As discussed in D49047 / D48987, shift-by-undef produces poison,
so we can't use undef vector elements in that case..
Note that we need to extend this for poison-generating flags,
and there's a proposal to create poison from FMF in D47963,
llvm-svn: 336562
As noted in D48987, there are many different ways for this transform to go wrong.
In particular, the poison potential for shifts means we have to more careful with those ops.
I added tests to make that behavior visible for all of the different cases that I could find.
This is a partial fix. To make this review easier, I did not make changes for the single binop
pattern (handled in foldSelectShuffleWith1Binop()). I also left out some potential optimizations
noted with TODO comments. I'll follow-up once we're confident that things are correct here.
The goal is to correct all marked FIXME tests to either avoid the shuffle transform or do it safely.
Note that distinguishing when the shuffle mask contains undefs and using getBinOpIdentity() allows
for some improvements to div/rem patterns, so there are wins along with the missed opportunities
and fixes.
Differential Revision: https://reviews.llvm.org/D49047
llvm-svn: 336546
r335553 with the non-trivial unswitching of switches.
The code correctly updated most aspects of the CFG and analyses, but
missed some crucial aspects:
1) When multiple cases have the same successor, we unswitch that
a single time and replace the switch with a direct branch. The CFG
here is correct, but the target of this direct branch may have had
a PHI node with multiple entries in it.
2) When we still have to clone a successor of the switch into an
unswitched copy of the loop, we'll delete potentially multiple edges
entering this successor, not just one.
3) We also have to delete multiple edges entering the successors in the
original loop when they have to be retained.
4) When the "retained successor" *also* occurs as a case successor, we
just assert failed everywhere. This doesn't happen very easily
because its always valid to simply drop the case -- the retained
successor for switches is always the default successor. However, it
is likely possible through some contrivance of different loop passes,
unrolling, and simplifying for this to occur in practice and
certainly there is nothing "invalid" about the IR so this pass needs
to handle it.
5) In the case of #4, we also will replace these multiple edges with
a direct branch much like in #1 and need to collapse the entries in
any PHI nodes to a single enrty.
All of this stems from the delightful fact that the same successor can
show up in multiple parts of the switch terminator, and each of these
are considered a distinct edge for the purpose of PHI nodes (and
iterating the successors and predecessors) but not for unswitching
itself, the dominator tree, or many other things. For the record,
I intensely dislike this "feature" of the IR in large part because of
the complexity it causes in passes like this. We already have a ton of
logic building sets and handling duplicates, and we just had to add
a bunch more.
I've added a complex test case that covers all five of the above failure
modes. I've also added a variation on it where #4 and #5 occur in loop
exit, adding fun where we have an LCSSA PHI node with "multiple entries"
despite have dedicated exits. There were no additional issues found by
this, but it seems a useful corner case to cover with testing.
One thing that working on all of this code has made painfully clear for
me as well is how amazingly inefficient our PHI node representation is
(in terms of the in-memory data structures and the APIs used to update
them). This code has truly marvelous complexity bounds because every
time we remove an entry from a PHI node we do a linear scan to find it
and then a linear update to the data structure to remove it. We could in
theory batch all of the PHI node updates into a single linear walk of
the operands making this much more efficient, but the APIs fight hard
against this and the fact that we have to handle duplicates in the
peculiar manner we do (removing all but one in some cases) makes even
implementing that very tedious and annoying. Anyways, none of this is
new here or specific to loop unswitching. All code in LLVM that updates
PHI node operands suffers from these problems.
llvm-svn: 336536
Summary:
PGOMemOPSize only modifies CFG in a couple of places; thus we can preserve the DominatorTree with little effort.
When optimizing SQLite with -O3, this patch can decrease 3.8% of the numbers of nodes traversed by DFS and 5.7% of the times DominatorTreeBase::recalculation is called.
Reviewers: kuhar, davide, dmgreen
Reviewed By: dmgreen
Subscribers: mzolotukhin, vsk, llvm-commits
Differential Revision: https://reviews.llvm.org/D48914
llvm-svn: 336522
In the 'detectCTLZIdiom' function support for loops that use LSHR instruction instead of ASHR has been added.
This supports creating ctlz from the following code.
int lzcnt(int x) {
int count = 0;
while (x > 0) {
count++;
x = x >> 1;
}
return count;
}
Patch by Olga Moldovanova
Differential Revision: https://reviews.llvm.org/D48354
llvm-svn: 336509
after trivial unswitching.
This PR illustrates that a fundamental analysis update was not performed
with the new loop unswitch. This update is also somewhat fundamental to
the core idea of the new loop unswitch -- we actually *update* the CFG
based on the unswitching. In order to do that, we need to update the
loop nest in addition to the domtree.
For some reason, when writing trivial unswitching, I thought that the
loop nest structure cannot be changed by the transformation. But the PR
helps illustrate that it clearly can. I've expanded this to a number of
different test cases that try to cover the different cases of this. When
we unswitch, we move an exit edge of a loop out of the loop. If this
exit edge changes which loop reached by an exit is the innermost loop,
it changes the parent of the loop. Essentially, this transformation may
hoist the inner loop up the nest. I've added the simple logic to handle
this reliably in the trivial unswitching case. This just requires
updating LoopInfo and rebuilding LCSSA on the impacted loops. In the
trivial case, we don't even need to handle dedicated exits because we're
only hoisting the one loop and we just split its preheader.
I've also ported all of these tests to non-trivial unswitching and
verified that the logic already there correctly handles the loop nest
updates necessary.
Differential Revision: https://reviews.llvm.org/D48851
llvm-svn: 336477
The replaceAllDbgUsesWith utility helps passes preserve debug info when
replacing one value with another.
This improves upon the existing insertReplacementDbgValues API by:
- Updating debug intrinsics in-place, while preventing use-before-def of
the replacement value.
- Falling back to salvageDebugInfo when a replacement can't be made.
- Moving the responsibiliy for rewriting llvm.dbg.* DIExpressions into
common utility code.
Along with the API change, this teaches replaceAllDbgUsesWith how to
create DIExpressions for three basic integer and pointer conversions:
- The no-op conversion. Applies when the values have the same width, or
have bit-for-bit compatible pointer representations.
- Truncation. Applies when the new value is wider than the old one.
- Zero/sign extension. Applies when the new value is narrower than the
old one.
Testing:
- check-llvm, check-clang, a stage2 `-g -O3` build of clang,
regression/unit testing.
- This resolves a number of mis-sized dbg.value diagnostics from
Debugify.
Differential Revision: https://reviews.llvm.org/D48676
llvm-svn: 336451
As discussed in D48987 and D48893, there are many different
ways to go wrong depending on the binop (and as shown here
we already do go wrong in some cases).
llvm-svn: 336450
Previously we only iterated over functions reachable from the set of
external functions in the module. But since some of the passes under
this (notably the always-inliner and coroutine lowerer) are required for
correctness, they need to run over everything.
This just adds an extra layer of iteration over the CallGraph to keep
track of which functions we've already visited and get the next batch of
SCCs.
Should fix PR38029.
llvm-svn: 336419
Better NaN handling for AMDGCN fmed3.
All operands are checked for NaN now. The checks
were moved before the canonicalization to provide
a better mapping from fclamp. Changed the behaviour
of fmed3(x,y,NaN) to return max(x,y) instead of
min(x,y) in light of this. Updated tests as a result
and added some new cases to cover the fix.
Patch by Alan Baker
llvm-svn: 336375
We have bailout hacks based on min/max in various places in instcombine
that shouldn't be necessary. The affected test was added for:
D48930
...which is a consequence of the improvement in:
D48584 (https://reviews.llvm.org/rL336172)
I'm assuming the visitTrunc bailout in this patch was added specifically
to avoid a change from SimplifyDemandedBits, so I'm just moving that
below the EvaluateInDifferentType optimization. A narrow min/max is still
a min/max.
llvm-svn: 336293
When creating `phi` instructions to resume at the scalar part of the loop,
copy the DebugLoc from the original phi over to the new one.
Differential Revision: https://reviews.llvm.org/D48769
llvm-svn: 336256
When zext is EvaluatedInDifferentType, InstCombine
drops the dbg.value intrinsic. This patch tries to
preserve said DI, by inserting the zext's old DI in the
resulting instruction. (Only for integer type for now)
Differential Revision: https://reviews.llvm.org/D48331
llvm-svn: 336254
As the test diffs show, the current users of getBinOpIdentity()
are InstCombine and Reassociate. SLP vectorizer is a candidate
for using this functionality too (D28907).
The InstCombine shuffle improvements are part of the planned
enhancements noted in D48830.
InstCombine actually has several other uses of getBinOpIdentity()
via SimplifyUsingDistributiveLaws(), but we don't call that for
any FP ops. Fixing that might be another part of removing the
custom reassociation in InstCombine that is only done for fadd+fmul.
llvm-svn: 336215
This is the last significant change suggested in PR37806:
https://bugs.llvm.org/show_bug.cgi?id=37806#c5
...though there are several follow-ups noted in the code comments
in this patch to complete this transform.
It's possible that a binop feeding a select-shuffle has been eliminated
by earlier transforms (or the code was just written like this in the 1st
place), so we'll fail to match the patterns that have 2 binops from:
D48401,
D48678,
D48662,
D48485.
In that case, we can try to materialize identity constants for the remaining
binop to fill in the "ghost" lanes of the vector (where we just want to pass
through the original values of the source operand).
I added comments to ConstantExpr::getBinOpIdentity() to show planned follow-ups.
For now, we only handle the 5 commutative integer binops (add/mul/and/or/xor).
Differential Revision: https://reviews.llvm.org/D48830
llvm-svn: 336196
Summary:
When salvaging a dbg.declare/dbg.addr we should not add
DW_OP_stack_value to the DIExpression
(see test/Transforms/InstCombine/salvage-dbg-declare.ll).
Consider this example
%vla = alloca i32, i64 2
call void @llvm.dbg.declare(metadata i32* %vla, metadata !1, metadata !DIExpression())
Instcombine will turn it into
%vla1 = alloca [2 x i32]
%vla1.sub = getelementptr inbounds [2 x i32], [2 x i32]* %vla, i64 0, i64 0
call void @llvm.dbg.declare(metadata [2 x i32]* %vla1.sub, metadata !19, metadata !DIExpression())
If the GEP can be eliminated, then the dbg.declare will be salvaged
and we should get
%vla1 = alloca [2 x i32]
call void @llvm.dbg.declare(metadata [2 x i32]* %vla1, metadata !19, metadata !DIExpression())
The problem was that salvageDebugInfo did not recognize dbg.declare
as being indirect (%vla1 points to the value, it does not hold the
value), so we incorrectly got
call void @llvm.dbg.declare(metadata [2 x i32]* %vla1, metadata !19, metadata !DIExpression(DW_OP_stack_value))
I also made sure that llvm::salvageDebugInfo and
DIExpression::prependOpcodes do not add DW_OP_stack_value to
the DIExpression in case no new operands are added to the
DIExpression. That way we avoid to, unneccessarily, turn a
register location expression into an implicit location expression
in some situations (see test11 in test/Transforms/LICM/sinking.ll).
Reviewers: aprantl, vsk
Reviewed By: aprantl, vsk
Subscribers: JDevlieghere, llvm-commits
Differential Revision: https://reviews.llvm.org/D48837
llvm-svn: 336191
unswitching loops.
Original patch trying to address this was sent in D47624, but that
didn't quite handle things correctly. There are two key principles used
to select whether and how to invalidate SCEV-cached information about
loops:
1) We must invalidate any info SCEV has cached before unswitching as we
may change (or destroy) the loop structure by the act of unswitching,
and make it hard to recover everything we want to invalidate within
SCEV.
2) We need to invalidate all of the loops whose CFGs are mutated by the
unswitching. Notably, this isn't the *entire* loop nest, this is
every loop contained by the outermost loop reached by an exit block
relevant to the unswitch.
And we need to do this even when doing trivial unswitching.
I've added more focused tests that directly check that SCEV starts off
with imprecise information and after unswitching (and simplifying
instructions) re-querying SCEV will produce precise information. These
tests also specifically work to check that an *outer* loop's information
becomes precise.
However, the testing here is still a bit imperfect. Crafting test cases
that reliably fail to be analyzed by SCEV before unswitching and succeed
afterward proved ... very, very hard. It took me several hours and
careful work to build these, and I'm not optimistic about necessarily
coming up with more to cover more elaborate possibilities. Fortunately,
the code pattern we are testing here in the pass is really
straightforward and reliable.
Thanks to Max Kazantsev for the initial work on this as well as the
review, and to Hal Finkel for helping me talk through approaches to test
this stuff even if it didn't come to much.
Differential Revision: https://reviews.llvm.org/D47624
llvm-svn: 336183
This patch changes order of transform in InstCombineCompares to avoid
performing transforms based on ranges which produce complex bit arithmetics
before more simple things (like folding with constants) are done. See PR37636
for the motivating example.
Differential Revision: https://reviews.llvm.org/D48584
Reviewed By: spatel, lebedev.ri
llvm-svn: 336172
Summary:
Comment on Transforms/LoopVersioning/incorrect-phi.ll: With the change
SCEV is able to prove that the loop doesn't wrap-self (due to zext i16
to i64), disabling the entire loop versioning pass. Removed the zext and
just use i64.
Reviewers: sanjoy
Subscribers: jlebar, hiraditya, javed.absar, bixia, llvm-commits
Differential Revision: https://reviews.llvm.org/D48409
llvm-svn: 336140
Summary: It is common to have the following min/max pattern during the intermediate stages of SLP since we only optimize at the end. This patch tries to catch such patterns and allow more vectorization.
%1 = extractelement <2 x i32> %a, i32 0
%2 = extractelement <2 x i32> %a, i32 1
%cond = icmp sgt i32 %1, %2
%3 = extractelement <2 x i32> %a, i32 0
%4 = extractelement <2 x i32> %a, i32 1
%select = select i1 %cond, i32 %3, i32 %4
Author: FarhanaAleen
Reviewed By: ABataev, RKSimon, spatel
Differential Revision: https://reviews.llvm.org/D47608
llvm-svn: 336130
This extends D48485 to allow another pair of binops (add/or) to be combined either
with or without a leading shuffle:
or X, C --> add X, C (when X and C have no common bits set)
Here, we need value tracking to determine that the 'or' can be reversed into an 'add',
and we've added general infrastructure to allow extending to other opcodes or moving
to where other passes could use that functionality.
Differential Revision: https://reviews.llvm.org/D48662
llvm-svn: 336128
Due to current limitations in constant analysis, we need flags
on add or mul to show propagation for the potential transform
suggested in these tests (no other binops currently report
identity constants).
llvm-svn: 336101
This version contains a fix to add values for which the state in ParamState change
to the worklist if the state in ValueState did not change. To avoid adding the
same value multiple times, mergeInValue returns true, if it added the value to
the worklist. The value is added to the worklist depending on its state in
ValueState.
Original message:
For comparisons with parameters, we can use the ParamState lattice
elements which also provide constant range information. This improves
the code for PR33253 further and gets us closer to use
ValueLatticeElement for all values.
Also, as we are using the range information in the solver directly, we
do not need tryToReplaceWithConstantRange afterwards anymore.
Reviewers: dberlin, mssimpso, davide, efriedma
Reviewed By: mssimpso
Differential Revision: https://reviews.llvm.org/D43762
llvm-svn: 336098
We were always using the opcodes of the first 2 scalars for the costs of the alternate opcode + shuffle. This made sense when we used SK_Alternate and opcodes were guaranteed to be alternating, but this fails for the more general SK_Select case.
This fix exposes an issue demonstrated by the fmul_fdiv_v4f32_const test - the SLM model has v4f32 fdiv costs which are more than twice those of the f32 scalar cost, meaning that the cost model determines that the vectorization is not performant. Unfortunately it completely ignores the fact that the fdiv by a constant will be changed into a fmul by InstCombine for a much lower cost vectorization. But at least we're seeing this now...
llvm-svn: 336095
Summary:
This patch introduce new intrinsic -
strip.invariant.group that was described in the
RFC: Devirtualization v2
Reviewers: rsmith, hfinkel, nlopes, sanjoy, amharc, kuhar
Subscribers: arsenm, nhaehnle, JDevlieghere, hiraditya, xbolva00, llvm-commits
Differential Revision: https://reviews.llvm.org/D47103
Co-authored-by: Krzysztof Pszeniczny <krzysztof.pszeniczny@gmail.com>
llvm-svn: 336073
This is a simple implementation of the unroll-and-jam classical loop
optimisation.
The basic idea is that we take an outer loop of the form:
for i..
ForeBlocks(i)
for j..
SubLoopBlocks(i, j)
AftBlocks(i)
Instead of doing normal inner or outer unrolling, we unroll as follows:
for i... i+=2
ForeBlocks(i)
ForeBlocks(i+1)
for j..
SubLoopBlocks(i, j)
SubLoopBlocks(i+1, j)
AftBlocks(i)
AftBlocks(i+1)
Remainder Loop
So we have unrolled the outer loop, then jammed the two inner loops into
one. This can lead to a simpler inner loop if memory accesses can be shared
between the now jammed loops.
To do this we have to prove that this is all safe, both for the memory
accesses (using dependence analysis) and that ForeBlocks(i+1) can move before
AftBlocks(i) and SubLoopBlocks(i, j).
Differential Revision: https://reviews.llvm.org/D41953
llvm-svn: 336062
Alternate opcode handling only supports binary operators, these tests demonstrate missed opportunities to vectorize some sitofp/uitofp and fptosi/fptoui style casts as well as some (successful) float bits manipulations
llvm-svn: 336060
This was discussed in D48401 as another improvement for:
https://bugs.llvm.org/show_bug.cgi?id=37806
If we have 2 different variable values, then we shuffle (select) those lanes,
shuffle (select) the constants, and then perform the binop. This eliminates a binop.
The new shuffle uses the same shuffle mask as the existing shuffle, so there's no
danger of creating a difficult shuffle.
All of the earlier constraints still apply, but we also check for extra uses to
avoid creating more instructions than we'll remove.
Additionally, we're disallowing the fold for div/rem because that could expose a
UB hole.
Differential Revision: https://reviews.llvm.org/D48678
llvm-svn: 335974
Summary:
An alternative to D48597.
Fixes [[ https://bugs.llvm.org/show_bug.cgi?id=37936 | PR37936 ]].
The problem is as follows:
1. `indvars` marks `%dec` as `NUW`.
2. `loop-instsimplify` runs `instsimplify`, which constant-folds `%dec` to -1 (D47908)
3. `loop-reduce` tries to do some further modification, but crashes
with an type assertion in cast, because `%dec` is no longer an `Instruction`,
If the runline is split into two, i.e. you first run `-indvars -loop-instsimplify`,
store that into a file, and then run `-loop-reduce`, there is no crash.
So it looks like the problem is due to `-loop-instsimplify` not discarding SCEV.
But in this case we can just not crash if it's not an `Instruction`.
This is just a local fix, unlike D48597, so there may very well be other problems.
Reviewers: mkazantsev, uabelho, sanjoy, silviu.baranga, wmi
Reviewed By: mkazantsev
Subscribers: evstupac, javed.absar, spatel, llvm-commits
Differential Revision: https://reviews.llvm.org/D48599
llvm-svn: 335950
Summary:
The InlinerFunctionImportStats will collect and dump stats regarding how
many function inlined into the module were imported by ThinLTO.
Reviewers: wmi, dexonsmith
Subscribers: mehdi_amini, inglorion, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D48729
llvm-svn: 335914
When rewriting an alloca partition copy the DL from the
old alloca over the the new one.
Differential Revision: https://reviews.llvm.org/D48640
llvm-svn: 335904
This is an enhancement to D48401 that was discussed in:
https://bugs.llvm.org/show_bug.cgi?id=37806
We can convert a shift-left-by-constant into a multiply (we canonicalize IR in the other
direction because that's generally better of course). This allows us to remove the shuffle
as we do in the regular opcodes-are-the-same cases.
This requires a small hack to make sure we don't introduce any extra poison:
https://rise4fun.com/Alive/ZGv
Other examples of opcodes where this would work are add+sub and fadd+fsub, but we already
canonicalize those subs into adds, so there's nothing to do for those cases AFAICT. There
are planned enhancements for opcode transforms such or -> add.
Note that there's a different fold needed if we've already managed to simplify away a binop
as seen in the test based on PR37806, but we manage to get that one case here because this
fold is positioned above the demanded elements fold currently.
Differential Revision: https://reviews.llvm.org/D48485
llvm-svn: 335888
SCCP does not change the CFG, so we can mark it as preserved.
Reviewers: dberlin, efriedma, davide
Reviewed By: davide
Differential Revision: https://reviews.llvm.org/D47149
llvm-svn: 335820
If a trunc has a user in a block which is not reachable from entry,
we can safely perform trunc elimination as if this user didn't exist.
llvm-svn: 335816
I think the intrinsics named 'avx512.mask.' should refer to the previous behavior of taking a mask argument in the intrinsic instead of using a 'select' or 'and' instruction in IR to accomplish the masking. This is more consistent with the goal that eventually we will have no intrinsics that have masking builtin. When we reach that goal, we should have no intrinsics named "avx512.mask".
llvm-svn: 335744
This patch adds a custom trunc store lowering for v4i8 vector types.
Since there is not v.4b register, the v4i8 is promoted to v4i16 (v.4h)
and default action for v4i8 is to extract each element and issue 4
byte stores.
A better strategy would be to extended the promoted v4i16 to v8i16
(with undef elements) and extract and store the word lane which
represents the v4i8 subvectores. The construction:
define void @foo(<4 x i16> %x, i8* nocapture %p) {
%0 = trunc <4 x i16> %x to <4 x i8>
%1 = bitcast i8* %p to <4 x i8>*
store <4 x i8> %0, <4 x i8>* %1, align 4, !tbaa !2
ret void
}
Can be optimized from:
umov w8, v0.h[3]
umov w9, v0.h[2]
umov w10, v0.h[1]
umov w11, v0.h[0]
strb w8, [x0, #3]
strb w9, [x0, #2]
strb w10, [x0, #1]
strb w11, [x0]
ret
To:
xtn v0.8b, v0.8h
str s0, [x0]
ret
The patch also adjust the memory cost for autovectorization, so the C
code:
void foo (const int *src, int width, unsigned char *dst)
{
for (int i = 0; i < width; i++)
*dst++ = *src++;
}
can be vectorized to:
.LBB0_4: // %vector.body
// =>This Inner Loop Header: Depth=1
ldr q0, [x0], #16
subs x12, x12, #4 // =4
xtn v0.4h, v0.4s
xtn v0.8b, v0.8h
st1 { v0.s }[0], [x2], #4
b.ne .LBB0_4
Instead of byte operations.
llvm-svn: 335735
This prevents InstCombine from creating mis-sized dbg.values when
replacing a sequence of casts with a simpler cast. For example, in:
(fptrunc (floor (fpext X))) -> (floorf X)
We no longer emit dbg.value(X) (with a 32-bit float operand) to describe
(fpext X) (which is a 64-bit float).
This was diagnosed by the debugify check added in r335682.
llvm-svn: 335696
It's not possible to get the fragment size of some dbg.values. Teach the
mis-sized dbg.value diagnostic to detect this scenario and bail out.
Tested with:
$ find test/Transforms -print -exec opt -debugify-each -instcombine {} \;
llvm-svn: 335695
Summary:
When recording uses we need to rewrite after cloning a loop we need to
check if the use is not dominated by the original def. The initial
assumption was that the cloned basic block will introduce a new path and
thus the original def will only dominate the use if they are in the same
BB, but as the reproducer from PR37745 shows it's not always the case.
This fixes PR37745.
Reviewers: haicheng, Ka-Ka
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D48111
llvm-svn: 335675
Not sure why this logic seems to be repeated in 2 different places,
one called by the other.
On AMDGPU addrspace(3) globals start allocating at 0, so these
checks will be incorrect (not that real code actually tries
to compare these addresses)
llvm-svn: 335649
I'm not sure why the code here is skipping calls since
TTI does try to do something for general calls, but it
at least should allow intrinsics.
Skip intrinsics that should not be omitted as calls, which
is by far the most common case on AMDGPU.
llvm-svn: 335645
Similar to other patches in this series:
https://reviews.llvm.org/rL335512https://reviews.llvm.org/rL335527https://reviews.llvm.org/rL335597https://reviews.llvm.org/rL335616
...this is filling a gap in analysis that is exposed by an unrelated select-of-constants transform.
I didn't see a way to unify the sext cases because each div/rem opcode results in a different fold.
Note that in this case, the backend might want to convert the select into math:
Name: sext urem
%e = sext i1 %x to i32
%r = urem i32 %y, %e
=>
%c = icmp eq i32 %y, -1
%z = zext i1 %c to i32
%r = add i32 %z, %y
llvm-svn: 335622
Since D46637 we are better at handling uniform/non-uniform constant Pow2 detection; this patch tweaks the SLP argument handling to support them.
As SLP works with arrays of values I don't think we can easily use the pattern match helpers here.
Differential Revision: https://reviews.llvm.org/D48214
llvm-svn: 335621
Note: I didn't add a hasOneUse() check because the existing,
related fold doesn't have that check. I suspect that the
improved analysis and codegen make these some of the rare
canonicalization cases where we allow an increase in
instructions.
llvm-svn: 335597
changeToUnreachable may remove PHI nodes from executable blocks we found values
for and we would fail to replace them. By changing dead blocks to unreachable after
we replaced constants in all executable blocks, we ensure such PHI nodes are replaced
by their known value before.
Fixes PR37780.
Reviewers: efriedma, davide
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D48421
llvm-svn: 335588
Summary:
This is a follow-up to r334830 and r335031.
In the valueCoversEntireFragment check we now also handle
the situation when there is a variable length array (VLA)
involved, and the length of the array has been reduced to
a constant.
The ConvertDebugDeclareToDebugValue functions that are related
to PHI nodes and load instructions now avoid inserting dbg.value
intrinsics when the value does not, for certain, cover the
variable/fragment that should be described.
In r334830 we assumed that the value always covered the entire
var/fragment and we had assertions in the code to show that
assumption. However, those asserts failed when compiling code
with VLAs, so we removed the asserts in r335031. Now when we
know that the valueCoversEntireFragment check can fail also for
PHI/Load instructions we avoid to insert the faulty dbg.value
intrinsic in such situations. Compared to the Store instruction
scenario we simply drop the dbg.value here (as the variable does
not change its value due to PHI/Load, so an earlier dbg.value
describing the variable should still be valid).
Reviewers: aprantl, vsk, efriedma
Reviewed By: aprantl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D48547
llvm-svn: 335580
Turn canonicalized subtraction back into (-1 - B) and combine it with (A + 1) into (A - B).
This is similar to the folding already done for (B ^ -1) + Const into (-1 + Const) - B.
Differential Revision: https://reviews.llvm.org/D48535
llvm-svn: 335579
unswitching of switches.
This works much like trivial unswitching of switches in that it reliably
moves the switch out of the loop. Here we potentially clone the entire
loop into each successor of the switch and re-point the cases at these
clones.
Due to the complexity of actually doing nontrivial unswitching, this
patch doesn't create a dedicated routine for handling switches -- it
would duplicate far too much code. Instead, it generalizes the existing
routine to handle both branches and switches as it largely reduces to
looping in a few places instead of doing something once. This actually
improves the results in some cases with branches due to being much more
careful about how dead regions of code are managed. With branches,
because exactly one clone is created and there are exactly two edges
considered, somewhat sloppy handling of the dead regions of code was
sufficient in most cases. But with switches, there are much more
complicated patterns of dead code and so I've had to move to a more
robust model generally. We still do as much pruning of the dead code
early as possible because that allows us to avoid even cloning the code.
This also surfaced another problem with nontrivial unswitching before
which is that we weren't as precise in reconstructing loops as we could
have been. This seems to have been mostly harmless, but resulted in
pointless LCSSA PHI nodes and other unnecessary cruft. With switches, we
have to get this *right*, and everything benefits from it.
While the testing may seem a bit light here because we only have two
real cases with actual switches, they do a surprisingly good job of
exercising numerous edge cases. Also, because we share the logic with
branches, most of the changes in this patch are reasonably well covered
by existing tests.
The new unswitch now has all of the same fundamental power as the old
one with the exception of the single unsound case of *partial* switch
unswitching -- that really is just loop specialization and not
unswitching at all. It doesn't fit into the canonicalization model in
any way. We can add a loop specialization pass that runs late based on
profile data if important test cases ever come up here.
Differential Revision: https://reviews.llvm.org/D47683
llvm-svn: 335553
This avoids creating unnecessary casts if the IP used to be a dbg info
intrinsic. Fixes PR37727.
Reviewers: vsk, aprantl, sanjoy, efriedma
Reviewed By: vsk, efriedma
Differential Revision: https://reviews.llvm.org/D47874
llvm-svn: 335513
We canonicalize to select with a zext-add and either zext-sub or sext-sub,
so this shows a pattern that's not conforming to the general trend.
llvm-svn: 335506
FDiv is replaced with multiplication by reciprocal and invariant
reciprocal is hoisted out of the loop, while multiplication remains
even if invariant.
Switch checks for all invariant operands and only invariant
denominator to fix the issue.
Differential Revision: https://reviews.llvm.org/D48447
llvm-svn: 335411
This gets rid of a bunch of weird special cases; instead, just use SCEV
rewriting for everything. In addition to being simpler, this fixes a
bug where we would use the wrong stride in certain edge cases.
The one bit I'm not quite sure about is the trip count handling,
specifically the FIXME about overflow. In general, I think we need to
widen the exit condition, but that's probably not profitable if the new
type isn't legal, so we probably need a check somewhere. That said, I
don't think I'm making the existing problem any worse.
As a followup to this, a bunch of IV-related code in root-finding could
be cleaned up; with SCEV-based rewriting, there isn't any reason to
assume a loop will have exactly one or two PHI nodes.
Differential Revision: https://reviews.llvm.org/D45191
llvm-svn: 335400
Enable tryToVectorizeList to support InstructionsState alternate opcode patterns at a root (build vector etc.) as well as further down the vectorization tree.
NOTE: This patch reduces some of the debug reporting if there are opcode mismatches - I can try to add it back if it proves a problem. But it could get rather messy trying to provide equivalent verbose debug strings via getSameOpcode etc.
Differential Revision: https://reviews.llvm.org/D48488
llvm-svn: 335364
SLP currently only accepts (F)Add/(F)Sub alternate counterpart ops to be merged into an alternate shuffle.
This patch relaxes this to accept any pair of BinaryOperator opcodes instead, assuming the target's cost model accepts the vectorization+shuffle.
Differential Revision: https://reviews.llvm.org/D48477
llvm-svn: 335349
This one shows another pattern that we'll need to match
in some cases, but the current ordering of folds allows
us to match this as 2 binops before simplification takes
place.
llvm-svn: 335347
AArch64 was only setting costs for SK_Transpose, which meant that many of the simpler shuffles (e.g. SK_Select and SK_PermuteSingleSrc for larger vector elements) was being severely overestimated by the default shuffle expansion.
This patch adds costs to help improve SLP performance and avoid a regression in reductions introduced by D48174.
I'm not very knowledgeable about AArch64 shuffle lowering so I've kept the extra costs to a minimum - someone who knows this code can add extra costs which should improve vectorization a lot more.
Differential Revision: https://reviews.llvm.org/D48172
llvm-svn: 335329
clear out deleted loops from the current queue beyond just the current
loop.
This is important because SimpleLoopUnswitch will now enqueue the same
loop to be re-processed. When it does this with the legacy PM, we don't
have a way of canceling the rest of the pipeline and so we can end up
deleting the loop before we reprocess it. =/
This change also makes it easy to support deleting other loops in the
queue to process, although I don't have any use cases for that.
Differential Revision: https://reviews.llvm.org/D48470
llvm-svn: 335317
With non-commutative binops, we could be using the same
variable value as operand 0 in 1 binop and operand 1 in
the other, so we have to check for that possibility and
bail out.
llvm-svn: 335312
Summary:
A reprise of D25849.
This crash was found through fuzzing some time ago and was documented in PR28879.
No check for load size has been added due to the following tests:
- Transforms/GVN/invariant.group.ll
- Transforms/GVN/pr10820.ll
These tests expect load sizes that are not a multiple of eight.
Thanks to @davide for the original patch.
Reviewers: nlopes, davide, RKSimon, reames, efriedma
Reviewed By: efriedma
Subscribers: davide, llvm-commits, Prazek
Differential Revision: https://reviews.llvm.org/D48330
llvm-svn: 335294
This is the simplest case from PR37806:
https://bugs.llvm.org/show_bug.cgi?id=37806
If we have a common variable operand used in a pair of binops with vector constants
that are vector selected together, then we can constant shuffle the constant vectors
to eliminate the shuffle instruction.
This has some tricky parts that are hopefully addressed in the tests and their
respective comments:
1. If the shuffle mask contains an undef element, then that lane of the result is
undef:
http://llvm.org/docs/LangRef.html#shufflevector-instruction
Therefore, we can replace the constant in that lane with an undef value except
for div/rem. With div/rem, an undef in the divisor would cause the whole op to
be undef. So I'm using the same hack as in D47686 - replace the undefs with '1'.
2. Intersect the wrapping and FMF of the original binops for the new binop. There
should be no extra poison or fast-math potential in the new binop that wasn't
possible in the original code.
3. Disregard other uses. Given that we're eliminating uses (shortening the
dependency chain), I think that's always the right IR canonicalization. But
I purposely chose the udiv test to demonstrate the scenario where both
intermediate values have other uses because that seems likely worse for
codegen with an expensive math op. This seems like a very rare possibility to
me, so I don't think it requires a backend patch first.
Differential Revision: https://reviews.llvm.org/D48401
llvm-svn: 335283
This reverts commit r335206.
As discussed here: https://reviews.llvm.org/rL333740, a fix will come
tomorrow. In the meanwhile, revert this to fix some bots.
llvm-svn: 335272
The previous code worked with vectors, but it failed when the
vector constants contained undef elements.
The matchers handle those cases.
llvm-svn: 335262
Summary:
Use the expanded features of the TableGen generic tables to avoid manually
adding the combinatorially exploded set of intrinsics. The
getAMDGPUImageDimIntrinsic lookup function is early-out,
i.e. non-AMDGPU intrinsics will never look at the underlying table.
Use a generic approach for getting the new intrinsic overload to keep the
code simple, and make the image dmask handling more generic:
- handle non-sampler image loads
- handle the case where the set of demanded elements is not a prefix
There is some overlap between this code and an optimization that happens
in the backend during code generation. They currently complement each other:
- only the codegen optimization can generate vec3 loads
- only the InstCombine optimization can handle D16
The InstCombine optimization also likely covers more cases since the
codegen optimization is fairly ad-hoc. Ideally, we'll remove the optimization
in codegen once the infrastructure for vec3 is in place (which will probably
take a long time).
Modify the test cases to use dimension-aware intrinsics. This makes it
easier to see that the test coverage for the new intrinsics is equivalent,
and the old style intrinsics will be removed in a follow-up commit anyway.
Change-Id: I4b91ea661413d13004956fe4ef7d13d41b8ce3ad
Reviewers: arsenm, rampitec, majnemer
Subscribers: kzhuravl, wdng, mgorny, yaxunl, dstuttard, tpr, t-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D48165
llvm-svn: 335230
This enables da-delinearize in Dependence Analysis for delinearizing array
accesses into multiple dimensions. This can help to increase the power of
Dependence analysis on multi-dimensional arrays and prevent having to fall
back to the slower and less accurate MIV tests. It adds static checks on the
bounds of the arrays to ensure that one dimension doesn't overflow into
another, and brings our code in line with our tests.
Differential Revision: https://reviews.llvm.org/D45872
llvm-svn: 335217
These were being over cautious for costs for one/two op general shuffles - VSHUFPD doesn't have to replicate the same shuffle in both lanes like VSHUFPS does.
llvm-svn: 335216
r335150 should resolve the issues with the clang-with-thin-lto-ubuntu
and clang-with-lto-ubuntu builders.
Original message:
This patch updates IPSCCP to use PredicateInfo to propagate
facts to true branches predicated by EQ and to false branches
predicated by NE.
As a follow up, we should be able to extend it to also propagate additional
facts about nonnull.
Reviewers: davide, mssimpso, dberlin, efriedma
Reviewed By: davide, dberlin
llvm-svn: 335206
conditions feeding a chain of `and`s or `or`s for a branch.
Much like with full non-trivial unswitching, we rely on the pass manager
to handle iterating until all of the profitable unswitches have been
done. This is to allow other more profitable unswitches to fire on any
of the cloned, simpler versions of the loop if viable.
Threading the partial unswiching through the non-trivial unswitching
logic motivated some minor refactorings. If those are too disruptive to
make it reasonable to review this patch, I can separate them out, but
it'll be somewhat timeconsuming so I wanted to send it for initial
review as-is. Feel free to tell me whether it warrants pulling apart.
I've tried to re-use (and factor out) logic form the partial trivial
unswitching, but not as much could be shared as I had haped. Still, this
wasn't as bad as I naively expected.
Some basic testing is added, but I probably need more. Suggestions for
things you'd like to see tested more than welcome. One thing I'd like to
do is add some testing that when we schedule this with loop-instsimplify
it effectively cleans up the cruft created.
Last but not least, this uncovered a bug that has been in loop cloning
the entire time for non-trivial unswitching. Specifically, we didn't
correctly add the outer-most cloned loop to the list of cloned loops.
This meant that LCSSA wouldn't be updated for it hypothetically, and
more significantly that we would never visit it in the loop pass
manager. I noticed this while checking loop-instsimplify by hand. I'll
try to separate this bugfix out into its own patch with a more focused
test. But it is just one line, so shouldn't significantly confuse the
review here.
After this patch, the only missing "feature" in this unswitch I'm aware
of us non-trivial unswitching of switches. I'll try implementing *full*
non-trivial unswitching of switches (which is at least a sound thing to
implement), but *partial* non-trivial unswitching of switches is
something I don't see any sound and principled way to implement. I also
have no interesting test cases for the latter, so I'm not really
worried. The rest of the things that need to be ported are bug-fixes and
more narrow / targeted support for specific issues.
Differential Revision: https://reviews.llvm.org/D47522
llvm-svn: 335203
Summary:
Two utils methods have essentially the same functionality. This is an attempt to merge them into one.
1. lib/Transforms/Utils/Local.cpp : MergeBasicBlockIntoOnlyPred
2. lib/Transforms/Utils/BasicBlockUtils.cpp : MergeBlockIntoPredecessor
Prior to the patch:
1. MergeBasicBlockIntoOnlyPred
Updates either DomTree or DeferredDominance
Moves all instructions from Pred to BB, deletes Pred
Asserts BB has single predecessor
If address was taken, replace the block address with constant 1 (?)
2. MergeBlockIntoPredecessor
Updates DomTree, LoopInfo and MemoryDependenceResults
Moves all instruction from BB to Pred, deletes BB
Returns if doesn't have a single predecessor
Returns if BB's address was taken
After the patch:
Method 2. MergeBlockIntoPredecessor is attempting to become the new default:
Updates DomTree or DeferredDominance, and LoopInfo and MemoryDependenceResults
Moves all instruction from BB to Pred, deletes BB
Returns if doesn't have a single predecessor
Returns if BB's address was taken
Uses of MergeBasicBlockIntoOnlyPred that need to be replaced:
1. lib/Transforms/Scalar/LoopSimplifyCFG.cpp
Updated in this patch. No challenges.
2. lib/CodeGen/CodeGenPrepare.cpp
Updated in this patch.
i. eliminateFallThrough is straightforward, but I added using a temporary array to avoid the iterator invalidation.
ii. eliminateMostlyEmptyBlock(s) methods also now use a temporary array for blocks
Some interesting aspects:
- Since Pred is not deleted (BB is), the entry block does not need updating.
- The entry block was being updated with the deleted block in eliminateMostlyEmptyBlock. Added assert to make obvious that BB=SinglePred.
- isMergingEmptyBlockProfitable assumes BB is the one to be deleted.
- eliminateMostlyEmptyBlock(BB) does not delete BB on one path, it deletes its unique predecessor instead.
- adding some test owner as subscribers for the interesting tests modified:
test/CodeGen/X86/avx-cmp.ll
test/CodeGen/AMDGPU/nested-loop-conditions.ll
test/CodeGen/AMDGPU/si-annotate-cf.ll
test/CodeGen/X86/hoist-spill.ll
test/CodeGen/X86/2006-11-17-IllegalMove.ll
3. lib/Transforms/Scalar/JumpThreading.cpp
Not covered in this patch. It is the only use case using the DeferredDominance.
I would defer to Brian Rzycki to make this replacement.
Reviewers: chandlerc, spatel, davide, brzycki, bkramer, javed.absar
Subscribers: qcolombet, sanjoy, nemanjai, nhaehnle, jlebar, tpr, kbarton, RKSimon, wmi, arsenm, llvm-commits
Differential Revision: https://reviews.llvm.org/D48202
llvm-svn: 335183
The idea of partial unswitching is to take a *part* of a branch's
condition that is loop invariant and just unswitching that part. This
primarily makes sense with i1 conditions of branches as opposed to
switches. When dealing with i1 conditions, we can easily extract loop
invariant inputs to a a branch and unswitch them to test them entirely
outside the loop.
As part of this, we now create much more significant cruft in the loop
body, so this relies on adding cleanup passes to the loop pipeline and
revisiting unswitched loops to do that cleanup before continuing to
process them.
This already appears to be more powerful at unswitching than the old
loop unswitch pass, and so I'd appreciate pretty careful review in case
I'm just missing some correctness checks. The `LIV-loop-condition` test
case is not unswitched by the old unswitch pass, but is with this pass.
Thanks to Sanjoy and Fedor for the review!
Differential Revision: https://reviews.llvm.org/D46706
llvm-svn: 335156
Using OrderedInstructions::dominates as comparator for instructions in
BBs without dominance relation can cause a non-deterministic order
between such instructions. That in turn can cause us to materialize
copies in a non-deterministic order. While this does not effect
correctness, it causes some minor non-determinism in the final generated
code, because values have slightly different labels.
Without this patch, running -print-predicateinfo on a reasonably large
module produces slightly different output on each run.
This patch uses the dominator trees DFSInNum to order instruction from
different BBs, which should enforce a deterministic ordering and
guarantee that dominated instructions come after the instructions that
dominate them.
Reviewers: dberlin, efriedma, davide
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D48230
llvm-svn: 335150
D47985 saw the old SK_Alternate 'alternating' shuffle mask replaced with the SK_Select mask which accepts either input operand for each lane, equivalent to a vector select with a constant condition operand.
This patch updates SLPVectorizer to make full use of this SK_Select shuffle pattern by removing the 'isOdd()' limitation.
The AArch64 regression will be fixed by D48172.
Differential Revision: https://reviews.llvm.org/D48174
llvm-svn: 335130
For both operands are unsigned, the following optimizations are valid, and missing:
1. X > Y && X != 0 --> X > Y
2. X > Y || X != 0 --> X != 0
3. X <= Y || X != 0 --> true
4. X <= Y || X == 0 --> X <= Y
5. X > Y && X == 0 --> false
unsigned foo(unsigned x, unsigned y) { return x > y && x != 0; }
should fold to x > y, but I found we haven't done it right now.
besides, unsigned foo(unsigned x, unsigned y) { return x < y && y != 0; }
Has been folded to x < y, so there may be a bug.
Patch by: Li Jia He!
Differential Revision: https://reviews.llvm.org/D47922
llvm-svn: 335129
These are the baseline tests for the functional change in D47922.
Patch by Li Jia He!
Differential Revision: https://reviews.llvm.org/D48000
llvm-svn: 335128
This patch replaces calls to X86-specific intrinsics with floor-ceil semantics
with calls to target-independent @llvm.floor.* and @llvm.ceil.* intrinsics. This
doesn't affect the resulting machine code, as those intrinsics are lowered to
the same instructions, but exposes these specific rounding cases to generic
optimizations.
Differential Revision: https://reviews.llvm.org/D48067
llvm-svn: 335039
LoopSimplifyCFG, being a loop pass, needs to preserve scalar
evolution. This invalidates SE for the loops altered during
block merging.
Differential Revision: https://reviews.llvm.org/D48258
llvm-svn: 335036
This patch adds logic to deal with the following constructions:
%iv = phi i64 ...
%trunc = trunc i64 %iv to i32
%cmp = icmp <pred> i32 %trunc, %invariant
Replacing it with
%iv = phi i64 ...
%cmp = icmp <pred> i64 %iv, sext/zext(%invariant)
In case if it is legal. Specifically, if `%iv` has signed comparison users, it is
required that `sext(trunc(%iv)) == %iv`, and if it has unsigned comparison
uses then we require `zext(trunc(%iv)) == %iv`. The current implementation
bails if `%trunc` has other uses than `icmp`, but in theory we can handle more
cases here (e.g. if the user of trunc is bitcast).
Differential Revision: https://reviews.llvm.org/D47928
Reviewed By: reames
llvm-svn: 335020
This reverts r334428. It incorrectly marks some multiplications as nuw. Tim
Shen is working on a proper fix.
Original commit message:
[SCEV] Add nuw/nsw to mul ops in StrengthenNoWrapFlags where safe.
Summary:
Previously we would add them for adds, but not multiplies.
llvm-svn: 335016
This reverts commit f976cf4cca0794267f28b54e468007fd476d37d9.
I am reverting this because it causes break in a few bots and its going
to take me sometime to look at this.
llvm-svn: 334993
Summary:
Simplify blockaddress usage before giving up in MergeBlockIntoPredecessor
This is a missing small optimization in MergeBlockIntoPredecessor.
This helps with one simplifycfg test which expects this case to be handled.
Reviewers: davide, spatel, brzycki, asbirlea
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D48284
llvm-svn: 334992
redundant-vf2-cost.ll is X86 specific. Moved from
test/Transforms/LoopVectorize/redundant-vf2-cost.ll to
test/Transforms/LoopVectorize/X86/redundant-vf2-cost.ll
llvm-svn: 334854
Summary:
When iterating users of a multiply in processUMulZExtIdiom, the
call to setOperand in the truncation case may replace the use
being visited; make sure the iterator has been advanced before
doing that replacement.
Reviewers: majnemer, davide
Reviewed By: davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D48192
llvm-svn: 334844
This is a minor fix for LV cost model, where the cost for VF=2 was
computed twice when the vectorization of the loop was forced without
specifying a VF.
Reviewers: xusx595, hsaito, fhahn, mkuper
Reviewed By: hsaito, xusx595
Differential Revision: https://reviews.llvm.org/D48048
llvm-svn: 334840
This is r334704 (which was reverted in r334732) with a fix for
types like x86_fp80. We need to use getTypeAllocSizeInBits and
not getTypeStoreSizeInBits to avoid dropping debug info for
such types.
Original commit msg:
> Summary:
> Do not convert a DbgDeclare to DbgValue if the store
> instruction only refer to a fragment of the variable
> described by the DbgDeclare.
>
> Problem was seen when for example having an alloca for an
> array or struct, and there were stores to individual elements.
> In the past we inserted a DbgValue intrinsics for each store,
> just as if the store wrote the whole variable.
>
> When handling store instructions we insert a DbgValue that
> indicates that the variable is "undefined", as we do not know
> which part of the variable that is updated by the store.
>
> When ConvertDebugDeclareToDebugValue is used with a load/phi
> instruction we assert that the referenced value is large enough
> to cover the whole variable. Afaict this should be true for all
> scenarios where those methods are used on trunk. If the assert
> blows in the future I guess we could simply skip to insert a
> dbg.value instruction.
>
> In the future I think we should examine which part of the variable
> that is accessed, and add a DbgValue instrinsic with an appropriate
> DW_OP_LLVM_fragment expression.
>
> Reviewers: dblaikie, aprantl, rnk
>
> Reviewed By: aprantl
>
> Subscribers: JDevlieghere, llvm-commits
>
> Tags: #debug-info
>
> Differential Revision: https://reviews.llvm.org/D48024
llvm-svn: 334830
Summary:
We already do it for splat constants, but not just values.
Also, undef cases are mostly non-functional.
The original commit was reverted because
it broke tests for amdgpu backend, which i didn't check.
Now, the backed was updated to recognize these new
patterns, so we are good.
https://bugs.llvm.org/show_bug.cgi?id=37603https://rise4fun.com/Alive/cplX
Reviewers: spatel, craig.topper, mareko, bogner, rampitec, nhaehnle, arsenm
Reviewed By: spatel, rampitec, nhaehnle
Subscribers: wdng, nhaehnle, llvm-commits
Differential Revision: https://reviews.llvm.org/D47980
llvm-svn: 334818
IEEE 754 defines the expected result on overflow. As far as I know,
hardware implementations (of f16), and compiler-rt (__floatuntisf)
correctly return +-Inf on overflow. And I can't think of any useful
transform that would take advantage of overflow being undefined here.
Differential Revision: https://reviews.llvm.org/D47807
llvm-svn: 334777
This patches teaches EarlyCSE to figure out that if `and i1 %x, %y` is true then both
`%x` and `%y` are true in the taken branch, and if `or i1 %x, %y` is false then both
`%x` and `%y` are false in non-taken branch. Fix for PR37635.
Differential Revision: https://reviews.llvm.org/D47574
Reviewed By: reames
llvm-svn: 334707
Summary:
Do not convert a DbgDeclare to DbgValue if the store
instruction only refer to a fragment of the variable
described by the DbgDeclare.
Problem was seen when for example having an alloca for an
array or struct, and there were stores to individual elements.
In the past we inserted a DbgValue intrinsics for each store,
just as if the store wrote the whole variable.
When handling store instructions we insert a DbgValue that
indicates that the variable is "undefined", as we do not know
which part of the variable that is updated by the store.
When ConvertDebugDeclareToDebugValue is used with a load/phi
instruction we assert that the referenced value is large enough
to cover the whole variable. Afaict this should be true for all
scenarios where those methods are used on trunk. If the assert
blows in the future I guess we could simply skip to insert a
dbg.value instruction.
In the future I think we should examine which part of the variable
that is accessed, and add a DbgValue instrinsic with an appropriate
DW_OP_LLVM_fragment expression.
Reviewers: dblaikie, aprantl, rnk
Reviewed By: aprantl
Subscribers: JDevlieghere, llvm-commits
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D48024
llvm-svn: 334704
IndVarSimplify sometimes makes transforms basing on users that are trivially dead. In particular,
if DCE wasn't run before it, there may be a dead `sext/zext` in loop that will trigger widening
transforms, however it makes no sense to do it.
This patch teaches IndVarsSimplify ignore the mist trivial cases of that.
Differential Revision: https://reviews.llvm.org/D47974
Reviewed By: sanjoy
llvm-svn: 334567
Name table occupies a big chunk of size in current binary format sample profile.
In order to reduce its size, the patch changes the sample writer/reader to
save/restore MD5Hash of names in the name table. Sample annotation phase will
also use MD5Hash of name to query samples accordingly.
Experiment shows compact binary format can reduce the size of sample profile by
2/3 compared with binary format generally.
Differential Revision: https://reviews.llvm.org/D47955
llvm-svn: 334447
Summary:
Previously we would add them for adds, but not multiplies.
Reviewers: sanjoy
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D48038
llvm-svn: 334428
Summary:
`%ret = add nuw i8 %x, C`
From [[ https://llvm.org/docs/LangRef.html#add-instruction | langref ]]:
nuw and nsw stand for “No Unsigned Wrap” and “No Signed Wrap”,
respectively. If the nuw and/or nsw keywords are present,
the result value of the add is a poison value if unsigned
and/or signed overflow, respectively, occurs.
So if `C` is `-1`, `%x` can only be `0`, and the result is always `-1`.
I'm not sure we want to use `KnownBits`/`LVI` here, because there is
exactly one possible value (all bits set, `-1`), so some other pass
should take care of replacing the known-all-ones with constant `-1`.
The `test/Transforms/InstCombine/set-lowbits-mask-canonicalize.ll` change *is* confusing.
What happening is, before this: (omitting `nuw` for simplicity)
1. First, InstCombine D47428/rL334127 folds `shl i32 1, %NBits`) to `shl nuw i32 -1, %NBits`
2. Then, InstSimplify D47883/rL334222 folds `shl nuw i32 -1, %NBits` to `-1`,
3. `-1` is inverted to `0`.
But now:
1. *This* InstSimplify fold `%ret = add nuw i32 %setbit, -1` -> `-1` happens first,
before InstCombine D47428/rL334127 fold could happen.
Thus we now end up with the opposite constant,
and it is all good: https://rise4fun.com/Alive/OA9https://rise4fun.com/Alive/sldC
Was mentioned in D47428 review.
Follow-up for D47883.
Reviewers: spatel, craig.topper
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D47908
llvm-svn: 334298
There could be more than one PHIs in exit block using same loop recurrence.
Don't assume there is only one and fix each user.
Differential Revision: https://reviews.llvm.org/D47788
llvm-svn: 334271
%ret = add nuw i8 %x, C
From langref:
nuw and nsw stand for “No Unsigned Wrap” and “No Signed Wrap”,
respectively. If the nuw and/or nsw keywords are present,
the result value of the add is a poison value if unsigned
and/or signed overflow, respectively, occurs.
So if C is -1, %x can only be 0, and the result is always -1.
https://rise4fun.com/Alive/sldC
Was mentioned in D47428 review.
llvm-svn: 334236
Summary:
`%r = shl nuw i8 C, %x`
As per langref:
```
If the nuw keyword is present, then the shift produces
a poison value if it shifts out any non-zero bits.
```
Thus, if the sign bit is set on `C`, then `%x` can only be `0`,
which means that `%r` can only be `C`.
Or in other words, set sign bit means that the signed value
is negative, so the constant is `<= 0`.
https://rise4fun.com/Alive/WMkhttps://rise4fun.com/Alive/udv
Was mentioned in D47428 review.
We already handle the `0` constant, https://godbolt.org/g/UZq1sJ, so this only handles negative constants.
Could use computeKnownBits() / LazyValueInfo,
but the cost-benefit analysis (https://reviews.llvm.org/D47891)
suggests it isn't worth it.
Reviewers: spatel, craig.topper
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D47883
llvm-svn: 334222
These weren't included in D19544 - probably just an oversight.
D40044 made it more likely that we'll have LLVM math intrinsics rather
than libcalls, so this bug was more easily exposed.
As the tests/code show, we already have the complete mappings for pow/exp/log.
I don't have any experience with SVML, so I don't know if anything else is
missing. It's also not clear to me that we should be doing this transform in
IR rather than DAG/isel, but that's a separate issue.
Differential Revision: https://reviews.llvm.org/D47610
llvm-svn: 334211
%r = shl nuw i8 C, %x
As per langref: If the nuw keyword is present, then the shift produces
a poison value if it shifts out any non-zero bits.
Thus, if the sign bit is set on C, then %x can only be 0,
which means that %r can only be C.
https://rise4fun.com/Alive/WMk
Was mentioned in D47428 review.
llvm-svn: 334200
We do the same thing in rewriteSingleStoreAlloca.
Fixes PR37632.
Reviewers: chandlerc, davide, efriedma
Reviewed By: davide
Differential Revision: https://reviews.llvm.org/D47825
llvm-svn: 334187
The bug report:
https://bugs.llvm.org/show_bug.cgi?id=36036
...requests a DAG change for this, but an IR canonicalization
probably handles most cases. If we still want to match this
pattern in the backend, there's a proposal for that too:
D47831
Alive proofs including nsw/nuw cases that were first noted in:
D46988
https://rise4fun.com/Alive/Kmp
This patch is largely copied from the existing code that was
initially added with:
D40984
...but I didn't see much gain from trying to share code.
llvm-svn: 334137
Summary:
This is [[ https://bugs.llvm.org/show_bug.cgi?id=37603 | PR37603 ]].
https://godbolt.org/g/VCMNpShttps://rise4fun.com/Alive/idM
When doing bit manipulations, it is quite common to calculate some bit mask,
and apply it to some value via `and`.
The typical C code looks like:
```
int mask_signed_add(int nbits) {
return (1 << nbits) - 1;
}
```
which is translated into (with `-O3`)
```
define dso_local i32 @mask_signed_add(int)(i32) local_unnamed_addr #0 {
%2 = shl i32 1, %0
%3 = add nsw i32 %2, -1
ret i32 %3
}
```
But there is a second, less readable variant:
```
int mask_signed_xor(int nbits) {
return ~(-(1 << nbits));
}
```
which is translated into (with `-O3`)
```
define dso_local i32 @mask_signed_xor(int)(i32) local_unnamed_addr #0 {
%2 = shl i32 -1, %0
%3 = xor i32 %2, -1
ret i32 %3
}
```
Since we created such a mask, it is quite likely that we will use it in `and` next.
And then we may get rid of `not` op by folding into `andn`.
But now that i have actually looked:
https://godbolt.org/g/VTUDmU
_some_ backend changes will be needed too.
We clearly loose `bzhi` recognition.
Reviewers: spatel, craig.topper, RKSimon
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D47428
llvm-svn: 334127
Before this patch, debugify would insert debug value intrinsics before the
terminating instruction in a block. This had the advantage of being simple,
but was a bit too simple/unrealistic.
This patch teaches debugify to insert debug values immediately after their
operand defs. This enables better testing of the compiler.
For example, with this patch, `opt -debugify-each` is able to identify a
vectorizer DI-invariance bug fixed in llvm.org/PR32761. In this bug, the
vectorizer produced different output with/without debug info present.
Reverting Davide's bugfix locally, I see:
$ ~/scripts/opt-check-dbg-invar.sh ./bin/opt \
.../SLPVectorizer/AArch64/spillcost-di.ll -slp-vectorizer
Comparing: -slp-vectorizer .../SLPVectorizer/AArch64/spillcost-di.ll
Baseline: /var/folders/j8/t4w0bp8j6x1g6fpghkcb4sjm0000gp/T/tmp.iYYeL1kf
With DI : /var/folders/j8/t4w0bp8j6x1g6fpghkcb4sjm0000gp/T/tmp.sQtQSeet
9,11c9,11
< %5 = getelementptr inbounds %0, %0* %2, i64 %0, i32 1
< %6 = bitcast i64* %4 to <2 x i64>*
< %7 = load <2 x i64>, <2 x i64>* %6, align 8, !tbaa !0
---
> %5 = load i64, i64* %4, align 8, !tbaa !0
> %6 = getelementptr inbounds %0, %0* %2, i64 %0, i32 1
> %7 = load i64, i64* %6, align 8, !tbaa !5
12a13
> store i64 %5, i64* %8, align 8, !tbaa !0
14,15c15
< %10 = bitcast i64* %8 to <2 x i64>*
< store <2 x i64> %7, <2 x i64>* %10, align 8, !tbaa !0
---
> store i64 %7, i64* %9, align 8, !tbaa !5
:: Found a test case ^
Running this over the *.ll files in tree, I found four additional examples
which compile differently with/without DI present. I plan on filing bugs for
these.
llvm-svn: 334118
These were added for D5603 / rL219542, and there's a proposal to
change one side in D47807.
These are tests of constant propagation, so they shouldn't have
ever been tested/housed under InstCombine.
llvm-svn: 334107
If no alignment is set, the abi/preferred alignment of structs will be
used which may be higher than required. This can lead to extra padding
and in the end an increase in data size.
Differential Revision: https://reviews.llvm.org/D47633
llvm-svn: 334099
We should never get different CodeGen based on whether the code is being
compiled in debug mode so we must skip over @llvm.dbg.value (and similar)
calls.
Should fix at least the worst part of PR37690.
llvm-svn: 334090
CodeGenPrepare pass move extension instructions close to load instructions in different BB, so they can be combined later. But the extension instructions can't move through logical and shift instructions in current implementation. This patch enables this enhancement, so we can eliminate more extension instructions.
Differential Revision: https://reviews.llvm.org/D45537
This is re-commit of r331783, which was reverted by r333305. The performance regression was caused by some unlucky alignment, not a code generation problem.
llvm-svn: 334049
When adjusting a cmp in order to canonicalize an abs/nabs select pattern we need
to use the type of the existing operand when creating a new operand not the
type of a select operand, as the two may be different.
This fixes PR37686.
llvm-svn: 334019
As noted in rL333782, we can be both better for optimization and
safer with this transform:
BinOp (shuffle V1, Mask), C --> shuffle (BinOp V1, NewC), Mask
The only potentially unsafe-to-speculate binops are integer div/rem.
All other binops are always safe (although I don't see a way to
assert that in code here).
For opcodes like shifts that can produce poison, it can't matter
here because we know the lanes with undef are dropped by the
subsequent shuffle.
Differential Revision: https://reviews.llvm.org/D47686
llvm-svn: 333962
entries to reach the target. Since these calls don't require type checks,
we can short-circuit them to their real targets, except in cases when they
can be pre-empted.
Differential Revision: https://reviews.llvm.org/D46326
llvm-svn: 333937
When checking a select to see if it matches an abs, allow the true/false values
to be a sign-extension of the comparison value instead of requiring that they're
directly the comparison value, as all the comparison cares about is the sign of
the value.
This fixes a regression due to r333702, where we were no longer generating ctlz
due to isKnownNonNegative failing to match such a pattern.
Differential Revision: https://reviews.llvm.org/D47631
llvm-svn: 333927
After r333856, opt -debugify would just stop emitting debug value
intrinsics after encountering a musttail call. This wasn't sufficient to
avoid verifier failures.
Debug value intrinicss for all instructions preceding a musttail call
must also be emitted before the musttail call.
llvm-svn: 333866
When we optimize select basing on fact that div by 0 is undef
we should not traverse the instruction which are not guaranteed to
transfer execution to next instruction. Guard intrinsic is an example.
Reviewers: spatel, craig.topper
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D47576
llvm-svn: 333864
There's a patchwork of existing transforms trying to handle
these cases, but as seen in the changed test, we weren't
catching them all.
llvm-svn: 333845
As noted in the review thread for rL333782, we could have
made a bug harder to hit if we were simplifying instructions
before trying other folds.
The shuffle transform in question isn't ever a simplification;
it's just a canonicalization. So I've renamed that to make that
clearer.
This is NFCI at this point, but I've regenerated the test file
to show the cosmetic value naming difference of using
instcombine's RAUW vs. the builder.
Possible follow-ups:
1. Move reassociation folds after simplifies too.
2. Refactor common code; we shouldn't have so much repetition.
llvm-svn: 333820
As noted in the review thread for rL333782, we're lacking coverage
for this transform, so add tests for each binop opcode with constant
operand.
llvm-svn: 333818
Summary:
I noticed this issue because we didn't put the primary cloned loop into
the `NonChildClonedLoops` vector and so never iterated on it. Once
I fixed that, it made it clear why I had to do a really complicated and
unnecesasry dance when updating the loops to remain in canonical form --
I was unwittingly working around the fact that the primary cloned loop
wasn't in the expected list of cloned loops. Doh!
Now that we include it in this vector, we don't need to return it and we
can consolidate the update logic as we correctly have a single place
where it can be handled.
I've just added a test for the iteration order aspect as every time
I changed the update logic partially or incorrectly here, an existing
test failed and caught it so that seems well covered (which is also
evidenced by the extensive working around of this missing update).
Reviewers: asbirlea, sanjoy
Subscribers: mcrosier, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D47647
llvm-svn: 333811
Summary:
Getelementptr returns a vector of pointers, instead of a single address,
when one or more of its arguments is a vector. In such case it is not
possible to simplify the expression by inserting a bitcast of operand(0)
into the destination type, as it will create a bitcast between different
sizes.
Reviewers: majnemer, mkuper, mssimpso, spatel
Reviewed By: spatel
Subscribers: lebedev.ri, llvm-commits
Differential Revision: https://reviews.llvm.org/D46379
llvm-svn: 333783
This bug:
https://bugs.llvm.org/show_bug.cgi?id=37648
...was created with the enhancement to this transform with rL332479.
The urem test shows the disaster potential: any undef divisor lane makes
the whole op undef.
The test diffs show that vector demanded elements turns some of the potential,
but not all, unused binop operands back into undef already.
llvm-svn: 333782
Summary:
Emit summaries for bitcode modules that are only destined for the
regular LTO portion of the build so they can participate in
summary-based dead stripping.
This change reduces the size of a nacl_helper build with cfi-icall
enabled by 7%, removing the majority of the overhead due to enabling
cfi-icall. The cfi-icall size increase was caused by compiling in lots
of unused code and cfi-icall generating jumptable references to unused
symbols that could no longer be removed by -Wl,-gc-sections. Increasing
the visibility of summary-based dead stripping prevented jumptable
entries being created for unused symbols from the regular LTO portion
of the build.
Reviewers: pcc
Reviewed By: pcc
Subscribers: dschuff, mehdi_amini, inglorion, eraman, llvm-commits, kcc
Differential Revision: https://reviews.llvm.org/D47594
llvm-svn: 333768
This patch updates IPSCCP to use PredicateInfo to propagate
facts to true branches predicated by EQ and to false branches
predicated by NE.
As a follow up, we should be able to extend it to also propagate additional
facts about nonnull.
Reviewers: davide, mssimpso, dberlin, efriedma
Reviewed By: davide, dberlin
Differential Revision: https://reviews.llvm.org/D45330
llvm-svn: 333740
Summary:
Loop idiom recognize tries to convert loops like
```
int foo(int x) {
int cnt = 0;
while (x) {
x >>= 1;
++cnt;
}
return cnt;
}
```
into calls to ctlz, but if x is initially negative this loop should be infinite.
It happens that the cases that motivated this change have an absolute value of x before the loop. So this patch restricts the transform to cases where we know x is positive. Note: We are relying on the absolute value of INT_MIN to be undefined so we can assume that the result is always positive.
Fixes PR37479
Reviewers: spatel, hfinkel, efriedma, javed.absar
Reviewed By: efriedma
Subscribers: dmgreen, llvm-commits
Differential Revision: https://reviews.llvm.org/D47348
llvm-svn: 333702
This is the planned enhancement to D47163 / rL333611.
We want to match cmp/select sizes because that will be recognized
as min/max more easily and lead to better codegen (especially for
vector types).
As mentioned in D47163, this improves some of the tests that would
also be folded by D46380, so we may want to adjust that patch to
match the new patterns where the extend op occurs after the select.
llvm-svn: 333689
Convert a vector load intrinsic into an llvm load instruction.
This is beneficial when the underlying object being addressed
comes from a constant, since we get constant-folding for free.
Differential Revision: https://reviews.llvm.org/D46273
llvm-svn: 333643
In post-commit review, Eric Christopher notes that many
new MSan warnings are being observed with this patch.
The probable reason is: if 'y' is undef here and we could
evaluate it twice and get different results.
We can't increase the number of uses of a value.
llvm-svn: 333631
Don't always:
cast (select (cmp x, y), z, C) --> select (cmp x, y), (cast z), C'
This is something that came up as far back as D26556, and I lost track of it.
I suspect that this transform is part of the underlying problem that is
inspiring some of the recent proposals that seek to match larger patterns
that include a cast op. Even if that's not true, this transform causes
problems for codegen (particularly with vector types).
A transform to actively match the size of cmp and select operand sizes should
follow. This patch just removes the harmful canonicalization in the other
direction.
Differential Revision: https://reviews.llvm.org/D47163
llvm-svn: 333611
Sanjay Patel