On x86: (shl V, 1) -> add V,V
Hardware support for vector-shift is sparse and in many cases we scalarize the
result. Additionally, on sandybridge padd is faster than shl.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143311 91177308-0d34-0410-b5e6-96231b3b80d8
If all of the inputs are zero/any_extended, create a new simple BV
which can be further optimized by other BV optimizations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143297 91177308-0d34-0410-b5e6-96231b3b80d8
fixes: Use a separate register, instead of SP, as the
calling-convention resource, to avoid spurious conflicts with
actual uses of SP. Also, fix unscheduling of calling sequences,
which can be triggered by pseudo-two-address dependencies.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143206 91177308-0d34-0410-b5e6-96231b3b80d8
Don't assume APInt::getRawData() would hold target-aware endianness nor host-compliant endianness. rawdata[0] holds most lower i64, even on big endian host.
FIXME: Add a testcase for big endian target.
FIXME: Ditto on CompileUnit::addConstantFPValue() ?
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143194 91177308-0d34-0410-b5e6-96231b3b80d8
it fixes the dragonegg self-host (it looks like gcc is miscompiled).
Original commit messages:
Eliminate LegalizeOps' LegalizedNodes map and have it just call RAUW
on every node as it legalizes them. This makes it easier to use
hasOneUse() heuristics, since unneeded nodes can be removed from the
DAG earlier.
Make LegalizeOps visit the DAG in an operands-last order. It previously
used operands-first, because LegalizeTypes has to go operands-first, and
LegalizeTypes used to be part of LegalizeOps, but they're now split.
The operands-last order is more natural for several legalization tasks.
For example, it allows lowering code for nodes with floating-point or
vector constants to see those constants directly instead of seeing the
lowered form (often constant-pool loads). This makes some things
somewhat more complicated today, though it ought to allow things to be
simpler in the future. It also fixes some bugs exposed by Legalizing
using RAUW aggressively.
Remove the part of LegalizeOps that attempted to patch up invalid chain
operands on libcalls generated by LegalizeTypes, since it doesn't work
with the new LegalizeOps traversal order. Instead, define what
LegalizeTypes is doing to be correct, and transfer the responsibility
of keeping calls from having overlapping calling sequences into the
scheduler.
Teach the scheduler to model callseq_begin/end pairs as having a
physical register definition/use to prevent calls from having
overlapping calling sequences. This is also somewhat complicated, though
there are ways it might be simplified in the future.
This addresses rdar://9816668, rdar://10043614, rdar://8434668, and others.
Please direct high-level questions about this patch to management.
Delete #if 0 code accidentally left in.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143188 91177308-0d34-0410-b5e6-96231b3b80d8
on every node as it legalizes them. This makes it easier to use
hasOneUse() heuristics, since unneeded nodes can be removed from the
DAG earlier.
Make LegalizeOps visit the DAG in an operands-last order. It previously
used operands-first, because LegalizeTypes has to go operands-first, and
LegalizeTypes used to be part of LegalizeOps, but they're now split.
The operands-last order is more natural for several legalization tasks.
For example, it allows lowering code for nodes with floating-point or
vector constants to see those constants directly instead of seeing the
lowered form (often constant-pool loads). This makes some things
somewhat more complicated today, though it ought to allow things to be
simpler in the future. It also fixes some bugs exposed by Legalizing
using RAUW aggressively.
Remove the part of LegalizeOps that attempted to patch up invalid chain
operands on libcalls generated by LegalizeTypes, since it doesn't work
with the new LegalizeOps traversal order. Instead, define what
LegalizeTypes is doing to be correct, and transfer the responsibility
of keeping calls from having overlapping calling sequences into the
scheduler.
Teach the scheduler to model callseq_begin/end pairs as having a
physical register definition/use to prevent calls from having
overlapping calling sequences. This is also somewhat complicated, though
there are ways it might be simplified in the future.
This addresses rdar://9816668, rdar://10043614, rdar://8434668, and others.
Please direct high-level questions about this patch to management.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143177 91177308-0d34-0410-b5e6-96231b3b80d8
MORESTACK_RET_RESTORE_R10; which are lowered to a RET and a RET
followed by a MOV respectively. Having a fake instruction prevents
the verifier from seeing a MachineBasicBlock end with a
non-terminator (MOV). It also prevents the rather eccentric case of a
MachineBasicBlock ending with RET but having successors nevertheless.
Patch by Sanjoy Das.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@143062 91177308-0d34-0410-b5e6-96231b3b80d8
discussions with Andy. Fundamentally, the previous algorithm is both
counter productive on several fronts and prioritizing things which
aren't necessarily the most important: static branch prediction.
The new algorithm uses the existing loop CFG structure information to
walk through the CFG itself to layout blocks. It coalesces adjacent
blocks within the loop where the CFG allows based on the most likely
path taken. Finally, it topologically orders the block chains that have
been formed. This allows it to choose a (mostly) topologically valid
ordering which still priorizes fallthrough within the structural
constraints.
As a final twist in the algorithm, it does violate the CFG when it
discovers a "hot" edge, that is an edge that is more than 4x hotter than
the competing edges in the CFG. These are forcibly merged into
a fallthrough chain.
Future transformations that need te be added are rotation of loop exit
conditions to be fallthrough, and better isolation of cold block chains.
I'm also planning on adding statistics to model how well the algorithm
does at laying out blocks based on the probabilities it receives.
The old tests mostly still pass, and I have some new tests to add, but
the nested loops are still behaving very strangely. This almost seems
like working-as-intended as it rotated the exit branch to be
fallthrough, but I'm not convinced this is actually the best layout. It
is well supported by the probabilities for loops we currently get, but
those are pretty broken for nested loops, so this may change later.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142743 91177308-0d34-0410-b5e6-96231b3b80d8
SHL inserts zeros from the right, thus even when the original
sign_extend_inreg value was of 1-bit, we need to sra.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142724 91177308-0d34-0410-b5e6-96231b3b80d8
ZExtPromotedInteger and SExtPromotedInteger based on the operation we legalize.
SetCC return type needs to be legalized via PromoteTargetBoolean.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142660 91177308-0d34-0410-b5e6-96231b3b80d8
it's a bit more plausible to use this instead of CodePlacementOpt. The
code for this was shamelessly stolen from CodePlacementOpt, and then
trimmed down a bit. There doesn't seem to be much utility in returning
true/false from this pass as we may or may not have rewritten all of the
blocks. Also, the statistic of counting how many loops were aligned
doesn't seem terribly important so I removed it. If folks would like it
to be included, I'm happy to add it back.
This was probably the most egregious of the missing features, and now
I'm going to start gathering some performance numbers and looking at
specific loop structures that have different layout between the two.
Test is updated to include both basic loop alignment and nested loop
alignment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142645 91177308-0d34-0410-b5e6-96231b3b80d8
canonical example I used when developing it, and is one of the primary
motivating real-world use cases for __builtin_expect (when burried under
a macro).
I'm working on more test cases here, but I'm trying to make sure both
that the pass is doing the right thing with the test cases and that they
aren't too brittle to changes elsewhere in the code generation pipeline.
Feedback and/or suggestions on how to test this are very welcome.
Especially feedback on whether testing the block comments is a good
strategy; I couldn't find any good examples to steal from but all the
other ideas I had were a lot uglier or more fragile.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142644 91177308-0d34-0410-b5e6-96231b3b80d8
When checking the availability of instructions using the TLI, a 'promoted'
instruction IS available. It means that the value is bitcasted to another type
for which there is an operation. The correct check for the availablity of an
instruction is to check if it should be expanded.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142542 91177308-0d34-0410-b5e6-96231b3b80d8
On spec/gcc, this caused a codesize improvement of ~1.9% for ARM mode and ~4.9% for Thumb(2) mode. This is
codesize including literal pools.
The pools themselves doubled in size for ARM mode and quintupled for Thumb mode, leaving suggestion that there
is still perhaps redundancy in LLVM's use of constant pools that could be decreased by sharing entries.
Fixes PR11087.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142530 91177308-0d34-0410-b5e6-96231b3b80d8
svn r139159 caused SelectionDAG::getConstant() to promote BUILD_VECTOR operands
with illegal types, even before type legalization. For this testcase, that led
to one BUILD_VECTOR with i16 operands and another with promoted i32 operands,
which triggered the assertion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142370 91177308-0d34-0410-b5e6-96231b3b80d8
.file filenumber "directory" "filename"
This removes one join+split of the directory+filename in MC internals. Because
bitcode files have independent fields for directory and filenames in debug info,
this patch may change the .o files written by existing .bc files.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142300 91177308-0d34-0410-b5e6-96231b3b80d8
The decision was to pack the bits. Currently no codegen supports this.
Currently, all of the bits in the vector are saved into the same address
in memory.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142149 91177308-0d34-0410-b5e6-96231b3b80d8
When spilling around an instruction with a dead def, remember to add a
value number for the def.
The missing value number wouldn't normally create problems since there
would be an incoming live range as well. However, due to another bug
we could spill a dead V_SET0 instruction which doesn't read any values.
The missing value number caused an empty live range to be created which
is dangerous since it doesn't interfere with anything.
This fixes part of PR11125.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141923 91177308-0d34-0410-b5e6-96231b3b80d8
release the stack segment and reset the stack pointer. Place the code in its own
MBB to make the verifier happy.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141859 91177308-0d34-0410-b5e6-96231b3b80d8
Now that MI->getRegClassConstraint() can also handle inline assembly,
don't bail when recomputing the register class of a virtual register
used by inline asm.
This fixes PR11078.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141836 91177308-0d34-0410-b5e6-96231b3b80d8
our current machine instruction defines a register with the same register class
as what's being replaced. This showed up in the SPEC 403.gcc benchmark, where it
would ICE because a tail call was expecting one register class but was given
another. (The machine instruction verifier catches this situation.)
<rdar://problem/10270968>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141830 91177308-0d34-0410-b5e6-96231b3b80d8
When widening a copy, we are reading a larger register that may not be
live. Use an <undef> flag to tell the register scavenger and machine
code verifier that we know the value isn't defined.
We now widen:
%S6<def> = COPY %S4<kill>, %D3<imp-def>
into:
%D3<def> = VMOVD %D2<undef>, pred:14, pred:%noreg, %S4<imp-use,kill>
This also keeps the <kill> flag on %S4 so we don't inadvertently kill a
live value in %S5.
Finally, ensure that ARMBaseInstrInfo::setExecutionDomain() preserves
the <undef> flag when converting VMOVD to VORR.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141746 91177308-0d34-0410-b5e6-96231b3b80d8
file. Since it should only be used when necessary propagate it through
the backend code generation and tweak testcases accordingly.
This helps with code like in clang's test/CodeGen/debug-info-line.c where
we have multiple #line directives within a single lexical block and want
to generate only a single block that contains each file change.
Part of rdar://10246360
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141729 91177308-0d34-0410-b5e6-96231b3b80d8
For example, MachineLICM should not hoist a load that is not guaranteed to be executed.
Radar 10254254.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141689 91177308-0d34-0410-b5e6-96231b3b80d8
For example, MachineLICM should not hoist a load that is not guaranteed to be executed.
Radar 10254254.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141569 91177308-0d34-0410-b5e6-96231b3b80d8
ARMII::AddrModeT1_s, we need to take into account that if the frame register is
ARM::SP, then the number of bits is 8. If it's not ARM::SP, then the number of
bits is 5.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141529 91177308-0d34-0410-b5e6-96231b3b80d8
In 64-bit mode, sub_8bit_hi sub-registers can only be used by NOREX
instructions. The COPY created from the EXTRACT_SUBREG DAG node cannot
target all GR8 registers, only those in GR8_NOREX.
TO enforce this, we ensure that all instructions using the
EXTRACT_SUBREG are GR8_NOREX constrained.
This fixes PR11088.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141499 91177308-0d34-0410-b5e6-96231b3b80d8
merging an lsl #2 that has multiple uses on A9. This shift is free, so there is
no problem merging it in multiple places. Other unprofitable shifts will not be
merged.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141247 91177308-0d34-0410-b5e6-96231b3b80d8
and the alignment is 0 (i.e., it's defined globally in one file and declared in
another file) it could get an alignment which is larger than the ABI allows for
that type, resulting in aligned moves being used for unaligned loads.
For instance, in file A.c:
struct S s;
In file B.c:
struct {
// something long
};
extern S s;
void foo() {
struct S p = s;
// ...
}
this copy is a 'memcpy' which is turned into a series of 'movaps' instructions
on X86. But this is wrong, because 'struct S' has alignment of 4, not 16.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140902 91177308-0d34-0410-b5e6-96231b3b80d8
Rewriting the entire loop nest now requires -enable-lsr-nested.
See PR11035 for some performance data.
A few unit tests specifically test nested LSR, and are now under a flag.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140762 91177308-0d34-0410-b5e6-96231b3b80d8
Lay some groundwork for converting to MC-based asm printer. This is the first
of probably many patches to bring the back-end back up-to-date with all of the
recent MC changes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140697 91177308-0d34-0410-b5e6-96231b3b80d8
This also enables domain swizzling for AVX code which required a few
trivial test changes.
The pass will be moved to lib/CodeGen shortly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140659 91177308-0d34-0410-b5e6-96231b3b80d8
I did not convert Atomics-32.ll and Atomics-64.ll by hand; the diff is autoupgrade output.
The wmb test is gone because there isn't any way to express wmb with the new atomic instructions; if someone really needs a non-asm way to write a wmb on Alpha, a platform-specific intrisic could be added.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140566 91177308-0d34-0410-b5e6-96231b3b80d8
Math is hard, and isScaledConstantInRange() always returned false for
negative constants. It was doing unsigned division of negative numbers
before casting back to signed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140425 91177308-0d34-0410-b5e6-96231b3b80d8
floating point add/sub of appropriate shuffle vectors. Does not
synthesize the 256 bit AVX versions because they work differently.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140332 91177308-0d34-0410-b5e6-96231b3b80d8
Few weeks ago, llvm completely inverted the debug info graph. Earlier each debug info node used to keep track of its compile unit, now compile unit keeps track of important nodes. One impact of this change is that the global variable's do not have any context, which should be checked before deciding to use AT_specification DIE.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140282 91177308-0d34-0410-b5e6-96231b3b80d8
Vector SetCC result types need to be type-legalized.
This code worked before because scalar result types are known to be legal.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140249 91177308-0d34-0410-b5e6-96231b3b80d8
On Windows x64, 128-bit arguments are not passed by reg but by indirect. eg.
maxpd:
vmovapd (%rcx), %xmm0
vmaxpd (%rdx), %xmm0, %xmm0
FIXME: I don't care YMM on x64 for now.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140143 91177308-0d34-0410-b5e6-96231b3b80d8
Modified ARMISelLowering::AdjustInstrPostInstrSelection to handle the
full gamut of CPSR defs/uses including instructins whose "optional"
cc_out operand is not really optional. This allowed removal of the
hasPostISelHook to simplify the .td files and make the implementation
more robust.
Fixes rdar://10137436: sqlite3 miscompile
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140134 91177308-0d34-0410-b5e6-96231b3b80d8
dag-combine optimization to implement the ext-load efficiently (using shuffles).
For example the type <4 x i8> is stored in memory as i32, but it needs to
find its way into a <4 x i32> register. Previously we scalarized the memory
access, now we use shuffles.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139995 91177308-0d34-0410-b5e6-96231b3b80d8
maxps and maxpd). This broke the sse41-blend.ll testcase by causing
maxpd to be produced rather than a cmp+blend pair, which is the reason
I tweaked it. Gives a small speedup on doduc with dragonegg when the
GCC vectorizer is used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139986 91177308-0d34-0410-b5e6-96231b3b80d8
take into consideration the presence of AVX. This change, together with
the SSEDomainFix enabled for AVX, makes AVX codegen to always (hopefully)
emit the same code as SSE for 128-bit vector ops. I don't
have a testcase for this, but AVX now beats SSE in performance for
128-bit ops in the majority of programas in the llvm testsuite
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139817 91177308-0d34-0410-b5e6-96231b3b80d8
An improper SlotIndex->VNInfo lookup was leading to unsafe copy removal.
Fixes PR10920 401.bzip2 miscompile with no IV rewrite.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139765 91177308-0d34-0410-b5e6-96231b3b80d8
However with this fix it does now.
Basically the operand order for the x86 target specific node
is not the same as the instruction, but since the intrinsic need that
specific order at the instruction definition, just change the order
during legalization. Also, there were some wrong invertions of condition
codes, such as GE => LE, GT => LT, fix that too. Fix PR10907.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139528 91177308-0d34-0410-b5e6-96231b3b80d8
assert("not implemented for target shuffle node");
to:
assert(0 && "not implemented for target shuffle node");
This causes a test failure in CodeGen/X86/palignr.ll which has
been marked as XFAIL for the time being.
Test failure filed at PR10901.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139454 91177308-0d34-0410-b5e6-96231b3b80d8
removing support for Mips1 and Mips2.
This change and the ones that follow have been discussed with and approved by
Bruno.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139344 91177308-0d34-0410-b5e6-96231b3b80d8
in Nadav's r139285 and r139287 commits.
1) Rename vsel.ll to a more descriptive name
2) Change the order of BLEND operands to "Op1, Op2, Cond", this is
necessary because PBLENDVB is already used in different places with
this order, and it was being emitted in the wrong way for vselect
3) Add AVX patterns and tests for the same SSE41 instructions
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139305 91177308-0d34-0410-b5e6-96231b3b80d8
(The fix for the related failures on x86 is going to be nastier because we actually need Acquire memoperands attached to the atomic load instrs, etc.)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139221 91177308-0d34-0410-b5e6-96231b3b80d8
Now the 'S' instructions, e.g. ADDS, treat S bit as optional operand as well.
Also fix isel hook to correctly set the optional operand.
rdar://10073745
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139157 91177308-0d34-0410-b5e6-96231b3b80d8
init.trampoline and adjust.trampoline intrinsics, into two intrinsics
like in GCC. While having one combined intrinsic is tempting, it is
not natural because typically the trampoline initialization needs to
be done in one function, and the result of adjust trampoline is needed
in a different (nested) function. To get around this llvm-gcc hacks the
nested function lowering code to insert an additional parent variable
holding the adjust.trampoline result that can be accessed from the child
function. Dragonegg doesn't have the luxury of tweaking GCC code, so it
stored the result of adjust.trampoline in the memory GCC set aside for
the trampoline itself (this is always available in the child function),
and set up some new memory (using an alloca) to hold the trampoline.
Unfortunately this breaks Go which allocates trampoline memory on the
heap and wants to use it even after the parent has exited (!). Rather
than doing even more hacks to get Go working, it seemed best to just use
two intrinsics like in GCC. Patch mostly by Sanjoy Das.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139140 91177308-0d34-0410-b5e6-96231b3b80d8
The explanation about a 0 argument being materialized as xor is no
longer valid. Rematerialization will check if EFLAGS is live before
clobbering it.
The code produced by X86TargetLowering::EmitLoweredSelect does not
clobber EFLAGS.
This causes one less testb instruction to be generated in the cmov.ll
test case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139057 91177308-0d34-0410-b5e6-96231b3b80d8
to be unreliable on platforms which require memcpy calls, and it is
complicating broader legalize cleanups. It is hoped that these cleanups
will make memcpy byval easier to implement in the future.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138977 91177308-0d34-0410-b5e6-96231b3b80d8
- On COFF the .lcomm directive has an alignment argument.
- On ELF we fall back to .local + .comm
Based on a patch by NAKAMURA Takumi.
Fixes PR9337, PR9483 and PR10128.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138976 91177308-0d34-0410-b5e6-96231b3b80d8
An instruction may define part of a register where the other bits are
undefined. In that case, it is safe to rematerialize the instruction.
For example:
%vreg2:ssub_0<def> = VLDRS <cp#0>, 0, pred:14, pred:%noreg, %vreg2<imp-def>
The extra <imp-def> operand indicates that the instruction does not read
the other parts of the virtual register, so a remat is safe.
This patch simply allows multiple def operands for the virtual register.
It is MI->readsVirtualRegister() that determines if we depend on a
previous value so remat is impossible.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138953 91177308-0d34-0410-b5e6-96231b3b80d8
An instruction that redefines only part of a larger register can never
be rematerialized since the virtual register value depends on the old
value in other parts of the register.
This was fixed for the inline spiller in r138794. This patch fixes the
problem for all register allocators, and includes a small test case.
<rdar://problem/10032939>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138944 91177308-0d34-0410-b5e6-96231b3b80d8
Added canClobberReachingPhysRegUse() to handle a particular pattern in
which a two-address instruction could be forced to interfere with
EFLAGS, causing a compare to be unnecessarilly cloned.
Fixes rdar://problem/5875261
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138924 91177308-0d34-0410-b5e6-96231b3b80d8
