Fix for https://bugs.llvm.org/show_bug.cgi?id=41477. On the x32 ABI
with stack probing a dynamic alloca will result in a WIN_ALLOCA_32
with a 32-bit size. The current implementation tries to copy it into
RAX, resulting in a physreg copy error. Fix this by copying to EAX
instead. Also fix incorrect opcodes or registers used in subs.
[X86] Don't turn (and (shl X, C1), C2) into (shl (and X, (C1 >> C2), C2) if the original AND can represented by MOVZX.
The MOVZX doesn't require an immediate to be encoded at all. Though it does use
a 2 byte opcode so its the same size as a 1 byte immediate. But it has a
separate source and dest register so can help avoid copies.
Sam Clegg [Fri, 19 Apr 2019 22:43:32 +0000 (22:43 +0000)]
[WebAssembly] FastISel: Don't fallback to SelectionDAG after BuildMI in selectCall
My understanding is that once BuildMI has been called we can't fallback
to SelectionDAG.
This change moves the fallback for when getRegForValue() fails for
that target of an indirect call. This was failing in -fPIC mode when
the callee is GlobalValue.
Summary:
This emits labels around heapallocsite calls and S_HEAPALLOCSITE debug
info in codeview. Currently only changes FastISel, so emitting labels still
needs to be implemented in SelectionDAG.
These are general queries, so they should not die when given
a degenerate input like an all undef mask. Callers should be
able to deal with an op that will eventually be simplified away.
[LTO] Add plumbing to save stats during LTO on Darwin.
Gold and ld on Linux already support saving stats, but the
infrastructure is missing on Darwin. Unfortunately it seems like the
configuration from lib/LTO/LTO.cpp is not used.
This patch adds a new LTOStatsFile option and adds plumbing in Clang to
use it on Darwin, similar to the way remarks are handled.
Currnetly the handling of LTO flags seems quite spread out, with a bunch
of duplication. But I am not sure if there is an easy way to improve
that?
Igor Kudrin [Fri, 19 Apr 2019 10:12:56 +0000 (10:12 +0000)]
[llvm-symbolizer] Make the output with -output-style=GNU closer to addr2line's
This patch addresses two differences in the output of llvm-symbolizer
and GNU's addr2line:
* llvm-symbolizer prints an empty line after the report for an address.
* With "-f -i=0", llvm-symbolizer replaces the name of an inlined
function with the name from the symbol table, i. e., the top caller
function in the inlining chain. addr2line preserves the name of the
inlined function.
Summary:
The basic idea here is to make it possible to use
MachineInstr::mayAlias also when the MachineInstr
is const (or the "Other" MachineInstr is const).
The addition of const in MachineInstr::mayAlias
then rippled down to the need for adding const
in several other places, such as
TargetTransformInfo::getMemOperandWithOffset.
James Molloy [Fri, 19 Apr 2019 09:00:55 +0000 (09:00 +0000)]
[PATCH] [MachineScheduler] Check pending instructions when an instruction is scheduled
Pending instructions that may have been blocked from being available by the HazardRecognizer may no longer may not be blocked any more when an instruction is scheduled; pending instructions should be re-checked in this case.
This is primarily aimed at VLIW targets with large parallelism and esoteric constraints.
No testcase as no in-tree targets have this behavior.
Fangrui Song [Fri, 19 Apr 2019 07:57:51 +0000 (07:57 +0000)]
[MergeFunc] removeUsers: call remove() only on direct users
removeUsers uses a work list to collect indirect users and call remove()
on those functions. However it has a bug (`if (!Visited.insert(UU).second)`).
Actually, we don't have to collect indirect users.
After the merge of F and G, G's callers will be considered (added to
Deferred). If G's callers can be merged, G's callers' callers will be
considered.
Update the test unnamed-addr-reprocessing.ll to make it clear we can
still merge indirect callers.
Piotr Sobczak [Fri, 19 Apr 2019 06:19:14 +0000 (06:19 +0000)]
[AMDGPU] Ignore non-SUnits edges
Summary:
Ignore edges to non-SUnits (e.g. ExitSU) when checking
for low latency instructions.
When calling the function isLowLatencyInstruction(),
an ExitSU could be on the list of successors, not necessarily
a regular SU. In other places in the code there is a check
"Succ->NodeNum >= DAGSize" to prevent further processing of
ExitSU as "Succ->getInstr()" is NULL in such a case.
Also, 8 out of 9 cases of "SUnit *Succ = SuccDep.getSUnit())"
has the guard, so it is clearly an omission here.
[CallSite removal] Move the legacy PM, call graph, and some inliner
code to `CallBase`.
This patch focuses on the legacy PM, call graph, and some of inliner and legacy
passes interacting with those APIs from `CallSite` to the new `CallBase` class.
No interesting changes.
[X86] Make sure we copy the HandleSDNode back to N before executing the default code after the switch in matchAddressRecursively
Summary:
There are two places where we create a HandleSDNode in address matching in order to handle the case where N is changed by CSE. But if we end up not matching, we fall back to code at the bottom of the switch that really would like N to point to something that wasn't CSEd away. So we should make sure we copy the handle back to N on any paths that can reach that code.
This appears to be the true reason we needed to check DELETED_NODE in the negation matching. In pr32329.ll we had two subtracts back to back. We recursed through the first subtract, and onto the second subtract. The second subtract called matchAddressRecursively on its LHS which caused that subtract to CSE. We ultimately failed the match and ended up in the default code. But N was pointing at the old node that had been deleted, but the default code didn't know that and took it as the base register. Then we unwound back to the first subtract and tried to access this bogus base reg requiring the check for deleted node. With this patch we now use the CSE result as the base reg instead.
matchAdd has been broken since sometime in 2015 when it was pulled out of the switch into a helper function. The assignment to N at the end was still there, but N was passed by value and not by reference so the update didn't go anywhere.
Ali Tamur [Fri, 19 Apr 2019 02:26:56 +0000 (02:26 +0000)]
[llvm] Prevent duplicate files in debug line header in dwarf 5: another attempt
Another attempt to land the changes in debug line header to prevent duplicate
files in Dwarf 5. I rolled back my previous commit because of a mistake in
generating the object file in a test. Meanwhile, I addressed some offline
comments and changed the implementation; the largest difference is that
MCDwarfLineTableHeader does not keep DwarfVersion but gets it as a parameter. I
also merged the patch to fix two lld tests that will strt to fail into this
patch.
Original Commit:
https://reviews.llvm.org/D59515
Original Message:
Motivation: In previous dwarf versions, file name indexes started from 1, and
the primary source file was not explicit. Dwarf 5 standard (6.2.4) prescribes
the primary source file to be explicitly given an entry with an index number 0.
The current implementation honors the specification by just duplicating the
main source file, once with index number 0, and later maybe with another
index number. While this is compliant with the letter of the standard, the
duplication causes problems for consumers of this information such as lldb.
(Some files are duplicated, where only some of them have a line table although
all refer to the same file)
With this change, dwarf 5 debug line section files always start from 0, and
the zeroth entry is not duplicated whenever possible. This requires different
handling of dwarf 4 and dwarf 5 during generation (e.g. when a function returns
an index zero for a file name, it signals an error in dwarf 4, but not in dwarf
5) However, I think the minor complication is worth it, because it enables all
consumers (lldb, gdb, dwarfdump, objdump, and so on) to treat all files in the
file name list homogenously.
MergeFunc: preserve COMDAT information when creating a thunk
We would previously drop the COMDAT on the thunk we generated when replacing a
function body with the forwarding thunk. This would result in a function that
may have been multiply emitted and multiply merged to be emitted with the same
name without the COMDAT. This is a hard error with PE/COFF where the COMDAT is
used for the deduplication of Value Witness functions for Swift.
Adrian Prantl [Thu, 18 Apr 2019 21:22:50 +0000 (21:22 +0000)]
Implement sys::fs::copy_file using the macOS copyfile(3) API
to support APFS clones.
This patch adds a Darwin-specific implementation of
llvm::sys::fs::copy_file() that uses the macOS copyfile(3) API to
support APFS copy-on-write clones, which should be faster and much
more space efficient.
[GlobalISel][AArch64] Legalize/select G_(S/Z/ANY)_EXT for v8s8s
This adds legalization for G_SEXT, G_ZEXT, and G_ANYEXT for v8s8s.
We were falling back on G_ZEXT in arm64-vabs.ll before, preventing us from
selecting the @llvm.aarch64.neon.sabd.v8i8 intrinsic.
This adds legalizer support for those 3, which gives us selection via the
importer. Update the relevant tests (legalize-ext.mir, select-int-ext.mir) and
add a GISel line to arm64-vabs.ll.
[BlockExtractor] Extend the file format to support the grouping of basic blocks
Prior to this patch, each basic block listed in the extrack-blocks-file
would be extracted to a different function.
This patch adds the support for comma separated list of basic blocks
to form group.
When the region formed by a group is not extractable, e.g., not single
entry, all the blocks of that group are left untouched.
Let us see this new format in action (comments are not part of the
file format):
;; funcName bbName[,bbName...]
foo bb1 ;; Extract bb1 in its own function
foo bb2,bb3 ;; Extract bb2,bb3 in their own function
bar bb1,bb4 ;; Extract bb1,bb4 in their own function
bar bb2 ;; Extract bb2 in its own function
Assuming all regions are extractable, this will create one function and
thus one call per region.
Simon Pilgrim [Thu, 18 Apr 2019 17:23:09 +0000 (17:23 +0000)]
[X86] combineVectorTruncationWithPACKUS - remove split/concatenation of mask
combineVectorTruncationWithPACKUS is currently splitting the upper bit bit masking into 128-bit subregs and then concatenating them back together.
This was originally done to avoid regressions that caused existing subregs to be concatenated to the larger type just for the AND masking before being extracted again. This was fixed by @spatel (notably rL303997 and rL347356).
This also lets SimplifyDemandedBits do some further improvements before it hits the recursive depth limit.
My only annoyance with this is that we were broadcasting some xmm masks but we seem to have lost them by moving to ymm - but that's a known issue as the logic in lowerBuildVectorAsBroadcast isn't great.
Philip Reames [Thu, 18 Apr 2019 17:01:19 +0000 (17:01 +0000)]
[LoopPred] Fix a blatantly obvious bug in r358684
The bug is that I didn't check whether the operand of the invariant_loads were themselves invariant. I don't know how this got missed in the patch and review. I even had an unreduced test case locally, and I remember handling this case, but I must have lost it in one of the rebases. Oops.
Stefan Granitz [Thu, 18 Apr 2019 16:37:07 +0000 (16:37 +0000)]
[CMake] Allow custom extensions for externalized debug info
Summary:
Extra flexibility for emitting debug info to external files (remains Darwin only for now).
LLDB needs this functionality to emit a LLDB.framework.dSYM instead of LLDB.dSYM when building the framework, because the latter could conflict with the driver's lldb.dSYM when emitted in the same directory on case-insensitive file systems.
Philip Reames [Thu, 18 Apr 2019 16:33:17 +0000 (16:33 +0000)]
[LoopPredication] Allow predication of loop invariant computations (within the loop)
The purpose of this patch is to eliminate a pass ordering dependence between LoopPredication and LICM. To understand the purpose, consider the following snippet of code inside some loop 'L' with IV 'i'
A = _a.length;
guard (i < A)
a = _a[i]
B = _b.length;
guard (i < B);
b = _b[i];
...
Z = _z.length;
guard (i < Z)
z = _z[i]
accum += a + b + ... + z;
Today, we need LICM to hoist the length loads, LoopPredication to make the guards loop invariant, and TrivialUnswitch to eliminate the loop invariant guard to establish must execute for the next length load. Today, if we can't prove speculation safety, we'd have to iterate these three passes 26 times to reduce this example down to the minimal form.
Using the fact that the array lengths are known to be invariant, we can short circuit this iteration. By forming the loop invariant form of all the guards at once, we remove the need for LoopPredication from the iterative cycle. At the moment, we'd still have to iterate LICM and TrivialUnswitch; we'll leave that part for later.
As a secondary benefit, this allows LoopPred to expose peeling oppurtunities in a much more obvious manner. See the udiv test changes as an example. If the udiv was not hoistable (i.e. we couldn't prove speculation safety) this would be an example where peeling becomes obviously profitable whereas it wasn't before.
A couple of subtleties in the implementation:
- SCEV's isSafeToExpand guarantees speculation safety (i.e. let's us expand at a new point). It is not a precondition for expansion if we know the SCEV corresponds to a Value which dominates the requested expansion point.
- SCEV's isLoopInvariant returns true for expressions which compute the same value across all iterations executed, regardless of where the original Value is located. (i.e. it can be in the loop) This implies we have a speculation burden to prove before expanding them outside loops.
- invariant_loads and AA->pointsToConstantMemory are two cases that SCEV currently does not handle, but meets the SCEV definition of invariance. I plan to sink this part into SCEV once this has baked for a bit.
[SDA] Bug fix: Use IPD outside the loop as divergence bound
Summary:
The immediate post dominator of the loop header may be part of the divergent loop.
Since this /was/ the divergence propagation bound the SDA would not detect joins of divergent paths outside the loop.
Philip Reames [Thu, 18 Apr 2019 16:10:21 +0000 (16:10 +0000)]
Fix a bug in SCEV's isSafeToExpand around speculation safety
isSafeToExpand was making a common, but dangerously wrong, mistake in assuming that if any instruction within a basic block executes, that all instructions within that block must execute. This can be trivially shown to be false by considering the following small example:
bb:
add x, y <-- InsertionPoint
call @throws()
udiv x, y <-- SCEV* S
br ...
It's clearly not legal to expand S above the throwing call, but the previous logic would do so since S dominates (but not properlyDominates) the block containing the InsertionPoint.
Since iterating instructions w/in a block is expensive, this change special cases two cases: 1) S is an operand of InsertionPoint, and 2) InsertionPoint is the terminator of it's block. These two together are enough to keep all current optimizations triggering while fixing the latent correctness issue.
As best I can tell, this is a silent bug in current ToT. Given that, there's no tests with this change. It was noticed in an upcoming optimization change (D60093), and was reviewed as part of that. That change will include the test which caused me to notice the issue. I'm submitting this seperately so that anyone bisecting a problem gets a clear explanation.
Pavel Labath [Thu, 18 Apr 2019 14:57:31 +0000 (14:57 +0000)]
MinidumpYAML: Add support for ModuleList stream
Summary:
This patch adds support for yaml (de)serialization of the minidump
ModuleList stream. It's a fairly straight forward-application of the
existing patterns to the ModuleList structures defined in previous
patches.
One thing, which may be interesting to call out explicitly is the
addition of "new" allocation functions to the helper BlobAllocator
class. The reason for this was, that there was an emerging pattern of a
need to allocate space for entities, which do not have a suitable
lifetime for use with the existing allocation functions. A typical
example of that was the "size" of various lists, which is only available
as a temporary returned by the .size() method of some container. For
these cases, one can use the new set of allocation functions, which
will take a temporary object, and store it in an allocator-managed
buffer until it is written to disk.
with the more general icmp lowering so it can pick up more cases through bitcasts - notably vXi8 cases which use vXi16 shifts+masks, this patch can remove the mask and use pcmpgtb(0,x) for the sra.
James Henderson [Thu, 18 Apr 2019 09:13:30 +0000 (09:13 +0000)]
[llvm-objcopy][llvm-strip] Add switch to allow removing referenced sections
llvm-objcopy currently emits an error if a section to be removed is
referenced by another section. This is a reasonable thing to do, but is
different to GNU objcopy. We should allow users who know what they are
doing to have a way to produce the invalid ELF. This change adds a new
switch --allow-broken-links to both llvm-strip and llvm-objcopy to do
precisely that. The corresponding sh_link field is then set to 0 instead
of an error being emitted.
I cannot use llvm-readelf/readobj to test the link fields because they
emit an error if any sections, like the .dynsym, cannot be properly
loaded.
Kang Zhang [Thu, 18 Apr 2019 07:24:15 +0000 (07:24 +0000)]
[PowerPC] Fix wrong ElemSIze when calling isConsecutiveLS()
Summary:
This issue from the bugzilla: https://bugs.llvm.org/show_bug.cgi?id=41177
When the two operands for BUILD_VECTOR are same, we will get assert error.
llvm::SDValue combineBVOfConsecutiveLoads(llvm::SDNode*, llvm::SelectionDAG&):
Assertion `!(InputsAreConsecutiveLoads && InputsAreReverseConsecutive) &&
"The loads cannot be both consecutive and reverse consecutive."' failed.
This error caused by the wrong ElemSIze when calling isConsecutiveLS(). We
should use `getScalarType().getStoreSize();` to get the ElemSize instread of
`getScalarSizeInBits() / 8`.
Tim Renouf [Thu, 18 Apr 2019 05:27:01 +0000 (05:27 +0000)]
[AMDGPU] Avoid DAG combining assert with fneg(fadd(A,0))
fneg combining attempts to turn it into fadd(fneg(A), fneg(0)), but
creating the new fadd folds to just fneg(A). When A has multiple uses,
this confuses it and you get an assert. Fixed.
Richard Trieu [Thu, 18 Apr 2019 01:39:45 +0000 (01:39 +0000)]
Fix bad compare function over FusionCandidate.
Reverse the checking of the domiance order so that when a self compare happens,
it returns false. This makes compare function have strict weak ordering.
Adrian Prantl [Thu, 18 Apr 2019 00:01:05 +0000 (00:01 +0000)]
Implement sys::fs::copy_file using the macOS copyfile(3) API
to support APFS clones.
This patch adds a Darwin-specific implementation of
llvm::sys::fs::copy_file() that uses the macOS copyfile(3) API to
support APFS copy-on-write clones, which should be faster and much
more space efficient.
...so if we can fold the shift op into LEA in the 1st pattern, then we
should be able to do the same in the 2nd pattern (unnecessary 'movzbl'
is a separate bug I think).
We don't want to do this any sooner though because that would conflict
with generic transforms that try to narrow the width of the shift.
Nick Desaulniers [Wed, 17 Apr 2019 22:21:10 +0000 (22:21 +0000)]
[AsmPrinter] hoist %a output template to base class for ARM+Aarch64
Summary:
X86 is quite complicated; so I intend to leave it as is. ARM+Aarch64 do
basically the same thing (Aarch64 did not correctly handle immediates,
ARM has a test llvm/test/CodeGen/ARM/2009-04-06-AsmModifier.ll that uses
%a with an immediate) for a flag that should be target independent
anyways.
This patch adds a basic loop fusion pass. It will fuse loops that conform to the
following 4 conditions:
1. Adjacent (no code between them)
2. Control flow equivalent (if one loop executes, the other loop executes)
3. Identical bounds (both loops iterate the same number of iterations)
4. No negative distance dependencies between the loop bodies.
The pass does not make any changes to the IR to create opportunities for fusion.
Instead, it checks if the necessary conditions are met and if so it fuses two
loops together.
The pass has not been added to the pass pipeline yet, and thus is not enabled by
default. It can be run stand alone using the -loop-fusion option.
Nick Desaulniers [Wed, 17 Apr 2019 18:22:48 +0000 (18:22 +0000)]
[AsmPrinter] defer %c to base class for ARM, PPC, and Hexagon. NFC
Summary:
None of these derived classes do anything that the base class cannot.
If we remove these case statements, then the base class can handle them
just fine.
Steven Wu [Wed, 17 Apr 2019 17:38:09 +0000 (17:38 +0000)]
[ThinLTO] Fix ThinLTOCodegenerator to export llvm.used symbols
Summary:
Reapply r357931 with fixes to ThinLTO testcases and llvm-lto tool.
ThinLTOCodeGenerator currently does not preserve llvm.used symbols and
it can internalize them. In order to pass the necessary information to the
legacy ThinLTOCodeGenerator, the input to the code generator is
rewritten to be based on lto::InputFile.
Now ThinLTO using the legacy LTO API will requires data layout in
Module.
"internalize" thinlto action in llvm-lto is updated to run both
"promote" and "internalize" with the same configuration as
ThinLTOCodeGenerator. The old "promote" + "internalize" option does not
produce the same output as ThinLTOCodeGenerator.
Philip Reames [Wed, 17 Apr 2019 17:37:58 +0000 (17:37 +0000)]
[InstCombine] Factor out unreachable inst idiom creation [NFC]
In InstCombine, we use an idiom of "store i1 true, i1 undef" to indicate we've found a path which we've proven unreachable. We can't actually insert the unreachable instruction since that would require changing the CFG. We leave that to simplifycfg later.
This just factors out that idiom creation so we don't duplicate the same mostly undocument idiom creation in multiple places.
[LVI][CVP] Constrain values in with.overflow branches
If a branch is conditional on extractvalue(op.with.overflow(%x, C), 1)
then we can constrain the value of %x inside the branch based on
makeGuaranteedNoWrapRegion(). We do this by extending the edge-value
handling in LVI. This allows CVP to then fold comparisons against %x,
as illustrated in the tests.
projects / llvm / log