Jakub Kuderski [Wed, 2 Aug 2017 18:17:52 +0000 (18:17 +0000)]
[Dominators] Teach LoopDeletion to use the new incremental API
Summary:
This patch makes LoopDeletion use the incremental DominatorTree API.
We modify LoopDeletion to perform the deletion in 5 steps:
1. Create a new dummy edge from the preheader to the exit, by adding a conditional branch.
2. Inform the DomTree about the new edge.
3. Remove the conditional branch and replace it with an unconditional edge to the exit. This removes the edge to the loop header, making it unreachable.
4. Inform the DomTree about the deleted edge.
5. Remove the unreachable block from the function.
Creating the dummy conditional branch is necessary to perform incremental DomTree update.
We should consider using the batch updater when it's ready.
Hiroshi Inoue [Wed, 2 Aug 2017 18:16:32 +0000 (18:16 +0000)]
[StackColoring] Update AliasAnalysis information in stack coloring pass (part 2)
This patch is update after the first patch (https://reviews.llvm.org/rL309651) based on the post-commit comments.
Stack coloring pass need to maintain AliasAnalysis information when merging stack slots of different types.
Actually, there is a FIXME comment in StackColoring.cpp
// FIXME: In order to enable the use of TBAA when using AA in CodeGen,
// we'll also need to update the TBAA nodes in MMOs with values
// derived from the merged allocas.
But, TBAA has been already enabled in CodeGen without fixing this pass.
The incorrect TBAA metadata results in recent failures in bootstrap test on ppc64le (PR33928) by allowing unsafe instruction scheduling.
Although we observed the problem on ppc64le, this is a platform neutral issue.
This patch makes the stack coloring pass maintains AliasAnalysis information when merging multiple stack slots.
Nirav Dave [Wed, 2 Aug 2017 16:35:58 +0000 (16:35 +0000)]
[DAG] Improve candidate pruning in store merge failure case. NFCI
During store merge we construct a sorted list of consecutive store
candidates and consider subsequences for merging into a single
store. For each subsequence we check if the stored value type is legal
the merged store would have valid and fast and if the constructed
value to be stored is valid. The only properties that affect this
check between subsequences is the size of the subsequence, the
alignment of the first store, the alignment of the stored load value
(when merging stores-of-loads), and whether the merged value is a
constant zero.
If we do not find a viable mergeable subsequence starting from the
first store of length N, we know that a subsequence starting at a
later store of length N will also fail unless the new store's
alignment, the new load's alignment (if we're merging store-of-loads),
or we've dropped stores of nonzero value and could construct a merged
stores of zero (for merging constants).
As a result if we fail to find a valid subsequence starting from the
first store we can safely skip considering subsequences that start
with subsequent stores unless one of the above properties is
true. This significantly (2x) improves compile time in some
pathological cases.
Alexey Bataev [Wed, 2 Aug 2017 13:25:26 +0000 (13:25 +0000)]
[SLP] Fix for PR31880: shuffle and vectorize repeated scalar ops on extracted elements
Summary:
Currently most of the time vectors of extractelement instructions are
treated as scalars that must be gathered into vectors. But in some
cases, like when we have extractelement instructions from single vector
with different constant indeces or from 2 vectors of the same size, we
can treat this operations as shuffle of a single vector or blending of 2
vectors.
```
define <2 x i8> @g(<2 x i8> %x, <2 x i8> %y) {
%x0 = extractelement <2 x i8> %x, i32 0
%y1 = extractelement <2 x i8> %y, i32 1
%x0x0 = mul i8 %x0, %x0
%y1y1 = mul i8 %y1, %y1
%ins1 = insertelement <2 x i8> undef, i8 %x0x0, i32 0
%ins2 = insertelement <2 x i8> %ins1, i8 %y1y1, i32 1
ret <2 x i8> %ins2
}
```
can be converted to something like
```
define <2 x i8> @g(<2 x i8> %x, <2 x i8> %y) {
%1 = shufflevector <2 x i8> %x, <2 x i8> %y, <2 x i32> <i32 0, i32 3>
%2 = mul <2 x i8> %1, %1
ret <2 x i8> %2
}
```
Currently this type of conversion is considered as high cost
transformation.
I intend to use this to test PIC support in GlobalISel for ARM.
This is difficult to test outside of that context, since the existing
MIR tests usually rely on parser support as well, and that seems a bit
trickier to add. We could try to add a unit test, but the setup for that
seems rather convoluted and overkill.
We do test however that the parser reports a nice error when
encountering a target-specific constant pool.
Daniel Sanders [Wed, 2 Aug 2017 11:03:36 +0000 (11:03 +0000)]
[globalisel][tablegen] Do not merge memoperands from instructions that weren't in the match.
Summary:
Fix a bug discovered in an out-of-tree target where memoperands from
pseudo-instructions that weren't part of the match were being merged into the
result instructions as part of GIR_MergeMemOperands.
This bug was caused by a change to the handling of State.MIs between rules when
the state machine tables were fused into a single table. Previously, each rule
would reset State.MIs using State.MIs.resize(1) but this is no longer done, as a
result stale data is occasionally left in some elements of State.MIs. Most
opcodes aren't affected by this but GIR_MergeMemOperands merges all memoperands
from the intructions recorded in State.MIs into the result instruction.
Suppose for example, we processed but rejected the following pattern:
(signextend (load x))
at this point, State.MIs contains the signextend and the load. Now suppose we
process and accept this pattern:
(add x, y)
at this point, State.MIs contains the add as well as the (now irrelevant) load.
When GIR_MergeMemOperands is processed, the memoperands from that irrelevant
load will be merged into the result instruction even though it was not part of
the match.
Bringing back the State.MIs.resize(1) would fix the problem but it would limit
our ability to optimize the table in the future. Instead, this patch fixes the
problem by explicitly stating which instructions should be merged into the result.
There's no direct test case in this commit because a test case would be very brittle.
However, at the time of writing this should fix the failures in
http://green.lab.llvm.org/green/job/Compiler_Verifiers_GlobalISEL/ as well as a
failure in test/CodeGen/ARM/GlobalISel/arm-isel.ll when expensive checks are enabled.
Craig Topper [Wed, 2 Aug 2017 06:35:15 +0000 (06:35 +0000)]
[InstCombine] Add test cases for 'or' and 'xor' to match the vector 'and' of 'sext' of 'cmp' test.
When the 'and' test was originally added it was intended to make sure we didn't change it to a sext of and of cmp. But since then the test was changed to expect it to be turned into 'select cmp1, sext cmp2, 0'. Then another optimization was added to turn the select into 'sext (and cmp1, cmp2)' which is exactly the transformation that was being blocked when the test case started.
Looks like 'or' gets optimized in a similar way, but not 'xor'.
[PM] Fix a bug where through CGSCC iteration we can get
infinite-inlining across multiple runs of the inliner by keeping a tiny
history of internal-to-SCC inlining decisions.
This is still a bit gross, but I don't yet have any fundamentally better
ideas and numerous people are blocked on this to use new PM and ThinLTO
together.
The core of the idea is to detect when we are about to do an inline that
has a chance of re-splitting an SCC which we have split before with
a similar inlining step. That is a critical component in the inlining
forming a cycle and so far detects all of the various cyclic patterns
I can come up with as well as the original real-world test case (which
comes from a ThinLTO build of libunwind).
I've added some tests that I think really demonstrate what is going on
here. They are essentially state machines that march the inliner through
various steps of a cycle and check that we stop when the cycle is closed
and that we actually did do inlining to form that cycle.
A lot of thanks go to Eric Christopher and Sanjoy Das for the help
understanding this issue and improving the test cases.
The biggest "yuck" here is the layering issue -- the CGSCC pass manager
is providing somewhat magical state to the inliner for it to use to make
itself converge. This isn't great, but I don't honestly have a lot of
better ideas yet and at least seems nicely isolated.
I have tested this patch, and it doesn't block *any* inlining on the
entire LLVM test suite and SPEC, so it seems sufficiently narrowly
targeted to the issue at hand.
We have come up with hypothetical issues that this patch doesn't cover,
but so far none of them are practical and we don't have a viable
solution yet that covers the hypothetical stuff, so proceeding here in
the interim. Definitely an area that we will be back and revisiting in
the future.
Dehao Chen [Wed, 2 Aug 2017 01:28:31 +0000 (01:28 +0000)]
Update the new PM pipeline to make ICP aware if it is SamplePGO build.
Summary: In ThinLTO backend compile, OPTOptions are not set so that the ICP in ThinLTO backend does not know if it is a SamplePGO build, in which profile count needs to be annotated directly on call instructions. This patch cleaned up the PGOOptions handling logic and passes down PGOOptions to ThinLTO backend.
Previous change "Turn s_and_saveexec_b64 into s_and_b64 if
result is unused" introduced asan use-after-poison error.
Instruction was analyzed after eraseFromParent() calls.
Matt Arsenault [Wed, 2 Aug 2017 00:43:42 +0000 (00:43 +0000)]
DAG: Undo and->or combine with FrameIndexes
This pattern shows up when lowering byval copies on AMDGPU.
The byval object access is split into 4-byte chunks, adding a
constant offset to the FixedStack base. When some of the offsets
turn into ors, this prevents combining the constant offsets.
This makes it not apparent that the object is there when matching
addressing modes, so it ends up using a scratch wave offset
relative access and the lengthy frame index expansion for that.
Matthias Braun [Wed, 2 Aug 2017 00:28:10 +0000 (00:28 +0000)]
X86: Do not use llc -march in tests.
`llc -march` is problematic because it only switches the target
architecture, but leaves the operating system unchanged. This
occasionally leads to indeterministic tests because the OS from
LLVM_DEFAULT_TARGET_TRIPLE is used.
However we can simply always use `llc -mtriple` instead. This changes
all the tests to do this to avoid people using -march when they copy and
paste parts of tests.
See also the discussion in https://reviews.llvm.org/D35287
Petr Hosek [Tue, 1 Aug 2017 23:46:33 +0000 (23:46 +0000)]
Reland "[llvm][llvm-objcopy] Added support for outputting to binary in llvm-objcopy"
This change adds the "-O binary" flag which directs llvm-objcopy to
output the object file to the same format as GNU objcopy does when given
the flag "-O binary". This was done by splitting the Object class into
two subclasses ObjectELF and ObjectBianry which each output a different
format but relay on the same code to read in the Object in Object.
[AMDGPU] Turn s_and_saveexec_b64 into s_and_b64 if result is unused
With SI_END_CF elimination for some nested control flow we can now
eliminate saved exec register completely by turning a saveexec version
of instruction into just a logical instruction.
Add a pass to remove redundant S_OR_B64 instructions enabling lanes in
the exec. If two SI_END_CF (lowered as S_OR_B64) come together without any
vector instructions between them we can only keep outer SI_END_CF, given
that CFG is structured and exec bits of the outer end statement are always
not less than exec bit of the inner one.
This needs to be done before the RA to eliminate saved exec bits registers
but after register coalescer to have no vector registers copies in between
of different end cf statements.
Matthias Braun [Tue, 1 Aug 2017 22:20:49 +0000 (22:20 +0000)]
ARM: Do not use llc -march in tests.
`llc -march` is problematic because it only switches the target
architecture, but leaves the operating system unchanged. This
occasionally leads to indeterministic tests because the OS from
LLVM_DEFAULT_TARGET_TRIPLE is used.
However we can simply always use `llc -mtriple` instead. This changes
all the tests to do this to avoid people using -march when they copy and
paste parts of tests.
See also the discussion in https://reviews.llvm.org/D35287
Matthias Braun [Tue, 1 Aug 2017 22:20:41 +0000 (22:20 +0000)]
PowerPC: Do not use llc -march in tests.
`llc -march` is problematic because it only switches the target
architecture, but leaves the operating system unchanged. This
occasionally leads to indeterministic tests because the OS from
LLVM_DEFAULT_TARGET_TRIPLE is used.
However we can simply always use `llc -mtriple` instead. This changes
all the tests to do this to avoid people using -march when they copy and
paste parts of tests.
This patch:
- Removes -march if the .ll file already has a matching `target triple`
directive or -mtriple argument.
- In all other cases changes -march=ppc32/-march=ppc64 to
-mtriple=ppc32--/-mtriple=ppc64--
See also the discussion in https://reviews.llvm.org/D35287
Adrian Prantl [Tue, 1 Aug 2017 21:45:24 +0000 (21:45 +0000)]
Remove PrologEpilogInserter's usage of DBG_VALUE's offset field
In the last half-dozen commits to LLVM I removed code that became dead
after removing the offset parameter from llvm.dbg.value gradually
proceeding from IR towards the backend. Before I can move on to
DwarfDebug and friends there is one last side-called offset I need to
remove: This patch modifies PrologEpilogInserter's use of the
DBG_VALUE's offset argument to use a DIExpression instead. Because the
PrologEpilogInserter runs at the Machine level I had to play a little
trick with a named llvm.dbg.mir node to get the DIExpressions to print
in MIR dumps (which print the llvm::Module followed by the
MachineFunction dump).
I also had to add rudimentary DwarfExpression support to CodeView and
as a side-effect also fixed a bug (CodeViewDebug::collectVariableInfo
was supposed to give up on variables with complex DIExpressions, but
would fail to do so for fragments, which are also modeled as
DIExpressions).
With this last holdover removed we will have only one canonical way of
representing offsets to debug locations which will simplify the code
in DwarfDebug (and future versions of CodeViewDebug once it starts
handling more complex expressions) and make it easier to reason about.
This patch is NFC-ish: All test case changes are for assembler
comments and the binary output does not change.
Vedant Kumar [Tue, 1 Aug 2017 21:23:26 +0000 (21:23 +0000)]
[llvm-cov] Allow specifying distinct architectures for each loaded binary
The coverage tool needs to know which slice to look at when it's handed
a universal binary. Some projects need to look at aggregate coverage
reports for a variety of slices in different binaries: this patch adds
support for these kinds of projects to llvm-cov.
Martin Storsjo [Tue, 1 Aug 2017 21:13:54 +0000 (21:13 +0000)]
[AArch64] Rewrite stack frame handling for win64 vararg functions
The previous attempt, which made do with a single offset in
computeCalleeSaveRegisterPairs, wasn't quite enough. The previous
attempt only worked as long as CombineSPBump == true (since the
offset would be adjusted later in fixupCalleeSaveRestoreStackOffset).
Instead include the size for the fixed stack area used for win64
varargs in calculations in emitPrologue/emitEpilogue. The stack
consists of mainly three parts;
- AFI->getLocalStackSize()
- AFI->getCalleeSavedStackSize()
- FixedObject
Most of the places in the code which previously used the CSStackSize
now use PrologueSaveSize instead, which is the sum of the latter
two, while some cases which need exactly the middle one use
AFI->getCalleeSavedStackSize() explicitly instead of a local variable.
In addition to moving the offsetting into emitPrologue/emitEpilogue
(which fixes functions with CombineSPBump == false), also set the
frame pointer to point to the right location, where the frame pointer
and link register actually are stored. In addition to the prologue/epilogue,
this also requires changes to resolveFrameIndexReference.
Add tests for a function that keeps a frame pointer and another one
that uses a VLA.
Matt Arsenault [Tue, 1 Aug 2017 20:49:41 +0000 (20:49 +0000)]
AMDGPU: Fix handling of div_scale with undef inputs
The src0 register must match src1 or src2, but if these
were undefined they could end up using different implicit_defed
virtual registers. Force these to use one undef vreg or pick the
defined other register.
Also fixes producing invalid nodes without the right number of
inputs when src2 is undef.
Matt Arsenault [Tue, 1 Aug 2017 19:54:18 +0000 (19:54 +0000)]
AMDGPU: Initial implementation of calls
Includes a hack to fix the type selected for
the GlobalAddress of the function, which will be
fixed by changing the default datalayout to use
generic pointers for 0.
Reid Kleckner [Tue, 1 Aug 2017 19:45:09 +0000 (19:45 +0000)]
[DebugInfo] Don't turn dbg.declare into DBG_VALUE for static allocas
Summary:
We already have information about static alloca stack locations in our
side table. Emitting instructions for them is inefficient, and it only
happens when the address of the alloca has been materialized within the
current block, which isn't often.
Craig Topper [Tue, 1 Aug 2017 17:18:14 +0000 (17:18 +0000)]
[X86] Use BEXTR/BEXTRI for 64-bit 'and' with a large mask
Summary: The 64-bit 'and' with immediate instruction only supports a 32-bit immediate. So for larger constants we have to load the constant into a register first. If the immediate happens to be a mask we can use the BEXTRI instruction to perform the masking. We already do something similar using the BZHI instruction from the BMI2 instruction set.
Manoj Gupta [Tue, 1 Aug 2017 15:39:12 +0000 (15:39 +0000)]
[X86] Fix a crash in FEntryInserter Pass.
Summary:
FEntryInserter pass unconditionally derefs the first Instruction
in the first Basic Block. The pass crashes when the first
BasicBlock is empty. Fix the crash by not dereferencing the basic
Block iterator. This fixes an issue observed when building Linux kernel
4.4 with clang.
Craig Topper [Tue, 1 Aug 2017 15:31:24 +0000 (15:31 +0000)]
[AVX-512] Don't use unmasked VMOVDQU8/16 for 8-bit or 16-bit element stores even when BWI instructions are supported. Always use VMOVDQA32/VMOVDQU32.
We were already using the 32 bit element opcode if BWI isn't enabled, but there's no reason to change opcode if we have BWI. We will still use the 8/16 opcodes for masked stores though.
This allows us to use the aligned opcode when we can which makes our test output more consistent between different modes. It also reduces the number of isel patterns we need.
This is a slight inconsistency with loads which default to 64 bit element opcodes. I'll probably rectify that in a future patch.
Craig Topper [Tue, 1 Aug 2017 15:10:25 +0000 (15:10 +0000)]
[InstCombine] Remove explicit check for impossible condition. Replace with assert
Summary:
As far as I can tell the earlier call getLimitedValue will guaranteed ShiftAmt is saturated to BitWidth-1 preventing it from ever being equal or greater than BitWidth.
At one point in the past the getLimitedValue call was only passed BitWidth not BitWidth - 1. This would have allowed the equality case to get here. And in fact this check was initially added as just BitWidth == ShiftAmt, but was changed shortly after to include > which should have never been possible.
Tobias Grosser [Tue, 1 Aug 2017 14:40:55 +0000 (14:40 +0000)]
[PostDom] document the current handling of infinite loops and unreachables
Summary:
As we are in the process of changing the behavior of how the post-dominator tree
is computed, make sure we have some more test coverage in this area.
Current inconsistencies:
- Newly unreachable nodes are not added as new roots, in case the PDT is updated
but not rebuilt.
- Newly unreachable loops are not added to the CFG at all (neither when
building from scratch nor when updating the CFG). This is inconsistent with
the fact that unreachables are added to the PDT, but unreachable loops not.
On the other side, PDT relationships are not loosened at the moment in
cases where new unreachable loops are built.
This commit is providing additional test coverage for
https://reviews.llvm.org/D35851
Nirav Dave [Tue, 1 Aug 2017 13:45:35 +0000 (13:45 +0000)]
[DAG] Extend visitSCALAR_TO_VECTOR optimization to truncated vector.
Summary:
Allow SCALAR_TO_VECTOR of EXTRACT_VECTOR_ELT to reduce to
EXTRACT_SUBVECTOR of vector shuffle when output is smaller. Marginally
improves vector shuffle computations.
This patch enables control flow optimization for
variations of BBIT instruction. In this case
optimization removes unnecessary branch after
BBIT instruction.
Support itineraries in TargetSubtargetInfo::getSchedInfoStr - Now if the given instr does not have sched model then we try to calculate the latecy/throughput with help of itineraries.
Differential Revision https://reviews.llvm.org/D35997
Petr Hosek [Tue, 1 Aug 2017 05:18:30 +0000 (05:18 +0000)]
[llvm][llvm-objcopy] Added support for outputting to binary in llvm-objcopy
This change adds the "-O binary" flag which directs llvm-objcopy to
output the object file to the same format as GNU objcopy does when given
the flag "-O binary". This was done by splitting the Object class into
two subclasses ObjectELF and ObjectBianry which each output a different
format but relay on the same code to read in the Object in Object.
Davide Italiano [Tue, 1 Aug 2017 05:14:45 +0000 (05:14 +0000)]
[MetaRenamer] Leave `@main` alone.
To the best of my knowledge -metarenamer is used in two cases:
1) obfuscate names, when e.g. they contain informations that
can't be shared.
2) Improve clarity of the textual IR for testcases.
One of the usecases if getting the output of `opt` and passing it
to the lli interpreter to run the test. If metarenamer renames
@main, lli can't find an entry point.
projects / llvm / log