Revert "[InstCombine] Canonicalize `(sitofp x)` -> `(uitofp x)` if `x >= 0`"

This reverts commit d80d5b923c6f611590a12543bdb33e0c16044d44.

It wasn't a particularly important transform to begin with

Revert "[InstCombine] Canonicalize `(sitofp x)` -> `(uitofp x)` if `x >= 0`"

This reverts commit d80d5b923c6f611590a12543bdb33e0c16044d44.

It wasn't a particularly important transform to begin with and caused
some codegen regressions on targets that prefer `sitofp` so dropping.

Might re-visit along with adding `nneg` flag to `uitofp` so its easily
reversable for the backend.

show more ...

[InstCombine] Fold `fpto{s|u}i non-norm` to zero (#85569)

This patch enables more optimization after canonicalizing `fmul X, 0.0`
into a copysign.
I decide to implement this fold in InstCombine be

[InstCombine] Fold `fpto{s|u}i non-norm` to zero (#85569)

This patch enables more optimization after canonicalizing `fmul X, 0.0`
into a copysign.
I decide to implement this fold in InstCombine because
`computeKnownFPClass` may be expensive.

Alive2: https://alive2.llvm.org/ce/z/ASM8tQ

show more ...

[InstCombine] Canonicalize `(sitofp x)` -> `(uitofp x)` if `x >= 0`

Just a standard canonicalization.

Proofs: https://alive2.llvm.org/ce/z/9W4VFm

Closes #82404

[InstCombine] Simplify `zext nneg i1 X` to zero (#85043)

Alive2: https://alive2.llvm.org/ce/z/Wm6kCk

[InstCombine] Fix shift calculation in InstCombineCasts (#84027)

Fixes #84025.

[InstCombine] Pick bfloat over half when shrinking ops that started with an fpext from bfloat (#82493)

This fixes the case where we would shrink an frem to half and then
bitcast to bfloat, producin

[InstCombine] Pick bfloat over half when shrinking ops that started with an fpext from bfloat (#82493)

This fixes the case where we would shrink an frem to half and then
bitcast to bfloat, producing invalid results. The transformation was
written under the assumption that there is only one type with a given
bit width.

Also add a strategic assert to CastInst::CreateFPCast to turn this
miscompilation into a crash.

show more ...

[PatternMatch] Add a matching helper `m_ElementWiseBitCast`. NFC. (#80764)

This patch introduces a matching helper `m_ElementWiseBitCast`, which is
used for matching element-wise int <-> fp casts.

[PatternMatch] Add a matching helper `m_ElementWiseBitCast`. NFC. (#80764)

This patch introduces a matching helper `m_ElementWiseBitCast`, which is
used for matching element-wise int <-> fp casts.
The motivation of this patch is to avoid duplicating checks in
https://github.com/llvm/llvm-project/pull/80740 and
https://github.com/llvm/llvm-project/pull/80414.

show more ...

[InstCombine] Try to fold trunc(shuffle(zext)) to just a shuffle (#78636)

Tries to remove extra trunc/ext instruction for shufflevector
instructions.

Differential Review: https://github.com/llvm/ll

[InstCombine] Try to fold trunc(shuffle(zext)) to just a shuffle (#78636)

Tries to remove extra trunc/ext instruction for shufflevector
instructions.

Differential Review: https://github.com/llvm/llvm-project/pull/78636

show more ...

Revert "[InstCombine] Try to fold trunc(shuffle(zext)) to just a shuffle (#78636)"

This reverts commit 4d11f04b20f0bd7488e19e8f178ba028412fa519.

This breaks some programs as mentioned in #78636

[InstCombine] Try to fold trunc(shuffle(zext)) to just a shuffle (#78636)

Tries to remove extra trunc/ext instruction for shufflevector
instructions.

[ValueTracking] Convert isKnownNonNegative() to use SimplifyQuery (NFC)

[InstCombine] Fix incorrect nneg inference on shift amount

Whether this is valid depends on the bit widths of the involved
integers.

Fixes https://github.com/llvm/llvm-project/issues/72927.

[InstCombine] Infer nneg flag from shift users (#71947)

This patch sets `nneg` flag when the zext is only used by a shift.

Alive2: https://alive2.llvm.org/ce/z/h3xKjP
Compile-time impact:
https

[InstCombine] Infer nneg flag from shift users (#71947)

This patch sets `nneg` flag when the zext is only used by a shift.

Alive2: https://alive2.llvm.org/ce/z/h3xKjP
Compile-time impact:
https://llvm-compile-time-tracker.com/compare.php?from=bd611264993f64decbce178d460caf1d1cb05f59&to=26bc473b239010bb24ff1bc39d58b42ecbbc4730&stat=instructions:u

This is an alternative to #71906.

show more ...

[InstCombine] Avoid uses of ConstantExpr::getLShr()

Use the constant folding API instead.

[InstCombine] Infer zext nneg flag (#71534)

Use KnownBits to infer the nneg flag on zext instructions.

Currently we only set nneg when converting sext -> zext, but don't set
it when we have a ze

[InstCombine] Infer zext nneg flag (#71534)

Use KnownBits to infer the nneg flag on zext instructions.

Currently we only set nneg when converting sext -> zext, but don't set
it when we have a zext in the first place. If we want to use it in
optimizations, we should make sure the flag inference is consistent.

show more ...

[InstCombine] Favour `CreateZExtOrTrunc` in `narrowFunnelShift` (NFC)

Use `CreateZExtOrTrunc`, reduce test and regenerate checks.

[InstCombine] Zero-extend shift amounts in narrow funnel shift ops

An issue arose when handling shift amounts while performing
narrowed funnel shifts simplification. Specifically, shift
amounts were

[InstCombine] Zero-extend shift amounts in narrow funnel shift ops

An issue arose when handling shift amounts while performing
narrowed funnel shifts simplification. Specifically, shift
amounts were incorrectly truncated when their type was
narrower than the target bit width. This has been addressed
by zero-extending `ShAmt` in such cases.

Fixes: https://github.com/llvm/llvm-project/issues/71463.

Proof: https://alive2.llvm.org/ce/z/5draKz.

show more ...

[InstCombine] Fold `(ptrtoint (ptrmask p0, m0))` -> `(and (ptrtoint p0), m0)`

`and` is generally more supported so if we have a `ptrmask` anyways
might as well use `and`.

Differential Revision: htt

[InstCombine] Fold `(ptrtoint (ptrmask p0, m0))` -> `(and (ptrtoint p0), m0)`

`and` is generally more supported so if we have a `ptrmask` anyways
might as well use `and`.

Differential Revision: https://reviews.llvm.org/D156640

Closes #67166

show more ...

[InstCombine] Fix canAlwaysEvaluateInTy() with constant exprs

The m_ZExtOrSExt / m_Trunc in the following code can match constant
expressions, which we don't want here. Make sure we bail out early
f

[InstCombine] Fix canAlwaysEvaluateInTy() with constant exprs

The m_ZExtOrSExt / m_Trunc in the following code can match constant
expressions, which we don't want here. Make sure we bail out early
for non-immediate constants.

show more ...

[InstCombine] Avoid use of ConstantExpr::getIntegerCast()

Require that constants are ImmConstant for this transform, as we
may otherwise generate constant expressions, which are not
necessarily free.

[InstCombine] Avoid some uses of ConstantExpr::getIntegerCast() (NFC)

Use IRBuilder or ConstantFolding instead.

[InstCombine] Infer nneg on zext when forming from non-negative sext (#70706)

Builds on #67982 which recently introduced the nneg flag on a zext
instruction. InstCombine is one of our largest canon

[InstCombine] Infer nneg on zext when forming from non-negative sext (#70706)

Builds on #67982 which recently introduced the nneg flag on a zext
instruction. InstCombine is one of our largest canonicalizers of zext
from non-negative sext instructions, so set the flag there.

show more ...

[InstCombine] Remove redundant cast of GEP fold (NFC)

With opaque pointers, zero-index GEPs will be eliminated in
general.

[InstCombine] Fold zext-of-icmp with no shift (#68503)

This regression triggers after commit f400daa to fix infinite loop
issue.

In this case, we can known the shift count is 0, so it will not b

[InstCombine] Fold zext-of-icmp with no shift (#68503)

This regression triggers after commit f400daa to fix infinite loop
issue.

In this case, we can known the shift count is 0, so it will not be
triggered by the form of (iN (~X) u>> (N - 1)) in commit 21d3871, of
which N indicates the data type bitwidth of X.

Fixes https://github.com/llvm/llvm-project/issues/68465.

show more ...

[InstCombine] Avoid use of ConstantExpr::getSExt() (NFC)

Use the constant folding API instead.