Max Kazantsev [Tue, 12 Feb 2019 09:59:44 +0000 (09:59 +0000)]
[IndVars] Fix corner case with unreachable Phi inputs. PR40454
Logic in `getInsertPointForUses` doesn't account for a corner case when `Def`
only comes to a Phi user from unreachable blocks. In this case, the incoming
value may be arbitrary (and not even available in the input block) and break
the loop-related invariants that are asserted below.
In fact, if we encounter this situation, no IR modification is needed. This
Phi will be simplified away with nearest cleanup.
George Rimar [Tue, 12 Feb 2019 09:50:04 +0000 (09:50 +0000)]
[llvm-readobj] - Simplify .gnu.version_r dumping a bit.
Current implementation takes "Number of needed versions" from DT_VERNEEDNUM
dynamic tag entry. Though it would be a bit simpler to take it from sh_info
section header field directly:
https://docs.oracle.com/cd/E19683-01/816-1386/chapter6-94076/index.html
Max Kazantsev [Tue, 12 Feb 2019 09:37:00 +0000 (09:37 +0000)]
[LoopSimplifyCFG] Change logic of dead loops removal to avoid hitting asserts
The function `LI.erase` has some invariants that need to be preserved when it
tries to remove a loop which is not the top-level loop. In particular, it
requires loop's preheader to be strictly in loop's parent. Our current logic
of deletion of dead blocks may erase the information about preheader before we
handle the loop, and therefore we may hit this assertion.
This patch changes the logic of loop deletion: we make them top-level loops
before we actually erase them. This allows us to trigger the simple branch of
`erase` logic which just detatches blocks from the loop and does not try to do
some complex stuff that need this invariant.
George Rimar [Tue, 12 Feb 2019 09:08:59 +0000 (09:08 +0000)]
[yaml2obj/obj2yaml] - Move `Info` field out from `Section` class.
ELFYAML.h contains a `Section` class which is a base for a few other
sections classes that are used for mapping different section types.
`Section` has a `StringRef Info` field used for storing sh_info.
At the same time, sh_info has very different meanings for sections and
cannot be processed in a similar way generally,
for example ELFDumper does not handle it in `dumpCommonSection`
but do that in `dumpGroup` and `dumpCommonRelocationSection` respectively.
At this moment, we have and handle it as a string, because that was possible for
the current use case. But also it can simply be a number:
For SHT_GNU_verdef is "The number of version definitions within the section."
The patch moves `Info` field out to be able to have it as a number.
With that change, each class will be able to decide what type and purpose
of the sh_info field it wants to use.
I also had to edit 2 test cases. This is because patch fixes a bug. Previously we
accepted yaml files with Info fields for all sections (for example, for SHT_DYNSYM too).
But we do not handle it and the resulting objects had zero sh_info fields set for
such sections. Now it is accepted only for sections that supports it.
Max Kazantsev [Tue, 12 Feb 2019 07:48:07 +0000 (07:48 +0000)]
[LoopSimplifyCFG] Pay respect to LCSSA when removing dead blocks
Utility function that we use for blocks deletion always unconditionally removes
one-input Phis. In LoopSimplifyCFG, it can lead to breach of LCSSA form.
This patch alters this function to keep them if needed.
Eli Friedman [Tue, 12 Feb 2019 00:33:25 +0000 (00:33 +0000)]
[LoopReroll] Fix reroll root legality checking.
The code checked that the first root was an appropriate distance from
the base value, but skipped checking the other roots. This could lead to
rerolling a loop that can't be legally rerolled (at least, not without
rewriting the loop in a non-trivial way).
David Blaikie [Tue, 12 Feb 2019 00:00:38 +0000 (00:00 +0000)]
DebugInfo: Split DWARF + gmlt + no-split-dwarf-inlining shouldn't emit anything to the .dwo file
This configuration (due to r349207) was intended not to emit any DWO CU,
but a degenerate CU was still being emitted - containing a header and a
DW_TAG_compile_unit with no attributes.
Under that situation, emit nothing to the .dwo file. (since this is a
dynamic property of the input the .dwo file is still emitted, just with
nothing in it (so a valid, but empty, ELF file) - if some other CU
didn't satisfy this criteria, its DWO CU would still go there, etc)
Philip Reames [Mon, 11 Feb 2019 23:34:33 +0000 (23:34 +0000)]
Be conservative about unordered accesses for the moment
Background: As described in https://reviews.llvm.org/D57601, I'm working towards separating volatile and atomic in the MMO uses for atomic instructions.
In https://reviews.llvm.org/D57593, I fixed a bug where isUnordered was returning the wrong result, but didn't account for the fact I was getting slightly ahead of myself. While both uses of isUnordered are correct (as far as I can tell), we don't have tests to demonstrate this and being aggressive gets in the way of having the removal of volatile truly be non-functional. Once D57601 lands, I will return to these call sites, revert this patch, and add the appropriate tests to show the expected behaviour.
Daniel Sanders [Mon, 11 Feb 2019 23:02:02 +0000 (23:02 +0000)]
[tblgen] Add a timer covering the time spent reading the Instruction defs
This patch adds a -time-regions option to tablegen that can enable timers
(currently only one) that assess the performance of tablegen itself. This
can be useful for identifying scaling problems with tablegen backends.
This particular timer has allowed me to ignore time that is not attributed
the GISel combiner pass. It's useful by itself but it is particularly
useful in combination with https://reviews.llvm.org/D52954 which causes
this period of time to be annotated within Xcode Instruments which in turn
allows profile samples and recorded allocations attributed to reading
instructions to be filtered out.
Ana Pazos [Mon, 11 Feb 2019 22:10:08 +0000 (22:10 +0000)]
[LegalizeTypes] Expand FNEG to bitwise op for IEEE FP types
Summary:
Except for custom floating point types x86_fp80 and ppc_fp128,
expand Y = FNEG(X) to Y = X ^ sign mask to avoid library call.
Using bitwise operation can improve code size and performance.
Scott Linder [Mon, 11 Feb 2019 22:01:13 +0000 (22:01 +0000)]
[IRReader] Expose getLazyIRModule
Currently there is no way to lazy-load an in-memory IR module without
first writing it to disk. This patch just exposes the existing
implementation of getLazyIRModule.
Craig Topper [Mon, 11 Feb 2019 20:38:10 +0000 (20:38 +0000)]
[X86] Correct the memory operand for the FLD emitted in FP_TO_INTHelper for 32-bit SSE targets.
We were using DstTy, but that represents the integer type we are converting to which is i64 in this
case. The FLD is part of an intermediate step to get from the SSE registers to the x87 registers.
If the floating point type is f32, the memory operand should reflect a 4 byte access not an 8 byte
access. The store we used to get from SSE to the stack is using the corect size.
While there, consistenly use TheVT in place of Op.getOperand(0).getValueType() throughout the function.
Matt Davis [Mon, 11 Feb 2019 20:30:53 +0000 (20:30 +0000)]
[llvm-cxxfilt] Split and demangle stdin input
Summary:
Originally, llvm-cxxfilt would treat a line as a single mangled item to be demangled.
If a mangled name appears in the middle of that string, that name would not be demangled.
GNU c++filt splits and demangles every word in a string that is piped to it via stdin.
Prior to this patch llvm-cxxfilt would never split strings piped to it.
This patch replicates the GNU behavior and splits strings that are piped to it via stdin.
Alina Sbirlea [Mon, 11 Feb 2019 19:51:21 +0000 (19:51 +0000)]
[MemorySSA] Remove verifyClobberSanity.
Summary:
This verification may fail after certain transformations due to
BasicAA's fragility. Added a small explanation and a testcase that
triggers the assert in checkClobberSanity (before its removal).
Addresses PR40509.
Michael Kruse [Mon, 11 Feb 2019 19:45:44 +0000 (19:45 +0000)]
Refactor setAlreadyUnrolled() and setAlreadyVectorized().
Loop::setAlreadyUnrolled() and
LoopVectorizeHints::setLoopAlreadyUnrolled() both add loop metadata that
stops the same loop from being transformed multiple times. This patch
merges both implementations.
In doing so we fix 3 potential issues:
* setLoopAlreadyUnrolled() kept the llvm.loop.vectorize/interleave.*
metadata even though it will not be used anymore. This already caused
problems such as http://llvm.org/PR40546. Change the behavior to the
one of setAlreadyUnrolled which deletes this loop metadata.
* setAlreadyUnrolled() used to create a new LoopID by calling
MDNode::get with nullptr as the first operand, then replacing it by
the returned references using replaceOperandWith. It is possible
that MDNode::get would instead return an existing node (due to
de-duplication) that then gets modified. To avoid, use a fresh
TempMDNode that does not get uniqued with anything else before
replacing it with replaceOperandWith.
* LoopVectorizeHints::matchesHintMetadataName() only compares the
suffix of the attribute to set the new value for. That is, when
called with "enable", would erase attributes such as
"llvm.loop.unroll.enable", "llvm.loop.vectorize.enable" and
"llvm.loop.distribute.enable" instead of the one to replace.
Fortunately, function was only called with "isvectorized".
Sanjay Patel [Mon, 11 Feb 2019 19:26:27 +0000 (19:26 +0000)]
[InstCombine] Fix matchRotate bug when one operand is a ConstantExpr shift
This bug seems to be harmless in release builds, but will cause an error in UBSAN
builds or an assertion failure in debug builds.
When it gets to this opcode comparison, it assumes both of the operands are BinaryOperators,
but the prior m_LogicalShift will also match a ConstantExpr. The cast<BinaryOperator> will
assert in a debug build, or reading an invalid value for BinaryOp from memory with
((BinaryOperator*)constantExpr)->getOpcode() will cause an error in a UBSAN build.
The test I added will fail without this change in debug/UBSAN builds, but not in release.
Bjorn Pettersson [Mon, 11 Feb 2019 19:23:30 +0000 (19:23 +0000)]
[SelectionDAGBuilder] Add restrictions to EmitFuncArgumentDbgValue
Summary:
This patch fixes PR40587.
When a dbg.value instrinsic is emitted to the DAG
by using EmitFuncArgumentDbgValue the resulting
DBG_VALUE is hoisted to the beginning of the entry
block. I think the idea is to be able to locate
a formal argument already from the start of the
function.
However, EmitFuncArgumentDbgValue only checked that
the value that was used to describe a variable was
originating from a function parameter, not that the
variable itself actually was an argument to the
function. So when for example assigning a local
variable "local" the value from an argument "a",
the assocated DBG_VALUE instruction would be hoisted
to the beginning of the function, even if the scope
for "local" started somewhere else (or if "local"
was mapped to other values earlier in the function).
This patch adds some logic to EmitFuncArgumentDbgValue
to check that the variable being described actually
is an argument to the function. And that the dbg.value
being lowered already is in the entry block. Otherwise
we bail out, and the dbg.value will be handled as an
ordinary dbg.value (not as a "FuncArgumentDbgValue").
A tricky situation is when both the variable and
the value is related to function arguments, but not
neccessarily the same argument. We make sure that we
do not describe the same argument more than once as
a "FuncArgumentDbgValue". This solution works as long
as opt has injected a "first" dbg.value that corresponds
to the formal argument at the function entry.
Alina Sbirlea [Mon, 11 Feb 2019 19:07:15 +0000 (19:07 +0000)]
[LICM&MSSA] Limit store hoisting.
Summary:
If there is no clobbering access for a store inside the loop, that store
can only be hoisted if there are no interfearing loads.
A more general verification introduced here: there are no loads that are
not optimized to an access outside the loop.
Addresses PR40586.
Jordan Rupprecht [Mon, 11 Feb 2019 18:05:48 +0000 (18:05 +0000)]
[DebugInfo] Fix /usr/lib/debug llvm-symbolizer lookup with relative paths
Summary:
rL189250 added a realpath call, and rL352916 because realpath breaks assumptions with some build systems. However, the /usr/lib/debug case has been clarified, falling back to /usr/lib/debug is currently broken if the obj passed in is a relative path. Adding a call to use absolute paths when falling back to /usr/lib/debug fixes that while still not making any realpath assumptions.
This also adds a --fallback-debug-path command line flag for testing (since we probably can't write to /usr/lib/debug from buildbot environments), but was also verified manually:
Lookups went to relative/path/to/dwarfdump-test.elf-x86-64, relative/path/to/.debug/dwarfdump-test.elf-x86-64, and then finally /usr/lib/debug/absolute/path/to/dwarfdump-test.elf-x86-64.
Andrea Di Biagio [Mon, 11 Feb 2019 17:55:47 +0000 (17:55 +0000)]
[MCA][Scheduler] Track resources that were found busy when issuing an instruction.
This is a follow up of r353706. When the scheduler fails to issue a ready
instruction to the underlying pipelines, it now updates a mask of 'busy resource
units'. That information will be used in future to obtain the set of
"problematic" resources in the case of bottlenecks caused by resource pressure.
No functional change intended.
Roland Froese [Mon, 11 Feb 2019 17:29:14 +0000 (17:29 +0000)]
[PowerPC] Avoid scalarization of vector truncate
The PowerPC code generator currently scalarizes vector truncates that would fit in a vector register, resulting in vector extracts, scalar operations, and vector merges. This patch custom lowers a vector truncate that would fit in a register to a vector shuffle instead.
David Greene [Mon, 11 Feb 2019 15:40:02 +0000 (15:40 +0000)]
Add recipes for migrating downstream branches of git mirrors
Add some common recipes for downstream users developing on top of the
existing git mirrors. These instructions show how to migrate local
branches to the monorepo.
Andrea Di Biagio [Mon, 11 Feb 2019 14:53:04 +0000 (14:53 +0000)]
[MCA] Return a mask of busy resources from method ResourceManager::checkAvailability(). NFCI
In case of bottlenecks caused by pipeline pressure, we want to be able to
correctly report the set of problematic pipelines. This is a first step towards
adding support for bottleneck hints in llvm-mca (see PR37494). No functional
change intended.
Neil Henning [Mon, 11 Feb 2019 14:44:14 +0000 (14:44 +0000)]
[AMDGPU] Fix DPP sequence in atomic optimizer.
This commit fixes the DPP sequence in the atomic optimizer (which was
previously missing the row_shr:3 step), and works around a read_register
exec bug by using a ballot instead.
Sam Parker [Mon, 11 Feb 2019 11:35:42 +0000 (11:35 +0000)]
[ARM] Add v8m.base pattern for add negative imm
The v8m.base ISA contains movw, which can operate on an unsigned
16-bit value. Add the pattern that converts an add with a negative
value, that could fit into 16-bits when negated, into a sub with that
positive value.
Diana Picus [Mon, 11 Feb 2019 10:30:22 +0000 (10:30 +0000)]
test-release.sh: Add option to use ninja
Allow the use of ninja instead of make. This is useful on some
platforms where we'd like to be able to limit the number of link jobs
without slowing down the other steps of the release.
This patch adds a -use-ninja command line option, which sets the
generator to Ninja both for LLVM and the test-suite. It also deals with
some differences between make and ninja:
* DESTDIR handling - ninja doesn't like this to be listed after the
target, but both make and ninja can handle it before the command
* Verbose mode - ninja uses -v, make uses VERBOSE=1
* Keep going mode - make has a -k mode, which builds as much as possible
even when failures are encountered; for ninja we need to set a hard
limit (we use 100 since most people won't look at 100 failures anyway)
[DWARF] LLVM ERROR: Broken function found, while removing Debug Intrinsics.
Check that when SimplifyCFG is flattening a 'br', all their debug intrinsic instructions are removed, including any dbg.label referencing a label associated with the basic blocks being removed.
As the test case involves a CFG transformation, move it to the correct location.
Sjoerd Meijer [Mon, 11 Feb 2019 09:37:42 +0000 (09:37 +0000)]
[ARM] LoadStoreOptimizer: reoder limit
The whole design of generating LDMs/STMs is fragile and unreliable: it depends on
rescheduling here in the LoadStoreOptimizer that isn't register pressure aware
and regalloc that isn't aware of generating LDMs/STMs.
This patch adds a (hidden) option to control the total number of instructions that
can be re-ordered. I appreciate this looks only a tiny bit better than a hard-coded
constant, but at least it allows more easy experimentation with different values
for now. Ideally we calculate this reorder limit based on some heuristics, and take
register pressure into account. I might be looking into that next.
Michal Gorny [Mon, 11 Feb 2019 09:07:07 +0000 (09:07 +0000)]
[llvm] [cmake] Use current directory in GenerateVersionFromVCS
Find dependent scripts of GenerateVersionFromVCS in current directory
rather than ../../cmake/modules. I do not see any reason why the former
would not work and The latter is incorrect when GenerateVersionFromVCS
is used from install directory (i.e. in stand-alone builds).
Chandler Carruth [Mon, 11 Feb 2019 07:42:30 +0000 (07:42 +0000)]
[CallSite removal] Migrate the statepoint GC infrastructure to use the
`CallBase` class rather than `CallSite` wrappers.
I pushed this change down through most of the statepoint infrastructure,
completely removing the use of CallSite where I could reasonably do so.
I ended up making a couple of cut-points: generic call handling
(instcombine, TLI, SDAG). As soon as it hit truly generic handling with
users outside the immediate code, I simply transitioned into or out of
a `CallSite` to make this a reasonable sized chunk.
Simon Pilgrim [Sun, 10 Feb 2019 17:04:00 +0000 (17:04 +0000)]
[DAGCombine] Simplify funnel shifts with undef/zero args to bitshifts
Now that we have SimplifyDemandedBits support for funnel shifts (rL353539), we need to simplify funnel shifts back to bitshifts in cases where either argument has been folded to undef/zero.
Sanjay Patel [Sun, 10 Feb 2019 15:22:06 +0000 (15:22 +0000)]
[x86] narrow 256-bit horizontal ops via demanded elements
256-bit horizontal math ops are an x86 monstrosity (and thankfully have
not been extended to 512-bit AFAIK).
The two 128-bit halves operate on separate halves of the inputs. So if we
don't demand anything in the upper half of the result, we can extract the
low halves of the inputs, do the math, and then insert that result into a
256-bit output.
All of the extract/insert is free (ymm<-->xmm), so we're left with a
narrower (cheaper) version of the original op.
In the affected tests based on:
https://bugs.llvm.org/show_bug.cgi?id=33758
https://bugs.llvm.org/show_bug.cgi?id=38971
...we see that the h-op narrowing can result in further narrowing of other
math via existing generic transforms.
I originally drafted this patch as an exact pattern match starting from
extract_vector_elt, but I thought we might see diffs starting from
extract_subvector too, so I changed it to a more general demanded elements
solution. There are no extra existing regression test improvements from
that switch though, so we could go back.
Sanjay Patel [Sun, 10 Feb 2019 14:29:57 +0000 (14:29 +0000)]
[TargetLowering] refactor setcc folds to fix another miscompile (PR40657)
SimplifySetCC still has much room for improvement, but this should
fix the remaining problem examples from:
https://bugs.llvm.org/show_bug.cgi?id=40657
George Rimar [Sun, 10 Feb 2019 08:35:38 +0000 (08:35 +0000)]
[yaml2obj] - Fix .dynamic section entries writing for 32bit targets.
This was introduced by me in r353613.
I tried to fix Big-endian bot and replaced
uintX_t -> ELFT::Xword. But ELFT::Xword is a packed<uint64_t>,
so it is always 8 bytes and that was obviously incorrect.
My intention was to use something like packed<uint> actually, which
size is target dependent.
Patch fixes this bug and adds a test case, since no bots seems reported this.
Simon Pilgrim [Sat, 9 Feb 2019 20:34:59 +0000 (20:34 +0000)]
[X86] CombineOr - fold to generic funnel shifts
As discussed on D57389, this is a first step towards moving the SHLD/SHRD matching code to DAGCombiner using FSHL/FSHR instead.
There's a bit of work to do before I can do that, so this just folds to FSHL/FSHR in the existing code (handling the different SHRD/FSHR argument ordering), which fixes the issue we had with i16 shift amounts not being correctly masked.
projects / llvm / log