This patch adds the support for and/or in `getFreelyInvertedImpl` using
DeMorgan's Law:
```
(~(A | B)) -> (~A & ~B)
(~(A & B)) -> (~A | ~B)
```
Alive2: https://alive2.llvm.org/ce/z/Uig8-j
The instruction itself can be considered good for minbitwidth casting,
even if one of the operand checks returns false.
Reviewers: RKSimon
Reviewed By: RKSimon
Pull Request: https://github.com/llvm/llvm-project/pull/84363
The instruction itself can be considered good for minbitwidth casting,
even if one of the operand checks returns false.
Reviewers: RKSimon
Reviewed By: RKSimon
Pull Request: https://github.com/llvm/llvm-project/pull/84363
We can treat a shift by constant as a multiply by a power of 2
and we can treat an or disjoint as a 'add nsw nuw'.
I've added a helper struct similar to a struct used in
ScalarEvolution.cpp
to represent the opcode, operands, and NSW/NUW flags for normal
add/sub/mul
and shl/or that are being treated as mul/add.
I don't think we need to teach cloneIVUser about this. It will continue
to clone them using cloneBitwiseIVUser. After the cloning we will ask
for the SCEV expression for the cloned IV user and verify that it
matches
the AddRec returned by getExtendedOperandRecurrence. Since SCEV also
knows how to convert shl to mul and or disjoint to add nsw nuw, this
should
usually match. If it doesn't match, the cloned IV user will be deleted.
getArithmeticInstrCost is used by both LoopVectorizer and SLPVectorizer
to compute the cost of frem, which becomes a call cost on AArch64 when
TLI has a vector library function.
Add tests that do SLP vectorization for code that contains 2x double and
4x float frem instructions.
Try to transform the powi(X, Y) * X into powi(X, Y+1) with Ofast
For this case, when the Y is 3, then powi(X, 4) is replaced by
X2 = X * X; X2 * X2 in the further step.
Similar to D109954, who requires reassoc.
Fixes https://github.com/llvm/llvm-project/issues/69862.
This improves overall analysis for minbitwidth in SLP. It allows to
analyze the trees with store/insertelement root nodes. Also, instead of
using single minbitwidth, detected from the very first analysis stage,
it tries to detect the best one for each trunc/ext subtree in the graph
and use it for the subtree.
Results in better code and less vector register pressure.
Metric: size..text
Program size..text
results results0 diff
test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test 92549.00 92609.00 0.1%
test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 663381.00 663493.00 0.0%
test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 663381.00 663493.00 0.0%
test-suite :: MultiSource/Benchmarks/Bullet/bullet.test 307182.00 307214.00 0.0%
test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test 1394420.00 1394484.00 0.0%
test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test 1394420.00 1394484.00 0.0%
test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2040257.00 2040273.00 0.0%
test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0%
test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test 909944.00 909768.00 -0.0%
SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar
instructions remain scalar (good).
Spec2017/x264 - the whole function idct4x4dc is vectorized using <16
x i16> instead of <16 x i32>, also zext/trunc are removed. In other
places last vector zext/sext removed and replaced by
extractelement + scalar zext/sext pair.
MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by
reduce or <4 x i8>
Spec2017/imagick - Removed extra zext from 2 packs of the operations.
Spec2017/parest - Removed extra zext, replaced by extractelement+scalar
zext
Spec2017/blender - the whole bunch of vector zext/sext replaced by
extractelement+scalar zext/sext, some extra code vectorized in smaller
types.
Spec2006/gobmk - fixed cost estimation, some small code remains scalar.
Original Pull Request: https://github.com/llvm/llvm-project/pull/84334
The patch has the same functionality (no test changes, no changes in
benchmarks) as the original patch, just has some compile time
improvements + fixes for xxhash unittest, discovered earlier in the
previous version of the patch.
Reviewers:
Pull Request: https://github.com/llvm/llvm-project/pull/84536
This patch canonicalizes `extractvalue (select Cond, TV, FV)` into
`select Cond, (extractvalue TV), (extractvalue FV)`. The latter form may
enable more optimizations.
Loop guards that apply to loop SCEV bounds allow IRCE for cases with
compound loop bounds such as:
if (K > 0 && M > 0)
for (i = 0; i < min(K, M); i++) {...}
if (K > 0 && M > 0)
for (i = min(K, M); i >= 0; i--) {...}
Otherwise SCEV couldn't prove that loops have safe bounds in these
cases.
Co-authored-by: Aleksander Popov <apopov@azul.com>
The function does exist, but it is a plain wrapper over regular ldexp(),
so there's no benefit in calling it over regular ldexp(). Therefore,
treat it as missing.
This fixes builds of Wine for aarch64 with Clang in mingw mode, which
regressed recently in 8d976c7f20fe8d92fe6f54af411594e15fac25ae. That
commit unlocked transforming calls to ldexp into ldexpf, for some
codepaths within Wine. Wine can use compilers in mingw mode without
the regular mingw runtime libraries, which caused this to fail.
(However, if the transformation to use ldexpf() would have made sense,
the right fix would have been for Wine to provide a similar
ldexpf->ldexp wrapper just like mingw does.)
The TranspBlocks set was used to cache aliasing decision for all
processed loads in the parent loop. This is incorrect, because each load
can access a different location, which means one load not being modified
in a block doesn't translate to another load not being modified in the
same block.
All loads access the same underlying object, so we could perhaps use a
location without size for all loads and retain the cache, but that would
mean we loose precision.
For now, just drop the cache.
Fixes https://github.com/llvm/llvm-project/issues/84807
PR: https://github.com/llvm/llvm-project/pull/84835
With special case with Add constant is 0.
Original message:
We can support these by changing the sext promotion to -zext(-C) and
replacing a sgt check with ugt. Reframing the logic in terms of how the
unsigned range are affected. More comments in the patch.
The new cases check isLegalAddImmediate to avoid some
regressions in lit tests.
