimisation to account for TFE/LWE being
enabled (requires extra registers to be maintained for error return value).
5. An extra pass to zero initialize the error value return - this is because if
the error does not occur, the register is not written and thus must be zeroed
before use. Also added a new (on by default) option to ensure ALL return values
are zero-initialized that is required for sparse texture support.
6. Disable the inst_combine optimization in the presence of tfe/lwe (later TODO
for this to re-enable and handle correctly).
There's an additional fix now to avoid a dmask=0
For an image intrinsic with tfe where all result channels except tfe
were unused, I was getting an image instruction with dmask=0 and only a
single vgpr result for tfe. That is incorrect because the hardware
assumes there is at least one vgpr result, plus the one for tfe.
Fixed by forcing dmask to 1, which gives the desired two vgpr result
with tfe in the second one.
The TFE or LWE result is returned from the intrinsics using an aggregate
type. Look in the test code provided to see how this works, but in essence IR
code to invoke the intrinsic looks as follows:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15,
i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0
%v.err = extractvalue {<4 x float>, i32} %v, 1
Differential revision: https://reviews.llvm.org/D48826
Change-Id: If222bc03642e76cf98059a6bef5d5bffeda38dda
llvm-svn: 347871
This reverts commits r347776 and r347778.
The first one, r347776, caused significant compile time regressions
for certain input files, see PR39836 for details.
llvm-svn: 347867
Lack of an attribute means that the function hasn't been checked for what vector width it requires. So if the caller or the callee doesn't have the attribute we should make sure the combined function after inlining does not have the attribute.
If the caller already doesn't have the attribute we can just avoid adding it. Otherwise if the callee doesn't have the attribute just remove the caller's attribute.
llvm-svn: 347841
Packing the flags into one bitcode word will save effort in
adding new flags in the future.
Differential Revision: https://reviews.llvm.org/D54755
llvm-svn: 347806
In PR39807 we incorrectly handle circumstances where calls are common'd
from conditional blocks into the parent BB. Calls that can be inlined
must always have DebugLocs, however we strip them during commoning, which
the IR verifier asserts on.
Fix this by using applyMergedLocation: it will perform the same DebugLoc
stripping of conditional Locs, but will also generate an unknown location
DebugLoc that satisfies the requirement for inlinable calls to always have
locations.
Some of the prior logic for selecting a DebugLoc is now likely redundant;
I'll generate a follow-up to remove it (involves editing more regression
tests).
Differential Revision: https://reviews.llvm.org/D54997
llvm-svn: 347782
This commit caused failures because it failed to correctly handle cases where
we hoist a phi, then hoist a use of that phi, then have to rehoist that use. We
need to make sure that we rehoist the use to _after_ the hoisted phi, which we
do by always rehoisting to the immediate dominator instead of just rehoisting
everything to the original preheader.
An option is also added to control whether control flow is hoisted, which is
off in this commit but will be turned on in a subsequent commit.
Differential Revision: https://reviews.llvm.org/D52827
llvm-svn: 347776
Combine
sat(sat(X + C1) + C2) -> sat(X + (C1+C2))
and
sat(sat(X - C1) - C2) -> sat(X - (C1+C2))
if the sign of C1 and C2 matches.
In the unsigned case we can compute C1+C2 with saturating arithmetic,
and InstSimplify will reduce this just to the saturation value. For
the signed case, we cannot perform the simplification if the result
of the addition overflows.
This change is part of https://reviews.llvm.org/D54534.
llvm-svn: 347773
Canonicalize ssub.sat(X, C) to ssub.sat(X, -C) if C is constant and
not signed minimum. This will help further optimizations to apply.
This change is part of https://reviews.llvm.org/D54534.
llvm-svn: 347772
Always-overflow was already determined for unsigned addition, but
not subtraction. This patch establishes parity.
This allows us to perform some additional simplifications for
signed saturating subtractions.
This change is part of https://reviews.llvm.org/D54534.
llvm-svn: 347771
If ValueTracking can determine that the add/sub can newer overflow,
replace it with the corresponding nuw/nsw add/sub.
Additionally, for the unsigned case, if ValueTracking determines
that the add/sub always overflows, replace the result with the
saturation value.
This change is part of https://reviews.llvm.org/D54534.
llvm-svn: 347770
If a saturating add intrinsic has one constant argument, make sure
it is on the RHS. This will simplify further transformations.
This change is part of https://reviews.llvm.org/D54534.
llvm-svn: 347769
Summary:
If the original reduction root instruction was vectorized, it might be
removed from the tree. It means that the insertion point may become
invalidated and the whole vectorization of the reduction leads to the
incorrect output result.
The ReductionRoot instruction must be marked as externally used so it
could not be removed. Otherwise it might cause inconsistency with the
cost model and we may end up with too optimistic optimization.
Reviewers: RKSimon, spatel, hfinkel, mkuper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D54955
llvm-svn: 347759
InlineCost also treats them as free and the current implementation
can cause assertion failures if PHI nodes are moved outside the region
from entry BBs to the region.
It also updates the code to use the instructionsWithoutDebug iterator.
Reviewers: davidxl, davide, vsk, graham-yiu-huawei
Reviewed By: davidxl
Differential Revision: https://reviews.llvm.org/D54748
llvm-svn: 347683
It fixes a bug that doesn't update Phi inputs of the only live successor that
is in the list of block's successors more than once.
Thanks @uabelho for finding this.
Differential Revision: https://reviews.llvm.org/D54849
Reviewed By: anna
llvm-svn: 347640
Summary:
Removing ncompatible attributes at indirect-call promoted callsites, not removing it results in
at least a IR verification error.
Reviewers: davidxl, xur, mssimpso
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D54913
llvm-svn: 347605
This reverts commit r346970.
It was causing PR39774, a crash in slp-vectorizer on a rather simple loop
with just a bunch of 'and's in the body.
llvm-svn: 347541
OriginalOp of a Predicate refers to the original IR value,
before renaming. While solving in IPSCCP, we have to use
the operand of the ssa_copy instead, to avoid missing
updates for nested conditions on the same IR value.
Fixes PR39772.
llvm-svn: 347524
Support funnel shifts in InstCombine demanded bits simplification.
If the shift amount is constant, we can determine both the demanded
bits of the operands, as well as the known bits of the result.
If one of the operands has no demanded bits, it will be replaced
by undef and the funnel shift will be simplified into a simple shift
due to the simplifications added in D54778.
Differential Revision: https://reviews.llvm.org/D54869
llvm-svn: 347515
This changeset is modeled after Intel's submission for SVML. It enables
trigonometry functions vectorization via SLEEF: http://sleef.org/.
* A new vectorization library enum is added to TargetLibraryInfo.h: SLEEF.
* A new option is added to TargetLibraryInfoImpl - ClVectorLibrary: SLEEF.
* A comprehensive test case is included in this changeset.
* In a separate changeset (for clang), a new vectorization library argument is
added to -fveclib: -fveclib=SLEEF.
Trigonometry functions that are vectorized by sleef:
acos
asin
atan
atanh
cos
cosh
exp
exp2
exp10
lgamma
log10
log2
log
sin
sinh
sqrt
tan
tanh
tgamma
Patch by Stefan Teleman
Differential Revision: https://reviews.llvm.org/D53927
llvm-svn: 347510
The following simplifications are implemented:
* `fshl(X, 0, C) -> shl X, C%BW`
* `fshl(X, undef, C) -> shl X, C%BW` (assuming undef = 0)
* `fshl(0, X, C) -> lshr X, BW-C%BW`
* `fshl(undef, X, C) -> lshr X, BW-C%BW` (assuming undef = 0)
* `fshr(X, 0, C) -> shl X, (BW-C%BW)`
* `fshr(X, undef, C) -> shl X, BW-C%BW` (assuming undef = 0)
* `fshr(0, X, C) -> lshr X, C%BW`
* `fshr(undef, X, C) -> lshr, X, C%BW` (assuming undef = 0)
The simplification is only performed if the shift amount C is constant,
because we can explicitly compute C%BW and BW-C%BW in this case.
Differential Revision: https://reviews.llvm.org/D54778
llvm-svn: 347505
When removing edges, we also update Phi inputs and may end up removing
a Phi if it has only one input. We should not do it for edges that leave the current
loop because these Phis are LCSSA Phis and need to be preserved.
Thanks @dmgreen for finding this!
Differential Revision: https://reviews.llvm.org/D54841
llvm-svn: 347484
The MergeFunctions pass was originally intended to emit aliases
instead of thunks where possible (unnamed_addr). However, for a
long time this functionality was behind a flag hardcoded to false,
bitrotted and was eventually removed in r309313.
Originally the functionality was first disabled in r108417 due to
lack of support for aliases in Mach-O. I believe that this is no
longer the case nowadays, but not really familiar with this area.
In the interest of being conservative, this patch reintroduces the
aliasing functionality behind a default disabled -mergefunc-use-aliases
flag.
Differential Revision: https://reviews.llvm.org/D53285
llvm-svn: 347407
LVI was symbolically executing binary operators only when the RHS was
constant, missing the case where we have a ConstantRange for the RHS,
but not an actual constant. Tested using check-all and by
bootstrapping. Compile time is not impacted measurably.
Differential Revision: https://reviews.llvm.org/D19859
llvm-svn: 347379
These are part of D54666, so adding them here before the patch to
show the baseline (currently unoptimized) results.
Patch by: @nikic (Nikita Popov)
llvm-svn: 347331
This patch fixes PR39695.
The original LoopSink only considers memory alias in loop body. But PR39695 shows that instructions following sink candidate in preheader should also be checked. This is a conservative patch, it simply adds whole preheader block to alias set. It may lose some optimization opportunity, but I think that is very rare because: 1 in the most common case st/ld to the same address, the load should already be optimized away. 2 usually preheader is not very large.
Differential Revision: https://reviews.llvm.org/D54659
llvm-svn: 347325
This patch fixes the issue noticed in D54532.
The problem is that cst_pred_ty-based matchers like m_Zero() currently do not match
scalar undefs (as expected), but *do* match vector undefs. This may lead to optimization
inconsistencies in rare cases.
There is only one existing test for which output changes, reverting the change from D53205.
The reason here is that vector fsub undef, %x is no longer matched as an m_FNeg(). While I
think that the new output is technically worse than the previous one, it is consistent with
scalar, and I don't think it's really important either way (generally that undef should have
been folded away prior to reassociation.)
I've also added another test case for this issue based on InstructionSimplify. It took some
effort to find that one, as in most cases undef folds are either checked first -- and in the
cases where they aren't it usually happens to not make a difference in the end. This is the
only case I was able to come up with. Prior to this patch the test case simplified to undef
in the scalar case, but zeroinitializer in the vector case.
Patch by: @nikic (Nikita Popov)
Differential Revision: https://reviews.llvm.org/D54631
llvm-svn: 347318
The initial version of patch lacked Phi nodes updates in destinations of removed
edges. This version contains this update and tests on this situation.
Differential Revision: https://reviews.llvm.org/D54021
llvm-svn: 347289
Summary:
Currently, when vectorizing stores to uniform addresses, the only
instance we prevent vectorization is if there are multiple stores to the
same uniform address causing an unsafe dependency.
This patch teaches LAA to avoid vectorizing loops that have an unsafe
cross-iteration dependency between a load and a store to the same uniform address.
Fixes PR39653.
Reviewers: Ayal, efriedma
Subscribers: rkruppe, llvm-commits
Differential Revision: https://reviews.llvm.org/D54538
llvm-svn: 347220
The general approach taken is to make note of loop invariant branches, then when
we see something conditional on that branch, such as a phi, we create a copy of
the branch and (empty versions of) its successors and hoist using that.
This has no impact by itself that I've been able to see, as LICM typically
doesn't see such phis as they will have been converted into selects by the time
LICM is run, but once we start doing phi-to-select conversion later it will be
important.
Differential Revision: https://reviews.llvm.org/D52827
llvm-svn: 347190
This patch introduces infrastructure and the simplest case for constant-folding
of branch and switch instructions within loop into unconditional branches.
It is useful as a cleanup for such passes as loop unswitching that sometimes
produce such branches.
Only the simplest case supported in this patch: after the folding, no block
should become dead or stop being part of the loop. Support for more
sophisticated cases will go separately in follow-up patches.
Differential Revision: https://reviews.llvm.org/D54021
Reviewed By: anna
llvm-svn: 347183
Fix all of the missing debug location errors in CVP found by debugify.
This includes the missing-location-after-udiv-truncation case described
in llvm.org/PR38178.
llvm-svn: 347147
We need to control exponential behavior of loop-unswitch so we do not get
run-away compilation.
Suggested solution is to introduce a multiplier for an unswitch cost that
makes cost prohibitive as soon as there are too many candidates and too
many sibling loops (meaning we have already started duplicating loops
by unswitching).
It does solve the currently known problem with compile-time degradation
(PR 39544).
Tests are built on top of a recently implemented CHECK-COUNT-<num>
FileCheck directives.
Reviewed By: chandlerc, mkazantsev
Differential Revision: https://reviews.llvm.org/D54223
llvm-svn: 347097
We were adding the entire scalarization extraction cost for reductions, which returns the total cost of extracting every element of a vector type.
For reductions we don't need to do this - we just need to extract the 0'th element after the reduction pattern has completed.
Fixes PR37731
Differential Revision: https://reviews.llvm.org/D54585
llvm-svn: 346970
There's a potential miscompile here. It's unlikely in the real
world because this transform is guarded with shouldChangeType(),
but this test file doesn't include a standard data-layout for
some reason (despite including a custom 1), so we can see the bug.
llvm-svn: 346966
This is a problem seen in common rotate idioms as noted in:
https://bugs.llvm.org/show_bug.cgi?id=34924
Note that we are not canonicalizing standard IR (shifts and logic) to the intrinsics yet.
(Although I've written this before...) I think this is the last step before we enable
that transform. Ie, we could regress code by doing that transform without this
simplification in place.
In PR34924, I questioned whether this is a valid transform for target-independent IR,
but I convinced myself this is ok. If we're speculating a funnel shift by turning cmp+br
into select, then SimplifyCFG has already determined that the transform is justified.
It's possible that SimplifyCFG is not taking into account profile or other metadata,
but if that's true, then it's a bug independent of funnel shifts.
Also, we do have CGP code to restore a guard like this around an intrinsic if it can't
be lowered cheaply. But that isn't necessary for funnel shift because the default
expansion in SelectionDAGBuilder includes this same cmp+select.
Differential Revision: https://reviews.llvm.org/D54552
llvm-svn: 346960
A call to @llvm.trap can be expected to be cold (i.e. unlikely to be
reached in a normal program execution).
Outlining paths which unconditionally trap is an important memory
saving. As the hot/cold splitting pass (imho) should not treat all
noreturn calls as cold, explicitly mark @llvm.trap cold so that it can
be outlined.
Split out of https://reviews.llvm.org/D54244.
Differential Revision: https://reviews.llvm.org/D54329
llvm-svn: 346885
Summary:
Previously we marked all vararg functions as non-inlinable in the
function summary, which prevented their importing. However, the
corresponding inliner restriction was loosened in r321940/r342675
to only apply to functions calling va_start. Adjust the summary
flag computation to match.
Reviewers: davidxl
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D54270
llvm-svn: 346883
Summary:
These asserts are based on the assumption that the order of true/false operands in a select and those in the compare would always be the same.
This fixes PR39595.
Reviewers: craig.topper, spatel, dmgreen
Reviewed By: craig.topper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D54359
llvm-svn: 346874
Regenerate this test using update_test_checks.py in preparation for an
upcomming commit, to make it not depend on the names of instructions.
llvm-svn: 346869
The shift amount of a funnel shift is modulo the scalar bitwidth:
http://llvm.org/docs/LangRef.html#llvm-fshl-intrinsic
...so we can use demanded bits analysis on that operand to simplify it
when we have a power-of-2 bitwidth.
This is another step towards canonicalizing {shift/shift/or} to the
intrinsics in IR.
Differential Revision: https://reviews.llvm.org/D54478
llvm-svn: 346814
LoopUtils.cpp contains a utility that splits an loop exit block, so that the new block contains only edges coming from the loop. In the case of nested loops, the exit path for the inner loop might also be the back-edge of the outer loop. The new block which is inserted on this path, is now a latch for the outer loop, and it needs to hold the loop metadata for the outer loop. (The test case gives a more concrete view of the situation.)
Patch by Chang Lin (clin1)
Differential Revision: https://reviews.llvm.org/D53876
llvm-svn: 346810
The cmp+branch variant of this pattern is shown in:
https://bugs.llvm.org/show_bug.cgi?id=34924
...and as discussed there, we probably can't transform
that without a rotate intrinsic. We do have that now
via funnel shift, but we're not quite ready to
canonicalize IR to that form yet. The case with 'select'
should already be transformed though, so that's this patch.
The sequence with negation followed by masking is what we
use in the backend and partly in clang (though that part
should be updated).
https://rise4fun.com/Alive/TplC
%cmp = icmp eq i32 %shamt, 0
%sub = sub i32 32, %shamt
%shr = lshr i32 %x, %shamt
%shl = shl i32 %x, %sub
%or = or i32 %shr, %shl
%r = select i1 %cmp, i32 %x, i32 %or
=>
%neg = sub i32 0, %shamt
%masked = and i32 %shamt, 31
%maskedneg = and i32 %neg, 31
%shl2 = lshr i32 %x, %masked
%shr2 = shl i32 %x, %maskedneg
%r = or i32 %shl2, %shr2
llvm-svn: 346807
The IEEE-754 Standard makes it clear that fneg(x) and
fsub(-0.0, x) are two different operations. The former is a bitwise
operation, while the latter is an arithmetic operation. This patch
creates a dedicated FNeg IR Instruction to model that behavior.
Differential Revision: https://reviews.llvm.org/D53877
llvm-svn: 346774
This patch updates DuplicateInstructionsInSplitBetween to update a DTU
instead of applying updates to the DT directly.
Given that there only are 2 users, also updated them in this patch to
avoid churn.
I slightly moved the code in CallSiteSplitting around to reduce the
places where we have to pass in DTU. If necessary, I could split those
changes in a separate patch.
This fixes missing DT updates when dealing with musttail calls in
CallSiteSplitting, by using DTU->deleteBB.
Reviewers: junbuml, kuhar, NutshellySima, indutny, brzycki
Reviewed By: NutshellySima
llvm-svn: 346769
This is a longer variant for the pattern handled in
rL346713
This one includes zexts.
Eventually, we should canonicalize all rotate patterns
to the funnel shift intrinsics, but we need a bit more
infrastructure to make sure the vectorizers handle those
intrinsics as well as the shift+logic ops.
https://rise4fun.com/Alive/FMn
Name: narrow rotateright
%neg = sub i8 0, %shamt
%rshamt = and i8 %shamt, 7
%rshamtconv = zext i8 %rshamt to i32
%lshamt = and i8 %neg, 7
%lshamtconv = zext i8 %lshamt to i32
%conv = zext i8 %x to i32
%shr = lshr i32 %conv, %rshamtconv
%shl = shl i32 %conv, %lshamtconv
%or = or i32 %shl, %shr
%r = trunc i32 %or to i8
=>
%maskedShAmt2 = and i8 %shamt, 7
%negShAmt2 = sub i8 0, %shamt
%maskedNegShAmt2 = and i8 %negShAmt2, 7
%shl2 = lshr i8 %x, %maskedShAmt2
%shr2 = shl i8 %x, %maskedNegShAmt2
%r = or i8 %shl2, %shr2
llvm-svn: 346716
The sub-pattern for the shift amount in a rotate can take on
several different forms, and there's apparently no way to
canonicalize those without seeing the entire rotate sequence.
This is the form noted in:
https://bugs.llvm.org/show_bug.cgi?id=39624https://rise4fun.com/Alive/qnT
%zx = zext i8 %x to i32
%maskedShAmt = and i32 %shAmt, 7
%shl = shl i32 %zx, %maskedShAmt
%negShAmt = sub i32 0, %shAmt
%maskedNegShAmt = and i32 %negShAmt, 7
%shr = lshr i32 %zx, %maskedNegShAmt
%rot = or i32 %shl, %shr
%r = trunc i32 %rot to i8
=>
%truncShAmt = trunc i32 %shAmt to i8
%maskedShAmt2 = and i8 %truncShAmt, 7
%shl2 = shl i8 %x, %maskedShAmt2
%negShAmt2 = sub i8 0, %truncShAmt
%maskedNegShAmt2 = and i8 %negShAmt2, 7
%shr2 = lshr i8 %x, %maskedNegShAmt2
%r = or i8 %shl2, %shr2
llvm-svn: 346713
This just identifies the intrinsics as candidates for vectorization.
It does not mean we will attempt to vectorize under normal conditions
(the test file is forcing vectorization).
The cost model must be fixed to show that the transform is profitable
in general.
Allowing vectorization with these intrinsics is required to avoid
potential regressions from canonicalizing to the intrinsics from
generic IR:
https://bugs.llvm.org/show_bug.cgi?id=37417
llvm-svn: 346661
This patch relaxes overconservative checks on whether or not we could write
memory before we execute an instruction. This allows us to hoist guards out of
loops even if they are not in the header block.
Differential Revision: https://reviews.llvm.org/D50891
Reviewed By: fedor.sergeev
llvm-svn: 346643
ComputeValueKnownInPredecessors has a "visited" set to prevent infinite
loops, since a value can be visited more than once. However, the
implementation didn't prevent the algorithm from taking exponential
time. Instead of removing elements from the RecursionSet one at a time,
we should keep around the whole set until
ComputeValueKnownInPredecessors finishes, then discard it.
The testcase is synthetic because I was having trouble effectively
reducing the original. But it's basically the same idea.
Instead of failing, we could theoretically cache the result instead.
But I don't think it would help substantially in practice.
Differential Revision: https://reviews.llvm.org/D54239
llvm-svn: 346562
After D45330, Dominators are required for IPSCCP and can be preserved.
This patch preserves DominatorTreeAnalysis in the new pass manager. AFAIK the legacy pass manager cannot preserve function analysis required by a module analysis.
Reviewers: davide, dberlin, chandlerc, efriedma, kuhar, NutshellySima
Reviewed By: chandlerc, kuhar, NutshellySima
Differential Revision: https://reviews.llvm.org/D47259
llvm-svn: 346486
We can stop recording conditions once we reached the immediate dominator
for the block containing the call site. Conditions in predecessors of the
that node will be the same for all paths to the call site and splitting
is not beneficial.
This patch makes CallSiteSplitting dependent on the DT anlysis. because
the immediate dominators seem to be the easiest way of finding the node
to stop at.
I had to update some exiting tests, because they were checking for
conditions that were true/false on all paths to the call site. Those
should now be handled by instcombine/ipsccp.
Reviewers: davide, junbuml
Reviewed By: junbuml
Differential Revision: https://reviews.llvm.org/D44627
llvm-svn: 346483
This patch adds logic to detect reductions across the inner and outer
loop by following the incoming values of PHI nodes in the outer loop. If
the incoming values take part in a reduction in the inner loop or come
from outside the outer loop, we found a reduction spanning across inner
and outer loop.
With this change, ~10% more loops are interchanged in the LLVM
test-suite + SPEC2006.
Fixes https://bugs.llvm.org/show_bug.cgi?id=30472
Reviewers: mcrosier, efriedma, karthikthecool, davide, hfinkel, dmgreen
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D43245
llvm-svn: 346438
Summary:
This fixes PR 37422
In ELF, non-weak symbols can also be non-prevailing. In this particular
PR, the __llvm_profile_* symbols are non-prevailing but weren't getting
dropped - causing multiply-defined errors with lld.
Also add a test, strong_non_prevailing.ll, to ensure that multiple
copies of a strong symbol are dropped.
To fix the test regressions exposed by this fix,
- do not mark prevailing copies for symbols with 'appending' linkage.
There's no one prevailing copy for such symbols.
- fix the prevailing version in dead-strip-fulllto.ll
- explicitly pass exported symbols to llvm-lto in fumcimport.ll and
funcimport_var.ll
Reviewers: tejohnson, pcc
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith,
dang, srhines, llvm-commits
Differential Revision: https://reviews.llvm.org/D54125
llvm-svn: 346436
Summary:
When the 3rd argument to these intrinsics is zero, lowering them
to shift instructions produces poison values, since we end up with
shift amounts equal to the number of bits in the shifted value. This
means we can only lower these intrinsics if we can prove that the
3rd argument is not zero.
Reviewers: arsenm
Reviewed By: arsenm
Subscribers: bnieuwenhuizen, jvesely, wdng, nhaehnle, llvm-commits
Differential Revision: https://reviews.llvm.org/D53739
llvm-svn: 346422
This eliminates the outlining penalty for llvm.trap/unreachable, because
callers no longer have to emit cleanup/ret instructions after calling an
outlined `noreturn` function.
rdar://45523626
llvm-svn: 346421
The patch has been reverted because it ended up prohibiting propagation
of a constant to exit value. For such values, we should skip all checks
related to hard uses because propagating a constant is always profitable.
Differential Revision: https://reviews.llvm.org/D53691
llvm-svn: 346397
LSR reassociates constants as unfolded offsets when the constants fit as
immediate add operands, which currently prevents such constants from being
combined later with loop invariant registers.
This patch modifies GenerateCombinations() to generate a second formula which
includes the unfolded offset in the combined loop-invariant register.
This commit fixes a bug in the original patch (committed at r345114, reverted
at r345123).
Differential Revision: https://reviews.llvm.org/D51861
llvm-svn: 346390
Summary:
MergeFunctions currently tries to process strong functions before
weak functions, because weak functions can simply call strong
functions, while a strong/weak function cannot call a weak function
(a backing strong function is needed).
This patch additionally tries to process external functions before
local functions, because we definitely have to keep the external
function, but may be able to drop the local one (and definitely
can if it is also unnamed_addr).
Unfortunately, this exposes an existing bug in the implementation:
The FnTree and FNodesInTree structures can currently go out of
sync in the case where two weak functions are merged, because the
function in FnTree/FNodesInTree is RAUWed. This leaves it behind in
FnTree (this is intended, as it is the strong backing function which
should be used for further merges), while it is replaced in
FNodesInTree (this is not intended).
This is fixed by switching FNodesInTree from using a ValueMap to
using a DenseMap of AssertingVH.
This exposes another minor issue: Currently FNodesInTree is not
cleared after MergeFunctions finishes running. Currently, this is
potentially dangerous (e.g. if something else wants to RAUW a function
with a non-function), but at the very least it is unnecessary/inefficient.
After the change to use AssertingVH it becomes more problematic,
because there are certainly passes that remove functions.
This issue is fixed by clearing FNodesInTree at the end of the pass.
Reviewers: jfb, whitequark
Reviewed By: whitequark
Subscribers: rkruppe, llvm-commits
Differential Revision: https://reviews.llvm.org/D53271
llvm-svn: 346386
Summary:
For unnamed_addr functions we RAUW instead of only replacing direct callers. However, functions in which replacements were performed currently are not added back to the worklist, resulting in missed merging opportunities.
Fix this by calling removeUsers() prior to RAUW.
Reviewers: jfb, whitequark
Reviewed By: whitequark
Subscribers: rkruppe, llvm-commits
Differential Revision: https://reviews.llvm.org/D53262
llvm-svn: 346385
If all the edge counts for a function are zero, skip count population and
annotation, as nothing will happen. This can save some compile time.
Differential Revision: https://reviews.llvm.org/D54212
llvm-svn: 346370
When partial unswitch operates on multiple conditions at once, .e.g:
if (Cond1 || Cond2 || NonInv) ...
it should infer (and replace) values for individual conditions only on one
side of unswitch and not another.
More precisely only these derivations hold true:
(Cond1 || Cond2) == false => Cond1 == Cond2 == false
(Cond1 && Cond2) == true => Cond1 == Cond2 == true
By the way we organize unswitching it means only replacing on "continue" blocks
and never on "unswitched" ones. Since trivial unswitch does not have "unswitched"
blocks it does not have this problem.
Fixes PR 39568.
Reviewers: chandlerc, asbirlea
Differential Revision: https://reviews.llvm.org/D54211
llvm-svn: 346350
If we simplify an instruction to itself, we do not need to add a user to
itself. For congruence classes with a defining expression, we already
use a similar logic.
Fixes PR38259.
Reviewers: davide, efriedma, mcrosier
Reviewed By: davide
Differential Revision: https://reviews.llvm.org/D51168
llvm-svn: 346335
By morphing the instruction rather than deleting and creating a new one,
we retain fast-math-flags and potentially other metadata (profile info?).
llvm-svn: 346331
This adds the llvm-side support for post-inlining evaluation of the
__builtin_constant_p GCC intrinsic.
Also fixed SCCPSolver::visitCallSite to not blow up when seeing a call
to a function where canConstantFoldTo returns true, and one of the
arguments is a struct.
Updated from patch initially by Janusz Sobczak.
Differential Revision: https://reviews.llvm.org/D4276
llvm-svn: 346322
The sibling fold for 'oge' --> 'ord' was already here,
but this half was missing.
The result of fabs() must be positive or nan, so asking
if the result is negative or nan is the same as asking
if the result is nan.
This is another step towards fixing:
https://bugs.llvm.org/show_bug.cgi?id=39475
llvm-svn: 346321
The lowering for a call to eh_typeid_for changes when it's moved from
one function to another.
There are several proposals for fixing this issue in llvm.org/PR39545.
Until some solution is in place, do not allow CodeExtractor to extract
calls to eh_typeid_for, as that results in serious miscompilations.
llvm-svn: 346256
When CodeExtractor moves instructions to a new function, debug
intrinsics referring to those instructions within the parent function
become invalid.
This results in the same verifier failure which motivated r344545, about
function-local metadata being used in the wrong function.
llvm-svn: 346255
This is another part of solving PR39475:
https://bugs.llvm.org/show_bug.cgi?id=39475
This might be enough to fix that particular issue, but as noted
with the FIXME, we're still dropping FMF on other folds around here.
llvm-svn: 346234
This patch makes LICM use `ICFLoopSafetyInfo` that is a smarter version
of LoopSafetyInfo that leverages power of Implicit Control Flow Tracking
to keep track of throwing instructions and give less pessimistic answers
to queries related to throws.
The ICFLoopSafetyInfo itself has been introduced in rL344601. This patch
enables it in LICM only.
Differential Revision: https://reviews.llvm.org/D50377
Reviewed By: apilipenko
llvm-svn: 346201
This reverts commit 2f425e9c7946b9d74e64ebbfa33c1caa36914402.
It seems that the check that we still should do the transform if we
know the result is constant is missing in this code. So the logic that
has been deleted by this change is still sometimes accidentally useful.
I revert the change to see what can be done about it. The motivating
case is the following:
@Y = global [400 x i16] zeroinitializer, align 1
define i16 @foo() {
entry:
br label %for.body
for.body: ; preds = %entry, %for.body
%i = phi i16 [ 0, %entry ], [ %inc, %for.body ]
%arrayidx = getelementptr inbounds [400 x i16], [400 x i16]* @Y, i16 0, i16 %i
store i16 0, i16* %arrayidx, align 1
%inc = add nuw nsw i16 %i, 1
%cmp = icmp ult i16 %inc, 400
br i1 %cmp, label %for.body, label %for.end
for.end: ; preds = %for.body
%inc.lcssa = phi i16 [ %inc, %for.body ]
ret i16 %inc.lcssa
}
We should be able to figure out that the result is constant, but the patch
breaks it.
Differential Revision: https://reviews.llvm.org/D51584
llvm-svn: 346198
This is NFCI for InstCombine because it calls InstSimplify,
so I left the tests for this transform there. As noted in
the code comment, we can allow this fold more often by using
FMF and/or value tracking.
llvm-svn: 346169
Summary:
This patch prevents MergeICmps to performn the transformation if the address operand GEP of the load instruction has a use outside of the load's parent block. Without this patch, compiler crashes with the given test case because the use of `%first.i` is still around when the basic block is erased from https://github.com/llvm-mirror/llvm/blob/master/lib/Transforms/Scalar/MergeICmps.cpp#L620. I think checking `isUsedOutsideOfBlock` with `GEP` is the original intention of the code, as the checking for `LoadI` is already performed in the same function.
This patch is incomplete though, as this makes the pass overly conservative and fails the test `tuple-four-int8.ll`. I believe what needs to be done is checking if GEP has a use outside of block that is not the part of "Comparisons" chain. Submit the patch as of now to prevent compiler crash.
Reviewers: courbet, trentxintong
Reviewed By: courbet
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D54089
llvm-svn: 346151
There was no coverage for at least 2 out of the 4 patterns because
of fcmp canonicalization. The tests and code should be moved to
InstSimplify in a follow-up because this doesn't create any new values.
llvm-svn: 346150
As stated in IEEE-754 and discussed in:
https://bugs.llvm.org/show_bug.cgi?id=38086
...the sign of zero does not affect any FP compare predicate.
Known regressions were fixed with:
rL346097 (D54001)
rL346143
The transform will help reduce pattern-matching complexity to solve:
https://bugs.llvm.org/show_bug.cgi?id=39475
...as well as improve CSE and codegen (a zero constant is almost always
easier to produce than 0x80..00).
llvm-svn: 346147
It looks like we correctly removed edge cases with 0.0 from D50714,
but we were a bit conservative because getBinOpIdentity() doesn't
distinguish between +0.0 and -0.0 and 'nsz' is effectively always
true for fcmp (see discussion in:
https://bugs.llvm.org/show_bug.cgi?id=38086
Without this change, we would get regressions by canonicalizing
to +0.0 in all fcmp, and that's a step towards solving:
https://bugs.llvm.org/show_bug.cgi?id=39475
llvm-svn: 346143
We currently seem to underestimate the size of functions with loops in them,
both in terms of absolute code size and in the difficulties of dealing with
such code. (Calls, for example, can be tail merged to further reduce
codesize). At -Oz, we can then increase code size by inlining small loops
multiple times.
This attempts to penalise functions with loops at -Oz by adding a CallPenalty
for each top level loop in the function. It uses LI (and hence DT) to calculate
the number of loops. As we are dealing with minsize, the inline threshold is
small and functions at this point should be relatively small, making the
construction of these cheap.
Differential Revision: https://reviews.llvm.org/D52716
llvm-svn: 346134
Using TargetTransformInfo allows the splitting pass to factor in the
code size cost of instructions as it decides whether or not outlining is
profitable.
This did not regress the overall amount of outlining seen on the handful
of internal frameworks I tested.
Thanks to Jun Bum Lim for suggesting this!
Differential Revision: https://reviews.llvm.org/D53835
llvm-svn: 346108
In PR39475:
https://bugs.llvm.org/show_bug.cgi?id=39475
..we may fail to recognize/simplify fabs() in some cases because we do not
canonicalize fcmp with a -0.0 operand.
Adding that canonicalization can cause regressions on min/max FP tests, so
that's this patch: for the purpose of determining whether something is min/max,
let the value returned by the select determine how we treat a 0.0 operand in the fcmp.
This patch doesn't actually change the -0.0 to +0.0. It just changes the analysis, so
we don't fail to recognize equivalent min/max patterns that only differ in the
signbit of 0.0.
Differential Revision: https://reviews.llvm.org/D54001
llvm-svn: 346097
This patch gives the IR ComputeNumSignBits the same functionality as the
DAG version (the code is derived from the existing code).
This an extension of the single input shuffle analysis added with D53659.
Differential Revision: https://reviews.llvm.org/D53987
llvm-svn: 346071
Summary:
-mldst-motion creates a new phi node without any debug info. Use the merged debug location from the incoming stores to fix this.
Fixes PR38177. The test case here is (somewhat) simplified from:
```
struct S {
int foo;
void fn(int bar);
};
void S::fn(int bar) {
if (bar)
foo = 1;
else
foo = 0;
}
```
Reviewers: dblaikie, gbedwell, aprantl, vsk
Reviewed By: vsk
Subscribers: vsk, JDevlieghere, llvm-commits
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D54019
llvm-svn: 346027
Model this function more closely after the BasicTTIImpl version, with
separate handling of loads and stores. For loads, the set of actually loaded
vectors is checked.
This makes it more readable and just slightly more accurate generally.
Review: Ulrich Weigand
https://reviews.llvm.org/D53071
llvm-svn: 345998
Fix PR39417, PR39497
The loop vectorizer may generate runtime SCEV checks for overflow and stride==1
cases, leading to execution of original scalar loop. The latter is forbidden
when optimizing for size. An assert introduced in r344743 triggered the above
PR's showing it does happen. This patch fixes this behavior by preventing
vectorization in such cases.
Differential Revision: https://reviews.llvm.org/D53612
llvm-svn: 345959
Inner-loop only reductions require additional checks to make sure they
form a load-phi-store cycle across inner and outer loop. Otherwise the
reduction value is not properly preserved. This patch disables
interchanging such loops for now, as it causes miscompiles in some
cases and it seems to apply only for a tiny amount of loops. Across the
test-suite, SPEC2000 and SPEC2006, 61 instead of 62 loops are
interchange with inner loop reduction support disabled. With
-loop-interchange-threshold=-1000, 3256 instead of 3267.
See the discussion and history of D53027 for an outline of how such legality
checks could look like.
Reviewers: efriedma, mcrosier, davide
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D53027
llvm-svn: 345877
When rewriting loop exit values, IndVars considers this transform not profitable if
the loop instruction has a loop user which it believes cannot be optimized away.
In current implementation only calls that immediately use the instruction are considered
as such.
This patch extends the definition of "hard" users to any side-effecting instructions
(which usually cannot be optimized away from the loop) and also allows handling
of not just immediate users, but use chains.
Differentlai Revision: https://reviews.llvm.org/D51584
Reviewed By: etherzhhb
llvm-svn: 345814
Unlike its legacy counterpart new pass manager's LoopUnrollPass does
not provide any means to select which flavors of unroll to run
(runtime, peeling, partial), relying on global defaults.
In some cases having ability to run a restricted LoopUnroll that
does more than LoopFullUnroll is needed.
Introduced LoopUnrollOptions to select optional unroll behaviors.
Added 'unroll<peeling>' to PassRegistry mainly for the sake of testing.
Reviewers: chandlerc, tejohnson
Differential Revision: https://reviews.llvm.org/D53440
llvm-svn: 345723
For some unclear reason rewriteLoopExitValues considers recalculation
after the loop profitable if it has some "soft uses" outside the loop (i.e. any
use other than call and return), even if we have proved that it has a user inside
the loop which we think will not be optimized away.
There is no existing unit test that would explain this. This patch provides an
example when rematerialisation of exit value is not profitable but it passes
this check due to presence of a "soft use" outside the loop.
It makes no sense to recalculate value on exit if we are going to compute it
due to some irremovable within the loop. This patch disallows applying this
transform in the described situation.
Differential Revision: https://reviews.llvm.org/D51581
Reviewed By: etherzhhb
llvm-svn: 345708
optsize using masked wide loads
Under Opt for Size, the vectorizer does not vectorize interleave-groups that
have gaps at the end of the group (such as a loop that reads only the even
elements: a[2*i]) because that implies that we'll require a scalar epilogue
(which is not allowed under Opt for Size). This patch extends the support for
masked-interleave-groups (introduced by D53011 for conditional accesses) to
also cover the case of gaps in a group of loads; Targets that enable the
masked-interleave-group feature don't have to invalidate interleave-groups of
loads with gaps; they could now use masked wide-loads and shuffles (if that's
what the cost model selects).
Reviewers: Ayal, hsaito, dcaballe, fhahn
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D53668
llvm-svn: 345705
InstCombine features an optimization that essentially replaces:
if (a)
free(a)
into:
free(a)
Right now, this optimization is gated by the minsize attribute and therefore
we only perform it if we can prove that we are going to be able to eliminate
the branch and the destination block.
However when casts are involved the optimization would fail to apply, because
the optimization was not smart enough to realize that it is possible to also
move the casts away from the destination block and that is harmless to the
performance since they are just noops.
E.g.,
foo(int *a)
if (a)
free((char*)a)
Wouldn't be optimized by instcombine, because
- We would refuse to hoist the `bitcast i32* %a to i8` in the source block
- We would fail to see that `bitcast i32* %a to i8` and %a are the same value.
This patch fixes both these problems:
- It teaches the pattern matching of the comparison how to look
through casts.
- It checks that whether the additional instruction in the destination block
can be hoisted and are harmless performance-wise.
- It hoists all the code of the destination block in the source block.
Differential Revision: D53356
llvm-svn: 345644
shuffle (insert ?, Scalar, IndexC), V1, Mask --> insert V1, Scalar, IndexC'
The motivating case is at least a couple of steps away: I noticed that
SLPVectorizer does not analyze shuffles as well as sequences of
insert/extract in PR34724:
https://bugs.llvm.org/show_bug.cgi?id=34724
...so SLP may fail to vectorize when source code has shuffles to start
with or instcombine has converted insert/extract to shuffles.
Independent of that, an insertelement is always a simpler op for IR
analysis vs. a shuffle, so we should transform to insert when possible.
I don't think there's any codegen concern here - if a target can't insert
a scalar directly to some fixed element in a vector (x86?), then this
should get expanded to the insert+shuffle that we started with.
Differential Revision: https://reviews.llvm.org/D53507
llvm-svn: 345607
This commit is a combination of two patches:
* "Fix in getScalarizationOverhead()"
If target returns false in TTI.prefersVectorizedAddressing(), it means the
address registers will not need to be extracted. Therefore, there should
be no operands scalarization overhead for a load instruction.
* "Don't pass the instruction pointer from getMemInstScalarizationCost."
Since VF is always > 1, this is a cost query for an instruction in the
vectorized loop and it should not be evaluated within the scalar
context of the instruction.
Review: Ulrich Weigand, Hal Finkel
https://reviews.llvm.org/D52351https://reviews.llvm.org/D52417
llvm-svn: 345603
This fixes an assertion when constant folding a GEP when the part of the offset
was in i32 (IndexSize, as per DataLayout) and part in the i64 (PointerSize) in
the newly created test case.
Differential Revision: https://reviews.llvm.org/D52609
llvm-svn: 345585
It can be profitable to outline single-block cold regions because they
may be large.
Allow outlining single-block regions if they have over some threshold of
non-debug, non-terminator instructions. I chose 3 as the threshold after
experimenting with several internal frameworks.
In practice, reducing the threshold further did not give much
improvement, whereas increasing it resulted in substantial regressions.
Differential Revision: https://reviews.llvm.org/D53824
llvm-svn: 345524
As K has to dominate I, IIUC I's range metadata must be a subset of
K's. After Eli's recent clarification to the LangRef, loading a value
outside of the range is undefined behavior.
Therefore if I's range contains elements outside of K's range and we would load
one such value, K would cause undefined behavior.
In cases like hoisting/sinking, we still want the most generic range
over all code paths to/from the hoist/sink point. As suggested in the
patches related to D47339, I will refactor the handling of those
scenarios and try to decouple it from this function as follow up, once
we switched to a similar handling of metadata in most of
combineMetadata.
I updated some tests checking mostly the merging of metadata to keep the
metadata of to dominating load. The most interesting one is probably test8 in
test/Transforms/JumpThreading/thread-loads.ll. It contained a comment
about the alias metadata preventing us to eliminate the branch, but it
seem like the actual problem currently is that we merge the ranges of
both loads and cannot eliminate the icmp afterwards. With this patch, we
manage to eliminate the icmp, as the range of the first load excludes 8.
Reviewers: efriedma, nlopes, davide
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D51629
llvm-svn: 345456
The motivating case is from PR37549:
https://bugs.llvm.org/show_bug.cgi?id=37549
The analysis improvement allows us to form a vector 'select' out of
bitwise logic (the use of ComputeNumSignBits was added at rL345149).
The smaller test shows another InstCombine improvement - we use
ComputeNumSignBits to add 'nsw' to shift-left. But the negative
test shows an example where we must not add 'nsw' - when the shuffle
mask contains undef elements.
Differential Revision: https://reviews.llvm.org/D53659
llvm-svn: 345429
Summary:
The visitICmp analysis function would record compares of pointer types, as size 0. This causes the resulting memcmp() call to have the wrong total size.
Found with "self-build" of clang/LLVM on Windows.
Reviewers: christylee, trentxintong, courbet
Reviewed By: courbet
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D53536
llvm-svn: 345413
This patch adds support of `llvm.experimental.guard` intrinsics to non-trivial
simple loop unswitching. These intrinsics represent implicit control flow which
has pretty much the same semantics as usual conditional branches. The
algorithm of dealing with them is following:
- Consider guards as unswitching candidates;
- If a guard is considered the best candidate, turn it into a branch;
- Apply normal unswitching algorithm on this branch.
The patch has no compile time effect on code that does not contain any guards.
Differential Revision: https://reviews.llvm.org/D53744
Reviewed By: chandlerc
llvm-svn: 345387
Replacing BinaryOperator::isFNeg(...) to avoid regressions when we
separate FNeg from the FSub IR instruction.
Differential Revision: https://reviews.llvm.org/D53650
llvm-svn: 345295
This mirrors what we already do for AArch64 as the cores are similar.
As discussed in the review, enabling the machine scheduler causes
more variations in performance changes so it is not enabled for now.
This patch improves LNT scores by a geomean of 1.57% at -O3.
Differential Revision: https://reviews.llvm.org/D53562
llvm-svn: 345272
The current splitting algorithm works in three stages:
1) Identify cold blocks, then
2) Use forward/backward propagation to mark hot blocks, then
3) Grow a SESE region of blocks *outside* of the set of hot blocks and
start outlining.
While testing this pass on Apple internal frameworks I noticed that some
kinds of control flow (e.g. loops) are never outlined, even though they
unconditionally lead to / follow cold blocks. I noticed two other issues
related to how cold regions are identified:
- An inconsistency can arise in the internal state of the hotness
propagation stage, as a block may end up in both the ColdBlocks set
and the HotBlocks set. Further inconsistencies can arise as these sets
do not match what's in ProfileSummaryInfo.
- It isn't necessary to limit outlining to single-exit regions.
This patch teaches the splitting algorithm to identify maximal cold
regions and outline them. A maximal cold region is defined as the set of
blocks post-dominated by a cold sink block, or dominated by that sink
block. This approach can successfully outline loops in the cold path. As
a side benefit, it maintains less internal state than the current
approach.
Due to a limitation in CodeExtractor, blocks within the maximal cold
region which aren't dominated by a single entry point (a so-called "max
ancestor") are filtered out.
Results:
- X86 (LNT + -Os + externals): 134KB of TEXT were outlined compared to
47KB pre-patch, or a ~3x improvement. Did not see a performance impact
across two runs.
- AArch64 (LNT + -Os + externals + Apple-internal benchmarks): 149KB
of TEXT were outlined. Ditto re: performance impact.
- Outlining results improve marginally in the internal frameworks I
tested.
Follow-ups:
- Outline more than once per function, outline large single basic
blocks, & try to remove unconditional branches in outlined functions.
Differential Revision: https://reviews.llvm.org/D53627
llvm-svn: 345209
Summary:
The current default of appending "_"+entry block label to the new
extracted cold function breaks demangling. Change the deliminator from
"_" to "." to enable demangling. Because the header block label will
be empty for release compile code, use "extracted" after the "." when
the label is empty.
Additionally, add a mechanism for the client to pass in an alternate
suffix applied after the ".", and have the hot cold split pass use
"cold."+Count, where the Count is currently 1 but can be used to
uniquely number multiple cold functions split out from the same function
with D53588.
Reviewers: sebpop, hiraditya
Subscribers: llvm-commits, erik.pilkington
Differential Revision: https://reviews.llvm.org/D53534
llvm-svn: 345178
The original patch was committed here:
rL344609
...and reverted:
rL344612
...because it did not properly check/test data types before calling
ComputeNumSignBits().
The tests that caused bot failures for the previous commit are
over-reaching front-end tests that run the entire -O optimizer
pipeline:
Clang :: CodeGen/builtins-systemz-zvector.c
Clang :: CodeGen/builtins-systemz-zvector2.c
I've added a negative test here to ensure coverage for that case.
The new early exit check also tests the type of the 'B' parameter,
so we don't waste time on matching if either value is unsuitable.
Original commit message:
This is part of solving PR37549:
https://bugs.llvm.org/show_bug.cgi?id=37549
The patterns shown here are a special case of something
that we already convert to select. Using ComputeNumSignBits()
catches that case (but not the more complicated motivating
patterns yet).
The backend has hooks/logic to convert back to logic ops
if that's better for the target.
llvm-svn: 345149
This work is to avoid regressions when we seperate FNeg from the FSub IR instruction.
Differential Revision: https://reviews.llvm.org/D53205
llvm-svn: 345146
masked-interleaving is enabled
Enable interleave-groups under fold-tail scenario for Opt for size compilation;
D50480 added support for vectorizing loops of arbitrary trip-count without a
remiander, which in turn makes everything in the loop conditional, including
interleave-groups if any. It therefore invalidated all interleave-groups
because we didn't have support for vectorizing predicated interleaved-groups
at the time. In the meantime, D53011 introduced this support, so we don't
have to invalidate interleave-groups when masked-interleaved support is enabled.
Reviewers: Ayal, hsaito, dcaballe, fhahn
Reviewed By: hsaito
Differential Revision: https://reviews.llvm.org/D53559
llvm-svn: 345115
LSR reassociates constants as unfolded offsets when the constants fit as
immediate add operands, which currently prevents such constants from being
combined later with loop invariant registers.
This patch modifies GenerateCombinations() to generate a second formula which
includes the unfolded offset in the combined loop-invariant register.
Differential Revision: https://reviews.llvm.org/D51861
llvm-svn: 345114
in the same round of SCC update.
In https://reviews.llvm.org/rL309784, inline history is added to prevent
infinite inlining across multiple run of inliner and SCC update, but the
history will only be kept when new SCC is actually generated during SCC update.
We found a case that SCC can be split and then merge into itself in the same
round of SCC update, so the same SCC will be pop out from UR.CWorklist and
then added back immediately, without any new SCC generated, that is why the
existing patch cannot catch the infinite inline case.
What the patch does is even if no new SCC is generated, if only the current
SCC appears in UR.CWorklist again, then keep the inline history.
Differential Revision: https://reviews.llvm.org/D52915
llvm-svn: 345103
Outlined code is cold by assumption, so it makes sense to optimize it
for minimal code size rather than performance.
After r344869 moved the splitting pass to the end of the IR pipeline,
this does not result in much of a code size reduction. This is probably
because a comparatively small number backend transforms make use of the
MinSize hint.
Running LNT on x86_64, I see that 33/1020 binaries shrink for a total of
919 bytes of TEXT reduction. I didn't measure a significant performance
impact.
Differential Revision: https://reviews.llvm.org/D53518
llvm-svn: 345072
Summary:
TryToShrinkGlobalToBoolean, when possible, will split store <value> + load <value> into store <bool> + select <bool ? value : 0>. This preserves DebugLoc during that pass.
Fixes PR37959. The test case here is the simplified .ll for:
```
static int foo;
int bar() {
foo = 5;
return foo;
}
```
Reviewers: dblaikie, gbedwell, aprantl
Reviewed By: dblaikie
Subscribers: mehdi_amini, JDevlieghere, dexonsmith, llvm-commits
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D53531
llvm-svn: 345046
We need to update this code before introducing an 'fneg' instruction in IR,
so we might as well kill off the integer neg/not queries too.
This is no-functional-change-intended for scalar code and most vector code.
For vectors, we can see that the 'match' API allows for undef elements in
constants, so we optimize those cases better.
Ideally, there would be a test for each code diff, but I don't see evidence
of that for the existing code, so I didn't try very hard to come up with new
vector tests for each code change.
Differential Revision: https://reviews.llvm.org/D53533
llvm-svn: 345042
Expand arithmetic reduction to include mul/and/or/xor instructions.
This patch just fixes the SLPVectorizer - the effective reduction costs for AVX1+ are still poor (see rL344846) and will need to be improved before SLP sees this as a valid transform - but we can already see the effect on SSE2 tests.
This partially helps PR37731, but doesn't fix it all as it still falls over on the extraction/reduction order for some reason.
Differential Revision: https://reviews.llvm.org/D53473
llvm-svn: 345037
Nikita Popov