Fold trig functions call of poison to poison.
This includes sin, cos, asin, acos, atan, atan2, sinh, cosh, sincos,
sincospi.
Test cases are fixed and also added to
llvm/test/Transforms/InstSimplify/fold-intrinsics.ll just like in
https://github.com/llvm/llvm-project/pull/146750
This consolidates the "fold poison arg to poison result" constant
folding logic for intrinsics, based on a common
intrinsicPropagatesPoison() helper, which is also used for poison
propagation reasoning in ValueTracking. This ensures that the set of
supported intrinsics is consistent.
This add ucmp, scmp, smul.fix, smul.fix.sat, canonicalize and sqrt to
the intrinsicPropagatesPoison list, as these were handled by
ConstantFolding but not ValueTracking. The ctpop test is an example of
the converse, where it was handled by ValueTracking but not
ConstantFolding.
Similarly to what it is being done to match simple recurrence cycle
relations, attempt to match value-accumulating recurrences of kind:
```
%umax.acc = phi i8 [ %umax, %backedge ], [ %a, %entry ]
%umax = call i8 @llvm.umax.i8(i8 %umax.acc, i8 %b)
```
Preliminary work to let InstCombine avoid folding such recurrences,
so that simple loop-invariant computation may get hoisted. Minor
opportunity to refactor out code as well.
When the original predicate is ordered and both operands are non-NaN,
`Ordered` should be set to true. This variable still matters even if
both operands are non-NaN because FMF only applies to select, not fcmp.
Closes https://github.com/llvm/llvm-project/issues/143123.
Having a finite Depth (or recursion limit) for computeKnownBits is very
limiting, but is currently a load-bearing necessity, as all KnownBits
are recomputed on each call and there is no caching. As a prerequisite
for an effort to remove the recursion limit altogether, either using a
clever caching technique, or writing a easily-invalidable KnownBits
analysis, make the Depth argument in APIs in ValueTracking uniformly the
last argument with a default value. This would aid in removing the
argument when the time comes, as many callers that currently pass 0
explicitly are now updated to omit the argument altogether.
add `GenericFloatingPointPredicateUtils` in order to generalize
effects of floating point comparisons on `KnownFPClass` for both IR and
MIR.
---------
Co-authored-by: Matt Arsenault <arsenm2@gmail.com>
For now use the same treatment as minnum/maxnum, but these should
diverge. alive2 seems happy with this, except for some preexisting bugs
with weird denormal modes.
For GEPs, we have three bit widths involved: The pointer bit width, the
index bit width, and the bit width of the GEP operands.
The correct behavior here is:
* We need to sextOrTrunc the GEP operand to the index width *before*
multiplying by the scale.
* If the index width and pointer width differ, GEP only ever modifies
the low bits. Adds should not overflow into the high bits.
I'm testing this via unit tests because it's a bit tricky to test in IR
with InstCombine canonicalization getting in the way.
Rename the function to reflect its correct behavior and to be consistent
with `Module::getOrInsertFunction`. This is also in preparation of
adding a new `Intrinsic::getDeclaration` that will have behavior similar
to `Module::getFunction` (i.e, just lookup, no creation).
According to IEEE Std 754-2019, `sqrt` returns nan when the input is
negative (except for -0). In this case, we cannot make assumptions about
sign bit of the result.
Fixes https://github.com/llvm/llvm-project/issues/92217
Prior to #85863, the required parameters of llvm::isKnownNonZero were
Value and DataLayout. After, they are Value, Depth, and SimplifyQuery,
where SimplifyQuery is implicitly constructible from DataLayout. The
change to move Depth before SimplifyQuery needed callers to be updated
unnecessarily, and as commented in #85863, we actually want Depth to be
after SimplifyQuery anyway so that it can be defaulted and the caller
does not need to specify it.
If a function only exits for certain input values we can still derive
that an argument is "returned". We can also derive range metadata that
describe the possible value range returned by the function. However, it
turns out that those two analyses can result in conflicting information.
Example:
declare i16 @foo(i16 returned)
...
%A = call i16 @foo(i16 4095), !range !{i16 32, i16 33}
To avoid "Bits known to be one AND zero?" assertion failures we know
make sure to discard the known bits for this kind of scenario.
This patch extends `propagatesPoison` to handle more integer intrinsics.
It will turn more logical ands/ors into bitwise ands/ors.
See also https://reviews.llvm.org/D99671.
Shufflevector semantics have changed so that poison mask elements
return poison rather than undef. Reflect this in the
canCreateUndefOrPoison() implementation.
It's not safe for InstCombine to add disjoint metadata when converting
Add to Or otherwise.
I've added noundef attribute to preserve existing test behavior.
In the failure case we return null, which callers are checking. We were
also returning an fcNone which was unused. It's more consistent to
return fcAllFlags as any possible value, such that the value is always
directly usable without checking the returned value.
When processing assumes, we also handle assumes on ptrtoint of the
value. In canonical IR, these will have the same size as the value.
However, in non-canonical IR there may be an implicit zext or
trunc, which results in a bit width mismatch. We currently handle
this by adjusting bitwidth everywhere, but this is fragile and I'm
pretty sure that the way we do this is incorrect for some predicates,
because we effectively end up commuting an ext/trunc and an icmp.
Instead, add an m_PtrToIntSameSize() matcher that will only handle
bitwidth preserving cases. For the bitwidth-changing cases, wait
until they have been canonicalized.
The original handling for this was added purely to prevent crashes
in an earlier implementation which failed to account for this
entirely.
Use a unit test since I don't see any existing uses try to make use of
the high bits of a pointer.
This will also assert if the metadata type doesn't match the pointer
width, but I consider that a defect in the verifier and shouldn't be
handled.
AMDGPU allocates LDS globals by assigning !absolute_symbol with the
final fixed address. Tracking the high bits are 0 may help with
addressing mode matching.
For always poison shifts, any KnownBits return value is valid.
Currently we return unknown, but returning zero is generally more
profitable. We had some code in ValueTracking that tried to do this,
but was actually dead code.
Differential Revision: https://reviews.llvm.org/D150648
