Mask/Bool vectors are often bitcast to/from scalar integers, in particular when concatenating mask results, often this is due to the difficulties of working with vector of bools on C/C++. On x86 this typically involves the MOVMSK/KMOV instructions.
To concatenate bool masks, these are typically cast to scalars, which are then zero-extended, shifted and OR'd together.
This patch attempts to match these scalar concatenation patterns and convert them to vector shuffles instead. This in turn often assists with further vector combines, depending on the cost model.
Reapplied patch from #119559 - fixed use after free issue.
Fixes#111431
Instead of going through getOrAddLiveIn to get a VPValue for the vector
trip count retrieve it directly from VPlan via getVectorTripCount.
Small simplification following 0e70289f373.
Noticed while investigating a crash in #119559 - we don't account for I being replaced and its Type being reallocated. So hoist the checks to the start of the loop.
Consider the following loop:
```
int rdx = init;
for (int i = 0; i < n; ++i)
rdx = (a[i] > b[i]) ? i : rdx;
```
We can vectorize this loop if `i` is an increasing induction variable.
The final reduced value will be the maximum of `i` that the condition
`a[i] > b[i]` is satisfied, or the start value `init`.
This patch added new RecurKind enums - IFindLastIV and FFindLastIV.
---------
Co-authored-by: Alexey Bataev <5361294+alexey-bataev@users.noreply.github.com>
This patch introduces a new method:
void Vectorizer::mergeEquivalenceClasses(EquivalenceClassMap &EQClasses)
const
The method is called at the end of
Vectorizer::collectEquivalenceClasses() and is needed to merge
equivalence classes that differ only by their underlying objects (UO1
and UO2), where UO1 is 1-level-indirection underlying base for UO2. This
situation arises due to the limited lookup depth used during the search
of underlying bases with llvm::getUnderlyingObject(ptr).
Using any fixed lookup depth can result into creation of multiple
equivalence classes that only differ by 1-level indirection bases.
The new approach merges equivalence classes if they have adjacent bases
(1-level indirection). If a series of equivalence classes form ladder
formed of 1-step/level indirections, they are all merged into a single
equivalence class. This provides more opportunities for the load-store
vectorizer to generate better vectors.
---------
Signed-off-by: Klochkov, Vyacheslav N <vyacheslav.n.klochkov@intel.com>
A more lightweight variant of
https://github.com/llvm/llvm-project/pull/109193,
which dispatches to multiple exit blocks via the middle blocks.
The patch also introduces a bit of required scaffolding to enable
early-exit vectorization, including an option. At the moment, early-exit
vectorization doesn't come with legality checks, and is only used if the
option is provided and the loop has metadata forcing vectorization. This
is only intended to be used for testing during bring-up, with @david-arm
enabling auto early-exit vectorization plugging in the changes from
https://github.com/llvm/llvm-project/pull/88385.
PR: https://github.com/llvm/llvm-project/pull/112138
Mask/Bool vectors are often bitcast to/from scalar integers, in particular when concatenating mask results, often this is due to the difficulties of working with vector of bools on C/C++. On x86 this typically involves the MOVMSK/KMOV instructions.
To concatenate bool masks, these are typically cast to scalars, which are then zero-extended, shifted and OR'd together.
This patch attempts to match these scalar concatenation patterns and convert them to vector shuffles instead. This in turn often assists with further vector combines, depending on the cost model.
Fixes#111431
foldInsExtVectorToShuffle is likely to be inserting into an undef value, so make sure we've canonicalized this to the RHS in the folded shuffle to help further VectorCombine folds.
Minor tweak to help #34072
To reduce repeated code, TreeEntry::setOperandsInOrder will be replaced
by VLOperands.
Arg_size will be provided to make sure other operands will not be
reorderd when VL[0] is IntrinsicInst (because APO is a boolean value).
In addition, BoUpSLP::reorderInputsAccordingToOpcode will also be
removed since it is simple.
When checking for the poison elements in the matches node, need to
consider the register number, when clearing the corresponding mask
element.
Fixes#119393
Add getMainOp and getAltOp.
Use `InstructionsState &` instead of `const InstructionsState &`.
Use `!S.isAltShuffle()` instead of `S.MainOp == S.AltOp`.
foldShuffleOfShuffles already handles "shuffle (shuffle x, undef), (shuffle y, undef)" patterns, this patch relaxes the requirement so it can handle cases where only a single operand is a shuffle (and the other can be any other value and will be kept in place).
Fixes#86068
Adjust the assertion in fixupIVUsers to only require a unique exit block
if there are any values to fix up. This enables the bring up of
multi-exit loop vectorization without requiring a scalar epilogue.
Split off as suggested from
https://github.com/llvm/llvm-project/pull/112138.
Use getUniqueLatchExitBlock instead of getUniqueExitBlock in preparation
for multi-exit vectorization *without* requiring a scalar epilogue.
Split off as suggested from
https://github.com/llvm/llvm-project/pull/112138
Update the code to create induction resume PHIs to also create a resume
phi for the canonical induction during epilogue vectorization. This
unifies the code for handling induction resume values and removes the
need to explicitly create manually resume PHI and return it during
epilogue creation.
Overall it helps to move the code for updating the canonical induction
resume value to the place where all other header phi resume values are
updated.
This is NFC, modulo order of the created phis.
This patch fixes:
llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:2699:49: error:
captured structured bindings are a C++20 extension
[-Werror,-Wc++20-extensions]
When VF has a fixed width and equals the number of iterations, and we are not
tail folding by masking, comparison instruction and induction operation will be DCEed later.
Ignoring the costs of these instructions improves the cost model.
Don't use TCK_RecipThroughput independently in every VectorCombine fold.
Some prep work to allow a potential future patch to use VectorCombine to optimise for code size for -Os/Oz builds (setting TCK_CodeSize instead of TCK_RecipThroughput).
There's still more cleanup to do as a lot of get*Cost calls are relying on the default TargetCostKind value (usually TCK_RecipThroughput but not always).
This moves printing of the final VPlan to ::execute. This ensures the
final VPlan is printed, including recipes that get introduced by late,
lowering transforms and skeleton construction.
Split off from https://github.com/llvm/llvm-project/pull/114292, to
simplify the diff.
If the perfect diamond match was detected for the postponed buildvectors
and the vector for the previous node comes after the current node, need
to move the vector register before the current inserting point to
prevent compiler crash.
Fixes#119002
This reverts commit f09b16e2671cbcdf7cb7dc7ed705db092a9deda1.
The crash when building llvm-test-suite with stage2 should have been
fixed by 1091fad31a83d5ab87eb6fa11fe3bdb3f0d152ea.
This reverts commit 0678e2058364ec10b94560d27ec7138dfa003287.
This reverts commit 1091fad31a83d5ab87eb6fa11fe3bdb3f0d152ea.
Causes crashes in llvm-test-suite when using stage 2 clang.
getSExtValue assumes the result fits in 64 bits, but this may not be the
case for indcutions with wider types. Instead, directly perform the
compare on the APInt for the ConstantInt.
Fixes https://github.com/llvm/llvm-project/issues/118850.
Towards #118048
I was building llvm (clang and lld) for webassembly and came across
these warnings. Not sure if they are seen in our builds too.
```
/Users/anutosh491/work/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:6924:67: warning: comparison of integers of different signs: 'typename iterator_traits<user_iterator_impl<User>>::difference_type' (aka 'long') and 'unsigned int' [-Wsign-compare]
6924 | if (std::distance(LI->user_begin(), LI->user_end()) !=
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^
6925 | LI->getNumUses())
| ~~~~~~~~~~~~~~~~
[ 79%] Building CXX object lib/Transforms/Instrumentation/CMakeFiles/LLVMInstrumentation.dir/PGOInstrumentation.cpp.o
/Users/anutosh491/work/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:9754:43: warning: comparison of integers of different signs: 'typename iterator_traits<Value *const *>::difference_type' (aka 'long') and 'unsigned int' [-Wsign-compare]
9754 | count(Slice, Slice.front()) ==
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^
9755 | (isa<UndefValue>(Slice.front()) ? VF - 1 : 1)) {
```
This PR tries to address those warnings.
Updated ILV.createInductionResumeValues (now createInductionResumeVPValue)
to directly update the VPIRInstructions wrapping the original phis with the
created resume values.
This is the first step towards modeling them completely in VPlan.
Subsequent patches will move creation of the resume values completely
into VPlan.
Depends on https://github.com/llvm/llvm-project/pull/109975.
PR: https://github.com/llvm/llvm-project/pull/110577
The DAG maintains a chain of MemDGNodes that links together all the
nodes that may touch memroy.
Whenever a new instruction gets created we need to make sure that this
chain gets updated. If the new instruction touches memory then its
corresponding MemDGNode should be inserted into the chain.
To reduce repeated code, TreeEntry::setOperandsInOrder will be replaced
by VLOperands.
Arg_size will be provided to make sure other operands will not be
reorderd when VL[0] is IntrinsicInst (because APO is a boolean value).
In addition, BoUpSLP::reorderInputsAccordingToOpcode will also be
removed since it is simple.
Introduce a general recipe to generate a scalar phi. Lower
VPCanonicalIVPHIRecipe and VPEVLBasedIVRecipe to VPScalarIVPHIrecipe
before plan execution, avoiding the need for duplicated ::execute
implementations. There are other cases that could benefit, including
in-loop reduction phis and pointer induction phis.
Builds on a similar idea as
https://github.com/llvm/llvm-project/pull/82270.
PR: https://github.com/llvm/llvm-project/pull/114305
