|  | Commit message (Collapse) | Author | Age | Files | Lines | 
|---|
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary:
Avoid "count" if possible -> use "find" to check for the existence of keys.
Passed llvm test suite.
Reviewers: fhahn, dcaballe, mkuper, rengolin
Reviewed By: fhahn
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D51054
llvm-svn: 340563 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary:
Follow up change to rL339703, where we now vectorize loops with non-phi
instructions used outside the loop. Note that the cyclic dependency
identification occurs when identifying reduction/induction vars.
We also need to identify that we do not allow users where the PSCEV information
within and outside the loop are different. This was the fix added in rL307837
for PR33706.
Reviewers: Ayal, mkuper, fhahn
Subscribers: javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D50778
llvm-svn: 340278 | 
| | 
| 
| 
| 
| 
| 
| | Clarifying the comment about PSCEV and external IV users by referencing
the bug in question.
llvm-svn: 339722 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary:
This patch teaches the loop vectorizer to vectorize loops with non
header phis that have have outside uses.  This is because the iteration
dependence distance for these phis can be widened upto VF (similar to
how we do for induction/reduction) if they do not have a cyclic
dependence with header phis. When identifying reduction/induction/first
order recurrence header phis, we already identify if there are any cyclic
dependencies that prevents vectorization.
The vectorizer is taught to extract the last element from the vectorized
phi and update the scalar loop exit block phi to contain this extracted
element from the vector loop.
This patch can be extended to vectorize loops where instructions other
than phis have outside uses.
Reviewers: Ayal, mkuper, mssimpso, efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D50579
llvm-svn: 339703 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary:
Reworked the previously committed patch to insert shuffles for reused
extract element instructions in the correct position. Previous logic was
incorrect, and might lead to the crash with PHIs and EH instructions.
Reviewers: efriedma, javed.absar
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D50143
llvm-svn: 339166 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary:
If the ExtractElement instructions can be optimized out during the
vectorization and we need to reshuffle the parent vector, this
ShuffleInstruction may be inserted in the wrong place causing compiler
to produce incorrect code.
Reviewers: spatel, RKSimon, mkuper, hfinkel, javed.absar
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D49928
llvm-svn: 338380 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | The patch introduces loop analysis (VPLoopInfo/VPLoop) for VPBlockBases.
This analysis will be necessary to perform some H-CFG transformations and
detect and introduce regions representing a loop in the H-CFG.
Reviewers: fhahn, rengolin, mkuper, hfinkel, mssimpso
Reviewed By: fhahn 
Differential Revision: https://reviews.llvm.org/D48816
llvm-svn: 338346 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | The patch introduces dominator analysis for VPBlockBases and extend
VPlan's GraphTraits specialization with the required interfaces. Dominator
analysis will be necessary to perform some H-CFG transformations and
to introduce VPLoopInfo (LoopInfo analysis on top of the VPlan representation).
Reviewers: fhahn, rengolin, mkuper, hfinkel, mssimpso
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D48815
llvm-svn: 338310 | 
| | 
| 
| 
| 
| 
| | sed -Ei 's/[[:space:]]+$//' include/**/*.{def,h,td} lib/**/*.{cpp,h}
llvm-svn: 338293 | 
| | 
| 
| 
| 
| 
| | This reverts commit r338106.
llvm-svn: 338109 | 
| | 
| 
| 
| 
| 
| 
| 
| | Reviewers: hsaito
Differential Revision: https://reviews.llvm.org/D49746
llvm-svn: 338106 | 
| | 
| 
| 
| | llvm-svn: 337992 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | In some cases LSV sees (load/store _ (select _ <pointer expression>
<pointer expression>)) patterns in input IR, often due to sinking and
other forms of CFG simplification, sometimes interspersed with
bitcasts and all-constant-indices GEPs. With this
patch`areConsecutivePointers` method would attempt to handle select
instructions. This leads to an increased number of successful
vectorizations.
Technically, select instructions could appear in index arithmetic as
well, however, we don't see those in our test suites / benchmarks.
Also, there is a lot more freedom in IR shapes computing integral
indices in general than in what's common in pointer computations, and
it appears that it's quite unreliable to do anything short of making
select instructions first class citizens of Scalar Evolution, which
for the purposes of this patch is most definitely an overkill.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D49428
llvm-svn: 337965 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary: truncateToMinimalBitWidths() doesn't handle all Instructions and the worst case is compiler crash via llvm_unreachable(). Fix is to add a case to handle PHINode and changed the worst case to NO-OP (from compiler crash).
Reviewers: sbaranga, mssimpso, hsaito
Reviewed By: hsaito
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D49461
llvm-svn: 337861 | 
| | 
| 
| 
| 
| 
| 
| 
| | This reapplies commit r337489 reverted by r337541
Additionally, this commit contains a speculative fix to the issue reported in r337541
(the report does not contain an actionable reproducer, just a stack trace)
llvm-svn: 337606 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | This reverts commit r337489.
It causes asserts to fire in some TensorFlow tests, e.g.
tensorflow/compiler/tests/gather_test.py on GPU.
Example stack trace:
Start test case: GatherTest.testHigherRank
assertion failed at third_party/llvm/llvm/lib/Support/APInt.cpp:819 in llvm::APInt llvm::APInt::trunc(unsigned int) const: width && "Can't truncate to 0 bits"
    @     0x5559446ebe10  __assert_fail
    @     0x55593ef32f5e  llvm::APInt::trunc()
    @     0x55593d78f86e  (anonymous namespace)::Vectorizer::lookThroughComplexAddresses()
    @     0x55593d78f2bc  (anonymous namespace)::Vectorizer::areConsecutivePointers()
    @     0x55593d78d128  (anonymous namespace)::Vectorizer::isConsecutiveAccess()
    @     0x55593d78c926  (anonymous namespace)::Vectorizer::vectorizeInstructions()
    @     0x55593d78c221  (anonymous namespace)::Vectorizer::vectorizeChains()
    @     0x55593d78b948  (anonymous namespace)::Vectorizer::run()
    @     0x55593d78b725  (anonymous namespace)::LoadStoreVectorizer::runOnFunction()
    @     0x55593edf4b17  llvm::FPPassManager::runOnFunction()
    @     0x55593edf4e55  llvm::FPPassManager::runOnModule()
    @     0x55593edf563c  (anonymous namespace)::MPPassManager::runOnModule()
    @     0x55593edf5137  llvm::legacy::PassManagerImpl::run()
    @     0x55593edf5b71  llvm::legacy::PassManager::run()
    @     0x55593ced250d  xla::gpu::IrDumpingPassManager::run()
    @     0x55593ced5033  xla::gpu::(anonymous namespace)::EmitModuleToPTX()
    @     0x55593ced40ba  xla::gpu::(anonymous namespace)::CompileModuleToPtx()
    @     0x55593ced33d0  xla::gpu::CompileToPtx()
    @     0x55593b26b2a2  xla::gpu::NVPTXCompiler::RunBackend()
    @     0x55593b21f973  xla::Service::BuildExecutable()
    @     0x555938f44e64  xla::LocalService::CompileExecutable()
    @     0x555938f30a85  xla::LocalClient::Compile()
    @     0x555938de3c29  tensorflow::XlaCompilationCache::BuildExecutable()
    @     0x555938de4e9e  tensorflow::XlaCompilationCache::CompileImpl()
    @     0x555938de3da5  tensorflow::XlaCompilationCache::Compile()
    @     0x555938c5d962  tensorflow::XlaLocalLaunchBase::Compute()
    @     0x555938c68151  tensorflow::XlaDevice::Compute()
    @     0x55593f389e1f  tensorflow::(anonymous namespace)::ExecutorState::Process()
    @     0x55593f38a625  tensorflow::(anonymous namespace)::ExecutorState::ScheduleReady()::$_1::operator()()
*** SIGABRT received by PID 7798 (TID 7837) from PID 7798; ***
llvm-svn: 337541 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | This is mostly a preparation work for adding a limited support for
select instructions. It proved to be difficult to do due to size and
irregularity of Vectorizer::isConsecutiveAccess, this is fixed here I
believe.
It also turned out that these changes make it simpler to finish one of
the TODOs and fix a number of other small issues, namely:
1. Looking through bitcasts to a type of a different size (requires
careful tracking of the original load/store size and some math
converting sizes in bytes to expected differences in indices of GEPs).
2. Reusing partial analysis of pointers done by first attempt in proving
them consecutive instead of starting from scratch. This added limited
support for nested GEPs co-existing with difficult sext/zext
instructions. This also required a careful handling of negative
differences between constant parts of offsets.
3. Handing a case where the first pointer index is not an add, but
something else (a function parameter for instance).
I observe an increased number of successful vectorizations on a large
set of shader programs. Only few shaders are affected, but those that
are affected sport >5% less loads and stores than before the patch.
Reviewed By: rampitec
Differential-Revision: https://reviews.llvm.org/D49342
llvm-svn: 337489 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | pointers.
Summary: Currently, isConsecutiveAccess() detects two pointers(PtrA and PtrB) as consecutive by
         comparing PtrB with BaseDelta+PtrA. This works when both pointers are factorized or
         both of them are not factorized. But isConsecutiveAccess() fails if one of the
         pointers is factorized but the other one is not.
         Here is an example:
         PtrA = 4 * (A + B)
         PtrB = 4 + 4A + 4B
         This patch uses getMinusSCEV() to compute the distance between two pointers.
         getMinusSCEV() allows combining the expressions and computing the simplified distance.
Author: FarhanaAleen
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D49516
llvm-svn: 337471 | 
| | 
| 
| 
| 
| 
| 
| 
| | BoUpSLP::getEntryCost. NFCI.
Pulled out from D49225, we have a lot of repeated scalar cost calculations, often with arguments that don't look the same but turn out to be.
llvm-svn: 337390 | 
| | 
| 
| 
| 
| 
| | TTI::getMinMaxReductionCost typically can't handle pointer types - until this is changed its better to limit horizontal reduction to integer/float vector types only.
llvm-svn: 337280 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | (REAPPLIED-2)
We currently only support binary instructions in the alternate opcode shuffles.
This patch is an initial attempt at adding cast instructions as well, this raises several issues that we probably want to address as we continue to generalize the alternate mechanism:
1 - Duplication of cost determination - we should probably add scalar/vector costs helper functions and get BoUpSLP::getEntryCost to use them instead of determining costs directly.
2 - Support alternate instructions with the same opcode (e.g. casts with different src types) - alternate vectorization of calls with different IntrinsicIDs will require this.
3 - Allow alternates to be a different instruction type - mixing binary/cast/call etc.
4 - Allow passthrough of unsupported alternate instructions - related to PR30787/D28907 'copyable' elements.
Reapplied with fix to only accept 2 different casts if they come from the same source type (PR38154).
Differential Revision: https://reviews.llvm.org/D49135
llvm-svn: 336989 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| | instructions. (REAPPLIED)"
This reverts commit r336812, which broke compilation of a number
of projects, see PR38154.
llvm-svn: 336949 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | (REAPPLIED)
We currently only support binary instructions in the alternate opcode shuffles.
This patch is an initial attempt at adding cast instructions as well, this raises several issues that we probably want to address as we continue to generalize the alternate mechanism:
1 - Duplication of cost determination - we should probably add scalar/vector costs helper functions and get BoUpSLP::getEntryCost to use them instead of determining costs directly.
2 - Support alternate instructions with the same opcode (e.g. casts with different src types) - alternate vectorization of calls with different IntrinsicIDs will require this.
3 - Allow alternates to be a different instruction type - mixing binary/cast/call etc.
4 - Allow passthrough of unsupported alternate instructions - related to PR30787/D28907 'copyable' elements.
Reapplied with fix to only accept 2 different casts if they come from the same source type.
Differential Revision: https://reviews.llvm.org/D49135
llvm-svn: 336812 | 
| | 
| 
| 
| 
| 
| 
| 
| | cast instructions.
Reverting due to buildbot failures
llvm-svn: 336806 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | We currently only support binary instructions in the alternate opcode shuffles.
This patch is an initial attempt at adding cast instructions as well, this raises several issues that we probably want to address as we continue to generalize the alternate mechanism:
1 - Duplication of cost determination - we should probably add scalar/vector costs helper functions and get BoUpSLP::getEntryCost to use them instead of determining costs directly.
2 - Support alternate instructions with the same opcode (e.g. casts with different src types) - alternate vectorization of calls with different IntrinsicIDs will require this.
3 - Allow alternates to be a different instruction type - mixing binary/cast/call etc.
4 - Allow passthrough of unsupported alternate instructions - related to PR30787/D28907 'copyable' elements.
Differential Revision: https://reviews.llvm.org/D49135
llvm-svn: 336804 | 
| | 
| 
| 
| 
| 
| | Differential Revision: https://reviews.llvm.org/D48968
llvm-svn: 336667 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | This patch introduces a VPValue in VPBlockBase to represent the condition
bit that is used as successor selector when a block has multiple successors.
This information wasn't necessary until now, when we are about to introduce
outer loop vectorization support in VPlan code gen.
Reviewers: fhahn, rengolin, mkuper, hfinkel, mssimpso
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D48814
llvm-svn: 336554 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | from opcodes. NFCI.
This is an early step towards matching Instructions by attributes other than the opcode. This will be necessary for cast/call alternates which share the same opcode but have different types/intrinsicIDs etc. - which we could vectorize as long as we split them using the alternate mechanism.
Differential Revision: https://reviews.llvm.org/D48945
llvm-svn: 336344 | 
| | 
| 
| 
| | llvm-svn: 336291 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| | When creating `phi` instructions to resume at the scalar part of the loop,
copy the DebugLoc from the original phi over to the new one.
Differential Revision: https://reviews.llvm.org/D48769
llvm-svn: 336256 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | Summary: It is common to have the following min/max pattern during the intermediate stages of SLP since we only optimize at the end. This patch tries to catch such patterns and allow more vectorization.
         %1 = extractelement <2 x i32> %a, i32 0
         %2 = extractelement <2 x i32> %a, i32 1
         %cond = icmp sgt i32 %1, %2
         %3 = extractelement <2 x i32> %a, i32 0
         %4 = extractelement <2 x i32> %a, i32 1
         %select = select i1 %cond, i32 %3, i32 %4
Author: FarhanaAleen
Reviewed By: ABataev, RKSimon, spatel
Differential Revision: https://reviews.llvm.org/D47608
llvm-svn: 336130 | 
| | 
| 
| 
| 
| 
| 
| 
| | getEntryCost
This code is only used by alternate opcodes so the InstructionsState has already confirmed that every Value is an Instruction, plus we use cast<Instruction> which will assert on failure.
llvm-svn: 336102 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | handle SK_Select patterns.
We were always using the opcodes of the first 2 scalars for the costs of the alternate opcode + shuffle. This made sense when we used SK_Alternate and opcodes were guaranteed to be alternating, but this fails for the more general SK_Select case.
This fix exposes an issue demonstrated by the fmul_fdiv_v4f32_const test - the SLM model has v4f32 fdiv costs which are more than twice those of the f32 scalar cost, meaning that the cost model determines that the vectorization is not performant. Unfortunately it completely ignores the fact that the fdiv by a constant will be changed into a fmul by InstCombine for a much lower cost vectorization. But at least we're seeing this now...
llvm-svn: 336095 | 
| | 
| 
| 
| 
| 
| 
| 
| | getEntryCost/vectorizeTree. NFCI.
Add assertions - we're already assuming this in how we use the AltOpcode and treat everything as BinaryOperators.
llvm-svn: 336092 | 
| | 
| 
| 
| 
| 
| | instead of an opcode. NFCI.
llvm-svn: 336069 | 
| | 
| 
| 
| 
| 
| 
| 
| | helper. NFCI.
This is a basic step towards matching more general instructions types than just opcodes.
llvm-svn: 336068 | 
| | 
| 
| 
| | llvm-svn: 336063 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | Since D46637 we are better at handling uniform/non-uniform constant Pow2 detection; this patch tweaks the SLP argument handling to support them.
As SLP works with arrays of values I don't think we can easily use the pattern match helpers here.
Differential Revision: https://reviews.llvm.org/D48214
llvm-svn: 335621 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | Enable tryToVectorizeList to support InstructionsState alternate opcode patterns at a root (build vector etc.) as well as further down the vectorization tree.
NOTE: This patch reduces some of the debug reporting if there are opcode mismatches - I can try to add it back if it proves a problem. But it could get rather messy trying to provide equivalent verbose debug strings via getSameOpcode etc.
Differential Revision: https://reviews.llvm.org/D48488
llvm-svn: 335364 | 
| | 
| 
| 
| 
| 
| 
| 
| | InstructionsState. NFCI.
All calls were extracting the InstructionsState Opcode/AltOpcode values so we might as well pass it directly
llvm-svn: 335359 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | SLP currently only accepts (F)Add/(F)Sub alternate counterpart ops to be merged into an alternate shuffle.
This patch relaxes this to accept any pair of BinaryOperator opcodes instead, assuming the target's cost model accepts the vectorization+shuffle.
Differential Revision: https://reviews.llvm.org/D48477
llvm-svn: 335349 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | call tree
As described in D48359, this patch pushes InstructionsState down the BoUpSLP call hierarchy instead of the corresponding raw OpValue. This makes it easier to track the alternate opcode etc. and avoids us having to call getAltOpcode which makes it difficult to support more than one alternate opcode.
Differential Revision: https://reviews.llvm.org/D48382
llvm-svn: 335170 | 
| | 
| 
| 
| 
| 
| | A future patch will have isOneOf use InstructionsState.
llvm-svn: 335142 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | This is part of a move towards generalizing the alternate opcode mechanism and not just supporting (F)Add/(F)Sub counterparts.
The patch embeds the AltOpcode in the InstructionsState instead of calling getAltOpcode so often.
I'm hoping to eventually remove all uses of getAltOpcode and handle alternate opcode selection entirely within getSameOpcode, that will require us to use InstructionsState throughout the BoUpSLP call hierarchy (similar to some of the changes in D28907), which I will begin in future patches.
Differential Revision: https://reviews.llvm.org/D48359
llvm-svn: 335134 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | SK_Select shuffle pattern
D47985 saw the old SK_Alternate 'alternating' shuffle mask replaced with the SK_Select mask which accepts either input operand for each lane, equivalent to a vector select with a constant condition operand.
This patch updates SLPVectorizer to make full use of this SK_Select shuffle pattern by removing the 'isOdd()' limitation.
The AArch64 regression will be fixed by D48172.
Differential Revision: https://reviews.llvm.org/D48174
llvm-svn: 335130 | 
| | 
| 
| 
| | llvm-svn: 335110 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | getArithmeticInstrCost calls (NFC)
The getArithmeticInstrCost calls for shuffle vectors entry costs specify TargetTransformInfo::OperandValueKind arguments, but are just using the method's default values. This seems to be a copy + paste issue and doesn't affect the costs in anyway. The TargetTransformInfo::OperandValueProperties default arguments are already not being used.
Noticed while working on D47985.
Differential Revision: https://reviews.llvm.org/D48008
llvm-svn: 335045 | 
| | 
| 
| 
| 
| 
| | Minor step towards making the alternate opcode system work with a wider range of opcode pairs.
llvm-svn: 335032 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| 
| | This patch introduces a VPInstructionToVPRecipe transformation, which
allows us to generate code for a VPInstruction based VPlan re-using the
existing infrastructure.
Reviewers: dcaballe, hsaito, mssimpso, hfinkel, rengolin, mkuper, javed.absar, sguggill
Reviewed By: dcaballe
Differential Revision: https://reviews.llvm.org/D46827
llvm-svn: 334969 | 
| | 
| 
| 
| 
| 
| 
| 
| 
| 
| | Ensure we keep track of the input vectors in all cases instead of just for SK_Select.
Ideally we'd reuse the shuffle mask pattern matching in TargetTransformInfo::getInstructionThroughput here to easily add support for all TargetTransformInfo::ShuffleKind without mass code duplication, I've added a TODO for now but D48236 should help us here.
Differential Revision: https://reviews.llvm.org/D48023
llvm-svn: 334958 |