as well as sext(C + x + ...) -> (D + sext(C-D + x + ...))<nuw><nsw>
similar to the equivalent transformation for zext's
if the top level addition in (D + (C-D + x * n)) could be proven to
not wrap, where the choice of D also maximizes the number of trailing
zeroes of (C-D + x * n), ensuring homogeneous behaviour of the
transformation and better canonicalization of such AddRec's
(indeed, there are 2^(2w) different expressions in `B1 + ext(B2 + Y)` form for
the same Y, but only 2^(2w - k) different expressions in the resulting `B3 +
ext((B4 * 2^k) + Y)` form, where w is the bit width of the integral type)
This patch generalizes sext(C1 + C2*X) --> sext(C1) + sext(C2*X) and
sext{C1,+,C2} --> sext(C1) + sext{0,+,C2} transformations added in
r209568 relaxing the requirements the following way:
1. C2 doesn't have to be a power of 2, it's enough if it's divisible by 2
a sufficient number of times;
2. C1 doesn't have to be less than C2, instead of extracting the entire
C1 we can split it into 2 terms: (00...0XXX + YY...Y000), keep the
second one that may cause wrapping within the extension operator, and
move the first one that doesn't affect wrapping out of the extension
operator, enabling further simplifications;
3. C1 and C2 don't have to be positive, splitting C1 like shown above
produces a sum that is guaranteed to not wrap, signed or unsigned;
4. in AddExpr case there could be more than 2 terms, and in case of
AddExpr the 2nd and following terms and in case of AddRecExpr the
Step component don't have to be in the C2*X form or constant
(respectively), they just need to have enough trailing zeros,
which in turn could be guaranteed by means other than arithmetics,
e.g. by a pointer alignment;
5. the extension operator doesn't have to be a sext, the same
transformation works and profitable for zext's as well.
Apparently, optimizations like SLPVectorizer currently fail to
vectorize even rather trivial cases like the following:
double bar(double *a, unsigned n) {
double x = 0.0;
double y = 0.0;
for (unsigned i = 0; i < n; i += 2) {
x += a[i];
y += a[i + 1];
}
return x * y;
}
If compiled with `clang -std=c11 -Wpedantic -Wall -O3 main.c -S -o - -emit-llvm`
(!{!"clang version 7.0.0 (trunk 337339) (llvm/trunk 337344)"})
it produces scalar code with the loop not unrolled with the unsigned `n` and
`i` (like shown above), but vectorized and unrolled loop with signed `n` and
`i`. With the changes made in this commit the unsigned version will be
vectorized (though not unrolled for unclear reasons).
How it all works:
Let say we have an AddExpr that looks like (C + x + y + ...), where C
is a constant and x, y, ... are arbitrary SCEVs. Let's compute the
minimum number of trailing zeroes guaranteed of that sum w/o the
constant term: (x + y + ...). If, for example, those terms look like
follows:
i
XXXX...X000
YYYY...YY00
...
ZZZZ...0000
then the rightmost non-guaranteed-zero bit (a potential one at i-th
position above) can change the bits of the sum to the left (and at
i-th position itself), but it can not possibly change the bits to the
right. So we can compute the number of trailing zeroes by taking a
minimum between the numbers of trailing zeroes of the terms.
Now let's say that our original sum with the constant is effectively
just C + X, where X = x + y + .... Let's also say that we've got 2
guaranteed trailing zeros for X:
j
CCCC...CCCC
XXXX...XX00 // this is X = (x + y + ...)
Any bit of C to the left of j may in the end cause the C + X sum to
wrap, but the rightmost 2 bits of C (at positions j and j - 1) do not
affect wrapping in any way. If the upper bits cause a wrap, it will be
a wrap regardless of the values of the 2 least significant bits of C.
If the upper bits do not cause a wrap, it won't be a wrap regardless
of the values of the 2 bits on the right (again).
So let's split C to 2 constants like follows:
0000...00CC = D
CCCC...CC00 = (C - D)
and represent the whole sum as D + (C - D + X). The second term of
this new sum looks like this:
CCCC...CC00
XXXX...XX00
----------- // let's add them up
YYYY...YY00
The sum above (let's call it Y)) may or may not wrap, we don't know,
so we need to keep it under a sext/zext. Adding D to that sum though
will never wrap, signed or unsigned, if performed on the original bit
width or the extended one, because all that that final add does is
setting the 2 least significant bits of Y to the bits of D:
YYYY...YY00 = Y
0000...00CC = D
----------- <nuw><nsw>
YYYY...YYCC
Which means we can safely move that D out of the sext or zext and
claim that the top-level sum neither sign wraps nor unsigned wraps.
Let's run an example, let's say we're working in i8's and the original
expression (zext's or sext's operand) is 21 + 12x + 8y. So it goes
like this:
0001 0101 // 21
XXXX XX00 // 12x
YYYY Y000 // 8y
0001 0101 // 21
ZZZZ ZZ00 // 12x + 8y
0000 0001 // D
0001 0100 // 21 - D = 20
ZZZZ ZZ00 // 12x + 8y
0000 0001 // D
WWWW WW00 // 21 - D + 12x + 8y = 20 + 12x + 8y
therefore zext(21 + 12x + 8y) = (1 + zext(20 + 12x + 8y))<nuw><nsw>
This approach could be improved if we move away from using trailing
zeroes and use KnownBits instead. For instance, with KnownBits we could
have the following picture:
i
10 1110...0011 // this is C
XX X1XX...XX00 // this is X = (x + y + ...)
Notice that some of the bits of X are known ones, also notice that
known bits of X are interspersed with unknown bits and not grouped on
the rigth or left.
We can see at the position i that C(i) and X(i) are both known ones,
therefore the (i + 1)th carry bit is guaranteed to be 1 regardless of
the bits of C to the right of i. For instance, the C(i - 1) bit only
affects the bits of the sum at positions i - 1 and i, and does not
influence if the sum is going to wrap or not. Therefore we could split
the constant C the following way:
i
00 0010...0011 = D
10 1100...0000 = (C - D)
Let's compute the KnownBits of (C - D) + X:
XX1 1 = carry bit, blanks stand for known zeroes
10 1100...0000 = (C - D)
XX X1XX...XX00 = X
--- -----------
XX X0XX...XX00
Will this add wrap or not essentially depends on bits of X. Adding D
to this sum, however, is guaranteed to not to wrap:
0 X
00 0010...0011 = D
sX X0XX...XX00 = (C - D) + X
--- -----------
sX XXXX XX11
As could be seen above, adding D preserves the sign bit of (C - D) +
X, if any, and has a guaranteed 0 carry out, as expected.
The more bits of (C - D) we constrain, the better the transformations
introduced here canonicalize expressions as it leaves less freedom to
what values the constant part of ((C - D) + x + y + ...) can take.
Reviewed By: mzolotukhin, efriedma
Differential Revision: https://reviews.llvm.org/D48853
llvm-svn: 337943
if the top level addition in (D + (C-D + x + ...)) could be proven to
not wrap, where the choice of D also maximizes the number of trailing
zeroes of (C-D + x + ...), ensuring homogeneous behaviour of the
transformation and better canonicalization of such expressions.
This enables better canonicalization of expressions like
1 + zext(5 + 20 * %x + 24 * %y) and
zext(6 + 20 * %x + 24 * %y)
which get both transformed to
2 + zext(4 + 20 * %x + 24 * %y)
This pattern is common in address arithmetics and the transformation
makes it easier for passes like LoadStoreVectorizer to prove that 2 or
more memory accesses are consecutive and optimize (vectorize) them.
Reviewed By: mzolotukhin
Differential Revision: https://reviews.llvm.org/D48853
llvm-svn: 337859
SCEV tries to constant-fold arguments of trunc operands in SCEVAddExpr, and when it does
that, it passes wrong flags into the recursion. It is only valid to pass flags that are proved for
narrow type into a computation in wider type if we can prove that trunc instruction doesn't
actually change the value. If it did lose some meaningful bits, we may end up proving wrong
no-wrap flags for sum of arguments of trunc.
In the provided test we end up with `nuw` where it shouldn't be because of this bug.
The solution is to conservatively pass `SCEV::FlagAnyWrap` which is always a valid thing to do.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D49471
llvm-svn: 337435
It's a bit neater to write T.isIntOrPtrTy() over `T.isIntegerTy() ||
T.isPointerTy()`.
I used Python's re.sub with this regex to update users:
r'([\w.\->()]+)isIntegerTy\(\)\s*\|\|\s*\1isPointerTy\(\)'
llvm-svn: 336462
Summary:
Comment on Transforms/LoopVersioning/incorrect-phi.ll: With the change
SCEV is able to prove that the loop doesn't wrap-self (due to zext i16
to i64), disabling the entire loop versioning pass. Removed the zext and
just use i64.
Reviewers: sanjoy
Subscribers: jlebar, hiraditya, javed.absar, bixia, llvm-commits
Differential Revision: https://reviews.llvm.org/D48409
llvm-svn: 336140
We can have AddRec with loops having many predecessors.
This changes an assert to an early return.
Differential Revision: https://reviews.llvm.org/D48766
llvm-svn: 335965
Summary:
This initiates a discussion on changing Polly accordingly while re-applying r335197 (D48338).
I have never worked on Polly. The proposed change to param_div_div_div_2.ll is not educated, but just patterns that match the output.
All LLVM files are already reviewed in D48338.
Reviewers: jdoerfert, bollu, efriedma
Subscribers: jlebar, sanjoy, hiraditya, llvm-commits, bixia
Differential Revision: https://reviews.llvm.org/D48453
llvm-svn: 335292
Summary:
Try to match udiv and urem patterns, and sink zext down to the leaves.
I'm not entirely sure why some unrelated tests change, but the added <nsw>s seem right.
Reviewers: sanjoy
Subscribers: jlebar, hiraditya, bixia, llvm-commits
Differential Revision: https://reviews.llvm.org/D48338
llvm-svn: 335197
This reverts r334428. It incorrectly marks some multiplications as nuw. Tim
Shen is working on a proper fix.
Original commit message:
[SCEV] Add nuw/nsw to mul ops in StrengthenNoWrapFlags where safe.
Summary:
Previously we would add them for adds, but not multiplies.
llvm-svn: 335016
Summary:
Sending for presubmit review out of an abundance of caution; it would be
bad to mess this up.
Reviewers: sanjoy
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D48238
llvm-svn: 334875
Summary:
Obviates the need for mask/clear/setFlags helpers.
There are some expressions here which can be simplified, but to keep
this easy to review, I have not simplified them in this patch.
No functional change.
Reviewers: sanjoy
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D48237
llvm-svn: 334874
Summary:
Specifically, we transform
zext(2^K * (trunc X to iN)) to iM ->
2^K * (zext(trunc X to i{N-K}) to iM)<nuw>
This is helpful because pulling the 2^K out of the zext allows further
optimizations.
Reviewers: sanjoy
Subscribers: hiraditya, llvm-commits, timshen
Differential Revision: https://reviews.llvm.org/D48158
llvm-svn: 334737
Summary:
Previously we would do this simplification only if it did not introduce
any new truncs (excepting new truncs which replace other cast ops).
This change weakens this condition: If the number of truncs stays the
same, but we're able to transform trunc(X + Y) to X + trunc(Y), that's
still simpler, and it may open up additional transformations.
While we're here, also clean up some duplicated code.
Reviewers: sanjoy
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D48160
llvm-svn: 334736
Summary:
Previously we would add them for adds, but not multiplies.
Reviewers: sanjoy
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D48038
llvm-svn: 334428
An expression like
(zext i2 {(trunc i32 (1 + %B) to i2),+,1}<%while.body> to i32)
will become zero exactly when the nested value becomes zero in its type.
Strip injective operations from the input value in howFarToZero to make
the value simpler.
Differential Revision: https://reviews.llvm.org/D47951
llvm-svn: 334318
The DEBUG() macro is very generic so it might clash with other projects.
The renaming was done as follows:
- git grep -l 'DEBUG' | xargs sed -i 's/\bDEBUG\s\?(/LLVM_DEBUG(/g'
- git diff -U0 master | ../clang/tools/clang-format/clang-format-diff.py -i -p1 -style LLVM
- Manual change to APInt
- Manually chage DOCS as regex doesn't match it.
In the transition period the DEBUG() macro is still present and aliased
to the LLVM_DEBUG() one.
Differential Revision: https://reviews.llvm.org/D43624
llvm-svn: 332240
During simplification umax we trigger isKnownPredicate twice. As a first attempt it
tries the induction. To do that it tries to get post increment of SCEV.
Re-writing the SCEV may result in simplification of umax. If the SCEV contains a lot
of umax operations this recursion becomes very slow.
The added test demonstrates the slow behavior.
To resolve this we use only simple ways to check whether the predicate is known.
Reviewers: sanjoy, mkazantsev
Reviewed By: sanjoy
Subscribers: lebedev.ri, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D46046
llvm-svn: 331949
This patch was temporarily reverted because it has exposed bug 37229 on
PowerPC platform. The bug is unrelated to the patch and was just a general
bug in the optimization done for PowerPC platform only. The bug was fixed
by the patch rL331410.
This patch returns the disabled commit since the bug was fixed.
llvm-svn: 331427
See r331124 for how I made a list of files missing the include.
I then ran this Python script:
for f in open('filelist.txt'):
f = f.strip()
fl = open(f).readlines()
found = False
for i in xrange(len(fl)):
p = '#include "llvm/'
if not fl[i].startswith(p):
continue
if fl[i][len(p):] > 'Config':
fl.insert(i, '#include "llvm/Config/llvm-config.h"\n')
found = True
break
if not found:
print 'not found', f
else:
open(f, 'w').write(''.join(fl))
and then looked through everything with `svn diff | diffstat -l | xargs -n 1000 gvim -p`
and tried to fix include ordering and whatnot.
No intended behavior change.
llvm-svn: 331184
The invocation of getExact in ScalarEvolution::getBackedgeTakenInfo is used
only for getting statistic and for assert.
Even if statistics is disabled, the code related to it will be eliminated
the invocation to getExact itself will not be eliminated
because it may have side-effects like creation of new SCEVs.
So do invocation only when we collect statistics or executes asserts.
Reviewers: mkazantsev, sanjoy, javed.absar
Reviewed By: javed.absar
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D46178
llvm-svn: 331099
Add new umin creation method which accepts a list of operands.
SCEV does not represents umin which is required in getExact, so
it transforms umin to umax with not. As a result the transformation of
tree of max to max with several operands does not work.
We just use the new introduced method for creation umin from several operands.
Reviewers: sanjoy, mkazantsev
Reviewed By: sanjoy
Subscribers: javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D46047
llvm-svn: 331015
This reverts commit 023c8be90980e0180766196cba86f81608b35d38.
This patch triggers miscompile of zlib on PowerPC platform. Most likely it is
caused by some pre-backend PPC-specific pass, but we don't clearly know the
reason yet. So we temporally revert this patch with intention to return it
once the problem is resolved. See bug 37229 for details.
llvm-svn: 330893
In the function `simplifyOneLoop` we optimistically assume that changes in the
inner loop only affect this very loop and have no impact on its parents. In fact,
after rL329047 has been merged, we can now calculate exit counts for outer
loops which may depend on inner loops. Thus, we need to invalidate all parents
when we do something to a loop.
There is an evidence of incorrect behavior of `simplifyOneLoop`: when we insert
`SE->verify()` check in the end of this funciton, it fails on a bunch of existing
test, in particular:
LLVM :: Transforms/LoopUnroll/peel-loop-not-forced.ll
LLVM :: Transforms/LoopUnroll/peel-loop-pgo.ll
LLVM :: Transforms/LoopUnroll/peel-loop.ll
LLVM :: Transforms/LoopUnroll/peel-loop2.ll
Note that previously we have fixed issues of this variety, see rL328483.
This patch makes this function invalidate the outermost loop properly.
Differential Revision: https://reviews.llvm.org/D45937
Reviewed By: chandlerc
llvm-svn: 330576
This patch teaches SCEV how to prove implications for SCEVUnknown nodes that are Phis.
If we need to prove `Pred` for `LHS, RHS`, and `LHS` is a Phi with possible incoming values
`L1, L2, ..., LN`, then if we prove `Pred` for `(L1, RHS), (L2, RHS), ..., (LN, RHS)` then we can also
prove it for `(LHS, RHS)`. If both `LHS` and `RHS` are Phis from the same block, it is sufficient
to prove the predicate for values that come from the same predecessor block.
The typical case that it handles is that we sometimes need to prove that `Phi(Len, Len - 1) >= 0`
given that `Len > 0`. The new logic was added to `isImpliedViaOperations` and only uses it and
non-recursive reasoning to prove the facts we need, so it should not hurt compile time a lot.
Differential Revision: https://reviews.llvm.org/D44001
Reviewed By: anna
llvm-svn: 329150
The patch changes the usage of dominate to properlyDominate
to satisfy the condition !(a < a) while using std::max.
It is actually NFC due to set data structure is used to keep
the Loops and no two identical loops can be in collection.
So in reality there is no difference between usage of
dominate and properlyDominate in this particular case.
However it might be changed so it is better to fix it.
llvm-svn: 329051
Current implementation of `computeExitLimit` has a big piece of code
the only purpose of which is to prove that after the execution of this
block the latch will be executed. What it currently checks is actually a
subset of situations where the exiting block dominates latch.
This patch replaces all these checks for simple particular cases with
domination check over loop's latch which is the only necessary condition
of taking the exiting block into consideration. This change allows to
calculate exact loop taken count for simple loops like
for (int i = 0; i < 100; i++) {
if (cond) {...} else {...}
if (i > 50) break;
. . .
}
Differential Revision: https://reviews.llvm.org/D44677
Reviewed By: efriedma
llvm-svn: 329047
Summary:
r327219 added wrappers to std::sort which randomly shuffle the container before sorting.
This will help in uncovering non-determinism caused due to undefined sorting
order of objects having the same key.
To make use of that infrastructure we need to invoke llvm::sort instead of std::sort.
Note: This patch is one of a series of patches to replace *all* std::sort to llvm::sort.
Refer D44363 for a list of all the required patches.
Reviewers: sanjoy, dexonsmith, hfinkel, RKSimon
Reviewed By: dexonsmith
Subscribers: david2050, llvm-commits
Differential Revision: https://reviews.llvm.org/D44944
llvm-svn: 328925
Currently, `getExact` fails if it sees two exit counts in different blocks. There is
no solid reason to do so, given that we only calculate exact non-taken count
for exiting blocks that dominate latch. Using this fact, we can simply take min
out of all exits of all blocks to get the exact taken count.
This patch makes the calculation more optimistic with enforcing our assumption
with asserts. It allows us to calculate exact backedge taken count in trivial loops
like
for (int i = 0; i < 100; i++) {
if (i > 50) break;
. . .
}
Differential Revision: https://reviews.llvm.org/D44676
Reviewed By: fhahn
llvm-svn: 328611
This patch teaches `computeConstantDifference` handle calculation of constant
difference between `(X + C1)` and `(X + C2)` which is `(C2 - C1)`.
Differential Revision: https://reviews.llvm.org/D43759
Reviewed By: anna
llvm-svn: 328609
Summary:
Revert r325687 workaround for PR36032 since
a fix was committed in r326154.
Reviewers: sbaranga
Differential Revision: http://reviews.llvm.org/D44768
From: Evgeny Stupachenko <evstupac@gmail.com>
<evgeny.v.stupachenko@intel.com>
llvm-svn: 328257
This is re-land of https://reviews.llvm.org/rL327362 with a fix
and regression test.
The crash was due to it is possible that for found MDL loop,
LHS or RHS may contain an invariant unknown SCEV which
does not dominate the MDL. Please see regression
test for an example.
Reviewers: sanjoy, mkazantsev, reames
Reviewed By: mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44553
llvm-svn: 327822
Methods `computeExitLimitFromCondCached` and `computeExitLimitFromCondImpl` take
true and false branches as parameters and only use them for asserts and for identifying
whether true/false branch belongs to the loop (which can be done once earlier). This fact
complicates generalization of exit limit computation logic on guards because the guards
don't have blocks to which they go in case of failure explicitly.
The motivation of this patch is that currently this part of SCEV knows nothing about guards
and only works with explicit branches. As result, it fails to prove that a loop
for (i = 0; i < 100; i++)
guard(i < 10);
exits after 10th iteration, while in the equivalent example
for (i = 0; i < 100; i++)
if (i >= 10) break;
SCEV easily proves this fact. We are going to change it in near future, and this is why
we need to make these methods operate on more abstract level.
This patch refactors this code to get rid of these parameters as meaningless and prepare
ground for teaching these methods to work with guards as well as they work with explicit
branching instructions.
Differential Revision: https://reviews.llvm.org/D44419
llvm-svn: 327615
