In the last change to IRCE the BPI is ignored if BFI is present, however
BFI and BPI have a different thresholds. Specifically BPI approach checks only
latch exit probability so it is expected if the loop has only one exit block (latch)
the behavior with BFI and BPI should be the same,
BPI approach by default uses threshold 10, so it considers the loop with estimated
number of iterations less then 10 should not be considered for IRCE optimization.
BFI approach uses the default value 3 and this is inconsistent.
The CL modifies the code to use the same threshold for both approaches..
The test is updated due to it has two side-exits (except latch) and each of them has a
probability 1/16, so BFI estimates the number of runtime iteration is about to 7
(1/16 + 1/16 + some for latch) and test fails.
Reviewers: mkazantsev, ebrevnov
Reviewed By: mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D91230
There are 1-2 potential follow-up NFC commits to reduce
this further on the way to generalizing this for vectors.
The operand replacing path should be dead code because demanded
bits handles that more generally (D91415).
I noticed an add example like the one from D91343, so here's a similar patch.
The logic is based on existing code for the single-use demanded bits fold.
But I only matched a constant instead of using compute known bits on the
operands because that was the motivating patterni that I noticed.
I think this will allow removing a special-case (but incomplete) dedicated
fold within visitAnd(), but I need to untangle the existing code to be sure.
https://rise4fun.com/Alive/V6fP
Name: add with low mask
Pre: (C1 & (-1 u>> countLeadingZeros(C2))) == 0
%a = add i8 %x, C1
%r = and i8 %a, C2
=>
%r = and i8 %x, C2
Differential Revision: https://reviews.llvm.org/D91415
This patch turns VPWidenGEPRecipe into a VPValue and uses it
during VPlan construction and codegeneration instead of the plain IR
reference where possible.
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D84683
This is used to test RemoveRedundantDbgInstrs(), which is used by other
passes.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D91477
See discussion in https://bugs.llvm.org/show_bug.cgi?id=45073 / https://reviews.llvm.org/D66324#2334485
the implementation is known-broken for certain inputs,
the bugreport was up for a significant amount of timer,
and there has been no activity to address it.
Therefore, just completely rip out all of misexpect handling.
I suspect, fixing it requires redesigning the internals of MD_misexpect.
Should anyone commit to fixing the implementation problem,
starting from clean slate may be better anyways.
This reverts commit 7bdad08429,
and some of it's follow-ups, that don't stand on their own.
dependency. NFC
Use findSingleDependency in place of FindDependencies and stop passing a
set of Instructions around. Modify FindDependencies to return a boolean
flag which indicates whether the dependencies it has found are all
valid.
Like inlineCallIfPossible and InlinerPass, after inlining mergeAttributesForInlining
should be called to merge callee's attributes to caller. But it is not called in
AlwaysInliner, causes caller's attributes inconsistent with inlined code.
Attached test case demonstrates that attribute "min-legal-vector-width"="512" is
not merged into caller without this patch, and it causes failure in SelectionDAG
when lowering the inlined AVX512 intrinsic.
Differential Revision: https://reviews.llvm.org/D91446
Handle the emission of the add in a single place, instead of three
different ones.
Don't emit an unnecessary add with zero to start with. It will get
dropped by InstCombine, but we may as well not create it in the
first place. This also means that InstCombine does not need to
specially handle this extra add.
This is conceptually NFC, but can affect worklist order etc.
Use exact component name in add_ocaml_library.
Make expand_topologically compatible with new architecture.
Fix quoting in is_llvm_target_library.
Fix LLVMipo component name.
Write release note.
This patch adds a new !annotation metadata kind which can be used to
attach annotation strings to instructions.
It also adds a new pass that emits summary remarks per function with the
counts for each annotation kind.
The intended uses cases for this new metadata is annotating
'interesting' instructions and the remarks should provide additional
insight into transformations applied to a program.
To motivate this, consider these specific questions we would like to get answered:
* How many stores added for automatic variable initialization remain after optimizations? Where are they?
* How many runtime checks inserted by a frontend could be eliminated? Where are the ones that did not get eliminated?
Discussed on llvm-dev as part of 'RFC: Combining Annotation Metadata and Remarks'
(http://lists.llvm.org/pipermail/llvm-dev/2020-November/146393.html)
Reviewed By: thegameg, jdoerfert
Differential Revision: https://reviews.llvm.org/D91188
No longer rely on an external tool to build the llvm component layout.
Instead, leverage the existing `add_llvm_componentlibrary` cmake function and
introduce `add_llvm_component_group` to accurately describe component behavior.
These function store extra properties in the created targets. These properties
are processed once all components are defined to resolve library dependencies
and produce the header expected by llvm-config.
Differential Revision: https://reviews.llvm.org/D90848
If we cannot prove that the check is trivially true, but can prove that it either
fails on the 1st iteration or never fails, we can replace it with first iteration check.
Differential Revision: https://reviews.llvm.org/D88527
Reviewed By: skatkov
There might be some demanded/known bits way to generalize this,
but I'm not seeing it right now.
This came up as a regression when I was looking at a different
demanded bits improvement.
https://rise4fun.com/Alive/5fl
Name: general
Pre: ((-1 << countTrailingZeros(C1)) & C2) == 0
%a1 = add i8 %x, C1
%a2 = and i8 %x, C2
%r = sub i8 %a1, %a2
=>
%r = and i8 %a1, ~C2
Name: test 1
%a1 = add i8 %x, 192
%a2 = and i8 %x, 10
%r = sub i8 %a1, %a2
=>
%r = and i8 %a1, -11
Name: test 2
%a1 = add i8 %x, -108
%a2 = and i8 %x, 3
%r = sub i8 %a1, %a2
=>
%r = and i8 %a1, -4
Summary:
Refactor SinkAdHoistLICMFlags from a struct to a class with accessors and constructors to allow other
classes to construct flags with meaningful defaults while not exposing LICM internal details.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: asbirlea (Alina Sbirlea)
Differential Revision: https://reviews.llvm.org/D90482
Sometimes the an instruction we are trying to widen is used by the IV
(which means the instruction is the IV increment). Currently this may
prevent its widening. We should ignore such user because it will be
dead once the transform is done anyways.
Differential Revision: https://reviews.llvm.org/D90920
Reviewed By: fhahn
In the existing logic, for a given alloca, as long as its pointer value is stored into another location, it's considered as escaped.
This is a bit too conservative. Specifically, in non-optimized build mode, it's often to have patterns of code that first store an alloca somewhere and then load it right away.
These used should be handled without conservatively marking them escaped.
This patch tracks how the memory location where an alloca pointer is stored into is being used. As long as we only try to load from that location and nothing else, we can still
consider the original alloca not escaping and keep it on the stack instead of putting it on the frame.
Differential Revision: https://reviews.llvm.org/D91305
InstCombine canonicalizes 'sub nuw' instructions to 'add' without the
`nuw` flag. The typical case where we see it is decrementing induction
variables. For them, IndVars fails to prove that it's legal to widen them,
and inserts unprofitable `zext`'s.
This patch adds recognition of such pattern using SCEV.
Differential Revision: https://reviews.llvm.org/D89550
Reviewed By: fhahn, skatkov
We need to be able to call function pointers. Inline the dispatch
function.
Also inline the context projection function.
Transfer debug locations from the suspend point to the inlined functions.
Use the function argument index instead of the function argument in
coro.id.async. This solves any spurious use issues.
Coerce the arguments of the tail call function at a suspend point. The LLVM
optimizer seems to drop casts leading to a vararg intrinsic.
rdar://70097093
Differential Revision: https://reviews.llvm.org/D91098
Previously the inliner did a bit of a hack by adding ref edges for all
new edges introduced by performing an inline before calling
updateCGAndAnalysisManagerForPass(). This was because
updateCGAndAnalysisManagerForPass() didn't handle new non-trivial call
edges.
This adds handling of non-trivial call edges to
updateCGAndAnalysisManagerForPass(). The inliner called
updateCGAndAnalysisManagerForFunctionPass() since it was handling adding
newly introduced edges (so updateCGAndAnalysisManagerForPass() would
only have to handle promotion), but now it needs to call
updateCGAndAnalysisManagerForCGSCCPass() since
updateCGAndAnalysisManagerForPass() is now handling the new call edges
and function passes cannot add new edges.
We follow the previous path of adding trivial ref edges then letting promotion
handle changing the ref edges to call edges and the CGSCC updates. So
this still does not allow adding call edges that result in an addition
of a non-trivial ref edge.
This is in preparation for better detecting devirtualization. Previously
since the inliner itself would add ref edges,
updateCGAndAnalysisManagerForPass() would think that promotion and thus
devirtualization had happened after any sort of inlining.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D91046
Before the change, DFSan always does the propagation. W/o
origin tracking, it is harder to understand such flows. After
the change, the flag is off by default.
Reviewed-by: morehouse
Differential Revision: https://reviews.llvm.org/D91234
This reverts commits:
* [LoopVectorizer] NFCI: Calculate register usage based on TLI.getTypeLegalizationCost.
b873aba394.
* [LoopVectorizer] Silence warning in GetRegUsage.
9ff701100a.
We have a frequent pattern where we're merging two KnownBits to get the common/shared bits, and I just fell for the gotcha where I tried to use the & operator to merge them........
This patch silences the warning:
error: lambda capture 'DL' is not used [-Werror,-Wunused-lambda-capture]
auto GetRegUsage = [&DL, &TTI=TTI](Type *Ty, ElementCount VF) {
~^~~
1 error generated.
Introduced in:
https://reviews.llvm.org/rGb873aba3943c067a5efd5303cbdf5aeb0732cf88
This is more accurate than dividing the bitwidth based on the element count by the
maximum register size, as it can just reuse whatever has been calculated for
legalization of these types.
This change is also necessary when calculating register usage for scalable vectors, where
the legalization of these types cannot be done based on the widest register size, because
that does not take the 'vscale' component into account.
Reviewed By: SjoerdMeijer
Differential Revision: https://reviews.llvm.org/D91059
delete abs/nabs handling in earlycse pass to avoid bugs related to
hashing values. After abs/nabs is canonicalized to intrinsics in D87188,
we should get CSE ability for abs/nabs back.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D90734
Tracking local variables across suspend points is still somewhat incomplete.
Consider this coroutine snippet:
```
resumable foo() {
int x[10] = {};
int a = 3;
co_await std::experimental::suspend_always();
a++;
x[0] = 1;
a += 2;
x[1] = 2;
a += 3;
x[2] = 3;
}
```
Can't manage to print `a` or `x` if they turn out to be allocas during
CoroSplit (which happens if you build this code with `-O0` prior to this
commit):
```
* thread #1, queue = 'com.apple.main-thread', stop reason = step over
frame #0: 0x0000000100003729 main-noprint`foo() at main-noprint.cpp:43:5
40 co_await std::experimental::suspend_always();
41 a++;
42 x[0] = 1;
-> 43 a += 2;
44 x[1] = 2;
45 a += 3;
46 x[2] = 3;
(lldb) p x
error: <user expression 21>:1:1: use of undeclared identifier 'x'
x
^
```
The generated IR contains a `llvm.dbg.declare` for `x` in it's initialization
basic block. After CoroSplit, the `llvm.dbg.declare` might not dominate all of
`x` uses and we lose debugging quality.
Add `llvm.dbg.value`s to all relevant basic blocks such that if later
transformations break the dominance the reliable debug info is already in
place. For instance, this BB:
```
await.ready:
...
%arrayidx = getelementptr inbounds [10 x i32], [10 x i32]* %x.reload.addr, i64 0, i64 0, !dbg !760
...
%arrayidx19 = getelementptr inbounds [10 x i32], [10 x i32]* %x.reload.addr, i64 0, i64 1, !dbg !763
...
%arrayidx21 = getelementptr inbounds [10 x i32], [10 x i32]* %x.reload.addr, i64 0, i64 2, !dbg !766
```
becomes:
```
await.ready:
...
call void @llvm.dbg.value(metadata [10 x i32]* %x.reload.addr, metadata !751, metadata !DIExpression()), !dbg !753
...
%arrayidx = getelementptr inbounds [10 x i32], [10 x i32]* %x.reload.addr, i64 0, i64 0, !dbg !760
...
%arrayidx19 = getelementptr inbounds [10 x i32], [10 x i32]* %x.reload.addr, i64 0, i64 1, !dbg !763
...
%arrayidx21 = getelementptr inbounds [10 x i32], [10 x i32]* %x.reload.addr, i64 0, i64 2, !dbg !766
```
Differential Revision: https://reviews.llvm.org/D90772
This is a prep step for widening induction variables in LoopFlatten if this is
posssible (D90640), to avoid having to perform certain overflow checks. Since
IndVarSimplify may already widen induction variables, we want to run
LoopFlatten just before IndVarSimplify. This is a minor reshuffle as both
passes were already close after each other.
Differential Revision: https://reviews.llvm.org/D90402
This converts LoopFlatten from a LoopPass to a FunctionPass so that we don't
run into problems of a loop pass deleting a (inner)loop.
Differential Revision: https://reviews.llvm.org/D90940
This patch turns VPWidenSelectRecipe into a VPValue and uses it
during VPlan construction and codegeneration instead of the plain IR
reference where possible.
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D84682
This was missing as discovered by the SystemZ multistage bot:
http://lab.llvm.org:8011/#/builders/8, where wrong code resulted when this
extension was not performed.
Thanks for review by Ulrich Weigand and Roman Lebedev.
Differential Revision: https://reviews.llvm.org/D90760
We already do not unroll loops with vector instructions under MVE, but
that does not include the remainder loops that the vectorizer produces.
These remainder loops will be rarely executed and are not worth
unrolling, as the trip count is likely to be low if they get executed at
all. Luckily they get llvm.loop.isvectorized to make recognizing them
simpler.
We have wanted to do this for a while but hit issues with low overhead
loops being reverted due to difficult registry allocation. With recent
changes that seems to be less of an issue now.
Differential Revision: https://reviews.llvm.org/D90055
The LoopDistribute pass is missing from the LTO pipeline, so
-enable-loop-distribute has no effect during post-link. The pre-link
loop distribution doesn't seem to survive the LTO pipeline either.
With this patch (and -flto -mllvm -enable-loop-distribute) we see a 43%
uplift on SPEC 2006 hmmer for AArch64. The rest of SPECINT 2006 is
unaffected.
Differential Revision: https://reviews.llvm.org/D89896
Interfaces changed to take `ElementCount` as parameters:
* LoopVectorizationPlanner::buildVPlans
* LoopVectorizationPlanner::buildVPlansWithVPRecipes
* LoopVectorizationCostModel::selectVectorizationFactor
This patch is NFC for fixed-width vectors.
Reviewed By: dmgreen, ctetreau
Differential Revision: https://reviews.llvm.org/D90879
For consistency with the IRBuilder, OpenMPIRBuilder has method names starting with 'Create'. However, the LLVM coding style has methods names starting with lower case letters, as all other OpenMPIRBuilder already methods do. The clang-tidy configuration used by Phabricator also warns about the naming violation, adding noise to the reviews.
This patch renames all `OpenMPIRBuilder::CreateXYZ` methods to `OpenMPIRBuilder::createXYZ`, and updates all in-tree callers.
I tested check-llvm, check-clang, check-mlir and check-flang to ensure that I did not miss a caller.
Reviewed By: mehdi_amini, fghanim
Differential Revision: https://reviews.llvm.org/D91109
Prior to D89768, any alloca that's used after suspension points will be put on to the coroutine frame, and hence they will always be reloaded in the resume function.
However D89768 introduced a more precise way to determine whether an alloca should live on the frame. Allocas that are only used within one suspension region (hence does not need to live across suspension points) will not be put on the frame. They will remain local to the resume function.
When creating the new entry for the .resume function, the existing logic only moved all the allocas from the old entry to the new entry. This covers every alloca from the old entry. However allocas that's defined afer coro.begin are put into a separate basic block during CoroSplit (the PostSpill basic block). We need to make sure these allocas are moved to the new entry as well if they are used.
This patch walks through all allocas, and check if they are still used but are not reachable from the new entry, if so, we move them to the new entry.
Differential Revision: https://reviews.llvm.org/D90977
Introduce struct FlattenInfo to group some of the bookkeeping. Besides this
being a bit of a clean-up, it is a prep step for next additions (D90640). I
could take things a bit further, but thought this was a good first step also
not to make this change too large.
Differential Revision: https://reviews.llvm.org/D90408
This patch turns VPWidenCall into a VPValue and uses it
during VPlan construction and codegeneration instead of the plain IR
reference where possible.
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D84681
Feeding vector values to `InstCombiner::OptimizeOverflowCheck` produces a scalar boolean flag if it proves the overflow check can be eliminated.
This causes `InstCombiner::CreateOverflowTuple` to crash as it correctly expects a vector of i1 values instead.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D89628
As raised by @nlopes on D90382 - if this is not a rotate then the select was blocking poison from the 'shift-by-zero' non-TVal, but a funnel shift won't - so freeze it.
From C11 and C++11 onwards, a forward-progress requirement has been
introduced for both languages. In the case of C, loops with non-constant
conditionals that do not have any observable side-effects (as defined by
6.8.5p6) can be assumed by the implementation to terminate, and in the
case of C++, this assumption extends to all functions. The clang
frontend will emit the `mustprogress` function attribute for C++
functions (D86233, D85393, D86841) and emit the loop metadata
`llvm.loop.mustprogress` for every loop in C11 or later that has a
non-constant conditional.
This patch modifies LoopDeletion so that only loops with
the `llvm.loop.mustprogress` metadata or loops contained in functions
that are required to make progress (`mustprogress` or `willreturn`) are
checked for observable side-effects. If these loops do not have an
observable side-effect, then we delete them.
Loops without observable side-effects that do not satisfy the above
conditions will not be deleted.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86844
Currently, LoopDeletion refuses to remove dead loops with no exit blocks
because it cannot statically determine the control flow after it removes
the block. This leads to miscompiles if the loop is an infinite loop and
should've been removed.
Differential Revision: https://reviews.llvm.org/D90115
Results of convergent operations are implicitly affected by the
enclosing control flows and should not be hoisted out of arbitrary
loops.
Patch by Xiaoqing Wu <xiaoqing_wu@apple.com>
Differential Revision: https://reviews.llvm.org/D90361
The `llvm.coro.suspend.async` intrinsic takes a function pointer as its
argument that describes how-to restore the current continuation's
context from the context argument of the continuation function. Before
we assumed that the current context can be restored by loading from the
context arguments first pointer field (`first_arg->caller_context`).
This allows for defining suspension points that reuse the current
context for example.
Also:
llvm.coro.id.async lowering: Add llvm.coro.preprare.async intrinsic
Blocks inlining until after the async coroutine was split.
Also, change the async function pointer's context size position
struct async_function_pointer {
uint32_t relative_function_pointer_to_async_impl;
uint32_t context_size;
}
And make the position of the `async context` argument configurable. The
position is specified by the `llvm.coro.id.async` intrinsic.
rdar://70097093
Differential Revision: https://reviews.llvm.org/D90783
This patch changes the type of Start, End in VFRange to be an ElementCount
instead of `unsigned`. This is done as preparation to make VPlans for
scalable vectors, but is otherwise NFC.
Reviewed By: dmgreen, fhahn, vkmr
Differential Revision: https://reviews.llvm.org/D90715
This moves WidenIV from IndVarSimplify to Utils/SimplifyIndVar so that we have
createWideIV available as a generic helper utility. I.e., this is not only
useful in IndVarSimplify, but could be useful for loop transformations. For
example, motivation for this refactoring is the loop flatten transformation: if
induction variables in a loop nest can be widened, we can avoid having to
perform certain overflow checks, enabling this transformation.
Differential Revision: https://reviews.llvm.org/D90421
When replacing an assume(false) with a store, we have to be more careful
with the order we insert the new access. This patch updates the code to
look at the accesses in the block to find a suitable insertion point.
Alterantively we could check the defining access of the assume, but IIRC
there has been some discussion about making assume() readnone, so
looking at the access list might be more future proof.
Fixes PR48072.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D90784
This patch adds the `async` lowering of coroutines.
This will be used by the Swift frontend to lower async functions. In
contrast to the `retcon` lowering the frontend needs to be in control
over control-flow at suspend points as execution might be suspended at
these points.
This is very much work in progress and the implementation will change as
it evolves with the frontend. As such the documentation is lacking
detail as some of it might change.
rdar://70097093
Reapply with fix for memory sanitizer failure and sphinx failure.
Differential Revision: https://reviews.llvm.org/D90612
This patch adds the `async` lowering of coroutines.
This will be used by the Swift frontend to lower async functions. In
contrast to the `retcon` lowering the frontend needs to be in control
over control-flow at suspend points as execution might be suspended at
these points.
This is very much work in progress and the implementation will change as
it evolves with the frontend. As such the documentation is lacking
detail as some of it might change.
rdar://70097093
Differential Revision: https://reviews.llvm.org/D90612
This is slightly better compile-time wise,
since we avoid potentially-costly knownbits analysis that will
ultimately not allow us to actually do anything with said `add`.
This reverts commit 59b22e495c.
That commit broke building for ARM and AArch64, reproducible like this:
$ cat apedec-reduced.c
a;
b(e) {
int c;
unsigned d = f();
c = d >> 32 - e;
return c;
}
g() {
int h = i();
if (a)
h = h << a | b(a);
return h;
}
$ clang -target aarch64-linux-gnu -w -c -O3 apedec-reduced.c
clang: ../lib/Transforms/InstCombine/InstructionCombining.cpp:3656: bool llvm::InstCombinerImpl::run(): Assertion `DT.dominates(BB, UserParent) && "Dominance relation broken?"' failed.
Same thing for e.g. an armv7-linux-gnueabihf target.
There is already an API in BasicBlock that checks and returns the musttail call if it precedes the return instruction.
Use it instead of manually checking in each place.
Differential Revision: https://reviews.llvm.org/D90693
InstCombine is quite aggressive in doing the opposite transform,
folding `add` of operands with no common bits set into an `or`,
and that not many things support that new pattern..
In this case, teaching Reassociate about it is easy,
there's preexisting art for `sub`/`shl`:
just convert such an `or` into an `add`:
https://rise4fun.com/Alive/Xlyv
The LoopDistribute pass is missing from the LTO pipeline, so
-enable-loop-distribute has no effect during post-link. The pre-link
loop distribution doesn't seem to survive the LTO pipeline either.
With this patch (and -flto -mllvm -enable-loop-distribute) we see a 43%
uplift on SPEC 2006 hmmer for AArch64. The rest of SPECINT 2006 is
unaffected.
Differential Revision: https://reviews.llvm.org/D89896
This relaxes one-use restriction on that `sub` fold,
since apparently the addition of Negator broke
preexisting `C-(C2-X) --> X+(C-C2)` (with C=0) fold.
Vectors where all elements have the same known constant range are treated as a
single constant range in the lattice. When bitcasting such vectors, there is a
mis-match between the width of the lattice value (single constant range) and
the original operands (vector). Go to overdefined in that case.
Fixes PR47991.
The fold currently only handles rotation patterns, but with the maturation of backend funnel shift handling we can now realistically handle all funnel shift patterns.
This should allow us to begin resolving PR46896 et al.
Differential Revision: https://reviews.llvm.org/D90625
If we know that some check will not be executed on the last iteration, we can use this
fact to eliminate its check.
Differential Revision: https://reviews.llvm.org/D88210
Reviwed By: ebrevnov
This patch enhances computeOutliningColdRegionsInfo() to allow it to
consider regions containing a single basic block and a single
predecessor as candidate for partial inlining.
Reviewed By: fhann
Differential Revision: https://reviews.llvm.org/D89911
This reverts the revert commit 408c4408fa.
This version of the patch includes a fix for a crash caused by
treating ICmp/FCmp constant expressions as instructions.
Original message:
On some targets, like AArch64, vector selects can be efficiently lowered
if the vector condition is a compare with a supported predicate.
This patch adds a new argument to getCmpSelInstrCost, to indicate the
predicate of the feeding select condition. Note that it is not
sufficient to use the context instruction when querying the cost of a
vector select starting from a scalar one, because the condition of the
vector select could be composed of compares with different predicates.
This change greatly improves modeling the costs of certain
compare/select patterns on AArch64.
I am also planning on putting up patches to make use of the new argument in
SLPVectorizer & LV.
Similar to -fprofile-generate=, add -fmemory-profile= which takes a
directory path. This is passed down to LLVM via a new module flag
metadata. LLVM in turn provides this name to the runtime via the new
__memprof_profile_filename variable.
Additionally, always pass a default filename (in $cwd if a directory
name is not specified vi the = form of the option). This is also
consistent with the behavior of the PGO instrumentation. Since the
memory profiles will generally be fairly large, it doesn't make sense to
dump them to stderr. Also, importantly, the memory profiles will
eventually be dumped in a compact binary format, which is another reason
why it does not make sense to send these to stderr by default.
Change the existing memprof tests to specify log_path=stderr when that
was being relied on.
Depends on D89086.
Differential Revision: https://reviews.llvm.org/D89087
This patch updates DSE + MemorySSA to use the same check as the legacy
implementation to determine if a location is killed by a free call.
This changes the existing behavior so that a free does not kill
locations before the start of the freed pointer.
This should fix PR48036.
This reverts the revert commit a1b53db324.
This patch includes a fix for a reported issue, caused by
matchSelectPattern returning UMIN for selects of pointers in
some cases by looking to some connected casts.
For now, ensure integer instrinsics are only returned for selects of
ints or int vectors.
This is the last of the rotate->funnel shift InstCombine generalizations for PR46896
We still have foldGuardedRotateToFunnelShift to deal with in AggressiveInstCombine
Differential Revision: https://reviews.llvm.org/D90382
Previously, !noalias and !alias.scope metadata on the call site was
applied as part of CloneAliasScopeMetadata(), which short-circuits
if the callee does not use any noalias metadata itself. However,
these two things have no relation to each other.
Consistently apply !noalias and !alias.scope metadata by integrating
this into an existing function that handled !llvm.access.group and
!llvm.mem.parallel_loop_access metadata. The handling for all of
these metadata kinds essentially the same.
This reverts commit 1922570489.
This appears to cause a crash in the following example
a, b, c;
l() {
int e = a, f = l, g, h, i, j;
float *d = c, *k = b;
for (;;)
for (; g < f; g++) {
k[h] = d[i];
k[h - 1] = d[j];
h += e << 1;
i += e;
}
}
clang -cc1 -triple i386-unknown-linux-gnu -emit-obj -target-cpu pentium-m -O1 -vectorize-loops -vectorize-slp reduced.c
llvm::Type *llvm::Type::getWithNewBitWidth(unsigned int) const: Assertion `isIntOrIntVectorTy() && "Original type expected to be a vector of integers or a scalar integer."' failed.
Add support for match-all tags and GOT-free runtime calls, which
are both required for the kernel to be able to support outlined
checks. This requires extending the access info to let the backend
know when to enable these features. To make the code easier to maintain
introduce an enum with the bit field positions for the access info.
Allow outlined checks to be enabled with -mllvm
-hwasan-inline-all-checks=0. Kernels that contain runtime support for
outlined checks may pass this flag. Kernels lacking runtime support
will continue to link because they do not pass the flag. Old versions
of LLVM will ignore the flag and continue to use inline checks.
With a separate kernel patch [1] I measured the code size of defconfig
+ tag-based KASAN, as well as boot time (i.e. time to init launch)
on a DragonBoard 845c with an Android arm64 GKI kernel. The results
are below:
code size boot time
before 92824064 6.18s
after 38822400 6.65s
[1] https://linux-review.googlesource.com/id/I1a30036c70ab3c3ee78d75ed9b87ef7cdc3fdb76
Depends on D90425
Differential Revision: https://reviews.llvm.org/D90426
This is a workaround for poor heuristics in the backend where we can
end up materializing the constant multiple times. This is particularly
bad when using outlined checks because we materialize it for every call
(because the backend considers it trivial to materialize).
As a result the field containing the shadow base value will always
be set so simplify the code taking that into account.
Differential Revision: https://reviews.llvm.org/D90425
CallInst::updateProfWeight() creates branch_weights with i64 instead of i32.
To be more consistent everywhere and remove lots of casts from uint64_t
to uint32_t, use i64 for branch_weights.
Reviewed By: davidxl
Differential Revision: https://reviews.llvm.org/D88609
This patch modifies two for loops to use the range based syntax.
Since they are equivalent, this patch is tagged NFC.
Differential Revision: https://reviews.llvm.org/D90069
- As convergent intrinsics/calls could only be moved to
control-equivalent blocks, or more precisely the same divergent
branch, PRE needs to skip them.
Differential Revision: https://reviews.llvm.org/D90391
Currently isOverwrite returns OW_MaybePartial even for accesss known not to overlap. This is not a big problem for legacy implementation (since isPartialOverwrite follows isOverwrite and clarifies the result). Contrary SSA based version does a lot of work to later find out that accesses don't overlap. Besides negative impact on compile time we quickly reach MemorySSAPartialStoreLimit and miss optimization opportunities.
Note: In fact, I think it would be cleaner implementation if isOverwrite returned fully clarified result in the first place whithout need to call isPartialOverwrite. This can be done as a follow up. What do you think?
Reviewed By: fhahn, asbirlea
Differential Revision: https://reviews.llvm.org/D90371
As per the comment in VPRecipeBase, clients should not rely on
getVPRecipeID, as it may change in the future. It should only be used in
classof implementations. Use isa instead in getFirstNonPhi.
On some targets, like AArch64, vector selects can be efficiently lowered
if the vector condition is a compare with a supported predicate.
This patch adds a new argument to getCmpSelInstrCost, to indicate the
predicate of the feeding select condition. Note that it is not
sufficient to use the context instruction when querying the cost of a
vector select starting from a scalar one, because the condition of the
vector select could be composed of compares with different predicates.
This change greatly improves modeling the costs of certain
compare/select patterns on AArch64.
I am also planning on putting up patches to make use of the new argument in
SLPVectorizer & LV.
Reviewed By: dmgreen, RKSimon
Differential Revision: https://reviews.llvm.org/D90070
Currently we fail to eliminate some noop stores if there is a kill-able
store between the starting def and the load. This is because we
eliminate noop stores first.
In practice it seems like eliminating noop stores after the main
elimination for a def covers slightly more cases.
This patch improves the number of stores slightly in 2 cases for X86 -O3
-flto
Same hash: 235 (filtered out)
Remaining: 2
Metric: dse.NumRedundantStores
Program base patch diff
test-suite...ce/Benchmarks/PAQ8p/paq8p.test 2.00 3.00 50.0%
test-suite...006/453.povray/453.povray.test 18.00 21.00 16.7%
There might be other phase ordering issues, but it appears that they do
not show up in the test-suite/SPEC2000/SPEC2006. We can always tune the
ordering later.
Partly fixes PR47887.
Reviewed By: asbirlea, zoecarver
Differential Revision: https://reviews.llvm.org/D89650
And use it to model LLVM IR's `ptrtoint` cast.
This is essentially an alternative to D88806, but with no chance for
all the problems it caused due to having the cast as implicit there.
(see rG7ee6c402474a2f5fd21c403e7529f97f6362fdb3)
As we've established by now, there are at least two reasons why we want this:
* It will allow SCEV to actually model the `ptrtoint` casts
and their operands, instead of treating them as `SCEVUnknown`
* It should help with initial problem of PR46786 - this should eventually allow us
to not loose pointer-ness of an expression in more cases
As discussed in [[ https://bugs.llvm.org/show_bug.cgi?id=46786 | PR46786 ]], in principle,
we could just extend `SCEVUnknown` with a `is ptrtoint` cast, because `ScalarEvolution::getPtrToIntExpr()`
should sink the cast as far down into the expression as possible,
so in the end we should always end up with `SCEVPtrToIntExpr` of `SCEVUnknown`.
But i think that it isn't the best solution, because it doesn't really matter
from memory consumption side - there probably won't be *that* many `SCEVPtrToIntExpr`s
for it to matter, and it allows for much better discoverability.
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D89456
The existing logic in determining whether an alloca should live on the frame only looks explicit def-use relationships. However a value defined by an alloca may be implicitly needed across suspension points, either because an alias has across-suspension-point def-use relationship, or escaped by store/call/memory intrinsics. To properly handle all these cases, we have to properly visit the alloca pointer up-front. Thie patch extends the exisiting alloca use visitor to determine whether an alloca should live on the frame.
Differential Revision: https://reviews.llvm.org/D89768
Use -0.0 instead of 0.0 as the start value. The previous use of 0.0
was fine for all existing uses of this function though, as it is
always generated with fast flags right now, and thus nsz.
Some architectures do not have general vector select instructions (e.g.
AArch64). But some cmp/select patterns can be vectorized using other
instructions/intrinsics.
One example is using min/max instructions for certain patterns.
This patch updates the cost calculations for selects in the SLP
vectorizer to consider using min/max intrinsics.
This patch does not change SLP vectorizer's codegen itself to actually
generate those intrinsics, but relies on the backends to lower the
vector cmps & selects. This keeps things simple on the SLP side and
works well in practice for AArch64.
This exposes additional SLP vectorization opportunities in some
benchmarks on AArch64 (-O3 -flto).
Metric: SLP.NumVectorInstructions
Program base slp diff
test-suite...ications/JM/ldecod/ldecod.test 502.00 697.00 38.8%
test-suite...ications/JM/lencod/lencod.test 1023.00 1414.00 38.2%
test-suite...-typeset/consumer-typeset.test 56.00 65.00 16.1%
test-suite...6/464.h264ref/464.h264ref.test 804.00 822.00 2.2%
test-suite...006/453.povray/453.povray.test 3335.00 3357.00 0.7%
test-suite...CFP2000/177.mesa/177.mesa.test 2110.00 2121.00 0.5%
test-suite...:: External/Povray/povray.test 2378.00 2382.00 0.2%
Reviewed By: RKSimon, samparker
Differential Revision: https://reviews.llvm.org/D89969
When we promote pointer arguments we did compute a wrong offset and use
a wrong type for the array case.
Bug reported and reduced by Whitney Tsang <whitneyt@ca.ibm.com>.
The following constraints hold for swifterror values:
A swifterror value (either the parameter or the alloca) can only
be loaded and stored from, or used as a swifterror argument.
This patch updates instcombine to not try to convert a bitcast of a
function into a bitcast of a swifterror argument.
Reviewed By: rjmccall
Differential Revision: https://reviews.llvm.org/D90258
This reverts commit e038b60d91.
This reverts commit a0d84d8031.
This revert was a mistake. The reason of the failures was
"Use uint64_t for branch weights instead of uint32_t"
Differential Revision: https://reviews.llvm.org/D87832
Instead of getting the defining access we should be able to use
getClobberingMemoryAccess to skip non-aliasing MemoryDefs. No additional
checks should be needed, because we only remove the starting def if it
matches the defining access of the load. All we need to worry about is
that there are no (may)alias stores between the starting def and the
load and getClobberingMemoryAccess should guarantee that.
Partly fixes PR47887.
This improves the number of redundant stores removed in some cases
(numbers below for MultiSource, SPEC2000, SPEC2006 on X86 with -flto
-O3).
Same hash: 226 (filtered out)
Remaining: 11
Metric: dse.NumRedundantStores
Program base patch1 diff
test-suite...:: External/Povray/povray.test 1.00 5.00 400.0%
test-suite...chmarks/MallocBench/gs/gs.test 1.00 3.00 200.0%
test-suite...0/253.perlbmk/253.perlbmk.test 21.00 37.00 76.2%
test-suite...0.perlbench/400.perlbench.test 24.00 37.00 54.2%
test-suite.../Applications/SPASS/SPASS.test 3.00 4.00 33.3%
test-suite...006/453.povray/453.povray.test 15.00 18.00 20.0%
test-suite...T2006/445.gobmk/445.gobmk.test 27.00 29.00 7.4%
test-suite.../CINT2006/403.gcc/403.gcc.test 136.00 137.00 0.7%
test-suite.../CINT2000/176.gcc/176.gcc.test 6.00 6.00 0.0%
test-suite.../Benchmarks/Bullet/bullet.test NaN 3.00 nan%
test-suite.../Benchmarks/Ptrdist/bc/bc.test NaN 1.00 nan%
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D89647
The types of SEH aren't x86(-32) vs x64 but rather stack-based exception chaining
vs table-based exception handling. x86-32 is the only arch for which Windows
uses the former. 32-bit ARM would use what is called Win64SEH today, which
is a bit confusing so instead let's just rename it to be a bit more clear.
Reviewed By: compnerd, rnk
Differential Revision: https://reviews.llvm.org/D90117
This patch removes extraneous calls to setEdgeProbability introduced
in c91487769d.
The follow-up patch, a7b662d0f4, has
since fixed BranchProbabilityInfo::eraseBlock, so we don't need to
worry about getting stale values from getEdgeProbability.
Also, since getEdgeProbability(BB, BB->getSingleSuccessor()) returns
edge probability 1/1 by default for BB with exactly one successor
edge, we don't need to explicitly call setEdgeProbability.
This patch introduces almost no functional change, but we do end up
reducing debug messages from setEdgeProbability.
Differential Revision: https://reviews.llvm.org/D90284
Before we used to only mark unreachable static functions as dead if all
uses were known dead. Now we optimistically assume uses to be dead until
proven otherwise.
If we are looking at a call site argument it might be a load or call
which is in a different context than the call site argument. We cannot
simply use the call site argument range for the call or load.
Bug reported and reduced by Whitney Tsang <whitneyt@ca.ibm.com>.
In the AANoAlias logic we determine if a pointer may have been captured
before a call. We need to look at other uses in the call not uses of the
call.
The new code is not perfect as it does not allow trivial cases where the
call has multiple arguments but it is at least not unsound and a TODO
was added.
This patch teaches the jump threading pass to set edge probabilities
whenever the pass creates new basic blocks.
Without this patch, the compiler sometimes produces non-deterministic
results. The non-determinism comes from the jump threading pass using
stale edge probabilities in BranchProbabilityInfo. Specifically, when
the jump threading pass creates a new basic block, we don't initialize
its outgoing edge probability.
Edge probabilities are maintained in:
DenseMap<Edge, BranchProbability> Probs;
in class BranchProbabilityInfo, where Edge is an ordered pair of
BasicBlock * and a successor index declared as:
using Edge = std::pair<const BasicBlock *, unsigned>;
Probs maps edges to their corresponding probabilities.
Now, we rarely remove entries from this map, so if we happen to
allocate a new basic block at the same address as a previously deleted
basic block with an edge probability assigned, the newly created basic
block appears to have an edge probability, albeit a stale one.
This patch fixes the problem by explicitly setting edge probabilities
whenever the jump threading pass creates new basic blocks.
Differential Revision: https://reviews.llvm.org/D90106
Summary:
This patch adds support for passing in the original delcaration name in the
source file to the libomptarget runtime. This will allow the runtime to provide
more intelligent debugging messages. This patch takes the original expression
parsed from the OpenMP map / update clause and provides a textual
representation if it was explicitly mapped, otherwise it takes the name of the
variable declaration as a fallback. The information in passed to the runtime in
a global array of strings that matches the existing ident_t source location
strings using ";name;filename;column;row;;". See
clang/test/OpenMP/target_map_names.cpp for an example of the generated output
for a given map clause.
Reviewers: jdoervert
Differential Revision: https://reviews.llvm.org/D89802
These logically belong together since it's a base commit plus
followup fixes to less common build configurations.
The patches are:
Revert "CfgInterface: rename interface() to getInterface()"
This reverts commit a74fc48158.
Revert "Wrap CfgTraitsFor in namespace llvm to please GCC 5"
This reverts commit f2a06875b6.
Revert "Try to make GCC5 happy about the CfgTraits thing"
This reverts commit 03a5f7ce12.
Revert "Introduce CfgTraits abstraction"
This reverts commit c0cdd22c72.
This patch changes MergeBlockIntoPredecessor to skip the call to
RemoveRedundantDbgInstrs, in effect partially reverting D71480 due to
some compile-time issues spotted in LoopUnroll and SimplifyCFG.
The call to RemoveRedundantDbgInstrs appears to have changed the
worst-case behavior of the merging utility. Loosely speaking, it seems
to have gone from O(#phis) to O(#insts).
It might not be possible to mitigate this by scanning a block to
determine whether there are any debug intrinsics to remove, since such a
scan costs O(#insts).
So: skip the call to RemoveRedundantDbgInstrs. There's surprisingly
little fallout from this, and most of it can be addressed by doing
RemoveRedundantDbgInstrs later. The exception is (the block-local
version of) SimplifyCFG, where it might just be too expensive to call
RemoveRedundantDbgInstrs.
Differential Revision: https://reviews.llvm.org/D88928
This reverts commit c6ca26c0bf.
This breaks stage2 builds due to hitting this assert:
```
Assertion failed: (WeightSum <= UINT32_MAX && "Expected weights to scale down to 32 bits"), function calcMetadataWeights
```
when compiling AArch64RegisterBankInfo.cpp in LLVM.
-Oz normally does not allow loop header duplication so this loop wouldn't be
vectorized. However the vectorization pragma should override this and allow
for loop rotation.
rdar://problem/49281061
Original patch by Adam Nemet.
Reviewed By: Meinersbur
Differential Revision: https://reviews.llvm.org/D59832
GVN Load PRE can split the backedge causing breaking the loop structure where the latch
contains the conditional branch with for example induction variable.
Different optimizations expect this form of the loop, so it is better to preserve it for some time.
This CL adds an option to control an ability to split backedge.
Default value is true so technically it is NFC and current behavior is not changed.
Reviewers: fedor.sergeev, mkazantsev, nikic, reames, fhahn
Reviewed By: mkazasntsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D89854
Even if the exact exit count is unknown, we can still prove that this
exit will not be taken. If we can prove that the predicate is monotonic,
fulfilled on first & last iteration, and no overflow happened in between,
then the check can be removed.
Differential Revision: https://reviews.llvm.org/D87832
Reviewed By: apilipenko
CallInst::updateProfWeight() creates branch_weights with i64 instead of i32.
To be more consistent everywhere and remove lots of casts from uint64_t
to uint32_t, use i64 for branch_weights.
Reviewed By: davidxl
Differential Revision: https://reviews.llvm.org/D88609
`-separate-const-offset-from-gep` has not yet be ported, so some tests are not updated.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D90149
Prepend the module name hash with a fixed string ".__uniq." which helps tools
that consume sampled profiles and attribute it to functions to understand
that this symbol belongs to a unique internal linkage type symbol.
Symbols with suffixes can result from various optimizations in the compiler.
Function Multiversioning, function splitting, parameter constant propogation,
unique internal linkage names.
External tools like sampled profile aggregators combine profiles from multiple
runs of a binary. They use various heuristics with symbols that have suffixes
to try and attribute the profile to the right function instance. For instance
multi-versioned symbols like foo.avx, foo.sse4.2, etc even though different
should be attributed to the same source function if a single function is
versioned, using attribute target_clones (supported in GCC but yet to land in
LLVM). Similarly, functions that are split (split part having a .cold suffix)
could have profiles for both the original and split symbols but would be
aggregated and attributed to the original function that was split.
Unique internal linkage functions however have different source instances and
the aggregator must not put them together but attribute it to the appropriate
function instance. To be sure that we are dealing with a symbol of a unique
internal linkage function, we would like to prepend the hash with a known
string ".__uniq." which these tools can check to understand the suffix type.
Differential Revision: https://reviews.llvm.org/D89617
This fixes the bug 47945. It is legal to have a PHI with values
from from the same block, but values must stay the same. In this
case it is illegal to merge different values.
Differential Revision: https://reviews.llvm.org/D89978
The warning would fire when calling canReplaceGEPIdxWithZero on a GEP
whose source element type is a scalable vector. The size of scalable
vector types is not known, so this optimization cannot be performed.
This patch fixes the issue by:
- bailing out early in this routine if the GEP instruction's source
element type is a scalable vector.
- making use of getFixedSize -- this removes the dependency on the
deprecated interface.
Reviewed By: fpetrogalli
Differential Revision: https://reviews.llvm.org/D89968
The warning would fire when calling isDereferenceableAndAlignedInLoop
with a scalable load. Calling isDereferenceableAndAlignedInLoop with a
scalable load would result in the use of the now deprecated implicit
cast of TypeSize to uint64_t through the overloaded operator.
This patch fixes this issue by:
- no longer considering vector loads as candidates in
canVectorizeWithIfConvert. This doesn't make sense in the context of
identifying scalar loads to vectorize.
- making use of getFixedSize inside isDereferenceableAndAlignedInLoop --
this removes the dependency on the deprecated interface, and will
trigger an assertion error if the function is ever called with a
scalable type.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D89798
I'm not certain InstCombinerImpl::matchBSwapOrBitReverse needs to filter the or(op0(),op1()) ops - there are just too many cases that recognizeBSwapOrBitReverseIdiom/collectBitParts handle now (and quickly).
The exit blocks of the versioned and non-versioned loops are not dedicated and thus the two loops are not in simplify form.
Insert dummy exit blocks after loop versioning with `formDedicatedExits()` to preserve the simplify form for subsequence passes.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D89569
As discussed on PR35155, this extends narrowFunnelShift (recently renamed from narrowRotate) to support basic funnel shift patterns.
Unlike matchFunnelShift we don't include the computeKnownBits limitation as extracting the pattern from the zext/trunc layers should be a indicator of reasonable funnel shift codegen, in D89139 we demonstrated how to efficiently promote funnel shifts to wider types.
Differential Revision: https://reviews.llvm.org/D89542
Duplicated callsites share the same callee profile if the original callsite was inlined. The sharing also causes the profile of callee's callee to be shared. This breaks the assert introduced ealier by D84997 in a tricky way.
To illustrate, I'm using an abstract example. Say we have three functions `A`, `B` and `C`. A calls B twice and B calls C once. Some optimize performed prior to the sample profile loader duplicates first callsite to `B` and the program may look like
```
A()
{
B(); // with nested profile B1 and C1
B(); // duplicated, with nested profile B1 and C1
B(); // with nested profile B2 and C2
}
```
For some reason, the sample profile loader inliner then decides to only inline the first callsite in `A` and transforms `A` into
```
A()
{
C(); // with nested profile C1
B(); // duplicated, with nested profile B1 and C1
B(); // with nested profile B2 and C2.
}
```
Here is what happens next:
1. Failing to inline the callsite `C()` results in `C1`'s samples returned to `C`'s base (outlined) profile. In the meantime, `C1`'s head samples are updated to `C1`'s entry sample. This also affects the profile of the middle callsite which shares `C1` with the first callsite.
2. Failing to inline the middle callsite results in `B1` returned to `B`'s base profile, which in turn will cause `C1` merged into `B`'s base profile. Note that the nest `C` profile in `B`'s base has a non-zero head sample count now. The value actually equals to `C1`'s entry count.
3. Failing to inline last callsite results in `B2` returned to `B`'s base profile. Note that the nested `C` profile in `B`'s base now has an entry count equal to the sum of that of `C1` and `C2`, with the head count equal to that of `C1`. This will trigger the assert later on.
4. Compiling `B` using `B`'s base profile. Failing to inline `C` there triggers the returning of the nested `C` profile. Since the nested `C` profile has a non-zero head count, the returning doesn't go through. Instead, the assert goes off.
It's good that `C1` is only returned once, based on using a non-zero head count to ensure an inline profile is only returned once. However C2 is never returned. While it seems hard to solve this perfectly within the current framework, I'm just removing the broken assert. This should be reasonably fixed by the upcoming CSSPGO work where counts returning is based on context-sensitivity and a distribution factor for callsite probes.
The simple example is extracted from one of our internal services. In reality, why the original callsite `B()` and duplicate one having different inline behavior is a magic. It has to do with imperfect counts in profile and extra complicated inlining that makes the hotness for them different.
Reviewed By: wenlei
Differential Revision: https://reviews.llvm.org/D90056
This doesn't support -structurizecfg-skip-uniform-regions since that
would require porting LegacyDivergenceAnalysis.
The NPM doesn't support adding a non-analysis pass as a dependency of
another, so I had to add -lowerswitch to some tests or pin them to the
legacy PM.
This is the only RegionPass in tree, so I simply copied the logic for
finding all Regions from the legacy PM's RGManager into
StructurizeCFG::run().
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D89026
This change introduces a GC parseable lowering for element atomic
memcpy/memmove intrinsics. This way runtime can provide an
implementation which can take a safepoint during copy operation.
See "GC-parseable element atomic memcpy/memmove" thread on llvm-dev
for the background and details:
https://groups.google.com/g/llvm-dev/c/NnENHzmX-b8/m/3PyN8Y2pCAAJ
Differential Revision: https://reviews.llvm.org/D88861
It's currently ambiguous in IR whether the source language explicitly
did not want a stack a stack protector (in C, via function attribute
no_stack_protector) or doesn't care for any given function.
It's common for code that manipulates the stack via inline assembly or
that has to set up its own stack canary (such as the Linux kernel) would
like to avoid stack protectors in certain functions. In this case, we've
been bitten by numerous bugs where a callee with a stack protector is
inlined into an __attribute__((__no_stack_protector__)) caller, which
generally breaks the caller's assumptions about not having a stack
protector. LTO exacerbates the issue.
While developers can avoid this by putting all no_stack_protector
functions in one translation unit together and compiling those with
-fno-stack-protector, it's generally not very ergonomic or as
ergonomic as a function attribute, and still doesn't work for LTO. See also:
https://lore.kernel.org/linux-pm/20200915172658.1432732-1-rkir@google.com/https://lore.kernel.org/lkml/20200918201436.2932360-30-samitolvanen@google.com/T/#u
Typically, when inlining a callee into a caller, the caller will be
upgraded in its level of stack protection (see adjustCallerSSPLevel()).
By adding an explicit attribute in the IR when the function attribute is
used in the source language, we can now identify such cases and prevent
inlining. Block inlining when the callee and caller differ in the case that one
contains `nossp` when the other has `ssp`, `sspstrong`, or `sspreq`.
Fixes pr/47479.
Reviewed By: void
Differential Revision: https://reviews.llvm.org/D87956
matchBSwapOrBitReverse was hardcoded to just match bswaps - we're going to need to expose the ability to match bitreverse as well, so make this part of the function call.
This patch copies @vsk's fix to instcombine from D85555 over to mem2reg. The
motivation and rationale are exactly the same: When mem2reg removes an alloca,
it erases the dbg.{addr,declare} instructions which refer to the alloca. It
would be better to instead remove all debug intrinsics which describe the
contents of the dead alloca, namely all dbg.value(<dead alloca>, ...,
DW_OP_deref)'s.
As far as I can tell, prior to D80264 these `dbg.value+deref`s would have been
silently dropped instead of being made `undef`, so we're just returning to
previous behaviour with these patches.
Testing:
`llvm-lit llvm/test` and `ninja check-clang` gave no unexpected failures. Added
3 tests, each of which covers a dbg.value deletion path in mem2reg:
mem2reg-promote-alloca-1.ll
mem2reg-promote-alloca-2.ll
mem2reg-promote-alloca-3.ll
The first is based on the dexter test inlining.c from D89543. This patch also
improves the debugging experience for loop.c from D89543, which suffers
similarly after arg promotion instead of inlining.
Use isKnownXY comparators when one of the operands can be with
scalable vectors or getFixedSize() for all the other cases.
This patch also does bug fixes for getPrimitiveSizeInBits by using
getFixedSize() near the places with the TypeSize comparison.
Differential Revision: https://reviews.llvm.org/D89703
We want to have a caching version of symbolic BE exit count
rather than recompute it every time we need it.
Differential Revision: https://reviews.llvm.org/D89954
Reviewed By: nikic, efriedma
An alwaysinline function may not get inlined in inliner-wrapper due to
the inlining order.
Previously for the following, the inliner would first inline @a() into @b(),
```
define void @a() {
entry:
call void @b()
ret void
}
define void @b() alwaysinline {
entry:
br label %for.cond
for.cond:
call void @a()
br label %for.cond
}
```
making @b() recursive and unable to be inlined into @a(), ending at
```
define void @a() {
entry:
call void @b()
ret void
}
define void @b() alwaysinline {
entry:
br label %for.cond
for.cond:
call void @b()
br label %for.cond
}
```
Running always-inliner first makes sure that we respect alwaysinline in more cases.
Fixes https://bugs.llvm.org/show_bug.cgi?id=46945.
Reviewed By: davidxl, rnk
Differential Revision: https://reviews.llvm.org/D86988
This reverts commit 26ee8aff2b.
It's necessary to insert bitcast the pointer operand of a lifetime
marker if it has an opaque pointer type.
rdar://70560161
When performing a call slot optimization to a GEP destination, it
will currently usually fail, because the GEP is directly before the
memcpy and as such does not dominate the call. We should move it
above the call if that satisfies the domination requirement.
I think that a constant-index GEP is the only useful thing to move
here, as otherwise isDereferenceablePointer couldn't look through
it anyway. As such I'm not trying to generalize this further.
Differential Revision: https://reviews.llvm.org/D89623
Make member function const where possible, use LLVM_DEBUG to print debug traces
rather than a custom option, pass by reference to avoid null checking, ...
Reviewed By: fhann
Differential Revision: https://reviews.llvm.org/D89895
When InstCombine removes an alloca, it erases the dbg.{addr,declare}
instructions which refer to the alloca. It would be better to instead
remove all debug intrinsics which describe the contents of the dead
alloca, namely all dbg.value(<dead alloca>, ..., DW_OP_deref)'s.
This effectively undoes work performed in an InstCombine run earlier in
the pipeline by LowerDbgDeclare, which inserts DW_OP_deref dbg.values
before CallInst users of an alloca. The motivating example looks like:
```
define void @foo(i32 %0) {
%a = alloca i32 ; This alloca is erased.
store i32 %0, i32* %a
dbg.value(i32 %0, "arg0") ; This dbg.value survives.
dbg.value(i32* %a, "arg0", DW_OP_deref)
call void @trivially_inlinable_no_op(i32* %a)
ret void
}
```
If the DW_OP_deref dbg.value is not erased, it becomes dbg.value(undef)
after inlining, making "arg0" unavailable. But we already have dbg.value
descriptions of the alloca's value (from LowerDbgDeclare), so the
DW_OP_deref dbg.value cannot serve its purpose of describing an
initialization of the alloca by some callee. It invalidates other useful
dbg.values, causing large gaps in location coverage, so we should delete
it (even though doing so may cause stale dbg.values to appear, if
there's a dead store to `%a` in @trivially_inlinable_no_op).
OTOH, it wouldn't be correct to delete all dbg.value descriptions of an
alloca. Note that it's possible to describe a variable that takes on
different pointer values, e.g.:
```
void use(int *);
void t(int a, int b) {
int *local = &a; // dbg.value(i32* %a.addr, "local")
local = &b; // dbg.value(i32* undef, "local")
use(&a); // (note: %b.addr is optimized out)
local = &a; // dbg.value(i32* %a.addr, "local")
}
```
In this example, the alloca for "b" is erased, but we need to describe
the value of "local" as <unavailable> before the call to "use". This
prevents "local" from appearing to be equal to "&a" at the callsite.
rdar://66592859
Differential Revision: https://reviews.llvm.org/D85555
Use BFI if it is available and BPI otherwise.
This is a promised follow-up after D89541.
Reviewers: ebrevnov, mkazantsev
Reviewed By: ebrevnov
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D89773
PreservedCFGCheckerInstrumentation was saying that LowerMatrixIntrinsics
didn't properly preserve CFG even though it claimed to. The legacy pass
says it doesn't. Match the legacy pass's preserved analyses.
Reviewed By: thakis
Differential Revision: https://reviews.llvm.org/D89175
For GC parseable element atomic memcpy/memmove we'll need to
shuffle statepoint arguments. Make it possible by storing the
arguments as Value *, not Use *.
Avoid having to instantiate and compile a subset of the dominator tree logic
separately for each node type. More importantly, this allows generic
algorithms to be built on top of dominator trees without writing them as
templates -- such algorithms can now use opaque CfgBlockRef and
CfgInterface instead.
A type-erased implementation of dominator trees could be written in
terms of CfgInterface as well, but doing so would change the current
trade-off: it would slightly reduce code size at the cost of a slight
runtime overhead.
This patch does not change the trade-off, as it only does type-erasure
where basic blocks can be treated in a fully opaque way, i.e. it only
moves methods that don't require iteration over CFG successors and
predecessors.
v5:
- rename generic_{begin,end,children} back without the generic_ prefix
and refer explictly to base class methods in NewGVN, which wants to
mutate the order of dominator tree node children directly
v6:
- style change: iDom -> idom; it's arguable whether this is really
invalid, since it is actually standard camelCase, but clang-tidy
complains about it so... *shrug*
- rename {to,from}Generic -> {wrap,unwrap}Ref
Change-Id: Ib860dc04cf8bb093d8ed00be7def40d662213672
Differential Revision: https://reviews.llvm.org/D83089
isMemTerminator checks if the current def is a memory terminator that
terminates the memory pointed to by DefLoc. We do not have to add any of
their users to the worklist, because the follow-on users cannot read the
memory in question.
This leads to more stores eliminated in the presence of lifetime calls.
Previously we added the users of those intrinsics to the worklist,
limiting elimination.
In terms of removed stores, this gives a nice boost on some benchmarks
(MultiSource/SPEC2000/SPEC2006 on X86 with -flto -O3):
Same hash: 205 (filtered out)
Remaining: 32
Metric: dse.NumFastStores
Program base patch diff
test-suite...000/197.parser/197.parser.test 4.00 8.00 100.0%
test-suite...rolangs-C++/family/family.test 4.00 7.00 75.0%
test-suite...marks/7zip/7zip-benchmark.test 1722.00 2189.00 27.1%
test-suite...CFP2000/177.mesa/177.mesa.test 30.00 38.00 26.7%
test-suite :: External/Nurbs/nurbs.test 44.00 49.00 11.4%
test-suite...lications/sqlite3/sqlite3.test 115.00 128.00 11.3%
test-suite...006/447.dealII/447.dealII.test 2715.00 3013.00 11.0%
test-suite...ProxyApps-C++/CLAMR/CLAMR.test 237.00 261.00 10.1%
test-suite...tions/lambda-0.1.3/lambda.test 40.00 44.00 10.0%
test-suite...3.xalancbmk/483.xalancbmk.test 1366.00 1475.00 8.0%
test-suite...abench/jpeg/jpeg-6a/cjpeg.test 13.00 14.00 7.7%
test-suite...oxyApps-C++/miniFE/miniFE.test 43.00 46.00 7.0%
test-suite...lications/ClamAV/clamscan.test 230.00 246.00 7.0%
test-suite...006/450.soplex/450.soplex.test 284.00 299.00 5.3%
test-suite...nsumer-jpeg/consumer-jpeg.test 21.00 22.00 4.8%
The CfgTraits abstraction simplfies writing algorithms that are
generic over the type of CFG, and enables writing such algorithms
as regular non-template code that operates on opaque references
to CFG blocks and values.
Implementations of CfgTraits provide operations on the concrete
CFG types, e.g. `IrCfgTraits::BlockRef` is `BasicBlock *`.
CfgInterface is an abstract base class which provides operations
on opaque types CfgBlockRef and CfgValueRef. Those opaque types
encapsulate a `void *`, but the meaning depends on the concrete
CFG type. For example, MachineCfgTraits -- for use with MachineIR
in SSA form -- encodes a Register inside CfgValueRef. Converting
between concrete references and opaque/generic ones is done by
CfgTraits::{fromGeneric,toGeneric}. Convenience methods
CfgTraits::{un}wrap{Iterator,Range} are available as well.
Writing algorithms in terms of CfgInterface adds some overhead
(virtual method calls, plus in same cases it removes the
opportunity to inline iterators), but can be much more convenient
since generic algorithms can be written as non-templates.
This patch adds implementations of CfgTraits for all CFGs on
which dominator trees are calculated, so that the dominator
tree can be ported to this machinery. Only IrCfgTraits (LLVM IR)
and MachineCfgTraits (Machine IR in SSA form) are complete, the
other implementations are limited to the absolute minimum
required to make the upcoming dominator tree changes work.
v5:
- fix MachineCfgTraits::blockdef_iterator and allow it to iterate over
the instructions in a bundle
- use MachineBasicBlock::printName
v6:
- implement predecessors/successors for all CfgTraits implementations
- fix error in unwrapRange
- rename toGeneric/fromGeneric into wrapRef/unwrapRef to have naming
that is consistent with {wrap,unwrap}{Iterator,Range}
- use getVRegDef instead of getUniqueVRegDef
v7:
- std::forward fix in wrapping_iterator
- fix typos
v8:
- cleanup operators on CfgOpaqueType
- address other review comments
Change-Id: Ia75f4f268fded33fca11218a7d578c9aec1f3f4d
Differential Revision: https://reviews.llvm.org/D83088
This adds the LLVM IR attribute `mustprogress` as defined in LangRef through D86233. This attribute will be applied to functions with in languages like C++ where forward progress is guaranteed. Functions without this attribute are not required to make progress.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D85393
IRCE has some overhead for runtime checks and in case number of iteration is small
the overhead can kill the benefit from optimizations.
This CL bases on BlockFrequencyInfo of pre-header and header to estimate the
number of loop iterations. If it is less than irce-min-estimated-iters we do not transform the loop.
Probably it is better to make more complex cost model but for simplicity it seems the be enough.
The usage of BFI is added only for new pass manager and tries to use it efficiently.
Reviewers: ebrevnov, dantrushin, asbirlea, mkazantsev
Reviewed By: mkazantsev
Subscribers: llvm-commits, fhahn
Differential Revision: https://reviews.llvm.org/D89541
The main tricky thing here is forward-declaring the enum:
we have to specify it's underlying data type.
In particular, this avoids the danger of switching over the SCEVTypes,
but actually switching over an integer, and not being notified
when some case is not handled.
I have updated most of such switches to be exaustive and not have
a default case, where it's pretty obvious to be the intent,
however not all of them.
If we switch over an enum, compiler can easily issue a diagnostic
if some case is not handled. However with an if cascade that isn't so.
Experimental evidence suggests new behavior to be superior.
Fixes a number of stage2 buildbots that were failing when I generalized the m_ConstantInt() logic - that didn't match for pointer types but m_Zero() does......
Scalar cases were already being handled by foldLogOpOfMaskedICmps (so this was dead code), but refactoring to support non-uniform vectors will take some time, so tweak this fold in the meantime.
This broke Chromium's PGO build, it seems because hot-cold-splitting got turned
on unintentionally. See comment on the code review for repro etc.
> This patch adds -f[no-]split-cold-code CC1 options to clang. This allows
> the splitting pass to be toggled on/off. The current method of passing
> `-mllvm -hot-cold-split=true` to clang isn't ideal as it may not compose
> correctly (say, with `-O0` or `-Oz`).
>
> To implement the -fsplit-cold-code option, an attribute is applied to
> functions to indicate that they may be considered for splitting. This
> removes some complexity from the old/new PM pipeline builders, and
> behaves as expected when LTO is enabled.
>
> Co-authored by: Saleem Abdulrasool <compnerd@compnerd.org>
> Differential Revision: https://reviews.llvm.org/D57265
> Reviewed By: Aditya Kumar, Vedant Kumar
> Reviewers: Teresa Johnson, Aditya Kumar, Fedor Sergeev, Philip Pfaffe, Vedant Kumar
This reverts commit 273c299d5d.
All existing SCEV cast types operate on integers.
D89456 will add SCEVPtrToIntExpr cast expression type.
I believe this is best for consistency.
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D89455
isNoopIntrinsic returns true for some intrinsics that are modeled in
MemorySSA but do not actually read or write any memory and do not block
DSE. Such intrinsics should not be considered as read-clobbers.
This patch adds metadata !noundef and makes load instructions can optionally have it.
A load with !noundef always return a well-defined value (has no undef bit or isn't poison).
If the loaded value isn't well defined, the behavior is undefined.
This metadata can be used to encode the assumption from C/C++ that certain reads of variables should have well-defined values.
It is helpful for optimizing freeze instructions away, because freeze can be removed when its operand has well-defined value, and showing that a load from arbitrary location is well-defined is usually hard otherwise.
The same information can be encoded with llvm.assume with operand bundle; using metadata is chosen because I wasn't sure whether code motion can be freely done when llvm.assume is inserted from clang instead.
The existing codebase already is stripping unknown metadata when doing code motion, so using metadata is UB-safe as well.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D89050
We can not bitcast pointers across different address spaces, and VectorCombine
should be careful when it attempts to find the original source of the loaded
data.
Differential Revision: https://reviews.llvm.org/D89577
Logic of widenWithVariantUse is split into check and transform
part, unlike any other transform in IndVars. We want to pass some
extra flags from analysis to transform part and standartize
the code at once, so merging them together.
Variable ExtendOperExpr only exists to check whether it is a SCEV ext.
We create it as SCEV ext right here, so semantically this check is
trivially true. In theory, it may fail if SCEV is smart enough and can
simplify the expression. However, no matter whether it is an ext or not,
we never use this fact for further reasoning. So this code is currently
useless and in theory may become harmful with SCEV's development.
We do not expect any behavior changes with removing it. If it caused
negative changes, the patch should be reverted.
Some facts have already been checked in widenWithVariantUse and then
checked again in widenWithVariantUseCodegen. The latter is redundant,
we can replace it with asserts.
Prep work for PR35155 - renamed narrowRotate to narrowFunnelShift, rewrote some comments and adjusted code to collect separate shift values, although we bail if they don't match (still only rotations are only actually folded).
I'm trying to match matchFunnelShift as much as possible in case we finally get to merge these one day.
After investigation by @asbirlea, the issue that caused the
revert appears to be an issue in the original source, rather
than a problem with the compiler.
This patch enables MemorySSA DSE again.
This reverts commit 915310bf14.
This patch adds -f[no-]split-cold-code CC1 options to clang. This allows
the splitting pass to be toggled on/off. The current method of passing
`-mllvm -hot-cold-split=true` to clang isn't ideal as it may not compose
correctly (say, with `-O0` or `-Oz`).
To implement the -fsplit-cold-code option, an attribute is applied to
functions to indicate that they may be considered for splitting. This
removes some complexity from the old/new PM pipeline builders, and
behaves as expected when LTO is enabled.
Co-authored by: Saleem Abdulrasool <compnerd@compnerd.org>
Differential Revision: https://reviews.llvm.org/D57265
Reviewed By: Aditya Kumar, Vedant Kumar
Reviewers: Teresa Johnson, Aditya Kumar, Fedor Sergeev, Philip Pfaffe, Vedant Kumar
This is an initial cleanup of the way LoopVersioning interacts with LAA.
Currently LoopVersioning has 2 ways of initializing things:
1. Passing LAI and passing UseLAIChecks = true
2. Passing UseLAIChecks = false, followed by calling setSCEVChecks and
setAliasChecks.
Both ways of initializing lead to the same result and the duplication
seems more complicated than necessary.
This patch removes the UseLAIChecks flag from the constructor and the
setSCEVChecks & setAliasChecks helpers and move initialization
exclusively to the constructor.
This simplifies things, by providing a single way to initialize
LoopVersioning and reducing duplication.
Reviewed By: Meinersbur, lebedev.ri
Differential Revision: https://reviews.llvm.org/D84406
Following up D81682 and D83903, remove the code for the old value profiling
buckets, which have been replaced with the new, extended buckets and disabled by
default.
Also syncing InstrProfData.inc between compiler-rt and llvm.
Differential Revision: https://reviews.llvm.org/D88838
Replace m_ConstantInt with m_APInt to support uniform vectors (with no undef elements)
Adding non-undef support would involve some refactoring of the MaskOps struct but this might still be worth it.
This reverts commit 25a97c3a43.
We have other constant folds that fold undef funnel shift amounts to 0 - so we need to be consistent.
If we end up with regressions where we lose a splat shift amount pattern we'll have to investigate other canonicalizations, but matchFunnelShift currently protects us from that.
This was broken by 16295d521e, when
instructions started being handled and not just constant
expressions. This was re-inserting an equivalent bitcast to the
original memcpy operand, which made a non-functional IR change on
every iteration.
This also fixes a secondary problem where it was inserting
addrspacecasts which may not have been legal (i.e. it changed the
source address space). Start visiting all pointer users and fail out
if we can't process them. Also start handling the relevant memory
intrinsic users. These cases can be dealt with by running
InferAddressSpaces separately.
This reverts the revert commit 710aceb645
and includes a fix for a memsan failure.
Original message:
This patch turns VPMemoryInstructionRecipe into a VPValue and uses it
during VPlan construction and codegeneration instead of the plain IR
reference where possible.
m_SpecificInt doesn't accept undef elements in a vector splat value - tweak specific_intval to optionally allow undefs and add the m_SpecificIntAllowUndef variants.
Allows us to remove the m_APIntAllowUndef + comparison hack inside matchFunnelShift
By always performing a modulo on the shift amount constants this was causing undef amounts being replaced with zero, meaning we were losing funnel shift by splat (with undef) patterns.
Tweaked the shift amount bounds check to support (passthrough) undefs, and use Constant::mergeUndefsWith to preserve the undefs after folding.
While we haven't encountered an earth-shattering problem with this yet,
by now it is pretty evident that trying to model the ptr->int cast
implicitly leads to having to update every single place that assumed
no such cast could be needed. That is of course the wrong approach.
Let's back this out, and re-attempt with some another approach,
possibly one originally suggested by Eli Friedman in
https://bugs.llvm.org/show_bug.cgi?id=46786#c20
which should hopefully spare us this pain and more.
This reverts commits 1fb6104293,
7324616660,
aaafe350bb,
e92a8e0c74.
I've kept&improved the tests though.
LV fails with assertion checking that UF > 0. We already set UF to 1 if it is 0 except the case when IC > MaxInterleaveCount. The fix is to set UF to 1 for that case as well.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D87679
Replace m_SpecificInt with m_APIntAllowUndef to matching splats containing undefs, then use ConstantExpr::mergeUndefsWith to merge the undefs together in the result.
The undef funnel shift amounts are getting replaced with zero later on - I'll address this in a later patch, otherwise we lose potential shift by splat value patterns.
D85703 will need to create shallow wrappers in order to track the spmd icv. We need to make it available.
Differential Revision: https://reviews.llvm.org/D89342
-loop-extract-single is just -loop-extract on one loop.
-loop-extract depended on -break-crit-edges and -loop-simplify in the
legacy PM, but the NPM doesn't allow specifying pass dependencies like
that, so manually add those passes to the RUN lines where necessary.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D89016
While promotion currently always has an AST available, it is only
relevant for invalidation purposes in LoopPromoter, so we do not
need to have it as a hard dependency.
This adds an -enable-memcpyopt-memoryssa option that currently does
nothing apart from requiring MSSA as a dependency. The tests are
split to run both with the option disabled and enabled. I went with
this rather than the separate directory DSE uses, as I found it
convenient to have a direct side-by-side comparison of differences.
Differential Revision: https://reviews.llvm.org/D89206
moveUp() moves instructions, so we should move the corresponding
memory accesses as well. We should also move the store instruction
itself: Even though we'll end up removing it later, this gives us
a correct MemoryDef to replace.
The implementation is somewhat more complicated than it should be,
because we also handle the case where P does not have a memory
access due to a degnerate AA pipeline. Hopefully, the need for this
will go away in the future, when the rest of the pass is based on
MSSA.
Differential Revision: https://reviews.llvm.org/D88778
If the memcpy operands are the same (which is allowed since D86815)
then the memcpy is effectively a no-op and the partially overlapping
memset is not dead.
Differential Revision: https://reviews.llvm.org/D89192
MemCpyOpt can shorten a memset if it is later partially overwritten
by a memcpy. It checks that the destination is not read in between,
but we also need to make sure that the destination cannot be observed
via unwinding.
Differential Revision: https://reviews.llvm.org/D89190
The previous code added the scope on each iteration, so that the
same scope was represented many times in the same !noalias metadata.
That's legal, and semantically equivalent to only storing the scope
once, but it's also wasteful and may pessimize further optimization
if AATags get intersected naively, as done by the AliasSetTracker.
Based on the recent patches D88475 and D88429 where we are losing undef values due to extension/comparisons.
I've added a Constant::mergeUndefsWith method that merges the undef scalar/elements from another Constant into a specific Constant.
Differential Revision: https://reviews.llvm.org/D88687
This relands commit 1c021c64ca which was
reverted in commit 17cec6a11a because
an assertion was being triggered, since `BuildConstantFromSCEV()`
wasn't updated to handle the case where the constant we want to truncate
is actually a pointer. I was unsuccessful in coming up with a test case
where we'd end there with constant zext/sext of a pointer,
so i didn't handle those cases there until there is a test case.
Original commit message:
While we indeed can't treat them as no-ops, i believe we can/should
do better than just modelling them as `unknown`. `inttoptr` story
is complicated, but for `ptrtoint`, it seems straight-forward
to model it just as a zext-or-trunc of unknown.
This may be important now that we track towards
making inttoptr/ptrtoint casts not no-op,
and towards preventing folding them into loads/etc
(see D88979/D88789/D88788)
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D88806
This patch turns VPMemoryInstructionRecipe into a VPValue and uses it
during VPlan construction and codegeneration instead of the plain IR
reference where possible.
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D84680
> While we indeed can't treat them as no-ops, i believe we can/should
> do better than just modelling them as `unknown`. `inttoptr` story
> is complicated, but for `ptrtoint`, it seems straight-forward
> to model it just as a zext-or-trunc of unknown.
>
> This may be important now that we track towards
> making inttoptr/ptrtoint casts not no-op,
> and towards preventing folding them into loads/etc
> (see D88979/D88789/D88788)
>
> Reviewed By: mkazantsev
>
> Differential Revision: https://reviews.llvm.org/D88806
It caused the following assert during Chromium builds:
llvm/lib/IR/Constants.cpp:1868:
static llvm::Constant *llvm::ConstantExpr::getTrunc(llvm::Constant *, llvm::Type *, bool):
Assertion `C->getType()->isIntOrIntVectorTy() && "Trunc operand must be integer"' failed.
See code review for a link to a reproducer.
This reverts commit 1c021c64ca.
60b852092c introduced SCEV verification to
deleteDeadLoop, but it appears this check is currently a bit over-eager
and some users of deleteDeadLoop appear to only patch up SE after
calling it (e.g. PR47753).
Remove the extra check for now. We can consider adding it back after we
tracked down the source of the inconsistency for PR47753.
If value tracking can confirm that a shift value is less than the type bitwidth then we can more confidently fold general or(shl(a,x),lshr(b,sub(bw,x))) patterns to a funnel/rotate intrinsic pattern without causing bad codegen regressions in the backend (see D89139).
Reapplied after the shift canonicalization in rG02295e6d1a15 which removed the need to flip the shift values.
Differential Revision: https://reviews.llvm.org/D88783
After rG02295e6d1a15 we no longer need to invert the shift values for fshr - this is just hidden at the moment as funnel shifts only ever match for constant values so never use the fshr "Sub on SHL" path.
While we indeed can't treat them as no-ops, i believe we can/should
do better than just modelling them as `unknown`. `inttoptr` story
is complicated, but for `ptrtoint`, it seems straight-forward
to model it just as a zext-or-trunc of unknown.
This may be important now that we track towards
making inttoptr/ptrtoint casts not no-op,
and towards preventing folding them into loads/etc
(see D88979/D88789/D88788)
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D88806
I have introduced a new template PolySize class, where the template
parameter determines the type of quantity, i.e. for an element
count this is just an unsigned value. The ElementCount class is
now just a simple derivation of PolySize<unsigned>, whereas TypeSize
is more complicated because it still needs to contain the uint64_t
cast operator, since there are still many places in the code that
rely upon this implicit cast. As such the class also still needs
some of it's own operators.
I've tried to minimise the amount of code in the base PolySize
class, which led to a couple of changes:
1. In some places we were relying on '==' operator comparisons
between ElementCounts and the scalar value 1. I didn't put this
operator in the new PolySize class, and thought it was actually
clearer to use the isScalar() function instead.
2. I removed the isByteSized function and replaced it with calls
to isKnownMultipleOf(8).
I've also renamed NextPowerOf2 to be coefficientNextPowerOf2 so
that it's more consistent with coefficientDivideBy.
Differential Revision: https://reviews.llvm.org/D88409
And another step towards transforms not introducing inttoptr and/or
ptrtoint casts that weren't there already.
As we've been establishing (see D88788/D88789), if there is a int<->ptr cast,
it basically must stay as-is, we can't do much with it.
I've looked, and the most source of new such casts being introduces,
as far as i can tell, is this transform, which, ironically,
tries to reduce count of casts..
On vanilla llvm test-suite + RawSpeed, @ `-O3`, this results in
-33.58% less `IntToPtr`s (19014 -> 12629)
and +76.20% more `PtrToInt`s (18589 -> 32753),
which is an increase of +20.69% in total.
However just on RawSpeed, where i know there are basically
none `IntToPtr` in the original source code,
this results in -99.27% less `IntToPtr`s (2724 -> 20)
and +82.92% more `PtrToInt`s (4513 -> 8255).
which is again an increase of 14.34% in total.
To me this does seem like the step in the right direction,
we end up with strictly less `IntToPtr`, but strictly more `PtrToInt`,
which seems like a reasonable trade-off.
See https://reviews.llvm.org/D88860 / https://reviews.llvm.org/D88995
for some more discussion on the subject.
(Eventually, `CastInst::isNoopCast()`/`CastInst::isEliminableCastPair`
should be taught about this, yes)
Reviewed By: nlopes, nikic
Differential Revision: https://reviews.llvm.org/D88979
This expands upon the inloop reductions added in e9761688e41cb9e976,
allowing them to be inserted into tail folded loops. Reductions are
generates with the form:
x = select(mask, vecop, zero)
v = vecreduce.add(x)
c = add chain, v
Where zero here is chosen as the identity value for add reductions. The
backend is then expected to fold the select and the vecreduce into a
single predicated instruction.
Most of the code is fairly straight forward, except for the creation of
blockmasks which need to ensure they are created in dominance order. The
order they are added is altered to be after any phis, keeping the
requirements for the underlying IR.
Differential Revision: https://reviews.llvm.org/D84451
As shown in the affected test, we could increase instruction
count without this limitation. There's another test with extra
use that shows we still convert directly to a real "sext" if
possible.
If value tracking can confirm that a shift value is less than the type bitwidth then we can more confidently fold general or(shl(a,x),lshr(b,sub(bw,x))) patterns to a funnel/rotate intrinsic pattern without causing bad codegen regressions in the backend (see D89139).
Differential Revision: https://reviews.llvm.org/D88783
This exposes the helper for other power-of-2 instcombine folds that I'm intending to add vector support to.
The helper only operated on power-of-2 constants so getExactLogBase2 is a more accurate name.
This patch is a refactoring of how we process spills and allocas during CoroSplit.
In the previous implementation, everything that needs to go to the heap is put into Spills, including all the values defined by allocas.
And the way to identify a Spill, is to check whether there exists a use-def relationship that crosses suspension points.
This approach is fundamentally confusing, and unfortunately, incorrect.
First of all, allocas are always process differently than spills, hence it's quite confusing to put them together. It's a much cleaner to separate them and process them separately.
Doing so simplify lots of code and makes the logic more clear and easier to reason about.
Secondly, use-def relationship is insufficient to decide whether a value defined by AllocaInst needs to go to the heap.
There are many cases where a value defined by AllocaInst can implicitly be used across suspension points without a direct use-def relationship.
For example, you can store the address of an alloca into the heap, and load that address after suspension. Or you can escape the address into an object through a function call.
Or you can have a PHINode that takes two allocas, and this PHINode is used across suspension point (when this happens, the existing implementation will spill the PHINode, a.k.a a stack adddress to the heap!).
All these issues suggest that we need to separate spill and alloca in order to properly implement this.
This patch does not yet fix these bugs, however it sets up the code in a better shape so that we can start fixing them in the next patch.
The core idea of this patch is to add a new struct called FrameDataInfo, which contains all Spills, all Allocas, and a map from each definition to its layout index in the frame (FieldIndexMap).
Spills and Allocas are identified, stored and processed independently. When they are initially added to the frame, we record their field index through FieldIndexMap. When the frame layout is finalized, we update each index into their final layout index.
In doing so, I also cleaned up a few things and also discovered a few other bugs.
Cleanups:
1. Found out that PromiseFieldId is not used, delete it.
2. Previously, SpillInfo is a vector, which is strange because every def can have multiple users. This patch cleans it up by turning it into a map from def to users.
3. Previously, a frame Field struct contains a list of Spills that field corresponds to. This isn't necessary since we only need the layout index for each given definition. This patch removes that list. Instead, we connect each field and definition using the FieldIndexMap.
4. All the loops that process Spills are simplified now because we use a map instead of a vector.
Bugs:
It seems that we are only keeping llvm.dbg.declare intrinsics in the .resume part of the function. The ramp function will no longer has it. This means we are dropping some debug information in the ramp function.
The next step is to start fixing the bugs where the implementation fails to identify some allocas that should live on the frame.
Differential Revision: https://reviews.llvm.org/D88872
MemCpyOpt can hoist stores while load+store pairs into memcpy.
This hoisting can currently result in stores being executed that
weren't guaranteed to execute in the original problem.
Differential Revision: https://reviews.llvm.org/D89154
Summary:
The current instruction sink pass uses findNearestCommonDominator of all users to find block to sink the instruction to.
However, a user may be in a dead block, which will result in unexpected behavior.
This patch handles such cases by skipping dead blocks. This patch fixes:
https://bugs.llvm.org/show_bug.cgi?id=47415
Reviewers:
MaskRay, arsenm
Differential Revision:
https://reviews.llvm.org/D89166
If a module has many values that need to be resolved by
ResolvedUndefsIn, compilation takes quadratic time overall. Solve should
do a small amount of work, since not much is added to the worklists each
time markOverdefined is called. But ResolvedUndefsIn is linear over the
length of the function/module, so resolving one undef at a time is
quadratic in general.
To solve this, make ResolvedUndefsIn resolve every undef value at once,
instead of resolving them one at a time. This loses a little
optimization power, but can be a lot faster.
We still need a loop around ResolvedUndefsIn because markOverdefined
could change the set of blocks that are live. That should be uncommon,
hopefully. We could optimize it by tracking which blocks transition from
dead to live, instead of iterating over the whole module to find them.
But I'll leave that for later. (The whole function will become a lot
simpler once we start pruning branches on undef.)
The regression test changes seem minor. The specific cases in question
could probably be optimized with a bit more work, but they seem like
edge cases that don't really matter.
Fixes an "infinite" compile issue my team found on an internal workoad.
Differential Revision: https://reviews.llvm.org/D89080
There are cases that generated OpenMP code consists of multiple,
consecutive OpenMP parallel regions, either due to high-level
programming models, such as RAJA, Kokkos, lowering to OpenMP code, or
simply because the programmer parallelized code this way. This
optimization merges consecutive parallel OpenMP regions to: (1) reduce
the runtime overhead of re-activating a team of threads; (2) enlarge the
scope for other OpenMP optimizations, e.g., runtime call deduplication
and synchronization elimination.
This implementation defensively merges parallel regions, only when they
are within the same BB and any in-between instructions are safe to
execute in parallel.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D83635
In the NPM, a pass cannot depend on another non-analysis pass. So pin
the test that tests that -lowerswitch is run automatically to legacy PM.
Reviewed By: sameerds
Differential Revision: https://reviews.llvm.org/D89051
There might be a better way to specify the pre-conditions,
but this is hopefully clearer than the way it was written:
https://rise4fun.com/Alive/Jhk3
Pre: C2 < 0 && isShiftedMask(C2) && (C1 == C1 & C2)
%a = and %x, C2
%r = add %a, C1
=>
%a2 = add %x, C1
%r = and %a2, C2
IV widening is sometimes a strictly harmful transform (some examples
of this are shown in tests 11, 12 in widen-loop-comp.ll). One of the
reasons of this is that sometimes SCEV fails to prove some facts after
part of guards has been widened.
Though each single such case looks like a bug that can be addressed,
it seems that disabling of IV widening may be profitable in some cases.
We want to have an option to do so. By default, existing behavior is
preserved and IV widening is on.
Annoyingly vectors aren't supported by shouldChangeType(), but we have precedents for always performing this on vector types (e.g. narrowBinOp).
Differential Revision: https://reviews.llvm.org/D89067
Use cast<> as we immediately dereference the pointer afterwards - cast<> will assert if we fail.
Prevents clang static analyzer warning that we could deference a null pointer.
Pre-conditions seem to be optimal, but we don't need a use check
because we are only replacing an add with a sub.
https://rise4fun.com/Alive/hzN
Pre: (~C1 | C2 == -1) && isPowerOf2(C2+1)
%m = and i8 %x, C1
%f = xor i8 %m, C2
%r = add i8 %f, C3
=>
%r = sub i8 C2 + C3, %m
Use cast<> as we immediately dereference the pointer afterwards - cast<> will assert if we fail.
Prevents clang static analyzer warning that we could deference a null pointer.
Complete basic PR46895 fixes by refactoring D87452/D88402 to allow us to match non-uniform constant values.
We still don't handle non-uniform vectors that contain undef elements, but that can wait until we have a decent generic mechanism for this.
Differential Revision: https://reviews.llvm.org/D88420
Use SCEV to salvage additional @llvm.dbg.value that have turned into
referencing undef after transformation (and traditional
salvageDebugInfo). Before transformation compute SCEV for each
@llvm.dbg.value in the loop body and store it (along side its current
DIExpression). After transformation update those @llvm.dbg.value now
referencing undef by comparing its stored SCEV to the SCEV of the
current loop-header PHI-nodes. Allow match with offset by inserting
compensation code in the DIExpression.
Includes fix for the nullptr deref that caused the original commit
to be reverted in 9d63029770.
Fixes : PR38815
Differential Revision: https://reviews.llvm.org/D87494
First step towards extending the existing rotation support to full funnel shift handling now that the backend legalization support has improved.
This enables us to match the shift by constant cases, which are pretty trivial to expand again if necessary.
D88420 will add non-uniform support for funnel shifts as well once its been finalized.
Differential Revision: https://reviews.llvm.org/D88834
The existing code ignores undef values which matches m_SpecificInt_ICMP, although m_SpecificInt_ICMP returns false for an all-undef constant, I've added test coverage at rGfe0197e194a64f9 to show that undef folding should already have dealt with that case.
We currently collect the ICmp and Add from an induction variable,
marking them as dead so that vplan values are not created for them. This
extends that to include any single use trunk from the ICmp, which allows
the Add to more readily be removed too.
This can help with costing vplan nodes, as the ICmp and Add are more
reliably removed and are not double-counted.
Differential Revision: https://reviews.llvm.org/D88873
In some cases, we can negate instruction if only one of it's operands
negates. Previously, we assumed that constants would have been
canonicalized to RHS already, but that isn't guaranteed to happen,
because of InstCombine worklist visitation order,
as the added test (previously-hanging) shows.
So if we only need to negate a single operand,
we should ensure ourselves that we try constant operand first.
Do that by re-doing the complexity sorting ourselves,
when we actually care about it.
Fixes https://bugs.llvm.org/show_bug.cgi?id=47752
And another step towards transformss not introducing inttoptr and/or
ptrtoint casts that weren't there already.
In this case, when load/store uses have conflicting types,
instead of falling back to the iN, we can try to use allocated sub-type.
As disscussed, this isn't the best idea overall (we shouldn't rely on
allocated type), but it works fine as a temporary measure.
I've measured, and @ `-O3` as of vanilla llvm test-suite + RawSpeed,
this results in +0.05% more bitcasts, -5.51% less inttoptr
and -1.05% less ptrtoint (at the end of middle-end opt pipeline)
See https://bugs.llvm.org/show_bug.cgi?id=47592
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D88788
The old function attribute deduction pass ignores reads of constant
memory and we need to copy this behavior to replace the pass completely.
First step are constant globals. TBAA can also describe constant
accesses and there are other possibilities. We might want to consider
asking the alias analyses that are available but for now this is simpler
and cheaper.
If the function is not assumed `noreturn` we should not wait for an
update to mark the call site as "may-return".
This has two kinds of consequences:
- We have less iterations in many tests.
- We have less deductions based on "known information" (since we ask
earlier, point 1, and therefore assumed information is not "known"
yet).
The latter is an artifact that we might want to tackle properly at some
point but which is not easily fixable right now.
The call slot optimization has some home-grown code for checking
whether the destination is dereferenceable. Replace this with the
generic isDereferenceableAndAlignedPointer() helper.
I'm not checking alignment here, because that is currently handled
separately and may be an enforced alignment for allocas. The clean
way of integrating that part would probably be to accept a callback
in isDereferenceableAndAlignedPointer() for the actual isAligned check,
which would then have a chance to use an enforced alignment instead.
This allows the destination to be a GEP (among other things), though
the two open TODOs may prevent it from working in practice.
Differential Revision: https://reviews.llvm.org/D88805
When performing call slot optimization for a non-local destination,
we need to check whether there may be throwing calls between the
call and the copy. Otherwise, the early write to the destination
may be observable by the caller.
This was already done for call slot optimization of load/store,
but not for memcpys. For the sake of clarity, I'm moving this check
into the common optimization function, even if that does need an
additional instruction scan for the load/store case.
As efriedma pointed out, this check is not sufficient due to
potential accesses from another thread. This case is left as a TODO.
Differential Revision: https://reviews.llvm.org/D88799
When we assume a return value is dead we might still visit return
instructions via `Attributor::checkForAllReturnedValuesAndReturnInsts(..)`.
When we do so the "returned value" is potentially simplified to `undef`
as it is the assumed "returned value". This is a problem if there was a
preexisting `noundef` attribute that will only be removed as we manifest
the `undef` return value. We should not use this combination to derive
`unreachable` though. Two test cases fixed.
In AAMemoryBehaviorFloating we used to track benign uses in a SetVector.
With this change we look through benign uses eagerly to reduce the
number of elements (=Uses) we look at during an update.
The test does actually not fail prior to this commit but I already wrote
it so I kept it.
We know V is a IntToPtrInst or PtrToIntInst type so we know its a CastInst - so use cast<> directly.
Prevents clang static analyzer warning that we could deference a null pointer.
This still only gets used for scalar types but now always uses ConstantExpr in preparation for vector support - it was using APInt methods in some places.
Use context to prove that load can be safely executed at a point where load is being hoisted.
Postpone the decision about safety of speculative load execution till the moment we know
where we hoist load and check safety at that context.
Reviewers: nikic, fhahn, mkazantsev, lebedev.ri, efriedma, reames
Reviewed By: reames, mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D88725
This reverts commit 20797989ea.
This patch (https://reviews.llvm.org/D69257) cannot complete a stage2
build due to the change:
```
CI->getCalledFunction()->getName().contains("longjmp")
```
There are several concrete issues here:
- The callee may not be a function, so `getCalledFunction` can assert.
- The called value may not have a name, so `getName` can assert.
- There's no distinction made between "my_longjmp_test_helper" and the
actual longjmp libcall.
At a higher level, there's a serious layering problem here. The
splitting pass makes policy decisions in a general way (e.g. based on
attributes or profile data). Special-casing certain names breaks the
layering. It subverts the work of library maintainers (who may now need
to opt-out of unexpected optimization behavior for any affected
functions) and can lead to inconsistent optimization behavior (as not
all llvm passes special-case ".*longjmp.*" in the same way).
The patch may need significant revision to address these issues.
But the immediate issue is that this crashes while compiling llvm's unit
tests in a stage2 build (due to the `getName` problem).
(it was introduced in https://lists.llvm.org/pipermail/llvm-dev/2015-January/080956.html)
This canonicalization seems dubious.
Most importantly, while it does not create `inttoptr` casts by itself,
it may cause them to appear later, see e.g. D88788.
I think it's pretty obvious that it is an undesirable outcome,
by now we've established that seemingly no-op `inttoptr`/`ptrtoint` casts
are not no-op, and are no longer eager to look past them.
Which e.g. means that given
```
%a = load i32
%b = inttoptr %a
%c = inttoptr %a
```
we likely won't be able to tell that `%b` and `%c` is the same thing.
As we can see in D88789 / D88788 / D88806 / D75505,
we can't really teach SCEV about this (not without the https://bugs.llvm.org/show_bug.cgi?id=47592 at least)
And we can't recover the situation post-inlining in instcombine.
So it really does look like this fold is actively breaking
otherwise-good IR, in a way that is not recoverable.
And that means, this fold isn't helpful in exposing the passes
that are otherwise unaware of these patterns it produces.
Thusly, i propose to simply not perform such a canonicalization.
The original motivational RFC does not state what larger problem
that canonicalization was trying to solve, so i'm not sure
how this plays out in the larger picture.
On vanilla llvm test-suite + RawSpeed, this results in
increase of asm instructions and final object size by ~+0.05%
decreases final count of bitcasts by -4.79% (-28990),
ptrtoint casts by -15.41% (-3423),
and of inttoptr casts by -25.59% (-6919, *sic*).
Overall, there's -0.04% less IR blocks, -0.39% instructions.
See https://bugs.llvm.org/show_bug.cgi?id=47592
Differential Revision: https://reviews.llvm.org/D88789
... by using MapVector. The issue was caused by 63182c2ac0.
Also use stable_partition instead of partition to get stable results
across different STL implementations.
When retrying the "simplify with operand replaced" select
optimization without poison flags, also handle inbounds on GEPs.
Of course, this particular example would also be safe to transform
while keeping inbounds, but the underlying machinery does not
know this (yet).
Use m_Specific instead of m_Value followed by an equality check - we already do this for the similar folds above, it looks like an oversight in rG2b459fe7e1e where the original pattern match code looked a little different.
This reverts commit a3caf7f610.
The ReleaseLTO-g test-suite configuration has been failing
to build since this commit, because clang segfaults while
building 7zip.
Update the code responsible for deleting VPBBs and recipes to properly
update users and release operands.
This is another preparation for D84680 & following patches towards
enabling modeling def-use chains in VPlan.
Use SCEV to salvage additional @llvm.dbg.value that have turned into
referencing undef after transformation (and traditional
salvageDebugInfo). Before transformation compute SCEV for each
@llvm.dbg.value in the loop body and store it (along side its current
DIExpression). After transformation update those @llvm.dbg.value now
referencing undef by comparing its stored SCEV to the SCEV of the
current loop-header PHI-nodes. Allow match with offset by inserting
compensation code in the DIExpression.
Fixes : PR38815
Differential Revision: https://reviews.llvm.org/D87494
This adds a helper to convert a VPRecipeBase pointer to a VPUser, for
recipes that inherit from VPUser. Once VPRecipeBase directly inherits
from VPUser this helper can be removed.
When updating operands of a VPUser, we also have to adjust the list of
users for the new and old VPValues. This is required once we start
transitioning recipes to become VPValues.
Do not instrument user-defined ELF sections (whose names resemble valid
C identifiers). They may have special use semantics and modifying them
may break programs. This is e.g. the case with NetBSD __link_set API
that expects these sections to store consecutive array elements.
Differential Revision: https://reviews.llvm.org/D76665
Add basic vector handling to recognizeBSwapOrBitReverseIdiom/collectBitParts - this works at the element level, all vector element operations must match (splat constants etc.) and there is no cross-element support (insert/extract/shuffle etc.).
If we're bswap'ing some bytes and zero'ing the remainder we can perform this as a bswap+mask which helps us match 'partial' bswaps as a first step towards folding into a more complex bswap pattern.
Reapplied with early-out if recognizeBSwapOrBitReverseIdiom collects a source wider than the result type.
Differential Revision: https://reviews.llvm.org/D88578
Next to erasing the instruction, we also always want to remove
it from MSSA and MD. Use a common function to do so.
This is a refactoring split out from D26739.
The removal of the cpy instruction is left to the caller of
performCallSlotOptzn(), including the invalidation of MD. Both
call-sites already do this.
Also handle incrementation of NumMemCpyInstr consistently at the
call-site. One of the call-site was already doing this, which
ended up incrementing the statistic twice.
This fix was part of D26739.
While looping through all args or all return values, we may mark a use
of a later iteration as live. Previously when we got to that later value
it would ignore that and continue adding to Uses instead of marking it
live. For example, when looping through arg#0 and arg#1,
MarkValue(arg#0, Live) may cause some use of arg#1 to be live, but
MarkValue(arg#1, MaybeLive) will not notice that and continue adding
into Uses.
Now MarkValue(RA, MaybeLive) will MarkLive(RA) if any use is live.
Fixes PR47444.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D88529
If we're bswap'ing some bytes and zero'ing the remainder we can perform this as a bswap+mask which helps us match 'partial' bswaps as a first step towards folding into a more complex bswap pattern.
Differential Revision: https://reviews.llvm.org/D88578
If we try to coerce a vector of non-integral pointers to a narrower type (either narrower vector or single pointer), we use inttoptr and violate the semantics of non-integral pointers. In theory, we can handle many of these cases, we just need to use a different code idiom to convert without going through inttoptr and back.
This shows up as wrong code bugs, and in some cases, crashes due to failed asserts. Modeled after a change which has lived downstream for a couple years, though completely rewritten to be more idiomatic.
If we allow the non-integral pointers to become memset and memcpy, we loose the ability to reason about pointer propagation. This patch is modeled on changes we've carried downstream for a long time, figured it was worth being equally conservative for other users. There is room to refine the semantics and handling here if anyone is motivated.
When replacing X == Y ? f(X) : Z with X == Y ? f(Y) : Z, make sure
that Y cannot be undef. If it may be undef, we might end up picking
a different value for undef in the comparison and the select
operand.
Summary:
Adds support for "following" memory through MSSA PHI arguments. This will help catch more noop stores that exist between blocks.
Originally part of D79391.
Reviewers: fhahn, jfb, asbirlea
Differential Revision: https://reviews.llvm.org/D82588
Make sure we're using getScalarSizeInBits instead of cast<IntegerType> to get Type bit widths.
This is preliminary cleanup before we can start adding vector support to the bswap/bitreverse (element level) matching.
This is a simple pass that flattens nested loops. The intention is to optimise
loop nests like this, which together access an array linearly:
for (int i = 0; i < N; ++i)
for (int j = 0; j < M; ++j)
f(A[i*M+j]);
into one loop:
for (int i = 0; i < (N*M); ++i)
f(A[i]);
It can also flatten loops where the induction variables are not used in the
loop. This can help with codesize and runtime, especially on simple cpus
without advanced branch prediction.
This is only worth flattening if the induction variables are only used in an
expression like i*M+j. If they had any other uses, we would have to insert a
div/mod to reconstruct the original values, so this wouldn't be profitable.
This partially fixes PR40581 as this pass triggers on one of the two cases. I
will follow up on this to learn LoopFlatten a few more (small) tricks. Please
note that LoopFlatten is not yet enabled by default.
Patch by Oliver Stannard, with minor tweaks from Dave Green and myself.
Differential Revision: https://reviews.llvm.org/D42365
The legacy pass's default constructor sets UseCommandLine = true and
goes down a separate testing route. Match that in the NPM pass.
This fixes all tests in llvm/test/Transforms/WholeProgramDevirt under NPM.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D88588
There doesn't seem to be any good reason for having a separate path for when we bswap/bitreverse at a smaller size than the destination size - so merge these to make the instruction generation a lot clearer.
Now that VPUser is not inheriting from VPValue, we can take the next
step and turn the recipes that already manage their operands via VPUser
into VPUsers directly. This is another small step towards traversing
def-use chains in VPlan.
This is NFC with respect to the generated code, but makes the interface
more powerful.
Fix a number of WShadow warnings (I was used as the instruction and index......) and fix cases to match style.
Also, replaced the Bit APInt mask check in AND instructions with a direct APInt[] bit check.
PR39793 demonstrated an issue where we fail to recognize 'partial' bswap patterns of the lower bytes of an integer source.
In fact, most of this is already in place collectBitParts suitably tags zero bits, so we just need to correctly handle this case by finding the zero'd upper bits and reducing the bswap pattern just to the active demanded bits.
Differential Revision: https://reviews.llvm.org/D88316
I think we initially made this fold conservative to be safer, but we do not
need the alignment attribute/metadata limitation because the masked load
intrinsic itself specifies the alignment. A normal vector load is better for
IR transforms and should be no worse in codegen than the masked alternative.
If it is worse for some target, the backend can reverse this transform.
Differential Revision: https://reviews.llvm.org/D88505
Attempt to fold trunc (*shr (trunc A), C) --> trunc(*shr A, C) iff the shift amount if small enough that all zero/sign bits created by the shift are removed by the last trunc.
Helps fix the regressions encountered in D88316.
I've tweaked a couple of shift values as suggested by @lebedev.ri to ensure we have coverage of shift values close (above/below) to the max limit.
Differential Revision: https://reviews.llvm.org/D88429
This patch adds noundef to the returned pointers of allocators (malloc, calloc, ...)
and the pointer argument of free.
The returned pointer of allocators cannot be poison or (partially) undef.
Since the pointer that is given to free should precisely have zero offset,
it cannot be poison or (partially) undef too.
For the size arguments of allocators, noundef wasn't attached simply because
I wasn't sure whether attaching it is okay or not.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D87984
This came from @lebedev.ri's suggestion to use m_SpecificInt_ICMP for D88429 - since I was going to change the m_APInt to m_Constant for that patch I thought I would do it for the only other user of the APInt first.
I've added a ConstantExpr::getUMin helper - its trivial to add UMAX/SMIN/SMAX but thought I'd wait until we have use cases.
Differential Revision: https://reviews.llvm.org/D88475
With branch protection the jump to the jump table entries requires a landing pad.
Reviewed By: eugenis, tamas.petz
Differential Revision: https://reviews.llvm.org/D81251
When removing an overflow intrinsic the Changed status in SimplifyIndvar
was not set, leading to the IndVarSimplify pass returning an incorrect
status.
This was caught using the check introduced by D80916.
As pointed out in the code review, a similar bug may exist for
eliminateTrunc().
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D85971
After D71539, we need to forget the loop before setting the incoming
values of phi nodes in exit blocks, because we are looking through those
phi nodes now and the SCEV expression could depend on the loop phi. If
we update the phi nodes before forgetting the loop, we miss those users
during invalidation.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D88167
From preconditions it is known that either A dominates B or
B dominates A. If A does not dominate B, we do not really need
to check it. Assert should be enough. Should save some compile
time.
When removing exiting loop conditions, we only consider checks for
which we know the exact exit count. We could also eliminate checks for
which the condition is always true/false.
Differential Revision: https://reviews.llvm.org/D87344
Reviewed By: lebedev.ri, reames
For a call site which had both constant deopt operands and nonnull arguments, we were missing the opportunity to recognize the later by bailing early.
This is somewhat of a speculative fix. Months ago, I'd had a private report of performance and compile time regressions from the deopt operand folding. I never received a test case. However, the only possibility I see was that after that change CVP missed the nonnull fold, and we end up with a pass ordering/missed simplification issue. So, since it's a real issue, fix it and hope.
When ASan and e.g. Dead Virtual Function Elimination are enabled, the
latter will rely on type metadata to determine if certain virtual calls can be
removed. However, ASan currently does not copy type metadata, which can cause
virtual function calls to be incorrectly removed.
Differential Revision: https://reviews.llvm.org/D88368
Fixes minor bug in D88402 where we were using the original shift constant (with undefs) instead of one with the splat values (re)splatted to all elements.
bug 45566 shows the process of building coroutine frame won't consider
that the lifetimes of different local variables are not overlapped,
which means the compiler could generates smaller frame.
This patch calculate the lifetime range of each alloca by StackLifetime
class. Then the patch build non-overlapped sets for allocas whose
lifetime ranges are not overlapped. We use the largest type in a
non-overlapped set as the field type in the frame. In insertSpills
process, if we find the type of field is not the same with the alloca,
we cast the pointer to the field type to the pointer to the alloca type.
Since the lifetime range of alloca in one non-overlapped set is not
overlapped with each other, it should be ok to reuse the storage space
in the frame.
Test plan: check-llvm, check-clang, cppcoro, folly
Reviewers: junparser, lxfind, modocache
Differential Revision: https://reviews.llvm.org/D87596
Add a flag to getPredicateAt() that allows making use of the block
value. This allows us to take into account range information from
the current block, rather than only information that is threaded
over edges, making the icmp simplification in CVP a lot more
powerful.
I'm not changing getPredicateAt() to use the block value
unconditionally to avoid any impact on the JumpThreading pass,
which is somewhat picky about LVI query order.
Most test changes here are just icmps that now get dropped (while
previously only a result used in a return was replaced). The three
tests in icmp.ll show some representative improvements. Some of
the folds this enables have been covered by IPSCCP in the meantime,
but LVI can reason about some cases which are hard to support in
IPSCCP, such as in test_br_cmp_with_offset.
The compile-time time cost of doing this is fairly minimal, with
a ~0.05% CTMark regression for ReleaseThinLTO:
https://llvm-compile-time-tracker.com/compare.php?from=709d03f8af4da4204849a70f01798e7cebba2e32&to=6236fd503761f43c99f4537121e057a01056f185&stat=instructions
This is because the block values will typically already be queried
and cached by other CVP optimizations anyway.
Differential Revision: https://reviews.llvm.org/D69686
If -enable-constraint-elimination is specified, add it to the -O2/-O3 pipeline.
(-O1 uses a separate function now.)
Reviewed By: fhahn, aeubanks
Differential Revision: https://reviews.llvm.org/D88365
Require CxtI in getConstant() and getConstantRange() APIs.
Accordingly drop the BB parameter, as it is implied by
CxtI->getParent().
This makes sure we don't forget to pass the context instruction,
and makes the API contract clearer (also clean up the comments to
that effect -- the value holds at the context instruction, not
the end of the block).
This fold was the only place not passing the context instruction.
The tests worked around that fact by introducing a basic block split,
which is now no longer necessary.
There appears to be a mis-compile with MemorySSA-backed DSE in
combination with llvm.lifetime.end. It currently appears like
DSE is doing the right thing and the llvm.lifetime.end markers
are incorrect. The reverted patch uncovers the mis-compile.
This patch temporarily switches back to the legacy DSE
implementation, while we investigate.
This reverts commit 9d172c8e9c.
When looking for memory defs killed by memory terminators the code
currently incorrectly ignores the size argument of llvm.lifetime.end.
This patch updates the code to use isMemTerminator and updates
isMemTerminator to use isOverwrite() to make sure locations that are
outside the range marked as dead by llvm.lifetime.end are not
considered. Note that isOverwrite is only used for llvm.lifetime.end,
because free-like functions make the whole underlying object dead.
This matches the legacy PM pass by having one constructor use command
line flags, and the other use parameters to the pass.
This fixes all tests under Transforms/LowerTypeTests using NPM.
Reviewed By: ychen, pcc
Differential Revision: https://reviews.llvm.org/D87845
This patch adds handling of rotation patterns with constant shift amounts - the next bit will be how we want to support non-uniform constant vectors.
Differential Revision: https://reviews.llvm.org/D87452
Pulled from D87452, this is a fixed version of the collectBitParts fshl/fshr handling which as @nikic noticed wasn't checking for different providers or had correct bit ordering (which was hid by only testing shift amounts of bitwidth/2).
Differential Revision: https://reviews.llvm.org/D88292
Issue Details:
In order to support coroutine splitting, any multi-value PHI node in a coroutine is split into multiple blocks with single-value PHI Nodes, which then allows a subsequent transform to generate `reload` instructions as required (i.e., to reload the value if required if the coroutine has been resumed). This causes issues with EH pads (`catchswitch` and `catchpad`) as all pads within a `catchswitch` must have the same unwind destination, but the coroutine splitting logic may modify them to each have a unique unwind destination if there is a PHI node in the unwind `cleanuppad` that is set from values in the `catchswitch` and `cleanuppad` blocks.
Fix Details:
During splitting, if such a PHI node is detected, then create a "dispatcher" `cleanuppad` as well as the blocks with single-value PHI Nodes: thus the "dispatcher" is the unwind destination and it will detect which predecessor called it and then branch to the appropriate single-value PHI node block, which will then branch back to the original `cleanuppad` block.
Reviewed By: GorNishanov, lxfind
Differential Revision: https://reviews.llvm.org/D88059
Summary:
This patch add support for printing analysis messages relating to data
globalization on the GPU. This occurs when data is shared between the
threads in a GPU context and must be pushed to global or shared memory.
Reviewers: jdoerfert
Subscribers: guansong hiraditya llvm-commits ormris sstefan1 yaxunl
Tags: #OpenMP #LLVM
Differential Revision: https://reviews.llvm.org/D88243
Introduce a helper which can be used to update the debug location of an
Instruction after the instruction is hoisted. This can be used to safely
drop a source location as recommended by the docs.
For more context, see the discussion in https://reviews.llvm.org/D60913.
Differential Revision: https://reviews.llvm.org/D85670
These were only really used for 2 things. One was to check if the operand matches the phi if it exists. The other was for the createOp method to build the reduction.
For the first case we still have the operation we just need to know how to index its operands. So I've modified getLHS/getRHS to just use the opcode/kind to know how to find the right operands on an instruction that is now passed in.
For the other case we had to create an OperationData object to set the LHS/RHS values and copy the opcode/kind from another object. We would then just call createOp on that temporary object. Instead I've made LHS/RHS arguments to createOp and removed all these temporary objects.
Differential Revision: https://reviews.llvm.org/D88193
This seems to fit the CGSCC updates model better than calling
addNewFunctionInto{Ref,}SCC() on newly created/outlined functions.
Now addNewFunctionInto{Ref,}SCC() are no longer necessary.
However, this doesn't work on newly outlined functions that aren't
referenced by the original function. e.g. if a() was outlined into b()
and c(), but c() is only referenced by b() and not by a(), this will
trigger an assert.
This also fixes an issue I was seeing with newly created functions not
having passes run on them.
Ran check-llvm with expensive checks.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87798
All of the callers already have an Instruction *. Many of them
from a dyn_cast.
Also update the OperationData constructor to use a Instruction&
to remove a dyn_cast and make it clear that the pointer is non-null.
Differential Revision: https://reviews.llvm.org/D88132
D87691 reordered some checks, which turned out to be unsafe. More
specifically, when examining a store instruction, the check against
getOrCreateResult should be done before attempting to call
isSameMemGeneration. Otherwise a crash in MSSA walker can occur.
This patch restores the order of these calls to what it was originally.
This refactors VPuser to not inherit from VPValue to facilitate
introducing operations that introduce multiple VPValues (e.g.
VPInterleaveRecipe).
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D84679
In this patch I've fixed some warnings that arose from the implicit
cast of TypeSize -> uint64_t. I tried writing a variety of different
cases to show how this optimisation might work for scalable vectors
and found:
1. The optimisation does not work for cases where the cast type
is scalable and the allocated type is not. This because we need to
know how many times the cast type fits into the allocated type.
2. If we pass all the various checks for the case when the allocated
type is scalable and the cast type is not, then when creating the
new alloca we have to take vscale into account. This leads to
sub-optimal IR that is worse than the original IR.
3. For the remaining case when both the alloca and cast types are
scalable it is hard to find examples where the optimisation would
kick in, except for simple bitcasts, because we typically fail the
ABI alignment checks.
For now I've changed the code to bail out if only one of the alloca
and cast types is scalable. This means we continue to support the
existing cases where both types are fixed, and also the specific case
when both types are scalable with the same size and alignment, for
example a simple bitcast of an alloca to another type.
I've added tests that show we don't attempt to promote the alloca,
except for simple bitcasts:
Transforms/InstCombine/AArch64/sve-cast-of-alloc.ll
Differential revision: https://reviews.llvm.org/D87378
A conversion from `pow` to `sqrt` shall not call an `errno`-setting
`sqrt` with -//infinity//: the `sqrt` will set `EDOM` where the `pow`
call need not.
This patch avoids the erroneous (pun not intended) transformation by
applying the restrictions discussed in the thread for
https://lists.llvm.org/pipermail/llvm-dev/2020-September/145051.html.
The existing tests are updated (depending on emphasis in the checks for
library calls, avoidance of overlap, and overall coverage):
- to add `ninf`, retaining the intended library call,
- to use the intrinsic, retaining the use of `select`, or
- to expect the replacement to not occur.
The following is tested:
- The pow intrinsic folds to a `select` instruction to
handle -//infinity//.
- The pow library call folds, with `ninf`, to `sqrt` without the
`select` instruction associated with handling -//infinity//.
- The pow library call does not fold to `sqrt` without `ninf`.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D87877
As an exhaustive test shows, this logic is fully identical to the old
implementation, with exception of the case where both of the operands
had empty ranges:
```
TEST_F(ConstantRangeTest, CVP_UDiv) {
unsigned Bits = 4;
EnumerateConstantRanges(Bits, [&](const ConstantRange &CR0) {
if(CR0.isEmptySet())
return;
EnumerateConstantRanges(Bits, [&](const ConstantRange &CR1) {
if(CR0.isEmptySet())
return;
unsigned MaxActiveBits = 0;
for (const ConstantRange &CR : {CR0, CR1})
MaxActiveBits = std::max(MaxActiveBits, CR.getActiveBits());
ConstantRange OperandRange(Bits, /*isFullSet=*/false);
for (const ConstantRange &CR : {CR0, CR1})
OperandRange = OperandRange.unionWith(CR);
unsigned NewWidth = OperandRange.getUnsignedMax().getActiveBits();
EXPECT_EQ(MaxActiveBits, NewWidth) << CR0 << " " << CR1;
});
});
}
```
This is a continuation of 8d487668d0,
the logic is pretty much identical for SRem:
Name: pos pos
Pre: C0 >= 0 && C1 >= 0
%r = srem i8 C0, C1
=>
%r = urem i8 C0, C1
Name: pos neg
Pre: C0 >= 0 && C1 <= 0
%r = srem i8 C0, C1
=>
%r = urem i8 C0, -C1
Name: neg pos
Pre: C0 <= 0 && C1 >= 0
%r = srem i8 C0, C1
=>
%t0 = urem i8 -C0, C1
%r = sub i8 0, %t0
Name: neg neg
Pre: C0 <= 0 && C1 <= 0
%r = srem i8 C0, C1
=>
%t0 = urem i8 -C0, -C1
%r = sub i8 0, %t0
https://rise4fun.com/Alive/Vd6
Now, this new logic does not result in any new catches
as of vanilla llvm test-suite + RawSpeed.
but it should be virtually compile-time free,
and it may be important to be consistent in their handling,
because if we had a pair of sdiv-srem, and only converted one of them,
-divrempairs will no longer see them as a pair,
and thus not "merge" them.
The current code for handling pow(x, y) where y is an integer plus 0.5
is not explicitly guarded against attempting to transform the case where
abs(y) is exactly 0.5.
The latter case is meant to be handled by `replacePowWithSqrt`. Indeed,
if the pow(x, integer+0.5) case proceeds past a certain point, it will
hit an assertion by attempting to form pow(x, 0) using `getPow`.
This patch adds an explicit check to prevent attempting the
pow(x, integer+0.5) transformation on pow(x, +/-0.5) as suggested during
the review of D87877. This has the effect of retaining the shrinking of
`pow` to `powf` when the `sqrt` libcall cannot be formed.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D88066
Refactored __tgt_target_data_begin_mapper_<issue|wait> to receive the handle as an input/output argument.
This given the compiler warning of returning the handle as copy.
Differential Revision: https://reviews.llvm.org/D88029
This provides a convenient way to print VPValues and recipes in a
debugger. In particular it saves the user from instantiating
VPSlotTracker to print recipes or values.
Changes TTI function getIntImmCostInst to take an additional Instruction parameter,
which enables us to be able to check it is part of a min(max())/max(min()) pattern that will match SSAT.
We can then mark the constant used as free to prevent it being hoisted so SSAT can still be generated.
Required minor changes in some non-ARM backends to allow for the optional parameter to be included.
Differential Revision: https://reviews.llvm.org/D87457
Extend the handling of memory intrinsics to also include non-
target-specific intrinsics, in particular masked loads and stores.
Invent "isHandledNonTargetIntrinsic" to distinguish between intrin-
sics that should be handled natively from intrinsics that can be
passed to TTI.
Add code that handles masked loads and stores and update the
testcase to reflect the results.
Differential Revision: https://reviews.llvm.org/D87340
SimplifyCFG's options should always be overridden by command line flags,
but they mistakenly weren't in the default constructor.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D87718
1. Store intrinsic ID in ParseMemoryInst instead of a boolean flag
"IsTargetMemInst". This will make it easier to add support for
target-independent intrinsics.
2. Extract the complex multiline conditions from EarlyCSE::processNode
into a new function "getMatchingValue".
Differential Revision: https://reviews.llvm.org/D87691
This pass is like DeadCodeEliminationPass, but only does one pass
through a function instead of iterating on users of eliminated
instructions.
DeadCodeEliminationPass should be used in all cases.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87933
The implementation of gather() should be reduced too,
but this change by itself makes things a little clearer:
we don't try to gather to a different type or
number-of-values than whatever is passed in as the value
list itself.
If some leaves have the same instructions to be vectorized, we may
incorrectly evaluate the best order for the root node (it is built for the
vector of instructions without repeated instructions and, thus, has less
elements than the root node). In this case we just can not try to reorder
the tree + we may calculate the wrong number of nodes that requre the
same reordering.
For example, if the root node is \<a+b, a+c, a+d, f+e\>, then the leaves
are \<a, a, a, f\> and \<b, c, d, e\>. When we try to vectorize the first
leaf, it will be shrink to \<a, b\>. If instructions in this leaf should
be reordered, the best order will be \<1, 0\>. We need to extend this
order for the root node. For the root node this order should look like
\<3, 0, 1, 2\>. This patch allows extension of the orders of the nodes
with the reused instructions.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D45263
LSR claims to MemorySSA, but we also have to make sure it is preserved
when splitting critical edges. This can be done by passing MSSAU to
SplitCriticalEdge.
Fixes PR47557.
We do similar factorization folds in SimplifyUsingDistributiveLaws,
but that drops no-wrap properties. Propagating those optimally may
help solve:
https://llvm.org/PR47430
The propagation is all-or-nothing for these patterns: when all
3 incoming ops have nsw or nuw, the 2 new ops should have the
same no-wrap property:
https://alive2.llvm.org/ce/z/Dv8wsU
This also solves:
https://llvm.org/PR47584
The test (currently crashing) is reduced from the example provided
in the post-commit discussion in D87149.
Differential Revision: https://reviews.llvm.org/D87965
I want to export this function, and the current API was a bit
weird: It took an additional Alignment argument that didn't really
have anything to do with what the function does. Drop it, and
perform a max at the callsite.
Also rename it to tryEnforceAlignment().
Currently SCEVExpander creates inttoptr for non-integral pointers if the
base is a null constant for example. This results in invalid IR.
This patch changes InsertNoopCastOfTo to emit a GEP & bitcast to convert
to a non-integral pointer. First, a GEP of i8* null is generated and the
integral value is used as index. The GEP is then bitcasted to the target
type.
This was exposed by D71539.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D87827
When address sanitizing a function, stack unpinsoning code is inserted before each ret instruction. However if the ret instruciton is preceded by a musttail call, such transformation broke the musttail call contract and generates invalid IR.
This patch fixes the issue by moving the insertion point prior to the musttail call if there is one.
Differential Revision: https://reviews.llvm.org/D87777
We cannot iterate on scalable vector, the number of elements is unknown at compile-time.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D87918
If some leaves have the same instructions to be vectorized, we may
incorrectly evaluate the best order for the root node (it is built for the
vector of instructions without repeated instructions and, thus, has less
elements than the root node). In this case we just can not try to reorder
the tree + we may calculate the wrong number of nodes that requre the
same reordering.
For example, if the root node is \<a+b, a+c, a+d, f+e\>, then the leaves
are \<a, a, a, f\> and \<b, c, d, e\>. When we try to vectorize the first
leaf, it will be shrink to \<a, b\>. If instructions in this leaf should
be reordered, the best order will be \<1, 0\>. We need to extend this
order for the root node. For the root node this order should look like
\<3, 0, 1, 2\>. This patch allows extension of the orders of the nodes
with the reused instructions.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D45263
This switches to using DSE + MemorySSA by default again, after
fixing the issues reported after the first commit.
Notable fixes fc82006331, a0017c2bc2.
This reverts commit 3a59628f3c.
Enable canonicalization of SPF_ABS and SPF_NABS to the abs intrinsic.
To be conservative, the one-use check on the comparison is retained,
this may be relaxed if all goes well.
It's pretty likely that this will uncover places that missing
handling for the abs() intrinsic. Please report any seen performance
regressions.
Differential Revision: https://reviews.llvm.org/D87188
Summary: Allow unroll and jam loops forced by user.
LoopUnrollAndJamPass is still disabled by default in the NPM pipeline,
and can be controlled by -enable-npm-unroll-and-jam.
Reviewed By: Meinersbur, dmgreen
Differential Revision: https://reviews.llvm.org/D87786
As discussed in:
https://llvm.org/PR47558
...there are several potential fixes/follow-ups visible
in the test case, but this is the quickest and safest
fix of the perf regression.
When a spill definition is before CoroBegin, we cannot spill it to the frame immediately after the definition. We have to spill it after the frame is ready.
The current implementation handles it properly for any other kinds of instructions except for PhINode and InvokeInst, which could also be defined before CoroBegin.
This patch fixes it by moving the CoroBegin dominance check earlier, so that it covers all cases.
Added a test.
Differential Revision: https://reviews.llvm.org/D87810
This is one (small) part of improving PR41312:
https://llvm.org/PR41312
As shown there and in the smaller tests here, if we have some member of the
reduction values that does not match the others, we want to push it to the
end (bring the matching members forward and together).
In the regression tests, we have 5 candidates for the 4 slots of the reduction.
If the one "wrong" compare is grouped with the others, it prevents forming the
ideal v4i1 compare reduction.
Differential Revision: https://reviews.llvm.org/D87772
Reapply after fixing SimplifyWithOpReplaced() to never return
the original value, which would lead to an infinite loop in this
transform.
-----
For selects of the type X == Y ? A : B, check if we can simplify A
by using the X == Y equality and replace the operand if that's
possible. We already try to do this in InstSimplify, but will only
fold if the result of the simplification is the same as B, in which
case the select can be dropped entirely. Here the select will be
retained, just one operand simplified.
As we are performing an actual replacement here, we don't have
problems with refinement / poison values.
Differential Revision: https://reviews.llvm.org/D87480
https://bugs.llvm.org/show_bug.cgi?id=45932
assert(OutlinedFunctionCost >= Cloner.OutlinedRegionCost && "Outlined function cost should be no less than the outlined region") getting triggered in computeBBInlineCost.
Intrinsics like "assume" are considered regular function calls while computing costs.
This patch enables computeBBInlineCost to queries TTI for intrinsic call cost.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D87132
D65060 was reverted because it introduced non-determinism by using BFI counts from already freed blocks. The parent of this revision fixes that by using a VH callback on blocks to prevent this from happening and makes sure BFI data is passed correctly in LoopStandardAnalysisResults.
This re-introduces the previous optimization of using BFI data to prevent LICM from hoisting/sinking if the instruction will end up moving to a colder block.
Internally at Facebook this change results in a ~7% win in a CPU related metric in one of our big services by preventing hoisting cold code into a hot pre-header like the added test case demonstrates.
Testing:
ninja check
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87551
~~D65060 uncovered that trying to use BFI in loop passes can lead to non-deterministic behavior when blocks are re-used while retaining old BFI data.~~
~~To make sure BFI is preserved through loop passes a Value Handle (VH) callback is registered on blocks themselves. When a block is freed it now also wipes out the accompanying BFI entry such that stale BFI data can no longer persist resolving the determinism issue. ~~
~~An optimistic approach would be to incrementally update BFI information throughout the loop passes rather than only invalidating them on removed blocks. The issues with that are:~~
~~1. It is not clear how BFI information should be incrementally updated: If a block is duplicated does its BFI information come with? How about if it's split/modified/moved around? ~~
~~2. Assuming we can address these problems the implementation here will be a massive undertaking. ~~
~~There's a known need of BFI in LICM analysis which requires correct but not incrementally updated BFI data. A follow-up change can register BFI in all loop passes so this preserved but potentially lossy data is available to any loop pass that wants it.~~
See: D75341 for an identical implementation of preserving BFI via VH callbacks. The previous statements do still apply but this change no longer has to be in this diff because it's already upstream 😄 .
This diff also moves BFI to be a part of LoopStandardAnalysisResults since the previous method using getCachedResults now (correctly!) statically asserts (D72893) that this data isn't static through the loop passes.
Testing
Ninja check
Reviewed By: asbirlea, nikic
Differential Revision: https://reviews.llvm.org/D86156
Call instructions with musttail tag must be optimized as a tailcall, otherwise could lead to incorrect program behavior.
When TSAN is instrumenting functions, it broke the contract by adding a call to the tsan exit function inbetween the musttail call and return instruction, and also inserted exception handling code.
This happend throguh EscapeEnumerator, which adds exception handling code and returns ret instructions as the place to insert instrumentation calls.
This becomes especially problematic for coroutines, because coroutines rely on tail calls to do symmetric transfers properly.
To fix this, this patch moves the location to insert instrumentation calls prior to the musttail call for ret instructions that are following musttail calls, and also does not handle exception for musttail calls.
Differential Revision: https://reviews.llvm.org/D87620
For scalable type, the aggregated size is unknown at compile-time.
Skip instructions with scalable type to ensure the list of instructions
for vectorizeSimpleInstructions does not contains any scalable-vector instructions.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D87550
Invalid IR in unreachable code is technically valid IR. In this case,
the address space of the value was never inferred, and we tried to
rewrite it with an invalid address space value which would assert.
This patch is a first draft of a new pass that adds a more flexible way
to eliminate compares based on more complex constraints collected from
dominating conditions.
In particular, it aims at simplifying conditions of the forms below
using a forward propagation approach, rather than instcomine-style
ad-hoc backwards walking of def-use chains.
if (x < y)
if (y < z)
if (x < z) <- simplify
or
if (x + 2 < y)
if (x + 1 < y) <- simplify assuming no wraps
The general approach is to collect conditions and blocks, sort them by
dominance and then iterate over the sorted list. Conditions are turned
into a linear inequality and add it to a system containing the linear
inequalities that hold on entry to the block. For blocks, we check each
compare against the system and see if it is implied by the constraints
in the system.
We also keep a stack of processed conditions and remove conditions from
the stack and the constraint system once they go out-of-scope (= do not
dominate the current block any longer).
Currently there still are the least the following areas for improvements
* Currently large unsigned constants cannot be added to the system
(coefficients must be represented as integers)
* The way constraints are managed currently is not very optimized.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D84547
Similar to the tsan suppression in
`Utils/VNCoercion.cpp:getLoadLoadClobberFullWidthSize` (rL175034; load widening used by GVN),
the D81766 optimization should be suppressed under tsan due to potential
spurious data race reports:
struct A {
int i;
const short s; // the load cannot be vectorized because
int modify; // it overlaps with bytes being concurrently modified
long pad1, pad2;
};
// __tsan_read16 does not know that some bytes are undef and accessing is safe
Similarly, under asan, users can mark memory regions with
`__asan_poison_memory_region`. A widened load can lead to a spurious
use-after-poison error. hwasan/memtag should be similarly suppressed.
`mustSuppressSpeculation` suppresses asan/hwasan/tsan but not memtag, so
we need to exclude memtag in `vectorizeLoadInsert`.
Note, memtag suppression can be relaxed if the load is aligned to the
its granule (usually 16), but that is out of scope of this patch.
Reviewed By: spatel, vitalybuka
Differential Revision: https://reviews.llvm.org/D87538
Forward declare AAResults instead of the (old) AliasAnalysis type.
Remove includes from SLPVectorizer.cpp that are already included in SLPVectorizer.h.
The FailureReason input parameter maybe null, we check this in all other cases in the method but this one was missed somehow.
Fixes clang-tidy warning.
The "takeName" logic in ScalarizerVisitor::gather did not consider
that the value vector could refer to non-instructions, such as
global variables. This patch make sure that we avoid changing the
name of a value if it isn't an instruction.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D87685
I've amended the isLoadInvariantInLoop function to bail out for
scalable vectors for now since the invariant.start intrinsic is only
ever generated by the clang frontend for thread locals or struct
and class constructors, neither of which support sizeless types.
In addition, the intrinsic itself does not currently support the
concept of a scaled size, which makes it impossible to compare
the sizes of different scalable objects, e.g. <vscale x 32 x i8>
and <vscale x 16 x i8>.
Added new tests here:
Transforms/LICM/AArch64/sve-load-hoist.ll
Transforms/LICM/hoisting.ll
Differential Revision: https://reviews.llvm.org/D87227
As to not conflict with the legacy PM example passes under
llvm/lib/Transforms/Hello, this is under HelloNew. This makes the
CMakeLists.txt and general directory structure less confusing for people
following the example.
Much of the doc structure was taken from WritinAnLLVMPass.rst.
This adds a HelloWorld pass which simply prints out each function name.
More will follow after this, e.g. passes over different units of IR, analyses.
https://llvm.org/docs/WritingAnLLVMPass.html contains a lot more.
Relanded with missing "Support" dependency in LLVMBuild.txt.
Reviewed By: ychen, asbirlea
Differential Revision: https://reviews.llvm.org/D86979
As to not conflict with the legacy PM example passes under
llvm/lib/Transforms/Hello, this is under HelloNew. This makes the
CMakeLists.txt and general directory structure less confusing for people
following the example.
Much of the doc structure was taken from WritinAnLLVMPass.rst.
This adds a HelloWorld pass which simply prints out each function name.
More will follow after this, e.g. passes over different units of IR, analyses.
https://llvm.org/docs/WritingAnLLVMPass.html contains a lot more.
Reviewed By: ychen, asbirlea
Differential Revision: https://reviews.llvm.org/D86979
This is consistent with the clang option added in
7ed8124d46, and the comments on the
runtime patch in D87120.
Differential Revision: https://reviews.llvm.org/D87622
For selects of the type X == Y ? A : B, check if we can simplify A
by using the X == Y equality and replace the operand if that's
possible. We already try to do this in InstSimplify, but will only
fold if the result of the simplification is the same as B, in which
case the select can be dropped entirely. Here the select will be
retained, just one operand simplified.
As we are performing an actual replacement here, we don't have
problems with refinement / poison values.
Differential Revision: https://reviews.llvm.org/D87480
AliasAnalysis/MemoryLocation does not account for loops. Two
MemoryLocation can be must-overwrite, even if the first one writes
multiple locations in a loop.
This patch prevents removing such stores, by only considering candidates
that are known to be loop invariant, or executed in the same BB.
Currently the invariant check is quite conservative and only considers
Alloca and Alloca-like instructions and arguments as invariant base pointers.
It also considers GEPs with all constant indices and invariant bases as
invariant.
This can be improved in the future, but the current implementation has
only minor impact on the total number of stores eliminated (25903 vs
26047 for the baseline). There are some 2-10% swings for some individual
benchmarks. In roughly half of the cases, the number of stores removed
increases actually, because we skip candidates that are unlikely to be
valid candidates early.
The function LoopIdiomRecognize::isLegalStore looks for stores in loops
that could be transformed into memset or memcpy. However, the algorithm
currently requires that we know how big the store is at runtime, i.e.
that the store size will not overflow an unsigned integer. For scalable
vectors we cannot guarantee this so I have changed the code to bail out
for now. In addition, even if we add a way to query the maximum value of
vscale in future we will still need to update the algorithm to cope with
non-constant strides. The additional cost associated with calculating
the memset and memcpy arguments will need to be taken into account as
well.
This patch also fixes up an implicit TypeSize -> uint64_t cast,
thereby removing a warning. I've added tests here showing a fixed
width vector loop being transformed into memcpy, and a scalable
vector loop remaining unchanged:
Transforms/LoopIdiom/memcpy-vectors.ll
Differential Revision: https://reviews.llvm.org/D87439
This fixes PR47297.
When ProcessBlock() was able to constant fold the terminator's
condition, but not do any more transformations, the function would
return false, which would lead to the JumpThreading pass returning an
incorrect modified status. This patch makes so that ProcessBlock()
returns true in such cases. This will trigger an unnecessary invocation
of ProcessBlock() in such cases, but this should be rare to occur.
This was caught using the check introduced by D80916.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D87392
Turns out this was use-after-move of function_ref, which is trivially
copyable and movable, so the move did nothing and use after move was
safe.
But since this function_ref is being copied into a std::function, change
the function_ref to be std::function to avoid extra layers of type
erasure indirection - and then it's a real use after move, and fix that
by referring to the moved-to member variable rather than the moved-from
parameter.
gcov is an "Edge Profiling with Edge Counters" application according to
Optimally Profiling and Tracing Programs (1994).
The minimum number of counters necessary is |E|-(|V|-1). The unmeasured edges
form a spanning tree. Both GCC --coverage and clang -fprofile-generate leverage
this optimization. This patch implements the optimization for clang --coverage.
The produced .gcda files are much smaller now.
i.e. change the work flow from
* .gcno for function A
* .gcno for function B
* .gcno for function C
* .gcda for function A
* .gcda for function B
* .gcda for function C
to
* .gcno for function A
* .gcda for function A
* .gcno for function B
* .gcda for function B
* .gcno for function C
* .gcda for function C
Currently there is duplicate logic in .gcno & .gcda processing: how functions
are filtered, which edges are instrumented, etc. This refactor enables simplification.
Since we always process .gcno, in -fprofile-arcs -fno-test-coverage mode,
__llvm_internal_gcov_emit_function_args.0 will have non-zero checksums.
When deleting stores at the end of a function, we have to do PHI
translation, otherwise we might miss reads in different iterations of a
loop. See multiblock-loop-carried-dependence.ll for details.
This fixes a mis-compile and surprisingly also increases the number of
eliminated stores from 26047 to 26572 for MultiSource/SPEC2000/SPEC2006
on X86 with -O3 -flto. This is most likely because we save budget by not
exploring through MemoryPhis, which are less likely to result in valid
candidates for elimination.
The issue was reported post-commit for fb109c42d9.
This allows the backend to tell the vectorizer to produce inloop
reductions through a TTI hook.
For the moment on ARM under MVE this means allowing integer add
reductions of the correct size. In the future this can include integer
min/max too, under -Os.
Differential Revision: https://reviews.llvm.org/D75512
NOTE: There is a mailing list discussion on this: http://lists.llvm.org/pipermail/llvm-dev/2019-December/137632.html
Complemantary to the assumption outliner prototype in D71692, this patch
shows how we could simplify the code emitted for an alignemnt
assumption. The generated code is smaller, less fragile, and it makes it
easier to recognize the additional use as a "assumption use".
As mentioned in D71692 and on the mailing list, we could adopt this
scheme, and similar schemes for other patterns, without adopting the
assumption outlining.
This is a followup to D86834, which partially fixed this issue in
InstSimplify. However, InstCombine repeats the same transform while
dropping poison flags -- which does not cover cases where poison is
introduced in some other way.
The fix here is a bit more comprehensive, because things are quite
entangled, and it's hard to only partially address it without
regressing optimization. There are really two changes here:
* Export the SimplifyWithOpReplaced API from InstSimplify, with an
added AllowRefinement flag. For replacements inside the TrueVal
we don't actually care whether refinement occurs or not, the
replacement is always legal. This part of the transform is now
done in InstSimplify only. (It should be noted that the current
AllowRefinement check is not sufficient -- that's an issue we
need to address separately.)
* Change the InstCombine fold to work by temporarily dropping
poison generating flags, running the fold and then restoring the
flags if it didn't work out. This will ensure that the InstCombine
fold is correct as long as the InstSimplify fold is correct.
Differential Revision: https://reviews.llvm.org/D87445
The test example based on PR47450 shows that we can
match non-byte-sized shifts, but those won't ever be
bswap opportunities. This isn't a full fix (we'd still
match if the shifts were by 8-bits for example), but
this should be enough until there's evidence that we
need to do more (this is a borderline case for
vectorization in the first place).
When inlining functions containing allocas of scalable vectors we
cannot specify the size in the lifetime markers, since we don't
know this at compile time.
Added new test here:
test/Transforms/Inline/AArch64/sve-alloca-merge.ll
Differential Revision: https://reviews.llvm.org/D87139
In addition to calculate hash consistently by swapping SELECT's
operands, we also need to inverse the select pattern favor to match the
original logic.
[EarlyCSE] Equivalent SELECTs should hash equally
DenseMap<SimpleValue> assumes that, if its isEqual method returns true
for two elements, then its getHashValue method must return the same value
for them. This invariant is broken when one SELECT node is a min/max
operation, and the other can be transformed into an equivalent min/max by
inverting its predicate and swapping its operands. This patch fixes an
assertion failure that would occur intermittently while compiling the
following IR:
define i32 @t(i32 %i) {
%cmp = icmp sle i32 0, %i
%twin1 = select i1 %cmp, i32 %i, i32 0
%cmpinv = icmp sgt i32 0, %i
%twin2 = select i1 %cmpinv, i32 0, i32 %i
%sink = add i32 %twin1, %twin2
ret i32 %sink
}
Differential Revision: https://reviews.llvm.org/D86843
The tests have been updated and I plan to move them from the MSSA
directory up.
Some end-to-end tests needed small adjustments. One difference to the
legacy DSE is that legacy DSE also deletes trivially dead instructions
that are unrelated to memory operations. Because MemorySSA-backed DSE
just walks the MemorySSA, we only visit/check memory instructions. But
removing unrelated dead instructions is not really DSE's job and other
passes will clean up.
One noteworthy change is in llvm/test/Transforms/Coroutines/ArgAddr.ll,
but I think this comes down to legacy DSE not handling instructions that
may throw correctly in that case. To cover this with MemorySSA-backed
DSE, we need an update to llvm.coro.begin to treat it's return value to
belong to the same underlying object as the passed pointer.
There are some minor cases MemorySSA-backed DSE currently misses, e.g. related
to atomic operations, but I think those can be implemented after the switch.
This has been discussed on llvm-dev:
http://lists.llvm.org/pipermail/llvm-dev/2020-August/144417.html
For the MultiSource/SPEC2000/SPEC2006 the number of eliminated stores
goes from ~17500 (legayc DSE) to ~26300 (MemorySSA-backed). More numbers
and details in the thread on llvm-dev.
Impact on CTMark:
```
Legacy Pass Manager
exec instrs size-text
O3 + 0.60% - 0.27%
ReleaseThinLTO + 1.00% - 0.42%
ReleaseLTO-g. + 0.77% - 0.33%
RelThinLTO (link only) + 0.87% - 0.42%
RelLO-g (link only) + 0.78% - 0.33%
```
http://llvm-compile-time-tracker.com/compare.php?from=3f22e96d95c71ded906c67067d75278efb0a2525&to=ae8be4642533ff03803967ee9d7017c0d73b0ee0&stat=instructions
```
New Pass Manager
exec instrs. size-text
O3 + 0.95% - 0.25%
ReleaseThinLTO + 1.34% - 0.41%
ReleaseLTO-g. + 1.71% - 0.35%
RelThinLTO (link only) + 0.96% - 0.41%
RelLO-g (link only) + 2.21% - 0.35%
```
http://195.201.131.214:8000/compare.php?from=3f22e96d95c71ded906c67067d75278efb0a2525&to=ae8be4642533ff03803967ee9d7017c0d73b0ee0&stat=instructions
Reviewed By: asbirlea, xbolva00, nikic
Differential Revision: https://reviews.llvm.org/D87163
DenseMap<SimpleValue> assumes that, if its isEqual method returns true
for two elements, then its getHashValue method must return the same value
for them. This invariant is broken when one SELECT node is a min/max
operation, and the other can be transformed into an equivalent min/max by
inverting its predicate and swapping its operands. This patch fixes an
assertion failure that would occur intermittently while compiling the
following IR:
define i32 @t(i32 %i) {
%cmp = icmp sle i32 0, %i
%twin1 = select i1 %cmp, i32 %i, i32 0
%cmpinv = icmp sgt i32 0, %i
%twin2 = select i1 %cmpinv, i32 0, i32 %i
%sink = add i32 %twin1, %twin2
ret i32 %sink
}
Differential Revision: https://reviews.llvm.org/D86843
Bail from maskIsAllZeroOrUndef and maskIsAllOneOrUndef prior to iterating over the number of
elements for scalable vectors.
Assert that the mask type is not scalable in possiblyDemandedEltsInMask .
Assert that the types are correct in all three functions.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D87424
Previously we could match fcmp+select to a reduction if the fcmp had
the nonans fast math flag. But if the select had the nonans fast
math flag, InstCombine would turn it into a fminnum/fmaxnum intrinsic
before SLP gets to it. Seems fairly likely that if one of the
fcmp+select pair have the fast math flag, they both would.
My plan is to start vectorizing the fmaxnum/fminnum version soon,
but I wanted to get this code out as it had some of the strangest
fast math flag behaviors.
PGOInstrumentation runs `SplitIndirectBrCriticalEdges` but some IndirectBrInst
critical edge cannot be split. `getInstrBB` will crash when calling `SplitCriticalEdge`, e.g.
int foo(char *p) {
void *targets[2];
targets[0] = &&indirect;
targets[1] = &&end;
for (;; p++)
if (*p == 7) {
indirect:
goto *targets[p[1]]; // the self loop is critical in -O
}
end:
return 0;
}
Skip such critical edges to prevent a crash.
Reviewed By: davidxl, lebedev.ri
Differential Revision: https://reviews.llvm.org/D87435
The argument promotion pass currently fails to copy function annotations
over to the modified function after promoting arguments.
This patch copies the original function annotation to the new function.
Reviewed By: fhann
Differential Revision: https://reviews.llvm.org/D86630
See discussion in D87149. Dropping volatile stores here is legal
per LLVM semantics, but causes issues for real code and may result
in a change to LLVM volatile semantics. Temporarily treat volatile
stores as "not guaranteed to transfer execution" in just this place,
until this issue has been resolved.
MemoryLocation has been taught about memcpy.inline, which means we can
get the memory locations read and written by it. This means DSE can
handle memcpy.inline
This patch enables inserting freeze when JumpThreading converts a select to
a conditional branch when it is run in LTO.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D85534
As code size is the only thing we care about at minsize, query the
cost of materialising immediates when calculating the cost of a SCEV
expansion. We also modify the CostKind to TCK_CodeSize for minsize,
instead of RecipThroughput.
Differential Revision: https://reviews.llvm.org/D76434
This patch fixes pr45956 (https://bugs.llvm.org/show_bug.cgi?id=45956 ).
To minimize its impact to the quality of generated code, I suggest enabling
this only for LTO as a start (it has two JumpThreading passes registered).
This patch contains a flag that makes JumpThreading enable it.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D84940
The test in PR47457 demonstrates a situation when candidate load's pointer's SCEV
is no loger a SCEVAddRec after loop versioning. The code there assumes that it is
always a SCEVAddRec and crashes otherwise.
This patch makes sure that we do not consider candidates for which this requirement
is broken after the versioning.
Differential Revision: https://reviews.llvm.org/D87355
Reviewed By: asbirlea
Atomic stores are modeled as MemoryDef to model the fact that they may
not be reordered, depending on the ordering constraints.
Atomic stores that are monotonic or weaker do not limit re-ordering, so
we do not have to treat them as potential read clobbers.
Note that llvm/test/Transforms/DeadStoreElimination/MSSA/atomic.ll
already contains a set of negative test cases.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87386
The entry block is split at the first instruction where `shouldKeepInEntry`
returns false. The created basic block has a br jumping to the original entry
block. The new basic block causes the function label line and the other entry
block lines to be covered by different basic blocks, which can affect line
counts with special control flows (fork/exec in the entry block requires
heuristics in llvm-cov gcov to get consistent line counts).
int main() { // BB0
return 0; // BB2 (due to entry block splitting)
}
// BB1 is the exit block (since gcov 4.8)
This patch adds a synthetic entry block (like PGOInstrumentation and GCC) and
inserts an edge from the synthetic entry block to the original entry block. We
can thus remove the tricky `shouldKeepInEntry` and entry block splitting. The
number of basic blocks does not change, but the emitted .gcno files will be
smaller because we can save one GCOV_TAG_LINES tag.
// BB0 is the synthetic entry block with a single edge to BB2
int main() { // BB2
return 0; // BB2
}
// BB1 is the exit block (since gcov 4.8)
This is the initial part of the implementation of the C++20 likelihood
attributes. It handles the attributes in an if statement.
Differential Revision: https://reviews.llvm.org/D85091
If a function had at most one return block, the pass would return false
regardless if an unified unreachable block was created.
This patch fixes that by refactoring runOnFunction into two separate
helper functions for handling the unreachable blocks respectively the
return blocks, as suggested by @bjope in a review comment.
This was caught using the check introduced by D80916.
Reviewed By: serge-sans-paille
Differential Revision: https://reviews.llvm.org/D85818
This patch follows D85345 and adds more noundef attributes to return values/arguments of library functions
that are mostly about accessing the file system or processes.
A few functions like `chmod` or `times` use typedef `mode_t` and `clock_t`.
They are neither struct nor union, so they cannot contain undef even if they're lowered to iN in IR. So, it is fine to add noundef to them.
- clock_t's actual type is size_t (C17, 7.27.1.3), so it isn't struct or union.
- For mode_t, either int or long is used in practice because programmers use bit manipulation. So, I think it is okay that it's never aggregate in practice.
After this patch, the remaining library functions are those that eagerly participate in optimizations: they can be removed, reordered, or
introduced by a transformation from primitive IR operations.
For them, a few testings is needed, since it may not be valid to add noundef anymore even if C standard says it's okay.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D85894
This patch adds isGuaranteedNotToBePoison and programUndefinedIfUndefOrPoison.
isGuaranteedNotToBePoison will be used at D75808. The latter function is used at isGuaranteedNotToBePoison.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D84242
The MemorySSAWrapperPass depends on AAResultsWrapperPass and if
MemorySSA is preserved but AAResultsWrapperPass is not, this could lead
to a crash when updating the last user of the MemorySSAWrapperPass.
Alternatively AAResultsWrapperPass could be marked preserved by GVN, but
I am not sure if that would be safe. I am not sure what is required in
order to preserve AAResultsWrapperPass. At the moment, it seems like a
couple of passes that do similar transforms to GVN are preserving it.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87137
This commit cleans up the ::initialize method of various AAs in the
following ways:
- If an associated function is required, give up on declarations.
This was discovered as a real problem when lots of llvm.dbg.XXX
call sites were assumed `noreturn` until proven otherwise. That
does not make any sense and caused huge regressions and missed
deductions.
- Require more associated declarations for function interface AAs.
- Use the IRAttribute::initialize to determine if function interface
AAs can be used in IPO, don't replicate the checks (especially
isFunctionIPOAmendable) all over the place. Arguably the function
declaration check should be moved to some central place to.
If we have a callback, call site arguments were already associated with
the callback callee. Now we also associate the function with the
callback callee, thus we know ensure that the following holds true (if
all return nonnull):
`getAssociatedArgument()->getParent() == getAssociatedFunction()`
To test this an early exit from
`AAMemoryBehaviorCallSiteArgument::initialize``
is included as well. Without the change to getAssociatedFunction() this
kind of early exit for declarations would cause callback call site
arguments to miss out.
As we handle callback calls we need to disambiguate the call site
argument number from the callee argument number. While always equal in
non-callback calls, a callback comes with a partial parameter-argument
mapping so there is no implicit correspondence. Here we split
`IRPosition::getArgNo()` into two public functions, `getCallSiteArgNo()`
and `getCalleeArgNo()`. Usages are adjusted to pick the right one for
their purpose. This fixed some problems that would have been exposed as
we more aggressively optimize callbacks.
While operand bundles carry unpredictable semantics, we know some of
them and can therefore "ignore" them. In this case we allow to look at
the declaration of `llvm.assume` when asked for the attributes at a call
site. The assume operand bundles we have do not invalidate the
declaration attributes.
We cannot test this in isolation because the llvm.assume attributes are
determined by the parser. However, a follow up patch will provide test
coverage.
In `MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4.cpp` we initialized
attributes until stack frame ~35k caused space to run out. The initial
size 1024 is pretty much random.
The get{Return,Unwind,Unreachable}Block functions in
UnifyFunctionExitNodes have not been used for many years,
so just remove them.
Reviewed By: bjope
Differential Revision: https://reviews.llvm.org/D87078
If we know that the abs operand is known negative, we can replace
it with a neg.
To avoid computing known bits twice, I've removed the fold for the
non-negative case from InstSimplify. Both the non-negative and the
negative case are handled by InstCombine now, with one known bits call.
Differential Revision: https://reviews.llvm.org/D87196
D66230 attempted to fix a problem where when there are allocas used before CoroBegin.
It keeps allocas and their uses stay in put if there are no escapse/changes to the data before CoroBegin.
Unfortunately that's incorrect.
Consider this code:
%var = alloca i32
%1 = getelementptr .. %var; stays put
%f = call i8* @llvm.coro.begin
store ... %1
After this fix, %1 will now stay put, however if a store happens after coro.begin and hence modifies the content, this change will not be reflected in the coroutine frame (and will eventually be DCEed).
To generalize the problem, if any alias ptr is created before coro.begin for an Alloca and that alias ptr is latter written into after coro.begin, it will lead to incorrect behavior.
There are also a few other minor issues, such as incorrect dominate condition check in the ptr visitor, unhandled memory intrinsics and etc.
Ths patch attempts to fix some of these issue, and make it more robust to deal with aliases.
While visiting through the alloca pointer, we also keep track of all aliases created that will be used after CoroBegin. We track the offset of each alias, and then reacreate these aliases after CoroBegin using these offset.
It's worth noting that this is not perfect and there will still be cases we cannot handle. I think it's impractical to handle all cases given the current design.
This patch makes it more robust and should be a pure win.
In the meantime, we need to think about what how to completely elimiante these issues, likely through the route as @rjmccall mentioned in D66230.
Differential Revision: https://reviews.llvm.org/D86859
In GenerateConstantOffsetsImpl, we may generate non canonical Formula
if BaseRegs of that Formula is updated and includes a recurrent expr reg
related with current loop while its ScaledReg is not.
Patched by: mdchen
Reviewed By: qcolombet
Differential Revision: https://reviews.llvm.org/D86939
The CloneFunctionInto has implicit requirements with regards to the
linkage and visibility of the function. We now update these after we did
the CloneFunctionInto on the copy with the same linkage and visibility
as the original.
Deleting or replacing anything is certainly a modification. This caused
a later assertion in IPSCCP when compiling 400.perlbench with the new PM.
I'm not sure how to test this.
This was reverted in 503deec218
because it caused gigantic increase (3x) in branch mispredictions
in certain benchmarks on certain CPU's,
see https://reviews.llvm.org/D84108#2227365.
It has since been investigated and here are the results:
https://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20200907/827578.html
> It's an amazingly severe regression, but it's also all due to branch
> mispredicts (about 3x without this). The code layout looks ok so there's
> probably something else to deal with. I'm not sure there's anything we can
> reasonably do so we'll just have to take the hit for now and wait for
> another code reorganization to make the branch predictor a bit more happy :)
>
> Thanks for giving us some time to investigate and feel free to recommit
> whenever you'd like.
>
> -eric
So let's just reland this.
Original commit message:
I've been looking at missed vectorizations in one codebase.
One particular thing that stands out is that some of the loops
reach vectorizer in a rather mangled form, with weird PHI's,
and some of the loops aren't even in a rotated form.
After taking a more detailed look, that happened because
the loop's headers were too big by then. It is evident that
SimplifyCFG's common code hoisting transform is at fault there,
because the pattern it handles is precisely the unrotated
loop basic block structure.
Surprizingly, `SimplifyCFGOpt::HoistThenElseCodeToIf()` is enabled
by default, and is always run, unlike it's friend, common code sinking
transform, `SinkCommonCodeFromPredecessors()`, which is not enabled
by default and is only run once very late in the pipeline.
I'm proposing to harmonize this, and disable common code hoisting
until //late// in pipeline. Definition of //late// may vary,
here currently i've picked the same one as for code sinking,
but i suppose we could enable it as soon as right after
loop rotation happens.
Experimentation shows that this does indeed unsurprizingly help,
more loops got rotated, although other issues remain elsewhere.
Now, this undoubtedly seriously shakes phase ordering.
This will undoubtedly be a mixed bag in terms of both compile- and
run- time performance, codesize. Since we no longer aggressively
hoist+deduplicate common code, we don't pay the price of said hoisting
(which wasn't big). That may allow more loops to be rotated,
so we pay that price. That, in turn, that may enable all the transforms
that require canonical (rotated) loop form, including but not limited to
vectorization, so we pay that too. And in general, no deduplication means
more [duplicate] instructions going through the optimizations. But there's still
late hoisting, some of them will be caught late.
As per benchmarks i've run {F12360204}, this is mostly within the noise,
there are some small improvements, some small regressions.
One big regression i saw i fixed in rG8d487668d09fb0e4e54f36207f07c1480ffabbfd, but i'm sure
this will expose many more pre-existing missed optimizations, as usual :S
llvm-compile-time-tracker.com thoughts on this:
http://llvm-compile-time-tracker.com/compare.php?from=e40315d2b4ed1e38962a8f33ff151693ed4ada63&to=c8289c0ecbf235da9fb0e3bc052e3c0d6bff5cf9&stat=instructions
* this does regress compile-time by +0.5% geomean (unsurprizingly)
* size impact varies; for ThinLTO it's actually an improvement
The largest fallout appears to be in GVN's load partial redundancy
elimination, it spends *much* more time in
`MemoryDependenceResults::getNonLocalPointerDependency()`.
Non-local `MemoryDependenceResults` is widely-known to be, uh, costly.
There does not appear to be a proper solution to this issue,
other than silencing the compile-time performance regression
by tuning cut-off thresholds in `MemoryDependenceResults`,
at the cost of potentially regressing run-time performance.
D84609 attempts to move in that direction, but the path is unclear
and is going to take some time.
If we look at stats before/after diffs, some excerpts:
* RawSpeed (the target) {F12360200}
* -14 (-73.68%) loops not rotated due to the header size (yay)
* -272 (-0.67%) `"Number of live out of a loop variables"` - good for vectorizer
* -3937 (-64.19%) common instructions hoisted
* +561 (+0.06%) x86 asm instructions
* -2 basic blocks
* +2418 (+0.11%) IR instructions
* vanilla test-suite + RawSpeed + darktable {F12360201}
* -36396 (-65.29%) common instructions hoisted
* +1676 (+0.02%) x86 asm instructions
* +662 (+0.06%) basic blocks
* +4395 (+0.04%) IR instructions
It is likely to be sub-optimal for when optimizing for code size,
so one might want to change tune pipeline by enabling sinking/hoisting
when optimizing for size.
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D84108
This reverts commit 503deec218.
For intrinsics supported by ConstantRange, compute the result range
based on the argument ranges. We do this independently of whether
some or all of the input ranges are full, as we can often still
constrain the result in some way.
Differential Revision: https://reviews.llvm.org/D87183
This was supposed to be an NFC cleanup, but there's
a real logic difference (did not drop 'nsw') visible
in some tests in addition to an efficiency improvement.
This is because in the case where we have 2 GEPs,
the code was *always* swapping the operands and
negating the result. But if we have 2 GEPs, we
should *never* need swapping/negation AFAICT.
This is part of improving flags propagation noticed
with PR47430.
Normal dead code elimination ignores assume intrinsics, so we fail to
delete assumes that are not meaningful (and potentially worse if they
cause conflicts with other assumptions).
The motivating example in https://llvm.org/PR47416 suggests that we
might have problems upstream from here (difference between C and C++),
but this should be a cheap way to make sure we remove more dead code.
Differential Revision: https://reviews.llvm.org/D87149
To enable the cost of constants, the helper function has been
reorganised:
- A struct has been introduced to hold SCEV operand information so
that we know the user of the operand, as well as the operand index.
The Worklist now uses instead instead of a bare SCEV.
- The costing of each SCEV, and collection of its operands, is now
performed in a helper function.
Differential Revision: https://reviews.llvm.org/D86050
Modify FoldBranchToCommonDest to consider the cost of inserting
instructions when attempting to combine predicates to fold blocks.
The threshold can be controlled via a new option:
-simplifycfg-branch-fold-threshold which defaults to '2' to allow
the insertion of a not and another logical operator.
Differential Revision: https://reviews.llvm.org/D86526
Similar to D87168, but for abs. If we have a dominating x >= 0
condition, then we know that abs(x) is x. This fold is in
InstCombine, because we need to create a sub instruction for
the x < 0 case.
Differential Revision: https://reviews.llvm.org/D87184
When a switch case is folded into default's case, that's an IR change that
should be reported, update ConstantFoldTerminator accordingly.
Differential Revision: https://reviews.llvm.org/D87142
When checking call sites, give special handling to indirect call, as the
callee may be unknown and can lead to nullptr dereference later. Assume
conservatively that the ICV always changes in such case.
Reviewed By: sstefan1
Differential Revision: https://reviews.llvm.org/D87104
This patch updates MemCpyOpt to preserve MemorySSA. It uses the
MemoryDef at the insertion point of the builder and inserts the new def
after that def.
In some cases, we just modify a memory instruction. In that case, get
the defining access, then remove the memory access and add a new one.
If the defining access is in a different block, insert a new def at the
beginning of the current block, otherwise after the defining access.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86651
As far as I am aware, the placement of MergedLoadStoreMotion in the
pipeline is not heavily tuned currently. It seems to not matter much if
we do it after DSE in the LTO pipeline (no binary changes for -O3 -flto
on MultiSource/SPEC2000/SPEC2006). Moving it after DSE however has a
major benefit: MemorySSA is constructed by LICM and is consumed by DSE,
so if MergedLoadStoreMotion happens after DSE, we do not need to
preserve MemorySSA in it.
If there are any concerns with this move, I can also update
MergedLoadStoreMotion to preserve MemorySSA.
This patch together with D86651 (preserve MemSSA in MemCpyOpt) and
D86534 (preserve MemSSA in GVN) are the remaining patches to bring down
compile-time for DSE + MemorySSA to the levels outlined in
http://lists.llvm.org/pipermail/llvm-dev/2020-August/144417.html
Once they land, we should be able to start with flipping the switch on
enabling DSE + MmeorySSA.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86967
Preserve MemorySSA if it is available before running GVN.
DSE with MemorySSA will run closely after GVN. If GVN and 2 other
passes preserve MemorySSA, DSE can re-use MemorySSA used by LICM
when doing LTO.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86534
This is a followup to 1ccfb52a61, which made a number of changes
including the apparently innocuous reordering of required passes in
MemCpyOptimizer. This however altered the creation order of BasicAA vs
Phi Values analysis, meaning BasicAA did not pick up PhiValues as a
cached result. Instead if we require MemoryDependence first it will
require PhiValuesAnalysis allowing BasicAA to use it for better results.
I don't claim this is an excellent design, but it fixes a nasty little
regressions where a query later in JumpThreading was getting worse
results.
Differential Revision: https://reviews.llvm.org/D87027
Currently IPSCCP (and others like CVP/GVN) blindly propagate pointer
equalities. In certain cases, that leads to dereferenceable pointers
being replaced, as in the example test case.
I think this is not allowed, as it introduces an access of an
un-dereferenceable pointer. Note that the pointer is inbounds, but one
past the last element, so it is valid, but not dereferenceable.
This patch is mostly to highlight the issue and start a discussion.
Currently it only checks for specifically looking
one-past-the-last-element pointers with array typed bases.
This causes the mis-compile outlined in
https://stackoverflow.com/questions/55754313/is-this-gcc-clang-past-one-pointer-comparison-behavior-conforming-or-non-standar
In the test case, if we replace %p with the GEP for the store, we
subsequently determine that the store and the load cannot alias, because
they are to different underlying objects.
Note that Alive2 seems to think that the replacement is valid:
https://alive2.llvm.org/ce/z/2rorhk
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D85332
First, shuffle cost for scalable type is not known for scalable type;
Second, we cannot reason if the narrowed shuffle mask for scalable type
is a splat or not.
E.g., Bitcast splat vector from type <vscale x 4 x i32> to <vscale x 8 x i16>
will involve narrowing shuffle mask <vscale x 4 x i32> zeroinitializer to
<vscale x 8 x i32> with element sequence of <0, 1, 0, 1, ...>, which cannot be
reasoned if it's a valid splat or not.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D86995
In IPSCCP when a function is optimized to return undef, it should clear the returned attribute for all its input arguments
and its corresponding call sites.
The bug is exposed when the value of an input argument of the function is assigned to a physical register and
because of the argument having a returned attribute, the value of this physical register will continue to be used
as the function return value right after the call instruction returns, even if the value that this register holds may
be clobbered during the function call. This potentially results in incorrect values being used afterwards.
Reviewed By: jdoerfert, fhahn
Differential Revision: https://reviews.llvm.org/D84220
When marking a global variable constant, and simplifying users using
CleanupConstantGlobalUsers(), the pass could incorrectly return false if
there were still some uses left, and no further optimizations was done.
This was caught using the check introduced by D80916.
This fixes PR46749.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D85837
These transforms will now be performed irrespective of the number of uses for the expression "1.0/sqrt(X)":
1.0/sqrt(X) * X => X/sqrt(X)
X * 1.0/sqrt(X) => X/sqrt(X)
We already handle more general cases, and we are intentionally not creating extra (and likely expensive)
fdiv ops in IR. This pattern is the exception to the rule because we always expect the Backend to reduce
X/sqrt(X) to sqrt(X), if it has the necessary (reassoc) fast-math-flags.
Ref: DagCombiner optimizes the X/sqrt(X) to sqrt(X).
Differential Revision: https://reviews.llvm.org/D86726
This is an enhancement to D81766 to allow loading the minimum target
vector type into an IR vector with a different number of elements.
In one of the motivating tests from PR16739, SLP creates <2 x float>
load ops mixed with <4 x float> insert ops, so we want to handle that
pattern in addition to potential oversized vectors created by the
vectorizers.
For now, we are assuming the insert/extract subvector with undef is
free because there is no exact corresponding TTI modeling for that.
Differential Revision: https://reviews.llvm.org/D86160
This patch makes it possible for AAUB to use information from AANoUndef.
This is the next patch of D86983
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86984
When the associated value is undef, we immediately forced to indicate a pessimistic fixpoint so far.
This patch changes the initialization to check the attribute given in IR at first and to indicate an optimistic fixpoint when it is given.
This change will enable us to catch , for example, the following case in AAUB.
```
call void @foo(i32 noundef undef)
```
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86983
Summary:
Analyses are preserved in MemCpyOptimizer.
Get analyses before running the pass and store the pointers, instead of
using lambdas and getting them every time on demand.
Reviewers: lenary, deadalnix, mehdi_amini, nikic, efriedma
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74494
Interleave for small loops that have reductions inside,
which breaks dependencies and expose.
This gives very significant performance improvements for some benchmarks.
Because small loops could be in very hot functions in real applications.
Differential Revision: https://reviews.llvm.org/D81416
Loop Idiom Recognize Pass (LIRP) attempts to transform loops with subscripted arrays
into memcpy/memset function calls. In some particular situation, this transformation
introduces negative impacts. For example: https://bugs.llvm.org/show_bug.cgi?id=47300
This patch will enable users to disable a particular part of the transformation, while
he/she can still enjoy the benefit brought about by the rest of LIRP. The default
behavior stays unchanged: no part of LIRP is disabled by default.
Reviewed By: etiotto (Ettore Tiotto)
Differential Revision: https://reviews.llvm.org/D86262
getValuesInOffloadArrays goes through the offload arrays in __tgt_target_data_begin_mapper getting the values stored in them before the call is issued.
call void @__tgt_target_data_begin_mapper(arg0, arg1,
i8** %offload_baseptrs, i8** %offload_ptrs, i64* %offload_sizes,
...)
Diferential Revision: https://reviews.llvm.org/D86300
The 1st try was reverted because I missed an assert that
needed softening.
As discussed in D86798 / rG09652721 , we were potentially
returning a different result for whether an Instruction
is commutable depending on if we call the base class or
derived class method.
This requires relaxing asserts in GVN, but that pass
seems to be working otherwise.
NewGVN requires more work because it uses different
code paths for numbering binops and calls.
For an instruction in the basic block BB, SinkingPass enumerates basic blocks
dominated by BB and BB's successors. For each enumerated basic block,
SinkingPass uses `AllUsesDominatedByBlock` to check whether the basic
block dominates all of the instruction's users. This is inefficient.
Use the nearest common dominator of all users to avoid enumerating the
candidate. The nearest common dominator may be in a parent loop which is
not beneficial. In that case, find the ancestors in the dominator tree.
In the case that the instruction has no user, with this change we will
not perform unnecessary move. This causes some amdgpu test changes.
A stage-2 x86-64 clang is a byte identical with this change.
As discussed in D86798 / rG09652721 , we were potentially
returning a different result for whether an Instruction
is commutable depending on if we call the base class or
derived class method.
This requires relaxing an assert in GVN, but that pass
seems to be working otherwise.
NewGVN requires more work because it uses different
code paths for numbering binops and calls.
addRuntimeChecks uses SCEVExpander, which relies on the DT/LoopInfo to
be up-to-date. Changing the CFG afterwards may invalidate some inserted
instructions, especially LCSSA phis.
Reorder the code to first update the CFG and then create the runtime
checks. This should not have any impact on the generated code, as we
adjust the CFG and generate runtime checks together.
Fixes PR47343.
As discussed in D86798, it's not clear if the caller code
works with a more liberal definition of "commutative" that
includes intrinsics like min/max. This makes the binop
restriction (current functionality is unchanged) explicit
until the code is audited/tested.
The problem with module slice has been addressed in D86319
Introduce two new AAs. AAICVTrackerFunctionReturned which checks if a
function can have a unique ICV value after it is finished, and
AAICVCallSiteReturned which checks AAICVTrackerFunctionReturned for a
call site. This enables us to check the value of a call and if it
changes the ICV. This also changes the approach in
`getReplacementValues()` to a worklist-based approach so we can explore
all relevant BBs.
Differential Revision: https://reviews.llvm.org/D85544
Summary:
The module slice describes which functions we can analyze and transform
while working on an SCC as part of the Attributor-CGSCC pass. So far we
simply restricted it to the SCC.
Reviewers: jdoerfert
Differential Revision: https://reviews.llvm.org/D86319
This is the next patch of D86842
When we check `noundef` attribute violation at callsites, we do not have to require `nonnull` in the following two cases.
1. An argument is known to be simplified to undef
2. An argument is known to be dead
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86845
Even though `noundef` IR attribute might be attached to non-void type values, AANoUndef is mistakenly identified for pointer type values only.
This patch fixes that.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86737
Move bail out when optimizing for size before runtime check generation.
In that case, we do not use the result of the expansion, the expanded
instruction will be dead and cleaned up later.
By doing the check before expanding the runtime-checks, we can save a
bit of unnecessary work.
EarlyCSE has a mode to verify the invariant that hash equality equals
key equality, but EliminateDuplicatePHINodes() doesn't.
I've verified that this would have caught the stage2-stage3 mismatches
5ec2b757cc revert has fixed,
that were introduced last time in 3e69871ab5.
This patch fixes AANoUndef manifestation.
We should not manifest noundef for positions that will be changed to undef.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86835
Handling the new min/max intrinsics is the motivation, but it
turns out that we have a bunch of other intrinsics with this
missing bit of analysis too.
The FP min/max tests show that we are intersecting FMF,
so that part should be safe too.
As noted in https://llvm.org/PR46897 , there is a commutative
property specifier for intrinsics, but no corresponding function
attribute, and so apparently no uses of that bit. We may want to
remove that next.
Follow-up patches should wire up the Instruction::isCommutative()
to this IntrinsicInst specialization. That requires updating
callers to be aware of the more general commutative property
(not just binops).
Differential Revision: https://reviews.llvm.org/D86798
The original take 1 was 6102310d81,
which taught InstSimplify to do that, which seemed better at time,
since we got EarlyCSE support for free.
However, it was proven that we can not do that there,
the simplified-to PHI would not be reachable from the original PHI,
and that is not something InstSimplify is allowed to do,
as noted in the commit ed90f15efb
that reverted it:
> It appears to cause compilation non-determinism and caused stage3 mismatches.
Then there was take 2 3e69871ab5,
which was InstCombine-specific, but it again showed stage2-stage3 differences,
and reverted in bdaa3f86a0.
This is quite alarming.
Here, let's try to change how we find existing PHI candidate:
due to the worklist order, and the way PHI nodes are inserted
(it may be inserted as the first one, or maybe not), let's look at *all*
PHI nodes in the block.
Effects on vanilla llvm test-suite + RawSpeed:
```
| statistic name | baseline | proposed | Δ | % | \|%\| |
|----------------------------------------------------|-----------|-----------|-------:|---------:|---------:|
| asm-printer.EmittedInsts | 7942329 | 7942457 | 128 | 0.00% | 0.00% |
| assembler.ObjectBytes | 254295632 | 254312480 | 16848 | 0.01% | 0.01% |
| correlated-value-propagation.NumPhis | 18412 | 18347 | -65 | -0.35% | 0.35% |
| early-cse.NumCSE | 2183283 | 2183267 | -16 | 0.00% | 0.00% |
| early-cse.NumSimplify | 550105 | 541842 | -8263 | -1.50% | 1.50% |
| instcombine.NumAggregateReconstructionsSimplified | 73 | 4506 | 4433 | 6072.60% | 6072.60% |
| instcombine.NumCombined | 3640311 | 3644419 | 4108 | 0.11% | 0.11% |
| instcombine.NumDeadInst | 1778204 | 1783205 | 5001 | 0.28% | 0.28% |
| instcombine.NumPHICSEs | 0 | 22490 | 22490 | 0.00% | 0.00% |
| instcombine.NumWorklistIterations | 2023272 | 2024400 | 1128 | 0.06% | 0.06% |
| instcount.NumCallInst | 1758395 | 1758802 | 407 | 0.02% | 0.02% |
| instcount.NumInvokeInst | 59478 | 59502 | 24 | 0.04% | 0.04% |
| instcount.NumPHIInst | 330557 | 330545 | -12 | 0.00% | 0.00% |
| instcount.TotalBlocks | 1077138 | 1077220 | 82 | 0.01% | 0.01% |
| instcount.TotalFuncs | 101442 | 101441 | -1 | 0.00% | 0.00% |
| instcount.TotalInsts | 8831946 | 8832606 | 660 | 0.01% | 0.01% |
| simplifycfg.NumHoistCommonCode | 24186 | 24187 | 1 | 0.00% | 0.00% |
| simplifycfg.NumInvokes | 4300 | 4410 | 110 | 2.56% | 2.56% |
| simplifycfg.NumSimpl | 1019813 | 999767 | -20046 | -1.97% | 1.97% |
```
So it fires 22490 times, which is less than ~24k the take 1 did,
but more than what take 2 did (22228 times)
.
It allows foldAggregateConstructionIntoAggregateReuse() to actually work
after PHI-of-extractvalue folds did their thing. Previously SimplifyCFG
would have done this PHI CSE, of all places. Additionally, allows some
more `invoke`->`call` folds to happen (+110, +2.56%).
All in all, expectedly, this catches less things overall,
but all the motivational cases are still caught, so all good.
While the original variant with doing this in InstSimplify (rightfully)
caused questions and ultimately was detected to be a culprit
of stage2-stage3 mismatch, it was expected that
InstCombine-based implementation would be fine.
But apparently it's not, as
http://lab.llvm.org:8011/builders/clang-with-thin-lto-ubuntu/builds/24095/steps/compare-compilers/logs/stdio
suggests.
Which suggests that somewhere in InstCombine there is a loop
over nondeterministically sorted container, which causes
different worklist ordering.
This reverts commit 3e69871ab5.
As we've established, if it takes more than two iterations
(one to perform folding and one to ensure that no folding opportunities
remain) per function, then there are worklist management issues.
So it may be interesting to keep track of it.
The original take was 6102310d81,
which taught InstSimplify to do that, which seemed better at time,
since we got EarlyCSE support for free.
However, it was proven that we can not do that there,
the simplified-to PHI would not be reachable from the original PHI,
and that is not something InstSimplify is allowed to do,
as noted in the commit ed90f15efb
that reverted it :
> It appears to cause compilation non-determinism and caused stage3 mismatches.
However InstCombine already does many different optimizations,
so it should be a safe place to do it here.
Note that we still can't just compare incoming values ranges,
because there is no guarantee that these PHI's we'd simplify to
were already re-visited and sorted.
However coming up with a test is problematic.
Effects on vanilla llvm test-suite + RawSpeed:
```
| statistic name | baseline | proposed | Δ | % | |%| |
|----------------------------------------------------|-----------|-----------|-------:|---------:|---------:|
| instcombine.NumPHICSEs | 0 | 22228 | 22228 | 0.00% | 0.00% |
| asm-printer.EmittedInsts | 7942329 | 7942456 | 127 | 0.00% | 0.00% |
| assembler.ObjectBytes | 254295632 | 254313792 | 18160 | 0.01% | 0.01% |
| early-cse.NumCSE | 2183283 | 2183272 | -11 | 0.00% | 0.00% |
| early-cse.NumSimplify | 550105 | 541842 | -8263 | -1.50% | 1.50% |
| instcombine.NumAggregateReconstructionsSimplified | 73 | 4506 | 4433 | 6072.60% | 6072.60% |
| instcombine.NumCombined | 3640311 | 3666911 | 26600 | 0.73% | 0.73% |
| instcombine.NumDeadInst | 1778204 | 1783318 | 5114 | 0.29% | 0.29% |
| instcount.NumCallInst | 1758395 | 1758804 | 409 | 0.02% | 0.02% |
| instcount.NumInvokeInst | 59478 | 59502 | 24 | 0.04% | 0.04% |
| instcount.NumPHIInst | 330557 | 330549 | -8 | 0.00% | 0.00% |
| instcount.TotalBlocks | 1077138 | 1077221 | 83 | 0.01% | 0.01% |
| instcount.TotalFuncs | 101442 | 101441 | -1 | 0.00% | 0.00% |
| instcount.TotalInsts | 8831946 | 8832611 | 665 | 0.01% | 0.01% |
| simplifycfg.NumInvokes | 4300 | 4410 | 110 | 2.56% | 2.56% |
| simplifycfg.NumSimpl | 1019813 | 999740 | -20073 | -1.97% | 1.97% |
```
So it fires ~22k times, which is less than ~24k the take 1 did.
It allows foldAggregateConstructionIntoAggregateReuse() to actually work
after PHI-of-extractvalue folds did their thing. Previously SimplifyCFG
would have done this PHI CSE, of all places. Additionally, allows some
more `invoke`->`call` folds to happen (+110, +2.56%).
All in all, expectedly, this catches less things overall,
but all the motivational cases are still caught, so all good.
instruction can decrement the reference count, not whether it can alter
it
This prevents the state transition from S_Use to S_CanRelease when doing
a bottom-up traversal and the transition from S_Retain to S_CanRelease
when doing a top-down traversal when the visited instruction can
increment the ref count but cannot decrement it. This allows the ARC
optimizer to remove retain/release pairs which were previously not
removed.
rdar://problem/21793154
There's a special case in hasAttribute for None when pImpl is null. If pImpl is not null we dispatch to pImpl->hasAttribute which will always return false for Attribute::None.
So if we just want to check for None its sufficient to just check that pImpl is null. Which can even be done inline.
This patch adds a helper for that case which I hope will speed up our getSubtargetImpl implementations.
Differential Revision: https://reviews.llvm.org/D86744
Since doInitialization() in the legacy pass modifies the module, the NPM
pass is a Module pass.
Reviewed By: ahatanak, ychen
Differential Revision: https://reviews.llvm.org/D86178
This patch fixes this crash https://gcc.godbolt.org/z/Ps8d1e
And gives SROA the ability to remove assumes if it allows promoting an alloca to register
Without removing assumes when it can't promote to register.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86570
This patch changes ElementCount so that the Min and Scalable
members are now private and can only be accessed via the get
functions getKnownMinValue() and isScalable(). In addition I've
added some other member functions for more commonly used operations.
Hopefully this makes the class more useful and will reduce the
need for calling getKnownMinValue().
Differential Revision: https://reviews.llvm.org/D86065
This changes getDomMemoryDef to check if a Current is a valid
candidate for elimination before checking for reads. Before the change,
we were spending a lot of compile-time in checking for read accesses for
Current that might not even be removable.
This patch flips the logic, so we skip Current if they cannot be
removed before checking all their uses. This is much more efficient in
practice.
It also adds a more aggressive limit for checking partially overlapping
stores. The main problem with overlapping stores is that we do not know
if they will lead to elimination until seeing all of them. This patch
limits adds a new limit for overlapping store candidates, which keeps
the number of modified overlapping stores roughly the same.
This is another substantial compile-time improvement (while also
increasing the number of stores eliminated). Geomean -O3 -0.67%,
ReleaseThinLTO -0.97%.
http://llvm-compile-time-tracker.com/compare.php?from=0a929b6978a068af8ddb02d0d4714a2843dd8ba9&to=2e630629b43f64b60b282e90f0d96082fde2dacc&stat=instructions
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86487
strspn, strncmp, strcspn, strcasecmp, strncasecmp, memcmp, memchr,
memrchr, memcpy, memmove, memcpy, mempcpy, strchr, strrchr, bcmp
should all only access memory through their arguments.
I broke out strcoll, strcasecmp, strncasecmp because the result
depends on the locale, which might get accessed through memory.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86724
Even if noundef is deduced for a position, we should not manifest it when the position is dead.
This is because the associated values with dead positions are replaced with undef values by AAIsDead.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86565
If we query an AA with `Attributor::getAAFor` in `AbstractAttribute::manifest`, the AA may be updated.
This patch makes use of the phase flag in Attributor, and handle `getAAFor` behavior according to the flag.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86635
Add a new flag that indicates which stage in the process we are in.
This flag is introduced for handling behavior of `getAAFor` according to the stage. (discussed in D86635)
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86678
See RFC for background:
http://lists.llvm.org/pipermail/llvm-dev/2020-June/142744.html
Note that the runtime changes will be sent separately (hopefully this
week, need to add some tests).
This patch includes the LLVM pass to instrument memory accesses with
either inline sequences to increment the access count in the shadow
location, or alternatively to call into the runtime. It also changes
calls to memset/memcpy/memmove to the equivalent runtime version.
The pass is modeled on the address sanitizer pass.
The clang changes add the driver option to invoke the new pass, and to
link with the upcoming heap profiling runtime libraries.
Currently there is no attempt to optimize the instrumentation, e.g. to
aggregate updates to the same memory allocation. That will be
implemented as follow on work.
Differential Revision: https://reviews.llvm.org/D85948
Currently we bail out early for strlen calls with a GEP operand, if none
of the GEP specific optimizations fire. But there could be later
optimizations that still apply, which we currently miss out on.
An example is that we do not apply the following optimization
strlen(x) == 0 --> *x == 0
Unless I am missing something, there seems to be no reason for bailing
out early there.
Fixes PR47149.
Reviewed By: lebedev.ri, xbolva00
Differential Revision: https://reviews.llvm.org/D85886
For DSE with MemorySSA it is beneficial to manually traverse the
defining access, instead of using a MemorySSA walker, so we can
better control the number of steps together with other limits and
also weed out invalid/unprofitable paths early on.
This patch requires a follow-up patch to be most effective, which I will
share soon after putting this patch up.
This temporarily XFAIL's the limit tests, because we now explore more
MemoryDefs that may not alias/clobber the killing def. This will be
improved/fixed by the follow-up patch.
This patch also renames some `Dom*` variables to `Earlier*`, because the
dominance relation is not really used/important here and potentially
confusing.
This patch allows us to aggressively cut down compile time, geomean
-O3 -0.64%, ReleaseThinLTO -1.65%, at the expense of fewer stores
removed. Subsequent patches will increase the number of removed stores
again, while keeping compile-time in check.
http://llvm-compile-time-tracker.com/compare.php?from=d8e3294118a8c5f3f97688a704d5a05b67646012&to=0a929b6978a068af8ddb02d0d4714a2843dd8ba9&stat=instructions
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86486
Currently, an undef value is reduced to 0 when it is added to a set of potential values.
This patch introduces a flag for under values. By this, for example, we can merge two states `{undef}`, `{1}` to `{1}` (because we can reduce the undef to 1).
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D85592
As discussed in
http://lists.llvm.org/pipermail/llvm-dev/2020-July/143801.html.
Currently no users outside of unit tests.
Replace all instances in tests of -constprop with -instsimplify.
Notable changes in tests:
* vscale.ll - @llvm.sadd.sat.nxv16i8 is evaluated by instsimplify, use a fake intrinsic instead
* InsertElement.ll - insertelement undef is removed by instsimplify in @insertelement_undef
llvm/test/Transforms/ConstProp moved to llvm/test/Transforms/InstSimplify/ConstProp
Reviewed By: lattner, nikic
Differential Revision: https://reviews.llvm.org/D85159
and indirect call promotion candidate.
Profile remapping is a feature to match a function in the module with its
profile in sample profile if the function name and the name in profile look
different but are equivalent using given remapping rules. This is a useful
feature to keep the performance stable by specifying some remapping rules
when sampleFDO targets are going through some large scale function signature
change.
However, currently profile remapping support is only valid for outline
function profile in SampleFDO. It cannot match a callee with an inline
instance profile if they have different but equivalent names. We found
that without the support for inline instance profile, remapping is less
effective for some large scale change.
To add that support, before any remapping lookup happens, all the names
in the profile will be inserted into remapper and the Key to the name
mapping will be recorded in a map called NameMap in the remapper. During
name lookup, a Key will be returned for the given name and it will be used
to extract an equivalent name in the profile from NameMap. So with the help
of the NameMap, we can translate any given name to an equivalent name in
the profile if it exists. Whenever we try to match a name in the module to
a name in the profile, we will try the match with the original name first,
and if it doesn't match, we will use the equivalent name got from remapper
to try the match for another time. In this way, the patch can enhance the
profile remapping support for searching inline instance and searching
indirect call promotion candidate.
In a planned large scale change of int64 type (long long) to int64_t (long),
we found the performance of a google internal benchmark degraded by 2% if
nothing was done. If existing profile remapping was enabled, the performance
degradation dropped to 1.2%. If the profile remapping with the current patch
was enabled, the performance degradation further dropped to 0.14% (Note the
experiment was done before searching indirect call promotion candidate was
added. We hope with the remapping support of searching indirect call promotion
candidate, the degradation can drop to 0% in the end. It will be evaluated
post commit).
Differential Revision: https://reviews.llvm.org/D86332
As FIXME said, they really should be checking for a single user,
not use, so let's do that. It is not *that* unusual to have
the same value as incoming value in a PHI node, not unlike
how a PHI may have the same incoming basic block more than once.
There isn't a nice way to do that, Value::users() isn't uniqified,
and Value only tracks it's uses, not Users, so the check is
potentially costly since it does indeed potentially involes
traversing the entire use list of a value.
This implements 2 different vectorisation fallback strategies if tail-folding
fails: 1) don't vectorise at all, or 2) vectorise using a scalar epilogue. This
can be controlled with option -prefer-predicate-over-epilogue, that has been
changed to take a numeric value corresponding to the tail-folding preference
and preferred fallback.
Patch by: Pierre van Houtryve, Sjoerd Meijer.
Differential Revision: https://reviews.llvm.org/D79783
This patch produces an edge-based interface in AAIsDead.
By this, we can query a set of basic blocks that are directly reachable from a given basic block.
This is specifically useful for implementation of AAReachability.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D85547
While since D86306 we do it's sibling fold for `insertvalue`,
we should also do this for `extractvalue`'s.
And unlike that one, the results here are, quite honestly, shocking,
as it can be observed here on vanilla llvm test-suite + RawSpeed results:
```
| statistic name | baseline | proposed | Δ | % | |%| |
|----------------------------------------------------|-----------|-----------|--------:|--------:|-------:|
| asm-printer.EmittedInsts | 7945095 | 7942507 | -2588 | -0.03% | 0.03% |
| assembler.ObjectBytes | 273209920 | 273069800 | -140120 | -0.05% | 0.05% |
| early-cse.NumCSE | 2183363 | 2183398 | 35 | 0.00% | 0.00% |
| early-cse.NumSimplify | 541847 | 550017 | 8170 | 1.51% | 1.51% |
| instcombine.NumAggregateReconstructionsSimplified | 2139 | 108 | -2031 | -94.95% | 94.95% |
| instcombine.NumCombined | 3601364 | 3635448 | 34084 | 0.95% | 0.95% |
| instcombine.NumConstProp | 27153 | 27157 | 4 | 0.01% | 0.01% |
| instcombine.NumDeadInst | 1694521 | 1765022 | 70501 | 4.16% | 4.16% |
| instcombine.NumPHIsOfExtractValues | 0 | 37546 | 37546 | 0.00% | 0.00% |
| instcombine.NumSunkInst | 63158 | 63686 | 528 | 0.84% | 0.84% |
| instcount.NumBrInst | 874304 | 871857 | -2447 | -0.28% | 0.28% |
| instcount.NumCallInst | 1757657 | 1758402 | 745 | 0.04% | 0.04% |
| instcount.NumExtractValueInst | 45623 | 11483 | -34140 | -74.83% | 74.83% |
| instcount.NumInsertValueInst | 4983 | 580 | -4403 | -88.36% | 88.36% |
| instcount.NumInvokeInst | 61018 | 59478 | -1540 | -2.52% | 2.52% |
| instcount.NumLandingPadInst | 35334 | 34215 | -1119 | -3.17% | 3.17% |
| instcount.NumPHIInst | 344428 | 331116 | -13312 | -3.86% | 3.86% |
| instcount.NumRetInst | 100773 | 100772 | -1 | 0.00% | 0.00% |
| instcount.TotalBlocks | 1081154 | 1077166 | -3988 | -0.37% | 0.37% |
| instcount.TotalFuncs | 101443 | 101442 | -1 | 0.00% | 0.00% |
| instcount.TotalInsts | 8890201 | 8833747 | -56454 | -0.64% | 0.64% |
| instsimplify.NumSimplified | 75822 | 75707 | -115 | -0.15% | 0.15% |
| simplifycfg.NumHoistCommonCode | 24203 | 24197 | -6 | -0.02% | 0.02% |
| simplifycfg.NumHoistCommonInstrs | 48201 | 48195 | -6 | -0.01% | 0.01% |
| simplifycfg.NumInvokes | 2785 | 4298 | 1513 | 54.33% | 54.33% |
| simplifycfg.NumSimpl | 997332 | 1018189 | 20857 | 2.09% | 2.09% |
| simplifycfg.NumSinkCommonCode | 7088 | 6464 | -624 | -8.80% | 8.80% |
| simplifycfg.NumSinkCommonInstrs | 15117 | 14021 | -1096 | -7.25% | 7.25% |
```
... which tells us that this new fold fires whopping 38k times,
increasing the amount of SimplifyCFG's `invoke`->`call` transforms by +54% (+1513) (again, D85787 did that last time),
decreasing total instruction count by -0.64% (-56454),
and sharply decreasing count of `insertvalue`'s (-88.36%, i.e. 9 times less)
and `extractvalue`'s (-74.83%, i.e. four times less).
This causes geomean -0.01% binary size decrease
http://llvm-compile-time-tracker.com/compare.php?from=4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=97dacca0111cb2ae678204e52a3cee00e3a69208&stat=size-text
and, ignoring `O0-g`, is a geomean -0.01%..-0.05% compile-time improvement
http://llvm-compile-time-tracker.com/compare.php?from=4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=97dacca0111cb2ae678204e52a3cee00e3a69208&stat=instructions
The other thing that tells is, is that while this is a massive win for `invoke`->`call` transform
`InstCombinerImpl::foldAggregateConstructionIntoAggregateReuse()` fold,
which is supposed to be dealing with such aggregate reconstructions,
fires a lot less now. There are two reasons why:
1. After this fold, as it can be seen in tests, we may (will) end up with trivially redundant PHI nodes.
We don't CSE them in InstCombine presently, which means that EarlyCSE needs to run and then InstCombine rerun.
2. But then, EarlyCSE not only manages to fold such redundant PHI's,
it also sees that the extract-insert chain recreates the original aggregate,
and replaces it with the original aggregate.
The take-aways are
1. We maybe should do most trivial, same-BB PHI CSE in InstCombine
2. I need to check if what other patterns remain, and how they can be resolved.
(i.e. i wonder if `foldAggregateConstructionIntoAggregateReuse()` might go away)
This is a reland of the original commit fcb51d8c24,
because originally i forgot to ensure that the base aggregate types match.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D86530
This reverts commit fcb51d8c24.
As buildbots report, there's apparently some missing check to ensure
that the types of incoming values match the type of PHI.
Let's revert for a moment.
While since D86306 we do it's sibling fold for `insertvalue`,
we should also do this for `extractvalue`'s.
And unlike that one, the results here are, quite honestly, shocking,
as it can be observed here on vanilla llvm test-suite + RawSpeed results:
```
| statistic name | baseline | proposed | Δ | % | |%| |
|----------------------------------------------------|-----------|-----------|--------:|--------:|-------:|
| asm-printer.EmittedInsts | 7945095 | 7942507 | -2588 | -0.03% | 0.03% |
| assembler.ObjectBytes | 273209920 | 273069800 | -140120 | -0.05% | 0.05% |
| early-cse.NumCSE | 2183363 | 2183398 | 35 | 0.00% | 0.00% |
| early-cse.NumSimplify | 541847 | 550017 | 8170 | 1.51% | 1.51% |
| instcombine.NumAggregateReconstructionsSimplified | 2139 | 108 | -2031 | -94.95% | 94.95% |
| instcombine.NumCombined | 3601364 | 3635448 | 34084 | 0.95% | 0.95% |
| instcombine.NumConstProp | 27153 | 27157 | 4 | 0.01% | 0.01% |
| instcombine.NumDeadInst | 1694521 | 1765022 | 70501 | 4.16% | 4.16% |
| instcombine.NumPHIsOfExtractValues | 0 | 37546 | 37546 | 0.00% | 0.00% |
| instcombine.NumSunkInst | 63158 | 63686 | 528 | 0.84% | 0.84% |
| instcount.NumBrInst | 874304 | 871857 | -2447 | -0.28% | 0.28% |
| instcount.NumCallInst | 1757657 | 1758402 | 745 | 0.04% | 0.04% |
| instcount.NumExtractValueInst | 45623 | 11483 | -34140 | -74.83% | 74.83% |
| instcount.NumInsertValueInst | 4983 | 580 | -4403 | -88.36% | 88.36% |
| instcount.NumInvokeInst | 61018 | 59478 | -1540 | -2.52% | 2.52% |
| instcount.NumLandingPadInst | 35334 | 34215 | -1119 | -3.17% | 3.17% |
| instcount.NumPHIInst | 344428 | 331116 | -13312 | -3.86% | 3.86% |
| instcount.NumRetInst | 100773 | 100772 | -1 | 0.00% | 0.00% |
| instcount.TotalBlocks | 1081154 | 1077166 | -3988 | -0.37% | 0.37% |
| instcount.TotalFuncs | 101443 | 101442 | -1 | 0.00% | 0.00% |
| instcount.TotalInsts | 8890201 | 8833747 | -56454 | -0.64% | 0.64% |
| instsimplify.NumSimplified | 75822 | 75707 | -115 | -0.15% | 0.15% |
| simplifycfg.NumHoistCommonCode | 24203 | 24197 | -6 | -0.02% | 0.02% |
| simplifycfg.NumHoistCommonInstrs | 48201 | 48195 | -6 | -0.01% | 0.01% |
| simplifycfg.NumInvokes | 2785 | 4298 | 1513 | 54.33% | 54.33% |
| simplifycfg.NumSimpl | 997332 | 1018189 | 20857 | 2.09% | 2.09% |
| simplifycfg.NumSinkCommonCode | 7088 | 6464 | -624 | -8.80% | 8.80% |
| simplifycfg.NumSinkCommonInstrs | 15117 | 14021 | -1096 | -7.25% | 7.25% |
```
... which tells us that this new fold fires whopping 38k times,
increasing the amount of SimplifyCFG's `invoke`->`call` transforms by +54% (+1513) (again, D85787 did that last time),
decreasing total instruction count by -0.64% (-56454),
and sharply decreasing count of `insertvalue`'s (-88.36%, i.e. 9 times less)
and `extractvalue`'s (-74.83%, i.e. four times less).
This causes geomean -0.01% binary size decrease
http://llvm-compile-time-tracker.com/compare.php?from=4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=97dacca0111cb2ae678204e52a3cee00e3a69208&stat=size-text
and, ignoring `O0-g`, is a geomean -0.01%..-0.05% compile-time improvement
http://llvm-compile-time-tracker.com/compare.php?from=4d5ca22b8adfb6643466e4e9f48ba14bb48938bc&to=97dacca0111cb2ae678204e52a3cee00e3a69208&stat=instructions
The other thing that tells is, is that while this is a massive win for `invoke`->`call` transform
`InstCombinerImpl::foldAggregateConstructionIntoAggregateReuse()` fold,
which is supposed to be dealing with such aggregate reconstructions,
fires a lot less now. There are two reasons why:
1. After this fold, as it can be seen in tests, we may (will) end up with trivially redundant PHI nodes.
We don't CSE them in InstCombine presently, which means that EarlyCSE needs to run and then InstCombine rerun.
2. But then, EarlyCSE not only manages to fold such redundant PHI's,
it also sees that the extract-insert chain recreates the original aggregate,
and replaces it with the original aggregate.
The take-aways are
1. We maybe should do most trivial, same-BB PHI CSE in InstCombine
2. I need to check if what other patterns remain, and how they can be resolved.
(i.e. i wonder if `foldAggregateConstructionIntoAggregateReuse()` might go away)
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D86530
This patch helps getGuaranteedNonPoisonOp find multiple non-poison operands.
Instead of special-casing llvm.assume, I think it is also a viable option to
add noundef to Intrinsics.td. If it makes sense, I'll make a patch for that.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86477
The 1st attempt (rG557b890) was reverted because it caused miscompiles.
That bug is avoided here by changing the order of folds and as verified
in the new tests.
Original commit message:
InstCombine currently has odd rules for folding insert-extract chains to shuffles,
so we miss collapsing seemingly simple cases as shown in the tests here.
But poison makes this not quite as easy as we might have guessed. Alive2 tests to
show the subtle difference (similar to the regression tests):
https://alive2.llvm.org/ce/z/hp4hv3 (this is ok)
https://alive2.llvm.org/ce/z/ehEWaN (poison leakage)
SLP tends to create these patterns (as shown in the SLP tests), and this could
help with solving PR16739.
Differential Revision: https://reviews.llvm.org/D86460
This adapts LV to the new semantics of get.active.lane.mask as discussed in
D86147, which means that the LV now emits intrinsic get.active.lane.mask with
the loop tripcount instead of the backedge-taken count as its second argument.
The motivation for this is described in D86147.
Differential Revision: https://reviews.llvm.org/D86304
In getCastInstrCost when the instruction is a truncate we were relying
upon the implicit TypeSize -> uint64_t cast when asking if a given type
has the same size as a legal integer. I've changed the code to only
ask the question if the type is fixed length.
I have also changed InstCombinerImpl::SimplifyDemandedUseBits to bail
out for now if the type is a scalable vector.
I've added the following new tests:
Analysis/CostModel/AArch64/sve-trunc.ll
Transforms/InstCombine/AArch64/sve-trunc.ll
for both of these fixes.
Differential revision: https://reviews.llvm.org/D86432
Currently we repeatedly check the same uses for read clobbers in some
cases. We can avoid unnecessary checks by keeping track of the memory
accesses we already found read clobbers for. To do so, we just add
memory access causing read-clobbers to a set. Note that marking all
visited accesses as read-clobbers would be to pessimistic, as that might
include accesses not on any path to the actual read clobber.
If we do not find any read-clobbers, we can add all visited instructions
to another set and use that to skip the same accesses in the next call.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D75025
InstCombine currently has odd rules for folding insert-extract chains to shuffles,
so we miss collapsing seemingly simple cases as shown in the tests here.
But poison makes this not quite as easy as we might have guessed. Alive2 tests to
show the subtle difference (similar to the regression tests):
https://alive2.llvm.org/ce/z/hp4hv3 (this is ok)
https://alive2.llvm.org/ce/z/ehEWaN (poison leakage)
SLP tends to create these patterns (as shown in the SLP tests), and this could
help with solving PR16739.
Differential Revision: https://reviews.llvm.org/D86460
The "takeName" logic at the end of ScalarizerVisitor::finish
could end up renaming global variables when having simplified
and extractelement instruction to simply pick a single vector
element. If the input vector to the extractelement instruction
held pointers to global variables we ended up renaming the global
variable.
The patch make sure we only take the name of the replaced Op when
we have added new instructions that might need a useful name.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D86472
This reverts commit 2e43acfed8.
LLVMCoroutines (the library which contains Coroutines.h) depends on LLVMipo (the
library which contains SampleProfile.cpp). It is inappropriate for
SampleProfile.cpp to depent on Coroutines.h (circular dependency).
The test inverted dependencies as well:
llvm/test/Transforms/Coroutines/coro-inline.ll uses -sample-profile.
Using callCapturesBefore potentially improves the precision and the
number of stores we can remove. But in practice, it seems to have very
little impact in terms of stores removed. For example, for
SPEC2000/SPEC2006/MultiSource with -O3 -flto, ~50 more stores are
removed (out of ~26900 stores removed). But in terms of compile-time, it
is very expensive and the patch gives substantial compile-time
improvements: Geomean O3 -0.24%, ReleaseThinLTO -0.47%, ReleaseLTO-g
-0.39%.
http://llvm-compile-time-tracker.com/compare.php?from=612a0bff88ed906c83b82f079d4c49e5fecfb9d0&to=e6c86b96d20d97dd88e903a409bd8d39b6114312&stat=instructions
summary:
When callee coroutine function is inlined into caller coroutine
function before coro-split pass, llvm will emits "coroutine should
have exactly one defining @llvm.coro.begin". It seems that coro-early
pass can not handle this quiet well.
So we believe that unsplited coroutine function should not be inlined.
This patch fix such issue by not inlining function if it has attribute
"coroutine.presplit" (it means the function has not been splited) to
fix this issue
TestPlan: check-llvm
Reviewed By: wenlei
Differential Revision: https://reviews.llvm.org/D85812
Changes:
* Change `ToVectorTy` to deal directly with `ElementCount` instances.
* `VF == 1` replaced with `VF.isScalar()`.
* `VF > 1` and `VF >=2` replaced with `VF.isVector()`.
* `VF <=1` is replaced with `VF.isZero() || VF.isScalar()`.
* Replaced the uses of `llvm::SmallSet<ElementCount, ...>` with
`llvm::SmallSetVector<ElementCount, ...>`. This avoids the need of an
ordering function for the `ElementCount` class.
* Bits and pieces around printing the `ElementCount` to string streams.
To guarantee that this change is a NFC, `VF.Min` and asserts are used
in the following places:
1. When it doesn't make sense to deal with the scalable property, for
example:
a. When computing unrolling factors.
b. When shuffle masks are built for fixed width vector types
In this cases, an
assert(!VF.Scalable && "<mgs>") has been added to make sure we don't
enter coepaths that don't make sense for scalable vectors.
2. When there is a conscious decision to use `FixedVectorType`. These
uses of `FixedVectorType` will likely be removed in favour of
`VectorType` once the vectorizer is generic enough to deal with both
fixed vector types and scalable vector types.
3. When dealing with building constants out of the value of VF, for
example when computing the vectorization `step`, or building vectors
of indices. These operation _make sense_ for scalable vectors too,
but changing the code in these places to be generic and make it work
for scalable vectors is to be submitted in a separate patch, as it is
a functional change.
4. When building the potential VFs in VPlan. Making the VPlan generic
enough to handle scalable vectorization factors is a functional change
that needs a separate patch. See for example `void
LoopVectorizationPlanner::buildVPlans(unsigned MinVF, unsigned
MaxVF)`.
5. The class `IntrinsicCostAttribute`: this class still uses `unsigned
VF` as updating the field to use `ElementCount` woudl require changes
that could result in changing the behavior of the compiler. Will be done
in a separate patch.
7. When dealing with user input for forcing the vectorization
factor. In this case, adding support for scalable vectorization is a
functional change that migh require changes at command line.
Note that in some places the idiom
```
unsigned VF = ...
auto VTy = FixedVectorType::get(ScalarTy, VF)
```
has been replaced with
```
ElementCount VF = ...
assert(!VF.Scalable && ...);
auto VTy = VectorType::get(ScalarTy, VF)
```
The assertion guarantees that the new code is (at least in debug mode)
functionally equivalent to the old version. Notice that this change had been
possible because none of the methods that are specific to `FixedVectorType`
were used after the instantiation of `VTy`.
Reviewed By: rengolin, ctetreau
Differential Revision: https://reviews.llvm.org/D85794
Changes:
* Change `ToVectorTy` to deal directly with `ElementCount` instances.
* `VF == 1` replaced with `VF.isScalar()`.
* `VF > 1` and `VF >=2` replaced with `VF.isVector()`.
* `VF <=1` is replaced with `VF.isZero() || VF.isScalar()`.
* Add `<` operator to `ElementCount` to be able to use
`llvm::SmallSetVector<ElementCount, ...>`.
* Bits and pieces around printing the ElementCount to string streams.
* Added a static method to `ElementCount` to represent a scalar.
To guarantee that this change is a NFC, `VF.Min` and asserts are used
in the following places:
1. When it doesn't make sense to deal with the scalable property, for
example:
a. When computing unrolling factors.
b. When shuffle masks are built for fixed width vector types
In this cases, an
assert(!VF.Scalable && "<mgs>") has been added to make sure we don't
enter coepaths that don't make sense for scalable vectors.
2. When there is a conscious decision to use `FixedVectorType`. These
uses of `FixedVectorType` will likely be removed in favour of
`VectorType` once the vectorizer is generic enough to deal with both
fixed vector types and scalable vector types.
3. When dealing with building constants out of the value of VF, for
example when computing the vectorization `step`, or building vectors
of indices. These operation _make sense_ for scalable vectors too,
but changing the code in these places to be generic and make it work
for scalable vectors is to be submitted in a separate patch, as it is
a functional change.
4. When building the potential VFs in VPlan. Making the VPlan generic
enough to handle scalable vectorization factors is a functional change
that needs a separate patch. See for example `void
LoopVectorizationPlanner::buildVPlans(unsigned MinVF, unsigned
MaxVF)`.
5. The class `IntrinsicCostAttribute`: this class still uses `unsigned
VF` as updating the field to use `ElementCount` woudl require changes
that could result in changing the behavior of the compiler. Will be done
in a separate patch.
7. When dealing with user input for forcing the vectorization
factor. In this case, adding support for scalable vectorization is a
functional change that migh require changes at command line.
Differential Revision: https://reviews.llvm.org/D85794
Avoid computing InvisibleToCallerBefore/AfterRet up front. In most
cases, this information is not really needed. Instead, introduce helper
functions to compute and cache the result on demand.
Notably, this also does not use PointerMayBeCapturedBefore for
isInvisibleToCallerBeforeRet, as it requires the killing MemoryDef as
starting instruction, making the caching ineffective. But it appears the
use of PointerMayBeCapturedBefore has very limited benefits in practice
(e.g. on SPEC2000/SPEC2006/MultiSource there are no binary changes with
-O3 -flto). Refrain from using it for now, to limit-compile-time.
This gives some nice compile-time improvements:
http://llvm-compile-time-tracker.com/compare.php?from=db9345f6810f379a36752dc52caf5230585d0ebd&to=b4d091047e1b8a3d377d200137b79d03aca65663&stat=instructions
Limit elimination of stores at the end of a function to MemoryDefs with
a single underlying object, to save compile time.
In practice, the case with multiple underlying objects seems not very
important in practice. For -O3 -flto on MultiSource/SPEC2000/SPEC2006
this results in a total of 2 more stores being eliminated.
We can always re-visit that in the future.
Similar to the existing transform - peek through a select
to match a value and its negation.
https://alive2.llvm.org/ce/z/MXi5KG
define i8 @src(i1 %b, i8 %x) {
%0:
%neg = sub i8 0, %x
%sel = select i1 %b, i8 %x, i8 %neg
%abs = abs i8 %sel, 1
ret i8 %abs
}
=>
define i8 @tgt(i1 %b, i8 %x) {
%0:
%abs = abs i8 %x, 1
ret i8 %abs
}
Transformation seems to be correct!
This reverses the existing transform that would uniformly canonicalize any 'xor' after any shift. In the case of logical shifts, that turns a 'not' into an arbitrary 'xor' with constant, and that's probably not as good for analysis, SCEV, or codegen.
The SCEV motivating case is discussed in:
http://bugs.llvm.org/PR47136
There's an analysis motivating case at:
http://bugs.llvm.org/PR38781
I did draft a patch that would do the same for 'ashr' but that's questionable because it's just swapping the position of a 'not' and uncovers at least 2 missing folds that we would probably need to deal with as preliminary steps.
Alive proofs:
https://rise4fun.com/Alive/BBV
Name: shift right of 'not'
Pre: C2 == (-1 u>> C1)
%a = lshr i8 %x, C1
%r = xor i8 %a, C2
=>
%n = xor i8 %x, -1
%r = lshr i8 %n, C1
Name: shift left of 'not'
Pre: C2 == (-1 << C1)
%a = shl i8 %x, C1
%r = xor i8 %a, C2
=>
%n = xor i8 %x, -1
%r = shl i8 %n, C1
Name: ashr of 'not'
%a = ashr i8 %x, C1
%r = xor i8 %a, -1
=>
%n = xor i8 %x, -1
%r = ashr i8 %n, C1
Differential Revision: https://reviews.llvm.org/D86243
As disscussed in post-commit review starting with
https://reviews.llvm.org/D84108#2227365
while this appears to be mostly a win overall, especially code-size-wise,
this appears to shake //certain// code pattens in a way that is extremely
unfavorable for performance (+30% runtime regression)
on certain CPU's (i personally can't reproduce).
So until the behaviour is better understood, and a path forward is mapped,
let's back this out for now.
This reverts commit 1d51dc38d8.
- Adds a command line option to seed only selected functions.
- Makes seed allow listing exclusive to assertions enabled builds.
Reviewed By: sstefan1
Differential Revision: https://reviews.llvm.org/D86129
Currently, `AANoUndefImpl::initialize` mistakenly always indicates optimistic fixpoint for function returned position.
This is because an associated value is `Function` in the case, and `isGuaranteedNotToBeUndefOrPoison` returns true for Function.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86361
When trying to enable -debug-info-kind=constructor there was an assert
that occurs during debug info cloning ("mismatched subprogram between
llvm.dbg.value variable and !dbg attachment").
It appears that during llvm::CloneFunctionInto, a DISubprogram could be
duplicated when MapMetadata is called, and then added to the MD map again
when DIFinder gets a list of subprograms. This results in two different
versions of the DISubprogram.
This patch switches the order so that the DIFinder subprograms are
added before MapMetadata is called.
Fixes https://bugs.llvm.org/show_bug.cgi?id=46784
Differential Revision: https://reviews.llvm.org/D86185
If some of gc live value are not used in gc.relocate we can remove them
from gc-live bundle of statepoint instruction.
Also the CL removes duplicated Values in gc-live bundle.
Reviewers: reames, dantrushin
Reviewed By: dantrushin
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D85959
The only def for gc.relocate is a gc.statepoint. But real dependency is not
described by def-use chain. Instead this dependency is encoded by indecies
of operands in gc-live bundle of statepoint as integer constants in gc.relocate.
InstCombine operates by def-use chain. As a result when value in gc-live bundle
is simplified the gc.statepoint itself is not simplified but it might simplify dependent
gc.relocates. To trigger the optimization of gc.relocate we now unconditionally trigger
check of all dependent gc.relocates by adding them to worklist.
This CL handles of gc.relocates as process of gc.statepoint optimization considering
gc.statepoint and related gc.relocate as whole entity.
Reviewers: reames, dantrushin
Reviewed By: reames
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D85954
Recommit the patch after fixing an issue reported caused by the fact
that re-used values are also added to InsertedValues.
Additional tests have been added in 88818491b9
This reverts the revert commit 38884641f2.
Both AfterPass and AfterPassInvalidated pass instrumentation
callbacks get additional parameter of type PreservedAnalyses.
This patch was created by @fedor.sergeev. I have just slightly
changed it.
Reviewers: fedor.sergeev
Differential Revision: https://reviews.llvm.org/D81555
Before we speculatively execute a basic block, query the cost of
inserting the necessary select instructions against the phi folding
threshold. For non-trivial insertions, a more accurate decision can
probably be made during machine if-conversion. With minsize we query
the CodeSize cost, otherwise we use SizeAndLatency.
Differential Revision: https://reviews.llvm.org/D82438
As part of D84741, this adds a target hook for the
preferPredicatedReductionSelect option and makes use
of it under MVE, allowing us to tail predicate most
reduction loops.
Differential Revision: https://reviews.llvm.org/D85980
The normal scheme for tail folding reductions is to use:
loop:
p = phi(0, a)
mask = ...
x = masked_load(..., mask)
a = add(x, p)
s = select(mask, a, p)
This means we need to keep the register p and a alive out of the loop, plus
the mask. On a target with predicated operations we can instead generate
the phi as p = phi(0, s). This ensures the select in the loop and we can
fold select(m, add(a, b), c) to something like a vaddt c, a, b using the
m predicate. This in turn allows us to tail predicate the entire loop.
Differential Revision: https://reviews.llvm.org/D84741
Currently, although we handle `CallBase` case in updateImpl, we give up in initialize in the case.
That is problematic when we propagate a range from call site returned position to floating position.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86196
When removing a non-constant store to a global in
CleanupPointerRootUsers(), the GlobalOpt pass could incorrectly return
false.
This was caught using the check introduced by D80916.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D86149
Relanded since the buildbot issue was unrelated to this commit.
When hoisting simple values out from a loop, and an optsize attribute, a
convergent call, or an invoke instruction hindered the pass from
unswitching the loop, the pass would return an incorrect Modified
status.
This was caught using the check introduced by D80916.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86085
This reverts commit dfd447c220.
After I pushed this commit, llvm-sphinx-docs started failing, due to:
Warning, treated as error:
extension 'recommonmark' has no setup() function;
is it really a Sphinx extension module?
I don't see how this commit may have caused that, but I'm still
reverting it since I don't know how to proceed with that
troubleshooting.
When hoisting simple values out from a loop, and an optsize attribute, a
convergent call, or an invoke instruction hindered the pass from
unswitching the loop, the pass would return an incorrect Modified
status.
This was caught using the check introduced by D80916.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D86085
Comparison against null is a common pattern that usually is followed by
error handling code and the likes. We now use AANonNull to simplify
these comparisons optimistically in order to make more code dead early
on.
Reviewed By: uenoku
Differential Revision: https://reviews.llvm.org/D86145
`AADereferenceable::getAssumedDereferenceableBytes()` is actually
deducing `dereferenceable_or_null`. We should not use that information
to deduce `nonnull`, since it doesn't imply `nonnull`.
This commits breaks certain OpenMP codes (on power) because it expanded
the Attributor scope without telling the Attributor about the SCC
extend. See: https://reviews.llvm.org/D85544#2227611
This reverts commit b0b32e6490.
-force-attribute adds an attribute to function via command-line.
However, there was no counter-part to remove an attribute. This patch
adds -force-remove-attribute that removes an attribute from function.
Differential Revision: https://reviews.llvm.org/D85586
There's a potential motivating case to increase this limit in PR47191:
http://bugs.llvm.org/PR47191
But first we should make it less hacky. The limit in InstCombine is directly tied
to this value because an increase there can cause asserts in the underlying value
tracking calls if not changed together. The usage in VectorUtils is independent,
but the comment suggests that we should use the same value unless there's a known
reason to diverge. There are similar limits in codegen analysis, but I think we
should leave those independent in case we intentionally want the optimization
power/cost to be different there.
Differential Revision: https://reviews.llvm.org/D86113
D81345 appears to accidentally disables vectorization when explicitly
enabled. As PGSO isn't currently accessible from LoopAccessInfo, revert back to
the vectorization with versioning-for-unit-stride for PGSO.
Differential Revision: https://reviews.llvm.org/D85784
Kazu Hirata