Summary:
RewritePHIs algorithm used in building of CoroFrame inserts a placeholder
```
%placeholder = phi [%val]
```
on every edge leading to a block starting with PHI node with multiple incoming edges,
so that if one of the incoming values was spilled and need to be reloaded, we have a
place to insert a reload. We use SplitEdge helper function to split the incoming edge.
SplitEdge function does not deal with unwind edges comping into a block with an EHPad.
This patch adds an ehAwareSplitEdge function that can correctly split the unwind edge.
For landing pads, we clone the landing pad into every edge block and replace the original
landing pad with a PHI collection the values from all incoming landing pads.
For WinEH pads, we keep the original EHPad in place and insert cleanuppad/cleapret in the
edge blocks.
Reviewers: majnemer, rnk
Reviewed By: majnemer
Subscribers: EricWF, llvm-commits
Differential Revision: https://reviews.llvm.org/D31845
llvm-svn: 303172
There's no need (& a bit incorrect) to mask off the high bits of the
register reference when describing a simple bool value.
Reviewers: aprantl
Differential Revision: https://reviews.llvm.org/D31062
llvm-svn: 303117
ARM Neon has native support for half-sized vector registers (64 bits). This
is beneficial for example for 2D and 3D graphics. This patch adds the option
to lower MinVecRegSize from 128 via a TTI in the SLP Vectorizer.
*** Performance Analysis
This change was motivated by some internal benchmarks but it is also
beneficial on SPEC and the LLVM testsuite.
The results are with -O3 and PGO. A negative percentage is an improvement.
The testsuite was run with a sample size of 4.
** SPEC
* CFP2006/482.sphinx3 -3.34%
A pretty hot loop is SLP vectorized resulting in nice instruction reduction.
This used to be a +22% regression before rL299482.
* CFP2000/177.mesa -3.34%
* CINT2000/256.bzip2 +6.97%
My current plan is to extend the fix in rL299482 to i16 which brings the
regression down to +2.5%. There are also other problems with the codegen in
this loop so there is further room for improvement.
** LLVM testsuite
* SingleSource/Benchmarks/Misc/ReedSolomon -10.75%
There are multiple small SLP vectorizations outside the hot code. It's a bit
surprising that it adds up to 10%. Some of this may be code-layout noise.
* MultiSource/Benchmarks/VersaBench/beamformer/beamformer -8.40%
The opt-viewer screenshot can be seen at F3218284. We start at a colder store
but the tree leads us into the hottest loop.
* MultiSource/Applications/lambda-0.1.3/lambda -2.68%
* MultiSource/Benchmarks/Bullet/bullet -2.18%
This is using 3D vectors.
* SingleSource/Benchmarks/Shootout-C++/Shootout-C++-lists +6.67%
Noise, binary is unchanged.
* MultiSource/Benchmarks/Ptrdist/anagram/anagram +4.90%
There is an additional SLP in the cold code. The test runs for ~1sec and
prints out over 2000 lines. This is most likely noise.
* MultiSource/Applications/aha/aha +1.63%
* MultiSource/Applications/JM/lencod/lencod +1.41%
* SingleSource/Benchmarks/Misc/richards_benchmark +1.15%
Differential Revision: https://reviews.llvm.org/D31965
llvm-svn: 303116
This caused PR33053.
Original commit message:
> The new experimental reduction intrinsics can now be used, so I'm enabling this
> for AArch64. We will need this for SVE anyway, so it makes sense to do this for
> NEON reductions as well.
>
> The existing code to match shufflevector patterns are replaced with a direct
> lowering of the reductions to AArch64-specific nodes. Tests updated with the
> new, simpler, representation.
>
> Differential Revision: https://reviews.llvm.org/D32247
llvm-svn: 303115
Summary:
The following loops should be recognized:
i = 0;
while (n) {
n = n >> 1;
i++;
body();
}
use(i);
And replaced with builtin_ctlz(n) if body() is empty or
for CPUs that have CTLZ instruction converted to countable:
for (j = 0; j < builtin_ctlz(n); j++) {
n = n >> 1;
i++;
body();
}
use(builtin_ctlz(n));
Reviewers: rengolin, joerg
Differential Revision: http://reviews.llvm.org/D32605
From: Evgeny Stupachenko <evstupac@gmail.com>
llvm-svn: 303102
verifyMemoryCongruency() filters out trivially dead MemoryDef(s),
as we find them immediately dead, before moving from TOP to a new
congruence class.
This fixes the same problem for PHI(s) skipping MemoryPhis if all
the operands are dead.
Differential Revision: https://reviews.llvm.org/D33044
llvm-svn: 303100
Summary:
If the Worklist build causes an IR change this change flag currently factors into the flag for running another iteration of the iteration loop. But only changes during processing should trigger another loop.
This patch captures the worklist creation change flag into the outside the loop flag currently used for DbgDeclares and only sends that flag up to the caller. Rerunning the loop only depends on IC.run() now.
This uses the debug output of InstCombine to determine if one or two iterations run. I couldn't think of a better way to detect it since the second spurious iteration shoudn't make any visible changes. Just wasted computation.
I can do a pre-commit of the test case with the CHECK-NOT as a CHECK if this is an ok way to check this.
This is a subset of D31678 as I'm still not sure how to verify the analysis behavior for that.
Reviewers: davide, majnemer, spatel, chandlerc
Reviewed By: davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32453
llvm-svn: 302982
Tests with target intrinsics are inherently target specific, so it
doesn't actually make sense to run them if we've excluded their
target.
llvm-svn: 302979
I bet the change is correct but this test seems to expose some underlying
problem that manifest only on some buildbots, and I'm not able to reproduce
locally. Unfortunately I can't debug right now but I don't want to annoy
people with spurious failures, so I'll XFAIL until I can take a look (over
the weekend).
llvm-svn: 302976
Implemented frequency based cost/saving analysis
and related options.
The pass is now in a state ready to be turne on
in the pipeline (in follow up).
Differential Revision: http://reviews.llvm.org/D32783
llvm-svn: 302967
This code was missing a check for stores, so we were thinking the
congruency class didn't have any memory members, and reset the
memory leader.
Differential Revision: https://reviews.llvm.org/D33056
llvm-svn: 302905
invariant PHI inputs and to rewrite PHI nodes during the actual
unswitching.
The checking is quite easy, but rewriting the PHI nodes is somewhat
surprisingly challenging. This should handle both branches and switches.
I think this is now a full featured trivial unswitcher, and more full
featured than the trivial cases in the old pass while still being (IMO)
somewhat simpler in how it works.
Next up is to verify its correctness in more widespread testing, and
then to add non-trivial unswitching.
Thanks to Davide and Sanjoy for the excellent review. There is one
remaining question that I may address in a follow-up patch (see the
review thread for details) but it isn't related to the functionality
specifically.
Differential Revision: https://reviews.llvm.org/D32699
llvm-svn: 302867
Summary:
Don't use the metadata on call instructions for determining hotness
unless we are in sample PGO mode, where it is needed because profile
counts are not accurate. In instrumentation mode this is not necessary
and does more harm than good when calls have VP metadata that hasn't
been properly scaled after transformations or dropped after constant
prop based devirtualization (both should be fixed, but we don't need
to do this in the first place for instrumentation PGO).
This required adjusting a number of tests to distinguish between sample
and instrumentation PGO handling, and to add in profile summary metadata
so that getProfileCount can get the summary.
Reviewers: davidxl, danielcdh
Subscribers: aemerson, rengolin, mehdi_amini, Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D32877
llvm-svn: 302844
I ran the test-suite (including SPEC 2006) in PGO mode comparing cold
thresholds of 225 and 45. Here are some stats on the text size:
Out of 904 tests that ran, 197 see a change in text size. The average
text size reduction (of all the 904 binaries) is 1.07%. Of the 197
binaries, 19 see a text size increase, as high as 18%, but most of them
are small single source benchmarks. There are 3 multisource benchmarks
with a >0.5% size increase (0.7, 1.3 and 2.1 are their % increases). On
the other side of the spectrum, 31 benchmarks see >10% size reduction
and 6 of them are MultiSource.
I haven't run the test-suite with other values of inlinecold-threshold.
Since we have a cold callsite threshold of 45, I picked this value.
Differential revision: https://reviews.llvm.org/D33106
llvm-svn: 302829
The approach I followed was to emit the remark after getTreeCost concludes
that SLP is profitable. I initially tried emitting them after the
vectorizeRootInstruction calls in vectorizeChainsInBlock but I vaguely
remember missing a few cases for example in HorizontalReduction::tryToReduce.
ORE is placed in BoUpSLP so that it's available from everywhere (notably
HorizontalReduction::tryToReduce).
We use the first instruction in the root bundle as the locator for the remark.
In order to get a sense how far the tree is spanning I've include the size of
the tree in the remark. This is not perfect of course but it gives you at
least a rough idea about the tree. Then you can follow up with -view-slp-tree
to really see the actual tree.
llvm-svn: 302811
// (X ^ C1) | C2 --> (X | C2) ^ (C1&~C2)
This canonicalization was added at:
https://reviews.llvm.org/rL7264
By moving xors out/down, we can more easily combine constants. I'm adding
tests that do not change with this patch, so we can verify that those kinds
of transforms are still happening.
This is no-functional-change-intended because there's a later fold:
// (X^C)|Y -> (X|Y)^C iff Y&C == 0
...and demanded-bits appears to guarantee that any fold that would have
hit the fold we're removing here would be caught by that 2nd fold.
Similar reasoning was used in:
https://reviews.llvm.org/rL299384
The larger motivation for removing this code is that it could interfere with
the fix for PR32706:
https://bugs.llvm.org/show_bug.cgi?id=32706
Ie, we're not checking if the 'xor' is actually a 'not', so we could reverse
a 'not' optimization and cause an infinite loop by altering an 'xor X, -1'.
Differential Revision: https://reviews.llvm.org/D33050
llvm-svn: 302733
The new experimental reduction intrinsics can now be used, so I'm enabling this
for AArch64. We will need this for SVE anyway, so it makes sense to do this for
NEON reductions as well.
The existing code to match shufflevector patterns are replaced with a direct
lowering of the reductions to AArch64-specific nodes. Tests updated with the
new, simpler, representation.
Differential Revision: https://reviews.llvm.org/D32247
llvm-svn: 302678
The first test in this file is duplicated exactly in and.ll -> test33.
We have commuted and vector variants there too.
The second test is a composite of 2 folds. The first fold is tested
independently in add.ll -> flip_and_mask (including vector variant).
After that transform fires, the IR is identical to the first transform.
llvm-svn: 302676
The script at utils/update_test_checks.py has (had?) a bug when variables
start with the same sequence of letters (clearly, not all of the time).
llvm-svn: 302674
This is another step towards favoring 'not' ops over random 'xor' in IR:
https://bugs.llvm.org/show_bug.cgi?id=32706
This transformation may have occurred in longer IR sequences using computeKnownBits,
but that could be much more expensive to calculate.
As the scalar result shows, we do not currently favor 'not' in all cases. The 'not'
created by the transform is transformed again (unnecessarily). Vectors don't have
this problem because vectors are (wrongly) excluded from several other combines.
llvm-svn: 302659
This pass doesn't correctly handle testing for when it is legal to hoist
arbitrary instructions. The whitelist happens to make it safe, so before
it is removed the pass's legality checks will need to be enhanced.
Details have been added to the code review thread for the patch.
llvm-svn: 302640
Summary:
This fixes the immediate crash caused by introducing an incorrect inttoptr
before attempting the conversion. There may still be a legality
check missing somewhere earlier for non-integral pointers, but this change
seems necessary in any case.
Reviewers: sanjoy, dberlin
Reviewed By: dberlin
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32623
llvm-svn: 302587
The way we currently define congruency for two PHIExpression(s) is:
1) The operands to the phi functions are congruent
2) The PHIs are defined in the same BasicBlock.
NewGVN works under the assumption that phi operands are in predecessor
order, or at least in some consistent order. OTOH, is valid IR:
patatino:
%meh = phi i16 [ %0, %winky ], [ %conv1, %tinky ]
%banana = phi i16 [ %0, %tinky ], [ %conv1, %winky ]
br label %end
and the in-memory representations of the two SSA registers have an
inconsistent order. This violation of NewGVN assumptions results into
two PHIs found congruent when they're not. While we think it's useful
to have always a consistent order enforced, let's fix this in NewGVN
sorting uses in predecessor order before creating a PHI expression.
Differential Revision: https://reviews.llvm.org/D32990
llvm-svn: 302552
The comment says to avoid the case where zero bits are shifted into the truncated value,
but the code checks that the shift is smaller than the truncated value instead of the
number of bits added by the sign extension. Fixing this allows a shift by more than the
value size to be introduced, which is undefined behavior, so the shift is capped at the
value size minus one, which has the expected behavior of filling the value with the sign
bit.
Patch by Jacob Young!
Differential Revision: https://reviews.llvm.org/D32285
llvm-svn: 302548
Summary:
In first order recurrence vectorization, when the previous value is a phi node, we need to
set the insertion point to the first non-phi node.
We can have the previous value being a phi node, due to the generation of new
IVs as part of trunc optimization [1].
[1] https://reviews.llvm.org/rL294967
Reviewers: mssimpso, mkuper
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D32969
llvm-svn: 302532
Summary:
r284533 added hot and cold section prefixes based on profile
information, to enable grouping of hot/cold functions at link time.
However, it used "cold" as the prefix for cold sections, but gold only
recognizes "unlikely" (which is used by gcc for cold sections).
Therefore, cold sections were not properly being grouped. Switch to
using "unlikely"
Reviewers: danielcdh, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32983
llvm-svn: 302502
This is another step towards getting rid of dyn_castNotVal,
so we can recommit:
https://reviews.llvm.org/rL300977
As the tests show, we were missing the lshr case for constants
and both ashr/lshr vector splat folds. The ashr case with constant
was being performed inefficiently in 2 steps. It's also possible
there was a latent bug in that case because we can't do that fold
if the constant is positive:
http://rise4fun.com/Alive/Bge
llvm-svn: 302465
Transforms/IndVarSimplify/2011-10-27-lftrnull will fail if this regresses.
Transforms/GVN/PRE/2011-06-01-NonLocalMemdepMiscompile.ll has been changed to still test what it was
trying to test.
llvm-svn: 302446
This patch uses KnownOnes of the input of ctlz/cttz to bound the value that can be returned from these intrinsics. This makes these intrinsics more similar to the handling for ctpop which already uses known bits to produce a similar bound.
Differential Revision: https://reviews.llvm.org/D32521
llvm-svn: 302444
Summary:
Re-applying r301766 with a fix to a typo and a regression test.
The log message for r301766 was:
==================================================================================
InstructionSimplify: Canonicalize shuffle operands. NFC-ish.
Summary:
Apply canonicalization rules:
1. Input vectors with no elements selected from can be replaced with undef.
2. If only one input vector is constant it shall be the second one.
This allows constant-folding to cover more ad-hoc simplifications that
were in place and avoid duplication for RHS and LHS checks.
There are more rules we may want to add in the future when we see a
justification. e.g. mask elements that select undef elements can be
replaced with undef.
==================================================================================
Reviewers: spatel, RKSimon
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32863
llvm-svn: 302373
We can simplify (or (icmp X, C1), (icmp X, C2)) to 'true' or one of the icmps in many cases.
I had to check some of these with Alive to prove to myself it's right, but everything seems
to check out. Eg, the deleted code in instcombine was completely ignoring predicates with
mismatched signedness.
This is a follow-up to:
https://reviews.llvm.org/rL301260https://reviews.llvm.org/D32143
llvm-svn: 302370
Loop Idiom recognition was generating memset in a case that
would result generating a division operation to an unsafe location.
Differential Revision: https://reviews.llvm.org/D32674
llvm-svn: 302238
The sibling folds for 'and' with casts were added with https://reviews.llvm.org/rL273200.
This is a preliminary step for adding the 'or' variants for the folds added with https://reviews.llvm.org/rL301260.
The reason for the strange form with constant LHS in the 1st test is because there's another missing fold in that
case for the inverted predicate. That should be fixed when we add the ConstantRange functionality for 'or-of-icmps'
that already exists for 'and-of-icmps'.
I'm hoping to share more code for the and/or cases, so we won't have these differences. This will allow us to remove
code from InstCombine. It's also possible that we can remove some code here in InstSimplify. I think we have some
duplicated folds because patterns are not matched in a general way.
Differential Revision: https://reviews.llvm.org/D32876
llvm-svn: 302189
Change checkRippleForAdd from a heuristic to a full check -
if it is provable that the add does not overflow return true, otherwise false.
Patch by Yoav Ben-Shalom
Differential Revision: https://reviews.llvm.org/D32686
llvm-svn: 302093
Fixes PR31789 - When loop-vectorize tries to use these intrinsics for a
non-default address space pointer we fail with a "Calling a function with a
bad singature!" assertion. This patch solves this by adding the 'vector of
pointers' argument as an overloaded type which will determine the address
space.
Differential revision: https://reviews.llvm.org/D31490
llvm-svn: 302018
Summary:
Currently, loop deletion deletes loop where the only values
that are used outside the loop are loop-invariant.
This patch adds logic to delete loops where the loop is proven to be
never executed (i.e. the only predecessor of the loop preheader has a
constant conditional branch as terminator, and the preheader is not the
taken target). This will remove loops that become dead after
loop-unswitching generates constant conditional branches.
The next steps are:
1. moving the loop deletion implementation to LoopUtils.
2. Add logic in loop-simplifyCFG which will support changing conditional
constant branches to unconditional branches. If loops become unreachable in this
process, they can be removed using `deleteDeadLoop` function.
Reviewers: chandlerc, efriedma, sanjoy, reames
Reviewed by: sanjoy
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D32494
llvm-svn: 302015
We should always expect values to be named before running the module summary
analysis (see NameAnonGlobals pass), so it's fine if we crash in that case.
llvm-svn: 301991
This was originally checked in here:
https://reviews.llvm.org/rL301923
And reverted here:
https://reviews.llvm.org/rL301924
Because there's a clang test that would fail after this. I fixed/removed the
offending CHECK lines in:
https://reviews.llvm.org/rL301928
So let's try this again. Original commit message:
This is the fold that causes the infinite loop in BoringSSL
(https://github.com/google/boringssl/blob/master/crypto/cipher/e_rc2.c)
when we fix instcombine demanded bits to prefer 'not' ops as in https://reviews.llvm.org/D32255.
There are 2 or 3 problems with dyn_castNotVal, and I don't think we can
reinstate https://reviews.llvm.org/D32255 until dyn_castNotVal is completely eliminated.
1. As shown here, it transforms 'not' into random xor. This transform is harmful to SCEV and codegen because 'not' can often be folded while random xor cannot.
2. It does not transform vector constants. This is actually a good thing, but if you don't believe the above argument, then we shouldn't have excluded vectors.
3. It tries to avoid transforming not(not(X)). That's nice, but it doesn't match the greedy nature of instcombine. If we DeMorganize a pattern that has an extra 'not' in it: ~(~(~X) & Y) --> (~X | ~Y)
That's just another case of DeMorgan, so we should trust that we'll fold that pattern too: (~X | ~ Y) --> ~(X & Y)
Differential Revision: https://reviews.llvm.org/D32665
llvm-svn: 301929
This is the fold that causes the infinite loop in BoringSSL
(https://github.com/google/boringssl/blob/master/crypto/cipher/e_rc2.c)
when we fix instcombine demanded bits to prefer 'not' ops as in D32255.
There are 2 or 3 problems with dyn_castNotVal, and I don't think we can
reinstate D32255 until dyn_castNotVal is completely eliminated.
1. As shown here, it transforms 'not' into random xor. This transform is
harmful to SCEV and codegen because 'not' can often be folded while
random xor cannot.
2. It does not transform vector constants. This is actually a good thing,
but if you don't believe the above argument, then we shouldn't have
excluded vectors.
3. It tries to avoid transforming not(not(X)). That's nice, but it doesn't
match the greedy nature of instcombine. If we DeMorganize a pattern
that has an extra 'not' in it:
~(~(~X) & Y) --> (~X | ~Y)
That's just another case of DeMorgan, so we should trust that we'll fold
that pattern too:
(~X | ~ Y) --> ~(X & Y)
Differential Revision: https://reviews.llvm.org/D32665
llvm-svn: 301923
This change caused buildbot failures, apparently because we're not
passing around types that InstSimplify is used to seeing. I'm not overly
familiar with InstSimplify, so I'm reverting this until I can figure out
what exactly is wrong.
llvm-svn: 301885
In particular (since it wouldn't fit nicely in the summary):
(select (icmp eq V 0) P (getelementptr P V)) -> (getelementptr P V)
Differential Revision: https://reviews.llvm.org/D31435
llvm-svn: 301880
In the testcase attached, we believe %tmp1 implies %tmp4.
where:
br i1 %tmp1, label %bb2, label %bb7
br i1 %tmp4, label %bb5, label %bb7
because Wwhile looking at PredicateInfo stuffs we end up calling
isImpliedTrueByMatchingCmp() with the arguments backwards.
Differential Revision: https://reviews.llvm.org/D32718
llvm-svn: 301849
If we have ~(~X & Y), it only makes sense to transform it to (X | ~Y) when we do not need
the intermediate (~X & Y) value. In that case, we would need an extra instruction to
generate ~Y + 'or' (as shown in the test changes).
It's ok if we have multiple uses of ~X or Y, however. In those cases, we may not reduce the
instruction count or critical path, but we might improve throughput because we can generate
~X and ~Y in parallel. Whether that actually makes perf sense or not for a target is something
we can't answer in IR.
Differential Revision: https://reviews.llvm.org/D32703
llvm-svn: 301848
We may not be able to rewrite indirect branch target, but we also want to take it into
account when folding, i.e. if it and all its successor's predecessors go to the same
destination, we can fold, i.e. no need to thread.
llvm-svn: 301816
Summary: [JumpThread] Do RAUW in case Cond folds to a constant in the CFG
Reviewers: sanjoy
Reviewed By: sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32407
llvm-svn: 301804
retainAutoreleasedReturnValue that retains the returned value.
This commit fixes a bug in ARC optimizer where it moves a release
between a call and a retainAutoreleasedReturnValue, causing the returned
object to be released before the retainAutoreleasedReturnValue can
retain it.
This commit accomplishes that by doing a lookahead and checking whether
the call prevents the release from moving upwards. In the long term, we
should treat the region between the retainAutoreleasedReturnValue and
the call as a critical section and disallow moving anything there
(possibly using operand bundles).
rdar://problem/20449878
llvm-svn: 301724
Eliminates some more cases where some subset of the addressing
computation remains flat. Some cases with addrspacecasts
in nested constant expressions are still left behind however.
llvm-svn: 301704
Summary:
The motivation example is like below which has 13 cases but only 2 distinct targets
```
lor.lhs.false2: ; preds = %if.then
switch i32 %Status, label %if.then27 [
i32 -7012, label %if.end35
i32 -10008, label %if.end35
i32 -10016, label %if.end35
i32 15000, label %if.end35
i32 14013, label %if.end35
i32 10114, label %if.end35
i32 10107, label %if.end35
i32 10105, label %if.end35
i32 10013, label %if.end35
i32 10011, label %if.end35
i32 7008, label %if.end35
i32 7007, label %if.end35
i32 5002, label %if.end35
]
```
which is compiled into a balanced binary tree like this on AArch64 (similar on X86)
```
.LBB853_9: // %lor.lhs.false2
mov w8, #10012
cmp w19, w8
b.gt .LBB853_14
// BB#10: // %lor.lhs.false2
mov w8, #5001
cmp w19, w8
b.gt .LBB853_18
// BB#11: // %lor.lhs.false2
mov w8, #-10016
cmp w19, w8
b.eq .LBB853_23
// BB#12: // %lor.lhs.false2
mov w8, #-10008
cmp w19, w8
b.eq .LBB853_23
// BB#13: // %lor.lhs.false2
mov w8, #-7012
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_14: // %lor.lhs.false2
mov w8, #14012
cmp w19, w8
b.gt .LBB853_21
// BB#15: // %lor.lhs.false2
mov w8, #-10105
add w8, w19, w8
cmp w8, #9 // =9
b.hi .LBB853_17
// BB#16: // %lor.lhs.false2
orr w9, wzr, #0x1
lsl w8, w9, w8
mov w9, #517
and w8, w8, w9
cbnz w8, .LBB853_23
.LBB853_17: // %lor.lhs.false2
mov w8, #10013
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_18: // %lor.lhs.false2
mov w8, #-7007
add w8, w19, w8
cmp w8, #2 // =2
b.lo .LBB853_23
// BB#19: // %lor.lhs.false2
mov w8, #5002
cmp w19, w8
b.eq .LBB853_23
// BB#20: // %lor.lhs.false2
mov w8, #10011
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_21: // %lor.lhs.false2
mov w8, #14013
cmp w19, w8
b.eq .LBB853_23
// BB#22: // %lor.lhs.false2
mov w8, #15000
cmp w19, w8
b.ne .LBB853_3
```
However, the inline cost model estimates the cost to be linear with the number
of distinct targets and the cost of the above switch is just 2 InstrCosts.
The function containing this switch is then inlined about 900 times.
This change use the general way of switch lowering for the inline heuristic. It
etimate the number of case clusters with the suitability check for a jump table
or bit test. Considering the binary search tree built for the clusters, this
change modifies the model to be linear with the size of the balanced binary
tree. The model is off by default for now :
-inline-generic-switch-cost=false
This change was originally proposed by Haicheng in D29870.
Reviewers: hans, bmakam, chandlerc, eraman, haicheng, mcrosier
Reviewed By: hans
Subscribers: joerg, aemerson, llvm-commits, rengolin
Differential Revision: https://reviews.llvm.org/D31085
llvm-svn: 301649
Summary:
Skip memops if the total value profiled count is 0, we can't correctly
scale up the counts and there is no point anyway.
Reviewers: davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32624
llvm-svn: 301645
EarlyCSE should not just ignore assumes. It should use the fact that its condition is true for all dominated instructions.
Reviewers: sanjoy, reames, apilipenko, anna, skatkov
Reviewed By: reames, sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32482
llvm-svn: 301625
If a condition is calculated only once, and there are multiple guards on this condition, we should be able
to remove all guards dominated by the first of them. This patch allows EarlyCSE to try to find the condition
of a guard among the known values, and if it is true, remove the guard. Otherwise we keep the guard and
mark its condition as 'true' for future consideration.
Reviewers: sanjoy, reames, apilipenko, skatkov, anna, dberlin
Reviewed By: reames, sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32476
llvm-svn: 301623
Currently, this pass only focuses on *trivial* loop unswitching. At that
reduced problem it remains significantly better than the current loop
unswitch:
- Old pass is worse than cubic complexity. New pass is (I think) linear.
- New pass is much simpler in its design by focusing on full unswitching. (See
below for details on this).
- New pass doesn't carry state for thresholds between pass iterations.
- New pass doesn't carry state for correctness (both miscompile and
infloop) between pass iterations.
- New pass produces substantially better code after unswitching.
- New pass can handle more trivial unswitch cases.
- New pass doesn't recompute the dominator tree for the entire function
and instead incrementally updates it.
I've ported all of the trivial unswitching test cases from the old pass
to the new one to make sure that major functionality isn't lost in the
process. For several of the test cases I've worked to improve the
precision and rigor of the CHECKs, but for many I've just updated them
to handle the new IR produced.
My initial motivation was the fact that the old pass carried state in
very unreliable ways between pass iterations, and these mechansims were
incompatible with the new pass manager. However, I discovered many more
improvements to make along the way.
This pass makes two very significant assumptions that enable most of these
improvements:
1) Focus on *full* unswitching -- that is, completely removing whatever
control flow construct is being unswitched from the loop. In the case
of trivial unswitching, this means removing the trivial (exiting)
edge. In non-trivial unswitching, this means removing the branch or
switch itself. This is in opposition to *partial* unswitching where
some part of the unswitched control flow remains in the loop. Partial
unswitching only really applies to switches and to folded branches.
These are very similar to full unrolling and partial unrolling. The
full form is an effective canonicalization, the partial form needs
a complex cost model, cannot be iterated, isn't canonicalizing, and
should be a separate pass that runs very late (much like unrolling).
2) Leverage LLVM's Loop machinery to the fullest. The original unswitch
dates from a time when a great deal of LLVM's loop infrastructure was
missing, ineffective, and/or unreliable. As a consequence, a lot of
complexity was added which we no longer need.
With these two overarching principles, I think we can build a fast and
effective unswitcher that fits in well in the new PM and in the
canonicalization pipeline. Some of the remaining functionality around
partial unswitching may not be relevant today (not many test cases or
benchmarks I can find) but if they are I'd like to add support for them
as a separate layer that runs very late in the pipeline.
Purely to make reviewing and introducing this code more manageable, I've
split this into first a trivial-unswitch-only pass and in the next patch
I'll add support for full non-trivial unswitching against a *fixed*
threshold, exactly like full unrolling. I even plan to re-use the
unrolling thresholds, as these are incredibly similar cost tradeoffs:
we're cloning a loop body in order to end up with simplified control
flow. We should only do that when the total growth is reasonably small.
One of the biggest changes with this pass compared to the previous one
is that previously, each individual trivial exiting edge from a switch
was unswitched separately as a branch. Now, we unswitch the entire
switch at once, with cases going to the various destinations. This lets
us unswitch multiple exiting edges in a single operation and also avoids
numerous extremely bad behaviors, where we would introduce 1000s of
branches to test for thousands of possible values, all of which would
take the exact same exit path bypassing the loop. Now we will use
a switch with 1000s of cases that can be efficiently lowered into
a jumptable. This avoids relying on somehow forming a switch out of the
branches or getting horrible code if that fails for any reason.
Another significant change is that this pass actively updates the CFG
based on unswitching. For trivial unswitching, this is actually very
easy because of the definition of loop simplified form. Doing this makes
the code coming out of loop unswitch dramatically more friendly. We
still should run loop-simplifycfg (at the least) after this to clean up,
but it will have to do a lot less work.
Finally, this pass makes much fewer attempts to simplify instructions
based on the unswitch. Something like loop-instsimplify, instcombine, or
GVN can be used to do increasingly powerful simplifications based on the
now dominating predicate. The old simplifications are things that
something like loop-instsimplify should get today or a very, very basic
loop-instcombine could get. Keeping that logic separate is a big
simplifying technique.
Most of the code in this pass that isn't in the old one has to do with
achieving specific goals:
- Updating the dominator tree as we go
- Unswitching all cases in a switch in a single step.
I think it is still shorter than just the trivial unswitching code in
the old pass despite having this functionality.
Differential Revision: https://reviews.llvm.org/D32409
llvm-svn: 301576
Just calling dropAllReferences leaves pointers to the ConstantExpr
behind, so we would eventually crash with a null pointer dereference.
Differential Revision: https://reviews.llvm.org/D32551
llvm-svn: 301575
also a discussion about exactly what we should do prior to re-enabling
it.
The current bug is http://llvm.org/PR32821 and the discussion about this
is in the review thread for r300200.
llvm-svn: 301505
This patch is part of D28975's breakdown.
induction.ll encodes the specific (and rather arbitrary) numbers given to
predicated basic blocks by the unique naming mechanism, which makes it
sensitive to changes in LV's instruction generation order. This patch replaces
those specific numbers with a numeric pattern.
Differential Revision: https://reviews.llvm.org/D32404
llvm-svn: 301345
The code Sanjay Patel moved over from InstCombine doesn't work properly if the 'and' has both inputs as nots because we used a commuted op matcher on the 'and' first. But this will bind to the first 'not' on 'and' when there could be two 'not's. InstCombine could rely on DeMorgan to ensure the 'and' wouldn't have two 'not's eventually, but InstSimplify can't rely on that.
This patch matches the xor first then checks for the ands and allows a not of either operand of the xor.
Differential Revision: https://reviews.llvm.org/D32458
llvm-svn: 301329
The matching here wasn't able to handle all the possible commutes. It always assumed the not would be on the left of the xor, but that's not guaranteed.
Differential Revision: https://reviews.llvm.org/D32474
llvm-svn: 301316
One of the fast-math optimizations is to replace calls to standard double
functions with their float equivalents, e.g. exp -> expf. However, this can
cause infinite loops for the following:
float expf(float val) { return (float) exp((double) val); }
A similar inline declaration exists in the MinGW-w64 math.h header file which
when compiled with -O2/3 and fast-math generates infinite loops.
So this fix checks that the calling function to the standard double function
that is being replaced does not match the float equivalent.
Differential Revision: https://reviews.llvm.org/D31806
llvm-svn: 301304
We need to do this to prevent a miscompile which sinks an objc_retain
past an objc_release that releases the object objc_retain retains. This
happens because the top-down and bottom-up traversals each determines
the insert point for retain or release individually without knowing
where the other instruction is moved.
For example, when the following IR is fed to the ARC optimizer, the
top-down traversal decides to insert objc_retain right before
objc_release and the bottom-up traversal decides to insert objc_release
right after clang.arc.use.
(IR before ARC optimizer)
%11 = call i8* @objc_retain(i8* %10)
call void (...) @clang.arc.use(%0* %5)
call void @llvm.dbg.value(...)
call void @objc_release(i8* %6)
This reverses the order of objc_release and objc_retain, which causes
the object to be destructed prematurely.
(IR after ARC optimizer)
call void (...) @clang.arc.use(%0* %5)
call void @objc_release(i8* %6)
call void @llvm.dbg.value(...)
%11 = call i8* @objc_retain(i8* %10)
rdar://problem/30530580
llvm-svn: 301289
We can simplify (and (icmp X, C1), (icmp X, C2)) to one of the icmps in many cases.
I had to check some of these with Alive to prove to myself it's right, but everything
seems to check out. Eg, the code in instcombine was completely ignoring predicates with
mismatched signedness.
Handling or-of-icmps would be a follow-up step.
Differential Revision: https://reviews.llvm.org/D32143
llvm-svn: 301260
Summary:
Ensure that the new merge BB (which contains the rest of the original BB
after the mem op being optimized) gets a profile frequency, in case
there are additional mem ops later in the BB. Otherwise they get skipped
as the merge BB looks cold.
Reviewers: davidxl, xur
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32447
llvm-svn: 301244
This reverts commit r300732. This breaks a few tests.
I think the problem is related to adding more uses of
the condition that don't yet exist at this point.
llvm-svn: 301242
The current Loop Unroll implementation works with loops having a
single latch that contains a conditional branch to a block outside
the loop (the other successor is, by defition of latch, the header).
If this precondition doesn't hold, avoid unrolling the loop as
the code is not ready to handle such circumstances.
Differential Revision: https://reviews.llvm.org/D32261
llvm-svn: 301239
Summary:
llvm.invariant.group.barrier returns pointer that mustalias
pointer it takes. It can't be marked with `returned` attribute,
because it would be remove easily. The other reason is that
only Alias Analysis can know about this, because if any other
pass would know it, then the result would be replaced with it's
argument, which would be invalid.
We can think about returned pointer as something that mustalias, but
it doesn't have to be bitwise the same as the argument.
Reviewers: dberlin, chandlerc, hfinkel, sanjoy
Subscribers: reames, nlewycky, rsmith, anna, amharc
Differential Revision: https://reviews.llvm.org/D31585
llvm-svn: 301227
This is a straight cut and paste, but there's a bigger problem: if this
fold exists for simplifyOr, there should be a DeMorganized version for
simplifyAnd. But more than that, we have a patchwork of ad hoc logic
optimizations in InstCombine. There should be some structure to ensure
that we're not missing sibling folds across and/or/xor.
llvm-svn: 301213
When the location description of a source variable involves arithmetic
on the value itself, it needs to be marked with DW_OP_stack_value since it
is not describing the variable's location, but rather its value.
This is a follow-up to r297971 and fixes the source testcase quoted in
the comment in debuginfo-dce.ll.
rdar://problem/30725338
This reapplies r301093 without modifications.
llvm-svn: 301210
There is logic to track the expected number of instructions
produced. It thought in this case an instruction would
be necessary to negate the result, but here it folded
into a ConstantExpr fneg when the non-undef value operand
was cancelled out by the second fsub.
I'm not sure why we don't fold constant FP ops with undef currently,
but I think that would also avoid this problem.
llvm-svn: 301199
Summary:
Instead of keeping a variable indicating whether there are early exits
in the loop. We keep all the early exits. This improves LICM's ability to
move instructions out of the loop based on is-guaranteed-to-execute.
I am going to update compilation time as well soon.
Reviewers: hfinkel, sanjoy, efriedma, mkuper
Reviewed By: hfinkel
Subscribers: llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D32433
llvm-svn: 301196
Summary:
The return value of these intrinsics should always have 0 bits for
inactive threads. This means that when all arguments are constant
and the comparison evaluates to true, the intrinsic should return
the current exec mask.
Fixes some GL_ARB_shader_ballot tests.
Reviewers: arsenm
Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, llvm-commits, t-tye
Differential Revision: https://reviews.llvm.org/D32344
llvm-svn: 301195
We handled all of the commuted variants for plain xor already,
although they were scattered around and sometimes folded less
efficiently using distributive laws. We had no folds for not-xor.
Handling all of these patterns consistently is part of trying to
reinstate:
https://reviews.llvm.org/rL300977
llvm-svn: 301144
Summary:
In case all predecessor go to a single successor of current BB. We want to fold (not thread).
I failed to update the phi nodes properly in the last patch https://reviews.llvm.org/rL300657.
Phi nodes values are per predecessor in LLVM.
Reviewers: sanjoy
Reviewed By: sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32400
llvm-svn: 301139
There's probably some better way to write this that eliminates the
code duplication without hurting readability, but at least this
eliminates the logic holes and is hopefully slightly more efficient
than creating new instructions.
llvm-svn: 301129
When the location description of a source variable involves arithmetic
on the value itself, it needs to be marked with DW_OP_stack_value since it
is not describing the variable's location, but rather its value.
This is a follow-up to r297971 and fixes the source testcase quoted in
the comment in debuginfo-dce.ll.
rdar://problem/30725338
llvm-svn: 301093
The bug was introduced by r301018 "[InstCombine] fadd double (sitofp x), y check that the promotion is valid". The patch didn't expect that fadd can be on vectors not necessarily scalars. Add vector support along with the test.
llvm-svn: 301070
Fixes leaving intermediate flat addressing computations
where a GEP instruction's source is a constant expression.
Still leaves behind a trivial addrspacecast + gep pair that
instcombine is able to handle, which ideally could be folded
here directly.
llvm-svn: 301044
Doing these transformations check that the result of integer addition is representable in the FP type.
(fadd double (sitofp x), fpcst) --> (sitofp (add int x, intcst))
(fadd double (sitofp x), (sitofp y)) --> (sitofp (add int x, y))
This is a fix for https://bugs.llvm.org//show_bug.cgi?id=27036
Reviewed By: andrew.w.kaylor, scanon, spatel
Differential Revision: https://reviews.llvm.org/D31182
llvm-svn: 301018
Currently we choose PostBB as the single successor of QFB, but its possible that QTB's single successor is QFB which would make QFB the correct choice.
Differential Revision: https://reviews.llvm.org/D32323
llvm-svn: 300992
places based on it.
Existing constant hoisting pass will merge a group of contants in a small range
and hoist the const materialization code to the common dominator of their uses.
However, if the uses are all in cold pathes, existing implementation may hoist
the materialization code from cold pathes to a hot place. This may hurt performance.
The patch introduces BFI to the pass and selects the best insertion places based
on it.
The change is controlled by an option consthoist-with-block-frequency which is
off by default for now.
Differential Revision: https://reviews.llvm.org/D28962
llvm-svn: 300989
Phi nodes in non-header blocks are converted to select instructions after
if-conversion. This patch updates the cost model to account for the selects.
Differential Revision: https://reviews.llvm.org/D31906
llvm-svn: 300980
CodeExtractor looks up the dominator node corresponding to return blocks
when splitting them. If one of these blocks is unreachable, there's no
node in the Dom and CodeExtractor crashes because it doesn't check
for domtree node validity.
In theory, we could add just a check for skipping null DTNodes in
`splitReturnBlock` but the fix I propose here is slightly different. To the
best of my knowledge, unreachable blocks are irrelevant for the algorithm,
therefore we can just skip them when building the candidate set in the
constructor.
Differential Revision: https://reviews.llvm.org/D32335
llvm-svn: 300946
Before, we assumed that any ConstantInt offset was precisely the access width,
so we could use the "[rN]!" form. ISelLowering only ever created that kind, but
further simplification during combining could lead to unexpected constants and
incorrect codegen.
Should fix PR32658.
llvm-svn: 300878
This change is correct because the verifier requires that at most one
argument be marked 'sret'.
NFC, removes a use of AttributeList slot APIs.
llvm-svn: 300784
The most common case for a branch condition is
a single use compare. Directly invert the branch
predicate rather than adding a lot of xor i1 true
which the DAG will have to fold later.
This produces nicer to read structurizer output.
This produces some random changes in codegen
due to the DAG swapping branch conditions itself,
and then does a poor job of dealing with those
inverts.
llvm-svn: 300732
Summary:
See http://llvm.org/docs/LangRef.html#non-integral-pointer-type
The NewGVN test does not fail without these changes (perhaps it does
try to coerce pointers <-> integers to begin with?), but I added the
test case anyway.
Reviewers: dberlin
Subscribers: mcrosier, llvm-commits, Prazek
Differential Revision: https://reviews.llvm.org/D32208
llvm-svn: 300730
This patch simplifies the examples from D31509 and D31927 (PR30630) and catches
the basic identity shuffle tests that Zvi recently added.
I'm not sure if we have something like this in DAGCombiner, but we should?
It's worth noting that "MaxRecurse / RecursionLimit" is only 3 on entry at the moment.
We might want to bump that up if there are longer shuffle chains like this in the wild.
For now, we're ignoring shuffles that have undef mask elements because it's not
clear how those should be handled.
Differential Revision: https://reviews.llvm.org/D31960
llvm-svn: 300714
InstSimplify returned the wrong type when simplifying a vector GEP
and we ended up crashing when trying to replace all uses with the
new value. Fixes PR32697.
Differential Revision: https://reviews.llvm.org/D32180
llvm-svn: 300693
Summary: In case all predecessor go to a single successor of current BB. We want to fold (not thread).
Reviewers: efriedma, sanjoy
Reviewed By: sanjoy
Subscribers: dberlin, majnemer, llvm-commits
Differential Revision: https://reviews.llvm.org/D30869
llvm-svn: 300657
In tryToVectorizeList, under a very limited circumstance (when entered
from tryToVectorizePair), the values may be reordered (swapped) and the
SLP tree is built with the new order. This extends that to the case when
starting from phis in vectorizeChainsInBlock when there are exactly two
phis. The textual order of phi nodes shouldn't really matter. Without
this change, the loop body in the accompnaying test case is fully vectorized
when we swap the orde of the phis but not with this order. While this
doesn't solve the phi-ordering problem in a general way (for more than 2
phis), this is simple fix that piggybacks on an existing mechanism and
is useful in cases like multiplying two complex numbers.
Differential revision: https://reviews.llvm.org/D32065
llvm-svn: 300574
Summary:
* Add checks for store. That is needed because GlobalsAA is called
twice in the current pipeline with different sets of Function passes
following it. However, the loads are eliminated using instcombine
which happens everywhere. On the other hand, DeadStoreElimination is
performed only once so by checking for store we'll be able to catch
more cases when GlobalsAA is invalidated unintentionally.
* Add empty function above/below the test so that we don't depend on
the relative order of instcombine/dead-store-elimination and the
pass that invalidates the analysis (inside the same
FunctionPassManager).
Reviewers: kristof.beyls
Reviewed By: kristof.beyls
Subscribers: llvm-commits, n.bozhenov
Differential Revision: https://reviews.llvm.org/D32015
Patch by Andrei Elovikov <andrei.elovikov@intel.com>
llvm-svn: 300553
The DWARF specification knows 3 kinds of non-empty simple location
descriptions:
1. Register location descriptions
- describe a variable in a register
- consist of only a DW_OP_reg
2. Memory location descriptions
- describe the address of a variable
3. Implicit location descriptions
- describe the value of a variable
- end with DW_OP_stack_value & friends
The existing DwarfExpression code is pretty much ignorant of these
restrictions. This used to not matter because we only emitted very
short expressions that we happened to get right by accident. This
patch makes DwarfExpression aware of the rules defined by the DWARF
standard and now chooses the right kind of location description for
each expression being emitted.
This would have been an NFC commit (for the existing testsuite) if not
for the way that clang describes captured block variables. Based on
how the previous code in LLVM emitted locations, DW_OP_deref
operations that should have come at the end of the expression are put
at its beginning. Fixing this means changing the semantics of
DIExpression, so this patch bumps the version number of DIExpression
and implements a bitcode upgrade.
There are two major changes in this patch:
I had to fix the semantics of dbg.declare for describing function
arguments. After this patch a dbg.declare always takes the *address*
of a variable as the first argument, even if the argument is not an
alloca.
When lowering a DBG_VALUE, the decision of whether to emit a register
location description or a memory location description depends on the
MachineLocation — register machine locations may get promoted to
memory locations based on their DIExpression. (Future) optimization
passes that want to salvage implicit debug location for variables may
do so by appending a DW_OP_stack_value. For example:
DBG_VALUE, [RBP-8] --> DW_OP_fbreg -8
DBG_VALUE, RAX --> DW_OP_reg0 +0
DBG_VALUE, RAX, DIExpression(DW_OP_deref) --> DW_OP_reg0 +0
All testcases that were modified were regenerated from clang. I also
added source-based testcases for each of these to the debuginfo-tests
repository over the last week to make sure that no synchronized bugs
slip in. The debuginfo-tests compile from source and run the debugger.
https://bugs.llvm.org/show_bug.cgi?id=32382
<rdar://problem/31205000>
Differential Revision: https://reviews.llvm.org/D31439
llvm-svn: 300522
Summary: If there is suffix added in the function name (e.g. module hash added by thinLTO), we will not be able to find a match in profile as the suffix does not exist in profile. This patch build a map from function name to Function *. The map includes the entry for the stripped function name so that inlineHotFunctions can find the corresponding function to promote/inline.
Reviewers: davidxl, dnovillo, tejohnson
Reviewed By: davidxl
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D31952
llvm-svn: 300507
So, `cast<Instruction>` is not guaranteed to succeed. Change the
code so that we create a new constant and use it in the newly
created instruction, as it's done in other places in InstCombine.
OK'ed by Sanjay/Craig. Fixes PR32686.
llvm-svn: 300495
The splitIndirectCriticalEdges function generates and invalid CFG when the
'Target' basic block is a loop to itself. When this occurs, the code that
updates the predecessor terminator needs to update the terminator in the split
basic block.
This occurs when there is an edge from block D back to D. Since D is split in
to D0 and D1, the code needs to update the terminator in D1. But D1 is not in
the OtherPreds vector, so it was not getting updated.
Differential Revision: https://reviews.llvm.org/D32126
llvm-svn: 300480
Causes some VGPR usage improvements in shaderdb, but
introduces some SGPR spilling regressions due to random
scheduling changes later.
llvm-svn: 300453
This patch is a generalization of the improvement introduced in rL296898.
Previously, we were able to peel one iteration of a loop to get rid of a Phi that becomes
an invariant on the 2nd iteration. In more general case, if a Phi becomes invariant after
N iterations, we can peel N times and turn it into invariant.
In order to do this, we for every Phi in loop's header we define the Invariant Depth value
which is calculated as follows:
Given %x = phi <Inputs from above the loop>, ..., [%y, %back.edge].
If %y is a loop invariant, then Depth(%x) = 1.
If %y is a Phi from the loop header, Depth(%x) = Depth(%y) + 1.
Otherwise, Depth(%x) is infinite.
Notice that if we peel a loop, all Phis with Depth = 1 become invariants,
and all other Phis with finite depth decrease the depth by 1.
Thus, peeling N first iterations allows us to turn all Phis with Depth <= N
into invariants.
Reviewers: reames, apilipenko, mkuper, skatkov, anna, sanjoy
Reviewed By: sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31613
llvm-svn: 300446
When peeling loops basing on phis becoming invariants, we make a wrong loop size check.
UP.Threshold should be compared against the total numbers of instructions after the transformation,
which is equal to 2 * LoopSize in case of peeling one iteration.
We should also check that the maximum allowed number of peeled iterations is not zero.
Reviewers: sanjoy, anna, reames, mkuper
Reviewed By: mkuper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31753
llvm-svn: 300441
This patch adds new optimization (Folding cmp(sub(a,b),0) into cmp(a,b))
to instCombineCall pass and was written specific for X86 CMP intrinsics.
Differential Revision: https://reviews.llvm.org/D31398
llvm-svn: 300422
...when C1 differs from C2 by one bit and C1 <u C2:
http://rise4fun.com/Alive/Vuo
And move related folds to a helper function. This reduces code duplication and
will make it easier to remove the scalar-only restriction as a follow-up step.
llvm-svn: 300364
We currently only support folding a subtract into a select but not a PHI. This fixes that.
I had to fix an assumption in FoldOpIntoPhi that assumed the PHI node was always in operand 0. Now we pass it in like we do for FoldOpIntoSelect. But we still require some dancing to find the Constant when we create the BinOp or ConstantExpr. This is based code is similar to what we do for selects.
Since I touched all call sites, this also renames FoldOpIntoPhi to foldOpIntoPhi to match coding standards.
Differential Revision: https://reviews.llvm.org/D31686
llvm-svn: 300363
Switch from Euclid's algorithm to Stein's algorithm for computing GCD. This
avoids the (expensive) APInt division operation in favour of bit operations.
Remove all memory allocation from within the GCD loop by tweaking our `lshr`
implementation so it can operate in-place.
Differential Revision: https://reviews.llvm.org/D31968
llvm-svn: 300252
Summary:
Bug noticed by inspection.
Extend the test to handle invokes as well as calls, and rewrite it to
not depend on the inliner and other passes.
Also simplify the call site replacement code with CallSite, similar to
what I did to dead arg elimination and arg promotion (rL300235 and
rL300229).
Reviewers: danielcdh, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32041
llvm-svn: 300251
Summary: For iterative SamplePGO, an indirect call can be speculatively promoted to multiple direct calls and get inlined. All these promoted direct calls will share the same callsite location (offset+discriminator). With the current implementation, we cannot distinguish between different promotion candidates and its inlined instance. This patch adds callee_name to the key of the callsite sample map. And added helper functions to get all inlined callee samples for a given callsite location. This helps the profile annotator promote correct targets and inline it before annotation, and ensures all indirect call targets to be annotated correctly.
Reviewers: davidxl, dnovillo
Reviewed By: davidxl
Subscribers: andreadb, llvm-commits
Differential Revision: https://reviews.llvm.org/D31950
llvm-svn: 300240
Summary:
In first order recurrences where phi's are used outside the loop,
we should generate an additional vector.extract of the second last element from
the vectorized phi update.
This is because we require the phi itself (which is the value at the second last
iteration of the vector loop) and not the phi's update within the loop.
Also fix the code gen when we just unroll, but don't vectorize.
Fixes PR32396.
Reviewers: mssimpso, mkuper, anemet
Subscribers: llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D31979
llvm-svn: 300238
This is effectively a retry of:
https://reviews.llvm.org/rL299851
but now we have tests and an assert to make sure the bug
that was exposed with that attempt will not happen again.
I'll fix the code duplication and missing sibling fold next,
but I want to make this change as small as possible to reduce
risk since I messed it up last time.
This should fix:
https://bugs.llvm.org/show_bug.cgi?id=32524
llvm-svn: 300236
Noticed by inspection while doing attribute work. DAE, InstCombineCalls,
and ArgPromotion have a fair amount of duplicated code for hacking on
call sites, and you can find bugs by comparing them.
Add a test case for this.
llvm-svn: 300229
Summary:
* Add a bitreverse case in the demanded bits analysis pass.
* Add tests for the bitreverse (and bswap) intrinsic in the
demanded bits pass.
* Add a test case to the BDCE tests: that manipulations to
high-order bits are eliminated once the bits are reversed
and then right-shifted.
Reviewers: mkuper, jmolloy, hfinkel, trentxintong
Reviewed By: jmolloy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31857
llvm-svn: 300215
If we had these tests, the bug caused by https://reviews.llvm.org/rL299851 would have been caught sooner.
There's also an assert in the code that should have caught that bug, but the assert line itself has a bug.
llvm-svn: 300201
As discussed in:
https://bugs.llvm.org/show_bug.cgi?id=32486
...the canonicalization of vector select to shufflevector does not hold up
when undef elements are present in the condition vector.
Try to make the undef handling clear in the code and the LangRef.
Differential Revision: https://reviews.llvm.org/D31980
llvm-svn: 300092
Currently if we reach an instruction with multiples uses we know we can't do any optimizations to that instruction itself since we only have the demanded bits for one of the users. But if we know all of the bits are zero/one for that one user we can still go ahead and create a constant to give to that user.
This might then reduce the instruction to having a single use and allow additional optimizations on the other path.
This picks up an additional case that r300075 didn't catch.
Differential Revision: https://reviews.llvm.org/D31552
llvm-svn: 300084
If we are adding/subtractings 0s below the highest demanded bit we can just use the other operand and remove the operation.
My primary motivation is observing that we can call ShrinkDemandedConstant for the add/sub and create a 0 constant, rather than removing the add completely. In the case I saw, we modified the constant on an add instruction to a 0, but the add is not put into the worklist. So we didn't revisit it until the next InstCombine iteration. This caused an IR modification to remove add and a subsequent iteration to be ran.
With this change we get bypass the add in the first iteration and prevent the second iteration from changing anything.
Differential Revision: https://reviews.llvm.org/D31120
llvm-svn: 300075
One potential way to make InstCombine (very slightly?) faster is to recycle instructions
when possible instead of creating new ones. It's not explicitly stated AFAIK, but we don't
consider this an "InstSimplify". We could, however, make a new layer to house transforms
like this if that makes InstCombine more manageable (just throwing out an idea; not sure
how much opportunity is actually here).
Differential Revision: https://reviews.llvm.org/D31863
llvm-svn: 300067
Use '2>&1 |' and not '|&' to pipe debug output to FileCheck
Hopefully handles a "shell parser error" on
llvm-clang-x86_64-expensive-checks-win
test/Transforms/SLPVectorizer/SystemZ/SLP-cmp-cost-query.ll
llvm-svn: 300064
In getEntryCost(), make the scalar type for a compare instruction that of the
operands, not i1. This is needed in order to call getCmpSelInstrCost() for a
compare in a sensible way, the same way as the LoopVectorizer does.
New test: test/Transforms/SLPVectorizer/SystemZ/SLP-cmp-cost-query.ll
Review: Matthew Simpson
https://reviews.llvm.org/D31601
llvm-svn: 300061
The cost for a branch after vectorization is very different depending on if
the vectorizer will if-convert the block (branch is eliminated), or if
scalarized and predicated blocks will be produced (branch duplicated before
each block). There is also the case of remaining scalar branches, such as the
back-edge branch.
This patch handles these cases differently with TTI based cost estimates.
Review: Matthew Simpson
https://reviews.llvm.org/D31175
llvm-svn: 300058
Since SystemZ supports vector element load/store instructions, there is no
need for extracts/inserts if a vector load/store gets scalarized.
This patch lets Target specify that it supports such instructions by means of
a new TTI hook that defaults to false.
The use for this is in the LoopVectorizer getScalarizationOverhead() method,
which will with this patch produce a smaller sum for a vector load/store on
SystemZ.
New test: test/Transforms/LoopVectorize/SystemZ/load-store-scalarization-cost.ll
Review: Adam Nemet
https://reviews.llvm.org/D30680
llvm-svn: 300056
getArithmeticInstrCost(), getShuffleCost(), getCastInstrCost(),
getCmpSelInstrCost(), getVectorInstrCost(), getMemoryOpCost(),
getInterleavedMemoryOpCost() implemented.
Interleaved access vectorization enabled.
BasicTTIImpl::getCastInstrCost() improved to check for legal extending loads,
in which case the cost of the z/sext instruction becomes 0.
Review: Ulrich Weigand, Renato Golin.
https://reviews.llvm.org/D29631
llvm-svn: 300052
Summary:
Dead basic blocks may be forming a loop, for which SSA form is
fulfilled, but with a circular def-use chain. LoadCombine could
enter an infinite loop when analysing such dead code. This patch
solves the problem by simply avoiding to analyse all basic blocks
that aren't forward reachable, from function entry, in LoadCombine.
Fixes https://bugs.llvm.org/show_bug.cgi?id=27065
Reviewers: mehdi_amini, chandlerc, grosser, Bigcheese, davide
Reviewed By: davide
Subscribers: dberlin, zzheng, bjope, grandinj, Ka-Ka, materi, jholewinski, llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D31032
llvm-svn: 300034
Summary:
COFF requires that every comdat contain a symbol with the same name as
the comdat. ThinLTOBitcodeWriter renames symbols, which may cause this
requirement to be violated. This change avoids such violations by
renaming comdats if their leaders are renamed. It also keeps comdats
together when splitting modules.
Reviewers: pcc, mehdi_amini, tejohnson
Reviewed By: pcc
Subscribers: rnk, Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D31963
llvm-svn: 300019
In the vectorization of first order recurrence, we vectorize such
that the last element in the vector will be the one extracted to pass into the
scalar remainder loop. However, this is not true when there is a phi (other
than the primary induction variable) is used outside the loop.
In such a case, we need the value from the second last iteration (i.e.
the phi value), not the last iteration (which would be the phi update).
I've added a test case for this. Also see PR32396.
A follow up patch would generate the correct code gen for such cases,
and turn this vectorization on.
Differential Revision: https://reviews.llvm.org/D31910
Reviewers: mssimpso
llvm-svn: 299985
Analysis, it has Analysis passes, and once NewGVN is made an Analysis,
this removes the cross dependency from Analysis to Transform/Utils.
NFC.
llvm-svn: 299980
Before this patch, pass AddDiscriminators always avoided to assign
discriminators to intrinsic calls. This was done mainly for two reasons:
1) We wanted to minimize the number of based discriminators used.
2) We wanted to avoid non-deterministic discriminator assignment for
different debug levels.
Unfortunately, that approach was problematic for MemIntrinsic calls.
MemIntrinsic calls can be split by SROA into loads and stores, and each new
load/store instruction would obtain the debug location from the original
intrinsic call.
If we don't assign a discriminator to MemIntrinsic calls, then we cannot
correctly set the discriminator for the newly created loads and stores.
This may have a negative impact on the basic block weight computation
performed by the SampleLoader.
This patch fixes the issue by letting MemIntrinsic calls have a discriminator.
Differential Revision: https://reviews.llvm.org/D31900
llvm-svn: 299972
Summary:
In rL299692 I improved strip-dead-debug-info's ability to drop CUs that are not
referenced from the current module. However, in doing so I neglected to realize
that some SPs could be referenced entirely from inlined functions. It appears
I was not the only one to make this mistake, because DebugInfoFinder, doesn't
find those SPs either. Fix this in DebugInfoFinder and then use that to make
sure not to drop those CUs in strip-dead-debug-info.
Reviewers: aprantl
Reviewed By: aprantl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31904
llvm-svn: 299936
(h/t to Chandler for pointing this out)
The test in question was not at all testing what it was supposed to
test. We do not //care// about placing `!make.implicit` in inner
constant branch (since it will be folded away anyway). We care about
placing `!make.implicit` in the outer branch that switches between
either version of the loop.
Having said that, it is _correct_ to leave behind the `!make.implicit`
in the inner branch, but there is no need to do so.
llvm-svn: 299912
When allowed, we can hoist a division out of a loop in favor of a
multiplication by the reciprocal. Fixes PR32157.
Patch by vit9696!
Differential Revision: https://reviews.llvm.org/D30819
llvm-svn: 299911
LLVM makes several assumptions about address space 0. However,
alloca is presently constrained to always return this address space.
There's no real way to avoid using alloca, so without this
there is no way to opt out of these assumptions.
The problematic assumptions include:
- That the pointer size used for the stack is the same size as
the code size pointer, which is also the maximum sized pointer.
- That 0 is an invalid, non-dereferencable pointer value.
These are problems for AMDGPU because alloca is used to
implement the private address space, which uses a 32-bit
index as the pointer value. Other pointers are 64-bit
and behave more like LLVM's notion of generic address
space. By changing the address space used for allocas,
we can change our generic pointer type to be LLVM's generic
pointer type which does have similar properties.
llvm-svn: 299888
Summary: Now the SamplePGO support is more stable, we do not need so many verbose optimization remarks emitted.
Reviewers: dnovillo, davidxl
Reviewed By: davidxl
Subscribers: fhahn, llvm-commits
Differential Revision: https://reviews.llvm.org/D31826
llvm-svn: 299883
Summary:
While we don't want them aliasing with other pointers, there seems to
be no point in not having them clobber must-aliased'd pointers.
If some day, we split the aliasing and ordering chains, we'd make this
not aliasing but an ordering barrier (IE it doesn't affect it's
memory, but we can't hoist it above it).
Reviewers: hfinkel, george.burgess.iv
Subscribers: Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D31865
llvm-svn: 299865
This patch refactors and strengthens the type checks performed for interleaved
accesses. The primary functional change is to ensure that the interleaved
accesses have valid element types. The added test cases previously failed
because the element type is f128.
Differential Revision: https://reviews.llvm.org/D31817
llvm-svn: 299864
Also, make the same change in and-of-icmps and remove a hack for detecting that case.
Finally, add some FIXME comments because the code duplication here is awful.
This should fix the remaining IR problem noted in:
https://bugs.llvm.org/show_bug.cgi?id=32524
llvm-svn: 299851
We currently only fold scalar add of constants into selects. This improves this to support vectors too.
Differential Revision: https://reviews.llvm.org/D31683
llvm-svn: 299847
Summary: I noticed in the select folding code that we copied fast math flags, but did not do the same for the similar handling in phi nodes. This patch fixes that to do the same thing as select
Reviewers: spatel, davide, majnemer, hfinkel
Reviewed By: davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31690
llvm-svn: 299838
Summary:
Resolve indirect branch target when possible.
This potentially eliminates more basicblocks and result in better evaluation for phi and other things.
Reviewers: davide, efriedma, sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D30322
llvm-svn: 299830
In isUseTriviallyOptimizableToLiveOnEntry, pointsToConstantMemory needs to be
called on the load's pointer operand, not on the result of the load (which
might not even be a pointer).
llvm-svn: 299823
coro-split-after-phi.ll test was flaky due to non-determinism in
the coroutine frame construction that was sorting the spill
vector using a pointer to a def as a part of the key.
The sorting was intended to make sure that spills for the same def
are kept together, however, we populate the vector by processing
defs in order, so the spill entires will end up together anyways.
This change removes spill sorting and restores the determinism
in the test.
llvm-svn: 299809
Summary:
Fix a bug where we were inserting a spill in between the PHIs in the beginning of the block.
Consider this fragment:
```
begin:
%phi1 = phi i32 [ 0, %entry ], [ 2, %alt ]
%phi2 = phi i32 [ 1, %entry ], [ 3, %alt ]
%sp1 = call i8 @llvm.coro.suspend(token none, i1 false)
switch i8 %sp1, label %suspend [i8 0, label %resume
i8 1, label %cleanup]
resume:
call i32 @print(i32 %phi1)
```
Unless we are spilling the argument or result of the invoke, we were always inserting the spill immediately following the instruction.
The fix adds a check that if the spilled instruction is a PHI Node, select an appropriate insert point with `getFirstInsertionPt()` that
skips all the PHI Nodes and EH pads.
Reviewers: majnemer, rnk
Reviewed By: rnk
Subscribers: qcolombet, EricWF, llvm-commits
Differential Revision: https://reviews.llvm.org/D31799
llvm-svn: 299771
This patch reapplies r298620. The original patch was reverted because of two
issues. First, the patch exposed a bug in InstCombine that caused the Chromium
builds to fail (PR32414). This issue was fixed in r299017. Second, the patch
introduced a bug in the vectorizer's scalars analysis that caused test suite
builds to fail on SystemZ. The scalars analysis was too aggressive and marked a
memory instruction scalar, even though it was going to be vectorized. This
issue has been fixed in the current patch and several new test cases for the
scalars analysis have been added.
llvm-svn: 299770
Summary:
getModRefInfo is meant to answer the question "what impact does this
instruction have on a given memory location" (not even another
instruction).
Long debate on this on IRC comes to the conclusion the answer should be "nothing special".
That is, a noalias volatile store does not affect a memory location
just by being volatile. Note: DSE and GVN and memdep currently
believe this, because memdep just goes behind AA's back after it says
"modref" right now.
see line 635 of memdep. Prior to this patch we would get modref there, then check aliasing,
and if it said noalias, we would continue.
getModRefInfo *already* has this same AA check, it just wasn't being used because volatile was
lumped in with ordering.
(I am separately testing whether this code in memdep is now dead except for the invariant load case)
Reviewers: jyknight, chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31726
llvm-svn: 299741
The new codepath has been in the tree for years, and there isn't any
reason to use two codepaths here.
Differential Revision: https://reviews.llvm.org/D30596
llvm-svn: 299723
Summary:
Prior to this while it would delete the dead DIGlobalVariables, it would
leave dead DICompileUnits and everything referenced therefrom. For a bit
bitcode file with thousands of compile units those dead nodes easily
outnumbered the real ones. Clean that up.
Reviewed By: aprantl
Differential Revision: https://reviews.llvm.org/D31720
llvm-svn: 299692
memorydefs, not just stores. Along the way, we audit and fixup issues
about how we were tracking memory leaders, and improve the verifier
to notice more memory congruency issues.
llvm-svn: 299682
Summary:
LSV wants to know the maximum size that can be loaded to a vector register.
On X86, this always matches the maximum register width. Implement this
accordingly and add a test to make sure that LSV can vectorize up to the
maximum permissible width on X86.
Reviewers: delena, arsenm
Reviewed By: arsenm
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D31504
llvm-svn: 299589
This test case depends on the loop being vectorized without forcing the
vectorization factor. If the profitability ever changes in the future (due to
cost model improvements), the test may no longer work as intended. Instead of
checking the resulting IR, we should just check the instruction costs. The
costs will be computed regardless if vectorization is profitable.
llvm-svn: 299545
This is a latent bug that's been hanging around for a while. For a loop-invariant
pointer, expandBounds would return the range {Ptr, Ptr}, but this was interpreted
as a half-open range, not a closed range. So we ended up planting incorrect
bounds checks. Even worse, they were tautological, so we ended up incorrectly
executing the optimized loop.
llvm-svn: 299526
Fix a bug in ARC contract pass where an iterator that pointed to a
deleted instruction was dereferenced.
It appears that tryToContractReleaseIntoStoreStrong was incorrectly
assuming that a call to objc_retain would not immediately follow a call
to objc_release.
rdar://problem/25276306
llvm-svn: 299507
Currently we only fold with ConstantInt RHS. This generalizes to any Constant RHS.
Differential Revision: https://reviews.llvm.org/D31610
llvm-svn: 299466
This patch optimizes two memory intrinsic operations: memset and memcpy based
on the profiled size of the operation. The high level transformation is like:
mem_op(..., size)
==>
switch (size) {
case s1:
mem_op(..., s1);
goto merge_bb;
case s2:
mem_op(..., s2);
goto merge_bb;
...
default:
mem_op(..., size);
goto merge_bb;
}
merge_bb:
Differential Revision: http://reviews.llvm.org/D28966
llvm-svn: 299446
Otherwise, yamlize in YAMLTraits.h might be wrongly defined.
This makes some AMDGPU tests fail when LLVM_LINK_LLVM_DYLIB is set.
Differential Revision: https://reviews.llvm.org/D30508
llvm-svn: 299415
Summary:
Add a hook for simplification of shufflevector's with the following rules:
- Constant folding - NFC, as it was already being done by the default handler.
- If only one of the operands is constant, constant fold the shuffle if the
mask does not select elements from the variable operand - to show the hook is firing and affecting the test-cases.
Reviewers: RKSimon, craig.topper, spatel, sanjoy, nlopes, majnemer
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31525
llvm-svn: 299393
Summary:
Depends on D30928.
This adds support for coercion of stores and memory instructions that do not require insertion to process.
Another few tests down.
I added the relevant tests from rle.ll
Reviewers: davide
Subscribers: llvm-commits, Prazek
Differential Revision: https://reviews.llvm.org/D30929
llvm-svn: 299330
Disable bypassing if one of the operands looks like a hash value. Slow
division often occurs in hashtable implementations and fast division is
never taken there because a hash value is extremely unlikely to have
enough upper bits set to zero.
A value is considered to be hash-like if it is produced by
1) XOR operation
2) Multiplication by a constant wider than the shorter type
3) PHI node with all incoming values being hash-like
Differential Revision: https://reviews.llvm.org/D28200
llvm-svn: 299329
A common way to implement nearbyint is by fiddling with the floating
point environment and calling rint. This is used at least by the BSD
libm and musl. As such, canonicalizing the latter to the former will
create infinite loops for libm and generally pessimize performance, at
least when the generic C versions are used.
This change preserves the rint in the libcall translation and also
handles the domain truncation logic, so that rint with float argument
will be reduced to rintf etc.
llvm-svn: 299247
Summary: Currently the VP metadata was dropped when InstCombine converts a call to direct call. This patch converts the VP metadata to branch_weights so that its hotness is recorded.
Reviewers: eraman, davidxl
Reviewed By: davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31344
llvm-svn: 299228
Summary:
Triggered by commit r298620: "[LV] Vectorize GEPs".
If we encounter a vector GEP with scalar arguments, we splat the scalar
into a vector of appropriate size before we scatter the argument.
Reviewers: arsenm, mehdi_amini, bkramer
Reviewed By: arsenm
Subscribers: bjope, mssimpso, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D31416
llvm-svn: 299186
Since there is no sdiv in SCEV, an 'udiv' is a better canonical form than an 'sdiv' as the user of induction variable
Differential Revision: https://reviews.llvm.org/D31488
llvm-svn: 299118
Some of the GEP combines (e.g., descaling) can't handle vector GEPs. We have an
existing check that attempts to bail out if given a vector GEP. However, the
check only tests the GEP's pointer operand. A GEP results in a vector of
pointers if at least one of its operands is vector-typed (e.g., its pointer
operand could be a scalar, but its index could be a vector). We should just
check the type of the GEP itself. This should fix PR32414.
Reference: https://bugs.llvm.org/show_bug.cgi?id=32414
Differential Revision: https://reviews.llvm.org/D31470
llvm-svn: 299017
The vectorizer tries to replace truncations of induction variables with new
induction variables having the smaller type. After r295063, this optimization
was applied to all integer induction variables, including non-primary ones.
When optimizing the truncation of a non-primary induction variable, we still
need to transform the new induction so that it has the correct start value.
This should fix PR32419.
Reference: https://bugs.llvm.org/show_bug.cgi?id=32419
llvm-svn: 298882
Summary:
We are incorrectly folding selects into phi nodes when the incoming value of a phi
node is a constant vector. This optimization is done in `FoldOpIntoPhi` when the
select condition is a phi node with constant incoming values.
Without the fix, we are miscompiling (i.e. incorrectly folding the
select into the phi node) when the vector contains non-zero
elements.
This patch fixes the miscompile and we will correctly fold based on the
select vector operand (see added test cases).
Reviewers: majnemer, sanjoy, spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31189
llvm-svn: 298845
The first variant contains all current transformations except
transforming switches into lookup tables. The second variant
contains all current transformations.
The switch-to-lookup-table conversion results in code that is more
difficult to analyze and optimize by other passes. Most importantly,
it can inhibit Dead Code Elimination. As such it is often beneficial to
only apply this transformation very late. A common example is inlining,
which can often result in range restrictions for the switch expression.
Changes in execution time according to LNT:
SingleSource/Benchmarks/Misc/fp-convert +3.03%
MultiSource/Benchmarks/ASC_Sequoia/CrystalMk/CrystalMk -11.20%
MultiSource/Benchmarks/Olden/perimeter/perimeter -10.43%
and a couple of smaller changes. For perimeter it also results 2.6%
a smaller binary.
Differential Revision: https://reviews.llvm.org/D30333
llvm-svn: 298799
This moves it to the iterator facade utilities giving it full random
access semantics, etc. It can also now be used with standard algorithms
like std::all_of and std::any_of and range adaptors like llvm::reverse.
Also make the semantics of iterating match what every other iterator
uses and forbid decrementing past the begin iterator. This was used as
a hacky way to work around iterator invalidation. However, every
instance trying to do this failed to actually avoid touching invalid
iterators despite the clear documentation that the removed and all
subsequent iterators become invalid including the end iterator. So I've
added a return of the next iterator to removeCase and rewritten the
loops that were doing this to correctly follow the iterator pattern of
either incremneting or removing and assigning fresh values to the
iterator and the end.
In one case we were trying to go backwards to make this cleaner but it
doesn't actually work. I've made that code match the code we use
everywhere else to remove cases as we iterate. This changes the order of
cases in one test output and I moved that test to CHECK-DAG so it
wouldn't care -- the order isn't semantically meaningful anyways.
llvm-svn: 298791
Reason: breaks linking Chromium with LLD + ThinLTO (a pass crashes)
LLVM bug: https://bugs.llvm.org//show_bug.cgi?id=32413
Original change description:
[LV] Vectorize GEPs
This patch adds support for vectorizing GEPs. Previously, we only generated
vector GEPs on-demand when creating gather or scatter operations. All GEPs from
the original loop were scalarized by default, and if a pointer was to be stored
to memory, we would have to build up the pointer vector with insertelement
instructions.
With this patch, we will vectorize all GEPs that haven't already been marked
for scalarization.
The patch refines collectLoopScalars to more exactly identify the scalar GEPs.
The function now more closely resembles collectLoopUniforms. And the patch
moves vector GEP creation out of vectorizeMemoryInstruction and into the main
vectorization loop. The vector GEPs needed for gather and scatter operations
will have already been generated before vectoring the memory accesses.
Original Differential Revision: https://reviews.llvm.org/D30710
llvm-svn: 298735
Summary: Declarations need to be filtered out when counting functions.
Reviewers: eraman
Subscribers: Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D31336
llvm-svn: 298720
Summary: In DeadArgumentElimination, the call instructions will be replaced. We also need to set the prof weights so that function inlining can find the correct profile.
Reviewers: eraman
Reviewed By: eraman
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31143
llvm-svn: 298660
Library functions can have specific semantics that affect the behavior of
certain passes. DSE, for instance, gives special treatment to malloc-ed pointers
but not to pointers returned from an equivalently typed (but differently named)
function.
MetaRenamer ought not to alter program semantics, so library functions must
remain untouched.
Reviewers: mehdi_amini, majnemer, chandlerc, davide
Reviewed By: davide
Subscribers: davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D31304
llvm-svn: 298659
Summary: The current prefix based function layout algorithm only looks at function's entry count, which is not sufficient. A function should be grouped together if its entry count or any call edge count is hot.
Reviewers: davidxl, eraman
Reviewed By: eraman
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31225
llvm-svn: 298656
The new test asserts that scalarized memory operations get memcheck metadata
added even if the loop is only unrolled.
Differential Revision: https://reviews.llvm.org/D30972
llvm-svn: 298641
Summary:
The cumulative size of the bitcode files for a very large application
can be huge, particularly with -g. In a distributed build environment,
all of these files must be sent to the remote build node that performs
the thin link step, and this can exceed size limits.
The thin link actually only needs the summary along with a bitcode
symbol table. Until we have a proper bitcode symbol table, simply
stripping the debug metadata results in significant size reduction.
Add support for an option to additionally emit minimized bitcode
modules, just for use in the thin link step, which for now just strips
all debug metadata. I plan to add a cc1 option so this can be invoked
easily during the compile step.
However, care must be taken to ensure that these minimized thin link
bitcode files produce the same index as with the original bitcode files,
as these original bitcode files will be used in the backends.
Specifically:
1) The module hash used for caching is typically produced by hashing the
written bitcode, and we want to include the hash that would correspond
to the original bitcode file. This is because we want to ensure that
changes in the stripped portions affect caching. Added plumbing to emit
the same module hash in the minimized thin link bitcode file.
2) The module paths in the index are constructed from the module ID of
each thin linked bitcode, and typically is automatically generated from
the input file path. This is the path used for finding the modules to
import from, and obviously we need this to point to the original bitcode
files. Added gold-plugin support to take a suffix replacement during the
thin link that is used to override the identifier on the MemoryBufferRef
constructed from the loaded thin link bitcode file. The assumption is
that the build system can specify that the minimized bitcode file has a
name that is similar but uses a different suffix (e.g. out.thinlink.bc
instead of out.o).
Added various tests to ensure that we get identical index files out of
the thin link step.
Reviewers: mehdi_amini, pcc
Subscribers: Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D31027
llvm-svn: 298638
This patch adds support for vectorizing GEPs. Previously, we only generated
vector GEPs on-demand when creating gather or scatter operations. All GEPs from
the original loop were scalarized by default, and if a pointer was to be stored
to memory, we would have to build up the pointer vector with insertelement
instructions.
With this patch, we will vectorize all GEPs that haven't already been marked
for scalarization.
The patch refines collectLoopScalars to more exactly identify the scalar GEPs.
The function now more closely resembles collectLoopUniforms. And the patch
moves vector GEP creation out of vectorizeMemoryInstruction and into the main
vectorization loop. The vector GEPs needed for gather and scatter operations
will have already been generated before vectoring the memory accesses.
Differential Revision: https://reviews.llvm.org/D30710
llvm-svn: 298620
The code for generating scalar base pointers in vectorizeMemoryInstruction is
not needed. We currently scalarize all GEPs and maintain the scalarized values
in VectorLoopValueMap. The GEP cloning in this unneeded code is the same as
that in scalarizeInstruction. The test cases that changed as a result of this
patch changed because we were able to reuse the scalarized GEP that we
previously generated instead of cloning a new one.
Differential Revision: https://reviews.llvm.org/D30587
llvm-svn: 298615
Summary: ThinLTO will annotate the CFG twice. If the branch weight is set by the first annotation, we should not set the branch weight again in the second annotation because the first annotation is more accurate as there is less optimization that could affect debug info accuracy.
Reviewers: tejohnson, davidxl
Reviewed By: tejohnson
Subscribers: mehdi_amini, aprantl, llvm-commits
Differential Revision: https://reviews.llvm.org/D31228
llvm-svn: 298602
insertelement (insertelement X, Y, IdxC1), ScalarC, IdxC2 -->
insertelement (insertelement X, ScalarC, IdxC2), Y, IdxC1
As noted in the code comment and seen in the test changes, the motivation is that by pulling
constant insertion up, we may be able to constant fold some insertelement instructions.
Differential Revision: https://reviews.llvm.org/D31196
llvm-svn: 298520
Summary: Subtracts can have constants on the left side, but we don't shrink them based on demanded bits. This patch fixes that to match the right hand side.
Reviewers: davide, majnemer, spatel, sanjoy, hfinkel
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31119
llvm-svn: 298478
Currently the default C calling convention functions are treated
the same as compute kernels. Make this explicit so the default
calling convention can be changed to a non-kernel.
Converted with perl -pi -e 's/define void/define amdgpu_kernel void/'
on the relevant test directories (and undoing in one place that actually
wanted a non-kernel).
llvm-svn: 298444
This adds a parameter to @llvm.objectsize that makes it return
conservative values if it's given null.
This fixes PR23277.
Differential Revision: https://reviews.llvm.org/D28494
llvm-svn: 298430
Summary: Inliner should update the branch_weights annotation to scale it to proper value.
Reviewers: davidxl, eraman
Reviewed By: eraman
Subscribers: zzheng, llvm-commits
Differential Revision: https://reviews.llvm.org/D30767
llvm-svn: 298270
Summary:
In case we are loading on a phi-load in SimplifyPartiallyRedundantLoad.
Try to phi translate it into incoming values in the predecessors before
we search for available loads.
This needs https://reviews.llvm.org/D30524
Reviewers: davide, sanjoy, efriedma, dberlin, rengolin
Reviewed By: dberlin
Subscribers: junbuml, llvm-commits
Differential Revision: https://reviews.llvm.org/D30543
llvm-svn: 298217
Summary:
The reverse of an artbitrary bitpattern is also an arbitrary
bitpattern.
Reviewers: trentxintong, arsenm, majnemer
Reviewed By: majnemer
Subscribers: majnemer, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D31118
llvm-svn: 298201
Loop unswitching can be extremely harmful for a SIMT target. In case
if hoisted condition is not uniform a SIMT machine will execute both
clones of a loop sequentially. Therefor LoopUnswitch checks if the
condition is non-divergent.
Since DivergenceAnalysis adds an expensive PostDominatorTree analysis
not needed for non-SIMT targets a new option is added to avoid unneded
analysis initialization. The method getAnalysisUsage is called when
TargetTransformInfo is not yet available and we cannot use it here.
For that reason a new field DivergentTarget is added to PassManagerBuilder
to control the behavior and set this field from a target.
Differential Revision: https://reviews.llvm.org/D30796
llvm-svn: 298104
We were not handling getelemenptr instructions of vector type before.
Since getelemenptr instructions for vector types follow the same rule as
getelementptr instructions for non-vector types, we can just handle them
in the same way.
llvm-svn: 298028
[Reapplies r297971 and punting on finding a better API for findDbgValues()]
This patch improves debug info quality in InstCombine by looking at
values that are about to be deleted, checking whether there are any
dbg.value instrinsics referring to them, and potentially encoding the
semantics of the deleted instruction into the dbg.value's
DIExpression.
In the example in the testcase (which was extracted from XNU) there is a sequence of
%4 = load %struct.entry*, %struct.entry** %next2, align 8, !dbg !41
%5 = bitcast %struct.entry* %4 to i8*, !dbg !42
%add.ptr4 = getelementptr inbounds i8, i8* %5, i64 -8, !dbg !43
%6 = bitcast i8* %add.ptr4 to %struct.entry*, !dbg !44
call void @llvm.dbg.value(metadata %struct.entry* %6, i64 0, metadata !20, metadata !21), !dbg 34
When these instructions are eliminated by instcombine one after
another, we can still salvage the otherwise dead debug info:
- Bitcasts have no effect, so have the dbg.value point to operand(0)
- Loads can be expressed via a DW_OP_deref
- Constant gep instructions can be replaced by DWARF expression arithmetic
The API introduced by this patch is not specific to instcombine and
can be useful in other places, too.
rdar://problem/30725338
Differential Revision: https://reviews.llvm.org/D30919
llvm-svn: 297994
Matthew Simpson