If OpValue is non-null, we only consider operations similar to OpValue
when intersecting.
Differential Revision: https://reviews.llvm.org/D35292
llvm-svn: 308428
Summary:
When simplifying unconditional branches from empty blocks, we pre-test if the
BB belongs to a set of loop headers and keep the block to prevent passes from
destroying canonical loop structure. However, the current algorithm fails if
the destination of the branch is a loop header. Especially when such a loop's
latch block is folded into loop header it results in additional backedges and
LoopSimplify turns it into a nested loop which prevent later optimizations
from being applied (e.g., loop unrolling and loop interleaving).
This patch augments the existing algorithm by further checking if the
destination of the branch belongs to a set of loop headers and defer
eliminating it if yes to LateSimplifyCFG.
Fixes PR33605: https://bugs.llvm.org/show_bug.cgi?id=33605
Reviewers: efriedma, mcrosier, pacxx, hsung, davidxl
Reviewed By: efriedma
Subscribers: ashutosh.nema, gberry, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D35411
llvm-svn: 308422
Generate a single test to decide if there are enough iterations to jump to the
vectorized loop, or else go to the scalar remainder loop. This test compares the
Scalar Trip Count: if STC < VF * UF go to the scalar loop. If
requiresScalarEpilogue() holds, at-least one iteration must remain scalar; the
rest can be used to form vector iterations. So in this case the test checks
instead if (STC - 1) < VF * UF by comparing STC <= VF * UF, and going to the
scalar loop if so. Otherwise the vector loop is entered for at-least one vector
iteration.
This test covers the case where incrementing the backedge-taken count will
overflow leading to an incorrect trip count of zero. In this (rare) case we will
also avoid the vector loop and jump to the scalar loop.
This patch simplifies the existing tests and effectively removes the basic-block
originally named "min.iters.checked", leaving the single test in block
"vector.ph".
Original observation and initial patch by Evgeny Stupachenko.
Differential Revision: https://reviews.llvm.org/D34150
llvm-svn: 308421
functions.
In the prior commit, we provide ordering to the LCG between functions
and library function definitions that they might begin to call through
transformations. But we still would delete these library functions from
the call graph if they became dead during inlining.
While this immediately crashed, it also exposed a loss of information.
We shouldn't remove definitions of library functions that can still
usefully participate in the LCG-powered CGSCC optimization process. If
new call edges are formed, we want to have definitions to be called.
We can still remove these functions if truly dead using global-dce, etc,
but removing them during the CGSCC walk is premature.
This fixes a crash in the new PM when optimizing some unusual libraries
that end up with "internal" lib functions such as the code in the "R"
language's libraries.
llvm-svn: 308417
Summary: Currently, when GVN creates a load and when InstCombine creates a new store for unreachable Load, the DebugLoc info gets lost.
Reviewers: dberlin, davide, aprantl
Reviewed By: aprantl
Subscribers: davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D34639
llvm-svn: 308404
Summary:
ASan determines the stack layout from alloca instructions. Since
arguments marked as "byval" do not have an explicit alloca instruction, ASan
does not produce red zones for them. This commit produces an explicit alloca
instruction and copies the byval argument into the allocated memory so that red
zones are produced.
Submitted on behalf of @morehouse (Matt Morehouse)
Reviewers: eugenis, vitalybuka
Reviewed By: eugenis
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D34789
llvm-svn: 308387
Coverage hooks that take less-than-64-bit-integers as parameters need the
zeroext parameter attribute (http://llvm.org/docs/LangRef.html#paramattrs)
to make sure they are properly extended by the x86_64 ABI.
llvm-svn: 308296
In some particular cases eq/ne conditions can be turned into equivalent
slt/sgt conditions. This patch teaches parseLoopStructure to handle some
of these cases.
Differential Revision: https://reviews.llvm.org/D35010
llvm-svn: 308264
Rename the enum value from X86_64_Win64 to plain Win64.
The symbol exposed in the textual IR is changed from 'x86_64_win64cc'
to 'win64cc', but the numeric value is kept, keeping support for
old bitcode.
Differential Revision: https://reviews.llvm.org/D34474
llvm-svn: 308208
This reverts commit r308114 (and follow on fixes to test).
There is a linking failure in a ThinLTO bot:
http://green.lab.llvm.org/green/job/clang-stage2-configure-Rthinlto_build/3663/
(and undefined reference). It seems like it must be a second order
effect of the heuristic change I made, and may take some time to try
to reproduce locally and track down. Therefore, reverting for now.
llvm-svn: 308206
This restores r308078/r308079 with a fix for bot non-determinisim (make
sure we run llvm-lto in single threaded mode so the debug output doesn't get
interleaved).
llvm-svn: 308114
Summary:
If one side simplifies to the identity value for inner opcode, we can replace the value with just the operation that can't be simplified.
I've removed a couple now unneeded special cases in visitAnd and visitOr. There are probably other cases I missed.
Reviewers: spatel, majnemer, hfinkel, dberlin
Reviewed By: spatel
Subscribers: grandinj, llvm-commits, spatel
Differential Revision: https://reviews.llvm.org/D35451
llvm-svn: 308111
that appears to exhibit non-determinism and is flaking on the bots
pretty consistently.
r308078: [ThinLTO] Ensure we always select the same function copy to import
r308079: Require asserts in new test that uses debug flag
llvm-svn: 308095
Summary:
Check if the first eligible callee is under the instruction threshold.
Checking this on the first eligible callee ensures that we don't end
up selecting different callees to import when we invoke this routine
with different thresholds due to reaching the callee via paths that
are shallower or hotter (when there are multiple copies, i.e. with
weak or linkonce linkage). We don't want to leave the decision of which
copy to import up to the backend.
Reviewers: mehdi_amini
Subscribers: inglorion, fhahn, llvm-commits
Differential Revision: https://reviews.llvm.org/D35436
llvm-svn: 308078
Summary:
When checking for memory dependencies between calls using MemorySSA,
handle cases where the calls have no MemoryAccess associated with them
because the AA analysis being used has determined that the call does not
read/write memory.
Fixes PR33756
Reviewers: dberlin, davide
Subscribers: mcrosier, llvm-commits, Prazek
Differential Revision: https://reviews.llvm.org/D35317
llvm-svn: 308051
Add the following pattern to TryToUnfoldSelectInCurrBB()
bb:
%p = phi [0, %bb1], [1, %bb2], [0, %bb3], [1, %bb4], ...
%c = cmp %p, 0
%s = select %c, trueval, falseval
The Select in the above pattern will be unfolded and then jump-threaded. The
current implementation does not allow CMP in the middle of PHI and Select.
Differential Revision: https://reviews.llvm.org/D34762
llvm-svn: 308050
Summary:
DominatorTreeBase used to have IsPostDominators (bool) member to indicate if the tree is a dominator or a postdominator tree. This made it possible to switch between the two 'modes' at runtime, but it isn't used in practice anywhere.
This patch makes IsPostDominator a template argument. This way, it is easier to switch between different algorithms at compile-time based on this argument and design external utilities around it. It also makes it impossible to incidentally assign a postdominator tree to a dominator tree (and vice versa), and to further simplify template code in GenericDominatorTreeConstruction.
Reviewers: dberlin, sanjoy, davide, grosser
Reviewed By: dberlin
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D35315
llvm-svn: 308040
When iterating through loop
for (int i = INT_MAX; i > 0; i--)
We fail to generate the pre-loop for it. It happens because we use the
overflown value in a comparison predicate when identifying whether or not
we need it.
In old logic, we used SLE predicate against Greatest value which exceeds all
seen values of the IV and might be overflown. Now we use the GreatestSeen
value of this IV with SLT predicate.
Also added a test that ensures that a pre-loop is generated for such loops.
Differential Revision: https://reviews.llvm.org/D35347
llvm-svn: 308001
Summary:
When we runtime unroll with multiple exit blocks, we also need to update the
immediate dominators of the immediate successors of the exit blocks.
Reviewers: reames, mkuper, mzolotukhin, apilipenko
Reviewed by: mzolotukhin
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D35304
llvm-svn: 307909
This is an incremental change to the promotion feature.
There are two problems with the current behavior:
1) loops with multiple exiting blocks are totally disabled
2) a counter update can only be promoted one level up in
the loop nest -- which does help much for short trip
count inner loops inside a high trip-count outer loops.
Due to this limitation, we still saw very large profile
count fluctuations from run to run for the affected loops
which are usually very hot.
This patch adds the support for promotion counters iteratively
across the loop nest. It also turns on the promotion for
loops with multiple exiting blocks (with a limit).
For single-threaded applications, the performance impact is flat
on average. For instance, dealII improves, but povray regresses.
llvm-svn: 307863
Refactored the code and separated out a function
`canSafelyUnrollMultiExitLoop` to reduce redundant checks and make it
easier to add profitability heuristics later.
Added tests to runtime unrolling to make sure that unrolling for
multi-exit loops is not done unless the option
-unroll-runtime-multi-exit is true.
llvm-svn: 307843
Where is is needed (at the end of headers that define it), be
consistent about its use.
Also fix a few header guards that I found in the process.
Differential Revision: https://reviews.llvm.org/D34916
llvm-svn: 307840
Summary:
LoopRotate manually updates the DoomTree by iterating over all predecessors of a basic block and computing the Nearest Common Dominator.
When a predecessor happens to be unreachable, `DT.findNearestCommonDominator` returns nullptr.
This patch teaches LoopRotate to handle this case and fixes [[ https://bugs.llvm.org/show_bug.cgi?id=33701 | PR33701 ]].
In the future, LoopRotate should be taught to use the new incremental API for updating the DomTree.
Reviewers: dberlin, davide, uabelho, grosser
Subscribers: efriedma, mzolotukhin
Differential Revision: https://reviews.llvm.org/D35074
llvm-svn: 307828
This normally indicates mixed CFI + non-CFI compilation, and will
result in us treating the function in the same way as a function
defined outside of the LTO unit.
Part of PR33752.
Differential Revision: https://reviews.llvm.org/D35281
llvm-svn: 307744
[GlobalOpt] Remove unreachable blocks before optimizing a function.
While the change is presumably correct, it exposes a latent bug
in DI which breaks on of the CFI checks. I'll analyze it further
and try to understand what's going on.
llvm-svn: 307729
OpenCL 2.0 introduces the notion of memory scopes in atomic operations to
global and local memory. These scopes restrict how synchronization is
achieved, which can result in improved performance.
This change extends existing notion of synchronization scopes in LLVM to
support arbitrary scopes expressed as target-specific strings, in addition to
the already defined scopes (single thread, system).
The LLVM IR and MIR syntax for expressing synchronization scopes has changed
to use *syncscope("<scope>")*, where <scope> can be "singlethread" (this
replaces *singlethread* keyword), or a target-specific name. As before, if
the scope is not specified, it defaults to CrossThread/System scope.
Implementation details:
- Mapping from synchronization scope name/string to synchronization scope id
is stored in LLVM context;
- CrossThread/System and SingleThread scopes are pre-defined to efficiently
check for known scopes without comparing strings;
- Synchronization scope names are stored in SYNC_SCOPE_NAMES_BLOCK in
the bitcode.
Differential Revision: https://reviews.llvm.org/D21723
llvm-svn: 307722
This is fine as nothing in the code relies on leader and memory
leader being the same for a given congruency class. Ack'ed by
Dan.
Fixes PR33720.
llvm-svn: 307699
The loop structure for the outer loop does not contain the epilog
preheader when we try to unroll inner loop with multiple exits and
epilog code is generated. For now, we just bail out in such cases.
Added a test case that shows the problem. Without this bailout, we would
trip on assert saying LCSSA form is incorrect for outer loop.
llvm-svn: 307676
querying for analysis results on a function declaration rather than
a definition.
The only reason this worked previously is by chance -- because the way
we got alias analysis results with the legacy PM, we happened to not
compute a dominator tree and so we happened to not hit an assert even
though it didn't make any real sense. Now we bail out before trying to
compute alias analysis so that we don't hit these asserts.
llvm-svn: 307625
Summary:
As metioned in https://reviews.llvm.org/D34576, checkings in
`collectConstantCandidates` can be replaced by using
`llvm::canReplaceOperandWithVariable`.
The only special case is that `collectConstantCandidates` return false for
all `IntrinsicInst` but it is safe for us to collect constant candidates from
`IntrinsicInst`.
Reviewers: pirama, efriedma, srhines
Reviewed By: efriedma
Subscribers: llvm-commits, javed.absar
Differential Revision: https://reviews.llvm.org/D34921
llvm-svn: 307587
When unrolling under multiple exits which is under off-by-default option,
the assert that checks for VMap entry in loop exit values is too strong.
(assert if VMap entry did not exist, the value should be a
constant). However, values derived from
constants or from values outside loop, does not have a VMap entry too.
Removed the assert and added a testcase showcasing the property for
non-constant values.
llvm-svn: 307542
Summary:
This solves PR33641.
When removing a dead argument we must also handle possibly existing calls
to llvm.dbg.value that use the removed argument. Now we change the use
of the otherwise dead argument to an undef for some other pass to cleanup
later.
If the calls are left untouched, they will later on cause errors:
"function-local metadata used in wrong function"
since the ArgumentPromotion rewrites the code by creating a new function
with the wanted signature, but the metadata is not recreated so the new
function may then erroneously use metadata from the old function.
Reviewers: mstorsjo, rnk, arsenm
Reviewed By: rnk
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D34874
llvm-svn: 307521
the invalidation propagation logic from an SCC to a Function.
I wrote the infrastructure to test this but didn't actually use it in
the unit test where it was designed to be used. =[ My bad. Once
I actually added it to the test case I discovered that it also hadn't
been properly implemented, so I've implemented it. The logic in the FAM
proxy for an SCC pass to propagate invalidation follows the same ideas
as the FAM proxy for a Module pass, but the implementation is a bit
different to reflect the fact that it is forwarding just for an SCC.
However, implementing this correctly uncovered a surprising "bug" (it
was conservatively correct but relatively very expensive) in how we
handle invalidation when splitting one SCC into multiple SCCs. We did an
eager invalidation when in reality we should be deferring invaliadtion
for the *current* SCC to the CGSCC pass manager and just invaliating the
newly constructed SCCs. Otherwise we end up invalidating too much too
soon. This was exposed by the inliner test case that I've updated. Now,
we invalidate *just* the split off '(test1_f)' SCC when doing the CG
update, and then the inliner finishes and invalidates the '(test1_g,
test1_h)' SCC's analyses. The first few attempts at fixing this hit
still more bugs, but all of those are covered by existing tests. For
example, the inliner should also preserve the FAM proxy to avoid
unnecesasry invalidation, and this is safe because the CG update
routines it uses handle any necessary adjustments to the FAM proxy.
Finally, the unittests for the CGSCC pass manager needed a bunch of
updates where we weren't correctly preserving the FAM proxy because it
hadn't been fully implemented and failing to preserve it didn't matter.
Note that this doesn't yet fix the current crasher due to MemSSA finding
a stale dominator tree, but without this the fix to that crasher doesn't
really make any sense when testing because it relies on the proxy
behavior.
llvm-svn: 307487
I recently changed m_One and m_AllOnes to use Constant::isOneValue/isAllOnesValue which work on floating point values too. The original implementation looked specifically for ConstantInt scalars and splats. So I'm guessing we are accidentally trying to issue sext/zexts on floating point types now.
Hopefully I figure out how to reproduce the failure from the PR soon.
llvm-svn: 307486
The patch was reverted due to a bug. The bug was that if the IV is the 2nd operand of the icmp
instruction, then the "Pred" variable gets swapped and differs from the instruction's predicate.
In this patch we use the original predicate to do the transformation.
Also added a test case that exercises this situation.
Differentian Revision: https://reviews.llvm.org/D35107
llvm-svn: 307477
Previously the InstCombiner class contained a pointer to an IR builder that had been passed to the constructor. Sometimes this would be passed to helper functions as either a pointer or the pointer would be dereferenced to be passed by reference.
This patch makes it a reference everywhere including the InstCombiner class itself so there is more inconsistency. This a large, but mechanical patch. I've done very minimal formatting changes on it despite what clang-format wanted to do.
llvm-svn: 307451
Summary: For interative sample-pgo, if a hot call site is inlined in the profiling binary, we should inline it in before profile annotation in the backend. Before that, the compile phase first collects all GUIDs that needs to be imported and creates virtual "hot" call edge in the summary. However, "hot" is not good enough to guarantee the callsites get inlined. This patch introduces "critical" call edge, and assign much higher importing threshold for those edges.
Reviewers: tejohnson
Reviewed By: tejohnson
Subscribers: sanjoy, mehdi_amini, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D35096
llvm-svn: 307439
With the NFC refactoring in rL307417 (git SHA 987dd01), all the logic
is in place to support multiple exit/exiting blocks when prolog
remainder is generated.
This patch removed the assert that multiple exit blocks unrolling is only
supported when epilog remainder is generated.
Also, added test runs and checks with PROLOG prefix in
runtime-loop-multiple-exits.ll test cases.
llvm-svn: 307435
Summary:
This is an addon to the change rl304488 cloning fixes. (Originally rl304226 reverted rl304228 and reapplied rl304488 https://reviews.llvm.org/D33655)
rl304488 works great when DILocalVariables that comes from the inlined function has a 'unique-ed' type, but,
in the case when the variable type is distinct we will create a second DILocalVariable in the scope of the original function that was inlined.
Consider cloning of the following function:
```
define private void @f() !dbg !5 {
%1 = alloca i32, !dbg !11
call void @llvm.dbg.declare(metadata i32* %1, metadata !14, metadata !12), !dbg !18
ret void, !dbg !18
}
!14 = !DILocalVariable(name: "inlined", scope: !15, file: !6, line: 5, type: !17) ; came from an inlined function
!15 = distinct !DISubprogram(name: "inlined", linkageName: "inlined", scope: null, file: !6, line: 8, type: !7, isLocal: true, isDefinition: true, scopeLine: 9, isOptimized: false, unit: !0, variables: !16)
!16 = !{!14}
!17 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "some_struct", size: 32, align: 32)
```
Without this fix, when function 'f' is cloned, we will create another DILocalVariable for "inlined", due to its type being distinct.
```
define private void @f.1() !dbg !23 {
%1 = alloca i32, !dbg !26
call void @llvm.dbg.declare(metadata i32* %1, metadata !28, metadata !12), !dbg !30
ret void, !dbg !30
}
!14 = !DILocalVariable(name: "inlined", scope: !15, file: !6, line: 5, type: !17)
!15 = distinct !DISubprogram(name: "inlined", linkageName: "inlined", scope: null, file: !6, line: 8, type: !7, isLocal: true, isDefinition: true, scopeLine: 9, isOptimized: false, unit: !0, variables: !16)
!16 = !{!14}
!17 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "some_struct", size: 32, align: 32)
;
!28 = !DILocalVariable(name: "inlined", scope: !15, file: !6, line: 5, type: !29) ; OOPS second DILocalVariable
!29 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "some_struct", size: 32, align: 32)
```
Now we have two DILocalVariable for "inlined" within the same scope. This result in assert in AsmPrinter/DwarfDebug.h:131: void llvm::DbgVariable::addMMIEntry(const llvm::DbgVariable &): Assertion `V.Var == Var && "conflicting variable"' failed.
(Full example: See: https://bugs.llvm.org/show_bug.cgi?id=33492)
In this change we prevent duplication of types so that when a metadata for DILocalVariable is cloned it will get uniqued to the same metadate node as an original variable.
Reviewers: loladiro, dblaikie, aprantl, echristo
Reviewed By: loladiro
Subscribers: EricWF, llvm-commits
Differential Revision: https://reviews.llvm.org/D35106
llvm-svn: 307418
Minor refactoring to use the preexisting loop exit that's already
calculated. We do not need to recompute the loop exit in ConnectProlog.
Apart from avoiding redundant computation, this is required for
supporting multiple loop exits when Prolog remainder loops are generated.
llvm-svn: 307417
InferAddressSpaces does not check address space in collectFlatAddressExpressions,
which causes values with non flat address space put into Postorder and causes
assertion in cloneValueWithNewAddressSpace.
This patch fixes assertion in OpenCL 2.0 conformance test generic_address_space
subtest for amdgcn target.
Differential Revision: https://reviews.llvm.org/D34991
llvm-svn: 307349
Adds loop expansions for known-size and unknown-sized memcpy calls, allowing the
target to provide the operand types through TTI callbacks. The default values
for the TTI callbacks use int8 operand types and matches the existing behaviour
if they aren't overridden by the target.
Differential revision: https://reviews.llvm.org/D32536
llvm-svn: 307346
Revert "Copy arguments passed by value into explicit allocas for ASan."
Revert "[asan] Add end-to-end tests for overflows of byval arguments."
Build failure on lldb-x86_64-ubuntu-14.04-buildserver.
Test failure on clang-cmake-aarch64-42vma and sanitizer-x86_64-linux-android.
llvm-svn: 307345
ASan determines the stack layout from alloca instructions. Since
arguments marked as "byval" do not have an explicit alloca instruction, ASan
does not produce red zones for them. This commit produces an explicit alloca
instruction and copies the byval argument into the allocated memory so that red
zones are produced.
Patch by Matt Morehouse.
Differential revision: https://reviews.llvm.org/D34789
llvm-svn: 307342
Using profile information to guide consthoisting is generally helpful for
performance, so the patch turns it on by default. No compile time or perf
regression were found using spec2000 and spec2006 on x86. Some significant
improvement (>20%) was seen on internal benchmarks.
Differential Revision: https://reviews.llvm.org/D35063
llvm-svn: 307338
The patch is to adjust the strategy of frequency based consthoisting:
Previously when the candidate block has the same frequency with the existing
blocks containing a const, it will not hoist the const to the candidate block.
For that case, now we change the strategy to hoist the const if only existing
blocks have more than one block member. This is helpful for reducing code size.
Differential Revision: https://reviews.llvm.org/D35084
llvm-svn: 307328
This is the same as r304719 but for ThinLTO.
The substantial difference is that in this case we don't have
whole visibility, just the summary.
In the LTO case, when we got the resolution for the input file we
could just see if the linker told us whether a symbol was linker
redefined (using --wrap or --defsym) and switch the linkage directly
for the GV.
Here, we have the summary. So, we record that the linkage changed
from <whatever it was> to $weakany to prevent IPOs across this symbol
boundaries and actually just switch the linkage at FunctionImport time.
This patch should also fixes the lld bits (as all the scaffolding for
communicating if a symbol is linker redefined should be there & should
be the same), but I'll make sure to add some tests there as well.
Fixes PR33192.
Differential Revision: https://reviews.llvm.org/D35064
llvm-svn: 307303
Summary:
`Instruction::Switch`: only first operand can be set to a non-constant value.
`Instruction::InsertValue` both the first and the second operand can be set to a non-constant value.
`Instruction::Alloca` return true for non-static allocation.
Reviewers: efriedma
Reviewed By: efriedma
Subscribers: srhines, pirama, llvm-commits
Differential Revision: https://reviews.llvm.org/D34905
llvm-svn: 307294
Going through the Constant methods requires redetermining that the Constant is a ConstantInt and then calling isZero/isOne/isMinusOne.
llvm-svn: 307292
Currently, we do not support multiple exiting blocks to the
latch exit block. However, this bailout wasn't triggered when we had a
unique exit block (which is the latch exit), with multiple exiting
blocks to that unique exit.
Moved the bailout so that it's triggered in both cases and added
testcase.
llvm-svn: 307291
Summary: In this code we got to Dom by following the predecessor link of BB. So it stands to reason that BB should also show up as a successor of Dom's terminator right? There isn't a way to have the CFG connect in only one direction is there?
Reviewers: jmolloy, davide, mcrosier
Reviewed By: mcrosier
Subscribers: mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D35025
llvm-svn: 307276
Bswap isn't a simple operation so we need to make sure we are really removing a call to it before doing these simplifications.
For the case when both LHS and RHS are bswaps I've allowed it to be moved if either LHS or RHS has a single use since that at least allows us to move it later where it might find another bswap to combine with and it decreases the use count on the other side so maybe the other user can be optimized.
Differential Revision: https://reviews.llvm.org/D34974
llvm-svn: 307273
When the formulae search space is huge, LSR uses a series of heuristic to keep
pruning the search space until the number of possible solutions are within
certain limit.
The big hammer of the series of heuristics is NarrowSearchSpaceByPickingWinnerRegs,
which picks the register which is used by the most LSRUses and deletes the other
formulae which don't use the register. This is a effective way to prune the search
space, but quite often not a good way to keep the best solution. We saw cases before
that the heuristic pruned the best formula candidate out of search space.
To relieve the problem, we introduce a new heuristic called
NarrowSearchSpaceByFilterFormulaWithSameScaledReg. The basic idea is in order to
reduce the search space while keeping the best formula, we want to keep as many
formulae with different Scale and ScaledReg as possible. That is because the central
idea of LSR is to choose a group of loop induction variables and use those induction
variables to represent LSRUses. An induction variable candidate is often represented
by the Scale and ScaledReg in a formula. If we have more formulae with different
ScaledReg and Scale to choose, we have better opportunity to find the best solution.
That is why we believe pruning search space by only keeping the best formula with the
same Scale and ScaledReg should be more effective than PickingWinnerReg. And we use
two criteria to choose the best formula with the same Scale and ScaledReg. The first
criteria is to select the formula using less non shared registers, and the second
criteria is to select the formula with less cost got from RateFormula. The patch
implements the heuristic before NarrowSearchSpaceByPickingWinnerRegs, which is the
last resort.
Testing shows we get 1.8% and 2% on two internal benchmarks on x86. llvm nightly
testsuite performance is neutral. We also tried lsr-exp-narrow and it didn't help
on the two improved internal cases we saw.
Differential Revision: https://reviews.llvm.org/D34583
llvm-svn: 307269
It seems that the patch was reverted by mistake. Clang testing showed failure of the
MathExtras.SaturatingMultiply test, however I was unable to reproduce the issue on the
fresh code base and was able to confirm that the transformation introduced by the change
does not happen in the said test. This gives a strong confidence that the actual reason of
the failure of the initial patch was somewhere else, and that problem now seems to be
fixed. Re-submitting the change to confirm that.
llvm-svn: 307244
Summary:
GlobalExtensions is dereferenced twice, once for iteration and then a check if it is empty.
As a ManagedStatic this dereference forces it's construction which is unnecessary.
Reviewers: efriedma, davide, mehdi_amini
Reviewed By: mehdi_amini
Subscribers: chapuni, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D33381
llvm-svn: 307229
LLVM's definition of dominance allows instructions that are cyclic
in unreachable blocks, e.g.:
%pat = select i1 %condition, @global, i16* %pat
because any instruction dominates an instruction in a block that's
not reachable from entry.
So, remove unreachable blocks from the function, because a) there's
no point in analyzing them and b) GlobalOpt should otherwise grow
some more complicated logic to break these cycles.
Differential Revision: https://reviews.llvm.org/D35028
llvm-svn: 307215
This adds exact flags to AShr/LShr flags where we can statically
prove it is valid using the range of induction variables. This
allows further optimisations to remove extra loads.
Differential Revision: https://reviews.llvm.org/D34207
llvm-svn: 307157
This patch seems to cause failures of test MathExtras.SaturatingMultiply on
multiple buildbots. Reverting until the reason of that is clarified.
Differential Revision: https://reviews.llvm.org/rL307126
llvm-svn: 307135
-If there is a IndVar which is known to be non-negative, and there is a value which is also non-negative,
then signed and unsigned comparisons between them produce the same result. Both of those can be
seen in the same loop. To allow other optimizations to simplify them, we turn all instructions like
%c = icmp slt i32 %iv, %b
to
%c = icmp ult i32 %iv, %b
if both %iv and %b are known to be non-negative.
Differential Revision: https://reviews.llvm.org/D34979
llvm-svn: 307126
Summary: This makes it easier to find out which limitation prevented this pass from doing its work.
Reviewers: karthikthecool, mzolotukhin, efriedma, mcrosier
Reviewed By: mcrosier
Subscribers: mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D34940
llvm-svn: 307035
This reverts commit r306313. This breaks selfhost at -O3 and PR33652.
Let me know if you need additional information on reproducing the issue.
llvm-svn: 307021
We assumed the constant was a scalar when creating the replacement operand.
Also, improve tests for this fold and move the tests for this fold to their own file.
I'll move the related and missing tests to this file as a follow-up.
llvm-svn: 306985
I noticed this missed bswap optimization in the CGP memcmp() expansion,
and then I saw that we don't have the fold in InstCombine.
Differential Revision: https://reviews.llvm.org/D34763
llvm-svn: 306980
Summary:
I came across this while thinking about what would happen if one of the operands in this xor pattern was itself a inverted (A & ~B) ^ (~A & B)-> (A^B).
The patterns here assume that the (~a | ~b) will be demorganed to ~(a & b) first. Though I wonder if there's a multiple use case that would prevent the demorgan.
Reviewers: spatel
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34870
llvm-svn: 306967
This commit pretty much rolls back the logic added in r306495
as in the testcase provided we simplify an `icmp` looking through
a PHI that hasn't been mapped yet.
I think instsimplify shouldn't do threading over select/phis or
just looking through phis in general, but this is what we have
now. Also, add a test to prevent this from happening in case somebody
wants to modify this code again.
Briefly discussed with Kyle Butt (thanks Kyle!).
llvm-svn: 306938
With fix for use-after-free errors. We can't add the new branch and
remove the old one until we are done with the Builder constructed for
the block.
llvm-svn: 306937
Summary:
vectorizer-maximize-bandwidth is generally useful in terms of performance. I've tested the impact of changing this to default on speccpu benchmarks on sandybridge machines. The result shows non-negative impact:
spec/2006/fp/C++/444.namd 26.84 -0.31%
spec/2006/fp/C++/447.dealII 46.19 +0.89%
spec/2006/fp/C++/450.soplex 42.92 -0.44%
spec/2006/fp/C++/453.povray 38.57 -2.25%
spec/2006/fp/C/433.milc 24.54 -0.76%
spec/2006/fp/C/470.lbm 41.08 +0.26%
spec/2006/fp/C/482.sphinx3 47.58 -0.99%
spec/2006/int/C++/471.omnetpp 22.06 +1.87%
spec/2006/int/C++/473.astar 22.65 -0.12%
spec/2006/int/C++/483.xalancbmk 33.69 +4.97%
spec/2006/int/C/400.perlbench 33.43 +1.70%
spec/2006/int/C/401.bzip2 23.02 -0.19%
spec/2006/int/C/403.gcc 32.57 -0.43%
spec/2006/int/C/429.mcf 40.35 +0.27%
spec/2006/int/C/445.gobmk 26.96 +0.06%
spec/2006/int/C/456.hmmer 24.4 +0.19%
spec/2006/int/C/458.sjeng 27.91 -0.08%
spec/2006/int/C/462.libquantum 57.47 -0.20%
spec/2006/int/C/464.h264ref 46.52 +1.35%
geometric mean +0.29%
The regression on 453.povray seems real, but is due to secondary effects as all hot functions are bit-identical with and without the flag.
I started this patch to consult upstream opinions on this. It will be greatly appreciated if the community can help test the performance impact of this change on other architectures so that we can decided if this should be target-dependent.
Reviewers: hfinkel, mkuper, davidxl, chandlerc
Reviewed By: chandlerc
Subscribers: rengolin, sanjoy, javed.absar, bjope, dorit, magabari, RKSimon, llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D33341
llvm-svn: 306933
We aren't looking through any levels of IR here so I don't think we need the power of a matcher or the temporary variable it requires.
llvm-svn: 306885
Check if a single cast is preventing handling a first-order-recurrence Phi,
because the scheduling constraints it imposes on the first-order-recurrence
shuffle are infeasible; but they can be made feasible by moving the cast
downwards. Record such casts and move them when vectorizing the loop.
Differential Revision: https://reviews.llvm.org/D33058
llvm-svn: 306884
This patch appends the name of the function to the switch generated lookup
table. This will ease the visual debugging in identifying the function the table
is generated from.
Differential Revision: https://reviews.llvm.org/D34817
llvm-svn: 306867
Summary:
Runtime unrolling is done for loops with a single exit block and a
single exiting block (and this exiting block should be the latch block).
This patch adds logic to support unrolling in the presence of multiple exit
blocks (which also means multiple exiting blocks).
Currently this is under an off-by-default option and is supported when
epilog code is generated. Support in presence of prolog code will be in
a future patch (we just need to add more tests, and update comments).
This patch is essentially an implementation patch. I have not added any
heuristic (in terms of branches added or code size) to decide when
this should be enabled.
Reviewers: mkuper, sanjoy, reames, evstupac
Reviewed by: reames
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D33001
llvm-svn: 306846
It may be detrimental to vectorize loops with very small trip count, as various
costs of the vectorized loop body as well as enclosing overheads including
runtime tests and scalar iterations may outweigh the gains of vectorizing. The
current cost model measures the cost of the vectorized loop body only, expecting
it will amortize other costs, and loops with known or expected very small trip
counts are not vectorized at all. This patch allows loops with very small trip
counts to be vectorized, but under OptForSize constraints, which ensure the cost
of the loop body is dominant, having no runtime guards nor scalar iterations.
Patch inspired by D32451.
Differential Revision: https://reviews.llvm.org/D34373
llvm-svn: 306803
There are two conditions ORed here with similar checks and each contain two matches that must be true for the if to succeed. With the commutable match on the first half of the OR then both ifs basically have the same first part and only the second part distinguishs. With this change we move the commutable match to second half and make the first half unique.
This caused some tests to change because we now produce a commuted result, but this shouldn't matter in practice.
llvm-svn: 306800
It served us well, helped kick-start much of the vectorization efforts
in LLVM, etc. Its time has come and past. Back in 2014:
http://lists.llvm.org/pipermail/llvm-dev/2014-November/079091.html
Time to actually let go and move forward. =]
I've updated the release notes both about the removal and the
deprecation of the corresponding C API.
llvm-svn: 306797
In rL300494 there was an attempt to deal with excessive compile time on
invocations of getSign/ZeroExtExpr using local caching. This approach only
helps if we request the same SCEV multiple times throughout recursion. But
in the bug PR33431 we see a case where we request different values all the time,
so caching does not help and the size of the cache grows enormously.
In this patch we remove the local cache for this methods and add the recursion
depth limit instead, as we do for arithmetics. This gives us a guarantee that the
invocation sequence is limited and reasonably short.
Differential Revision: https://reviews.llvm.org/D34273
llvm-svn: 306785
The style guide states that the explicit `inline`
should not be used with inline methods. classof is
very common inline method with a fair amount on
inconsistency:
$ git grep classof ./include | grep inline | wc -l
230
$ git grep classof ./include | grep -v inline | wc -l
257
I chose to target this method rather the larger change
since this method is easily cargo-culted (I did it at
least once). I considered doing the larger change and
removing all occurrences but that would be a much larger
change.
Differential Revision: https://reviews.llvm.org/D33906
llvm-svn: 306731
Summary:
Indices for GEPs that index into a struct type should always be
constants. This added more checks in `collectConstantCandidates:` which make
sure constants for GEP pointer type are not hoisted.
This fixed Bug https://bugs.llvm.org/show_bug.cgi?id=33538
Reviewers: ributzka, rnk
Reviewed By: ributzka
Subscribers: efriedma, llvm-commits, srhines, javed.absar, pirama
Differential Revision: https://reviews.llvm.org/D34576
llvm-svn: 306704
Summary:
As discussed on the mailing list it is legal to propagate TBAA to loads/stores
from/to smaller regions of a larger load tagged with TBAA. Do so for
(load->extractvalue)=>(gep->load) and similar foldings.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D31954
llvm-svn: 306615
r306381 caused PR33613, by reversing the order in which insertelements were
generated per unroll part. This patch fixes PR33613 by retraining this order,
placing each set of insertelements per part immediately after the last scalar
being packed for this part. Includes a test case derived from PR33613.
Reference: https://bugs.llvm.org/show_bug.cgi?id=33613
Differential Revision: https://reviews.llvm.org/D34760
llvm-svn: 306575
Summary:
I was testing using this expansion logic in other cases besides
NVPTX, and found some runtime failures due to the lack of a check
for a zero length memcpy/memset before the loop. There is already
such a check in the memmove expansion code though.
Reviewers: hfinkel
Subscribers: jholewinski, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D34707
llvm-svn: 306541
Summary:
This commit allows matchSelectPattern to recognize clamp of float
arguments in the presence of FMF the same way as already done for
integers.
This case is a little different though. With integers, given the
min/max pattern is recognized, DAGBuilder starts selecting MIN/MAX
"automatically". That is not the case for float, because for them only
full FMINNAN/FMINNUM/FMAXNAN/FMAXNUM ISD nodes exist and they do care
about NaNs. On the other hand, some backends (e.g. X86) have only
FMIN/FMAX nodes that do not care about NaNS and the former NAN/NUM
nodes are illegal thus selection is not happening. So I decided to do
such kind of transformation in IR (InstCombiner) instead of
complicating the logic in the backend.
Reviewers: spatel, jmolloy, majnemer, efriedma, craig.topper
Reviewed By: efriedma
Subscribers: hiraditya, javed.absar, n.bozhenov, llvm-commits
Patch by Andrei Elovikov <andrei.elovikov@intel.com>
Differential Revision: https://reviews.llvm.org/D33186
llvm-svn: 306525
A slightly more efficient way to get constant, we avoid resolving in getSCEV and excessive
invocations, and we don't create a ConstantInt if 'true' branch is taken.
Differential Revision: https://reviews.llvm.org/D34672
llvm-svn: 306503
When simplifying an instruction that has been re-mapped, it should never
simplify to an instruction in the original function. In the edge case
where we are inlining a function into itself, the existing code led to
incorrect behavior. Replace the incorrect code with an assert verifying
that we never expect simplification to produce an instruction in the old
function, unless the functions are the same.
Differential Revision: https://reviews.llvm.org/D33850
llvm-svn: 306495
The check to see if we can propagate the nsw flag used m_ConstantInt(uint64_t*&) which doesn't work with splat vectors and has a restriction that the bitwidth of the ConstantInt must be 64-bits are less.
This patch changes it to use m_APInt to remove both these issues
Differential Revision: https://reviews.llvm.org/D34699
llvm-svn: 306457
BlockAddress are only valid within their function context, which does not
interact well with CodeExtractor. Detect this case and prevent it.
Differential Revision: https://reviews.llvm.org/D33839
llvm-svn: 306448
SROA assumes alloca address space is 0, which causes assertion. This patch fixes that.
Differential Revision: https://reviews.llvm.org/D34104
llvm-svn: 306440
This canonicalization was suggested in D33172 as a way to make InstCombine behavior more uniform.
We have this transform for icmp+br, so unless there's some reason that icmp+select should be
treated differently, we should do the same thing here.
The benefit comes from increasing the chances of creating identical instructions. This is shown in
the tests in logical-select.ll (PR32791). InstCombine doesn't fold those directly, but EarlyCSE
can simplify the identical cmps, and then InstCombine can fold the selects together.
The possible regression for the tests in select.ll raises questions about poison/undef:
http://lists.llvm.org/pipermail/llvm-dev/2017-May/113261.html
...but that transform is just as likely to be triggered by this canonicalization as it is to be
missed, so we're just pointing out a commutation deficiency in the pattern matching:
https://reviews.llvm.org/rL228409
Differential Revision: https://reviews.llvm.org/D34242
llvm-svn: 306435
Instead of getBackEdgeTakenCount, use getExitCount on the latch exiting block
(which is proven to be the only exiting block in the loop to be unrolled).
llvm-svn: 306410
Undoing revert 306338 after fixed bug: add metadata to the load instead of the
reverse shuffle added to it, retaining the original ValueMap implementation.
llvm-svn: 306381
This is based heavily on the work done ni D34285. I mostly wanted to do
test cleanup for the author to save them some time, but I had a really
hard time understanding why it was so hard to write better test cases
for these issues.
The problem is that because SROA does a second rewrite of the loads and
because we *don't* propagate !nonnull for non-pointer loads, we first
introduced invalid !nonnull metadata and then stripped it back off just
in time to avoid most ways of this PR manifesting. Moving to the more
careful utility only fixes this by changing the predicate to look at the
new load's type rather than the target type. However, that *does* fix
the bug, and the utility is much nicer including adding range metadata
to model the nonnull property after a conversion to an integer.
However, we have bigger problems because we don't actually propagate
*range* metadata, and the utility to do this extracted from instcombine
isn't really in good shape to do this currently. It *only* handles the
case of copying range metadata from an integer load to a pointer load.
It doesn't even handle the trivial cases of propagating from one integer
load to another when they are the same width! This utility will need to
be beefed up prior to using in this location to get the metadata to
fully survive.
And even then, we need to go and teach things to turn the range metadata
into an assume the way we do with nonnull so that when we *promote* an
integer we don't lose the information.
All of this will require a new test case that looks kind-of like
`preserve-nonnull.ll` does here but focuses on range metadata. It will
also likely require more testing because it needs to correctly handle
changes to the integer width, especially as SROA actively tries to
change the integer width!
Last but not least, I'm a little worried about hooking the range
metadata up here because the instcombine logic for converting from
a range metadata *to* a nonnull metadata node seems broken in the face
of non-zero address spaces where null is not mapped to the integer `0`.
So that probably needs to get fixed with test cases both in SROA and in
instcombine to cover it.
But this *does* extract the core PR fix from D34285 of preventing the
!nonnull metadata from being propagated in a broken state just long
enough to feed into promotion and crash value tracking.
On D34285 there is some discussion of zero-extend handling because it
isn't necessary. First, the new load size covers all of the non-undef
(ie, possibly initialized) bits. This may even extend past the original
alloca if loading those bits could produce valid data. The only way its
valid for us to zero-extend an integer load in SROA is if the original
code had a zero extend or those bits were undef. And we get to assume
things like undef *never* satifies nonnull, so non undef bits can
participate here. No need to special case the zero-extend handling, it
just falls out correctly.
The original credit goes to Ariel Ben-Yehuda! I'm mostly landing this to
save a few rounds of trivial edits fixing style issues and test case
formulation.
Differental Revision: D34285
llvm-svn: 306379
Summary:
EraseInst didn't report that it made IR changes through MadeChange.
It is essential that changes to the IR are reported correctly,
since for example ReassociatePass::run() will indicate that all
analyses are preserved otherwise.
And the CGPassManager determines if the CallGraph is up-to-date
based on status from InstructionCombiningPass::runOnFunction().
Reviewers: craig.topper, rnk, davide
Reviewed By: rnk, davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34616
llvm-svn: 306368
Summary:
vectorizer-maximize-bandwidth is generally useful in terms of performance. I've tested the impact of changing this to default on speccpu benchmarks on sandybridge machines. The result shows non-negative impact:
spec/2006/fp/C++/444.namd 26.84 -0.31%
spec/2006/fp/C++/447.dealII 46.19 +0.89%
spec/2006/fp/C++/450.soplex 42.92 -0.44%
spec/2006/fp/C++/453.povray 38.57 -2.25%
spec/2006/fp/C/433.milc 24.54 -0.76%
spec/2006/fp/C/470.lbm 41.08 +0.26%
spec/2006/fp/C/482.sphinx3 47.58 -0.99%
spec/2006/int/C++/471.omnetpp 22.06 +1.87%
spec/2006/int/C++/473.astar 22.65 -0.12%
spec/2006/int/C++/483.xalancbmk 33.69 +4.97%
spec/2006/int/C/400.perlbench 33.43 +1.70%
spec/2006/int/C/401.bzip2 23.02 -0.19%
spec/2006/int/C/403.gcc 32.57 -0.43%
spec/2006/int/C/429.mcf 40.35 +0.27%
spec/2006/int/C/445.gobmk 26.96 +0.06%
spec/2006/int/C/456.hmmer 24.4 +0.19%
spec/2006/int/C/458.sjeng 27.91 -0.08%
spec/2006/int/C/462.libquantum 57.47 -0.20%
spec/2006/int/C/464.h264ref 46.52 +1.35%
geometric mean +0.29%
The regression on 453.povray seems real, but is due to secondary effects as all hot functions are bit-identical with and without the flag.
I started this patch to consult upstream opinions on this. It will be greatly appreciated if the community can help test the performance impact of this change on other architectures so that we can decided if this should be target-dependent.
Reviewers: hfinkel, mkuper, davidxl, chandlerc
Reviewed By: chandlerc
Subscribers: rengolin, sanjoy, javed.absar, bjope, dorit, magabari, RKSimon, llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D33341
llvm-svn: 306336
Instead of providing access to the internal MapStorage holding all Values
associated with a given Key, used for setting or resetting them all together,
ValueMap keeps its MapStorage internal; its new interface allows getting,
setting or resetting a single Value, per part or per part-and-lane.
Follows the discussion in https://reviews.llvm.org/D32871.
Differential Revision: https://reviews.llvm.org/D34473
llvm-svn: 306331
The recommit fixes three bugs: The first one is to use CurrentBlock instead of
PREInstr's Parent as param of performScalarPREInsertion because the Parent
of a clone instruction may be uninitialized. The second one is stop PRE when
CurrentBlock to its predecessor is a backedge and an operand of CurInst is
defined inside of CurrentBlock. The same value defined inside of loop in last
iteration can not be regarded as available. The third one is an out-of-bound
array access in a flipped if guard.
Right now scalarpre doesn't have phi-translate support, so it will miss some
simple pre opportunities. Like the following testcase, current scalarpre cannot
recognize the last "a * b" is fully redundent because a and b used by the last
"a * b" expr are both defined by phis.
long a[100], b[100], g1, g2, g3;
__attribute__((pure)) long goo();
void foo(long a, long b, long c, long d) {
g1 = a * b;
if (__builtin_expect(g2 > 3, 0)) {
a = c;
b = d;
g2 = a * b;
}
g3 = a * b; // fully redundant.
}
The patch adds phi-translate support in scalarpre. This is only a temporary
solution before the newpre based on newgvn is available.
llvm-svn: 306313
metadata out of InstCombine and into helpers.
NFC, this just exposes the logic used by InstCombine when propagating
metadata from one load instruction to another. The plan is to use this
in SROA to address PR32902.
If anyone has better ideas about how to factor this or name variables,
I'm all ears, but this seemed like a pretty good start and lets us make
progress on the PR.
This is based on a patch by Ariel Ben-Yehuda (D34285).
llvm-svn: 306267
This was reverted in r306252, but I already had the bug fixed and was
just trying to form a test case.
The original commit factored the logic for forming dedicated exits
inside of LoopSimplify into a helper that could be used elsewhere and
with an approach that required fewer intermediate data structures. See
that commit for full details including the change to the statistic, etc.
The code looked fine to me and my reviewers, but in fact didn't handle
indirectbr correctly -- it left the 'InLoopPredecessors' vector dirty.
If you have code that looks *just* right, you can end up leaking these
predecessors into a subsequent rewrite, and crash deep down when trying
to update PHI nodes for predecessors that don't exist.
I've added an assert that makes the bug much more obvious, and then
changed the code to reliably clear the vector so we don't get this bug
again in some other form as the code changes.
I've also added a test case that *does* manage to catch this while also
giving some nice positive coverage in the face of indirectbr.
The real code that found this came out of what I think is CPython's
interpreter loop, but any code with really "creative" interpreter loops
mixing indirectbr and other exit paths could manage to tickle the bug.
I was hard to reduce the original test case because in addition to
having a particular pattern of IR, the whole thing depends on the order
of the predecessors which is in turn depends on use list order. The test
case added here was designed so that in multiple different predecessor
orderings it should always end up going down the same path and tripping
the same bug. I hope. At least, it tripped it for me without
manipulating the use list order which is better than anything bugpoint
could do...
llvm-svn: 306257
Recommit NFC patch (rL306157) where I missed incrementing the basic block iterator,
which caused loop deletion tests to hang due to infinite loop.
Had reverted it in rL306162.
rL306157 commit message:
Currently, the implementation of delete dead loops has a special case
when the loop being deleted is never executed. This special case
(updating of exit block's incoming values for phis) can be
run as a prepass for non-executable loops before performing
the actual deletion.
llvm-svn: 306254
http://rise4fun.com/Alive/i8Q
A narrow bitwise logic op is obviously better than math for value tracking,
and zext is better than sext. Typically, the 'not' will be folded into an
icmp predicate.
The IR difference would even survive through codegen for x86, so we would see
worse code:
https://godbolt.org/g/C14HMF
one_or_zero(int, int): # @one_or_zero(int, int)
xorl %eax, %eax
cmpl %esi, %edi
setle %al
retq
one_or_zero_alt(int, int): # @one_or_zero_alt(int, int)
xorl %ecx, %ecx
cmpl %esi, %edi
setg %cl
movl $1, %eax
subl %ecx, %eax
retq
llvm-svn: 306243
Summary:
InstCombine replaces large allocas with small globals consts causing buffer overflows
on valid code, see PR33372.
This fix permits this optimization only if the global is dereference for alloca size.
Fixes PR33372
Reviewers: eugenis, majnemer, chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34311
llvm-svn: 306194
This reverts commit r306157.
It caused some timeouts in clang tests. Perhaps unreachable loops have
far too many phi nodes.
Reverting and investigating.
llvm-svn: 306162
Currently, the implementation of delete dead loops has a special case
when the loop being deleted is never executed. This special case
(updating of exit block's incoming values for phis) can be
run as a prepass for non-executable loops before performing
the actual deletion.
llvm-svn: 306157
Summary:
Many languages have a three way comparison idiom where comparing two values
produces not a boolean, but a tri-state value. Typical values (e.g. as used in
the lcmp/fcmp bytecodes from Java) are -1 for less than, 0 for equality, and +1
for greater than.
We actually do a great job already of converting three way comparisons into
binary comparisons when the result produced has one a single use. Unfortunately,
such values can have more than one use, and in that case, our existing
optimizations break down.
The patch adds a peephole which converts a three-way compare + test idiom into a
binary comparison on the original inputs. It focused on replacing the test on
the result of the three way compare and does nothing about removing the three
way compare itself. That's left to other optimizations (which do actually kick
in commonly.)
We currently recognize one idiom on signed integer compare. In the future, we
plan to recognize and simplify other comparison idioms on
other signed/unsigned datatypes such as floats, vectors etc.
This is a resurrection of Philip Reames' original patch:
https://reviews.llvm.org/D19452
Reviewers: majnemer, apilipenko, reames, sanjoy, mkazantsev
Reviewed by: mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34278
llvm-svn: 306100
Currently JumpThreading can use LazyValueInfo to analyze an 'and' or 'or' of compare if the compare is fed by a livein of a basic block. This can be used to to prove the condition can't be met for some predecessor and the jump from that predecessor can be moved to the false path of the condition.
But if the compare is something that InstCombine turns into an add and a single compare, it can't be analyzed because the livein is now an input to the add and not the compare.
This patch adds a new method to LVI to get a ConstantRange on an edge. Then we teach jump threading to detect the add livein feeding a compare and to get the ConstantRange and propagate it.
Differential Revision: https://reviews.llvm.org/D33262
llvm-svn: 306085
I want to use the same logic as LoopSimplify to form dedicated exits in
another pass (SimpleLoopUnswitch) so I wanted to factor it out here.
I also noticed that there is a pretty significantly more efficient way
to implement this than the way the code in LoopSimplify worked. We don't
need to actually retain the set of unique exit blocks, we can just
rewrite them as we find them and use only a set to deduplicate.
This did require changing one part of LoopSimplify to not re-use the
unique set of exits, but it only used it to check that there was
a single unique exit. That part of the code is about to walk the exiting
blocks anyways, so it seemed better to rewrite it to use those exiting
blocks to compute this property on-demand.
I also had to ditch a statistic, but it doesn't seem terribly valuable.
Differential Revision: https://reviews.llvm.org/D34049
llvm-svn: 306081
Summary:
Currently, we incorrectly update exit blocks of loops when there are multiple
edges from a single exiting block to the exit block. This can happen when we
have switches as the terminator of the exiting blocks.
The fix here is to correctly update the phi nodes in the exit block, and remove
all incoming values *except* for one which is from the preheader.
Note: Currently, this error can manifest only while deleting non-executed loops. However, it
is possible to trigger this error in invariant loops, once we enhance the logic
around the exit conditions for the loop check.
Reviewers: chandlerc, dberlin, sanjoy, efriedma
Reviewed by: efriedma
Subscribers: mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D34516
llvm-svn: 306048
Summary:
InstCombine likes to turn (icmp eq (and X, C1), 0) into (icmp slt (trunc (X)), 0) sometimes. This breaks foldSelectICmpAndOr's ability to recognize (select (icmp eq (and X, C1), 0), Y, (or Y, C2))->(or (shl (and X, C1), C3), y).
This patch tries to recover this. I had to flip around some of the early out checks so that I could create a new And instruction during the compare processing without it possibly never getting used.
Reviewers: spatel, majnemer, davide
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34184
llvm-svn: 306029
If the components of the and/or had multiple uses, this transform created an additional instruction.
This patch makes sure we remove one of the components.
Differential Revision: https://reviews.llvm.org/D34498
llvm-svn: 306027
There are 2 parts to this patch made simultaneously to avoid a regression.
We're reversing the canonicalization that moves bitwise vector ops before bitcasts.
We're moving bitwise vector ops *after* bitcasts instead. That's the 1st and 3rd hunks
of the patch. The motivation is that there's only one fold that currently depends on
the existing canonicalization (see next), but there are many folds that would
automatically benefit from the new canonicalization.
PR33138 ( https://bugs.llvm.org/show_bug.cgi?id=33138 ) shows why/how we have these
patterns in IR.
There's an or(and,andn) pattern that requires an adjustment in order to continue matching
to 'select' because the bitcast changes position. This match is unfortunately complicated
because it requires 4 logic ops with optional bitcast and sext ops.
Test diffs:
1. The bitcast.ll and bitcast-bigendian.ll changes show the most basic difference -
bitcast comes before logic.
2. There are also tests with no diffs in bitcast.ll that verify that we're still doing
folds that were enabled by the previous canonicalization.
3. icmp-xor-signbit.ll shows the payoff. We don't need to adjust existing icmp patterns
to look through bitcasts.
4. logical-select.ll contains several tests for the or(and,andn) --> select fold to
verify that we are still handling those cases. The lone diff shows the movement of
the bitcast from the new canonicalization rule.
Differential Revision: https://reviews.llvm.org/D33517
llvm-svn: 306011
Summary:
vectorizer-maximize-bandwidth is generally useful in terms of performance. I've tested the impact of changing this to default on speccpu benchmarks on sandybridge machines. The result shows non-negative impact:
spec/2006/fp/C++/444.namd 26.84 -0.31%
spec/2006/fp/C++/447.dealII 46.19 +0.89%
spec/2006/fp/C++/450.soplex 42.92 -0.44%
spec/2006/fp/C++/453.povray 38.57 -2.25%
spec/2006/fp/C/433.milc 24.54 -0.76%
spec/2006/fp/C/470.lbm 41.08 +0.26%
spec/2006/fp/C/482.sphinx3 47.58 -0.99%
spec/2006/int/C++/471.omnetpp 22.06 +1.87%
spec/2006/int/C++/473.astar 22.65 -0.12%
spec/2006/int/C++/483.xalancbmk 33.69 +4.97%
spec/2006/int/C/400.perlbench 33.43 +1.70%
spec/2006/int/C/401.bzip2 23.02 -0.19%
spec/2006/int/C/403.gcc 32.57 -0.43%
spec/2006/int/C/429.mcf 40.35 +0.27%
spec/2006/int/C/445.gobmk 26.96 +0.06%
spec/2006/int/C/456.hmmer 24.4 +0.19%
spec/2006/int/C/458.sjeng 27.91 -0.08%
spec/2006/int/C/462.libquantum 57.47 -0.20%
spec/2006/int/C/464.h264ref 46.52 +1.35%
geometric mean +0.29%
The regression on 453.povray seems real, but is due to secondary effects as all hot functions are bit-identical with and without the flag.
I started this patch to consult upstream opinions on this. It will be greatly appreciated if the community can help test the performance impact of this change on other architectures so that we can decided if this should be target-dependent.
Reviewers: hfinkel, mkuper, davidxl, chandlerc
Reviewed By: chandlerc
Subscribers: rengolin, sanjoy, javed.absar, bjope, dorit, magabari, RKSimon, llvm-commits, mzolotukhin
Differential Revision: https://reviews.llvm.org/D33341
llvm-svn: 305960