[InstCombine] restrict select of bit-tests to constant shift amounts

This transform is responsible for a long-standing miscompile
as discussed in issue #47012 (was bugzilla #47668).

There was a pro

[InstCombine] restrict select of bit-tests to constant shift amounts

This transform is responsible for a long-standing miscompile
as discussed in issue #47012 (was bugzilla #47668).

There was a proposal to correct it in D88432, but that was
abandoned and there hasn't been any recent activity to fix
it AFAICT.

The original patch D45108 started with a constant-shift-only
restriction and only expanded during review, so I don't think
there's much risk of perf regression on the motivating code.

show more ...

[NFC][InstCombine] Refactor InstCombinerImpl::foldSelectIntoOp

Introduce a lambda function so that we remove a lot of code
duplication.

Differential Revision: https://reviews.llvm.org/D127493

[InstCombine] Use +0.0 instead of -0.0 as the FP identity for some folds

In foldSelectIntoOp we sometimes transform a select of a fadd into a
fadd of a select, where we select between data and an id

[InstCombine] Use +0.0 instead of -0.0 as the FP identity for some folds

In foldSelectIntoOp we sometimes transform a select of a fadd into a
fadd of a select, where we select between data and an identity value.
For both fadd and fsub the identity is always -0.0, but if the nsz
flag is set on the select instruction we can use +0.0 instead. Doing
so then triggers other optimisations, such as when folding the select
of masked load into a new masked load.

Differential Revision: https://reviews.llvm.org/D126774

show more ...

[NFC] format InstructionSimplify & lowerCaseFunctionNames

Clang-format InstructionSimplify and convert all "FunctionName"s to
"functionName". This patch does touch a lot of files but gets done with

[NFC] format InstructionSimplify & lowerCaseFunctionNames

Clang-format InstructionSimplify and convert all "FunctionName"s to
"functionName". This patch does touch a lot of files but gets done with
the cleanup of InstructionSimplify in one commit.

This is the alternative to the less invasive clang-format only patch: D126783

Reviewed By: spatel, rengolin

Differential Revision: https://reviews.llvm.org/D126889

show more ...

[InstCombine] Reuse icmp of and/or folds for logical and/or

Similarly to a change recently done for fcmps, add a flag that
indicates whether the and/or is logical to foldAndOrOfICmps, and
reuse the

[InstCombine] Reuse icmp of and/or folds for logical and/or

Similarly to a change recently done for fcmps, add a flag that
indicates whether the and/or is logical to foldAndOrOfICmps, and
reuse the function when folding logical and/or.

We were already calling some parts of it, but this gives us a
clearer indication of which parts may need poison-safe variants,
and would also allow to fold combinations of bitwise and logical
and/or.

This change should be close to NFC, because all folds this enables
were either already called previously, or can make use of implied
poison reasoning.

show more ...

[InstCombine] Allow undef vectors when foldSelectToCopysign

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D125671

[IR] create and use helper functions that test the signbit; NFCI

[InstCombine] Remove the undef-related workaround code in visitSelectInst

This patch removes an old hack in visitSelectInst that was written to avoid miscompilation bugs in loop unswitch.
(Added via

[InstCombine] Remove the undef-related workaround code in visitSelectInst

This patch removes an old hack in visitSelectInst that was written to avoid miscompilation bugs in loop unswitch.
(Added via https://reviews.llvm.org/D35811)

The legacy loop unswitch pass will be removed after D124376, and the new simple loop unswitch pass correctly uses freeze to avoid introducing UB after D124252.

Since the hack is not necessary anymore, this patch removes it.

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D124426

show more ...

[InstCombine] Fold logical and/or of range icmps with nowrap flags

This is an edge-case where we don't convert to bitwise and/or based
on implies poison reasoning, so explicitly try to perform the f

[InstCombine] Fold logical and/or of range icmps with nowrap flags

This is an edge-case where we don't convert to bitwise and/or based
on implies poison reasoning, so explicitly try to perform the fold
in logical form. The transform itself is poison-safe, as both icmps
are based on the same value and any nowrap flags are discarded as
part of the fold (https://alive2.llvm.org/ce/z/aCwC8b for the used
example).

show more ...

[InstCombine] 'round up integer': if bias is just right, just reuse instructions

This is only useful if we can't create new instruction
because %x.aligned has other uses and already sticks around.

[InstCombine] Fold 'round up integer' pattern (when alignment is a power of two)

But don't deal with non-splats.

The test coverage is sufficiently exhaustive,
and alive is happy about the changes t

[InstCombine] Fold 'round up integer' pattern (when alignment is a power of two)

But don't deal with non-splats.

The test coverage is sufficiently exhaustive,
and alive is happy about the changes there.

Example with constants: https://alive2.llvm.org/ce/z/EUaJ5- / https://alive2.llvm.org/ce/z/Bkng2X
General proof: https://alive2.llvm.org/ce/z/3RjJ5A

show more ...

[InstCombine] Update predicate when canonicalizing comparisons in canonicalizeClampLike.

canonicalizeClampLike canonicalizes the ule/ugt comparisons to ult/uge,
respectively. However, it does not up

[InstCombine] Update predicate when canonicalizing comparisons in canonicalizeClampLike.

canonicalizeClampLike canonicalizes the ule/ugt comparisons to ult/uge,
respectively. However, it does not update the variable holding the
comparison predicate type after doing this. Later code fails to handle
the non-canonical predicate type (specifically, the swap of
ThresholdLowIncl and ThresholdHighExcl when Pred0 has been canonicalized
from ugt to uge). This leads to the miscompile reported in PR53252. Fix
this by updating the comparison predicate after canonicalizing.

Fixes #53252

Differential Revision: https://reviews.llvm.org/D119690

show more ...

[InstCombine] Complete folding of fneg-of-fabs

This patch add a function foldSelectWithFCmpToFabs, and do more combine for
fneg-of-fabs.
With 'nsz':
fold (X < +/-0.0) ? X : -X or (X <= +/-0.0) ? X

[InstCombine] Complete folding of fneg-of-fabs

This patch add a function foldSelectWithFCmpToFabs, and do more combine for
fneg-of-fabs.
With 'nsz':
fold (X < +/-0.0) ? X : -X or (X <= +/-0.0) ? X : -X to -fabs(x)
fold (X > +/-0.0) ? X : -X or (X >= +/-0.0) ? X : -X to -fabs(x)

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D123830

show more ...

[InstCombine] canonicalize select with signbit test

This is part of solving issue #54750 - in that example
we have both forms of the compare and do not recognize
the equivalence.

[InstCombine] Fold icmp(X) ? f(X) : C

This diff extends foldSelectInstWithICmp to handle the case icmp(X) ? f(X) : C
when f(X) is guaranteed to be equal to C for all X in the exact range of the inve

[InstCombine] Fold icmp(X) ? f(X) : C

This diff extends foldSelectInstWithICmp to handle the case icmp(X) ? f(X) : C
when f(X) is guaranteed to be equal to C for all X in the exact range of the inverse predicate.
This addresses the issue https://github.com/llvm/llvm-project/issues/54089.

Differential revision: https://reviews.llvm.org/D123159

Test plan: make check-all

show more ...

[SVE] Extend support for folding select + masked gathers

Extend the work done in D106376 to include masked gathers

Differential Revision: https://reviews.llvm.org/D122896

[InstCombine] Fold two select patterns into and-or

select (~a | c), a, b -> and a, (or c, b) https://alive2.llvm.org/ce/z/bnDobs
select (~c & b), a, b -> and b, (or a, c) https://alive2.llvm.org/ce/

[InstCombine] Fold two select patterns into and-or

select (~a | c), a, b -> and a, (or c, b) https://alive2.llvm.org/ce/z/bnDobs
select (~c & b), a, b -> and b, (or a, c) https://alive2.llvm.org/ce/z/k2jJHJ

Differential Revision: https://reviews.llvm.org/D122152

show more ...

[InstCombine] Remove integer SPF of SPF folds (NFCI)

Now that we canonicalize to intrinsics, these folds should no
longer be needed. Only one fold that also applies to floating-point
min/max is reta

[InstCombine] Remove integer SPF of SPF folds (NFCI)

Now that we canonicalize to intrinsics, these folds should no
longer be needed. Only one fold that also applies to floating-point
min/max is retained.

show more ...

[InstCombine] Fold select based logic of fcmps with same operands when FMF is present.

If we have a logical and/or in select form and the true/false operand
is an fcmp with poison generating FMF, we

[InstCombine] Fold select based logic of fcmps with same operands when FMF is present.

If we have a logical and/or in select form and the true/false operand
is an fcmp with poison generating FMF, we won't be able to fold it
to an and/or instruction. This prevents us from optimizing the case
where it is a logical operation of two fcmps with identical operands.

This patch adds explicit checks for this case that doesn't rely on
converting to and/or to do the optimization. It reuses the existing
foldLogicOfFCmps, but adds a new flag to disable the other combine
that is inside that function.

FMF flags from the two FCmps are intersected using the logic added in
D121243. The FIXME has been updated to indicate that we can only use
a union for the non-select form.

This allows us to optimize cases like this from compare-fp-3.c in the
gcc torture suite with fast math.

void
test1 (float x, float y)
{
if ((x==y) && (x!=y))
link_error0();
}

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D121323

show more ...

[InstCombine] Slightly relax one-use check in abs canonicalization

Treat the icmp and sub symmetrically, and require that one of them
has one use, not the icmp in particular. This could be further
r

[InstCombine] Slightly relax one-use check in abs canonicalization

Treat the icmp and sub symmetrically, and require that one of them
has one use, not the icmp in particular. This could be further
relaxed in the abs (but not nabs) case to not check one-use at
all.

show more ...

[InstCombine] drop FMF in select->copysign transform

It is not correct to propagate flags from the select
to the new instructions:
https://alive2.llvm.org/ce/z/tNATrd
https://alive2.llvm.org/ce/z/Vw

[InstCombine] drop FMF in select->copysign transform

It is not correct to propagate flags from the select
to the new instructions:
https://alive2.llvm.org/ce/z/tNATrd
https://alive2.llvm.org/ce/z/VwcVzn

Fixes #54077

show more ...

[InstCombine] fold select-of-zero-or-ones with negated op

(X u< 2) ? -X : -1 --> sext (X != 0)
(X u> 1) ? -1 : -X --> sext (X != 0)

https://alive2.llvm.org/ce/z/U3y5Bb
https://alive2.llvm.org/ce/z/

[InstCombine] fold select-of-zero-or-ones with negated op

(X u< 2) ? -X : -1 --> sext (X != 0)
(X u> 1) ? -1 : -X --> sext (X != 0)

https://alive2.llvm.org/ce/z/U3y5Bb
https://alive2.llvm.org/ce/z/hgi-4p

This is part of solving:

show more ...

[InstCombine] Don't call matchSAddSubSat() for SPF (NFC)

Only call it for intrinsic min/max. The moved implementation is
unchanged apart from the one-use check: It is now hardcoded to
one-use, witho

[InstCombine] Don't call matchSAddSubSat() for SPF (NFC)

Only call it for intrinsic min/max. The moved implementation is
unchanged apart from the one-use check: It is now hardcoded to
one-use, without the two-use special case for SPF.

show more ...

[InstCombine] Remove SPF moveAddAfterMinMax() (NFC)

As SPF min/max is canonicalized to intrinsics before this point,
this change should be entirely NFC.

[InstCombine] Remove SPF moveNotAfterMinMax() (NFC)

This happens after SPF -> intrinsic canonicalization, and as such
should be entirely NFC.